Answers to Selected Exercises

Worked solutions and model analyses for the starred and odd-numbered exercises from each chapter. Forensic reasoning is learned by working evidence, not by reading conclusions — attempt every exercise, and especially every Case File entry, before reading its solution.

Exercise 1.1

Forensic science is the application of scientific methods and principles to questions a court of law must answer. Criminalistics is the narrower subset concerned specifically with recognizing, collecting, identifying, and comparing physical evidence (prints, fibers, firearms, drugs, DNA). All criminalistics is forensic science; not all forensic science is criminalistics (pathology, accounting, and psychology are forensic but not criminalistics).

Exercise 1.4

A class characteristic is shared by a group; an individual characteristic is effectively unique to one source. Original examples: class — a car's factory paint color shared by every vehicle of that make/year; individual — the specific pattern of scratches and a chipped layer on one repainted bumper. (Many answers are valid; the test is whether the feature could be shared by others or realistically belongs to one source.)

Exercise 1.9

"Consistent with" — same color, composition, cross-section — is a class association: those fibers are shared by every garment of that type. To move toward the individual end you would need features unlikely to be shared, e.g., an unusual dye lot, a manufacturing defect, or rare composite damage — and even then fibers rarely reach true individualization. Honest verb: consistent with / cannot exclude, not "match."

Exercise 1.12

(a) "The print is consistent with the defendant's finger and does not exclude him." (b) "The hairs are consistent with the defendant's; microscopic comparison cannot establish that they are his." (c) "I observed correspondence between the mark and the dentition, but bite-mark comparison has no validated basis for identifying a single biter, so I cannot attribute this mark to this person." Each replaces an individualization claim with what the method can defend.

Exercise 1.16

Two reasons the "they have nothing" argument can mislead even when literally true: (1) Most real scenes yield no usable forensic evidence — its absence is the norm and says nothing about guilt or innocence (the CSI effect manufactures the expectation). (2) A case can rest validly on non-forensic evidence (witnesses, documents, digital records); "no DNA" does not mean "no case." The honest point is that absence of forensic evidence is neither proof of innocence nor evidence of sloppiness.

Exercise 1.19

The single most important question is "What is the method's error rate, established by well-designed studies?" Its absence is itself a finding because a method that claims to match a mark to one unique source, yet has never measured how often that match is wrong, is asserting certainty it has not earned. With no measured error rate, the claim of reliability is an assumption, not a result (the core of the NAS/PCAST critique).

Exercise 1.24

Two defensible stand sentences: "The latent print is consistent with the defendant's known print; I found agreement in the ridge detail I could compare and no unexplained differences." / "I cannot state to a scientific certainty that it is his and no one else's; what I can say is that he cannot be excluded and that the degree of correspondence is what I would expect if he were the source." Neither overstates beyond what latent comparison supports.

Exercise 1.27

The trap is confirmation bias / anchoring: treating the first reasonable assumption (accidental fire) as the hypothesis to be confirmed rather than one of several to be tested. The chapter's correction: the evidence's job is to exclude the explanations it can and to support whatever remains only as far as honestly warranted — so the file opens with a question, not a theory, and we refuse to theorize before we have data.

Exercise 1.21 (odd)

Cynicism discards all forensic evidence as worthless; calibration asks of each piece "what kind, how strong, and how do we know?" and weights it accordingly. The book argues for calibration because the disciplines genuinely differ in validity — discarding DNA because bite marks are junk is as much an error as trusting bite marks because DNA is sound.

Exercise 1.25 (odd)

Clearing the innocent is forensic science's "finest hour" because its exclusion power is its most reliable and least overstated function: a true exclusion is often categorical, where an inclusion is only probabilistic. Tens of thousands are excluded pre-charge and never prosecuted, and post-conviction DNA has freed hundreds — outcomes that are pure benefit, with none of the overstatement risk that attends inclusion.

Chapter 2 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 2.1

A crime scene is any location where a crime may have occurred and where physical evidence of that crime may be found. The two "mays" matter because, at the moment of first response, no one yet knows for certain that a crime occurred (Mill Creek was treated as a probable accident) or where the boundaries of the relevant evidence lie. The scene is a working hypothesis about where the relevant past is still legible, not a fixed bounded place — which is why the first responder must protect its legibility before anyone has decided what the scene even is.

Exercise 2.3 †

Scene security is the set of actions that restrict access to a scene — perimeter, access control — so that evidence is neither lost, added, moved, nor contaminated before it can be documented and collected. The single most useful piece of paper is the entry/exit log (crime-scene log): a record of every person who crosses the perimeter, with time in, time out, and reason. It is the only honest answer to "who had access to this scene, and could any of them have moved or added this item?" A scene without a log is one whose contents can be challenged item by item.

Exercise 2.5

"Document before you disturb." (Five words.) Every irreversible act — measuring, collecting, walking where an item was — is preceded by a reversible record.

Exercise 2.7

Primary scene: where the principal criminal act occurred — e.g., the room where the victim was actually killed. Secondary scene: any other connected location with related evidence — e.g., where the body was dumped, the suspect's vehicle, or the route the perpetrator took. The richest evidence of what happened tends to be at the primary; the richest evidence of who and where they went is often spread across the secondaries.

Exercise 2.9

Line/strip — large open areas (a field, the ground around the cabin). Grid — same areas needing higher thoroughness; two perpendicular passes catch what one misses. Spiral — a scene with an obvious center (a body) or limited-access/underwater scenes one searcher must cover. Zone/quadrant — buildings, searched room by room. (Wheel/ray — small circular scenes, rarely ideal.)

Exercise 2.11

The skipped steps, in order: (a) overall and mid-range photographs establishing the knife's location relative to the room and nearby reference points; (b) a close-up with a scale; (c) a sketch measurement fixing the knife to two fixed points; only then (d) collection. Unrecoverable now: the knife's original location and orientation in the scene and its true dimensions from a scaled photo. The object survives, but its place in the world is gone — it has become an orphan item from nowhere. (The lab can still test the knife itself; what is lost is where it was.)

Exercise 2.12 †

Errors already committed before any evidence is touched: - Failure to secure / no perimeter (§2.7 #1) — the most damaging, because it is the parent of the rest: the scene was never controlled, so the others follow. - No entry/exit log (§2.7 #2) — now no one can say who had access; every item becomes challengeable. - Responder/visitor contamination (§2.7 #4) — four additional people (and a grieving relative) tracking material in and out of an undocumented scene, possibly moving items. - Risk of disturbing before documenting (§2.7 #3) — uncontrolled people will move things before the scene is recorded.

Ranking by damage: failure to secure (it causes the others) > responder/visitor contamination (silent, irreversible, can manufacture false associations) > no log (makes the contamination un-auditable) > disturbance before documentation. The deep point: these are not four independent mistakes but one (no control) cascading into four.

Exercise 2.13

Swab A (air-dried, paper): a clean, full profile is likely. Swab B (sealed wet in plastic): visibly moldy on arrival, yielding a partial/degraded profile or none. Mechanism: plastic traps moisture; moisture is the medium in which bacteria and fungi grow; those microbes digest DNA. Nine warm days turn the sealed bag into an incubator. The destroyer is microbial growth fueled by trapped moisture (warmth accelerates it). Identical evidence, opposite fates, caused entirely by the container.

Exercise 2.15 †

The observation tells you the body was moved after death: livor mortis (settled blood) pools at the lowest parts under gravity, so pooling on the back of a face-down body is inconsistent with the found position — therefore the position changed after death. This makes the location a possible secondary scene. What it does not yet tell you: the cause of death, the time of death precisely, who moved the body, or why — those are separate questions (Chapter 11 develops postmortem changes). It is one clean physical fact, not a narrative.

Exercise 2.17

The perpetrator's route to and from a scene — especially a remote one — often carries the richest evidence (discarded items, shoe and tire impressions, transfer material), and that route extends well beyond the body. A perimeter is trivial to shrink later if it proves too large, but evidence outside a too-small perimeter is already trampled by the public, press, and responders by the time anyone realizes the scene was bigger than it looked. Asymmetric cost ⇒ set the tape generously.

Exercise 2.19 †

With no scale and no sketch, you can describe the stain's appearance — color, apparent shape, approximate relative position within the frame — and say it appears consistent with (e.g.) a blood-like stain. You cannot state its true dimensions (no scale) or its exact location in the scene fixed to reference points (no sketch/mid-range context). Honest verbs: "the photograph shows a reddish stain of unknown true size"; you must not say "an 8 cm stain located 40 cm from the floor" — those are measurements the documentation never captured. The limit is the documentation's, not the evidence's.

Exercise 2.21

This is a staged scene (a homicide arranged to look like a suicide). The single principle by which it is detected: physical inconsistency between what the scene claims and what the evidence independently shows — here, a wound the deceased could not physically have inflicted and hands negative for gunshot residue. The stager controls appearance, not physics; the contradictions are testable. (Reading the offender's personality from the staging would be overreach — Chapter 28.)

Exercise 2.22 †

The chain of custody is worth exactly what it always is: it establishes the swab's identity and integrity in handling (same item, no undocumented alteration after bagging). It is powerless to fix what happened before bagging — the cross-contamination from one pair of gloves used across five items. A flawless chain on a sample contaminated at collection is a flawless record of a worthless result. Custody is not quality; do not let the unbroken chain launder the bad collection. The honest reading: the DNA association may be an artifact of handling, and the result cannot be relied upon for an inclusion.

Exercise 2.23

Supportable by physical evidence: that the scene shows specific arrangements and inconsistencies (e.g., what was moved, what is staged) — testable physical facts. Overreach: inferring the offender's personality and relationship to the victim ("organized, controlling, knew the victim") from the arrangement — that is criminal profiling, whose validity is deflated in Chapter 28. Keep staging analysis where the evidence can check it; the mind-reading part is storytelling.

Exercise 2.24 †

Rewrite: "The unbroken chain of custody establishes that the item presented is the same item collected at the scene, handled as documented and not altered in any undocumented way." It says nothing about correct collection, pre-collection contamination, or the soundness of the analysis — those are separate questions. (The original sentence overstates "custody" into "quality.")

Exercise 2.25

The claim "we found no evidence pointing away from the defendant" is hostage to how the scene was searched and what the searchers recognized as evidence. A search only finds what it covers and what the searcher perceives as relevant; an early theory can bias both (tunnel vision, §2.3). "No evidence" may mean "nothing the search was designed or disposed to find," not "nothing was there." The defense should ask: how was the scene searched, by what pattern, to what standard, and under what assumption about what happened?

Exercise 2.27

Not an error — preservation of life outranks every evidentiary concern (§2.1). The correct rule: disturb the scene only as much as saving the person requires, and remember exactly what you did. The one thing the paramedic must do is document the disturbance — report to the investigator precisely what was moved, cut, or introduced, and where things were originally — converting an unavoidable disturbance into usable information. The disturbance that destroys a case is the one no one hears about.

Exercise 2.28 †

Forensic argument for refusing: scene errors are irreversible and upstream of everything. Downgrading to "accidental" before the search is complete lowers the standard of care (the §2.1 bias), so evidence that would distinguish accident from homicide may go unrecognized and uncollected — and cannot be collected later, because the scene plays once. The validity of every downstream analysis (DNA, toxicology, fire science) is capped by the scene work; a sample never collected can never be tested, and a scene released early can never be re-secured in its original state. The cost of a few more hours is hours; the cost of premature downgrade can be the case — and, if a crime did occur, a guilty person freed or, via a contaminated record, an innocent one implicated.

Exercise 2.29

Obligation: document and disclose the lapse immediately — note what items, by whom, when, and that they were handled bare-handed an hour ago — so the contamination risk is on the record and can be reasoned about downstream. Hiding it is worse than the lapse itself because documented contamination is survivable (the lab and the court can account for it), whereas undocumented contamination enters the case invisibly, as if the introduced material had always been there, corrupting conclusions silently. Concealment converts a manageable error into an undetectable one — and destroys the analyst's credibility if discovered.

Exercise 2.30 † (Cold-case extension)

Model entry (the condition-of-the-scene note, not a theory of guilt):

Mill Creek scene — condition note. The structure was treated from the outset as a probable accidental fire, not a possible crime scene, with the result that scene control likely fell below a homicide standard. Specifically, there is reason to doubt that (1) a generous perimeter was set to include the gravel approach/route; (2) an entry/exit log was kept from the first responder forward; (3) the scene was fully documented (notes, scaled photos, measured sketch) before disturbance; and (4) the most transient and biological evidence was recognized, collected, dried, and packaged in paper rather than lost or mishandled. Consequence: some evidence was probably lost or contaminated, and the provenance of what survives is open to challenge. This note adds no information about who is responsible. It records only the ceiling on trust for all later analysis: the science to come is only as reliable as a scene that was processed on the assumption there was no case to process.

What it adds: a sober accounting of conditions. What it does not add: any fact about the perpetrator.

Exercise 2.31 (Cold-case extension)

Per §2.6, an investigator should hold open that the cabin is a secondary scene — that Diallo may have been killed elsewhere and his body moved to the cabin and burned — because assuming the cabin is primary smuggles in the very "accident here, that's the whole story" frame the chapter warns against. Evidence that will help decide: the autopsy (Chapter 11) — postmortem indicators such as livor mortis consistency, and the soot/airway findings that bear on whether death preceded the fire — and the fire science (Chapter 22), which speaks to whether the fire was set at that location. Until those speak, "primary" is an assumption, not a finding.

Exercise 2.33

To the skeptic: the laboratory can only analyze what the scene gives it, in the condition the scene gives it. The most sensitive DNA method cannot recover a sample that was never collected, restore the original position of a moved item, or un-contaminate a swab handled with a dirty glove — and it cannot tell, on its own, whether material was at the scene or carried there by the response. Every laboratory result therefore inherits a ceiling set in the first hours on the ground: collection, documentation, packaging, and chain of custody. Brilliant analysis of a contaminated or undocumented sample is brilliantly worthless. The lab is where evidence is read; the scene is where it is either preserved or lost forever.

Exercise 2.34 †

Three errors with one root: (2) no entry/exit log, (3) disturbing before documenting, and (4) responder contamination can all flow from failure to establish scene control / secure the scene early (#1) — and behind that, often, from tunnel vision (#6): a scene believed to hold no crime is a scene no one bothers to secure, log, or protect from traffic. The single root cause is the failure to control the scene immediately, frequently driven by an early frame ("accident," "natural death") that tells everyone there is nothing to control.

Exercise 2.35

Exclusion (clearing the innocent — e.g., a DNA non-match that says "not this person," Chapter 1 §1.6) is forensic science's surest, most morally important power. But a clean exclusion requires a sample trustworthy enough to rely on. A contaminated sample can introduce material that fails to exclude someone who should be excluded (a false inclusion), or degrade until it excludes no one; a lost sample excludes nobody at all. So the exclusion power the field most prizes is only as good as the scene work that fed it — when the scene is contaminated, the door that should swing cleanly shut on an innocent suspect can jam.

Exercise 2.37

Real forensic work involves so much paperwork because documentation and chain of custody are what make the eventual science trustworthy: the notes, scaled photos, and measured sketch let an analyst's result be tied to a specific item in a specific place, and the chain of custody proves that item is the same one, unaltered, from scene to courtroom. Without the paperwork, even perfect laboratory work hangs in the air, unanchored and challengeable; the paperwork is not bureaucracy around the science — it is the foundation the science stands on.

Exercise 2.38 † (model)

Scenario (correct in every respect), annotated:

The first officer confirms the occupant is beyond aid and exits the unstable structure (§2.1 — safety and preservation of life first). She sets crime-scene tape well beyond the cabin, taking in the gravel track (§2.1 — generous perimeter; the route is evidence), and starts an entry/exit log naming everyone who approaches (§2.1, §2.7 — the log). The lead investigator does a single hands-in-pockets walkthrough (§2.2 — initial walkthrough, observe before acting), then documents the scene with overall, mid-range, and scaled close-up photos, contemporaneous ink notes of the conditions, and a measured sketch fixing each item to two reference points (§2.2 — three independent records; document before you disturb). The team searches the building zone by zone and the grounds by grid, collecting the transient evidence first (§2.3 — recorded pattern; fragility first). The doorframe blood is air-dried and sealed in paper; glass fragments go in a sealed vial (§2.4 — paper for biology, plastic for the inert). Each item is sealed, initialed across the tape, and entered into a chain of custody for transport (§2.4, §2.5 — sealing and the unbroken chain).

Chapter 3 — Worked Solutions (daggered † and odd-numbered items)

Solutions to the daggered exercises and the odd-numbered items. Reasoning is shown, not just labels — the argument is the point.

Exercise 3.1

Locard's exchange principle: whenever two objects come into contact, there is a mutual transfer of material between them ("every contact leaves a trace"). The slogan is incomplete because it omits the bidirectional nature of the exchange: the perpetrator does not only leave material at the scene, he also carries material away from it. That second direction is why a suspect's clothing, shoes, vehicle, and hands are themselves scenes to be processed.

Exercise 3.2 †

(a) Transient evidence — temporary by nature, lost if not recorded immediately; e.g., a gasoline odor, the warmth of a car hood, a wet stain. (b) Conditional evidence — depends on the state/conditions at a moment; e.g., a door's locked/unlocked state, a body's position, a light on or off. (c) Pattern evidence — the arrangement/geometry carries the information; e.g., a bloodstain pattern, a shoe impression, glass fracture lines. (d) Transfer evidence — material moved by contact; e.g., hairs, fibers, glass fragments, soil, gunshot residue. (e) Associative evidence — any evidence that tends to link a person/object/place to the crime; e.g., a dropped wallet, a DNA profile matching a suspect, a vehicle placed by tire marks.

Exercise 3.3

A questioned sample is recovered from the scene/person with an unknown source — it is what you are trying to identify. A known/reference sample is from a verified, documented source, collected so the questioned item can be compared against it. The known/reference sample is the one whose defining feature is that you are certain where it came from.

Exercise 3.5 †

The four limits of Locard's principle: (1) the trace may be too small/sparse to detect (undetected and absent are operationally identical); (2) the trace may not persist (transfer ≠ persistence — it may wash off, fall off, or burn away before processing); (3) the trace may be undetectable against its background (e.g., a white cotton fiber on a white cotton sheet); (4) even a recovered, readable trace usually establishes only association, not the actor, the time of contact, or the intent.

Exercise 3.7

Degradation is the natural deterioration/decay of evidence over time and under environmental stress (heat, moisture, sunlight, microbes); nothing foreign is added — the sample simply decays until unreadable. Contamination is the introduction of foreign material (or transfer between items) that was not part of the evidence as deposited; it adds material and can manufacture a false association. Contamination adds foreign material; degradation is the sample decaying on its own.

Exercise 3.9

Smell of bleach — transient evidence. Handling: it cannot be collected; it must be documented immediately (noted, timed, location recorded), because it will dissipate within minutes to an hour. (It may also hint at an attempt to destroy biological evidence — a thread, not a conclusion.)

Exercise 3.10 †

Muddy shoe impression in soil — pattern evidence (and potentially transfer/associative). Handling: photograph with a scale first (the geometry is the evidence), then preserve by casting and/or lifting per protocol; protect it from rain and traffic immediately. The impression can show a shoe's class characteristics (size, tread pattern, brand) and possibly individual wear; it cannot identify the foot or person, or when it was made.

Exercise 3.11

Engine-hood temperature — transient evidence. Handling: cannot be collected; document immediately (note that the hood is warm, and when it was felt), because it changes continuously as the engine cools. A warm hood is conditional information about recent use — a thread, not proof of anything by itself.

Exercise 3.13

Upward-fanning spray of small bloodstains — pattern evidence (bloodstain pattern; also transfer/ associative as a biological deposit). Handling: photograph extensively with scales (it usually cannot be removed from the wall), document position and orientation, and collect representative stains for serology/ DNA with appropriate controls. The pattern may support a reconstruction of how blood traveled (Chapter 10), but BPA sits low on the validity spectrum and the inference is contestable; it does not name an actor.

Exercise 3.15 †

Fire debris suspected to contain an accelerant — transfer/associative (and a chemistry sample). Handling: this is the case where the packaging rule reverses — seal it in an airtight metal can or special nylon bag, NOT paper, because the accelerant is volatile and would evaporate through paper before the lab could analyze it (Chapter 22–23). Collect a comparison/substrate sample of unburned material where possible.

Exercise 3.17

Order: (1) Questioned — the unknown stain itself, swabbed for blood and DNA. (2) Known/reference — buccal swabs from the victim and each suspect, so the unknown can be compared to known people; this is what gives the stain a possible source. (3) Controls — a substrate control (swab of an adjacent unstained area of the same sill, to show the painted wood is not the source of the signal) and a reagent/negative control (the test run with no sample, to show the chemicals are clean). Optionally a positive control (to show the test works) and elimination samples (from anyone with legitimate access).

Exercise 3.18 †

You would insist that a positive control was run — the test performed on a sample known to contain blood. A presumptive negative is only meaningful if you have shown the test could have gone positive; if the positive control failed (stayed negative), the test itself was not working, and the "negative" on the evidence is uninterpretable — it tells you nothing about whether blood was present. (A reagent control also matters, but the positive control is the one that validates a negative result.)

Exercise 3.19

An unidentified profile is of limited immediate value because comparison evidence is meaningless without a known sample to compare against — a complete profile matched to nobody is just a profile. The single thing that converts it into a strong association is a known/reference sample from a specific person (via a suspect swab or a database hit, e.g., CODIS — Chapter 7) that the questioned profile can be compared to.

Exercise 3.21 †

Placing a bloodstained shirt and a knife in the same sack risks cross-transfer of material between the two items (e.g., blood or trace from the shirt onto the knife, or vice versa). This falls under contamination (a threat to evidence integrity, §3.6), and it violates the rule that each item must be packaged separately so that no two evidence items touch inside one container. The contamination could manufacture an association (blood "on the knife") that the collection process, not the crime, created.

Exercise 3.23 †

Overstatement: the sentence leaps from association (the print is on the sill — his finger contacted that surface) to identity of the actor ("he is the burglar"), dropping two unstated assumptions: that the print was deposited at the relevant time (prints can't be reliably dated — §3.4) and that touching the sill means committing the burglary. Defensible rewrite: "A latent print consistent with the suspect was found on the windowsill, indicating his finger contacted that surface at some point; it does not establish when the print was deposited or that he committed the burglary."

Exercise 3.25

Overstatement: from presence of gunshot residue to the action of firing. GSR transfers by handling a fired weapon, being near a discharge, or contact with a contaminated surface (Chapter 24) — presence is not proof she pulled the trigger, and it carries no timestamp. Defensible: "Gunshot residue was detected on her hands, which is consistent with her having discharged a firearm, been in proximity to one, or contacted a contaminated surface; it does not by itself establish that she fired the weapon."

Exercise 3.27

Overstatement: pry marks are typically class evidence (they show a type of tool — a pry bar of certain width/shape), so "match the suspect's pry bar" usually overstates to an individual source, and "so he broke in" leaps from a tool to an actor. Defensible: "The pry marks are consistent with a pry-bar-type tool [and, if individual marks exist, with the suspect's specific tool]; this supports forced entry by such a tool but does not establish who used it or when the marks were made."

Exercise 3.29

The logical error is treating a statement about the world (contact deposits a trace) as a guarantee about our knowledge (we will recover and correctly read that trace) — ignoring the four limits (detectability, persistence, background, and the association-not-actor gap). It connects to the Cognitive-Bias Watch in §3.1: "every contact leaves a trace" can quietly become "I will keep searching until I find the trace that fits my suspect," which is confirmation bias driving both the search and the interpretation. The safeguard is to search and interpret without being steered by whom the detective favors (Chapter 31).

Exercise 3.30 †

Two threats: (1) Bias in interpretation (§3.1). Being told "we know it's the partner — find something" primes the analyst to over-search and to read ambiguous results as "consistent," a confirmation-bias cascade; the safeguard is context management / blind analysis (Chapter 31), giving the analyst only domain-relevant information. (2) Integrity/contamination pressure (§3.6). Pressure to "find something" can lead to rushed or careless collection, skipped controls, or unconscious mishandling that contaminates or over-interprets the gas-can evidence; the safeguards are strict collection protocol, controls (substrate, reagent), elimination samples, and separation of known from questioned samples — applied regardless of the detective's theory.

Exercise 3.33 †

"The sample is fine, so the gap doesn't matter" wrongly collapses physical integrity into documentary integrity. Physical integrity is that the item is unaltered and uncontaminated; documentary integrity is the unbroken, recorded, sealed, dated trail proving the item is the same one from the scene and was not tampered with. A six-day unexplained gap plus undated tape means the chain can no longer be vouched for — even a physically perfect sample now invites the argument that it was switched, tampered with, or contaminated during the gap, and a defense will (rightly) raise it. Both halves of integrity are required; one cannot substitute for the other.

Exercise 3.35 †

(a) The pry-marked door is conditional + pattern evidence at a scene processed late; its value depended on whether its state (the lock position, the marks) was documented before it changed, because conditional evidence is destroyed the moment conditions are altered (and a fire scene is heavily disturbed by suppression and traffic). If no one photographed the door and marks before they were disturbed, the evidence may be irretrievably compromised. (b) The physical evidence can supply: that pry marks of a certain class (a tool of a given type/width) are present, and — if documented in place — the state of entry. It cannot supply: who made the marks, the specific individual tool (unless rare individual marks exist), or whether the marks predate the fire, absent further analysis. Honest status: the items are logged and categorized, but not yet interpreted — meaning unknown.

Chapter 4 — Worked Solutions (daggered † and odd-numbered exercises)

Model answers for the answers appendix. Daggered items: 2, 5, 11, 15, 19, 23, 26, 30, 32, 36. Odd items also worked below. Student wording will vary; award credit for the reasoning, especially the conformance-vs-correctness and necessary-vs-sufficient distinctions.

Exercise 4.1

A section (or unit) is a specialized sub-laboratory devoted to one discipline (DNA, drug chemistry, latent prints, etc.), with its own analysts, instruments, and standards; the laboratory is the whole facility housing several such sections. It is a "federation of labs" because the disciplines are as different from one another as forensic science's fields are (Ch. 1) — the DNA analyst, drug chemist, and firearms examiner share a building, not a method — so reliability and validity must be judged section by section, not for "the lab" as a monolith.

Exercise 4.2 †

Accreditation is the formal recognition, by an independent body, that a laboratory operates a documented quality-management system and is competent to perform the tests within its declared scope. It certifies that the lab has and follows a recognized quality system (conformance). It does not certify that any particular result is correct, nor that the underlying method is scientifically valid — a lab can be perfectly accredited and still faithfully run an invalid method such as bite-mark comparison.

Exercise 4.3

ISO/IEC 17025 specifies the general requirements for the competence of testing and calibration laboratories — documented procedures, validated methods, controlled records, trained/authorized personnel, calibrated equipment, traceable measurements, handling of nonconforming work, internal audits, and management review. It applies to testing and calibration labs generally (not only forensic ones). In the U.S., bodies such as ANAB (the ANSI National Accreditation Board) and A2LA grant accreditation to it.

Exercise 4.5 †

Proficiency testing is a formal exercise in which an analyst is given samples whose correct answers are known to the test provider and is graded on reaching the right result, used to measure competence. A declared test is known by the analyst to be a test; a blind test is slipped into the normal casework stream disguised as a real case. The difference matters because people behave differently under observation: a declared test measures performance when the analyst knows they are being graded and can slow down and double-check, while a blind test measures performance "in the wild," under ordinary backlog pressure — which is where real error lives. Blind testing is therefore far more revealing (and far less common).

Exercise 4.7

Any four of: accuracy (correct answer on known samples); precision/reproducibility (same answer when repeated, by the same analyst and across analysts/days/instruments); sensitivity / limit of detection (smallest amount reliably detected); specificity (responds to the target and not to look-alikes); robustness/ruggedness (holds up under small deliberate variations); known error rate / limitations (when and how often it fails).

Exercise 4.9

A staff elimination database is a set of DNA profiles of everyone who works in (or has access to) the lab. It solves the contamination-attribution problem: if a staff member's DNA turns up in a result, it can be recognized as contamination and excluded, rather than mistaken for a perpetrator's profile.

Exercise 4.11 †

The error is treating accreditation (a statement about process/conformance) as if it were a statement about the correctness of a specific result. Correction: accreditation means an outside body verified that the lab has and follows a recognized quality system within its declared scope — which makes a correct result more likely but does not guarantee it. The result's reliability still depends on the method's validity, whether the controls were run and passed, whether contamination was guarded against, and whether the analyst was insulated from bias. "Accredited" is the floor, not the proof.

Exercise 4.13

Every result in the batch must be treated as invalid and not reported until the cause is found, because a contaminated/failed negative control means something is wrong with the run (contaminated reagent, dirty surface, carryover) and no unknown in that batch can be trusted. The temptation under backlog pressure is to dismiss the anomaly as "obvious" and report the unknown anyway because "it's clearly the drug." Giving in leads down the road to §4.6: the control exists precisely to override the analyst's expectation, and an analyst who stops believing their controls has stopped doing science.

Exercise 4.15 †

The same molecular sensitivity that lets DNA typing detect the few cells a genuine perpetrator left behind will equally detect the few cells shed by an analyst, a technician, or a previous sample — so a contaminant is faithfully amplified and reported as if it belonged. The tie to a §4.5 control: an extraction blank (or reagent blank) — material carried through the whole process containing no evidence — must come up clean; if it shows DNA, contamination entered the process and the run is suspect. The blank is what converts an invisible contamination risk into a detectable one.

Exercise 4.17

• Positive control: a sample known to contain the target; it should test positive. If it does not, the test/reagents failed and results are invalid.
• Negative control: a sample known to be free of the target; it should test negative. If it tests positive, contamination or a reagent problem is present and the run is suspect.
• Reagent blank: the reagents run with no sample; they should come up clean. If they do not, the reagents themselves are contaminated.

Exercise 4.19 †

Honest rewrite: "Our laboratory is accredited, which means an independent body has verified that we follow a recognized quality system and are competent in this type of testing. That makes our results more reliable than work done without such a system, but accreditation does not by itself guarantee this particular result is correct — that also depends on the validated method, the controls we ran, and our safeguards against contamination and bias."

Exercise 4.21

The flaw, in §4.4 terms, is conflating developmental validation (the manufacturer showing the method can work in principle) with internal validation (this specific lab demonstrating that it can run the method correctly with its own instruments, reagents, and staff). The manufacturer's data is necessary background but not sufficient: until the lab performs and documents its own internal validation, it has not shown it has the technique under control, and using it on real cases is premature.

Exercise 4.23 †

"Contamination is impossible in our laboratory" is indefensible: contamination is always a possibility because methods are sensitive, samples and surfaces are shared, and humans shed cells. The honest version is that a well-run lab works to make contamination detectable and rare — through physical separation and unidirectional workflow, protective equipment, decontamination, blanks/controls, and elimination databases — so that when contamination occurs, a control catches it rather than a jury believing it. A blank that flags contamination is a success of the system; the real danger is undetected contamination.

Exercise 4.25

Example (blood swab): Intake — seal broken or mislabeled. Storage — biological evidence not refrigerated and degrades. Triage — mis-prioritized and delayed past a deadline. Examination — contaminated pipette or a skipped control. Technical/administrative review — a rubber-stamped review that catches nothing. Report/testimony — the result overstated on the stand. (Any reasonable item per stage.)

Exercise 4.26 †

Three safeguards and their gaps: - Accreditation — verifies a quality system exists and is followed (conformance), but does not guarantee any specific result is correct or the method valid. - Controls (QC) — verify a given run worked against known truth, but only catch problems they are designed to detect, and only if they are actually run and heeded (the scandals show controls being evaded). - Proficiency testing — measures competence on known-answer samples, but declared tests are taken under observation and may be unrepresentatively easy, so a high pass rate is not a real-world error rate. Each raises the floor; none guarantees the result — "necessary and insufficient."

Exercise 4.27

It means the same mechanism — interpretation drifting toward an expected/desired answer — operates at two scales. At the individual scale, an analyst who knows the wanted answer is biased toward it (fixed by context management and blind analysis, Ch. 31). At the institutional scale, an entire lab housed inside the prosecution's structure is structurally aligned with one side, shaping which results get questioned and what "a good job" feels like (fixed by structural independence, Ch. 38). The two reinforce each other; neither fix alone is sufficient.

Exercise 4.29

The validity spectrum gives a method its ceiling (how reliable it could be at best); the lab's quality systems determine whether a given result reaches that ceiling. DNA's developmental and internal validation, running controls, contamination defenses, and a quantified RMP let it approach its high ceiling — but a mislabeled tube, a contaminated reagent, or a misinterpreted mixture can still drop a real DNA result far below it. Position on the spectrum is a cap, not a guarantee; the lab is what closes (or squanders) the gap to it.

Exercise 4.30 †

New risks from a county/state split: a second chain of custody with an extra handoff (Ch. 2's controls — a seam where custody can break); transport and re-packaging exposing biological evidence to degradation or contamination (Ch. 3's integrity controls and Ch. 4's packaging/blanks); two separate queues compounding delay (Ch. 4 backlog); and divided accountability, where each institution's documentation must be complete for the whole to hold. Ch. 2 addresses custody and documentation; Ch. 3 addresses evidence integrity and packaging; Ch. 4 addresses backlog and contamination control.

Exercise 4.31

Problems: the note supplies domain-irrelevant, biasing context (it's "our guy"), pressures a predetermined result ("confirm"), and imposes a deadline that invites corner-cutting — every ingredient of §4.1/§4.5 bias and §4.6 misconduct. What you should do: run the validated test honestly with full controls, report whatever the science shows (including "cannot confirm" or "inconclusive"), and ideally have the biasing note removed from what you see (context management). What you must not do: let the expectation steer the interpretation, skip controls to hit the deadline, or report a "confirmation" the data does not support.

Exercise 4.32 †

The violated principle is the technical review — the independent second-analyst check that work is sound before a report issues (§4.1). Rubber-stamping it removes one of the building's most important controls; the risk is that errors (or fraud) pass undetected into reports and into court, exactly as happened in the §4.6 scandals. The analyst's obligation is to refuse to participate in a sham review, to insist reviews be genuine, and — if the practice continues — to report it through the appropriate channel, because a review that does not review is worse than none (it lends false assurance).

Exercise 4.33

The decision defeats the purpose of proficiency testing, which is to measure real-world competence, not to protect the lab's image. Declining blind testing because a hidden failure "might look bad" prioritizes appearance over accuracy and guarantees the lab never learns how it performs under ordinary conditions — the conditions in which error actually occurs. A high declared-test pass rate is precisely the reassurance §4.3 warns is least informative. An honest lab wants to find its hidden failures; that is the point.

Exercise 4.35

It reveals that declared proficiency testing measures capability under observation, not integrity or fitness during routine casework. An impaired analyst can summon enough performance to pass a known test while being unfit for, or dishonest in, day-to-day work (the Farak lesson). The safeguard is poorly matched to the risk: detecting impairment or misconduct requires blind testing, random re-analysis, supervision, and auditing of actual casework — not a periodic test the analyst knows is coming.

Exercise 4.37

(a) It is carryover / cross-contamination between samples — material potentially transferring from the gas can to the tool (or vice versa) via the shared, uncleaned bench and unchanged gloves. (b) A clean extraction blank shows that no contaminating DNA was introduced through the extraction process and reagents for the swabs tested — it documents that this particular route came up clean. It does not retroactively prove the earlier bench handling was flawless, prove the result's source, or rule out contamination by a route the blank doesn't monitor. The clean blank is reassuring evidence about one specific risk, not a guarantee about the whole history of the item.

Exercise 4.39

Because the chapter is about the machinery the evidence passes through, not about analyzing the evidence — and honest forensic reasoning refuses to conclude ahead of data (Ch. 1's "capital mistake to theorize before one has data"). After Ch. 4 we know which labs will handle the case, under what controls, with what backlog and what near-miss — exactly the context needed to weigh later results — but not a single test has been run. Declaring anyone included or excluded now would be theorizing without data; keeping the status at "process framed; nothing concluded" is the discipline the whole book teaches.

Chapter 5 — Worked Solutions (daggered † and odd-numbered items)

Exercise 5.1

The Frye standard: novel scientific evidence is admissible only if the method has gained general acceptance in its relevant scientific field. The single question it asks: do the experts in this field generally accept this method as reliable?

Exercise 5.2 †

The Daubert standard: under the Federal Rules of Evidence, the trial judge must independently determine whether proposed expert scientific testimony is both relevant and reliable before the jury hears it. Difference from Frye: under Frye the judge defers to the field's consensus (a sociological question — "do they accept it?"); under Daubert the judge must make an independent reliability assessment (a scientific question — "is it actually sound, and how do we know?"). Reputation gives way to reliability.

Exercise 5.3

FRE 702 is the Federal Rule of Evidence governing expert testimony. Beyond permitting a qualified expert to give an opinion, the post-2000 version requires that the testimony (a) be based on sufficient facts or data, (b) be the product of reliable principles and methods, and (c) reflect a reliable application of those principles and methods to the facts of the case. The third element — reliable application — is the one most often neglected in court.

Exercise 5.5

The four classic Daubert factors: (1) testability/falsifiability; (2) known or potential error rate; (3) peer review and publication; (4) general acceptance in the relevant scientific community. (Flexible and non-exclusive, per the Court.)

Exercise 5.6 †

An error rate is the frequency with which a method produces a wrong result — a false positive (declaring a match that isn't there) or a false negative (missing one that is). "We have never had a case overturned" fails to state one because (a) it is not a measured frequency derived from testing against known ground truth, and (b) it only counts errors that came to light — undetected errors and wrongful results never caught are exactly what it cannot see. A real error rate comes from blind studies where the correct answer was known in advance; "never caught wrong" is the absence of validation, dressed as a result.

Exercise 5.7

Peer review means scrutiny of a method or result by independent experts who did not develop it, typically as a condition of publication. Its absence tells you the method has never been exposed to hostile outside criticism — it has circulated only among its own practitioners, dodging precisely the test science relies on to catch error. (Note the limit: peer review is a marker, not a guarantee; bad work passes and good work is rejected.)

Exercise 5.9

The "Daubert trilogy": Daubert v. Merrell Dow Pharmaceuticals (1993, the reliability/gatekeeper standard), General Electric Co. v. Joiner (1997, the deferential "abuse of discretion" standard of appellate review), and Kumho Tire Co. v. Carmichael (1999, extending gatekeeping to all expert testimony).

Exercise 5.11

The four differences: (1) Clock — science is iterative and open-ended ("we don't know yet" is acceptable), the law is terminal and deadline-bound (a verdict must be reached now and, once reached, stands). (2) Goal — science seeks generalizable truth; the law seeks case-specific resolution (what happened to this person). (3) Standard of proof — science quantifies uncertainty; the law uses verbal standards ("beyond a reasonable doubt," "preponderance") it resists turning into numbers. (4) Posture toward authority — science is "take no one's word for it"; the law admits expert opinion because of who gives it (the ipse dixit temptation).

Exercise 5.12 †

Frye's circularity, with bite-mark matching as the example: forensic odontologists come to accept the method within their own community → on the strength of that "general acceptance" courts admit it → its repeated courtroom success becomes part of why the field continues to accept it → and the circle closes without anyone ever running a controlled validation study. It is "almost perfectly designed" to admit old junk because the "relevant field" is the small set of practitioners who make their living doing the technique and have every professional reason to believe in it — exactly the people least able to see its flaws. Frye measures consensus, and a whole field can be confidently wrong.

Exercise 5.13

Two directions: (a) Daubert can now exclude a method that Frye would have admitted — a generally accepted technique that is nonetheless unreliable (no testability, no error rate). (b) Daubert can, in principle, admit a method Frye would have excluded — a brand-new, not-yet-accepted technique that is demonstrably sound (published, measured error rate). The gate is supposed to be about quality, not seniority.

Exercise 5.15

The pattern disciplines argued their work was "experience/skill, not science" hoping to escape Daubert reliability scrutiny entirely, on the theory that the factors applied only to laboratory "science." Kumho Tire foreclosed it: gatekeeping applies to all expert testimony — technical and experience-based included — so an examiner cannot avoid the error-rate question by relabeling a scientific-sounding conclusion as seasoned intuition.

Exercise 5.16 †

General acceptance was demoted from being the sole test (under Frye) to being one factor among four (under Daubert). It still counts as evidence of reliability; it is simply no longer sufficient by itself or strictly necessary. The chapter argues the demotion often fails in practice because courts, faced with an unfamiliar technical record and a confident credentialed expert, frequently let "generally accepted in the forensic community" do the same heavy lifting it did under Frye — smuggling a method past the gate on reputation alone, contrary to Daubert's design.

Exercise 5.17

Foundational validity = whether the method, in general, reliably does what it claims (the question the four factors mostly target). Reliable application = whether the method was performed correctly in this case (contamination, mislabeling, misinterpretation). The post-2000 FRE 702 covers the second with its "reliable application... to the facts" requirement. It is the more neglected because courtroom attention concentrates on whether the method is valid, while a great deal of real-world error lives in how the method was executed — even a valid method like DNA can be ruined by a bad sample or a misread mixture (tie to Ch. 4).

Exercise 5.19 †

Running the toolmark offer through the factors: Testability — dodged; "I can tell by looking" plus twenty years' experience proposes no test that could prove the call wrong. Error rate — dodged; no measured, ground-truth figure is offered (experience is not an error rate). Peer review — dodged; "every examiner in the lab agreed" is in-house consensus, not independent published scrutiny. General acceptance — this is the only factor the statement addresses, and even then it conflates "accepted by my colleagues" with acceptance by the broader scientific community. The offer satisfies (at most) one factor — the weakest, most Frye-like one — and dodges the three that actually measure reliability. (Firearms/toolmark validity is PCAST-contested; developed in Ch. 15–16.)

Exercise 5.21

The four cross-examination questions, one per factor: Testability — "Is there any experimental result that could have shown your bite-mark conclusion to be wrong; has that test been run?" (Exposes that the core claim resists falsification.) Error rate — "What is the measured error rate of bite-mark identification, from blind studies against known ground truth?" (Exposes that there is no valid figure, or an alarmingly high one.) Peer review — "Has the specific identification claim — that these teeth and no others made this mark — been validated in the peer-reviewed literature, as opposed to relying on the general prestige of dentistry?" (Exposes the absence of published validation.) General acceptance — "Is bite-mark identification still accepted across the forensic-science community, given the exonerations and the NAS/PCAST findings?" (Exposes the eroding, practitioner-only acceptance.)

Exercise 5.22 †

Under Frye: inadmissible — the method is novel and not yet generally accepted, and acceptance is the whole test. Under Daubert: potentially admissible — it is testable, has a measured error rate from blind studies, and is peer-reviewed; three of four factors are strongly satisfied, and general acceptance is only one non-decisive factor (newness alone is not disqualifying). The contrast reveals the core difference: Frye rewards seniority/consensus and would keep out a sound new method, while Daubert rewards demonstrated reliability and can let it in. It is the mirror image of the junk-science problem — Frye keeps good new science out for the same reason it lets old junk in.

Exercise 5.23

(a) GC-MS confirmation of gasoline → clears an honest gate; weakest factor: essentially none of the four is weak (grounded in analytical chemistry) — the vulnerability is application (was the sample handled and run correctly), not foundational validity. (b) "Pour pattern" opinion without lab confirmation → should not survive; weakest factor: testability (the indicator has been tested and failed) and error rate (none valid). (c) Partial latent fingerprint → admissible but contestable; weakest factor: error rate (real, non-zero, and worsened for partial prints) and the subjectivity of the comparison. (d) Soil "consistent with" the scene → admissible but contestable; weakest factor: error rate / the probabilistic, class nature of the association (it supports "consistent with," not "came from").

Exercise 5.25 †

The logical move: substituting a method's legal history ("admitted for fifty years") for its scientific validity — the "grandfather problem." The sentence may be perfectly true and still tell you nothing about whether the method works, because admission is a legal act by generalist judges (often on precedent), not a scientific finding. A method can be admitted for a century without anyone ever running the validation study; longevity is evidence of the method's acceptance and entrenchment, not its accuracy. The honest follow-up is the four factors, none of which is "how long has it been admitted?"

Exercise 5.27

The endorsing epigraph is Justice Blackmun's: "Vigorous cross-examination, presentation of contrary evidence, and careful instruction on the burden of proof are the traditional and appropriate means of attacking shaky but admissible evidence." Why it can fail for weak forensic evidence: the line assumes the jury can reliably tell strong forensic evidence from weak once both are before it. The CSI effect (Ch. 1) says jurors are miscalibrated — primed by fictional infallibility to over-trust confident, credentialed forensic testimony and hear "consistent with" as "certain." Cross-examination only works as a safeguard if the audience can register the damage it does; against a confident examiner and a primed jury, "let the jury weigh it" can let weak evidence carry far more weight than it deserves.

Exercise 5.29

(a) "It's been peer-reviewed and published" → peer review factor (genuinely addressed, if the publication validated the specific claim). (b) "Everyone in my field does it this way" → general acceptance factor — and only the narrow, practitioner-only version of it. (c) "We ran blind proficiency tests and missed 1 in 50" → error rate factor (genuinely addressed — this is what a real error rate looks like). (d) "There's no experiment that could show me wrong — I just see the match" → testability factor, failed (the claim is unfalsifiable, which is disqualifying, not reassuring).

Exercise 5.31 †

It matters because the law's standards of proof are verbal ("beyond a reasonable doubt," "preponderance") and courts have resisted attaching numbers to them, while much forensic testimony is, at bottom, a probability claim (a random match probability, a likelihood ratio). So a numeric statement of uncertainty must be communicated to, and weighed by, a fact-finder using a non-numeric standard — and the translation is where distortion enters (the prosecutor's fallacy, over- and under-weighting). The gate decides whether the probabilistic evidence comes in at all; Chapter 9 takes up what happens when a number meets a verbal standard and how easily it is mis-mapped.

Exercise 5.33

The chapter stresses the Willingham investigators "were not frauds" to make its central point about how junk science enters court: not usually as deliberate deception but as the sincere, good-faith application of an unvalidated method by people trained to believe in it. This is more dangerous than fraud because (a) sincerity is persuasive — a confident, honest expert is more convincing than a liar; (b) the error is invisible to the people committing it, so they neither hedge nor welcome blind verification; and (c) a method that "cannot imagine being wrong" never measures its own error rate. Sincere belief substituting for measured validity is the engine of the bias (Ch. 31) and wrongful-conviction (Ch. 34) chapters.

Exercise 5.34 †

The honest resolution: you may testify to what your examination observed and to a comparison result at its true strength — "I found agreement at these features and no unexplained differences; the mark is consistent with the suspect's source," paired with whatever is known about how often such agreements occur by chance. You must refuse to testify that the method "individualizes" the mark to your suspect "to the exclusion of all others" or to a "reasonable degree of scientific certainty," because a method with no published error rate has not earned that claim — saying it would assert a precision the science cannot support, the exact overstatement this book is built to resist. That your supervisor and field accept the stronger language does not make it valid (Frye's circularity); Kumho Tire means the reliability question attaches whether you call your work science or skill. The ethical floor: claim only what the four factors can back.

Exercise 5.36 †

The completed admissibility column (illustrative entries): DNA from the gas can → "clears an honest gate"; weakest factor: application (sample integrity, mixture interpretation), not foundational validity. GC-MS accelerant confirmation → "clears an honest gate"; weakest factor: application. Fire-origin determination → split: the valid fire-dynamics + confirmed-residue path is "admissible/clears"; any folklore "pour pattern" reasoning is "should not survive" (fails testability/error rate). Fingerprint → "admissible but contestable"; weakest factor: error rate (worse for a partial). Soil/pollen → "admissible but contestable"; weakest factor: probabilistic/class error rate. Autopsy findings (cause/manner) → "admissible but contestable," strong where grounded in pathology, weaker on interpretive calls. The fire is the entry that most demands caution in this case because the fire is the heart of the matter and the chapter's anchor (Willingham) was executed on exactly the fire-folklore reasoning an early investigator might reach for here.

Exercise 5.37

A two-sentence response to the detective: "Right now the science can support only that the fire's burn patterns are consistent with — not proof of — a deliberately set fire, because 'crazed glass' and floor-level 'pour' patterns are routinely produced by an ordinary fire reaching flashover, with no accelerant; that exact reasoning is what executed Cameron Todd Willingham. A valid arson finding here has to run through fire-dynamics analysis of origin and cause and instrumental confirmation of ignitable-liquid residue (Chapters 22–23) — not through the indicators alone."

Exercise 5.39

A weak local Daubert gate and a weak local lab compound each other: the under-resourced lab is more likely to produce flawed or unvalidated work (or to lean on cheap, presumptive, folklore-prone methods), and a permissive local court is less likely to catch and exclude that work before it reaches a jury. The gate is supposed to be the backstop for lab error; the lab is supposed to generate evidence solid enough to survive the gate. When both are weak, flawed evidence is both more likely to be generated and more likely to be admitted — the two failures multiply rather than add. Chapter 38's reforms (independence, accreditation, blind verification) target both ends at once.

Chapter 6 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 6.1

A wrongful conviction is the conviction of a person who did not in fact commit the crime — a failure of factual accuracy. It differs from a conviction reversed for a procedural error (e.g., an illegal search, a bad jury instruction) because such a reversal concerns the legality of the process, not the defendant's actual guilt or innocence; a guilty person can win a procedural reversal, and an innocent person can be convicted through a perfectly lawful process.

Exercise 6.3 †

Junk science is a method or claim presented in court as valid science despite lacking a demonstrated scientific basis — no measured error rate, no validated procedure, no foundation beyond the practitioners' confidence. An honest examiner can be its vehicle because the method itself is hollow: a sincere bite-mark examiner who genuinely believes he can match teeth to a wound will testify truthfully to what he believes and still mislead the jury, because the belief was never validated. His sincerity makes him more persuasive and therefore more dangerous than an obvious fraud. Example: an examiner in good faith testifies that a defendant's teeth "match" a bite mark, when no validated method exists for that comparison.

Exercise 6.5

The 2009 NAS report found that, with the sole exception of nuclear DNA analysis, no forensic method had been rigorously shown to reliably and consistently demonstrate a connection between evidence and a specific individual or source.

Exercise 6.7

Foundational validity asks whether the method itself has been shown, by well-designed empirical (black-box) studies, to distinguish true matches from false ones with a measured, acceptable error rate — a property of the method, independent of any examiner. Validity as applied asks whether that (foundationally valid) method was actually used correctly in the specific case, by a competent examiner, following the validated procedure, within the conditions the studies covered. The first is about the method; the second is about this particular use of it. A foundationally valid method misapplied can still produce an invalid result.

Exercise 6.9

In rough order of frequency: (1) mistaken eyewitness identification — the most common single factor; (2) flawed or misapplied forensic science — present in roughly half of the DNA exonerations; (3) false confessions or admissions; (4) informants / jailhouse "snitches"; (5) official misconduct. These usually co-occur rather than appearing alone.

Exercise 6.11

The track record is not validation because in real casework you almost never learn the ground truth, so you cannot count your errors. The pilot who never finds out which of his landings would have crashed cannot estimate his crash rate from experience, however many flights he logs. "No known errors" means no errors were discovered — which, in a system that rarely surfaces ground truth, is exactly what you would observe whether the true error rate were zero or substantial. Validation requires designed studies with known answers so that errors become countable.

Exercise 6.12 †

DNA produced exonerations at scale because of two properties no prior forensic method combined: (1) retroactive applicability — it can be run on old, stored biological evidence in already-closed cases; and (2) clean, near-categorical exclusion — a mismatch at enough loci is, barring lab error, conclusive, so a result excluding the convicted person cannot be argued away the way an eyewitness account or a hair "match" can. Together they let DNA audit the system's past by speaking to ground truth.

Exercise 6.13

The two sub-categories are: (a) junk methods that should never have been admitted — e.g., bite-mark comparison — where the method itself has no validated basis, so even a careful examiner misleads; and (b) valid methods, overstated or misapplied — e.g., a sound serological method reported beyond what the data support, or a real comparison testified to with fabricated certainty — where the technique may be sound but the testimony is the failure.

Exercise 6.15

The additional element is the error rate measured by the validation (black-box) studies. Compliant testimony, e.g.: "The latent print is consistent with the defendant; in the published studies of this comparison, examiners declared erroneous matches at a rate on the order of about one in several hundred to one in a thousand comparisons. I cannot say it is his to the exclusion of all others, and I am not claiming a zero error rate." (Treat the figure as illustrative of the right magnitude.)

Exercise 6.16 †

The risk is contextual / confirmation bias (theme 3): once the analyst is told the "expected" answer and the suspect's identity, interpretation of an ambiguous print can drift toward the expected conclusion, with no dishonesty required. The safeguard is context management / blind analysis — withholding domain-irrelevant information (who the suspect is, that there's a confession) from the examiner, and using sequential unmasking. Developed in full in Chapter 31.

Exercise 6.17

They are biased because DNA exonerations are only possible where biological evidence (a) was generated by the crime, (b) was collected, and (c) survived in testable condition — which describes a particular slice of cases (many sexual assaults and homicides). Systematically missing are crimes that leave no testable biology — e.g., many robberies and assaults built on eyewitness ID — where a wrongly convicted person has no DNA "key" to their cell. So the countable exonerations undercount the true scope of wrongful conviction.

Exercise 6.19 †

Overstatements: (1) "zero error rate" — no human-judgment method has one; this is a refusal to measure. (2) "this print is the defendant's" — an individualization claim the method cannot support. (3) "to the exclusion of all others on Earth" — claims a comparison against every other source, which was never done. Defensible rewrite: "The latent print is consistent with the defendant; I found agreement in the features I compared and no unexplained differences. The validation studies for this method report a small but non-zero false-match rate. I cannot say it is his to the exclusion of all others."

Exercise 6.21

Two distinct errors: (1) a historical/legal error — for most of the twentieth century methods were admitted under a "general acceptance" standard, so admissibility reflects acceptance by the method's own practitioners, not proven accuracy; many junk methods became admissible without any validation study. (2) An error about what admissibility measures — that a court lets evidence in says the judge found it passes the legal gate, not that the underlying science is valid. Admissibility is a legal status, not a scientific verdict.

Exercise 6.23

This violates the PCAST 2016 recommendation, which held that an examiner from a foundationally valid discipline may report a match only by stating the relevant error rate from the validation studies — not by claiming "a reasonable degree of scientific certainty" (a phrase PCAST and others criticized as scientifically meaningless). The testimony should have reported the measured error rate and avoided both the certainty phrase and any individualization claim.

Exercise 6.25 †

Obligations and options: As a forensic scientist your duty runs to the evidence and the court, not to the prosecution or to institutional convenience (forward-ref Chapter 38 ethics; Chapter 30 the duty to the court). If a method lacks established foundational validity for the specific claim your reports make, continuing to report as before — even if courts still admit it — means knowingly producing testimony stronger than the science supports. Options, roughly in order: change the report language to match what the validation evidence supports (state limits, drop individualization, give the error rate or note its absence); document your objection in writing; escalate within the lab's quality system; seek guidance from the relevant standards body (e.g., OSAC, Chapter 38); and, if required to testify, confine your testimony to defensible claims regardless of pressure. "The courts still accept it" is an admissibility fact (Chapter 5), not a scientific justification.

Exercise 6.27

This reflects the CSI effect's first direction — the expectation that real cases come with forensic evidence (§1.2). Whether the reasoning was sound cannot be determined from the statement alone: if the non-forensic evidence was genuinely weak, demanding more before convicting is appropriate caution; if it was strong (e.g., credible eyewitnesses plus recovered property in a gloved-burglar case), then treating the ordinary absence of forensic evidence as doubt is the CSI-effect error. The point is that the absence of forensic evidence is normal and is not, by itself, exculpatory.

Exercise 6.29

A defensible position: primary moral responsibility lies with the system and the court that admitted the method, because the gatekeeping function (Chapter 5) exists precisely to keep unvalidated science from the jury; the sincere examiner is culpable to the degree they claimed more certainty than even the method's defenders supported, and the discipline bears collective responsibility for never running the validation studies. Good-faith belief mitigates individual blame but does not erase the harm, and it does not relieve the examiner of the duty to confine testimony to what is defensible. (Alternative positions are acceptable if they engage the same actors and the gatekeeping logic.)

Exercise 6.30 †

A strong answer: "Match my testimony's strength to the validation evidence — never claim more than the science supports, and welcome blind verification." Justification: it directly serves theme 1 (reserve strong claims; lean on exclusion) and theme 2 (the validity spectrum dictates the honest verb), and it pre-empts theme 3 (bias) by inviting independent checking. It is the single discipline that, applied consistently, would have prevented the overstatement at the heart of nearly every wrongful conviction in this chapter.

Exercise 6.31 †

Warning signs in the Renner confession: his youth (22); an extraordinarily long interrogation (about eleven hours); the interrogation was partly unrecorded (the first several hours); much of it occurred without counsel; and the confession contained crime-scene details that may have been supplied by the interrogators (contamination), so it cannot be treated as independent corroboration. Bias cascade: the early "accidental fire" frame, once it cracked, was replaced by a search for a convenient local; the detectives' expectation that Renner was responsible shaped how the interrogation was conducted and what details were fed to him; the resulting confession then appeared to confirm the suspicion that produced it — each step feeding the next, the case hardening around Renner before the physical evidence had been asked whether it excludes him.

Exercise 6.33

Treating a confession as a fixed point makes it the premise the rest of the evidence must confirm, which inverts the proper logic. A confession is itself a claim — one the exoneration data show is sometimes false — and it must be tested against the independent physical timeline, not used to interpret that timeline. If the physical evidence (e.g., cell-site placing Renner elsewhere; DNA excluding him) contradicts the confession, the science governs and the confession must yield. Using the confession as confirmation-bait risks exactly the cascade that convicts the innocent: every later result gets read in light of an assumed guilt the confession seemed to establish.

Exercise 6.35

Mapping Mill Creek onto the §6.3 causes at this stage: false confession — present and central (Renner). Official conduct / interrogation pressure — possible, given the length and partial non-recording. Flawed/overstated forensics — not yet in play (the physical analysis hasn't been done), but it is the danger to guard against going forward. Mistaken eyewitness ID — looms (a neighbor's "tall stranger's truck" surfaces in Chapter 32) but is not yet load-bearing. Informants — not in play. The danger is currently highest at the confession/interrogation node, and the standing instruction is to let the not-yet-collected physical evidence exclude Renner if it can, rather than letting the confession pre-color it.

Chapter 7 — Worked Solutions

Solutions to the daggered (†) and odd-numbered exercises. Section letters match exercises.md.

Exercise 7.1 (odd)

Acceptable one-sentence definitions: DNA — the molecule carrying genetic instructions, identical across a person's nucleated cells. STR — a short DNA sequence repeated a variable number of times at a fixed locus, the variation forensic typing reads. Locus — a named physical address in the genome targeted for typing. Allele — the specific version (repeat number) at a locus; two per autosomal locus. PCR — a technique that copies a target DNA region exponentially. Electropherogram — the peak-graph readout of a profile from capillary electrophoresis. DNA profile — the full set of alleles across all tested loci. CODIS — the U.S. national DNA database that searches unknown profiles against offender/casework profiles. RMP — the chance a random unrelated person coincidentally shares the profile.

Exercise 7.3 (odd)

More than ~99.9% of the human genome is identical between individuals. Forensic typing targets the varying, non-coding regions for two reasons: (a) practically, the variation is what allows discrimination between people — coding genes are too similar to tell people apart; (b) ethically, the non-coding "junk" regions reveal essentially nothing about health, traits, or appearance, so the test answers "whose cells?" while staying blind to private medical information.

Exercise 7.4 † (and the four stages)

The four major stages, with one failure mode each: (1) Extraction — liberate and purify DNA from cells; failure mode: co-purified inhibitors (denim dye, soil, heme) or contamination (stray analyst/carryover cells). (2) Quantification — measure how much human DNA is present; failure mode: proceeding with too little (incomplete profile) or too much (instrument overload). (3) PCR amplification — exponentially copy the target STR loci; failure mode: amplifying contaminant template along with evidence (exponential copying of any DNA present). (4) Capillary electrophoresis / electropherogram — sort fragments by size and read peaks; failure mode: interpretation errors at the edges (noise vs. real minor peak; off-ladder calls).

Exercise 7.5 (odd)

Not applicable as written (recall item). See 7.7 below.

Exercise 7.7 (odd)

CODIS holds, in separate indexes: convicted-offender (and often arrestee) profiles, forensic casework profiles (unknown crime-scene profiles), and other categories. An offender hit points to a known individual whose profile is on file (a candidate source — to be confirmed). A forensic hit links two unsolved crime scenes to the same unknown person, even when no one yet knows who that person is.

Exercise 7.8 †

As a fraction: 1/1,200,000,000. In plain terms: "you would expect to find this profile in roughly one person out of more than a billion — far more than the population of the United States." The book insists on both because the bare fraction is precise but cold and easily misheard as a statement about guilt; the human framing makes the rarity understood rather than merely impressive, and it is the antidote to a jury being overwhelmed into the prosecutor's fallacy.

Exercise 7.9 (odd)

PCR doubles the copies of its target each cycle, so one molecule becomes hundreds of millions after ~28–30 cycles — the "gift" that lets a trace become measurable. The "standing risk" is that the same exponential process copies whatever template is present, including a few contaminating molecules, so a single stray cell can be amplified into a detectable "profile" that was never relevant to the crime. Sensitivity and contamination-risk are the same property seen from two sides.

Exercise 7.11 † (and odd)

DNA sits at the top because of three properties the pattern methods lack: (1) an understood biological basis — we know why DNA varies (inheritance, mutation, recombination) and can model it, so the claim of rarity is grounded, not asserted; (2) a quantified error structure — the RMP is a calculation from measured population frequencies, and lab error has been studied, so the uncertainty is stated rather than hidden; (3) objectivity at the core measurement — fragment size is read by an instrument against a standard, not judged by eye, removing much of the examiner subjectivity that plagues fingerprints and toolmarks. Each buys defensibility: the method's central claim can be interrogated and checked.

Exercise 7.13 (odd)

A negative control is a blank sample carried through every step (extraction, PCR, CE) with no evidentiary DNA added. If a profile or peaks appear in that blank, DNA was introduced somewhere in the process — i.e., contamination — and the run is suspect. It is designed specifically to catch contamination introduced during laboratory processing (not coincidental matches and not relatives).

Exercise 7.15 † (and odd)

The coincidental-match risk is the chance an unrelated stranger happens to share the profile — that is the RMP, often astronomically small (1 in billions). The laboratory-error risk is the chance of a swapped tube, contamination, or a mislabel — a real, human error rate (example: two evidence tubes transposed during accessioning). The lab-error risk is realistically larger than a one-in-a-billion coincidence, and it does not shrink by typing more loci, because adding loci only drives down the coincidence probability; it does nothing about the independent chance that the wrong sample was tested or labeled. An honest expert keeps the two risks separate and never lets the astronomical coincidence figure stand in for the everyday risk of a mistake.

Exercise 7.16 †

The suspect is excluded on the strength of that one locus. At D3 the evidence source must carry alleles 15 and 16; the suspect carries 14 and 17 and so lacks both required alleles. Barring a documented laboratory error or a rare biological artifact, a true source cannot be missing an allele the evidence shows — so a single clean, reproducible mismatch excludes with near-certainty, regardless of agreement at every other locus. This is the asymmetry: a mismatch can refute the whole inclusion, while a match at one locus could never, on its own, prove a source.

Exercise 7.17 (odd)

With ~1 in 6 per locus and treating the loci as independent: $(1/6)^5 = 1/7{,}776$ — about 1 in 7,800 from five loci. (Numbers illustrative; real calculations use measured per-locus frequencies, not a flat 1-in-6.) The point is the shape: five modest rarities multiply into one substantial one, and twenty loci push the product into the billions or beyond.

Exercise 7.18 †

The detective has committed the prosecutor's fallacy — transposing "given innocence, the chance of a matching profile is 1 in a billion" into "given a match, the chance of innocence is 1 in a billion." Honest replacement: "This profile is found in approximately one in a billion unrelated individuals in the relevant population; that tells you how rare the profile is, not the probability that the defendant is innocent, which depends on the other evidence in the case."

Exercise 7.19 (odd)

A model answer renders the six fields for the hypothetical (tall peak at 12, short peak at 9). The key analytical move: the marked height imbalance between the two peaks is exactly what a single, balanced heterozygote should not show — so before calling 9 a true second allele, you would ask whether the short peak is (a) a genuine minor allele (possibly indicating a second contributor — i.e., a mixture, Ch. 8), (b) an artifact such as stutter or pull-up, or (c) noise near the analytical threshold. In a clean single-source reference, balanced peaks are expected; imbalance is a flag to investigate, not to call.

Exercise 7.21 †

The three questions the match cannot answer: When were the cells deposited? (No timestamp — could be from before the crime.) How did they get there? (Direct contact, secondary transfer, or planting — the profile cannot distinguish them.) What was the person doing, and did they do anything wrong? Each matters because a one-in-a-trillion match is a statement about the source of the cells, and a verdict is a statement about a person's conduct; innocent timing or transfer (cf. the Lukis Anderson case) can fully explain the match without any wrongdoing.

Exercise 7.23 †

Overstatement. "The DNA proves the defendant was at the scene" claims presence and certainty the profile does not support. Honest rewrite: "The defendant's DNA was found on [the object]; this indicates his cells were on that object, but does not establish when or how they got there, or that he was present at the scene when the crime occurred."

Exercise 7.25 (odd)

Statement 25 is defensible as written — it uses "cannot be excluded / consistent with," attaches the RMP, and limits it to "unrelated individuals." No rewrite needed; it is a model of honest reporting.

Exercise 7.26 †

Overstatement. "A CODIS hit identified the perpetrator" overstates in two ways: a CODIS hit identifies a candidate source of the DNA, not a perpetrator, and the hit is a lead that must be confirmed by direct re-typing of a fresh reference sample. Honest rewrite: "A CODIS search produced a candidate; a fresh reference sample was then obtained from that person and re-typed, and that direct comparison — reported with its random match probability — is the evidence."

Exercise 7.27 (odd)

Defensible, provided the independence assumption is genuinely warranted for those loci and the population frequencies are real. The statement correctly grounds the multiplication in independence and reports a profile frequency rather than a probability of guilt. (A reviewer would still confirm the corrections for population structure were applied.)

Exercise 7.28 †

Ethical feature: by reading only non-coding regions, the test is designed to be blind to the donor's health, disease risk, traits, and ancestry — it answers "whose cells?" without exposing private medical information, minimizing the intrusion the state imposes when it types someone's DNA. This is a deliberate, principled limitation, not merely a technical convenience. Reintroduced privacy risk: investigative genetic genealogy (Ch. 8) searches different markers in databases of ordinary people's ancestry data, exposing relatives who never consented and reaching information well beyond "whose cells" — re-opening the privacy questions that standard STR typing was built to avoid.

Exercise 7.31 †

Partly both. It is fair to insist that lab error is a real risk that the RMP does not cover and that it should be acknowledged. It is an overstatement to argue the jury should therefore ignore a one-in-a-trillion match entirely — lab error is a finite rate to be evaluated (quality records, controls, re-testing), not an infinite eraser of any result. An honest report states the match and its RMP and the separate, non-zero possibility of laboratory error, letting the jury weigh both rather than pretending either risk does not exist.

Exercise 7.32 †

Model Case File entry: "A partial DNA profile was recovered from the gas-can handle and searched against CODIS. The profile establishes only that someone's cells were on the handle; it does not, by itself, identify the source, indicate when the cells were deposited, or show that whoever touched the can set the fire. On a renovation site many people handle equipment for innocent reasons, and the sample — a touch deposit on a heat-affected object — may prove to be a degraded mixture. No person of interest is included or excluded on this evidence alone." (No names; all §7.6 limits applied.)

Exercise 7.34 †

A mixture lowers interpretive confidence because the clean rule of §7.3 — one or two balanced peaks per locus = one person's genotype — breaks down: with two or more contributors you may see three or four alleles at a locus, peaks of unequal height that must be sorted into contributors, and (with degradation and low template) alleles that drop out or stutter artifacts that drop in. The objective core measurement is intact, but assigning which alleles belong to which contributor reintroduces judgment and thus error. The first question to ask the lab: how many contributors are indicated, and is a single-source interpretation even appropriate? (The full method belongs to Ch. 8.)

Chapter 8 — Worked Solutions

Solutions to the daggered (†) and odd-numbered exercises. Daggered items get fuller treatment. Answers model the honest-verb discipline: exclude / consistent with / strongly supports / inconclusive, never "proves" unless earned.

Exercise 8.1

Touch DNA is genetic material recovered from the skin cells a person sheds onto an object simply by handling it. Trace DNA is the broader umbrella for any tiny, often invisible biological deposit from which a profile might be recovered — touch DNA is one kind of trace DNA, but trace also covers other minute deposits. Touch names the mechanism (contact/shedding); trace names the scale (tiny).

Exercise 8.2 †

Degradation is damage: environmental insults (heat, moisture, UV, microbes, time) break the long DNA molecules into short fragments, so loci whose target spans a break cannot be amplified — the large STR loci fail first, giving the "ski-slope." Low template is scarcity: too few cells / too little DNA to begin with, so the random lottery of early amplification dominates and produces stochastic effects (dropout, drop-in, peak imbalance). Yes, one sample can suffer both at once — the canonical example is the cold-case gas-can handle: a touch deposit (low template) that was also heat-exposed near a fire (degraded). When they combine you get faded large loci and unreliable calls at the surviving loci, often on top of a mixture — the hardest kind of sample.

Exercise 8.3

The three stochastic effects of low-template typing: allele dropout (a true allele fails to amplify and vanishes, so a heterozygote looks homozygous or a locus blanks); allele drop-in (a spurious allele appears, usually a stray contaminating molecule, making the profile look like it contains someone it does not); heightened peak-height imbalance (the two alleles of a heterozygote amplify to very unequal heights, so peak height stops being a reliable guide to who is present).

Exercise 8.5

A DNA mixture is a sample containing DNA from two or more contributors. From counting alleles you can be certain only of a minimum number of contributors — e.g., five alleles at one locus require at least three people (since one person contributes at most two), but you cannot rule out that there are four or more who happen to share alleles. So the count is a lower bound, never the exact true number.

Exercise 8.7 †

mtDNA can be recovered where nuclear STRs fail because of one biological fact: there are hundreds to thousands of mitochondria (and mtDNA copies) per cell, versus a single nuclear genome. When degradation destroys most copies, some mtDNA copies often survive simply because there were so many to begin with. This is why mtDNA is the method of choice for rootless hair shafts, old bones and teeth, and badly degraded tissue. The price (next exercise) is that mtDNA cannot individualize.

Exercise 8.9

A Y-STR is an STR typed on the male-only Y chromosome. Its signature use is the sexual-assault mixture where a small amount of male DNA is swamped by a large amount of female DNA: because Y-STRs ignore the female contribution entirely, they can pull a male profile out of that overwhelming female background (and help count the number of male contributors). The limitation is that the Y is shared down a paternal line, so a Y-STR match is to a lineage, not an individual.

Exercise 8.11

The four IGG steps: (1) generate a dense SNP profile from the crime-scene DNA (not the ~20 STR loci, which carry no genealogical information); (2) upload to a genealogy database that permits law-enforcement matching and obtain distant-relative (2nd–4th cousin) matches; (3) build family trees from conventional records and triangulate toward the common ancestors and candidate descendants; (4) narrow by demographics to a candidate, then confirm with a conventional STR comparison on a confirmed/abandoned sample. Only step 4's STR comparison is offered as courtroom identification; steps 1–3 generate a lead.

Exercise 8.13 †

What's right: every typed locus agreeing is genuinely consistent with the suspect being the source, and seven matching loci do carry weight. What's wrong: "perfect match" implies the rarity of a full 20-locus profile. A 7-locus profile has a much larger random match probability — many more people in the population share a 7-locus profile than share a 20-locus one — so the inclusion is weaker. Honest restatement: "The suspect is included as a possible source; the profile agrees at the 7 loci that typed, but it is a partial profile, and its rarity — and therefore the strength of the inclusion — is correspondingly lower than a full profile would give. It is consistent with the suspect, not an individualization."

Exercise 8.15

At three loci you see four alleles, so at least two contributors are certain (four alleles need a minimum of two people, two each). The true number could be higher — a third contributor could share alleles with the first two and hide in the four-allele picture, and degradation/dropout could mask a contributor's unique alleles. You are certain of two; you cannot exclude three or more. (This is why contributor count is a minimum, and why misjudging it skews every downstream interpretation.)

Exercise 8.16 †

Request Y-STR typing. Reason: the sample is a male-into-female mixture where the assailant's autosomal profile is drowned by the victim's, so standard autosomal typing yields essentially the victim and nothing usable; Y-STRs ignore the female DNA and can recover the male component. Major limitation: the Y is inherited essentially unchanged down a paternal line, so a Y-STR match is shared by the man's father, brothers, sons, and paternal male cousins — it identifies a paternal lineage, not the individual, and its match statistics are weaker than autosomal STRs. So it can strongly narrow and can exclude, but it cannot individualize.

Exercise 8.17

Try mtDNA (sequencing the hypervariable regions). It can establish that the bone is consistent with a maternal lineage — i.e., consistent with a putative relative through the maternal line — and it can cleanly exclude (a different mtDNA type rules a lineage out). It cannot individualize: a match is shared by all maternal relatives, and the match frequency is far weaker than nuclear STRs. So mtDNA here is for exclusion or for identification-by-lineage in combination with other evidence, not for a one-in-a-billion identification.

Exercise 8.19 †

Ranking: (A) the clean 20-locus single-source bloodstain is far stronger than (B) the low-template two-person mixture. (A) yields a full profile with a tiny random match probability and an unambiguous single source — it can strongly support the suspect as the source (and a mismatch would cleanly exclude). (B) is doubly compromised: it is a mixture (the contributor count is a minimum, allele assignment is ambiguous) and low-template (dropout/drop-in make calls unreliable), and the suspect is only a possible minor contributor. At best (B) supports a likelihood-ratio statement ("the evidence is X times more probable if he contributed"), never "this is his profile." Strength tracks how much interpretation stands between the data and the claim — and (B) requires far more.

Exercise 8.21

Overstatement: treating a lineage match as an individual identification. Honest rewrite: "The mtDNA on the hair is consistent with the defendant and with all of his maternal relatives, and with roughly the fraction of unrelated people who share this (sometimes common) type. It does not establish that this hair is the defendant's specifically."

Exercise 8.22 †

Two overstatements: (1) "found the suspect's DNA on the weapon" should specify it is touch DNA — a statement about cells, not contact; (2) "proves he used it" leaps from presence of cells to an act, ignoring secondary transfer and the unknown time of deposit. Honest rewrite: "Cells consistent with the defendant are present on the weapon. That establishes whose cells are present; it does not establish that the defendant handled the weapon, nor when the cells were deposited — secondary transfer is possible, and the profile alone does not distinguish direct from indirect deposit." (The Lukis Anderson case, Case Study 2, is the documented proof that this leap can be flatly false.)

Exercise 8.23

Overstatement: "genetic genealogy identified the killer." Honest rewrite: "Genetic genealogy generated an investigative lead — a name to investigate, from distant-relative matches and family-tree work. The identification offered as evidence was the conventional STR comparison between the crime-scene profile and the suspect's confirmed sample." Genealogy points; validated DNA identifies.

Exercise 8.25

Overstatement: "his alleles are all in there, so he's a contributor." In a mixture with several alleles per locus, a large fraction of the population might also "fit" into the crowd of alleles, so mere presence of a suspect's alleles is weak on its own. Honest rewrite: "The suspect cannot be excluded as a possible contributor; whether the evidence meaningfully supports his inclusion depends on a proper interpretation — ideally a likelihood ratio (Chapter 9) — and on how many people would also fit. Inclusion-by-eyeball is not strength of evidence."

Exercise 8.26 †

Overstatement: "same reliable DNA test, just more of it." Running extra amplification cycles (enhanced sensitivity / "LCN") is not covered by the validation that backs standard single-source typing; it amplifies stochastic effects (dropout, drop-in) and its reliability for casework has been seriously contested, including by a court inquiry into one such protocol. Honest rewrite: "The sample was typed at enhanced sensitivity because so little DNA was present. This is a more error-prone regime than standard typing — subject to dropout and drop-in — and the result must be interpreted and reported with those stochastic limits stated, not as if it were an ordinary single-source result."

Exercise 8.27

Two consent problems: (1) Consent at a distance — the uploader exposed not only her own genome but, partially, that of every relative who shares her DNA, including the suspect's relatives who never consented to anything. (2) Purpose mismatch / informed consent — she uploaded to find a half-sibling, a personal genealogy purpose, not to enable a criminal investigation; whether the database's terms permitted law-enforcement matching (and whether they changed after she uploaded) is a separate consent question. (One could also name innocent-relative intrusion as a third.)

Exercise 8.28 †

For limiting IGG to violent crime: the technique intrudes on the genetic privacy of many innocent relatives and creates a broad search capability; restricting it to the most serious offenses (murder, rape) keeps the intrusion proportionate to the gravity, which is how it was first justified (the Golden State Killer). Against any such limit: the line is arbitrary and unstable — if the method is reliable and lawful, why deny it to a serial burglar's victims, or to identify unknown remains (which almost everyone supports)? Each "obvious" exception invites the next (scope creep). A defensible landing: permit it under published policy with judicial or supervisory oversight, presumptively for serious violent crime and for identifying remains, with transparency about expansions — because the strongest objection is not to the method but to the absence of rules, which is fixable. (Credit any position that engages proportionality, scope creep, and the lead-vs-identification structure.)

Exercise 8.29

Two truthful direct-examination sentences: "DNA consistent with the defendant was recovered from the object." / "That result establishes that cells from the defendant are present on the object." One sentence to refuse even if invited: "Therefore the defendant handled the object" (or "was at the scene") — because the profile establishes presence of cells, not the act of contact or the time/means of deposit; secondary transfer is possible.

Exercise 8.31 †

The lab's protocol failure ties to theme two (the validity spectrum) by showing that a method's place on the spectrum is set not only by the underlying science but by the interpretation protocol layered on top: STR chemistry is rigorous, but a flawed mixture-interpretation rule (e.g., careless CPI/RMNE on ambiguous mixtures) can produce overstated inclusions, dragging the applied method down even though the typing is sound. "A mixture statistic is only as good as its interpretation protocol" means the number reported is a product of human-chosen rules, not a raw fact — so two labs can type the same mixture and report different strengths. The remedy (validated, transparent interpretation, e.g., properly validated probabilistic genotyping) is what moves the applied method back up the spectrum.

Exercise 8.33 †

Expected single-source undegraded reference (victim): full set of loci, all present, balanced heterozygote peaks of roughly equal height across the whole panel, no ski-slope, and at most two alleles per locus. Contrast with Figure 8.3 (degraded mixture): (1) the large loci are faded or absent (ski-slope of degradation) where the reference has full peaks; (2) several loci show more than two alleles and unequal heights (the mark of a mixture plus stochastic imbalance) where the single-source reference shows a clean one-or-two-allele pattern. Either difference alone flags the evidence sample as harder than a reference.

Exercise 8.34 †

(a) It establishes that at least two people's cells are on the handle — one component consistent with the victim (Diallo), plus an unknown minor contributor not in CODIS. It does not name the minor contributor, does not say when or how any cells were deposited (secondary transfer is live), and the "at least two" is a minimum count. (b) They turn to genealogy for the minor contributor because that is the unknown person and CODIS returned no hit — exactly the Golden State Killer situation, where the offender isn't in the criminal database, so you look for his relatives in a genealogy database instead. (c) "The DNA puts our suspect at the cabin" is premature on this chapter alone because (i) no suspect has even been named from the mixture yet, (ii) presence of cells ≠ presence of a person (transfer), and (iii) interpreting the mixture — computing a likelihood ratio for any named person — is Chapter 9's job, not this chapter's. This chapter only establishes "a mixture exists and a lead exists."

Exercise 8.35

The indispensable confirmatory step would be a conventional STR comparison between the crime-scene profile and a confirmed sample from the candidate (e.g., abandoned DNA), exactly as in the Golden State Killer case. Genealogy alone is only a lead — it points to a family branch and a candidate but is not offered as courtroom identification; the validated STR comparison is what could (or could not) actually associate the candidate with the evidence. Without it, a genealogical "candidate" means nothing evidentiary. (No solution to the case is implied — only the method.)

Exercise 8.37

Three points on DNA's internal validity gradient: (1) a clean single-source 20-locus STR profile — top of the spectrum, quantified, near-individualizing; (2) a low-template touch mixture — sound chemistry but treacherous (dropout/drop-in, secondary-transfer inference, contributor-count uncertainty), much weaker; (3) an mtDNA or Y-STR lineage match — valid but limited by biology to a family line, never an individual. Same discipline ("DNA"), three very different strengths — the chapter's proof that the validity spectrum runs inside the strongest method, not just between methods.

Exercise 8.38 †

The Golden State Killer case is honest progress, not just impressive progress, because of the lead-vs-identification structure. The new, less-validated step (genealogical inference from SNP matches) is used only to generate a lead; its error mode is "follow the wrong family branch" — a cost in effort and innocent-relative intrusion, not a wrongful conviction. The identification offered to the court is the conventional STR comparison, an old, validated, quantified method. This protects against the bite-mark failure mode (preview Ch. 16), where an unvalidated comparison was itself presented to juries as identification: IGG keeps the unvalidated step out of the courtroom-claim role and reserves the conclusion for the validated method. Confidence does not outrun proof.

Exercise 8.39

Three places interpretation (and thus bias) re-enters DNA, with safeguards: (1) touch-DNA noise floor — deciding whether low peaks are real alleles or noise → safeguard: interpret the evidence profile blind to the suspect's reference (context management, Ch. 31). (2) Mixture deconvolution — assigning alleles to contributors and counting them → safeguard: validated, transparent interpretation protocols / properly validated probabilistic genotyping (Ch. 9), and blind interpretation. (3) Genealogical tree-building — settling on a candidate family → safeguard: treat the lead as a hypothesis confirmed only by independent STR comparison, never as a conclusion to bend the case toward.

Exercise 8.40 †

Model correction: "Modern DNA is remarkably sensitive — it can develop a profile from a few shed skin cells — but sensitivity is not omnipotence. That same sensitivity makes secondary transfer a real problem: a profile tells you whose cells are on an object, not how they got there (an innocent, hospitalized man's DNA once reached a murder scene via paramedics). Many real samples are degraded (the large loci drop out, giving only a weaker partial profile) or are mixtures of several people (where the contributor count is only a minimum and the result may be inconclusive). And when the offender isn't in CODIS at all, even perfect DNA leaves a case cold until genealogy can find a relative. DNA is the strongest tool in forensic science — and precisely because it is, the honest practitioner is careful to say only what each result actually supports."

Chapter 9 — Worked Solutions (daggered † and odd-numbered exercises)

All numbers are illustrative teaching figures, not values from real cases.

Exercise 9.1

Allele frequency is the proportion of chromosomes in a reference population that carry a particular allele at a given genetic location. It is the raw material of a random match probability because the probability of a person's genotype at each locus is computed from the individual allele frequencies, and the whole-profile frequency is then the product of those per-locus frequencies across (approximately independent) loci. No allele frequencies, no match probability.

Exercise 9.3

The random match probability reports the probability of a match GIVEN that the matching person is not the source — i.e., P(profile matches | the person is a random, unrelated non-source). It is a statement about coincidence among unrelated people, not about the defendant's guilt.

Exercise 9.4 †

The prosecutor's fallacy is the error of treating the probability of the evidence given innocence as if it were the probability of innocence given the evidence. It confuses P(match | innocent) (e.g., the RMP) with P(innocent | match). These are different conditional probabilities, and the difference can be enormous — the four-legs/cow swap.

Exercise 9.5

The defense fallacy is the error of pointing to the many people in a large population who would match by chance and concluding the match is nearly worthless. It ignores the other evidence (opportunity, motive, connection to the victim) that brought this defendant — not the other hypothetical matchers — before the court.

Exercise 9.7 †

posterior odds = LR × prior odds. The forensic scientist supplies the LR (the weight of the DNA evidence, computed from the data and the two propositions). The jury supplies the prior (how likely the defendant was to be the source before the DNA, drawn from all the other, non-DNA evidence). The scientist must not supply the prior or the posterior; doing so usurps the jury's role and risks the prosecutor's fallacy.

Exercise 9.9 †

For a clean single-source match, the LR comparing "defendant is the source" to "an unrelated stranger is the source" is approximately 1 / RMP. If the defendant is the source, the probability of seeing his profile is ≈ 1; if a random unrelated person is the source, the probability of seeing the defendant's exact profile is the RMP. So LR ≈ 1 / (1/20,000,000) = 20,000,000. The evidence is about 20 million times more probable if the defendant is the source than if an unrelated stranger is.

Exercise 9.11 †

Combined profile frequency ≈ (1/10) × (1/25) × (1/4) × (1/40) = 1 / (10 × 25 × 4 × 40) = 1 / 40,000. Expressed in words: about one person in forty thousand would be expected to carry this four-locus profile by chance. (Real profiles use far more loci, driving the figure to one in billions or trillions — this is a deliberately small teaching example.)

Exercise 9.13 †

• (a) posterior odds = 100,000 × (1/100,000) = 1, i.e., even odds — a coin flip — despite a six-figure LR, because the prior is very weak.
• (b) posterior odds = 100,000 × (1/50) = 2,000 to 1, a strong case. The DNA (the LR) is identical in both; the conclusions differ entirely because the prior differs. This is the §9.5 lesson: the LR is the weight, not the verdict.

Exercise 9.14 †

Expected coincidental matchers ≈ population × RMP = 70,000,000 × (1/10,000,000) = 7 people. This does not give the probability the defendant is the source because (1) it does not account for the other evidence that singled out this defendant rather than the other six possible matchers (avoiding the defense fallacy), and (2) "probability he is the source" is a posterior that requires a prior (§9.5). The count of expected matchers is an input to honest reasoning, not the answer.

Exercise 9.15

With an identical twin, the 1-in-3-billion figure (computed against an unrelated person) is the wrong number, because identical twins share essentially their entire nuclear STR profile — the random-match probability for a twin is close to 1, not 1 in 3 billion. The DNA cannot distinguish them. Distinguishing the twins would require non-STR evidence: differing circumstances/alibi, fingerprints, or specialized methods that can detect the rare epigenetic or somatic differences between twins — none of which the ordinary profile provides.

Exercise 9.16 (a) †

(a) May not say it. This is the prosecutor's fallacy — it transposes P(match | innocent) into P(innocent | match). Correct rewrite: "A random, unrelated person would share this profile by chance with a probability of about 1 in 6 million." (b) is permissible (a strength-of-evidence LR statement). (c) is not permissible — it conflates DNA presence with both being at the scene and with conduct/timing. (d) is permissible (a correct RMP statement). (e) is not permissible — overstated individualization; reserve quantified claims for the quantified statistic and keep the verb on the evidence.

Exercise 9.17 †

The detective has committed the prosecutor's fallacy. "A billion to one" refers (at most) to the probability that a random unrelated person would coincidentally share the profile — a statement about coincidence — not the probability the defendant is the source or is guilty. Honest correction: "The DNA shows that a random unrelated person would share this profile only about once in a billion; that is strong support that the defendant rather than a random stranger contributed, but it is not the probability he is guilty, and it says nothing about how his DNA got there."

Exercise 9.19

Two validated programs disagreeing by a factor of a thousand on a complex mixture does not mean one is "broken." For complex mixtures, the LR is model-dependent: different programs make different modeling assumptions (about peak heights, drop-out, drop-in, number of contributors), and on a hard sample those assumptions diverge. The accurate description is that the result is an estimate sensitive to the model, that the sample may be near or beyond validated limits, and that the disagreement is itself information the court should hear — not proof of error or fraud.

Exercise 9.20 †

Three questions a cross-examiner should ask, and why: 1. What were the two propositions (Hp and Hd)? — An LR is meaningless without them; a poorly chosen defense proposition can inflate the number. 2. How many contributors did you assume, and why? — The assumed contributor number is a human input that drives the result and can be biased. 3. Was the sample within the software's validated limits? — Outside those limits the LR is unreliable and should not have been reported as valid. (Also fair: which program, which version, and may the defense examine its source code and validation data.)

Exercise 9.21

A model "Read the Evidence" block should keep WHAT IT DOESN'T honest: a 1-in-50-million RMP for the roommate's profile shows the profile is rare and the roommate cannot be excluded and is strongly associated — but a roommate's DNA in a shared dwelling has an obvious innocent explanation (he lives there; transfer and background DNA are expected), the statistic addresses presence not conduct or timing, and the relevant prior differs from a stranger's. THE INFERENCE: strong association awaiting a prior, not proof. THE LESSON: in a shared environment, even a rare match must be weighed against the everyday reasons that person's DNA would be present.

Exercise 9.22 †

PCAST 2016 reached a two-part verdict: probabilistic genotyping is foundationally valid for simple mixtures (few contributors, adequate DNA) within validated limits, but its validity for complex mixtures (many contributors, low template, heavy allele overlap) is not yet established. A single method can be valid in one regime and not another because foundational validity is demonstrated by studies under specified conditions; the evidence supports the method inside those conditions and is simply absent (or shows degraded performance) outside them. The honest analyst states which regime the actual sample occupies.

Exercise 9.23

Competing interests: the developer's trade-secret/intellectual-property claim over the source code versus the defendant's due-process right to confront and test the evidence used to convict him. It is a fairness question rather than a scientific one because the science of whether the program works is addressed by validation studies; the distinct problem is that an adversarial system built on the right to challenge evidence sits uneasily with a conviction resting partly on a calculation the defense is forbidden to inspect. Courts have split, and the issue remains live.

Exercise 9.25

The report should state that the sample (five contributors, very low template) falls beyond the software's validated limits, that the returned number is therefore not a reliable LR and should not be presented as a valid result, and that the result is inconclusive for evidential purposes. "We got a number, so we'll report it" is wrong because software run outside its validated envelope does not produce a defensible estimate; reporting it anyway lends false weight to an unvalidated computation.

Exercise 9.26 †

A model refusal: "I can't tell the jury there's a one-in-a-billion chance it's anyone else — that would state the probability of guilt, which is the prosecutor's fallacy and not something the science supports. What I can and will say is that a random unrelated person would share this profile only about once in a billion, which is strong support that the defendant rather than a stranger contributed — and I'll make clear that this speaks to presence, not to how the DNA got there or whether he is guilty."

Exercise 9.27

"Report the strength of the evidence, never the probability of guilt" prevents both errors because both errors share one feature: attaching the verb to the person instead of the evidence. The prosecutor's fallacy ("the chance he's innocent is…") and overstated individualization ("this is his and no one else's, to a certainty") both make a claim about the defendant that the data do not license. Keeping every statement about the evidence ("a random person would match with probability…," "the evidence supports… by a factor of…") structurally blocks the step into either error.

Exercise 9.28 †

• Over-trust: A jury primed by television for forensic certainty hears a carefully hedged probability as a verdict of guilt. Safeguard: state the limits explicitly — "this speaks to presence, not conduct; this is not the probability he is guilty."
• Under-value: A jury (or a defense argument recasting the LR as mere coincidence — the defense fallacy) dismisses genuinely strong evidence as worthless. Safeguard: explain the LR as strength-of- evidence and note that the DNA combines with the other evidence rather than standing alone.

Exercise 9.29

By the Bayesian framework, the conclusion is posterior odds = LR × prior odds. In a cold hit, the suspect was found only by searching millions of profiles, with no independent connection — so the appropriate prior that this person (rather than anyone else searchable) is the source can be much weaker than in a case where independent evidence pointed to the suspect first. Reporting the ordinary RMP as though the suspect had been independently chosen presents the same LR against an inappropriately strong implicit prior, overstating the evidence. What should be disclosed: that the suspect was identified by a database search, so the factfinder can weigh the weaker prior.

Exercise 9.31 (cold-case extension) — parts (a) and (c) are daggered

(a) Hp: Roy Keller is the minor contributor to the gas-can mixture. Hd: an unknown, unrelated person is the minor contributor. (The LR strongly supports Hp over Hd.) (b) A strong LR makes Keller consistent with having contributed, but Keller co-owns the property and the renovation, so his DNA on a gas can stored at the site has an innocent explanation (background/transfer DNA from ordinary access — the Chapter 8 transfer problem). The LR addresses presence, not how or when; and the relevant prior (a co-owner is expected to have left DNA around the property) is very different from a stranger's. So the number narrows the field without proving Keller handled the can on the night of the fire. (c) Case-file correction: "The detective's claim ('one-in-a-billion that it isn't Keller') is the prosecutor's fallacy — it transposes a statement about coincidence into a statement about Keller's guilt. The number means a random unrelated person would share this minor profile only about once in a billion; that is strong support that Keller rather than a stranger contributed, expressed as a likelihood ratio, not a probability that Keller is the source. It omits the prior — and the innocent-transfer explanation a co-owner of the property has."

Exercise 9.32 †

Two kinds of evidence that would raise the prior that Keller is the source: 1. Evidence of opportunity — e.g., cell-site or CCTV placing Keller at or near the cabin around the relevant time (later chapters), which makes it more likely a priori that his DNA was deposited in connection with the crime rather than by routine access. 2. Evidence of motive/means — e.g., the insurance/financial motive and access to the means of the crime (later chapters), which raises the prior odds that Keller, specifically, was involved. In Bayesian terms, the LR (the weight of the DNA) stays the same, but posterior odds = LR × prior odds, so a larger prior makes that same LR far more probative — the DNA that was merely "consistent with" Keller against a weak prior becomes strong corroboration against a strong one.

Chapter 10 — Worked Solutions (daggered † and odd-numbered exercises)

Solutions for the daggered items and the odd-numbered exercises. Where an exercise is open-ended, the model answer gives the points a strong response must hit, not the only acceptable wording.

Exercise 10.1

Serology is the forensic identification and characterization of blood and other body fluids. The two questions modern serology answers well: (1) is a biological fluid present, and (2) which fluid is it? (It largely ceded the old third question — whose is it? — to DNA.)

Exercise 10.2 †

A presumptive test is fast, sensitive, and inexpensive; a positive means the target is probably present (e.g., Kastle-Meyer pink → "peroxidase activity consistent with blood — keep going"), but it does not identify the substance because non-target materials can trigger it. A confirmatory test is more specific and establishes the substance is the target (e.g., a Takayama/Teichmann crystal test, or species/DNA testing → "this is human blood"). Positive presumptive = a lead to confirm; positive confirmatory = an identification you can defend.

Exercise 10.3

Every common presumptive blood test detects the peroxidase-like activity of the heme group (the iron-containing part of hemoglobin), not "blood" as such. Because other substances also have peroxidase activity (plant peroxidases like horseradish) or otherwise catalyze the reaction (some metal salts, oxidizers like bleach), those substances produce false positives — the tests are sensitive to the chemistry of heme, not unique to blood.

Exercise 10.5

Three luminol false positives (any three): bleach and other oxidizing cleaners; metal ions such as copper or iron; plant peroxidases (horseradish); also rust and some paints.

Exercise 10.7

Area of origin is the estimated 3-D region a spatter pattern came from. The area of convergence is the 2-D point/region where the stains' directions (projected back along the surface) intersect — it locates the source's position on the plane. Extending into the third dimension using each stain's impact angle lifts those lines into space; where they cross is the area of origin — the height-included estimate, always a region with uncertainty, not a coordinate.

Exercise 10.8 †

The relationship is $\sin(\text{angle of impact}) = \dfrac{\text{width of stain}}{\text{length of stain}}$. In plain words: a drop hitting straight down (90°) spreads evenly → a round stain (width ≈ length → sine ≈ 1). A drop hitting at a glancing angle skids along the surface → a long, thin ellipse (width much less than length → small sine → small angle). So a long, thin stain implies a shallow angle, and the narrow end/tail points the direction of travel.

Exercise 10.9

• Passive: formed by gravity alone — drips, flows, pools (e.g., a pool under a stationary wound).
• Transfer/contact: a bloody object pressed onto a surface (e.g., a bloody hand-print or a wipe/swipe).
• Spatter: blood projected through the air as droplets by a force (e.g., the fine droplets thrown when an object strikes a wet bloodstain).

Exercise 10.11

Many candidate stains → presumptively test each quickly. Negatives are reliably excluded (no detectable blood — don't waste DNA capacity on them). Positives ("could be blood") are the short list you then confirm and submit for DNA. The funnel lets you spend the expensive, sample-consuming steps only where screening says it's worth it, without fully testing all twelve.

Exercise 10.12 †

Overstatements in "this proves the defendant cleaned up a bloodbath": (1) luminol is presumptive, not confirmatory — a glow is not confirmed blood; (2) bleach is a known luminol false positive, and the suspect is known to have used bleach, so the glow may be the cleaner itself; (3) even if it were blood, luminol cannot say it is human, cannot say whose, cannot date it, and cannot establish the action ("cleaned up") or that a "bloodbath" occurred. Honest report line: "Luminol revealed chemiluminescence in the bathroom consistent with possible latent blood (or with oxidizing cleaners/other interferents); confirmatory testing for human blood and DNA analysis were requested. The reactivity does not establish the presence, source, quantity, or timing of blood."

Exercise 10.13

Ordered steps with justification: (1) Photograph/document in place, with and without scale — you cannot recollect a smeared pattern. (2) Presumptively test a small edge — confirm it's worth collecting without consuming the stain. (3) Collect by wet-then-dry swab (moist swab to lift, dry swab to capture remainder) — the right method for a dried stain on a non-absorbent painted surface. (4) Air-dry the swab, package in breathable paper — moisture breeds microbes that degrade DNA. (5) Take a substrate control (clean adjacent wall) and a positive control for the reagents — to prove the surface isn't causing the reaction and the test works. (6) Label, seal, initial, log chain of custody. Each step precedes the next because each protects the option the next step needs (the photo before contact; the small edge before DNA; controls before interpretation).

Exercise 10.14 †

Likely DNA condition: degraded, possibly severely — three warm days sealed wet means trapped moisture, bacterial growth, and mold, all of which fragment DNA; a profile may be partial or fail. Rule violated: biological evidence must be air-dried and packaged in breathable paper, never sealed in plastic (Ch. 2–3). Why: sealed plastic retains moisture; moisture + warmth = microbial activity and hydrolysis that break DNA into unusable fragments. The chemistry of "is it blood" might survive; the DNA — the part that answers whose — is what's lost.

Exercise 10.15

$\sin(\text{angle}) = 4/8 = 0.5$, so angle of impact $\approx 30°$. Three error sources despite the exact formula: (1) where the ellipse "ends" is a judgment call (irregular/scalloped edges), changing the measured length; (2) surface texture/absorbency distorts the stain shape from ideal; (3) the stain may not be a clean single impact (satellite spines, overlap), so the measured dimensions don't reflect one clean drop. The formula is exact; the inputs are uncertain.

Exercise 10.17

The stringing/tangent method projects each stain back along a straight line in its direction of travel and lifts it by its impact angle. But a blood droplet actually follows a parabolic, gravity-curved path (it falls as it flies). A straight back-projection therefore overshoots upward, placing the origin too high. The ignored fact is gravity's downward pull on the droplet in flight; software can partly correct using velocity assumptions, which carry their own uncertainty.

Exercise 10.18 †

Two qualified analysts calling the same stains "impact spatter" vs. "transfer" reveals that BPA's event/mechanism conclusions are subjective and not anchored to a measurable, reproducible standard — they sit in the contested region of the validity spectrum (the 2009 NAS finding). When conclusions can swing 180° on identical evidence depending on the reader, the method has not established foundational validity for that claim. The jury must be told plainly that this is contested expert opinion under uncertainty, not a measured finding, and should hear the competing interpretation and the basis for each. (This is the David Camm situation, §10.6.)

Exercise 10.19

Hazard: contextual bias — telling the analyst the desired conclusion ("we think he beat her — confirm it") before they interpret an ambiguous pattern lets expectation steer a subjective judgment. Safeguard: context management / blind analysis (Ch. 31) — the analyst measures the stains' features (directionality, angle, category) before and without knowing the case theory. The opinion is compromised even if correct because we can no longer distinguish "the analyst saw this in the blood" from "the analyst brought this to the blood"; reliability requires that the conclusion not be contaminated by the desired answer.

Exercise 10.21 †

Model Read the Evidence block (constructed example; 3 mm × 15 mm stain, tail up-and-right): - THE ITEM: a single elongated bloodstain, ~3 mm wide × ~15 mm long, on a vertical wall, with a small tail pointing up and to the right. - THE CONTEXT: one stain among a pattern; photographed with a scale before any contact; surface texture noted. - WHAT IT SHOWS: directionality — the drop was traveling toward the tail (up and to the right). Angle of impact $\approx \arcsin(3/15) = \arcsin(0.2) \approx 11.5°$ — a shallow, glancing strike. - WHAT IT DOESN'T: it does not establish what event projected the drop, when, with what weapon, or by whom; one stain cannot give an area of origin; surface/edge effects add uncertainty to the angle. - THE INFERENCE: a single drop traveling up-and-to-the-right, striking at a shallow angle — consistent with a low, lateral source, pending convergence with the rest of the pattern. Nothing about a person's actions. - THE LESSON: directionality and angle are defensible per-stain facts; the event is an inference that requires the whole pattern and still cannot narrate the crime.

Exercise 10.23

Sub-millimeter droplets densely clustered loosely suggest a higher-energy event (more atomization → smaller drops). But several mechanisms produce a fine mist: an impact striking a wet bloodstain, an expirated pattern (blood blown through nose/mouth), a gunshot-associated spatter, or blood flung as fine cast-off. The size alone cannot choose among them, because energy is not unique to one mechanism — which is exactly why §10.6 warns against reading a specific event off droplet size.

Exercise 10.25 †

"To a reasonable degree of scientific certainty" implies a measured, reproducible basis. For event-level BPA (which event/mechanism produced a pattern), the 2009 NAS report found the interpretations more subjective than scientific, with enormous uncertainties and inadequately established error rates. Attaching a certainty formula to a conclusion that is actually a contestable opinion misrepresents its epistemic status to the jury — it dresses an opinion as a finding, the precise overstatement the chapter (and Theme 4, the CSI effect) warns against. Directionality and angle measurements may merit confident language; the event narrative does not.

Exercise 10.27

A short certification course plus a paid practice testifying for whichever side hires you implicates (a) the NAS concern that BPA training and qualifications vary widely and are sometimes minimal — thin qualification undermines the weight of the opinion; and (b) the "hired gun" problem (preview Ch. 30) — an expert whose income depends on reaching the retaining side's preferred conclusion has a structural incentive that, combined with BPA's subjectivity, is especially corrosive. The duty is to the court and the evidence, not the client; the safeguard is transparency about training, method, and uncertainty.

Exercise 10.29

"What is it / whose is it?" (serology→DNA) and "what happened?" (BPA) must stay separate because they sit at different points on the validity spectrum; merging them lets weak event-reconstruction borrow the credibility of strong blood-ID. Real failure from merging: an analyst confirms a stain is the victim's blood (strong) and, in the same breath, narrates from the spatter exactly how the victim was attacked (weak/contested) — and the jury, hearing one confident voice, credits the narrative as if it were as solid as the identification.

Exercise 10.31 †

Cold-case extension. (a) The two facts establish sequence: the doorframe stain's lower edge is below the char line → blood was deposited before the fire reached that height; the low spatter is inconsistent with collapse-in-fire → consistent with blood projected by a force while liquid, before the fire. Together: force/bleeding occurred before the fire. (b) They do not establish: whose blood it is (DNA pending); who applied the force; or what happened (no blow-by-blow from BPA). (c) The next chapter is Chapter 11 (forensic pathology); the settling finding previewed in this book is no soot/smoke in the airways, which would show Diallo was not breathing when the fire burned — i.e., dead before the fire — far more decisively than any bloodstain.

Exercise 10.33

Cold-case extension. Two reasons luminol is poor in a fire scene specifically: (1) fire scenes are full of luminol interferents — metal ions, soot, rust, and especially oxidizers/cleaning residues — producing false glows; (2) heat and combustion degrade biological evidence and DNA, and luminol can degrade it further, so a glow may not yield a usable profile anyway. A justified limited use: when a specific, documented area (e.g., a swept-looking patch outside the main burn) needs screening to direct confirmatory and DNA testing — applied after photography, with controls, and understood as presumptive only.

Chapter 11 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 11.1

Forensic pathology is the branch of medicine that applies pathology (the study of disease and injury) to legal questions, chiefly determining the cause and manner of death. A forensic pathologist is a physician — a medical doctor who completed a pathology residency and then a forensic-pathology fellowship, typically board-certified in the subspecialty.

Exercise 11.2 †

A medical examiner is an appointed physician (ideally a board-certified forensic pathologist) with legal authority to investigate and certify reportable deaths on medical grounds. A coroner is an official, often elected, whose qualifications are set by statute and may require no medical training, and who hires a physician to perform autopsies while retaining authority to certify (in some places, even to overrule the physician). Ways the difference affects quality/independence (any two): (a) a coroner may decline to order an autopsy a pathologist would have required, missing a subtle homicide; (b) a non-physician may certify or overrule a cause/manner determination they are not trained to make; (c) an elected coroner may face political/local pressures a board-certified appointee is better insulated from; (d) resources and training vary widely. The 2009 NAS report recommended moving toward medical-examiner systems staffed by board-certified forensic pathologists and away from the coroner model.

Exercise 11.3

The five phases: (1) history and scene (context); (2) external examination (clothing, body chart, injuries — collecting trace before washing); (3) internal examination (cavities opened, organs examined/weighed/sectioned, wound tracks documented); (4) specimens and ancillary studies (toxicology, histology, radiography/CT, carboxyhemoglobin in fire deaths); (5) correlation and report (cause + manner + reasoning). The governing rule: observe and sample before you alter, because you get exactly one autopsy — each phase preserves information the next phase might destroy.

Exercise 11.5 †

The five conventional U.S. manners of death are natural, accident, suicide, homicide, and undetermined. Undetermined is reserved for when the evidence does not permit a confident classification (it is an honest classification, not a failure — the alternative, inventing a manner the evidence does not support, is the error).

Exercise 11.7

The postmortem interval (PMI) is the time elapsed since death. The early-window estimate (livor, rigor, algor, decomposition) is always a range because every one of those clocks runs on assumptions the scene can violate — above all the ambient temperature history, which is rarely known for the unobserved interval before discovery, plus body size, clothing, surface, air movement, and the decedent's pre-death state. Uncertain inputs propagate into an uncertain (hence ranged) output; a single confident time is a screenwriter's line, not a finding.

Exercise 11.9

Carboxyhemoglobin (COHb) is the compound formed when carbon monoxide — produced abundantly in structure fires — binds to hemoglobin in the blood (far more avidly than oxygen). A person breathing in a fire absorbs CO and the COHb saturation climbs; a person already dead does not. The blood sample is among the most important in a fire death because its measured level helps answer the pivotal question — was the person alive and breathing during the fire? — and a high level can itself establish the cause (CO poisoning/smoke inhalation), while a low level indicates the person was not breathing when it burned.

Exercise 11.11

Most parsimonious inference: the body was moved after death — it lay on its back long enough for blood to pool and fix there, then was repositioned face-down before discovery. The change that revealed it is livor mortis (lividity). "Fixed" means the pooled blood no longer shifts with position or blanches under finger pressure; fixation occurs after roughly 8–12 hours. Fixed lividity on a surface that is not the lowest surface in the found position is the classic sign of postmortem repositioning.

Exercise 11.12 †

A person alive when a fire starts breathes, and breathing does two things: it draws the fire's particulate soot down past the larynx into the trachea and deep bronchi (deposited on the moist lining), and it carries the fire's carbon monoxide into the blood, where it binds hemoglobin as carboxyhemoglobin, raising its saturation. So a living victim tends to show both deep-airway soot and elevated COHb. A person already dead does not breathe, so neither soot is inhaled into the deep airways nor CO is absorbed into the blood — hence neither finding. Each finding requires active respiration during the fire; their absence together indicates the person was not breathing when it burned.

Exercise 11.13

Conditions that speed rigor onset: warm ambient temperature and high pre-death muscular activity/exertion or fever (an extreme case is the historically described near-instant "cadaveric spasm" after intense exertion). A condition that slows it: cold. The pathologist must know the scene temperature and the decedent's pre-death activity because reading rigor without them can place the time of death badly wrong in either direction — a body that stiffened fast in heat could be mistaken for one dead much longer, and vice versa.

Exercise 11.15

It is possible — and expected — because cause and manner answer different questions. The cause (the injury or disease that started the lethal sequence) is a physical finding both pathologists can read from the body and agree on. The manner (the circumstance category) is an inference from the totality — the autopsy plus the scene, history, and circumstances — about human agency the body alone usually cannot reveal. The same fatal injury can be accident, suicide, or homicide; reasonable experts weighing the surrounding circumstances differently can therefore reach different manners while agreeing entirely on the cause. It is not incompetence; it reflects the manner's nature as a reasoned, revisable opinion.

Exercise 11.17 †

A head injury, by itself, often cannot distinguish assault from a fall because the injury records that blunt force was applied — roughly where, with what severity, sometimes a patterned shape — but not the agency behind it. A fall and a blow can produce overlapping injury patterns, and the wound does not contain the information about who or what delivered the force. Bridging the gap requires the scene and circumstances — the position of the body, the surfaces present, the account given, other injuries (e.g., defensive wounds), and the broader investigation. This is the cause-versus-manner distinction exactly: the body supports the cause (blunt-force head injury) firmly but supports the manner (accident vs. homicide) only weakly, through circumstances the body does not contain.

Exercise 11.19 †

For the gunshot case in Figure 11.1: cause = "gunshot wound of the torso" (the injury that started the lethal sequence); mechanism = "exsanguination" (the physiological failure — fatal blood loss — that actually ended life); candidate manners = natural is excluded, leaving accident, suicide, or homicide, which the body alone often cannot decide. The mechanism is the least informative for a court because many different causes share it — exsanguination follows a homicidal, accidental, or suicidal shooting (and many non-gunshot injuries) alike — so "died of exsanguination" tells a court almost nothing it needs to know.

Exercise 11.21

The deferral is honest because the determination is genuinely difficult: fire can fracture bone, producing heat artifacts (burned, contracted, fractured bone) that mimic blunt-force trauma, and distinguishing a true perimortem fracture from a heat-induced postmortem one is the specific expertise of the forensic anthropologist (Chapter 12). The pathologist states what can be supported now — a skull fracture is present, its character raises perimortem blunt-force trauma as a serious possibility — and refers the definitive call to the specialist, rather than overstating a conclusion the present finding cannot bear alone. Deferring an uncertain determination to the right expert is the discipline of the book, not a weakness of the autopsy.

Exercise 11.23

The sentence does well in at least three ways: (a) it gives a range ("roughly 12 to 24 hours"), not a false-precision single time; (b) it says "most consistent with," an honest verb that states the estimate as a supported inference rather than a fact; (c) it states its assumptions ("given assumptions about ambient temperature"), exposing exactly where the estimate could be wrong. (Bonus: "before examination," not "of death to the minute," locates the measurement honestly.)

Exercise 11.25

The overstatement misuses the manner determination: a skull fracture is evidence about the cause (blunt-force head injury), but "murder" is a manner (homicide), which is an opinion built from the totality of circumstances, not a fact the fracture announces — the same injury could, in principle, result from a fall. Defensible version: "The skull fracture is consistent with blunt-force head trauma as the cause of death; in my opinion, considered with the scene and the other findings, the manner is homicide — but that is a reasoned conclusion about the circumstances, not a fact established by the fracture alone."

Exercise 11.27

Petechiae (pinpoint hemorrhages) are not diagnostic of asphyxia because they are nonspecific: they can be absent in genuine asphyxial deaths and present (or mimicked) in others. Asphyxia interpretation is especially prone to overstatement because its findings are often subtle and nonspecific, strangulation in particular may leave little external mark, and the determination frequently rests on careful internal neck dissection and the exclusion of alternatives rather than a single unmistakable lesion. Reasoning from one nonspecific finding (petechiae) to a confident diagnosis (strangulation) skips the careful exclusion the method requires.

Exercise 11.29 †

Risk: being told "the partner confessed" before the autopsy lets that information anchor the analysis (contextual/confirmation bias, Chapter 31) — ambiguous findings get read toward the expected answer. Most vulnerable findings: the ambiguous judgment calls — antemortem vs. perimortem injury, the significance of a borderline neck hemorrhage in suspected strangulation, and the perimortem-versus-heat-artifact fracture call — because these are interpretive and the surrounding narrative can steer them. Safeguard: insofar as possible, the pathologist should form the medical findings before being saturated with the investigative theory, and should state explicitly when a conclusion rests on circumstances rather than the body (context management; sequential unmasking). A finding read blind is worth more than one read knowing the theory, even when they agree, because only the blind finding is uncontaminated by the expectation — agreement that might be bias-driven cannot be distinguished from genuine independent corroboration.

Exercise 11.31

Overwork is an error source because death investigation is a careful, time-intensive scientific act: a pathologist carrying a caseload above recommended limits has less time per case for thorough examination, specimen collection, histology, correlation, and documentation, and fatigue degrades judgment exactly where the hardest calls are made. This is the Chapter 4 lesson transposed to the morgue: just as the lab's accreditation, staffing, and quality systems set the ceiling on the evidence, the death-investigation office's resources and caseload set the ceiling on the death investigation. The national shortage of forensic pathologists makes this systemic, not merely individual.

Exercise 11.33 †

On the NAS 2009 / PCAST validity spectrum (more defensible → less): 1. Single-source nuclear DNA match (Chapter 7) — STRONGEST. Quantified, rigorously validated, known error structure; the field's gold standard. 2. "Dead before the fire" from absent airway soot + low carboxyhemoglobin (this chapter) — STRONG. Explicit respiratory mechanism, largely objective findings (soot present/absent below the larynx; COHb a measured number), two independent lines converging. Below a clean DNA match but well above subjective pattern methods. 3. "Victim was kneeling when struck" from bloodstain area-of-origin (Chapter 10) — CONTESTED/MIDDLE. Has a real geometric core but the event-level narrative ("kneeling," "struck") is interpretive, examiner-dependent, and (per the NAS report) often overstated, with error rates largely unmeasured. 4. "These teeth made this bite mark" (Chapter 16, previewed) — DISCREDITED. No validated basis for the specific identification claim; multiple exonerations. At the bottom. Justification turns on each conclusion's mechanism, objectivity, and measured error rate — the PCAST question.

Exercise 11.35

This chapter advances at least two themes. Theme 1 (exclusion over proof): the autopsy strongly supports "dead before the fire" and amends the manner to homicide, but it proves no perpetrator and names no one — the cause is firm, the manner is a revisable opinion, and the verbs stay honest ("strongly supports," "most consistent with," "in my opinion"). Theme 2 (the validity spectrum): the fire-death finding is explicitly placed high on the spectrum — above contested bloodstain event-reconstruction (Chapter 10) and far above discredited bite marks (Chapter 16) — because it has an explicit physiological mechanism, largely objective findings, and converging independent lines, while the manner determination is openly flagged as an opinion rather than a fact. (The chapter also touches Theme 3, the "scene history" and a known confession as bias channels in injury interpretation, and Theme 4, the television fantasy of an omniscient medical examiner.)

Exercise 11.36 †

Cold-case evidence-log entry (model): - (a) Strongly-supported conclusion + verb: the absence of inhaled soot in the deep airways together with a low carboxyhemoglobin level strongly supports that the decedent was not breathing during the fire — dead before the fire. - (b) Manner amendment as opinion: the manner of death is amended from accident toward homicide; this is the pathologist's reasoned opinion, built from the convergence of the airway/COHb findings with the skull fracture and the pre-fire blood evidence (Chapter 10) — not a fact the body announces. - (c) Deferred finding: whether the skull fracture is a true perimortem blunt-force injury or a heat artifact of the fire is deferred to Chapter 12 (forensic anthropology), because fire can fracture bone. - (d) Does NOT establish (any three): who caused the death; the exact time of death; that the gas-can DNA or any suspect is the source; that the fire was deliberately set (Chapter 22); whether the decedent was incapacitated first (Chapter 20). The autopsy reveals that a homicide occurred, never who.

Exercise 11.37

Three threads the autopsy hands forward as open questions: (1) the skull fracture → Chapter 12 (forensic anthropology): is it true perimortem blunt-force trauma or a heat artifact of the fire? (2) the toxicology specimens → Chapter 20 (postmortem toxicology): was the decedent incapacitated (e.g., sedated) before death? (3) the fire → Chapter 22 (arson): was the fire incendiary — deliberately set — rather than accidental? Each is a question the body raised but cannot answer alone.

Exercise 11.39

(Model short answer, ~170 words.) A real autopsy can establish "dead before the fire" with strong confidence because the finding rests on respiratory physiology: a living person in a fire breathes, drawing soot into the deep airways and absorbing carbon monoxide into the blood as carboxyhemoglobin; a person already dead does neither. When the deep airways are clear of soot and the carboxyhemoglobin is low, two independent findings converge on one conclusion — the person was not breathing when the fire burned. That is about as firm as forensic pathology gets, because the mechanism is explicit and the findings are largely objective. But the body cannot name who caused the death. The autopsy establishes the cause (most consistent with blunt-force head trauma) and supports a manner (homicide) as a reasoned opinion from the totality of circumstances — yet the wound does not record the hand that delivered it. Identifying a perpetrator requires other evidence: DNA, digital records, physical traces. The body tells us that a crime occurred, not who committed it.

Exercise 11.40 †

(Model short answer, ~190 words.) The "dead before the fire" determination rests on respiratory physiology with an explicit mechanism: breathing during a fire deposits soot in the deep airways and raises carboxyhemoglobin; not breathing does neither. Its findings are largely objective — soot is either below the larynx or it is not, and carboxyhemoglobin is a measured number — and two independent lines converge, each corroborating the other through the same mechanism. That is why it sits high on the validity spectrum and survives cross-examination robustly: the alternative explanations are few and can be considered and excluded, and the testing is standard. Bloodstain pattern event-reconstruction (Chapter 10), by contrast, has a real geometric core (directionality and impact angle) but its event-level conclusions — what action produced the pattern, in what sequence, by whom — rest on subjective interpretation that the 2009 NAS report found unvalidated, with error rates largely unmeasured, and that two qualified analysts can read in opposite ways (the David Camm case). The fire-death finding survives cross because it is mechanism-based and objective; the bloodstain narrative does not, because it is interpretive and contestable.

Chapter 12 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 12.1

Forensic anthropology — the application of physical/biological anthropology and human skeletal biology to legal questions, especially the recovery, identification, and analysis of decomposed, skeletonized, burned, or fragmentary remains. Biological profile — an estimate of a deceased person's sex, age, ancestry, and stature from the skeleton, which narrows an identification without individualizing it. Taphonomy — the study of everything that happens to remains between death and analysis (decomposition, weathering, scavenging, soil/water, burial, fire).

Exercise 12.2 †

The four components are sex, age at death, stature, and ancestry. Ranked from most to least reliable on a complete adult skeleton: (1) sex (adult pelvis, then skull — most reliable in mature adults); (2) age (tight for subadults via dental/epiphyseal indicators, looser for older adults — but well grounded overall); (3) stature (sound within population/sex-specific equations); (4) ancestry (least accurate and most contested). Accept age and stature in either order with justification; sex must be first and ancestry last.

Exercise 12.3

Antemortem trauma occurred before death and shows healing (rounded edges, bony callus, repair). Perimortem trauma occurred around the time of death on fresh bone with no healing (curved/beveled fractures, same color as bone). Postmortem trauma occurred after death on dry bone (straight/squared fractures, often off-color interior). The single defining features: healing (ante), fresh-bone fracture with no healing (peri), dry-bone fracture (post).

Exercise 12.5 †

Charring is the blackening of bone as its organic component carbonizes at moderate heat. Calcination is the later, chalky-white state after the organic component is gone, leaving only mineral. Calcination indicates the higher temperature. (The full color sequence: normal → brown → black/charred → gray → white/calcined.)

Exercise 12.7

The first step is a complete skeletal inventory — laying out, identifying, siding, and counting every element and fragment. It precedes any "reading" because you cannot estimate a feature from a bone you do not have, and any interpretation is invalid if two individuals are commingled. The inventory answers "how complete, and is it one person?" — the control against which every later claim is checked.

Exercise 12.9 †

Fresh (perimortem) bone retains its organic collagen framework and moisture, which makes it slightly elastic; under load it bends and hinges, fracturing along curved, beveled lines (like a green branch), and the broken surface is the same color as the rest of the bone because it broke while the bone was whole and fresh. Dry (postmortem) bone has lost collagen and moisture and is brittle; it snaps along straight, transverse, squared lines (like a dry stick), and the freshly exposed interior is often a different color from the weathered exterior. The controlling biological factor is the bone's collagen/moisture (organic) content.

Exercise 12.11

Honest rewrite: "The skeletal indicators are consistent with an adult of approximately [a range, e.g., 40–55] years of age." The original was wrong because a single number for an adult skeleton claims a precision the post-growth age markers cannot deliver — those markers (pubic symphysis, rib ends, suture closure, degenerative change) are slow and individually variable, so achievable precision is a wide range, not a point estimate.

Exercise 12.12 †

The three nested questions, in order: (1) Is it bone at all? (characteristic cortex-over-trabecular structure; rules out stone/wood/plastic). (2) Is it human or non-human? (gross morphology for whole bones; histology or molecular testing for fragments). (3) If human, is it forensically recent or archaeological/anatomical? The question made hardest by burning and fragmentation is #2 (human vs. non-human): calcination and breakage destroy the gross landmarks that would otherwise identify species, forcing reliance on histology or protein/DNA testing.

Exercise 12.13

A recovery is an excavation because the arrangement and context of the remains is itself evidence: whether the skeleton is articulated or scattered, its depth and soil layering, associated artifacts and projectiles, and present scavenging all bear on time since death, whether the body was moved, and which "fractures" are perimortem versus postmortem. A simple pickup destroys the depositional record — position, completeness (screening recovers small bones/teeth), and context — which cannot be regenerated in the lab (parallel to the crime scene, Chapter 2).

Exercise 12.15 †

(a) Stature: heat shrinks bone, so a femur measured after burning is shorter than in life; feeding a shrunken length into a stature regression yields an erroneously short stature estimate unless shrinkage is accounted for. (b) Sex: heat shrinks and warps the skull and pelvis, distorting the very dimensions and morphology used to estimate sex (e.g., subpubic angle, sciatic notch, cranial robustness); unaccounted-for shrinkage/warping can push an estimate toward the wrong sex or into spurious "indeterminate" or "male/female" calls. In both cases the fix is to recognize calcination/warping and adjust or caveat accordingly.

Exercise 12.17 †

This pattern — radiating fractures from a focal point, internal beveling, inward displacement at that point — is more consistent with a blunt-force blow than with heat. Justification (§12.6): blunt force delivers energy at a point, producing fractures with a focus (radiating outward, often concentric rings, inner-table beveling, inward displacement at impact). Heat fracturing, by contrast, is diffuse, typically curved/concentric, color-graded, and follows the heat gradient rather than radiating from a single impact. Report as "consistent with perimortem blunt-force trauma, distinguishable from heat-induced fracturing," and flag any heat-damaged margins as equivocal.

Exercise 12.19 †

A model answer (constructed teaching figure):

FIGURE (student) — "Grooved femur from a wooded scatter"        [constructed teaching example]
  THE ITEM        A femur from a surface scatter in woods; parallel grooves along one edge; the distal
                  end is missing.
  THE CONTEXT     Surface-deposited, exposed to weather and animals for an unknown period; recovered
                  with other scattered elements.
  WHAT IT SHOWS    The parallel grooves are consistent with rodent gnawing (paired-incisor scoring),
                  and the missing distal end is consistent with carnivore scavenging of a bone end —
                  i.e., postmortem taphonomic damage.
  WHAT IT DOESN'T  These features do not indicate perimortem trauma or a weapon; they say little about
                  cause of death and nothing about an assailant.
  THE INFERENCE    The damage is most consistent with POSTMORTEM animal activity, not injury. Scavenging
                  also explains why the inventory is incomplete (the distal end and possibly other
                  elements were removed/scattered).
  THE LESSON       Animal damage is a taphonomic mimic of trauma; subtract the world's signature before
                  reading the bone, and expect scavenging to scatter and remove elements.

Accept reasonable variants that correctly identify the grooves and missing end as postmortem taphonomy and avoid calling them trauma.

Exercise 12.21

A genuinely equivocal finding should be reported as equivocal — e.g., "the fracture margins are damaged by heat and the features are insufficient to determine perimortem versus postmortem timing." This is better than forcing a call because forcing one (a) overstates the evidence, exceeding what the method can support, and (b) is exactly where contextual bias enters — an ambiguous margin gets resolved toward the theory the case favors. Honest reporting of uncertainty is itself a forensic skill (preview Chapter 30).

Exercise 12.22 †

(a) Defensible — probabilistic, class-level, with the mature-adult caveat. (b) Overstated — ancestry stated as categorical fact; rewrite: "Cranial measurements are more consistent with one reference population than others; this is a probabilistic investigative aid, not a categorical biological fact, and 'race' does not map cleanly onto biological variation." (c) Defensible — consistent-with, distinguishes from heat. (d) Overstated — "at the moment of death" claims a timestamp the method cannot give; rewrite: "exhibits perimortem characteristics — consistent with injury around the time of death, before the bone dried." (e) Defensible — honest, scales confidence to surviving evidence.

Exercise 12.23

Three ways it exceeds real anthropology: (1) Occupation/handedness cannot be reliably read from a skull (at most, gross activity markers are weak and non-specific); this is fabricated precision. (2) Cause of death is the pathologist's/medical examiner's determination, not something the anthropologist announces from a skull alone. (3) The instant ("within a minute") confident certainty ignores the inventory, taphonomy, equivocal features, and the range/probability form of every honest anthropological output.

Exercise 12.24 †

Ancestry is most Daubert-vulnerable because (i) "race" is a social, not a clean biological, category (continuous/clinal variation), (ii) the statistical methods are constrained by their reference populations, and (iii) it is the least accurate, most contested component with the weakest claim to a measured error rate. Cross-examiner questions and honest answers: "Is race a biological category?" — "No, not a discrete one." "What reference populations did your software use, and was this individual's background among them?" — concede the dependence and possible non-representation. "What is the measured error rate of this classification?" — concede it is probabilistic and the validity is contested.

Exercise 12.25

The risk is contextual bias (Chapter 31): being told the investigators' conclusion ("the husband did it — just confirm the head injury") primes the examiner to read ambiguous fractures as blows and to reclassify postmortem damage as perimortem. The safeguard is context management / blind analysis: the anthropologist should reach a skeletal finding on the bone's physical features before being told the suspect's identity or the investigators' theory, and report equivocal findings as equivocal.

Exercise 12.27

The same methods scale: recover archaeologically (map, grid, screen), build the biological profile to narrow each set of remains, analyze perimortem trauma to document how victims died, and individualize by an independent method (dental, DNA). At mass scale the added problems are commingling and reconciliation (Chapter 35). The families are part of the mission because identification returns the dead to them and establishes the historical record — forensic science as a humanitarian act, not only a prosecutorial one.

Exercise 12.29 †

The two findings are independent: the pathologist's "no soot in the airways" (dead before the fire) comes from soft-tissue/airway examination; the anthropologist's "perimortem blunt-force fracture, not a fire artifact" comes from skeletal analysis. Because they rest on different evidence yet point the same way, their convergence raises confidence beyond either alone — a single finding could be an artifact or an error, but two independent lines agreeing is far harder to explain innocently. This is the weight-of- evidence / convergence reasoning the capstone (Chapter 39) formalizes.

Exercise 12.31 †

Skeletal trauma analysis sits in the broad middle of the validity spectrum: above bite-mark comparison (Chapter 16), which is discredited, and below single-source DNA (Chapter 7), the quantified gold standard. The single feature present in DNA and absent in trauma analysis that keeps trauma out of the top tier is a quantified, validated error rate (a number, with foundational validity per PCAST). Trauma analysis is well grounded in bone biomechanics but remains interpretive, yielding consistent-with conclusions without that number.

Exercise 12.33 †

The finding rules out the "the skull fracture was merely heat damage from the fire" theory — and with it the inference that the death was an accident. What remains unanswered about the perpetrator: the bone does not identify who struck the blow, the specific weapon (only a class of blunt force), or the exact timing beyond "perimortem"; heat-damaged margins remain equivocal. So the case is sharpened (homicide, trauma real) but no suspect is named or excluded by this evidence.

Exercise 12.35

A model two-sentence log entry: "Forensic-anthropological analysis found the cranial fracture exhibits perimortem blunt-force characteristics distinguishable from heat-induced fracturing, consistent with a focal blow around the time of death (corroborating the homicide finding). This skeletal evidence does not establish who inflicted the injury, the specific weapon, or the exact timing, and some heat-damaged margins are equivocal." (One sentence states the supported finding; one states the limits, using honest verbs.)

Exercise 12.36 †

The error is the prosecutor's fallacy (preview Chapter 9) compounded with confusing a class/consistent- with finding for an individualizing one (class vs. individual characteristics, Chapter 1). "Confirmed blunt-force trauma" establishes that a blow occurred, not who delivered it; leaping from "trauma confirmed" to "proof Keller killed him" treats a finding about the injury as a finding about the assailant. Correction to the detective: "The anthropology supports that a perimortem blunt-force injury occurred — it says nothing about who caused it. Tying it to Keller requires other converging evidence, and even then the honest verb is 'strongly supports,' never 'proves.'"

Chapter 13 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 13.1

Forensic entomology is the application of the study of insects and other arthropods to legal questions. Its single most common use is estimating the minimum time since death (minimum postmortem interval).

Exercise 13.3 †

ADD (accumulated degree days) is the total warmth-over-time available for insect development — the sum, over days, of the temperature above a species' developmental threshold. The casserole intuition: a dish needs a roughly fixed total amount of heat to cook, deliverable as a short time at high heat or a long time at low heat; likewise an insect needs a roughly fixed total of accumulated warmth to reach a given developmental stage, which it banks quickly in hot weather or slowly in cool weather. ADD measures that bank.

Exercise 13.5

Forensic botany is the application of plant science (plants, fragments, seeds, wood, algae, pollen) to legal questions, chiefly to associate a person or object with a place. Palynology is the study of pollen and spores, used forensically to compare pollen assemblages and associate an item with a location's flora. Both rest on Locard's exchange principle (Chapter 3): every contact leaves a trace — people and objects pick up and shed plant material from the places they pass through.

Exercise 13.7

A developmental threshold is the base temperature below which insect development effectively stops. When a day's average temperature falls below the threshold, that day contributes zero degree-days (never a negative number) — no usable developmental heat was banked that day.

Exercise 13.9 †

Required ≈ 150 ADD above a $10^{\circ}\text{C}$ threshold. - At $20^{\circ}\text{C}$: each day contributes $20 - 10 = 10$ degree-days. Days to reach 150 ADD $= 150 \div 10 = 15$ days → minimum PMI ≈ 15 days (reported as a range). - At $13^{\circ}\text{C}$: each day contributes $13 - 10 = 3$ degree-days. Days $= 150 \div 3 = 50$ days → minimum PMI ≈ 50 days. - Lesson: the estimate is acutely sensitive to the assumed temperature — a 7-degree difference more than triples the interval. This is exactly why temperature reconstruction is the dominant source of uncertainty (and why two experts who assume different temperatures get different windows; see Case Study 13.2).

Exercise 13.11

The maggot mass generated its own heat, so the larvae actually developed at a temperature higher than the surrounding air. If you use only ambient (cooler) temperature, you credit the insects with less daily warmth than they really banked, so you conclude it took them more days to reach their stage than it did — making the naive estimate too long (an overestimate of the interval). Mechanism: faster real development, slower assumed development → inflated day count.

Exercise 13.13

Larvae are difficult to identify to species; adults are far easier. Rearing live specimens to adulthood confirms the species, which fixes the correct developmental constants (threshold and required ADD). This guards against the misidentification error: wrong species → wrong constants → wrong interval. Preserved specimens (killed at the scene) freeze the developmental stage at the moment of collection; live ones secure the identity. You need both.

Exercise 13.15

Earliest → latest, with body state: 1. Blow-fly eggs in the openings — fresh stage (minutes to hours). 2. Third-stage blow-fly larvae — bloated / active decay (days). 3. Dermestid (hide) beetles — advanced decay → dry remains (tissue drying out; weeks). 4. Mites and moths on dry remains — dry / skeletal stage (weeks to months).

Exercise 13.17

The sentence is appropriately honest in at least three ways: (a) it says "no fewer than" — explicitly a minimum (a floor), not the time of death; (b) it gives a range ("approximately 9–13 days"), not a false-precision point; (c) it states its assumptions ("assuming the temperature reconstruction and species identification are correct"), exposing exactly where the estimate could be wrong. (Bonus: it says colonized, locating the clock's start at colonization, not death.)

Exercise 13.19

A burned body (the cold case) is a hard case because fire can kill/repel early colonizers and char the moist openings flies need, delaying colonization. Because colonization was delayed, the insects began their clock after a period in which the body was dead but uncolonized — so the insect age underestimates the true time since death. Direction of error: under-estimate. Hence the estimate is treated as corroborating, not leading: it sets a floor consistent with other evidence, but the fire makes it unreliable as the primary timekeeper.

Exercise 13.20 †

Problems with "Entomological analysis establishes that the victim died 9.7 days before discovery": 1. False precision. A single decimal-point figure misrepresents a method whose honest output is a range built on reconstructed, uncertain inputs. There is no basis for the tenths place. 2. "Died" vs. "was colonized." Entomology estimates time since colonization (a minimum PMI), not time of death. The word "establishes … died" overstates both the certainty and the quantity being measured. Honest rewrite: "Based on larval development and the reconstructed scene temperatures, insects colonized the remains no fewer than approximately 8–11 days before discovery — a minimum interval that would be longer if colonization was delayed."

Exercise 13.21

Overstatement: claiming pollen proves presence at one exact spot. Pollen evidence is associative class evidence: a profile can be consistent with / support contact with a place carrying that plant community, but many locations can share similar flora, and pollen also travels on the wind (background "rain"). It cannot individualize a precise location the way DNA individualizes a person. Defensible version: "The pollen on the boots is consistent with, and (given the assemblage's distinctiveness) supports, contact with a place carrying this plant community."

Exercise 13.23

Missing elements that should trigger distrust: (a) no error bars / range — an interval, not a point, is the honest form; (b) no species identification stated — without it the developmental constants are unjustified; (c) no temperature reconstruction described — the dominant input is hidden; (d) implicitly, no statement of assumptions or of whether colonization could have been delayed. A precise number stripped of all four is the signature of an overstated entomological claim.

Exercise 13.24 †

Risk: being told the "expected" date before analysis lets that number anchor the work (contextual/confirmation bias, Chapter 31) — ambiguous stages and specimens get read toward the wanted answer, an inconvenient beetle gets explained away. Safeguard: keep domain-irrelevant information (the suspect, the "expected" date) away from the analyst until the entomological estimate is fixed, then compare to the case facts (sequential unmasking / context management). A blind estimate is worth MORE than one computed knowing the answer — even when they agree — because only the blind estimate is uncontaminated by the expectation; agreement that might be bias-driven cannot be distinguished from genuine independent corroboration.

Exercise 13.25

Not a scandal — consistent with the method. Entomology's final interval is built from reconstructed, uncertain inputs (temperature history, colonization-delay assumptions, species data, specimen reading; §13.3–13.4). Reasonable experts can weigh those inputs differently and report different — even non-overlapping — ranges in good faith. Overlapping ranges in particular are a sign of honest uncertainty, not fraud. (Contrast a clean single-source DNA profile, whose tight error structure should not produce sharply divergent results between competent analysts — see Case Study 13.2.)

Exercise 13.27

Even with sound underlying plant science, the forensic application's reliability depends on the practitioner and the reference data. The scarcity of qualified palynologists and the thin error-rate literature mean a court has less independent basis to gauge how often the application errs, and fewer experts available to check a result — exactly the kind of validation gap the PCAST 2016 report flags as decisive (Chapters 5–6). Sound science applied by an unvalidated process, with no measured error rate, is still a method whose courtroom weight a careful judge should scrutinize.

Exercise 13.28 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited): 1. Single-source nuclear DNA — STRONG. Quantified, rigorously validated, known error structure; the field's gold standard (Chapter 7). 2. Accumulated-degree-day estimation — moderately strong / real-but-bounded. Core relationship (development vs. accumulated temperature) is grounded in physiology, is testable, and has been tested in rearing studies; yields an interval with stated uncertainty. Below DNA because the inputs are reconstructed and noisy, but well above the discredited methods. 3. Forensic palynology — lower-middle. Underlying botany is solid, but the forensic application has a thin validation record, few practitioners, and sparse error-rate studies; powerful for association, easily overstated into individualization. 4. Bite-mark comparison — DISCREDITED. No validated basis for the specific claim that one set of teeth made a mark; multiple exonerations (Chapter 16). At the bottom. Justification turns on each method's measured error rate and the rigor behind its core claim — the PCAST question.

Exercise 13.29

No soot in the airways (Chapter 11) shows the victim was dead before the fire. So during the fire the body was dead and (because fire chars openings and repels/kills early colonizers) likely uncolonized. The §13.4 caution is that fire delays colonization, making insect age an underestimate of time since death. The two findings are mutually consistent: the victim was already dead while the fire suppressed the insect clock from starting — which is precisely why, in this case, entomology corroborates rather than leads.

Exercise 13.31 †

Cold-case evidence-log entry (model): - (a) Defensible inference: the vehicle was at, or picked up material from, the cabin's distinctive environment. - (b) Honest verb: supports (resting on the distinctiveness of the assemblage and the quality of the control samples). - (c) Does NOT establish: who was driving; when the contact occurred; that a crime was committed; that the vehicle was at the precise clearing to the exclusion of all similar habitat. (Any three suffice.) - (d) Why no name yet: the pollen associates a vehicle with a place-type, not a person with the scene; attaching a driver requires other evidence (e.g., who controlled the vehicle, plus digital/soil/DNA links). To name a person now would be to claim more than the evidence has earned — the chapter's and the book's core discipline.

Exercise 13.33

Two evidence types that, with the vehicle association, begin to connect a person to the cabin: e.g., soil on a suspect's boots matching the cabin's distinctive soil (Chapter 24, previewed) and cell-site/location and deleted-message data placing a suspect's phone near the cabin (Chapter 25, previewed); touch-DNA on the gas can (Chapters 7–9) is a third. The pollen alone cannot make the leap because it associates the vehicle with a place, and says nothing about who controlled the vehicle or when — person-level and time-level links must come from other evidence.

Exercise 13.35 †

ADD vs. palynology — validity foundations (model answer). Accumulated-degree-day estimation rests on a physiological law — ectotherm development rate as a function of accumulated temperature — that is testable and has been measured in rearing studies; its output is a quantified interval with stated uncertainty. It is strong as honest, bounded science and is most easily overstated when dressed in false precision (a decimal-point time of death) or when a minimum is silently converted into the interval. Forensic palynology rests on solid botany — distinctive, durable, invisibly transferred pollen assemblages reflecting a place's flora — but its forensic application has a thin validation record, few qualified practitioners, and sparse error-rate studies; it is strong for association with a place-type and is most easily overstated when a profile "match" is presented as individualizing a precise location the way DNA individualizes a person. In short: ADD's risk is false precision about time; palynology's risk is false individualization about place.

Chapter 14 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 14.1

A friction ridge is a raised, continuous ridge of the hairless skin on the fingers, palms, soles, and toes (with furrows between), formed before birth and permanent for life. The two biological facts that make it useful for "who touched this?" are persistence (the pattern is fixed for life, barring deep scarring, so a print made today still matches the same finger years later) and variability (the detailed arrangement of ridges differs enormously between fingers, even between a person's own two thumbs and between identical twins).

Exercise 14.3 †

Minutiae (level 2 detail) are the specific points where friction ridges do something other than run straight and parallel. Three principal types: a ridge ending (a ridge simply stops), a bifurcation (one ridge splits into two), and a dot (a very short ridge). The configuration — not the count — carries the value because no single minutia is rare (bifurcations are common everywhere), but the relative positions, types, orientations, and the number of ridges counted between many minutiae form a combination that is highly improbable to recur. A dozen minutiae in agreement, in the same relative arrangement, with no unexplained differences, is a strong individual-characteristic argument; a dozen isolated bifurcations, considered one at a time, is not.

Exercise 14.5

An AFIS search returns a list — specifically a ranked candidate list of the stored records scored most similar to the query print. What it does not produce is an identification. The actual identification (if any) is made afterward by a human examiner performing ACE-V against the candidate exemplars, with all of that step's subjectivity and error intact.

Exercise 14.7

The three families: a loop (one or more ridges enter from one side, recurve, and exit the same side); a whorl (contains at least one ridge making a complete circuit); an arch (ridges enter one side, rise, and exit the other side, with no recurve or circuit). The arch is the rarest of the three.

Exercise 14.9 †

The dishonest word is "match." "Both prints are loops" is a statement about level 1 detail — the overall pattern, which is a class characteristic (Chapter 1, §1.3): a loop is shared by roughly six in ten fingers on Earth. Agreement at level 1 can exclude a source (a whorl latent cannot have come from an arch finger) or fail to exclude one, but it can never identify, because it narrows only to the majority of all human fingers. A "match" — an identification — requires level 2 agreement at minimum: a sufficient configuration of corresponding minutiae with no unexplained differences. Calling level-1 consistency a "match" smuggles individual-level certainty out of class-level evidence.

Exercise 14.11

The same residue that holds a latent print also holds touch DNA (Chapter 8), and several print-development chemicals can degrade or destroy that DNA. So the order of processing decides which evidence survives: aggressively dust or dye-stain an item first and you may destroy the DNA before it is collected; swab for touch DNA first (or choose DNA-compatible print chemistry) and you can preserve both. Cyanoacrylate (superglue) fuming is relatively DNA-friendly; some powders and dye stains are the general class that is not. The modern lab therefore sequences the workflow deliberately — often swabbing for touch DNA before aggressive print chemistry — rather than reflexively dusting. Get the sequence wrong and you trade one piece of evidence for another, permanently.

Exercise 14.13

(Even-numbered, not required; brief note for completeness.) Skip — see daggered/odd solutions. The number-one AFIS candidate is the closest stored record, not the source; the list always has a number one even when the true source is unenrolled, which is exactly why a high rank must be treated as a starting point for independent comparison, not as a result.

Exercise 14.15

Level 3 features (pore positions, ridge-edge shape) are deposition-sensitive: the same finger pressed twice will not reproduce its pore positions identically, because pressure, moisture, the surface, and the development method all alter how the fine detail records. On a clean, high-quality print, the level 3 detail that does appear is reliable enough to add discriminating power. On a smeared, poor print, the apparent level 3 features may be artifacts of how the latent distorted — so an examiner can "see" a pore agreement that reflects the smearing, not the finger. The mechanism is the same feature; the difference is whether the print preserved it faithfully or manufactured a false version of it.

Exercise 14.17

The sentence is appropriately disciplined in at least three ways: (a) it names the method ("I conducted an ACE-V comparison"), making the basis auditable; (b) it states what was found and its quantity ("agreement in pattern and in 12 minutiae in corresponding positions, with no unexplained differences"), rather than a bare conclusion; and (c) it frames the conclusion as an opinion about source ("in my opinion the latent and the exemplar originated from the same source"), not as a metaphysical certainty or a claim about "all other fingers on Earth." (Bonus: it does not attach a number like "100%" or "zero error rate.")

Exercise 14.19

No — not necessarily. §14.4 states that ACE-V has no defined, objective threshold for when "enough" agreement justifies an identification; the examiner makes a holistic judgment. On a difficult latent, two competent, honest examiners can legitimately weigh ambiguous detail differently and reach different conclusions — one "identification," one "inconclusive." That is expert judgment under genuine uncertainty, not proof of incompetence or dishonesty. Contrast a clean single-source DNA profile (Chapter 7), whose comparison yields a quantity with a tight, well-characterized error structure: two competent analysts there should not produce wildly divergent conclusions. The divergence on the latent reflects the method's location on the validity spectrum — a structured judgment with wider, less-quantified uncertainty — not a scandal.

Exercise 14.20 †

Two scientifically indefensible claims: 1. "Zero error rate." Empirical examiner-accuracy studies (e.g., the FBI/Noblis-type work) put the false-positive rate low but not zero; "zero" is a claim the method cannot support, and the NAS 2009 report specifically condemned it. 2. "To the exclusion of every other person on Earth." This is absolute individualization, which assumes that uniqueness of the source plus a partial-latent comparison establishes certainty — exactly the non-sequitur §14.1 dismantles. No examiner has compared this latent to every finger on Earth, and the conclusion is an opinion under a real error rate. Honest rewrite: "I conducted an ACE-V comparison and found agreement in pattern and in the minutiae I documented, with no differences I could not explain; in my opinion the latent and the exemplar originated from the same source. I cannot place a figure such as 'certainty' or 'zero error rate' on that opinion, because the method does not support one."

Exercise 14.21

The logical move is a non-sequitur (the chapter calls it out by structure in the Junk-Science Alert): "fingerprints are unique, therefore fingerprint identification is infallible" jumps from a premise about nature to a conclusion about a human task. The snowflake analogy: snowflakes are also said to be unique, but that does not mean you could pick one snowflake out of a blizzard from a blurry photograph of part of it. Uniqueness of the source (fingers, snowflakes) is a claim about the world; reliability of the comparison (an examiner matching a smeared partial latent to the right finger) is a claim about people doing a task — and only the second can be measured, only the second has a non-zero error rate, and only the second is what convicts someone.

Exercise 14.23

Questions a careful reader/cross-examiner should ask, and why each matters: (a) What was the quality of the latent? — a partial, distorted latent is where coincidental similarity and examiner error live (§14.3). (b) Who decided the points agreed, and was the analysis of the latent completed and documented before the exemplar was seen? — because the reference print can bias the reading of the latent (§14.4). (c) Was the verifier blind to the first examiner's conclusion? — non-blind verification is confirmation, not independence (§14.4). (d) What is the documented error rate for examiners on latents of this quality? — "14-point match" without an error rate hides the one number that locates the claim's reliability (§14.4, §14.6). A "14-point identification" stripped of these is the signature of an overstated print claim.

Exercise 14.25

Ordinary verification lets the second examiner know the first examiner already declared an identification before re-examining; blind verification withholds that conclusion so the second examiner reaches their own independently. When the verifier knows the answer, "the social and psychological pressure runs hard toward agreement" (§14.4) — verification becomes confirmation, and the independence is nominal. Therefore two examiners agreeing is not automatically twice as trustworthy: if the second knew the first's result, the second judgment may be the first judgment echoed, not a separate test. (This is precisely the Mayfield dynamic — multiple "agreements" that were one primed conclusion wearing several coats.)

Exercise 14.27

"A fraction of a percent" of false positives is reassuring about the typical case because it means a competent examiner, on the run of ordinary comparisons, is very rarely wrongly declaring an identification — the method is conservative. It is not a license for "zero," because a small fraction of a percent, multiplied across the enormous number of comparisons the system performs, is a real and recurring number of wrong identifications — each one a potential Mayfield. The asymmetry (false negatives considerably more common than false positives) tells you examiners tend to be conservative: they err much more often toward "I can't say / inconclusive" than toward "wrongly yes," preferring to miss a true identification than to manufacture a false one.

Exercise 14.28 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited): 1. Single-source nuclear DNA — STRONG. Quantified, rigorously validated, known and characterized error structure; the field's gold standard (Chapter 7). 2. Latent fingerprint comparison — foundationally valid, but a notch below. PCAST 2016 found properly designed studies show examiners can do the task at a measurable, usefully low error rate — far above the discredited methods. "Foundationally valid but with a non-zero error rate" means: the method works most of the time and has been tested, and its false-positive rate is small but real, so it cannot claim "zero" or absolute individualization; its validity also depends on examiners following bias-controlled procedure. 3. Microscopic hair comparison — weak / largely discredited for individualization (Chapter 19, previewed). The FBI's own review acknowledged systematic overstatement; it can support class-level association at best. 4. Bite-mark comparison — DISCREDITED (Chapter 16, previewed). No validated basis for the claim that one set of teeth made a particular mark; multiple exonerations. The bottom of the spectrum. Justification turns on each method's measured error rate and the rigor behind its core claim — the PCAST question.

Exercise 14.29

The autopsy's homicide finding (no soot in the airways → dead before the fire; blunt-force skull fracture, Chapter 11) established a defensible physical fact: the manner of death was not accidental. The gas-can latent's inconclusive result declined to claim anything about who handled the can — it neither identified nor excluded Keller. Both fit theme 1 (exclusion over proof): the autopsy excluded the accidental-fire hypothesis (a strong, well-grounded negative), while the latent honestly refused to prove a positive it could not support. Forensic science is at its most reliable when ruling possibilities out (homicide, not accident) and at its most disciplined when it declines to assert a positive (this is Keller's print) that the evidence has not earned.

Exercise 14.31 †

Cold-case evidence-log entry (model): - (a) Defensible conclusion: the partial latent on the gas-can handle is of insufficient quality and quantity of detail to identify a source; its pattern is consistent with a loop, which excludes very little. - (b) Honest verb (or refusal): inconclusive — neither an identification of anyone on the AFIS candidate list (including Keller) nor an exclusion. The only affirmative statement available is "consistent with a loop," which, being a class characteristic, attaches to roughly the majority of fingers and names no one. - (c) Does NOT establish (any three): that Keller's hand was on the can; that anyone on the candidate list is the source; when any contact occurred; that the print is even from the perpetrator rather than from innocent prior handling of the can. - (d) Why it would have been "Madrid in miniature": every condition that produced the Mayfield false positive was present here — a poor, partial latent; an AFIS candidate list that primed the analysts; and a suspect (Keller) the investigators wanted to place on the can — so declaring "Keller's print" from this latent would have repeated the exact bias cascade, manufacturing an identification the evidence could not support.

Exercise 14.33

Two evidence types that, combined, would begin to place Keller's hand or person at the cabin: e.g., soil on Keller's boots matching the cabin's distinctive soil (Chapter 24, previewed) and cell-site/location and deleted-message data placing his phone near the cabin against his alibi (Chapter 25, previewed); the DNA mixture consistent with him (Chapter 9) is a third, stronger-but-still-not-conclusive thread. The inconclusive print can do none of this on its own because it does not identify a source at all — it neither names Keller nor excludes him, so it adds no person-level or time-level information; person- and presence-level links must come entirely from other evidence.

Exercise 14.35 †

Biology vs. interpretation (model answer). The biological basis of fingerprints is sound: friction ridges form before birth through a developmental process so variable that no two fingers' full detail have ever been found alike, and they persist for life. That supports the undisputed application — verifying a known identity from a clean print. The interpretive method, comparing a smeared, partial latent to an exemplar via ACE-V, is a different thing entirely: a structured human judgment with no objective threshold, vulnerable to context and priming, carrying a real, measurable, non-zero error rate (false positives rare but not zero). Conflating the two — "the ridges are unique, therefore the examiner is infallible" — is the chapter's central error: it borrows the certainty of a fact about nature to dress up the reliability of a task performed by people. The honest version of the method sits above the discredited pattern disciplines (bite marks, hair) and well below single-source DNA — "foundationally valid" (PCAST 2016) but never the "gold standard" the marketing claimed — the fingerprint is a very good silver.

Chapter 15 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 15.1

Strictly, ballistics is the physics of projectiles in motion — how things fly when launched (internal, external, terminal phases). In the courtroom, "ballistics" is used loosely to mean the whole discipline of firearms examination, including the microscopic comparison of marks on bullets and cartridge cases. The part that is not really ballistics at all is the striation/toolmark comparison under the microscope — that is firearms identification, not the physics of a moving projectile. The chapter uses the loose sense "because everyone does," while keeping the distinction in mind.

Exercise 15.3 †

Rifling is the series of spiral grooves cut or formed into the interior of most firearm barrels, which forces the bullet to spin as it travels down the bore. The lands are the raised metal between the grooves (the high parts); the grooves are the cut-away spiral channels (the low parts) — exactly as on a landscape: high land, low groove. The naming trips up nearly everyone at first because the bullet ends up with the opposite pattern engraved into it (the lands cut grooves into the bullet's surface).

Exercise 15.5

Class characteristics are features shared by every barrel made to the same specification — caliber, number of lands and grooves, their width, and the direction and rate of twist. Examples: ".38 caliber"; "six lands and grooves, right-hand twist." Individual characteristics are microscopic irregularities unique (in theory) to one barrel — tool marks from manufacture, pits from rust, wear from use — that scrape a corresponding striation pattern onto a bullet. Examples: a particular pit-induced fine line; a distinctive group of striations from one worn barrel. Class excludes and narrows; individual is the (contested) basis for individualizing.

Exercise 15.7

Striations are the fine parallel lines scraped onto a fired bullet (and, generally, onto any surface a marked tool drags across) by the microscopic irregularities of the barrel's interior surface. In a barrel they are supposedly created by random tool marks, pits, and wear unique to that one bore — the "individual characteristics" of §15.2. Their correspondence between a questioned and a known bullet is what an examiner reads to argue same-source — and is precisely the reading whose reliability PCAST and the NAS found unmeasured.

Exercise 15.8 †

NIBIN/IBIS = the National Integrated Ballistic Information Network (the database, administered in the U.S. by the Bureau of Alcohol, Tobacco, Firearms and Explosives) and its imaging engine, the Integrated Ballistics Identification System. The single most important thing it does not do: it does not declare matches. It captures images of cartridge-case (and some bullet) marks, correlates them against the database, and returns a ranked list of candidate correlations — investigative leads that a firearms examiner must still confirm or reject by conventional comparison microscopy. "The database matched it" is the central error the section exists to prevent.

Exercise 15.9

Class characteristics on a fired bullet (caliber, number/width of lands and grooves, twist) are shared by every gun of that make and model, so they cannot point to one weapon — but they can exclude a candidate weapon cleanly (different caliber or twist = different gun) and narrow the field of possibilities dramatically. That exclusion-and-narrowing is a real, underrated power. What class characteristics can never do is individualize — say this gun and no other. The leap from "consistent class" to "this gun" requires the individual marks, and that leap is what the whole validity debate is about.

Exercise 15.11

The suspect's gun has not been identified as the source. What has been established: the recovered bullet's class characteristics (six lands/grooves, right-hand twist, the caliber) are consistent with the suspect's pistol — i.e., the pistol is not excluded on class grounds. What has not been established: that this specific pistol, and no other, fired the bullet. Many firearm models share those class features. Whether the individual marks (striations) correspond well enough to support an identification is a separate, harder, and more contestable judgment (§15.4). The honest verb at this stage is "consistent with / not excluded," never "identified."

Exercise 15.12 †

Conclusion: the suspect's revolver did not fire that bullet. The recovered bullet has five lands and grooves with a left-hand twist; the revolver produces six with a right-hand twist. These are class characteristics, and they are incompatible — a clean, defensible exclusion. It is the strongest kind of firearms conclusion because it does not depend on the subjective "sufficient agreement" judgment at all: it is a direct, objective mismatch of measurable class features, exactly the kind of exclusion §1.6 and §15.6 prize. (Theme 1: forensic evidence excludes far more cleanly than it individualizes.)

Exercise 15.13

An examiner test-fires the suspect weapon into a water tank or a box of dense gel because that stops the bullet without deforming it, recovering pristine bullets and cases carrying that gun's marks under controlled conditions. Firing into open air or a hard backstop would damage the bearing surface (the part that touched the barrel) and could lose the very striations needed for comparison. The test-fired bullet (or case) is the known — the reference specimen the questioned crime-scene item is compared against, questioned-versus-known, under the comparison microscope.

Exercise 15.15

Two problems with "read the surviving striations and call a match if enough line up" on a badly deformed bullet: (1) Deformation destroys the bearing surface, so striations are missing, not non-matching — and an examiner who counts "enough lines" from a fragment may be reading a small, unrepresentative remnant. (2) Missing detail cannot be distinguished from disagreement, and on a deformed bullet the temptation is to read marks that aren't really there or to over-weight a few coincidental correspondences. The honest response to a badly deformed bullet is usually "inconclusive" — a deformed bullet often "announces its own uselessness" (§15.2), and the danger it presents is false confidence.

Exercise 15.16 †

Two errors in "The database matched our casing to the gun": (1) NIBIN/IBIS does not declare matches — it returns candidate correlations, a ranked list of casings worth a human examiner's time; the algorithm nominated a candidate, it did not find a match. (2) No human confirmation has occurred — until a firearms examiner does the conventional comparison-microscope examination (and, ideally, does it as blindly as practical), nothing has been established beyond "this is a lead worth testing." What has actually been established at that moment: the imaging software scored these two items as similar enough to flag for examination — a hypothesis, not a finding, and certainly not something to present to a jury or the press as a match.

Exercise 15.17

Using a cartridge case's marks to determine the type of firearm relies on the positions and gross shapes of the marks — the round vs. elliptical firing-pin dent, the location of the ejector mark relative to the firing pin, the extractor position — which reliably indicate the kind of weapon and can exclude a candidate. Identifying the individual firearm relies on the fine texture of those marks (the breech-face micro-detail, individual striations) corresponding well enough to argue same-source. The first is on firm ground (class-level, validated, exclusionary); the second is the contested individualization claim that rests on subjective "sufficient agreement" without a validated error rate.

Exercise 15.19

Two legitimate reasons for "inconclusive": (a) the questioned bullet/casing is too deformed or low-quality to carry enough comparable detail; (b) the surfaces show agreement that could be subclass (shared across a batch) rather than individual, so the examiner cannot responsibly call same-source or different-source. The §15.4 worry is that "inconclusive" can become an escape hatch — a way to avoid the risk of an error on a hard comparison. Over-use makes a measured error rate look artificially low because an inconclusive is neither a hit nor a miss: it is excluded from the error count entirely, so a method that ducks its hardest calls appears more accurate than it is. (And in the real world, a wrongly withheld elimination can keep an innocent person in jeopardy.)

Exercise 15.21 †

(a) Honest stand sentence: "I found strong agreement in the class characteristics, and the fine breech-face detail corresponds with the test-fired case; in my opinion this supports the conclusion that the two cases were fired by the same firearm — stated as an opinion, with no validated random-match statistic available for these marks." (b) Overstated sentence: "These cases are identical and were fired by the defendant's gun, and that gun alone, to a scientific certainty." The honest version stays at supports / opinion; the overstatement asserts an identity to the exclusion of all others and a certainty the method has not earned (§15.4 Junk-Science Alert).

Exercise 15.23

Using the §15.1 At the Bench callout, two questions to ask before trusting a single-line, three-decimal trajectory: (1) How many bullet defects was it built from? A single defect with an assumed straight path is unreliable; two or more defects from the same shot are needed to constrain a path. (2) Was deflection (ricochet, glass, passing through a body) ruled out? An undetected deflection makes a precise-looking line meaningless. The honest output form of a trajectory analysis is a region with stated uncertainty — often a cone, not a line — because the entry and exit points carry measurement error. "Three decimal places" is a warning sign, not a sign of rigor.

Exercise 15.25

The three problematic parts of "fired by the defendant's pistol, to the exclusion of all other firearms, to a reasonable degree of scientific certainty": 1. "to the exclusion of all other firearms" — asserts a universal negative no one has tested against all firearms on Earth; the examiner has compared the questioned item to this gun, not ruled out every other. 2. "reasonable degree of scientific certainty" — a legal-sounding phrase with no agreed scientific meaning, which standards bodies and courts have urged examiners to abandon; it dresses a subjective opinion as objective certainty. 3. The implicit zero error rate / "identification as fact" — the sentence presents a same-source opinion (resting on subjective "sufficient agreement") as a fact of identity, implying a precision the discipline's measured error rate contradicts (§15.6, PCAST 2016). Honest rewrite: "I found agreement in the class characteristics and in the individual marks, exceeding what I have seen between different-source samples; in my opinion the recovered pistol is the source. I am stating an opinion, there is no validated random-match statistic for these marks, and I cannot exclude every other firearm in existence."

Exercise 15.27

Three separate inferential leaps in "the wound proves a .38, the defendant owns a .38, so the defendant's gun fired the fatal shot": 1. "The wound proves a .38." This is terminal ballistics (§15.1) — and a wound can, at best, suggest a caliber range, not prove an exact caliber; the leap overstates what a wound shows. (Wound → caliber is weak.) 2. ".38 → the defendant's .38." This conflates a class characteristic (caliber) with an individual identification; millions of guns are .38 caliber. Owning a gun of the same class does not make it the source (§15.2). (Class → individual is the unearned leap.) 3. "...fired the fatal shot" → the defendant fired it. Even if the weapon were identified, that speaks to the gun, not the hand (§15.1, §15.4 division of labor). Who pulled the trigger is a separate question entirely. The chain actually breaks at leap 2 (and is already weak at leap 1): caliber is a class feature, and the bridge from "a .38 was involved" to "this .38 fired it" requires individual-mark comparison the sentence never performs.

Exercise 15.29

Each piece of context can lower the threshold for perceiving "sufficient agreement": (a) the NIBIN lead naming the gun primes the examiner to expect a match, so ambiguous correspondences get read as agreement; (b) "the detective is confident this is our shooter" supplies a desired outcome, and a subjective criterion with a known wanted answer is a documented amplifier of error; (c) handing the casing together with that information means the examiner never forms an independent judgment first. The safeguard is context management / blind, sequential examination (Chapter 31): the questioned item is analyzed before, and shielded from, the known and the case context, so the human is testing the lead, not confirming it. (§15.4 calls this "theme three in its purest form": the threat is the analyst's mind, not the microscope.)

Exercise 15.30 †

A "verification" in which the second examiner is told the first examiner's conclusion before looking can confirm a shared error because the second examiner now approaches the comparison with the same expectation that may have produced the first error — the verification becomes a rubber stamp, not an independent test. If both were biased by the same context (a database lead, the detective's confidence, a subclass mark that fools two people the same way), agreement proves only that they agree, not that they are right. A genuine independent verification would require the second examiner to be blind to the first examiner's conclusion (and ideally to the case context) — reaching a judgment from the marks alone, so that agreement carries real independent weight.

Exercise 15.31

The "sufficient agreement" standard is defined, in part, relative to the best agreement the individual examiner has ever personally seen between different-source samples. That makes the threshold a function of one expert's accumulated, undocumented experience — it cannot be written down as a number, reproduced by another examiner, or audited by the court. To a jury it is therefore hard to present as "objective science," because the decisive comparison happens inside the examiner's trained judgment, not against a published, validated criterion. The examiner may be highly skilled and entirely honest, and the judgment may be good — but "trust my experience" is structurally different from "here is the measured error rate and the objective threshold," and the chapter insists the difference be made plain.

Exercise 15.33 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited): 1. Single-source nuclear DNA — STRONG. Quantified random match probability, rigorously validated, known error structure (Chapter 7). The summit. 2. Latent-fingerprint comparison — contested-but-relatively-strong. A comparison discipline (ACE-V) with a real, if subjective, basis and some measured error-rate studies; "the gold standard" that Mayfield (Chapter 14) showed is still human judgment under bias. 3. Firearms identification — contested middle. Class characteristics are validated and exclusionary; same-source grouping is strong; but the individualization claim rests on subjective "sufficient agreement" without a validated, quantified error rate (§15.6). 4. Bite-mark comparison — DISCREDITED. No validated basis for the specific claim; multiple exonerations (Chapter 16). The bottom. Firearms vs. fingerprints: reasonable reviewers disagree about the order because both are subjective comparison disciplines, but fingerprints have a somewhat larger error-rate literature and longer-validated feature set, while firearms has the cleaner class-level exclusionary power; PCAST treated both as feature-comparison methods needing more black-box validation, and some reviewers place firearms slightly below latent prints precisely because its individualization standard is even less quantified. The justification turns, as always, on the measured error rate and the rigor behind the core claim.

Exercise 15.35

This chapter advances at least two themes. Theme 1 (exclusion over proof): firearms evidence excludes cleanly on class grounds (wrong caliber/twist = different gun) and groups same-source items well, but its individualization claim is only a qualified opinion — and the cold-case casing's most important property is that, against the autopsy, it is irrelevant, teaching that recognizing irrelevance is itself a forensic skill. Theme 2 (the validity spectrum): the chapter explicitly places firearms identification in the contested middle — validated at the class level, subjective and unquantified at the individualization level — above the discredited bite mark (Chapter 16) and below DNA (Chapter 7), using the NAS 2009 and PCAST 2016 reports as the yardstick. (It also touches Theme 3, the subjective "sufficient agreement" decision as a model of bias, and Theme 4, the television "ballistics match" that means "certainty.")

Exercise 15.37 †

Cold-case evidence-log entry (model): - (a) Defensible status: a single fired 9 mm cartridge case was recovered from the cabin debris near the threshold; entered into NIBIN/IBIS (low-confidence leads only); against the autopsy finding of no gunshot wound, it is a probable red herring — present at the scene but inconsistent with the established manner of death (blunt-force trauma). - (b) Honest verb: irrelevant to the manner of death — it neither implicates nor excludes. (Not "matches," not "links.") - (c) Does NOT establish: that Diallo was shot (the body says he was not); when or by whom the case was fired (it may be old — a prior occupant, hunter, or target-shooter); that any person of interest was at the scene; or any link to the events of 18 October. - (d) Why statuses are unchanged: the casing implicates and excludes no one; because the autopsy removes any role for firearms in the death, the item carries no weight for or against Keller, Whitfield, Salas, or Renner. Pursuing it as a "ballistics match" would chase the case away from where the evidence points — the chapter's lesson in relevance.

Exercise 15.39

The casing illustrates theme one better than a dramatic "match" would because its forensic value lies entirely in an exclusion of relevance: the disciplined, correct action is to recognize that, against the autopsy, the item has nothing to do with the crime, and to say so — refusing to be led by an evidentiary item that looks important (a casing implies a gun implies a shooting). A flashy "match" would have tempted the investigation toward a false story (a shooting that never happened); the honest "this is irrelevant" protects the case from that error. Recognizing irrelevant evidence is "as much a forensic skill as any comparison" because the cost of chasing a red herring — wasted resources, a contaminated theory, possibly an innocent person implicated — is exactly the kind of harm the whole book is organized to prevent. Saying "this changes nothing" clearly is the forensic work here.

Chapter 16 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 16.1

Impression evidence is physical evidence in which the shape of one object is transferred to, and recorded on, the surface of another (a footwear impression in mud, a tire track in snow, a tool-gouge in a doorframe, a bite in skin). The single piece of physics it all rests on: when a harder surface contacts a softer one with enough force, the softer surface deforms to record the contours of the harder one.

Exercise 16.3 †

A cast is a three-dimensional reproduction of an impression, made by pouring a casting medium (typically dental stone) into a three-dimensional impression and letting it cure into a positive replica of the source. Negative vs. positive: where the shoe's sole stuck out (a lug, a ridge), it pushed into the soil, so the soil holds the inverse of the sole — a negative. Dental stone poured into that void fills the depressions and rises where the soil was pushed down, curing into a hard positive — a replica of the bottom of the shoe itself. This matters because you cannot directly lay a hole-in-the-ground beside a suspect's shoe; you can lay the positive cast beside it (and beside a test impression made with the shoe) and compare feature for feature.

Exercise 16.5

The four-rung hierarchy of conclusions, in order: 1. Exclusion — the known source could not have made the questioned impression (wrong size, wrong tread design, an unexplained difference). The cleanest and often most defensible call. 2. Class association ("could have made / is consistent with") — questioned and known share class characteristics; the source and every other object of its class cannot be excluded. 3. Association with individual characteristics ("identification" in some labs) — class agreement plus enough corresponding acquired damage that the examiner concludes the source made it. The contested top rung, where overstatement creeps in. 4. Inconclusive / unsuitable — the impression is too faint, partial, or smeared to support any of the above.

Exercise 16.7

Bite-mark analysis is described here as the claim that a forensic odontologist can examine a patterned injury on skin and determine which person's dentition made it, often identifying a single individual as the biter. It sits at the discredited end of the validity spectrum because there is no validated method establishing that human dentition is uniquely identifiable as expressed in a bite mark on skin, that skin reliably records that detail, or that examiners can reliably link a mark to one biter — and when tested, examiners performed at rates incompatible with their courtroom certainty.

Exercise 16.9

Two reasons the photograph is always taken before casting (and never skipped): 1. The photograph is the backup; the cast is the prize. Casting is slow and unforgiving — a cast poured too thin, or lifted before it cures, can shatter and take the only record of the impression with it. If the cast fails, the examination-quality photograph is the surviving record. 2. Casting (and even lifting) is partly destructive / alters the original. Pouring medium into the impression contacts and can disturb it; recovery should be preceded by a faithful, in-place record. Photographing first preserves the impression's location and detail before anything touches it.

Exercise 16.10 †

Setup of the examination-quality photograph: - Camera orientation: the lens is placed parallel to the plane of the impression (looking straight down on a print in the ground), so the image is free of perspective distortion and the scale reads true. - The scale: a ruler — ideally an L-shaped scale showing two axes — placed in the same plane as the impression, so measurements taken from the photograph are accurate. - The lighting: oblique light raked across the impression from a low angle (a flashlight held low and to the side). What oblique light accomplishes: a three-dimensional impression has depth, but under flat, straight-on light that depth is invisible — the tread detail washes out. Light raked in from a low angle throws the raised and recessed detail into relief by casting small shadows, making tread pattern and fine features leap out that flat light would hide. It is the single most valuable record of a 3-D impression.

Exercise 16.11

A useful tire impression may run for several feet, but a single cast can only span so much before it becomes unwieldy or fragile, so the examiner casts the impression in adjacent sections (a long impression in soil may take several pours). What must be recorded: the sequence and relative position of the sections (which abuts which, in what order along the track), so the lab can reassemble them and reconstruct the tire's repeating tread pattern — the full circumference is rarely captured in one impression, so the examiner works with whatever arc of tread is present and must know how the pieces fit.

Exercise 16.12 †

The possibility that must be ruled out is the subclass-characteristic trap (Chapter 1). A subclass characteristic is a feature finer than the broad class but still shared by a subset of tools made by the same manufacturing process — tools cut or ground by the same machine in the same batch can carry similar fine marks. A striated "match" built on subclass marks will associate the questioned mark with a whole production run of tools while looking like an individual identification, because the examiner sees fine corresponding striations and reads them as unique. Distinguishing genuinely individual striations from subclass artifacts is the toolmark examiner's central skill — and it is a judgment, not a measurement.

Exercise 16.13

Unlike a fired bullet (where the same barrel acts on each round in much the same way, with constrained geometry), a pry bar can be applied at any angle, with any force, on any material, so the resulting mark varies enormously from use to use. Two variables the examiner must reproduce when making test marks: the angle of application and the pressure/force (a third is the substrate/material the mark is made in). A poor reproduction of the questioned conditions yields a poor — and potentially misleading — comparison.

Exercise 16.15

Most → least defensible, with justification: 1. (b) "could not have been made by the suspect's shoe" (exclusion) — the most defensible call: a genuine, unexplained difference excludes, and that is the surest thing comparison evidence does. 2. (a) "consistent with the class characteristics" (class association) — well supported; it is, at bottom, manufacturing information, but it implicates the suspect's shoe and every other shoe of that design and size. 3. (c) "made by the suspect's shoe, to the exclusion of all others" (individual identification) — admitted in most courts but the contested top rung: an opinion about specificity, dependent on examiner judgment and a thin error-rate literature. 4. (d) "too smeared to evaluate" (inconclusive) — not a failure, but it establishes nothing; ranked last because it supports none of the others (it is the honest "no usable conclusion"). (Note: (d) is "least defensible" only in the sense of "least informative"; it is perfectly honest. The ordering is by evidentiary strength of the conclusion reached.)

Exercise 16.17

The sentence guards against the wrong lesson that a method which happened to point the wrong way was therefore "almost right" or merely "weak evidence" — as if a little more care would have fixed it. The correct lesson the figure teaches: the DNA exposed that the bite-mark method's confident identification carried no real reliability in the first place. It was not weak evidence pointing the right way; it was an unvalidated method pointing the wrong way. The fix is not "do bite marks more carefully"; it is to recognize the method does not have a validated basis for the claim it made.

Exercise 16.19

The hidden prior assumption is that the examiner can reliably establish, in the first place, that the patterned injury is a human bite mark and that bite marks on skin carry enough faithful, comparable detail to support any association with a specific dentition. "Cannot be excluded" only means something if there is a validated comparison in which exclusion is even meaningful — and bite-mark analysis lacks exactly that. The assumption is shaky because, when tested, examiners could not even reliably agree on the threshold question of whether a given injury was a human bite, let alone whose; so "cannot be excluded" smuggles in a reliability the method does not have.

Exercise 16.21

Overstatement: claiming a bite mark matches a defendant's teeth "to the exclusion of all others" — an individual identification. Bite-mark identification has no validated basis for that claim: skin is a poor, changing substrate, dentition has never been shown to be uniquely identifiable in a bite mark on skin, and examiners fail error-rate testing. What an examiner could honestly say under current standards: at most that an injury is consistent with a human bite, and possibly that a suspect cannot be excluded — never an individual identification. Why even that is contested: examiners cannot reliably distinguish a human bite from other patterned injuries, and a number of courts and scientific bodies now bar or sharply limit any bite-mark testimony at all — so even "consistent with a human bite" carries less weight than it sounds, and "cannot be excluded" rests on a comparison the method cannot validly perform.

Exercise 16.23

Missing elements that should trigger distrust of a footwear "identification": (a) the number of corresponding acquired (individual) features found — identification is an opinion about how many random marks correspond; (b) how distinctive those features are — a few common scuffs are not the same as several specific, well-placed gouges; (c) the error rate / validation basis against which the conclusion is offered — identification is a judgment, not a calculation, and the examiner should say so; (d) implicitly, whether class agreement alone was silently upgraded to identification. A confident "it's his shoe" stripped of feature count, distinctiveness, and error rate is the signature of an overstated impression claim.

Exercise 16.24 †

Risk: being handed the tool with "this is the one — confirm it" lets the wanted answer anchor the analysis (contextual/confirmation bias, Chapter 31). With a mark that varies by angle and force and striations open to interpretation, the examiner who expects a match counts ambiguous correspondences as agreement and explains away differences. Safeguard: keep domain-irrelevant information (who the suspect is, which tool the detective favors) away from the examiner; reach preliminary conclusions from the questioned mark before seeing the favored tool, and have a second examiner verify blind (context management / sequential unmasking). A comparison reached before knowing the favored answer is worth MORE — even if it agrees — because only that comparison is uncontaminated by the expectation; an agreement that might be bias-driven cannot be distinguished from genuine independent support.

Exercise 16.25

A genuine, unexplained difference between a questioned mark and a known source is, by the exclude-vs-associate asymmetry of §16.1, grounds for exclusion: a feature the known source could not have produced means the known source did not produce it. A bite mark reportedly showing more upper teeth than Brown possessed is such a feature — on its face, an exclusion of Brown. That the testimony "minimized rather than treated [it] as the exclusion it arguably was" means the discipline inverted its own logic: instead of letting the difference exclude, the examiner explained it away to preserve the match the prosecution wanted. A method that can convert a difference into a match has abandoned the one thing comparison evidence does best.

Exercise 16.27

A professional board (the ABFO), credentials, standardized guidelines, and a scoring vocabulary supply the appearance of standardization — but they do not supply the missing thing, which is validation: well-designed studies measuring whether examiners can actually do what they claim, with a known error rate. Standardized forms make practitioners consistent with one another in procedure; they do not show the procedure produces correct answers. Bite-mark analysis had the trappings of science — a board, guidelines, experts, overlay exhibits — and almost none of its substance, which is exactly the circular validation (admitted because its own practitioners vouch for it) the 2009 NAS report identified across the pattern disciplines.

Exercise 16.29

This chapter is "the validity spectrum compressed into a single chapter" because the same family of evidence — impressions and marks, all governed by one physics and one comparison logic — contains both genuinely defensible work and the field's most discredited discipline. At one end, footwear/tire class evidence (tread design, size, dimensions) is grounded in manufacturing fact; toolmark class evidence supports forced entry honestly. At the other end sits bite-mark identification, with no validated basis and a trail of exonerations. In between lie the contested individual-source claims (footwear, toolmark, firearms). The single separating question is the PCAST foundational-validity question: is there a validated method, with a known error rate, for the specific claim being made? For a tire's class characteristics, broadly yes; for "these teeth and no others made this wound," emphatically no.

Exercise 16.31 †

Cold-case evidence-log entry (model) — the cabin door's pry marks: - (a) Defensible inference: the door was forced; the gouges and compressed profile in the doorframe are consistent with a flat pry bar of a certain width and shape. This supports forced entry. - (b) Honest verb: consistent with (a class of tool) → supports (forced entry). A class-level conclusion. - (c) Does NOT establish: which pry bar (no validated individualization of a general toolmark to one tool, PCAST 2016); whose hand held it; when the door was forced; or that any particular person was present. (Any three suffice.) - (d) Why "forced entry supported" does not place a named person at the door: the marks date and characterize the act (a door forced from outside by a class of tool), not the actor. A class match to a pry bar is a fact about the door, not a name. Placing a person there requires other evidence (soil, DNA, digital/cell-site, CCTV) developed in later chapters; to attach a name now would claim more than the evidence has earned — the chapter's and the book's core discipline.

Exercise 16.33

Had an over-eager examiner found a patterned injury at Mill Creek and "matched" it to a person of interest, the cost would have been a conviction-grade exhibit built on an unvalidated method — exactly the trap that took Krone (a confident false identification) and Brown (a difference overridden into a match). A bite-mark "match" would have projected near-certainty onto skin, a substrate that cannot support it, and could have anchored the whole investigation on the wrong person while the real perpetrator went unaccountable. Given those two cases, "not using bite marks" is the disciplined choice, not a missed opportunity: a confident comparison from an unvalidated method is worse than no comparison at all, because it carries persuasive weight it has not earned. The most valuable thing bite-mark analysis contributes to this file is the discipline of not using it.

Exercise 16.35 †

Footwear/tire class evidence vs. bite-mark identification — validity foundations (model answer). Footwear and tire class evidence rests on manufacturing fact: a tread design and size are engineered into a finite production run, so "this impression is consistent with this model and size" is grounded information, and outsole/tread databases can yield a make and model. It is genuinely strong as class evidence — durable, measurable features arising from a known process — and is most easily overstated only when an examiner leaps from "consistent with this design and size" to "it's his shoe and no other" without the acquired-feature count to support it. Bite-mark identification rests on a stack of premises that were asserted, never validated: that dentition is unique, that skin (elastic, mobile, healing, distorting) faithfully records that uniqueness, and that examiners can reliably link a mark to one biter. What the 2009 NAS and 2016 PCAST reports identified as decisive and missing is exactly this: there are no well-designed studies showing examiners can accurately associate a bite mark with a specific person, with a known error rate — and when tested, examiners disagreed with the ground truth and each other, sometimes on whether a mark was even a human bite. In short: footwear/tire class evidence has a validated, manufacturing basis for the modest claim it makes; bite-mark identification has no validated basis for the enormous claim it made, which is why one is a workhorse and the other is discredited.

Chapter 17 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 17.1

Forensic odontology is the application of dental science to legal questions. Its one valid core contribution is dental identification — naming the dead by comparing antemortem dental records to a postmortem examination of the dentition. The one discredited application it is contrasted with is bite-mark comparison (owned by Chapter 16), which claims a specific person's teeth made a specific mark in skin.

Exercise 17.2 †

• Antemortem records — created during the person's life, by their dentist, orthodontist, oral surgeon, or institutional dental service; they are the known side of the comparison (the dental equivalent of a reference sample, Chapter 3). They include radiographs, charts, treatment/billing notes, and study models.
• Postmortem records — created from the body, after death, by the forensic odontologist (charting and radiographs of the dentition); they are the questioned side. Wherever practical, the postmortem record is made first and independently, before the antemortem file is studied in detail, to protect against confirmation bias.

Exercise 17.3

Dental identification is establishing the identity of a deceased person by comparing antemortem dental records to a postmortem examination of the dentition. It answers the prior investigative question — who is this body? — which must be settled before "who killed him?" can be asked: you cannot determine the cause and manner of one person's death, notify a family, settle an estate, or charge a suspect with his murder until you know whose body is on the table.

Exercise 17.5

Dental charting is the standardized, tooth-by-tooth, surface-by-surface notation of the state of the dentition (present, missing, or restored, and how). A shared convention is a precondition for comparison — not a convenience — because an antemortem chart made by one dentist and a postmortem chart made by an odontologist years later must describe the same mouth the same way for any concordance or discrepancy to be meaningful. Without a common framework, the two records are not comparable at all (and mismatched numbering systems are themselves a real source of error; §17.3).

Exercise 17.7

The four conclusion categories (ABFO logical structure): 1. Identification (positive/established) — concordant features of sufficient quantity and specificity, with no unexplainable discrepancies. 2. Possible / consistent — concordant, no unexplainable discrepancies, but the data are insufficient to be confident (sparse record, few restorations, no films). 3. Insufficient evidence — the records permit no conclusion (too little antemortem data, or remains too fragmentary). 4. Exclusion — the antemortem and postmortem data are irreconcilably inconsistent; an unexplainable discrepancy shows they are not the same person.

Exercise 17.9 †

Steps from recovery to conclusion, in order: 1. Recognize that dental ID is on the table and preserve the jaws — recover the cranium and jaws slowly, documenting in place; excavate (don't sweep) the debris, since small restorations, crowns, and tooth fragments are easily lost in ash; support or consolidate fragile jaws before lifting. 2. Postmortem examination — chart the dentition tooth-by-tooth and take postmortem radiographs using the same projections as common antemortem films. 3. Obtain a presumptive identity and request that person's antemortem dental records (the rate-limiting step). 4. Compare — ideally having charted the body first and independently; assess concordant features, classify discrepancies as explainable or unexplainable, and overlay antemortem and postmortem radiographs. 5. Reach a conclusion — identification / possible / insufficient / exclusion. The recovery hazard specific to burned teeth: after intense heat teeth become friable — they survive but turn brittle and can shatter at a touch — so the scene team must handle the jaws far more carefully than ordinary bone. "Slow is correct."

Exercise 17.11 †

The second decedent (extensive distinctive work + a complete recent film set) is readily identifiable; the first (perfect, untreated teeth + no dentist of record) may not be identifiable by odontology at all. The difference is in the record, not the teeth: dental identification is a comparison method, so its power is "teeth plus a record." Untreated teeth carry little documented, individually specific history, and with no dentist of record there is nothing on file to compare against — the dentition can be perfectly intact and still unidentifiable. Distinctive restorations and recent films give the comparison both discriminating features and an objective overlay anchor.

Exercise 17.13

Radiographic overlay compares (any three): root morphology (the shape and curvature of the roots), trabecular bone pattern (the internal spongy-bone texture around the teeth), sinus outlines (e.g., the maxillary sinus borders on upper films), the position and outline of restorations, and unerupted/impacted teeth. Overlay is more objective than comparing two written charts because it compares the actual imaged anatomy directly — superimposing one film on another to see whether structures coincide — rather than relying on two clinicians' notational descriptions, which can differ in style, completeness, and system even when describing the same mouth.

Exercise 17.15

The other two primary identifiers are fingerprints and DNA. The single property of teeth that earns dental identification its place among the primaries in fires and high-energy impacts is survivability: teeth and dental restorations frequently endure heat and force that destroy fingertips and degrade DNA, so the dentition is often the one richly informative tissue left.

Exercise 17.16 †

Figure 17.1, tooth #5 (present in the AM chart, absent in the body): - (a) It is an explainable discrepancy. Present→missing is biologically possible (a tooth present 14 months ago can be extracted since). - (b) The figure resolves it with a physical finding: the socket shows healed bone consistent with an extraction after the antemortem film — so the absence reflects later dental work, not the wrong body. - (c) The combination of concordant features — matching MOD amalgam on #3, the crown on #14, both third molars absent in both records, and the close radiographic overlay of root morphology and trabecular bone — supports a strong, concordant identification: very probably the presumed person. (No single feature identifies; the configuration does, and the lone discrepancy is explained.)

Exercise 17.17

The sentence is appropriately precise in at least three ways: (a) it specifies "concordant features of sufficient quantity and specificity," locating the strength in the combination, not one match; (b) it states "no unexplainable discrepancies," the exact logical condition for a positive identification (and the thing that would have excluded it); (c) it is framed as an opinion ("in my opinion they are from the same individual"), honestly marking the conclusion as an expert judgment rather than a metaphysical certainty. (Bonus: it speaks only to identity, not cause or manner of death.)

Exercise 17.19

A tooth recorded as crowned in the antemortem record that is, in the body, plainly sound and never treated, is an unexplainable discrepancy that excludes the identity — at high confidence. This one finding can outweigh several concordances pointing the other way because concordance is shared by an unknown number of similar mouths (so concordances only make an identity probable), whereas a genuine irreconcilable difference can come from only one place — the wrong body. (One caveat a careful examiner checks first: that the discrepancy is real and not a charting error in the antemortem record itself; if the record is verified, the exclusion stands.)

Exercise 17.20 †

The overstatement: claiming a suspect's teeth "match" a bite mark in skin "to a reasonable degree of dental certainty" is a bite-mark individualization claim — asserting these teeth made this mark — which the science cannot support (no established foundational validity; the skin substrate distorts; the uniqueness-and-faithful-transfer premise is unvalidated; §17.5, Chapter 16). Why the shared credential makes it dangerous: the same board-certified odontologist can validly identify a fire victim from dental records on Monday and make this invalid claim on Tuesday, and the jury hears one continuous expert authority with no way to know the first claim is well-founded and the second is not. The credential does not travel with the validity. The honest ceiling today: at most, an examiner may say a suspect cannot be excluded as a possible biter, or that an injury is consistent with a bite — and a growing position holds the comparison should not establish identity at all.

Exercise 17.21

"Dental identification is as certain as DNA" overstates in (at least) two ways: (1) error type — dental identification's conclusion is an examiner's judgment ("sufficient concordant features," "no unexplainable discrepancies" are not bright lines), whereas a single-source DNA match yields a quantified random match probability; (2) objective basis — DNA rests on a rigorously quantified statistical foundation, while dental ID, though sound, has no comparable quantified error rate, its errors dominated by human factors (charting mistakes, stale records, confirmation bias). Honest rewrite: "Dental identification is a valid, reliable method for establishing identity, sitting in the better-grounded middle of the validity spectrum — below the rigorously quantified certainty of single-source DNA, but well above discredited methods like bite-mark comparison."

Exercise 17.23

Provenance — whose the antemortem records are, who filed them, whether they were correctly attributed — is the chain on which the whole identification rests: a dental concordance only identifies the body as the person those records belong to. If the records were mis-filed, mislabeled, or (in a fraud scenario like the cold case's) deliberately substituted, even a strong concordance would point to the wrong identity. The chapter and the cold-case Case File keep this in view because the insurance-fraud theory itself is about a substituted body — so verifying that the records are genuinely the presumed person's is part of doing the identification honestly, not an afterthought. The dental match is necessary but not sufficient; the records' integrity must also hold.

Exercise 17.25

"Is the witness a qualified odontologist?" is the wrong question because the same credentialed expert can make a valid claim (identifying a body from dental records) and an invalid one (individualizing a bite mark in skin) — the credential certifies the expert, not the specific claim. The right question is: "Which question is this odontologist answering, and is that question one the method can validly answer?" A jury can rarely tell the two apart on its own because it hears one continuous expert authority in the same voice, with the same credentials, presented with the same confidence — the difference in validity is invisible without knowing the science behind each specific claim.

Exercise 17.27

"Slow is correct" is an ethical as well as a technical stance because the fragile, friable post-fire dentition is often the only identification route the body has left (fingerprints and face are gone). A rushed recovery that shatters the jaws or loses small restorations in swept-away ash can destroy the family's only chance of getting their relative named — and risks handing back the wrong body. The pressure of a grieving family or a press presence is exactly what tempts a fast recovery; yielding to it trades a person's identity for speed. What is lost if the jaws are handled like ordinary bone: irreplaceable concordant features (and the overlay anchor), and with them the identification itself.

Exercise 17.28 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited): 1. Single-source nuclear DNA — STRONG (top). Rigorously validated, quantified, known error structure (Chapter 7). 2. Dental identification — valid, better-grounded MIDDLE. Sound logical basis (documented, individually specific dental history), an objective anchor (radiographic overlay), and a clean exclusion rule. It is not at the top despite being valid because its conclusion is an examiner's judgment rather than a quantified probability, and its error is dominated by human factors (stale records, charting error, confirmation bias) rather than being statistically characterized. 3. Fingerprint comparison — upper-middle / valid-but-subjective. Strong discriminating basis and broad acceptance, but ACE-V is a human judgment with a real, measured error rate and famous failures (Mayfield, Chapter 14) — valid, applied by fallible humans. 4. Bite-mark comparison — DISCREDITED (bottom). No established foundational validity for the individualization claim; elastic/healing skin substrate; multiple exonerations (Chapter 16). Placement turns on each method's measured error rate and the rigor behind its core claim — the PCAST question — and dental ID's strength is logical-and-overlay-based rather than statistical.

Exercise 17.29

Neither finding depends on the other; they answer different questions. Dental identification answers who the body is (Marcus Diallo); the autopsy answers how he died ("dead before the fire," a blunt-force skull fracture → homicide, Chapter 11). The identity could be confirmed even if the manner of death were unknown — comparing the dentition to Diallo's records needs no autopsy conclusion — and conversely, the autopsy's "dead before the fire" finding is a fact about the body's physiology that holds regardless of whose body it is. In the cold case they are complementary, independent results: the dental ID gives the homicide a victim's name (so the death can be attributed to Marcus Diallo), but it neither establishes nor relies on the cause and manner.

Exercise 17.31 †

Cold-case evidence-log entry (model): - (a) Defensible conclusion / honest verb: the burned remains are Marcus Diallo — a positive dental identification (concordant antemortem/postmortem features, including close radiographic overlay of root morphology, sinus outline, and trabecular bone, with no unexplainable discrepancies). Verb: identification / "in my opinion the same individual." - (b) Theory this kills: the "someone else's body / faked-death / substituted-body" insurance-fraud theory — Diallo did not stage his death with a stand-in. - (c) Does NOT establish (any three): how he died; when he died; who is responsible; that the fire was arson; that the death is a homicide (those come from the autopsy/arson findings, not the teeth). - (d) Provenance caveat (logged alongside "no unexplainable discrepancies"): the conclusion assumes the antemortem records are genuinely Diallo's and correctly filed; the investigation verifies the records' source and chain, and the analyst keeps that dependency in view.

Exercise 17.33

Now that the body is confirmed as Marcus Diallo, the investigation can properly attribute to his death two findings already in the file, e.g.: the autopsy's "dead before the fire" + blunt-force skull fracture (Chapter 11) — i.e., Marcus Diallo was killed before the fire, a homicide; and the doorframe blood that predates the fire and spatter inconsistent with a collapse-in-fire (Chapter 10) — physical signs of force inflicted on Diallo before the burn. (The pre-fire timeline and the "dead before the fire" finding are the natural pairings.) Why none of this chapter's evidence points at a suspect: dental identification answers who the victim is, full stop. It says nothing about who controlled the scene, who set the fire, or who struck the blow — attributing responsibility requires the DNA, digital, financial, soil/pollen, and document threads from other chapters, converging at the capstone (Chapter 39).

Exercise 17.35 †

Dental identification vs. bite-mark comparison — validity foundations (model answer). Dental identification compares the body's dentition against a documented antemortem record of that same person's teeth, with a stable, mineralized substrate (teeth and restorations that do not distort), an objective anchor in radiographic overlay, and a clean exclusion rule; it sits in the valid, better-grounded middle of the spectrum, below quantified DNA but well above the discredited methods. Bite-mark comparison instead asks whether a set of teeth made a mark in human skin — an elastic, swelling, bruising, healing substrate that registers a bite differently by body region, movement, and time — against the unvalidated premise that a dentition is unique in the features a bite transfers and that skin faithfully records them; it has no comparable objective anchor and no established foundational validity, sitting at the discredited bottom. The single feature most responsible for the gap is the substrate-and-premise: identification compares a stable dentition to a record of that dentition; bite-mark comparison compares teeth to a distorting wound under a premise no study has validated.

Chapter 18 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 18.1

A questioned document is any document whose source, authenticity, or history is in dispute and is therefore subject to forensic examination. The paper, ink, and indentations count as "the document" just as much as the words because the discipline answers questions about the medium itself — what machine, ink, or paper produced it; whether parts were added or removed; what was impressed into the page from a sheet above. Many of the strongest findings (an alteration revealed under infrared, indented writing recovered by the ESDA) come from the physical medium, not from reading the words.

Exercise 18.3 †

A collected (non-request) exemplar is writing the person produced in the ordinary course of life before any dispute arose (old letters, signed checks, business records). Its chief virtue is that it is natural and unguarded — the writer had no motive to disguise; its chief peril is that you take what exists, which may not match the questioned writing in content, date, or writing conditions. A requested (dictated) exemplar is produced on demand, usually by dictating the questioned text and having the subject write it out, often repeatedly. Its chief virtue is control — you can match wording, instrument, and posture; its chief peril is that a subject who knows what is being collected may deliberately disguise their hand, or a suspected forger may have practiced the very writing in question.

Exercise 18.5

Indented writing is the pressure impression left in a sheet of paper by writing on a sheet that lay above it. The ESDA (electrostatic detection apparatus) is a non-destructive instrument that visualizes indented writing by developing the patterns of electrostatic charge held by the indentations. What the ESDA recovers: the content physically impressed in the page (text from a sheet that may since have been removed). What it cannot tell you: when the impression was made (it does not date the impression), and — crucially — who wrote the recovered text.

Exercise 18.7

A forensic document examiner (FDE) is a trained examiner — typically formed through a long apprenticeship and certified through a body such as the American Board of Forensic Document Examiners — who analyzes documents as physical and comparative evidence (authorship, alteration, indented writing, ink and paper). A graphologist practices graphology, the pseudoscientific claim that handwriting reveals personality, honesty, or character. The boundary matters in a courtroom because graphology has no validated basis and no place in forensic testimony; an "expert" who says a writer was "anxious and domineering" from the handwriting has wandered out of forensic science entirely, and letting the two be confused lends pseudoscience the credibility of the real discipline.

Exercise 18.9 †

One genuine exemplar is not enough because handwriting exhibits natural variation: a single writer's genuine writing is a range, not a fixed image — no one signs identically twice, and the slant, loops, and stroke lengths wander from instance to instance. With only one genuine sample, the examiner cannot tell whether a difference in the questioned signature is meaningful (a different writer) or merely the same writer on a different day. What the examiner needs instead is enough known writing to map the writer's range of variation, so that only differences falling outside that established range count as evidence of a different hand. (This is also the method's trap: an elastic "natural variation" can be stretched to dismiss real differences or to inflate trivial ones — which is why §18.3 ties it to bias.)

Exercise 18.11

A perfectly shape-matched forgery can be a worse forgery because matching the shapes closely requires the forger to draw slowly, copying the model — and slow drawing destroys the line quality of genuine writing. Genuine signing is a fluent motor act with smooth, rhythmic, naturally varying pressure; a drawn copy betrays itself through tremor, pen lifts, blunt/even pressure, and hesitation marks. So the better the shape match, the more the line quality tends to suffer. The giveaway feature is line quality, because it reveals the process of writing (fluent gesture vs. slow drawing), which is far harder to fake than the appearance.

Exercise 18.12 †

A "1935" letter written in a ballpoint ink not manufactured until after World War II establishes that the document could not have been written in 1935 — it cannot predate the material it is written in. It is a no-earlier-than boundary: the document was made no earlier than the ink's introduction date. It does not establish the actual date the document was written (it could have been made at any time after the ink existed), who wrote it, or why. First-appearance dating is powerful precisely as a one-sided boundary: it can conclusively impeach a claimed early date, but it cannot fix a precise date.

Exercise 18.13

At a line crossing, the examiner is trying to determine which ink line was written first — whether the upper line runs continuously over the lower, or how the pigments interact at the junction. It can matter to a case because sequence can show, for example, that a signature was added after the text it supposedly authorized, or that an entry was inserted later into a completed document. The honest answer is often "indeterminate" because line-crossing determination is sometimes reliable and sometimes genuinely cannot be resolved; an examiner who claims a certain sequence at every crossing is overstating the method, and "indeterminate" is the responsible report when the junction does not clearly show priority.

Exercise 18.15

Recovering content with the ESDA is more defensible than naming who wrote it because the ESDA recovers content that is physically present in the page — it makes visible writing that was impressed there, a demonstrable fact about the document, with no subjective same-hand judgment required. Reading recovered indented writing is closer to reading ordinary writing than to comparing two signatures. The moment you ask "whose hand?", you are back in the §18.2–18.3 problem: handwriting authorship comparison, the contested method with no quantified error rate that can exclude or find "consistent with" but cannot individualize to certainty.

Exercise 18.17 †

Figure 18.1 assigns four tells to the drawn (simulated) sample: 1. Pen lifts (gaps) mid-letter — supports: the signature was drawn, the hand stopping to check the model; does not support: who drew it. 2. Tremor in what should be a smooth curve — supports: slow, unfluent production (simulation); does not support: the identity of the forger. 3. Blunt, even pressure — supports: a drawn line rather than a fluently written one; does not support: authorship. 4. Hesitation marks where the writer paused — supports: copying a model (simulation); does not support: who the copier was. Across all four, the honest conclusion is "indications of simulation" (the signature was drawn rather than written) — a defensible, process-based call — and not any conclusion about which person produced it.

Exercise 18.19

This is close to demonstrative evidence because the jury can see the result with its own eyes: the room is darkened, the examiner switches to infrared, and the added "9" appears as a black mark in a field where the surrounding ink has gone invisible — there is little for the juror to take on faith. The physical fact that makes it work is that two inks that look identical in visible light can behave differently under infrared — one absorbs IR and stays dark while the other becomes transparent — so a digit added later in a different ink separates visibly from the original figure. The conclusion rests on the physics of how dyes interact with light, not on examiner intuition, which is why it sits at the strong end of the validity spectrum.

Exercise 18.21

The sentence "shows features consistent with the known writer and cannot be eliminated as the author" is appropriately honest in three ways: (a) "consistent with" is a comparison statement that keeps its uncertainty, not an individualization claim; (b) "cannot be eliminated" correctly frames the result as a failure to exclude, the method's surest register, rather than a positive proof of authorship; (c) it implicitly leaves room for other writers who might also be consistent, refusing the "to the exclusion of all others" overreach. It states the finding at a strength the method can bear.

Exercise 18.22 †

Problems with "Handwriting analysis proves the defendant, and no one else, wrote this note": 1. Individualization the method has not earned. "And no one else" is an individualization claim; handwriting comparison has no quantified, demonstrated error rate for naming one author to the exclusion of the world (§18.3). The method can exclude, flag simulation, and find "consistent with," but not this. 2. "Proves." Reserve "proves" for what is quantified (good single-source DNA). Handwriting authorship comparison rests on examiner judgment about whether differences fall within natural variation, with no measuring stick. Honest rewrite: "The questioned writing shows features consistent with the defendant's known writing and cannot be eliminated as having been written by the defendant; I cannot, from the handwriting alone, identify the defendant as the writer to the exclusion of all others."

Exercise 18.23

The pseudoscience smuggled in is graphology — the claim that handwriting reveals personality, honesty, or criminal propensity. It has no place in a courtroom because it has no validated basis: there is no demonstrated, reliable link between handwriting features and character or behavior, so testimony that a writer was "anxious, controlling, and prone to violence" from the writing alone is fortune-telling dressed as expertise. It differs from legitimate forensic document examination, which makes comparative and physical claims about a document (is this writing consistent with that exemplar; was this figure altered; was this signature drawn) — claims grounded in observable features and, for the physical half, in chemistry and physics — and which never purports to read character from script.

Exercise 18.25

Two things to be skeptical of in "The ink-dating expert proved the will was back-dated, so it's a fake": 1. The word "proved" applied to chemical ink-aging. Precise ink-aging years after the fact is destructive, controversial, and constrained (§18.4); naming a date with certainty overstates a contested method. A first-appearance boundary can impeach a claimed date, but a precise "added last year" is far weaker. 2. The leap from "the signature is recent" to "the will is a fake" (a prosecutor's-fallacy–style jump, Chapter 9). Even a sound finding that a line was added recently does not, by itself, establish fraud, who did it, or that the whole document is fake — there are alternative explanations, and "the document is a forgery" is a story the dating result has not, by itself, earned.

Exercise 18.27 †

Bias risk: Receiving the letter together with exemplars labeled "suspect" and a note that the suspect "has a clear motive" floods the examiner with domain-irrelevant context (Chapter 31). Knowing whose writing the "known" sample is and what conclusion is hoped for, the examiner goes looking through the elastic natural variation for agreement — and natural variation rewards the looking, so ambiguous features get read toward the wanted answer (confirmation/contextual bias). Safeguard: the examiner should compare the questioned writing against exemplars without knowing which belongs to the police's suspect, and should commit to documented observations before learning the desired answer (context management / sequential unmasking). A blind comparison is worth MORE than one done knowing the suspect — even when the two agree — because only the blind result is uncontaminated by the expectation; agreement that might be bias-driven cannot be distinguished from genuine independent corroboration.

Exercise 18.29

You should decline to testify that handwriting comparison "individualizes" a writer the way DNA individualizes a person because the two methods sit far apart on the validity spectrum: single-source DNA rests on a quantified random match probability with a known error structure, while handwriting authorship comparison has no objective standard for "how much agreement is enough" and no demonstrated, low error rate for individualization (§18.3). What you can honestly say instead: that the questioned writing is consistent with (or cannot be eliminated from, or shows indications of simulation relative to) the known writer — comparison statements that keep their uncertainty — or a frank "inconclusive." You can also testify to exclusion when a clear, reproducible difference falls outside any plausible natural variation, since exclusion is the method's surest voice.

Exercise 18.31

Not a scandal — consistent with how the method works. Handwriting authorship comparison rests on examiner judgment about whether features fall within an elastic natural variation, with no calibrated threshold for a conclusion. Two qualified examiners can weigh the same ambiguous features differently and reach different cautious conclusions in good faith — especially when both correctly stop short of individualization and report "consistent with" or "inconclusive." That two careful examiners differ while both staying within the method's defensible register reflects the method's genuine, acknowledged subjectivity — not fraud. (Contrast a clean single-source DNA profile, whose tight error structure should not produce divergent results between competent analysts.)

Exercise 18.32 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → contested): 1. Dating an ink alteration by infrared imaging — STRONG. Grounded in the physics of differential ink absorption/transmission under IR; non-destructive; the result is demonstrable (a jury can see the added digit appear and vanish). Behaves like analytical science. 2. Recovering indented writing with the ESDA — STRONG. Recovers content physically present in the page via surface electrostatics; no subjective same-hand judgment; demonstrable and non-destructive. 3. Stroke sequence at a line crossing — MIXED / BOUNDED. Sometimes reliably determinable by microscopy, often honestly indeterminate; defensible only when the junction clearly shows priority. 4. Concluding a specific person wrote a ransom note — CONTESTED MIDDLE. A subjective authorship comparison with no quantified error rate and no objective sufficiency standard; can exclude and flag simulation, but cannot individualize at the claimed strength. Why so far apart in one discipline? Because the tasks rest on different foundations: the physical/instrumental tasks rest on chemistry and physics with demonstrable outputs, while authorship comparison rests on examiner judgment about elastic natural variation. The discipline wears two coats, and the validity follows the foundation, not the job title.

Exercise 18.33

The split mirrors odontology (Chapter 17) exactly. In odontology, dental identification of the dead — matching antemortem and postmortem dental records — is a valid function resting on durable, documented dental history, while bite-mark comparison is the discredited function with no validated basis. In questioned documents, the physical-evidence half (ink chemistry, alterations, indented-writing recovery) is the strong, demonstrable function, while handwriting authorship comparison is the contested function. In both specialties, the lesson is the same: a single discipline can hold a rigorously defensible function and a weak-or-discredited one side by side, so you must locate the specific task on the spectrum, never trust the field's name as a whole.

Exercise 18.35 †

Cold-case evidence-log entry (model): - (a) Defensible inference: an insurance/partnership document was altered — most notably a beneficiary designation that does not sit naturally on the line — and indented writing from a vanished sheet was recovered from a surviving page. - (b) Honest verb: indicated (document tampering indicated; a possible altered beneficiary) — not proven. - (c) Does NOT establish: who made the alteration; when it was made; that the alteration is even connected to the death (it could reflect an ordinary business dispute); and, from a burned document, anything the fire has destroyed. (Any three suffice.) - (d) Why no name yet: the document evidence shows that a designation was likely changed and supplies a documentary hint of motive, but handwriting comparison cannot individualize the altering hand to certainty (§18.2–18.3), and a charred document gives the examiner less to work with, not more. Naming who altered it now would claim more than the evidence has earned — the chapter's and the book's core discipline. We are still narrowing, not concluding.

Exercise 18.37

Two evidence types that, with an altered beneficiary, would begin to connect a person to that motive: e.g., the financial records and insurance policies naming a specific beneficiary, plus that person's debts (Chapter 27, previewed — forensic accounting establishes who benefits and the motive's strength), and digital/cell-site evidence placing a particular person near the cabin against an alibi (Chapter 25, previewed). A touch-DNA mixture consistent with a contributor (Chapters 7–9) is a third. The altered document alone cannot make the leap because it shows that a beneficiary was likely changed and that someone stood to gain — but not whose hand altered it, when, or that the change is tied to the death; person-level and timing links must come from other evidence converging on the same individual.

Exercise 18.39 †

ESDA vs. handwriting-authorship comparison — validity foundations (model answer). Indented-writing recovery with the ESDA rests on surface physics — the differential electrostatic charge held by paper that pen pressure has compacted — and it recovers content that is physically present in the page; its output is a demonstrable, non-destructive image of impressed writing, with no subjective same-hand judgment required. It is strong as honest, bounded science, and is most easily overstated only if someone confuses recovering what was written with proving who wrote it, or treats the recovery as dating the impression (it cannot). Handwriting-authorship comparison rests on examiner judgment about whether features fall within an elastic natural variation, with no objective standard for sufficiency and no demonstrated low error rate for individualization; it is genuinely able to exclude and to flag simulation, but is most easily overstated when an examiner converts "consistent with" into "wrote it to the exclusion of all others." In the chapter's image, one discipline wears two coats: the ESDA coat is woven from physics and recovers facts; the authorship coat is woven from judgment and must keep its uncertainty.

Chapter 19 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 19.1

Trace evidence is physical evidence found in small or microscopic quantities, transferred between people, objects, and environments by contact (hairs, fibers, glass, paint, soil, residues). It is defined by quantity and origin rather than by chemistry because the forensically decisive facts about it are physical — how readily it transfers and how long it persists — not what substance it happens to be. A hair, a fiber, and a paint chip are utterly different materials but are all "trace evidence" because each is a tiny, contact-deposited residue the analyst must find, recover, and interpret.

Exercise 19.3 †

Transfer is the movement of material from one surface to another on contact; persistence is how long that transferred material remains before it is shed, washed, or worn away. They are different axes: a fiber can transfer easily but vanish in an hour, or transfer poorly but cling for days. The two together govern the meaning of a find: a readily-transferring, long-persisting fiber recovered from a suspect is weak evidence of recent, direct contact (it could have arrived many ways and stayed), whereas a poorly-transferring fiber that is normally gone within an hour, if found, points more strongly toward recent, substantial contact. You cannot reason from "a trace is present" to "this contact happened, then" without knowing the transfer and persistence behavior of that specific material.

Exercise 19.5

Hair morphology is the study of the physical structure of hair under a microscope. The three layers of the shaft are: the cuticle (the outer layer of overlapping scales, like roof shingles); the cortex (the bulk of the shaft, carrying pigment granules, air spaces, and structural detail); and the medulla (the central core, which may be continuous, interrupted, fragmented, or absent).

Exercise 19.7

A fiber is the smallest unit of a textile — a thread-like structure that combines into yarns and fabrics. A natural fiber comes from a plant or animal (e.g., cotton from a plant; wool from sheep). A manufactured/synthetic fiber is made from polymers (e.g., nylon or polyester). The distinction matters forensically because synthetics carry precisely measurable, engineered characteristics (polymer type, cross-section, dye chemistry), making them the workhorses of fiber evidence, whereas a very common natural fiber like cotton is nearly meaningless as an association.

Exercise 19.9

In rough order of preference: picking (forceps, for a visible hair/fiber, photographed and noted in place first); taping (a low-tack adhesive lift dabbed systematically across a garment, then mounted on acetate); scraping (debris worked loose over clean paper); and vacuuming (a filtered trace vacuum for large or inaccessible surfaces). The trade-off every method makes is thoroughness against selectivity: picking is the most selective but recovers only what you can already see; vacuuming is the most thorough but indiscriminate, sweeping up irrelevant background material along with the relevant. The recovery decision is made before anyone knows what was collected, which is exactly why contamination discipline matters.

Exercise 19.11

Every descriptor is a class characteristic. Nothing in the list is individual. It follows that a "match" of these features can support only non-exclusion ("cannot be excluded, along with a large number of others"), never identification — there is no individual feature, and no validated frequency, to elevate the claim.

Exercise 19.13 (odd)

Microscopy's honest modern job is to route hairs to DNA because the visual exam cannot identify a person — it can only describe class characteristics with no frequency behind them — whereas DNA (Chapters 7–8) can actually speak to source. The three things the examiner determines that decide which DNA test (if any) a hair can receive: (1) is it human? (no point testing animal hair); (2) does it have a root with adhering tissue? (good for nuclear STR typing, Chapter 7) or is it rootless? (mtDNA only, Chapter 8); and (3) which of many hairs at a scene are worth the cost of testing. Microscopy thus triages; DNA renders the source verdict.

Exercise 19.15

Two blue acrylic fibers that look identical to the eye may carry different dyes that absorb light differently. The bench step is microspectrophotometry, which records each fiber's color as a full absorption spectrum rather than as a subjective adjective. If the two spectra differ, the fibers are excluded as having a common source — full stop, no matter how alike they look. Exclusion here is "decisive" because it rests on an objective, reproducible measurement of a physical property; a single measured difference rules out a common source, whereas mere visual agreement only narrows.

Exercise 19.16 †

This association can be genuinely strong despite fibers being class evidence because the strength of a fiber association is not fixed — it scales with two things: (1) the rarity of the fibers and (2) the number of independent transfers. Here both are favorable: one fiber is an unusual custom color (rare → few sources could have shed it), and there are several different fiber types moving in both directions (a two-way exchange). The honest engine is coincidence becoming implausible: the probability that all of these independent transfers — multiple distinct fiber types, one rare, victim-to-suspect and suspect-to-victim — happened by innocent coincidence is small. No single fiber individualizes anyone; the weight comes from the multiplication of class-level associations, each of which only "cannot exclude," combining into something properly probabilistic.

Exercise 19.17

A single cotton fiber is nearly worthless because cotton is on essentially everyone and everything: an enormous number of sources could account for it, so "consistent with" carries almost no weight. A single custom-dyed trilobal nylon fiber from a particular carpet can carry real weight because few sources in the world could have shed it — the rarity is the point. This is "the value of the rare": the strength of a class-level association is inversely related to how common the trace is. The rarer the fiber, the smaller the chance the association arose by coincidence, and the more it constrains the field of possible sources — even though, like all fibers, it still does not individualize.

Exercise 19.19

Ordinary trace class evidence describes a type shared by many sources (a fiber consistent with a carpet, a hair consistent with a population); it can only narrow the field. A physical (jigsaw / fracture) match is different: a torn edge, a glass fragment that fits back into the broken pane, or two pieces of duct tape (Chapter 1) that reassemble along a unique fracture line. A fracture match can be near-individualizing because the irregular break is effectively unique — the two pieces were once one object, and the fit reconstructs that singular event. A hair "match" cannot do this because it shares only type features with countless other hairs; there is no unique reconstructable event, no fit, and no frequency. Example of class: the dark polyester fiber in the cold case. Example of fracture match: a glass fragment refitting a broken windowpane (§19.6).

Exercise 19.21 †

Figure 19.1 — what the exam shows: the hair is human, a head hair, naturally shed (club root), dark brown, medium diameter, with a fragmented medulla — and these features are shared with the reference hair within the range an examiner would call "consistent." What the exam does not show: whose hair it is; "dark brown, medium, fragmented medulla" describes a large fraction of the population; there is no database, no error rate, and no validated frequency, so "consistent" has no number behind it. The honest reading, in the correct verb: the hair cannot exclude the person of interest as a possible source — and that person is one of many it cannot exclude. This is "the ceiling of the visual exam" because the only stronger move available — assigning a probability or calling it a "match" — would require a frequency that does not exist; the next legitimate step is not a stronger adjective but DNA.

Exercise 19.23 (odd)

Honest sentence: "The questioned and known fibers are indistinguishable in polymer type, cross-section, diameter, and color; this is a common fiber type, it is class evidence, and I cannot exclude the jacket as a source — along with a large number of other dark polyester garments." Overstated sentence: "These fibers prove the defendant's jacket was pressed against the victim." The honest version attaches the limit (common type, class evidence, cannot-exclude-among-many) in the same breath; the overstatement converts a weak, common-fiber non-exclusion into a claim of unique contact the evidence cannot bear.

Exercise 19.25 (odd)

"A microscopic match" should be translated to the strongest claim the visual science supports: "the questioned hair cannot be excluded as having the same source as the known hair, sharing common class characteristics, with no statistical weight attaching to that association" — or, more bluntly, "I found no significant differences; this person, and many others, cannot be excluded." The word "match" is not available, because it implies an identification the method cannot deliver. The instrument-specific bias (§19.5) that made the "match" feel objective is the comparison microscope's design to present two specimens side by side, turning a vague "looks similar" into a vivid, immediate impression of "these go together" — an impression that feels like objectivity but, absent measurement and a frequency, is not.

Exercise 19.27 †

Overstatement: "Microscopic analysis places the defendant's hair on the victim's body" claims an individualization — that this specific hair is the defendant's and only the defendant's — which hair comparison cannot support. At most the exam can say the defendant cannot be excluded as one of many possible sources of a hair on the body. Defensible version: "A human head hair recovered from the victim shares common class characteristics with the defendant's reference hair; the defendant cannot be excluded as a possible source, along with an unknown but large number of other people, and no statistical weight attaches to the association." The earlier chapter's term for the illegitimate leap from "this type of hair" to "this person's hair" is individualization (Chapter 1, §1.4) — the myth of "a match."

Exercise 19.29 (odd)

The statement contains two overclaims. (1) "Hair evidence is worthless." The chapter's first honest caution: the lesson is not that hair evidence is worthless, but that visual hair comparison cannot identify a person — modern hair work, kept within its limits and paired with DNA, is genuinely useful (human/animal sorting, DNA triage, exclusion). The discredited thing is the overstated comparison, not hair evidence as such. (2) "Every defendant convicted with hair testimony is innocent." The second honest caution: the 2015 review faulted the testimony, not necessarily the verdicts — in many cases other, independent evidence also pointed to guilt, and an overstated hair conclusion does not by itself mean the defendant was innocent. The wrong is the overstatement, regardless of actual guilt.

Exercise 19.31 †

The prestige of the lab makes the scandal more instructive, not less, because it forecloses the comforting explanation that the errors came from incompetent analysts at a disreputable lab. These were examiners at the most prestigious forensic laboratory in the country, and the overstatement was nonetheless systemic — a whole discipline drifting, in the same direction, toward the conclusion the institution wanted. That is precisely the signature of cognitive bias (Chapter 31): not a few bad actors, but a structural drift. The hair scandal is "what happens when a method with no error rate meets a system with a strong preferred answer." The missing safeguard that would have caught the drift is blind verification with a measured error rate — context management / sequential unmasking, so the examiner does not know the wanted answer, and a second examiner verifies while also blind. Without those, there was nothing to catch the drift.

Exercise 19.33 (odd)

You should decline because a hair, by visual examination, shares only class characteristics with countless other hairs, and there is no validated frequency for any combination of them — so there is no basis, even in principle, for an individualization claim. DNA individualizes a person because it rests on a quantified, validated foundation with a known random match probability (Chapter 7); hair morphology has none of that. What you can honestly say instead: that the hair is human, a head hair (or whatever its class features are), that the person cannot be excluded as a possible source along with many others, that no statistical weight attaches, and that to tie the hair to a person would require DNA.

Exercise 19.35 (odd)

The ethical argument: forensic testimony is supposed to convey to the jury the actual strength of the evidence so they can weigh it correctly. Overstating weak evidence as strong deceives the fact-finder about how much the evidence should move them — it corrupts the deliberation regardless of the verdict's correctness. "He did it anyway" is not a defense because (a) the guilt was not established by the overstated evidence; the jury was led to convict partly on a claim the science could not support, so the process that produced the verdict was unsound even if the outcome happened to be right; and (b) a discipline that tolerates overstatement "when the defendant is guilty" has no way to confine the overstatement to guilty defendants — the same drift convicts the innocent (Tribble, Odom). The wrong is the misrepresentation, not the outcome (Theme 4: the CSI effect / over-trust cuts both ways).

Exercise 19.36 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited): 1. Single-source nuclear DNA — STRONG. Quantified, rigorously validated, with a known random match probability and error structure; the field's gold standard (Chapter 7). 2. Honest fiber comparison — class evidence done well. Anchored in objective, reproducible measurements (polymer, cross-section, color spectrum); exclusion is decisive; weight scales with rarity and the number of independent transfers. Below DNA because no fiber individualizes, but well above the discredited methods; PCAST treated it more gently than hair. 3. Microscopic hair comparison (for associating a hair with a person) — near the bottom. No validated frequency, no measured error rate; its strong forms were systematically overstated (2015 FBI review). Below contested toolmarks; above only fully discredited methods. 4. Bite-mark comparison — DISCREDITED. No validated basis for the claim that one set of teeth made a mark; multiple exonerations (Chapter 16). At the bottom. Why fiber and hair sit so far apart though "both are class evidence under the same microscope": fiber comparison rests on objective measurements (a difference in spectra excludes; rarity can be assessed), while visual hair comparison rests on subjective impression with no frequency — same instrument, but one discipline has a denominator and the other does not.

Exercise 19.37 (odd)

The comparison microscope is "the perfect emblem of the whole chapter" because the same instrument, pointed at fibers, supports a defensible, exclusion-anchored, class-level science, but pointed at hairs and operated by an examiner who knows the answer, it manufactured a generation of overstated testimony. The single sentence that captures the point (paraphrased): the tool was never the problem — the claims made with it were. The instrument is excellent for excluding and treacherous for confirming what you hoped to see; the failure lay in the language and the bias, not the optics.

Exercise 19.39 (odd)

The sentence — "a method with no measured error rate, operated inside a system that knows the answer it wants, will drift toward that answer and call the drift science" — is the common thread of three chapters. In Chapter 19 (hair), examiners with no error rate, inside an adversarial system with a preferred answer, drifted into systematic overstatement (the 2015 review). In Chapter 14 (Brandon Mayfield), the FBI's 100%-confident fingerprint "match" to the Madrid bombing was 100% wrong, driven partly by context and confirmation bias — the "gold standard" is still human judgment under bias. In Chapter 31 (the bias chapter), this becomes the explicit thesis: the chief threat to forensic accuracy is not bad equipment but the analyst who knows the wanted answer, and the fix is context management / blind verification / sequential unmasking. The hair scandal is one case; the principle is general.

Exercise 19.41 (odd) — Cold Case extension

Strengthening the fibers: a finding that the fibers were rarer than first thought (e.g., an unusual custom dye or cross-section, established against real reference data) and/or a two-way transfer — fibers from the victim's distinctive clothing found on Keller's jacket as well — would honestly increase the weight, because rarity and the number of independent transfers are exactly what scale a fiber association (§19.4). A set of diverse fibers transferred both ways would move the association from "weak/common" toward "properly probabilistic." Strengthening the hair: recovery of a rooted hair (with adhering tissue) attributable to the same source would allow nuclear STR typing (Chapter 7) rather than only mtDNA — and nuclear DNA, unlike mtDNA or visual comparison, can actually individualize. That single change (a root) is the difference between "cannot exclude, limited weight" and a potential identification.

Exercise 19.43 (odd) — Short writing (model)

A hair on a victim's body sounds like proof because it feels like a piece of the person who was there, shed in a struggle — and television has trained us to expect that a hair under a microscope yields a name. But by visual examination alone, a hair tells you a great deal about the hair and almost nothing about the person. An examiner can reliably say whether it is human, often what body region it came from, and something about its color, diameter, and treatment — but every one of those is a class characteristic, shared by an enormous number of people. There is no database and no validated frequency for hair features, so "consistent with" has no number behind it; the honest ceiling is "this person, along with many others, cannot be excluded." That is far weaker than proof. The one thing that can change it is DNA (Chapters 7–8): a rooted hair can be typed with nuclear STRs that genuinely individualize, and even a rootless hair can be tested for mitochondrial DNA, which can at least exclude. Without DNA, a hair is a thread to pull — not a name.

Exercise 19.44 † — Short writing (model)

Honest fiber comparison rests on objective, reproducible measurement: polymer type, cross-sectional shape, diameter, and — decisively — dye chemistry and color recorded as a full absorption spectrum. These can exclude cleanly (differing spectra rule out a common source) and characterize precisely, and the rarity of an unusual fiber can be assessed against reference data; its honest strength scales with rarity and the number of independent (ideally two-way) transfers. It is strongest when many diverse or rare fibers move both ways, and it states its class-level limit in the same breath as its strength. Microscopic hair comparison rests on subjective visual impression of class characteristics for which no population frequency exists: "consistent with" therefore has no denominator, cannot become a likelihood ratio, and cannot honestly exceed "cannot be excluded." Fiber comparison learned to state its conclusions at their true strength because it was anchored in numbers and exclusion was decisive; hair comparison, lacking any measured error rate or frequency and operating inside a system that wanted a "match," drifted for decades into the language of certainty — until DNA exonerations forced the 2015 reckoning. In short: fibers had a denominator and an exclusion discipline; hair had neither.

Chapter 20 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 20.1

Forensic toxicology is the application of the science of poisons (drugs, alcohol, and other chemicals) and their effects on the body to legal questions. Its three component activities are identification (what substance is present), quantitation (how much), and interpretation (what the concentration means for the legal question).

Exercise 20.2 †

The three questions, in order: 1. What is present? — identification (an analytical-chemistry question; the strong end of the discipline). 2. How much? — quantitation (a calibrated measurement with stated uncertainty; still strong). 3. So what? — interpretation (a human judgment: was the person impaired, incapacitated, poisoned, killed?). Certainty drains away because each step adds something the instrument cannot verify. Identification rests on the instrument and reference libraries (low, measurable error). Quantitation rests on calibration. But interpretation rests on tolerance, timing, route, drug combinations, and — after death — the corpse's own changing chemistry, none of which the instrument measures. The authority shifts from chemistry to pharmacology to inference, and the inference is the weakest because it depends on everything the first two leave out.

Exercise 20.3

A metabolite is a product the body makes when it chemically transforms a substance, usually on the way to eliminating it. Finding a metabolite in urine shows the parent drug was in the body and was processed, because the body had to take up the parent drug and act on it to produce the metabolite — so even after the parent drug itself has cleared the bloodstream and become undetectable, its metabolite remains as evidence that exposure occurred.

Exercise 20.5

An immunoassay is a screening test that uses antibodies engineered to bind a target drug or drug class, producing a measurable signal (color, fluorescence, absorbance change) when the target is present. Cross-reactivity is the binding of the antibody to other molecules of similar shape, which produces a false positive. Cross-reactivity makes the immunoassay presumptive only because the test reports that something class-shaped is present, not which specific compound — it cannot, by itself, distinguish the drug of interest from a structurally similar decongestant or other substance, so its positive is a lead requiring confirmation, not an identification.

Exercise 20.7

The dose makes the poison: no substance is poisonous or safe in itself; only a quantity is harmful or harmless (water in excess can kill; a lethal toxin in a small enough dose does nothing). The ranges: the therapeutic range is the band of concentrations producing the intended effect with acceptable safety (consistent with appropriate medical use); the toxic range is above therapeutic, where harmful effects (impairment, incapacitation, organ stress) appear without necessarily being fatal; the lethal range is the band associated with death.

Exercise 20.9

Vitreous humor is the clear gel inside the eyeball. It is the "protected archive" because (1) it is anatomically isolated — sealed in the eye, relatively shielded from the bacterial invasion and redistribution that disturb blood after death, so it decomposes more slowly and resists some postmortem artifacts; and (2) drugs reach it more slowly than they reach blood, so a vitreous level can preserve information about an earlier state. Its two forensic virtues are therefore isolation/durability and lag, which together make it especially valuable for the alcohol question (distinguishing antemortem drinking from postmortem production) and for some chemistries.

Exercise 20.11 †

A central (heart) blood concentration may overstate the antemortem circulating level because of postmortem redistribution: after death the barriers and gradients the living body maintained break down, and drugs that had concentrated in organs during life — especially the liver, lungs, and heart muscle, and especially lipophilic drugs with a large volume of distribution — leak back out into the adjacent central blood. The heart-blood concentration therefore rises above what was circulating at death. This is not an instrument error (the instrument measures the central-blood level accurately); it is that the central-blood level no longer represents the antemortem level. The single best alternative specimen is femoral (peripheral, leg-vein) blood, ideally tied off, because it is far less affected by redistribution from the central organs and is the most antemortem-representative blood available.

Exercise 20.13 †

The same measured BAC can map to different crash-time values because a single sample does not reveal which limb of the absorption–elimination curve the person was on at the time of the event. - Higher crash-time value: if the person was past peak and eliminating at the crash, then alcohol was eliminated between the crash and the draw; back-calculating (adding back the eliminated alcohol) gives a higher crash-time BAC than the measured value. - Lower crash-time value: if the person was still absorbing (on the rising limb) at the crash — e.g., they drank shortly before driving and had eaten little — then the BAC was still climbing toward its peak after the crash, so the crash-time value was lower than the later measurement; back-calculating as though they were eliminating would overstate it. To tell the two apart you need the drinking history: the time and quantity of the last drink(s), whether the person had eaten, and thus their likely absorption state at the moment of the event. Without it, the single sample cannot establish which limb it came from.

Exercise 20.15

A blood opioid concentration in the textbook "lethal range" does not, by itself, establish the opioid as the cause of death because (1) tolerance — a chronic, long-term user may function at, or survive, a concentration that is "lethal" for a naive person, so the range is a population guide, not a per-person verdict; and (2) the concentration may reflect postmortem redistribution (if drawn from central blood) rather than the antemortem circulating level — the death-time level may have been lower. (Either reason suffices; a third is that cause of death is an interpretation requiring the whole picture, not a number read off a table.)

Exercise 20.16 †

Reading each drug in isolation could miss the cause of death because of combined drug toxicity (synergy): several central-nervous-system depressants together — an opioid, a benzodiazepine, and alcohol, say — can produce a fatal additive or synergistic effect even when no single one is in its own lethal range. A toxicologist who checks each drug against its individual lethal threshold, finds none exceeded, and concludes "no toxicological cause of death" would miss a death caused by the combination. The interpretive concept is that the toxicologist must reason about the whole mixture of substances acting together in this person, not about each substance separately.

Exercise 20.17

Tolerance complicates interpretation because a chronic user's entire effect scale can be shifted to the right: a concentration that would be lethal for a naive person may be a routine, survivable maintenance level for them. So a "lethal-range" finding in a long-term user is ambiguous — it may not have been lethal for that individual. The additional, non-toxicological information that would help: evidence of chronic use — a prescription/medical record, a known history of use, physical findings such as needle tracks, and the scene context — which lets the toxicologist temper the textbook ranges against this person's likely tolerance.

Exercise 20.19

Before any adverse action, the presumptive immunoassay "amphetamines" flag must be confirmed by a different, more specific method (chromatography–mass spectrometry) that identifies the actual compound. The chemical reason the screen alone is unsafe to act on is cross-reactivity: the antibody can bind structurally similar molecules — here, an over-the-counter decongestant — producing a positive that has nothing to do with the drug of interest. The screen says "something class-shaped is present"; only confirmation can say "this specific substance, or not." Acting on the unconfirmed screen risks penalizing someone for a legal medication.

Exercise 20.21 †

(a) Honest sentence: "The measured blood alcohol concentration is reliable, but the value at the time of the crash depends on whether the driver was still absorbing or already eliminating; if past peak and eliminating, the crash-time level was likely somewhat higher than the measured value, whereas if still absorbing it may have been at or below it — so I state it as a range with those assumptions, not a single number." (b) Overstatement: "The driver's blood alcohol concentration at the time of the crash was [single back-calculated number]." The difference is that (a) exposes the dependence on the absorption–elimination state and the drinking history and reports a conditional range, while (b) launders the certainty of the measurement onto an inference that a single sample cannot support.

Exercise 20.23

The three kinds of statement (from the §20.1 "At the Bench" callout): - Instrument statement: "The drug was present at this level." (defensible from calibration and confirmation records). - Pharmacology statement: "That concentration is impairing." (rests on the published literature about the drug and on how well this person matches that population). - Inference statement: "Therefore he could not have walked to the next room." (the weakest — contingent on tolerance, the specific behavior, and everything the first two leave out). The weakest is the inference ("could not have walked"), because it depends on individual factors (tolerance, the specific motor task) that neither the instrument nor the general pharmacology establishes for this person. Blurring the three into one breath lends the certainty of the first to the third.

Exercise 20.25

A postmortem drug concentration with no sampling site stated should make a careful reader withhold an interpretation because the concentration's meaning depends on where the blood came from: a central (heart) sample may be artifactually elevated by postmortem redistribution, while a femoral (peripheral) sample is more antemortem-representative. Without the site, you cannot tell whether the number reflects the death-time level or a redistribution artifact. The single piece of information that would most change how the number is read is therefore the sampling site (central vs. peripheral).

Exercise 20.26 †

Problems with "Toxicology proves the defendant was high at the moment of the offense": 1. Conflates the three questions. Toxicology can identify and quantify a substance (strong), but "high at the moment of the offense" is an interpretation about impairment at a specific past time — the weakest, judgment-level claim, which the identification does not establish. 2. Present ≠ impairing, and the timing is unestablished. A confirmed presence (or even a concentration) does not equal impairment, because of tolerance and individual variation; and impairment at the moment of the offense requires knowing the concentration then, not at sampling, plus the person's response — none of which "toxicology proves." Honest rewrite: "A confirmed analysis identified [substance] at [concentration]; that level is consistent with impairment in the general population, but whether the defendant was impaired at the time of the offense depends on tolerance and timing I cannot fully characterize, so I can state it as consistent with impairment, not as proof."

Exercise 20.27

Before a roadside "positive for cocaine" is an identification, it must be confirmed by a laboratory using a different, more specific method (chromatography–mass spectrometry). The field test is a presumptive screen — sensitive by design and fooled by cross-reacting substances. Presumptive field tests have produced wrongful arrests because cheap color-change or single-strip tests have come up "positive" on substances that confirmatory analysis later showed were soap, candy, or over-the-counter medications; people have been arrested, jailed, and even induced to plead guilty on the strength of a screen that was simply wrong. The presence of a positive screen is not the presence of the drug — the confirmation is.

Exercise 20.29

Overstatement: that a urine-positive result "shows the driver was under the influence." Urine detects exposure over roughly the last few days (often via metabolites), not current effect; a positive can reflect use days earlier, long after any impairment has passed, and a metabolite is not the active parent drug. Defensible version: "The urine result indicates the driver was exposed to [drug] within roughly the last few days; it does not, by itself, establish that the drug was active in his bloodstream or that he was impaired at the time of the crash."

Exercise 20.31 †

Risk: being told "this is an obvious overdose — the family found him with pills" can bias the work in two ways (contextual/confirmation bias, Chapter 31). First, it can distort which tests are ordered: the toxicologist may run the panel that fits "overdose" and not look for other agents — and a substance no one thinks to look for is a substance that will not be found (§20.1). Second, it can distort how ambiguous results are read: a borderline concentration may be interpreted toward "fatal overdose," and inconvenient findings (a trauma, a competing cause) discounted. Safeguard: keep domain-irrelevant information away from the analyst until the analytical findings are fixed, run a sufficiently broad and case-appropriate screen rather than one narrowed by the assumed answer, and state explicitly which conclusions rest on the chemistry and which on the assumed narrative. An interpretation that quietly absorbed the "obvious overdose" framing is worth less than one that exposed its own assumptions.

Exercise 20.33

You should decline to let the certainty of the identification vouch for the impairment claim because they are statements of different kinds and different strengths (§20.1). The identification ("the drug was present at this concentration") is analytical chemistry — strong, documented, near the top of the validity spectrum. "Impaired at a specific past moment" is an interpretation that depends on the concentration at that moment (not at sampling), the person's tolerance, and their individual response — none of which the identification establishes. What you can honestly say instead: "I confirmed the substance at this concentration; that level is consistent with impairment in the general population, but I cannot state from the identification alone that this individual was impaired at that specific moment without information about tolerance and timing."

Exercise 20.35

The pattern — agreement on identification/quantitation, sharp disagreement on interpretation — is exactly what the §20.1/§20.6 framing predicts because detection is analytical chemistry (two competent analysts examining the same confirmed result will not meaningfully differ) while interpretation is a judgment built from tolerance, timing, sampling site, and combinations, where reasonable experts weighing those inputs differently reach different conclusions. A competent cross-examination should target the interpretive half: the assumptions about tolerance, the sampling site and redistribution, the timing, and the leap from "present at a level" to "impaired/incapacitated/killed" — because that is where the testimony is actually made and where reasonable people can differ. The identification is nearly unassailable; the interpretation is the battlefield.

Exercise 20.36 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited): 1. Single-source nuclear DNA — STRONG. Quantified, rigorously validated, known error structure; the gold standard (Chapter 7). 2. Confirmed GC-MS drug identification — STRONG (near DNA). Analytical chemistry, validated, low measurable error; the detection end of toxicology. This is the high-spectrum half of the discipline. 3. "Incapacitating for this person" interpretation — a JUDGMENT, well below the above. It rests on tolerance, timing, sampling site, and individual variation that no instrument measures; defensible as a bounded opinion ("strongly consistent with"), not as a measurement. This is the low-spectrum half within the same field. 4. Bite-mark comparison — DISCREDITED. No validated basis for the specific claim; multiple exonerations (Chapter 16); at the bottom. The justification turns on each claim's measured error rate and the rigor behind it — the PCAST question. Note that toxicology supplies BOTH item 2 (top) and item 3 (much lower): the same field occupies two positions, which is the chapter's thesis.

Exercise 20.37

Forensic toxicology occupies two positions because it does two different things. Its detection half — identifying a substance and measuring its concentration by confirmed instrumental analysis (e.g., "diazepam was confirmed by GC-MS in femoral blood at concentration N") — is near the top of the spectrum, grounded in validated analytical chemistry with low, measurable error. Its interpretation half — saying what that concentration means for impairment, incapacitation, or death (e.g., "that level was incapacitating for this person") — is a judgment that descends the spectrum, because it depends on tolerance, timing, drug combinations, and postmortem chemistry that no instrument measures. The discipline of this book is to keep the high-spectrum statement from vouching for the low-spectrum one: the strength of the confirmed identification must not be borrowed to make the interpretive opinion sound like a measurement. Each is presented at its true, different strength.

Exercise 20.39 †

Cold-case evidence-log entry (model): - (a) Defensible inference: at the time of the fatal events, Marcus Diallo was most consistent with being chemically incapacitated — sedated by a central-nervous-system depressant confirmed in peripheral blood at a toxic, incapacitating concentration — with a modest blood alcohol level also present (interpreted with postmortem cautions). - (b) Honest verb: strongly supports (for the incapacitation/state inference), built on a confirmed identification (top of the spectrum) plus a bounded interpretation of the concentration against the person and the sampling. - (c) Does NOT establish: who administered the sedative (it speaks to the victim's state, not any suspect); that the sedative was the cause of death (the cause was blunt-force trauma, Chapter 11 — the sedative is the means of overcoming him, not the killer); that homicide is independently proven by the toxicology (the autopsy proved that; toxicology corroborates); the exact time/route of administration. (Any three suffice.) - (d) Why "drugged before death" corroborates without proving homicide: the incapacitation finding explains how a fit 38-year-old was overcome without extensive defensive injury — he was sedated first — which is entirely consistent with the homicide the autopsy established (no soot in airways → dead before the fire; blunt-force skull fracture). But "consistent with / explains" is not "proves": the toxicology adds a coherent mechanism to the established homicide, it does not by itself demonstrate that a crime occurred (Chapter 11 did) or identify a perpetrator.

Exercise 20.41

Model answer (juror-facing). A confirmed drug identification is strong evidence because identifying a specific substance in the blood is analytical chemistry: the instrument compares the sample against known molecular fingerprints, with an error rate that can be measured and is low. So "this drug is present" is about as reliable as forensic science gets. But "present" is not "impaired." How a drug affects a person depends on how much they took, how long ago, their size and health, what else they took, and — crucially — their tolerance: a long-term user can function normally at a level that would incapacitate someone who had never taken it, and a level that is safe for one person can harm another. So a drug being in the blood tells you it was there, not that the person was impaired at a particular moment. This is exactly why the law often sets a per se alcohol limit (like 0.08): proving individual impairment from a number is hard, so the legislature simply makes the measurable number itself the offense — a tacit admission that "present at a level" and "impaired" are related but not the same thing.

Exercise 20.42 †

Model answer. Forensic toxicology has two halves that sit in very different places. Detection — identifying a substance and measuring its concentration by confirmed instrumental analysis (typically GC-MS after a presumptive screen) — rests on analytical chemistry: it is validated, quantified, and carries a low, measurable error rate, placing it near DNA at the top of the validity spectrum. Interpretation — saying what a concentration means for impairment, incapacitation, or cause of death — rests on pharmacology and judgment: it depends on tolerance, the timing between exposure and sampling, drug combinations, route, and (after death) postmortem redistribution and neoformation, none of which the instrument measures, so it sits well down the spectrum. Under cross-examination, an honest expert defends the detection robustly (calibration and confirmation records) but concedes the interpretive uncertainty: that "present" is not "impairing," that a postmortem central-blood level may not equal the antemortem level, that tolerance is unknown, and that a back-calculated value is an inference, not a measurement. The discipline is never to let the certainty of the identification vouch for the judgment of the interpretation.

Chapter 21 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 21.1

Forensic chemistry is the application of analytical chemistry to legal questions — chiefly identifying and characterizing substances whose identity matters to an investigation. The three families this chapter handles: controlled substances (drugs), explosives and their residues, and fire debris / accelerants.

Exercise 21.2 †

A presumptive color test is a fast chemical screen in which a reagent changes color with a class of substance, indicating the class is probably present (or, when negative, probably absent). A confirmatory test is a more specific, usually instrumental method that identifies the substance by interrogating the molecule itself. The single most important difference is what each may be reported as: a presumptive result may be reported only as "consistent with / indicated"; a confirmatory result may be reported as "identified as X" and may stand alone in court.

Exercise 21.3

A microcrystalline test adds a reagent to a trace of the suspected substance on a microscope slide, causing the target to form crystals of a characteristic shape and habit, read under magnification. It is still presumptive because (a) it depends on the analyst's subjective judgment in reading crystal morphology (a bias risk), and (b) crystal habit can be affected by impurities and cutting agents — it narrows the identification but does not deliver the molecular-level certainty of an instrument.

Exercise 21.5 †

Explosives residue is the trace chemical material left after an explosion or by handling an explosive — unconsumed explosive, characteristic breakdown products, and primer/propellant traces — recoverable in microscopic quantities. The absence of detected residue is not evidence that no explosive was used because the event itself consumes most of the explosive, and a clean scene, weathering, firefighting water, and the blast's efficiency can leave essentially nothing to find. You cannot reason from "we found no residue" to "there was no bomb" — a piece of evidence-strength honesty.

Exercise 21.7

Two families: low explosives (black/smokeless powder, pyrotechnics) — deflagrate; residues largely inorganic ions (nitrate, nitrite, chlorate, perchlorate) and metal cations. High explosives (TNT, RDX, PETN, many improvised organics) — detonate; the analytical target is organic molecules and their decomposition products. A single scene may carry either or both, which is why explosives work is rarely one test.

Exercise 21.9 †

A Marquis purple-black establishes: the powder is consistent with a class of opioids/related compounds — a screening lead, and a reason to run confirmatory chemistry. It does NOT establish: which specific compound (Marquis purple does not distinguish heroin from morphine from codeine); that the substance is in fact a controlled drug at all (unrelated compounds can give the same color); the quantity; or who handled it. Defensible report sentence: "Item tested presumptively positive (color test consistent with opioid-class substances); confirmatory instrumental analysis was requested." Anything beyond that, from a color test alone, overstates the chemistry.

Exercise 21.11

A new fentanyl analog can defeat an older opioid color screen because (1) the analog may give no color reaction or an unexpected one, so it is missed or misclassified by a test designed for an older compound; and (2) it typically appears in an adulterated mixture that the color test reads as a blur, unable to resolve a lethal component present at microgram potency. This makes instrumental confirmation more important because only a method that resolves and identifies each component (Chapter 23) can keep pace — exactly when the stakes (a potentially fatal component) are highest.

Exercise 21.12 †

The "chain of harm": field test "positive" → arrest/booking on a felony → (facing unaffordable bail and a distant trial) guilty plea before any lab confirmation → conviction entered → substance, if tested at all, analyzed only later → lab finds no controlled substance. The two points confirmatory testing would have prevented the wrongful conviction: (1) before charging/plea — confirming the substance would have shown there was no drug, so no felony charge to plead to; and (2) before the conviction is finalized — a mandatory confirmation step at any point pre-conviction would have caught the error while it was still reversible. The structural fix is "confirm before conviction."

Exercise 21.13

Fire-debris samples go into clean, unused metal cans or special nylon bags (never ordinary plastic) because common plastics can off-gas hydrocarbons that contaminate the sample or can let the volatile residue escape, either of which corrupts the analysis. The control sample of comparable unburned material guards against the background-pyrolysis confound: many ordinary building materials, carpet, and furnishings release hydrocarbons when they burn that can resemble parts of an ignitable liquid, and the control lets the lab distinguish a genuine accelerant from that background.

Exercise 21.15

A negative ion-chromatography result (no nitrate) addresses only the inorganic-ion / low-explosive world; it says essentially nothing about whether an organic high explosive (TNT, RDX, PETN, etc.) was present, because those are different molecules detected by different methods (GC-MS/HPLC, Raman/FTIR). The two-families distinction means a single negative on one method is not a negative for explosives generally — you must cover both worlds.

Exercise 21.16 †

The three control swabs (comparable steel from outside the blast zone, an unused swab from the lot, the collecting officer's gloves) carry as much weight as the positive because nitrate is environmentally common and a sensitive method finds it almost everywhere. The clean controls establish that the nitrate on the blast-zone swab is associated with the event, not with background or with contamination introduced during collection. The positive alone, with no controls, would be worth nearly nothing — an "artifact waiting to be overstated," a detection with no context to tell event from background.

Exercise 21.17

The report is honest in three ways: (a) it labels the color result "presumptively positive" — a screen, not an identification; (b) it specifies the confirmatory method ("by GC-MS") and reserves "identified" for it; (c) it keeps the two stages distinct and in order (screen → confirm), so the reader can see exactly what supports the identification. It reports each stage at its true strength.

Exercise 21.19

Still missing: an instrumental confirmation (GC-MS) that the volatiles are in fact gasoline, matched against the recognized chromatographic pattern — Chapter 23 supplies it. The chemistry alone cannot call the fire "arson" because an incendiary/arson finding requires fire-science analysis of origin and cause (Chapter 22) — multiple origins, fire dynamics, the ruling-out of accidental causes — not a single accelerant indication, and certainly not the discredited burn-pattern folklore (§21.5). "Accelerant indicated" is a chemistry status; "arson" is a fire-science conclusion.

Exercise 21.20 †

Problems with "The field test proves the substance was heroin": 1. A presumptive test does not "prove" / identify. A color test indicates consistency with a class; it cannot identify a specific compound, and innocent substances can trigger similar colors. "Proves" is precisely the overstatement the chapter forbids. 2. No confirmation / no orthogonal instrument. Identification requires a confirmatory method based on a different physical principle (GC-MS); a field test alone cannot bear the weight "proves" implies — and, as Case Study 21.1 shows, treating it that way has produced wrongful convictions. Honest rewrite: "The field test was consistent with an opioid-class substance and provided a basis to seize the item and request laboratory analysis; identification of the substance as heroin requires confirmatory instrumental analysis."

Exercise 21.21

Overstatement: converting a substance-and-place finding into a person finding, while ignoring contamination. Nitrate near a blast (a) is environmentally common and contamination-prone, so without clean controls it may not even be associated with the event, and (b) even when validly associated with the blast, says nothing about who handled the device. Defensible version: "Residue consistent with an oxidizer-based explosive was detected near the seat of the blast, against clean control samples; this concerns the substance and location and does not, by itself, identify who built or placed the device."

Exercise 21.23

Not simply correct. The roadside kit is not junk chemistry — a color test is a legitimate screen. It becomes a junk-science engine through misuse: a presumptive result treated as a confirmatory identification, with the confirmatory stage skipped. It is valid as triage and for developing probable cause (and as a reliable exclusion when negative); it is invalid as the sole basis for identifying a substance or convicting a person. The separating principle: a screen indicates a class and must be confirmed by an orthogonal instrument before it can be reported as an identification.

Exercise 21.25

"Confirm before conviction" is a scientific requirement because the presumptive/confirmatory pairing (§21.3) holds that a screen indicates a class and only an orthogonal instrument identifies the substance — so a conviction for possessing substance X logically requires that X was identified, not merely screened. Skipping confirmation does not save a step; it omits the step that turns "consistent with" into "is." The documented consequence (Case Study 21.1) is that innocent people are convicted on substances later found to be no drug at all. The cost saving is illusory against the cost of a wrongful conviction, and the requirement is rooted in the structure of the method, not in procedure.

Exercise 21.27

Not the same kind of disagreement. Two analysts arguing over a faint spot-test color are disagreeing about a subjective reading of a presumptive screen whose output is inherently class-level and condition-sensitive — which is exactly why the screen cannot identify and must be confirmed. Two DNA analysts examining a clean single-source profile should not sharply disagree, because the comparison yields a quantity with a tight, well-characterized error structure. The spot-test disagreement reveals the presumptive test's limits and argues directly for moving to an orthogonal instrument; it is the chapter's point in miniature.

Exercise 21.28 †

On the NAS 2009 / PCAST 2016 validity spectrum: 1. Confirmed GC-MS identification of cocaine — STRONG. Validated analytical chemistry, measurable error, instrument-checked; near DNA on the spectrum. 2. Roadside color test used as a screen directing lab work — VALID (as a screen). Legitimate triage and probable-cause tool; reliable exclusion when negative. 3. Roadside color test used as the sole identification — INVALID / overstatement. A class-level screen given a confirmatory weight it cannot bear; the failure mode that drove the Case Study 21.1 wrongful convictions. 4. Willingham-style burn-pattern "accelerant" claim with no chemistry — DISCREDITED folklore. A visual inference treated as chemical proof; fails testability and has no valid error rate. The same color test appears at two places (#2 and #3) because its validity depends entirely on its use: a sound screen, an invalid identification. Position on the spectrum is set by the claim being made, not the reagent.

Exercise 21.29

Both are the two-stage funnel: a fast, sensitive presumptive screen followed by a specific confirmatory step, with identification/source attribution reserved for the later stage. In Chapter 10 (serology), a Kastle-Meyer negative reliably excludes detectable blood. In this chapter, a drug color-test negative reliably indicates the screened-for class is absent (and a fire-debris field indication that is negative argues against an accelerant). In both, the presumptive negative is the more reliable result (Theme 1) and the presumptive positive says only "confirm."

Exercise 21.30

The chapter advances primarily Theme 2 (the validity spectrum) and Theme 4 (the CSI effect cutting both ways). Theme 2: it places confirmed instrumental chemistry near DNA while showing that presumptive tests are valid screens but invalid identifications — the same color test at two points by its use. Theme 4: the roadside field-test catastrophe (Case Study 21.1) is driven by the cultural faith that a "test" delivers a verdict, which lets a screen masquerade as proof. It also touches Theme 1 (the presumptive negative is the reliable result; absence of explosives residue is not absence of a bomb) and Theme 3 (bias in high-pressure explosives work and in an officer reading an ambiguous color).

Exercise 21.31 †

Cold-case evidence-log entry (model): - (a) Defensible status: accelerant indicated — consistent with gasoline, in the front-room fire debris, on the strength of a scene hydrocarbon odor and a handheld-detector indication (collection done correctly: clean cans, control sample, chain of custody). - (b) Honest verb: indicated (presumptive). NOT "confirmed," NOT "identified." - (c) Does NOT establish: that gasoline is confirmed (an instrumental claim, pending GC-MS, Chapter 23); that the fire is arson (a fire-science origin-and-cause conclusion, Chapter 22); who set the fire; or when the accelerant was placed. (Any three suffice.) - (d) Why not "gasoline confirmed" or "arson": "confirmed" is an instrumental word the chemistry has not yet earned (a smell and a sniffer are screens); "arson" is a fire-science conclusion requiring origin-and-cause analysis, not a single accelerant indication — and never the burn-pattern folklore that killed Willingham. Writing either now claims more than the evidence has earned, the book's core discipline.

Exercise 21.33

Two analyses that, with a confirmed accelerant, build toward an incendiary-fire conclusion: from this part, the GC-MS confirmation of gasoline (Chapter 23) turning "indicated" into "identified"; from an earlier chapter, the autopsy finding that the victim was dead before the fire (Chapter 11) plus the incapacitating sedative (Chapter 20), which together make an accidental fire implausible and a staged one likely. The accelerant indication alone is insufficient because (i) it is presumptive, not confirmed, and (ii) even a confirmed accelerant is not, by itself, "arson" — that requires fire-science origin-and-cause analysis (Chapter 22) ruling out accidental ignition.

Exercise 21.35 †

Instrumental vs. presumptive — validity foundations (model answer). Instrumental forensic chemistry (e.g., a confirmed GC-MS drug identification) rests on validated analytical chemistry: the instrument interrogates the molecule itself (its mass and fragmentation) by a physical principle independent of any color reaction, the method has a measurable error rate, and the result can be re-run and independently checked. It is strong because its core claim ("this substance is X") is testable and has been tested, which is why it sits near DNA on the validity spectrum; it is most easily overstated when an analyst reaches beyond the identified substance to claims about the person (cf. explosives residue), or when controls/documentation are weak. A presumptive color test rests on a class-level chemical reaction: a reagent reacting with a functional group common to a drug class, producing a color. It is strong as a screen — fast, cheap, sensitive, and a reliable exclusion when negative — and is most easily overstated when its class-level "consistent with" is read as a compound-level identification, especially at the roadside, where the confirmatory stage is skipped and the result is given a weight (conviction) it cannot bear. In short: the instrument's risk is overreaching from substance to person; the screen's risk is being mistaken for an identification.

Chapter 22 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 22.1

Fire investigation is the discipline of determining a fire's origin, cause, and development through systematic, scientific-method-based examination of the scene and physical evidence. The modern consensus standard that governs it is NFPA 921 (Guide for Fire and Explosion Investigations).

Exercise 22.3 †

The three legs of the fire triangle are fuel, an oxidizer (almost always atmospheric oxygen), and heat. Fire is a self-sustaining chemical reaction (combustion) that requires all three present at once; remove any one leg and the reaction stops and the fire goes out. (That is exactly how extinguishers work — by smothering the oxygen, cooling the heat, or cutting off the fuel.)

Exercise 22.5

The three modes of heat transfer: conduction (heat moving through a solid — e.g., a metal pipe carrying fire's heat into the next room); convection (heat carried by a moving fluid — in a fire, the buoyant column of hot gas rising, which is why fire climbs and the worst damage is often high above the fuel); and radiation (heat traveling as electromagnetic energy across space — e.g., feeling a bonfire's warmth on your face, and the mechanism that preheats nearby objects toward ignition). Convection and radiation do most of the work of spreading a room fire, and both push heat upward and outward.

Exercise 22.7

The four cause classifications are accidental, natural, incendiary (deliberately set — the fire-science term for arson), and undetermined. The chapter calls undetermined "the category the old folklore could never tolerate." It is often the only honest answer because a destroyed scene frequently does not allow the cause to be established to a reasonable degree of scientific certainty — and "undetermined" is the scientifically honest classification in that situation, not a failure.

Exercise 22.9

The heat release rate is how much energy a burning item puts out per unit time (kilowatts/megawatts). A single polyurethane-foam sofa or upholstered chair, once fully involved, can release heat at roughly one to two megawatts — comparable to a small gasoline fire — because modern furniture is essentially solid petroleum. This undermines "an unusually hot, fast fire proves an accelerant," because an intense fire is simply what a room full of synthetic furnishings normally produces. The science did not get more permissive; the fuel load of the average living room got far more flammable, and the old folklore never accounted for it.

Exercise 22.11 †

Crazed glass (a fine network of cracks in window glass) is caused by rapid cooling, not rapid heating — specifically, cold water from the fire hose striking hot glass. The thermal shock of fast cooling fractures the glass into the fine "crazed" network. Because firefighters spray water on hot glass at essentially every structure fire, crazing appears at almost every fire and is, if anything, an artifact of extinguishing the fire. It therefore says nothing about whether the fire was accelerated or deliberately set; the old claim that crazing showed an "unnaturally hot, fast" accelerant fire had the physics backward.

Exercise 22.13

The four-step self-confirming machine (§22.4): (1) an accidental fire reaches flashover, producing low burns and "pour-shaped" floor patterns; (2) the investigator, trained on the old indicators, reads those flashover artifacts as affirmative evidence of a poured accelerant; (3) reinforced by that "evidence," he inspects for accidental causes, finds none he can identify (the fire may have destroyed it), and invokes negative corpus — no accidental cause found, therefore arson; (4) he testifies sincerely that the physical evidence proves a deliberately set fire. Every step feels like diligence; not one is valid; and the honest classification, "undetermined," is never considered. No dishonest actor is required for this chain to convict an innocent person.

Exercise 22.14 †

The scientifically honest classification is undetermined. "Incendiary" is the wrong call because the only basis for it would be negative corpus — inferring arson purely from the failure to find an accidental cause — which is a logical fallacy, not a determination. The absence of an identified accidental cause does not establish a deliberate one, for two reasons: the accidental cause may have been destroyed by the fire, and the proper classification when the cause cannot be affirmatively established is undetermined. With no confirmed ignitable liquid and no multiple origins, there is no affirmative evidence of a set fire, so an incendiary finding is unsupported.

Exercise 22.15

Ventilation — the air supply — controls a fire as powerfully as fuel does: a fire starved of oxygen smolders, while one fed by an open door or broken window roars and burns most intensely where the air comes in. So burn patterns often track the path of ventilation rather than the point where a fire was "set," and an investigator who reads the most severe burning as "the origin" or "where accelerant was poured" may simply be reading where the air entered. Correctly accounting for ventilation is part of why modern origin determination requires fire dynamics, not pattern-matching.

Exercise 22.17

A negative ILR result does not prove no accelerant was used. A fire that burns hot enough and long enough can consume an ignitable liquid completely, leaving no detectable residue; a negative laboratory result therefore proves only that none was detected, not that none was present. This is the exclusion-versus-proof asymmetry from Chapter 1 (§1.6): ILR is strong evidence when present (it confirms an ignitable liquid) and weak evidence when absent (absence is consistent with both "no accelerant" and "accelerant fully consumed"). Absence does not exclude an accelerant.

Exercise 22.19 †

From Figure 22.1: the irregular, low, pool-shaped charring does show that this part of the floor burned intensely — a real, severe, low-level fire pattern. It does not show that a liquid was poured: controlled post-flashover burns produce indistinguishable patterns with no accelerant, and the eye cannot tell a gasoline pour from a flashover burn from a melted-and-burning-plastic pool. The single honest sentence: "This is low-level floor damage consistent with a fully developed, post-flashover fire; the presence of an ignitable liquid can be neither confirmed nor excluded without laboratory testing."

Exercise 22.21

Report A ("cause undetermined; the scene was too damaged to establish origin or cause to a reasonable degree of scientific certainty") is the more scientifically honest report. Report B ("incendiary; no accidental cause was found") is committing the negative corpus fallacy — inferring arson purely from the failure to find an accidental cause, with no affirmative evidence of a deliberately set fire. Report A correctly recognizes that a destroyed scene may not permit a cause determination and uses the honest classification; Report B treats its own inability to find something as proof of arson.

Exercise 22.23

A comparison (control) sample of unburned material from away from the suspected origin is collected so the lab can distinguish a genuine foreign ignitable liquid from the ordinary petroleum-derived background that modern synthetic materials emit when heated. Carpet, adhesives, and synthetic padding are made from petroleum; when heated they release hydrocarbons that a careless analysis (or an untrained nose) could mistake for "accelerant." The comparison sample is the control that guards against that false-positive error — it establishes what the normal material background looks like, so a true accelerant can be distinguished from it.

Exercise 22.25

The chain in "the crazed glass proves the fire burned unnaturally hot, which proves an accelerant, which proves arson" breaks at every link: (1) crazed glass is caused by rapid cooling (the fire hose), not by an unnaturally hot fire, so the first inference is simply false; (2) even a genuinely hot, fast fire does not prove an accelerant, because modern furnishings have enormous heat release rates and burn intensely on their own; (3) the presence of an accelerant — even if there were one — is not proof of arson without context (ignitable liquids have innocent reasons to be present). The slide cannot be rescued as written; the honest statement is that crazed glass indicates the fire was extinguished with water and supports no conclusion about cause.

Exercise 22.27

The chemist's statement — "the chemical pattern is consistent with gasoline" — is the validated analytical-chemistry claim, resting on GC-MS and reference patterns (near the top of the validity spectrum). The investigator's statement — "the gasoline was poured deliberately in this pattern to set the fire" — is the contestable scene interpretation, which is exactly where the old folklore creeps back in. Letting the first lend its strength to the second is a classic overstatement because they are two different claims with two different strengths: "it's gasoline" may be unassailable while "it was deliberately poured to set the fire" is a far weaker inference about distribution and intent that scene interpretation cannot establish to the same degree.

Exercise 22.28 †

Two reasons the television "instant arson" pronouncement is backward: (1) Cause depends on origin, and origin requires systematic examination, not a glance. A real determination forms hypotheses about where and how the fire started and tests them against all the data; you cannot read cause off a smoking floor in a minute. (2) The decisive evidence (a confirmed accelerant) is a laboratory finding, not a visual one. The scene investigator can only suspect and collect; confirming an ignitable liquid requires GC-MS in the lab (§22.5, Chapter 23). The television version skips the hypothesis-testing and the laboratory confirmation — the two things that make a determination valid — and substitutes the investigator's confidence, which is precisely the failure mode that convicted Willingham.

Exercise 22.29 †

Risk: an investigator told the homeowner "had money troubles and seemed off" arrives already suspecting arson, then looks for indicators — and because the indicators (crazed glass, low burns) are present at almost every fire, he finds them. Their ubiquity, which is exactly why they are worthless, makes them perfect confirmation fodder: whatever you suspect, the scene seems to confirm it, so the expectation selects the "evidence" and the "evidence" confirms the expectation. Safeguard: form the origin-and-cause hypothesis from the physics, test it against all the data (including data that would refute it), and keep the domain-irrelevant case context — the money troubles, the "feeling," even the dead occupant — out of the technical determination until it is fixed (context management / sequential unmasking, named formally in Chapter 31). The indicators are perfect confirmation fodder precisely because they are nearly universal: a method that finds its "signature" at every fire cannot distinguish the fire you suspect from any other.

Exercise 22.31

The same maturing fire science produced opposite endings because of the science-versus-law asymmetry (Chapter 5, §5.1): science is iterative and self-correcting, but the law's most severe punishment is irreversible. Han Tak Lee was still alive and imprisoned when the fire science was recognized as discredited, so the legal system — however slowly — could reopen his case and release him. Cameron Todd Willingham had already been executed when the same recognition matured; an execution cannot be undone. The variable was not the quality of the science used to convict (equally discredited in both) but whether the defendant was still alive when it was corrected. The lesson: pairing an irreversible penalty with a forensic method that may later be invalidated means the law's slowness to admit error can become lethal — the strongest argument in this book against the death penalty resting on unvalidated forensic science.

Exercise 22.33 †

On the NAS 2009 / PCAST 2016 validity spectrum, fire investigation spans the spectrum: - (a) The debunked-indicator method — DISCREDITED. Crazed glass, alligatoring, and "pour patterns" read by eye are untested visual rules with no measured error rate, refuted by controlled fire research (they are produced by ordinary post-flashover fires). This sits at the bottom of the spectrum, in the company of bite-mark comparison (Chapter 16), which likewise has no validated basis for its core claim and produced multiple exonerations. - (b) Modern fire-dynamics-plus-confirmed-ILR origin-and-cause work — SOUND/STRONG. Grounded in tested compartment-fire physics (flashover, heat release rate, ventilation), standardized by NFPA 921, and anchored by instrumental confirmation of ignitable liquids by GC-MS — validated analytical chemistry near the DNA/toxicology tier. The feature that separates them is whether the method's claims have been tested against ground truth and carry a measured error rate. The indicators were never validated and failed when finally tested; the modern, instrument-confirmed work is tested and reproducible. The discipline earned its higher standing by getting tested, not by getting more confident.

Exercise 22.35

The three findings cohere into a staged-homicide picture. No soot in the airways (Chapter 11) establishes Diallo was dead before the fire — he did not breathe smoke, so the fire did not kill him. The sedative at an incapacitating level (Chapter 20) explains how he could be subdued before death. This chapter's incendiary finding establishes that the fire that burned his already-dead body was deliberately set — consistent with an effort to destroy evidence and stage the scene as an accident. What the fire evidence specifically adds: that the fire was not accidental but a deliberate act, reinforcing "staging." What it specifically does not add: it does not identify who set the fire (arson is a determination about the fire's cause, not a person), and it does not, by itself, establish the sequence the other findings provide — it interlocks with them rather than proving them.

Exercise 22.36 †

Cold-case evidence-log entry (model): - (a) Cause classification and affirmative grounds: cause — incendiary (arson). Grounds: multiple, separate, unconnected points of origin (no single accidental fire could produce them) plus an ignitable-liquid (gasoline) distribution pattern (presumptively flagged by the chemistry bench, Chapter 21; GC-MS confirmation pending, Chapter 23). These are affirmative evidence of a deliberately set fire. - (b) Debunked indicators NOT relied upon, and why it matters: the finding does not rest on crazed glass, alligatoring, "pour patterns" read by eye, or negative corpus. This matters because each of those is the exact folklore that executed Willingham and imprisoned Lee; relying on them would make the cold case commit the error it critiques. The finding is honest because of how it was reached. - (c) Does NOT establish (any three): who set the fire; when (precisely) it was set; that the gasoline had no innocent origin until distribution/context is weighed; that the fire individualizes any person; or that the fire killed Diallo (the autopsy already showed it did not). - (d) Why no name attaches: arson is a finding about the fire's cause, not about a person. No name attaches to the accelerant; linking a person to the scene requires other evidence (soil — Chapter 24; digital/cell-site — Chapter 25; DNA — Chapters 7–9). To name an arsonist from the fire alone would claim more than the evidence has earned — the book's core discipline.

Exercise 22.37

The cold-case finding is deliberately contrasted with Willingham so it does not repeat the error it critiques. The valid version (cold case): arson rests on affirmative, testable, instrument-backed evidence — multiple unconnected origins (a structural fact no single accidental fire produces) and a gasoline pattern that the laboratory will confirm by GC-MS — plus an affirmative treatment of accidental causes (electrical ruled out by engineering analysis, Chapter 36), never negative corpus. The wrong (Willingham-style) version would have looked like: "the crazed glass and low burn patterns show a poured accelerant, and since we found no accidental cause, the fire was arson" — i.e., debunked visual indicators plus negative corpus, with no instrumental confirmation. The methodological difference is the entire lesson: both conclude "incendiary," but one is reached by validated, affirmative methods and the other by discredited folklore. How a finding is reached, not merely what it concludes, is what makes it honest.

Exercise 22.39

You would decline to put "crazed glass and alligatoring were also present" in the report even though they are loosely "consistent with" arson, because they are not valid evidence of anything about cause — crazed glass is an artifact of suppression (water on hot glass) and alligatoring only shows that wood burned. Including them would (1) launder discredited folklore into a report that is otherwise built on valid grounds, contaminating it; (2) invite the exact overstatement and Daubert vulnerability that the debunked indicators carry; and (3) risk the report being used, now or on appeal, to suggest the arson finding rested partly on folklore. Declining protects the integrity of the valid finding (multiple origins + confirmed ILR) and keeps the report from committing, even decoratively, the error that executed Willingham. A finding is only as clean as its weakest stated basis.

Chapter 23 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 23.1

Chromatography is a technique that separates the components of a mixture by exploiting how differently each component distributes between a moving mobile phase and a stationary phase. The two phases whose interaction does the separating are the mobile phase (a carrier gas in GC, a liquid solvent in HPLC) and the stationary phase (the coating/packing in the column).

Exercise 23.2 †

The three properties that make a result confirmatory: 1. Specificity — it responds to the target in a way other substances do not reproduce (ideally by reading molecular structure, since structure is what distinguishes one compound from another). Matters because a non-specific test confuses look-alikes. 2. Two orthogonal dimensions of information that fail differently (GC-MS: retention time and fragmentation pattern). Matters because an innocent coincidence on two unrelated measurements is vanishingly unlikely, where a coincidence on one is not. 3. Comparison to a same-day reference standard, with a clean blank. Matters because a result means nothing in the abstract — the standard proves the unknown behaves like the known, and the blank proves the signal is not contamination.

Exercise 23.3

A mass spectrum is a plot of the relative abundance of the charged fragments a molecule breaks into when ionized, arranged by their mass-to-charge ratio (m/z). The horizontal axis (m/z) gives each fragment's mass; the bar heights give how abundant each fragment is (tallest = base peak, scaled to 100%). The overall pattern is structure-specific and identifies the molecule; the heaviest major peak (the molecular ion) gives molecular weight.

Exercise 23.5

FTIR is Fourier-transform infrared spectroscopy: it identifies a substance by measuring how it absorbs infrared light. The physical property it exploits is that each type of chemical bond vibrates at a characteristic frequency (like a tiny tuned spring) and therefore absorbs infrared light at characteristic frequencies, producing a structural fingerprint.

Exercise 23.7

Analytical microscopy is the use of magnifying optical (and electron) instruments to observe, characterize, and compare structural features of trace evidence invisible or ambiguous to the naked eye. An ordinary light microscope is limited to roughly 1,000× useful magnification, because you cannot resolve detail finer than the wavelength of the visible light you are using — to go finer (and to read elements), the lab switches from light to electrons (SEM).

Exercise 23.9 †

Retention time is a class characteristic (Chapter 1): different compounds can coincidentally share the same retention time, so a match on retention time alone could be a coincidence — it narrows the field but does not identify. What is missing is the second, structure-specific dimension: the mass spectrum. A confirmatory identification needs the retention-time match plus a matching full fragmentation pattern against a same-day standard (the "two orthogonal dimensions" of §23.1). Reporting "confirmed" on retention time alone overstates a class-level result as an identification.

Exercise 23.11

A large, heat-sensitive opioid metabolite that decomposes when vaporized cannot be analyzed by gas chromatography, because GC requires the sample to be flash-vaporized intact — a molecule that decomposes on heating is destroyed before it can be separated (or gives misleading breakdown products). The lab should use HPLC (high-performance liquid chromatography), whose liquid mobile phase separates compounds without vaporizing them, so heavy and fragile molecules survive the analysis. (Often coupled as LC-MS for confirmation.)

Exercise 23.12 †

Gasoline is hundreds of hydrocarbons, so no single peak is "gasoline" — and a single peak (e.g., toluene) appears in many products and contexts, making it weak, ambiguous evidence on its own. The analyst confirms an accelerant from the whole chromatographic pattern — the characteristic distribution and relative heights of the aromatic "signature ridge," compared against a reference gasoline standard following a published ignitable-liquid classification scheme. The pattern-based approach guards against false identification from coincidental single components: a lone matching peak could come from countless innocent sources, but the full, weathered-gasoline pattern is far more specific. (It is the chromatographic version of "many features together, against a standard," not "one feature.")

Exercise 23.13

The non-destructive instrument well suited to identifying a bulk solid is FTIR (also Raman): it identifies the material from its infrared absorption without consuming or altering it, so the tiny sample survives for re-testing. Its main limit on mixtures is that overlapping spectra of several components blur into a composite that can be hard to disentangle — the situation where chromatography's separate-first approach wins. (If only one test is possible and the sample is a single bulk solid, the non-destructive structural fingerprint is the prudent choice.)

Exercise 23.15

Weakest → strongest identification, with justification: 1. A UV-Vis spectrum — weakest: broad, low-detail spectra shared by many substances; good for quantification, poor for identity. 2. A GC retention time — stronger but still a class characteristic; several compounds can share one, so it is one line of evidence, not an identification. 3. A full GC-MS run (retention time + matching mass spectrum) against a same-day standard — strongest: two orthogonal dimensions that fail differently, against a known standard with a clean blank — confirmatory of the compound's identity.

Exercise 23.16 †

(a) The flat baseline means the column is delivering nothing but carrier gas — no compound is eluting at that moment, so the detector reads near zero. (b) A peak's position (retention time) tells you which compound it might be — a class characteristic, confirmed against a standard; its height/area tells you roughly how much of that compound is present. (c) The "signature ridge" of aromatic peaks, not any single peak, is what identifies gasoline because any one peak (e.g., toluene) is common and ambiguous, while the relative distribution and heights of the whole aromatic family form a far more specific pattern — matched against a reference gasoline standard via the ignitable-liquid classification scheme.

Exercise 23.17

The base peak is the most abundant fragment (the tallest bar, scaled to 100%) — the fragment the molecule most readily breaks into; it is a strong clue to the molecule's family/structure (e.g., the m/z 91 base peak signals an alkylbenzene). The molecular ion is the intact, charged molecule (often the heaviest major peak); its m/z equals the compound's molecular weight, a hard constraint on identity. Each contributes differently: the base peak (and the whole fragmentation pattern) gives the structure-specific fingerprint; the molecular ion gives the molecular weight. Together with the full pattern they confirm identity against a library/standard.

Exercise 23.19

Honest sentence: "Gasoline was identified in the fire debris by GC-MS — the chromatographic pattern matched a weathered-gasoline standard, the mass spectra of the diagnostic peaks matched the reference and the spectral library, retention times matched the standard, and the blank was clean." Overstated sentence: "The GC-MS result proves the defendant poured gasoline in the cabin to set the fire." The first states the confirmed identity of the accelerant; the second leaps from a chemical fact (gasoline is present) to who, intent, and act — none of which the chemistry establishes.

Exercise 23.20 †

Two ways a high library "match score" can be confidently wrong: 1. Matched to the wrong library / a look-alike — a high score against an incomplete or inappropriate library can flag a structurally similar but different compound; the number is high and the identification is wrong. 2. A dominant component hiding a mixture — in a mixture, one major component can drive a high match while minor (and forensically important) components are masked; the score reflects the dominant species, not the whole sample. (Also: a contaminated or degraded sample can score high against the contaminant.) What the analyst must do instead of reading the percentage aloud: interrogate the match — Is the entire spectrum explained? Is there an unexplained peak? Was the blank clean? Does the result make sense in context? — and confirm against a same-day standard. The score is a hypothesis to be tested, not a conclusion.

Exercise 23.21

Two overstatements: (1) About elemental analysis — claiming an elemental match (lead/barium/antimony) identifies a unique source; elemental composition is class evidence (Chapter 1), telling you what class of particle it is, not which gun, cartridge, or person produced it. (2) About transfer — claiming the particle's presence proves the wearer fired a gun; GSR transfers and contaminates easily and can reach a sleeve by routes unrelated to firing (handling, environment, secondary transfer). Defensible version: "A particle with a morphology and elemental composition characteristic of gunshot residue was found on the sleeve; this is consistent with, but does not establish, that the wearer discharged a firearm, and it cannot show how the particle was deposited."

Exercise 23.23

A confident identification reported without a blank, a reference standard, or a chain of custody should be distrusted because those controls are where confirmation is actually won (§23.1, §23.3): the blank rules out contamination as the source of the signal; the same-day standard establishes that the unknown behaves identically to the known; the chain of custody (Chapter 2) establishes that the sample analyzed is the sample seized. An impressive chromatogram or spectrum with none of these is "garbage in, garbage out" waiting to happen — the instrument may be faithfully analyzing the wrong, contaminated, or misattributed thing.

Exercise 23.24 †

A flawless GC-MS run can still give a confident, wrong answer because the instrument analyzes whatever it is given, faithfully. Concrete example: a fire-debris can contaminated by a dirty sampling tool, or two submissions swapped at intake, will produce a perfect chromatogram and spectrum of the wrong sample — confidently identifying a substance that was never in the original evidence. Whose responsibility is the input: everyone upstream of the instrument — the collector (clean packaging), the lab intake (no mix-ups), the analyst (clean blank, proper standard). This means the method's validity (GC-MS is excellent) is not the same as the result's reliability, which depends on the integrity of the sample and the surrounding practice. Validity lives in the method and the practice.

Exercise 23.25

Risk: told which drug the sample "should" be, the analyst is anchored — an ambiguous spectrum or a marginal library match gets read toward the wanted answer (contextual/confirmation bias, Chapter 31), and an inconvenient unexplained peak gets explained away. Safeguard: keep domain-irrelevant information (the suspect, the "expected" answer) away from the analyst until the identification is fixed, then compare to case facts (context management / sequential unmasking). A blind interpretation is worth MORE than one made knowing the wanted answer — even when they agree — because only the blind result is uncontaminated by the expectation; agreement that might be bias-driven cannot be distinguished from genuine independent confirmation.

Exercise 23.27

Not evidence that GC-MS is unreliable. The two analysts agree on the core method's output — the spectrum matches the reference. Their disagreement is about an interpretation at the margins (whether a minor unexplained peak signals a second component) — a normal, resolvable scientific question handled by re-examination, additional testing, or a more careful extraction. This is fundamentally different from disputing the core method (as one would for bite marks, whose foundation is unsound). Disagreement about a marginal interpretation within a sound, high-validity method is honest science; it does not move GC-MS down the validity spectrum.

Exercise 23.28 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited), with the claim being ranked stated explicitly: 1. GC-MS identification of a compound — STRONG / confirmatory. Claim ranked: "this substance is X." Analytical chemistry, quantified, two orthogonal dimensions, reproducible; the gold standard for compound identity. 2. SEM-EDX elemental identification of a particle — strong as to composition, but class evidence. Claim ranked: "this particle is made of these elements / is characteristic of class Y." Objective and re-checkable, high-validity as to what the particle is made of; but it does not individualize a unique source, so the source-attribution claim is only class-level. 3. Microscopic hair comparison — WEAK. Claim ranked: "this hair is consistent with this person." The FBI admitted decades of overstatement (Chapter 19); valuable mainly for exclusion and as a route to mtDNA. 4. Bite-mark comparison — DISCREDITED. Claim ranked: "these teeth made this mark." No validated basis; multiple exonerations (Chapter 16). At the bottom. Justification turns on each method's measured error rate and the rigor behind its core claim — the PCAST question — and on being careful to rank the right claim (composition vs. source).

Exercise 23.29

GC-MS confirmation of gasoline establishes what is present in the debris — gasoline, identified instrumentally. Chapter 22's fire-science finding (multiple origins + ignitable-liquid burn pattern) establishes that the fire was incendiary — that the burn behavior indicates a set fire, not an accident. Neither alone establishes a homicide: "gasoline is present" and even "the fire was set" do not, by themselves, prove a person was killed unlawfully. The homicide conclusion rests on Chapter 11 (no soot in the airways → dead before the fire, plus blunt-force trauma). The chemistry confirms the accelerant; the fire science confirms the fire was set; the pathology establishes the death preceded the fire — three disciplines, each establishing its own piece, converging (Chapter 39).

Exercise 23.31

Both statements are true at once because validity is a property of the method and the practice. Instrumental analysis is "the high-validity end" because the methods (chromatography, mass spectrometry, spectroscopy, elemental analysis) rest on analytical chemistry and physics, are quantified and reproducible, and (for GC-MS) genuinely confirm a compound's identity. "The instrument is the easy part" because the hard, failure-prone work is everything around the instrument — clean collection and packaging, no sample mix-ups, a clean blank, the right standard, calibrated equipment, careful interpretation, honest testimony, a second reviewer. A perfect method on a compromised sample gives a confident wrong answer (Case Study 23.2). Holding both together — trusting the method and policing the practice — is the chapter's central discipline: high validity is a ceiling on reliability, not a guarantee of it.

Exercise 23.32 †

Cold-case evidence-log entry for the GC-MS fire-debris result (model): - (a) Defensible conclusion: gasoline is present in the fire debris from the cabin's front room. - (b) Honest verb: identified / confirmed (instrumentally, by GC-MS, against a same-day standard with a clean blank). This is one of the few places the strong verb is earned. - (c) Does NOT establish: who introduced the gasoline; when it was introduced or the fire set; why (intent); that the death was a homicide (that comes from Chapter 11). (Any three suffice.) - (d) Role relative to earlier findings: this confirms at high validity what Chapter 21 found presumptively (gasoline chemistry) and gives instrumental certainty to the accelerant underlying Chapter 22's incendiary-fire determination. It upgrades the certainty of the accelerant's identity; it does not extend the reach of any inference toward a person. It is corroborating/confirmatory, and load-bearing only as to what the debris contains.

Exercise 23.33

The chapter refuses to tie the SEM-EDX sleeve particle to a person for two reasons: (1) Class vs. individual — elemental composition is class evidence (Chapter 1, §23.5): it identifies the class of particle, not a unique source, so it cannot point to one person even in principle. (2) Transfer — particles transfer and contaminate easily and can reach a sleeve by routes unrelated to the event (handling, environment, secondary transfer), so the particle's presence does not establish how it got there. Additional evidence required before any person could be associated with the sleeve: establishing whose sleeve it is and its chain of custody; and converging person-level links from other evidence types (e.g., DNA on the garment, digital/location data, soil/pollen tying the same person to the scene). The particle is, at most, a consistent-with corroborator awaiting that context.

Exercise 23.35 †

Model answer (juror-facing, on "confirmed by GC-MS"). "Confirmed by GC-MS" means the laboratory did two independent things to the same substance: it separated it from everything else in the sample by gas chromatography — noting when it came through the instrument (its retention time) — and it broke the molecule into fragments and recorded their pattern by mass spectrometry, a structure-specific "fingerprint." When both the retention time and the whole fragmentation pattern match a known reference standard run the same day, against a clean blank, the chance that two different substances would agree on both is so small that the chemist can honestly say the compound is identified. That is one of the strongest statements in forensic science, and it deserves the trust it gets — for what it claims. But notice the boundary: the confirmation tells you what the substance is. It does not tell you who put it there, when, or why. "Gasoline is present in the debris, confirmed" is a hard fact; "therefore this person set the fire" is a separate question the chemistry never answers. A juror should accept the identification and demand other evidence for everything beyond it.

Chapter 24 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 24.1

Gunshot residue (GSR) is the particulate and chemical debris produced by the discharge of a firearm. Primer residue is its inorganic particulate fraction, formed when the metallic primer compounds vaporize on discharge and condense into particles. The diagnostic particle most commonly contains lead, barium, and antimony. Two things must be true to identify it confidently: the particle must have the right elemental composition (the three elements fused together) and the right morphology (a characteristic spheroid) — elements alone, or shape alone, are not enough.

Exercise 24.2 †

"Lead present" is weak because lead, barium, and antimony are each common in everyday life individually — lead in old paint, fishing weights, and wheel weights; barium and antimony in brake linings, certain lubricants, and industrial dusts. What is rare is a single microscopic particle that fused lead, barium, and antimony together while molten and cooled into a sphere. A bulk "lead present" reading detects an element that has countless innocent sources; the particle-by-particle SEM-EDX finding of the three-element spheroid detects the specific product of a firearm discharge. (Any two everyday sources suffice: e.g., lead in old paint; barium/antimony in brake dust.)

Exercise 24.3

The refractive index is a measure of how much glass bends (slows) light passing through it. It excludes well because two fragments with different indices cannot share a source — a single decisive difference is a clean exclusion. It associates weakly because float glass (nearly all flat window glass) is manufactured in enormous quantities within a very narrow index band, so two fragments with matching index are consistent with a common source and with two unrelated panes off the same vast production stream.

Exercise 24.5

Soil comparison is the examination and comparison of the physical, mineralogical, chemical, and biological properties of a questioned soil against known/control samples. Four independent comparison points: (1) color (moist and dry); (2) mineralogy (kinds/proportions of mineral grains); (3) particle-size distribution / texture (sand–silt–clay proportions); (4) biological components (pollen assemblage, diatoms, spores). (Also acceptable: density-gradient profile; pH/organic content; anthropogenic components.)

Exercise 24.7

The two families of cracks are radial cracks, which run outward from the point of impact like spokes from a hub, and concentric cracks, which form rings around the impact point.

Exercise 24.9 †

Three reasons the station-sampled positive is weaker than it sounds: (1) Police-transport/booking-area contamination — patrol-car seats, holding cells, and arresting officers' hands are GSR reservoirs that can deposit particles on the suspect during the very process of arrest and transport. (2) Secondary transfer — the suspect may have acquired particles by contact with a contaminated officer or surface, not by firing. (3) The three-hour delay means the sample was taken after extended exposure to those contaminated environments. Honest conclusion: "particles consistent with primer residue were present" — consistent with firing, handling, proximity, or transfer/contamination; the result does not establish that this person fired a weapon, and the contamination opportunities make the positive especially ambiguous.

Exercise 24.11

Both are class evidence, but Case B is far stronger because of distinctiveness and independence/number. Distinctiveness: a buried non-factory repaint in an unusual color is uncommon, so few unrelated vehicles would share it (the rare-feature principle). Independence/number: six matching layers are six independent features that must all coincide by chance for an innocent match — and the probability of sharing all six is the product of the individual probabilities, which is small. One common white layer (Case A) is shared by millions of vehicles and is a single feature; the multiplying rarity of Case B makes coincidence implausible (without individualizing).

Exercise 24.12 †

A difference in refractive index is an exclusion: the fragment did not come from that window, cleanly, because two pieces of glass from the same source share a refractive index. It is the strongest of the honest verbs. Had the indices been indistinguishable, you could conclude only consistency — "consistent with a common source" — because float glass of that index is mass-produced; you could not individualize the window. The contrast is the chapter's core asymmetry: a mismatch excludes; a match only narrows.

Exercise 24.13

Taking one sample from the center of a large lot is wrong because it characterizes neither the specific distinctive spot of interest (often the edge of the property, where a vehicle parked or a body lay) nor the background soil of the surrounding region. Without the distinctive-spot sample you may miss the very soil that matters; without the background samples you cannot tell whether a "match" is to a distinctive local soil or to the generic dirt of ten square miles. The investigator should take multiple known samples of the specific spot and the surrounding background, so the questioned soil's distinctiveness can be assessed against a baseline.

Exercise 24.15

Two visibly different soils layered on a shovel are a stratigraphy that can record a sequence of places visited — the order in which the shovel contacted different ground — which a single homogenized scrape would destroy by mixing everything together. It is read like the layers of a paint chip: the order and composition of the layers carry information the bulk material alone does not. (E.g., distinctive scene soil over ordinary yard soil suggests the scene was contacted after the yard.)

Exercise 24.16 †

The buried blue refinish layer adds distinctiveness: it records that the vehicle was repainted from blue to red, an uncommon history that few unrelated vehicles share, so its presence in both the questioned and known samples makes a coincidental common source much less likely. It is an extra independent, rare feature on top of the standard factory layers. Honest sentence: "The questioned and known paint are indistinguishable across all five layers, including an unusual buried blue refinish coat beneath the red basecoat; this strongly supports a common origin, though I cannot exclude another vehicle repainted in the same way."

Exercise 24.17

Can establish: (1) that there was a single point of impact (from the radial/concentric pattern); (2) which side the force came from (from the edge stress marks); and, with multiple holes, the sequence of breaks. Cannot establish: who applied the force; what object did it; when it happened; or any actor's identity. Honest verb for the directional conclusion: "consistent with" — e.g., "the fracture is consistent with force applied from the exterior face."

Exercise 24.19

The GSR sentence is appropriately honest in at least three ways: (a) it reports a count and confidence class ("three particles ... of the highest confidence class") rather than a bare "positive"; (b) it lists the multiple explanations consistent with the finding (discharged, handled, proximity, or transfer), rather than asserting the person fired; (c) it explicitly states the limit ("does not by itself establish which"), placing the finding and its boundary in the same breath. (Bonus: it says "consistent with," the honest verb, not "proves.")

Exercise 24.20 †

Problems with "Forensic analysis proves the soil on the defendant's boots came from the murder scene": 1. "Proves ... came from" overstates a class association. Soil is class evidence; it associates an object with an environment of a certain character, not with one spot to the exclusion of all similar soil. "Proves" and "came from [the exact scene]" claim individualization the method cannot support. 2. It omits the time problem and the innocent-explanation problem. Soil cannot say when the boots acquired it; if the defendant had legitimate access to the area (as Keller, a co-owner, did), the association is also consistent with an innocent visit. Honest rewrite: "The soil on the boots is consistent with, and — given that the controls show the matched mineral and pollen combination to be uncommon in the region — strongly supports, the boots having been in the scene's distinctive environment; it does not establish when the soil was acquired or that any crime was committed."

Exercise 24.21

The bare word "match" is dangerous because it does two jobs and the jury hears only the stronger one. The examiner usually means indistinguishable in the properties measured — a class statement. The jury hears the same object — an individualization. The word silently launders a probabilistic association into a certainty. The fix is to retire "match" for language that carries its own limit: "indistinguishable in [refractive index/density]," "consistent with a common source," "I cannot exclude," "strongly supports but does not individualize."

Exercise 24.23

Two reasons the television scene is backward: (1) A positive GSR result establishes presence, not firing. Particles transfer and contaminate; the swab being positive does not mean the suspect pulled a trigger — the inference from presence to action ignores the entire transfer problem. (2) Sampling at the station, after transport, invites contamination from the patrol car, the booking area, and the arresting officer; an honest workflow samples early (ideally at the scene), documents contamination opportunities, and reports the count with its limits — none of which the instant "you're the shooter" declaration does.

Exercise 24.25 †

(a) How it biases: Knowing the detective is "confident this is our shooter" can lead an honest examiner, at the margins, to resolve an ambiguous two-component particle as the three-element diagnostic one, or to report a borderline count with less hedging than it deserves — the expectation tilts close calls toward the wanted answer (contextual/confirmation bias, Chapter 31). (b) Safeguard: Keep the case narrative away from the particle examiner (context management); report the particle count and its limits, not the conclusion the detective wants. Where the contamination happens: crucially, GSR contamination usually happens before the sample reaches the lab (patrol car, booking area, officer's hands), where the analyst cannot see it — so the analyst can be entirely correct about the particles and the investigative inference can still be contaminated. That is why both procedural safeguards (early sampling, documentation, transport-environment controls) and the blind/context-managed examination matter.

Exercise 24.27

Decline because soil is class evidence: it associates an item with an environment of a particular character, and many locations can share a similar soil signature, so it cannot individualize a location the way a quantified DNA profile individualizes a person. What you can honestly say: "The soil is consistent with, and — given that the controls show the matched features to be uncommon in the region — strongly supports, having been in a place with this distinctive soil; it does not single out this exact spot to the exclusion of all similar soil, and it does not establish when the contact occurred."

Exercise 24.29 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited): 1. Single-source nuclear DNA — STRONG. Quantified, rigorously validated, known error structure; individualization with a stated random match probability (Chapter 7). 2. Multi-layer paint comparison — moderate / real-but-bounded. Objective analytical measurements (FTIR/SEM-EDX); a class association that can be strong when many unusual layers match, but it does not individualize absent a physical fit. Above the subjective pattern methods because the measurements are objective; below DNA because it is class-level without a validated frequency. 3. Single-layer common-paint comparison — weak. Same valid measurement, but one common feature → high probability of coincidental match → weak association ("consistent with"). 4. Bite-mark comparison — DISCREDITED. No validated basis for the claim that one set of teeth made a mark; multiple exonerations (Chapter 16). At the bottom. The determining feature: the rarity of the matched features (distinctiveness × number of independent features) and whether the core measurement is objective and the claim is supported by validated data — the PCAST question.

Exercise 24.31

(Open synthesis — model answer.) The chapter advances Theme 1 (exclusion over proof) and Theme 2 (the validity spectrum) most directly. Theme 1: every material is surest when it excludes (different paint layer, different refractive index, absent soil mineral, no GSR particles), and all associations are stated as probabilities — "consistent with," "strongly supports" — never "proves"; the cold-case soil is explicitly held to "places in the environment," not "proves guilt." Theme 2: the chapter locates GSR, paint, glass, and soil in the objective-measurement middle of the spectrum — above the discredited pattern methods because what they measure is sound and reproducible, below DNA because what they can conclude is a class association rarely accompanied by a validated frequency. (Also touched: cognitive bias, via the upstream-contamination GSR examiner; and the CSI effect, via the television "GSR test.")

Exercise 24.33

Keller's co-ownership bears on weight, not on the chemistry of the match. It does not weaken the analytical finding that his boot soil corresponds to the cabin's distinctive soil — the measurement stands. What it does is supply a strong innocent explanation for how and when the soil got there: a co-owner has legitimate, repeated reasons to walk the property, so the association is consistent with innocent presence as well as guilty presence, and it certainly cannot fix the soil's deposit to the time of the crime. An honest report must include this context alongside the finding, because omitting it would let "his boots carry the cabin's distinctive soil" be misread as "he was there for the killing." The soil places his boots in the environment; the co-ownership is exactly why that placement, alone, says nothing about guilt or timing.

Exercise 24.35

Sentence 1 (what the negative corroborates): The negative GSR result — no primer-residue particles where a shooting would have produced them — is consistent with the autopsy's finding of blunt-force trauma and no gunshot wound (Chapter 11) and with the conclusion that the stray cartridge case was a red herring (Chapter 15); all three agree that no one was shot. Sentence 2 (the §24.1 caution if it stood alone): Standing alone, a negative GSR would prove little, because GSR particles are lost within hours and many surfaces never bear them — so the negative is honestly usable here only because it corroborates a manner of death already established by the body, not as a load-bearing finding by itself.

Exercise 24.37 †

Glass refractive-index comparison rests on an objective physical property measured precisely (immersion/GRIM), reproducible between competent analysts; it is strong for exclusion (different index = different source, cleanly) and weak for association (float glass clusters in a narrow index band, so a match is "consistent with" only). It is most easily overstated when a "match" in refractive index is presented as proof of a common source, ignoring the mass-production stream. Soil comparison rests on a composite of many independent properties (color, mineralogy, texture, pollen, chemistry); it is strong for exclusion (a single decisive difference excludes) and can be strong for association when the soil is distinctive and the controls confirm it, but it is class evidence throughout — associating an object with an environment, not a point. It is most easily overstated when a distinctive-soil association is converted into individualization of a location or into proof of a time or a crime. In short: glass's risk is overreading a common-index match; soil's risk is overreading a distinctive-environment association into "this exact spot," a time, or guilt.

Chapter 25 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 25.1

Digital forensics is the application of scientific methods to the identification, preservation, acquisition, examination, and interpretation of data on digital devices for use as legal evidence. The five activities: identification (recognizing potential digital evidence), preservation (protecting it from change), acquisition (imaging/copying it soundly), examination (analyzing it), and interpretation (reporting what it does and does not establish).

Exercise 25.3 †

The paradox resolves once you separate the data from its handling. Digital evidence is unusually durable because so much of it is generated automatically and persists despite the user's efforts: "deleted" files survive until overwritten, and metadata is created behind the user's back — the data you don't know you're making is the data you can't erase. It is simultaneously unusually fragile because it is intangible and effortlessly changed by the act of examining it: a powered-on phone rewrites timestamps and syncs new data, and naively reading a disk alters it. There is no contradiction: the content tends to persist (durable against deletion), while the integrity is delicate (fragile against careless handling). The whole imaging-and-hashing apparatus exists to exploit the durability while protecting against the fragility.

Exercise 25.5 †

A hash value is a fixed-length string produced by a one-way cryptographic function — a digital fingerprint of data. The three forensically useful properties: (1) deterministic — the same input always yields the same hash; (2) avalanche sensitivity — changing a single bit of input changes the output completely and unpredictably; (3) practical irreversibility/collision-resistance — you cannot work backward from a hash to the data, and cannot feasibly construct different data with the same hash. Together these let an examiner prove that a copy is exact (hashes match) and that data has not changed since acquisition (the hash still matches).

Exercise 25.7

A write blocker is a hardware or software device between an evidence drive and the examiner's computer that allows reads but physically blocks writes to the evidence drive. It prevents the contamination that occurs when an operating system, left to its own devices, writes to a connected drive — indexing it, creating hidden system files, or updating timestamps — silently altering the evidence before anyone examines it. With the write blocker, the examiner can prove the original was read-only throughout acquisition.

Exercise 25.9

Cell-site/location data are historical carrier records of which cell towers and sectors a phone connected to, and when, used to infer the general geographic area in which the phone was located. The two words most often dropped in overstatement: "general" and "area" — courtroom testimony tends to convert "general geographic area" into a precise spot, which the data cannot support.

Exercise 25.11 †

Two ways the evidence may have been altered: (1) the operating system may have written to the drive on connection — creating hidden system/index files or thumbnails, changing the disk's contents; (2) simply accessing files updates their "last accessed" timestamps, corrupting the very metadata a timeline might depend on. This damages the case even with no incriminating find because it destroys the integrity guarantee: the defense can now argue the evidence was changed in custody, and the examiner can no longer prove the analyzed data matches what was seized. In digital forensics, a broken integrity chain taints everything on the device, not just one file — which is exactly why the write blocker exists.

Exercise 25.13

Using the analogy: deleting a file and emptying the recycle bin is like tearing a chapter's entry out of a book's table of contents — the listing is gone, so the system no longer knows where the chapter is, but the pages themselves are still in the book, intact, until something is printed over them. The operating system removes the file's index entry and marks its space "available for reuse," but the contents physically remain. The one event that genuinely destroys the data is overwriting — when the system actually writes new data onto that exact space. Deletion delists; only overwriting destroys.

Exercise 25.15

A carved JPEG most likely lacks: (1) its original filename — that lived in the filesystem index, which is gone; (2) a reliable timestamp (created/modified/accessed) — also stored in the index, so carving recovers the pixels but not the dates; (3) its folder/path context — where the file lived, what it was stored alongside, which user account owned it. All three are missing because carving recovers the file from raw space by its internal structure (header/footer), bypassing the filesystem entirely — so it gets the artifact without the surrounding record. (A fourth acceptable answer: the file may be partial/fragmented if its data were non-contiguous or partly overwritten.)

Exercise 25.17 †

A phone's GPS-derived location log is a direct positional fix the phone computed for its own mapping features, typically accurate to a few meters — it places the phone at a spot. Cell-site records only show which tower/sector the phone connected to, placing the phone somewhere in that tower's coverage area, which can be miles across in rural terrain, and the tower is not always the nearest one. These are fundamentally different: one is a point (meters), the other an area (potentially miles). Conflating them is the chapter's most consequential error because it converts "the phone was somewhere in this large region" into "the phone was at this exact place" — the precise overstatement that helped convict Lisa Roberts (Case Study 25.2). The honest verbs differ accordingly: GPS log → "was at"; cell-site → "was in the area of / consistent with."

Exercise 25.19

Honest: "The matching cryptographic hash values establish that this forensic image is an exact, unaltered, bit-for-bit copy of the seized device, and that the data has not changed since I acquired it." Overstated: "The matching hash proves this evidence is verified — so every file on it is authentic and true." The overstatement confuses integrity (did the data change? — what the hash actually proves) with authenticity and truth (is the data genuine and accurate? — what the hash says nothing about). A planted forgery hashes just as consistently as a genuine file.

Exercise 25.21

The EXIF GPS coordinate establishes, at honest strength, that the phone was at the cabin at 11:40 p.m. — a placement of the device, not provably its owner. The honest inference about the suspect is therefore limited: the phone associated with the suspect was at the cabin, if the suspect was the one holding it then. To connect the device's location to the person, you need additional evidence that the suspect actually had the phone at that time — for example, other activity on the phone only the suspect would generate, witness placement of the suspect with the phone, or corroborating physical evidence (soil, pollen, DNA) tying the person to the same place. The device-vs-person gap (Chapter 1's logic, restated for digital) must be bridged by other evidence, never assumed.

Exercise 25.23 †

Two overstatements: (1) area presented as a point — "at the cabin" converts a tower sector's broad coverage area into a specific spot the data cannot support; (2) phone presented as person — "the defendant's phone was at the cabin" slides toward "the defendant was at the cabin," when the records concern the device, not provably the person. (A third, the word "prove," asserts a certainty cell-site data never supplies.) Honest rewrite: "The cell records are consistent with the defendant's phone being somewhere in the coverage area that includes the cabin during the relevant window; they do not place the phone — or the defendant — at the cabin specifically, since a tower sector covers a broad area and a phone does not always use the nearest tower."

Exercise 25.25

The inference is unsound because recovery is not guaranteed and absence is not proof of non-existence. Messages that once existed may have been genuinely destroyed by overwriting (or by TRIM on a solid-state drive), or may be present only as fragments the tools failed to reconstruct, or may exist in a form/location the extraction did not reach. So "no incriminating messages were recovered" is consistent with several explanations — including that such messages existed and were overwritten. The honest report states what was recovered, what was fragmentary, and what is simply unknown; it never converts failure to recover into proof of absence.

Exercise 25.27

Two ways the television scene is backward: (1) speed and certainty — real cell-site analysis does not produce an exact GPS track in nine seconds from a single "ping"; obtaining records lawfully, reconstructing coverage, and honestly bounding the area is slow, and the output is a broad area, not a track. (2) area vs. point — a single tower connection yields a coverage area (possibly miles across), not a precise location, and the phone may not even have used the nearest tower; the fictional "exact GPS from one ping" claims a precision the method fundamentally lacks. This is the CSI effect (§1.2) in its credulity mode: teaching audiences to over-trust a method that is genuinely useful only for placing a phone in a broad area.

Exercise 25.29

The scope problem: a warrant authorizes searching the phone for evidence of insurance fraud, and that authorization defines the lawful boundary of the search. Material relating to an unrelated offense falls outside that scope. Mining the phone for the unrelated matter risks suppression of that evidence (it was obtained outside the warrant's authority) and can taint the broader search by making it look like an unlawful general "fishing expedition," jeopardizing even properly obtained evidence. The disciplined move is to stop, document the inadvertent observation, and seek a new warrant covering the new matter — keeping the original search demonstrably within its authorized scope (§25.6).

Exercise 25.31

Valuable side: commercial extraction tools can sometimes bypass a lock or recover data (deleted, encrypted-at-rest, or app-buried) that no manual method could reach, making otherwise-inaccessible lawful evidence available. Problem side: their internal methods are proprietary trade secrets — not published or peer-reviewed — so the defense cannot fully audit how a given artifact was produced; like all software they have bugs, version differences, and parsing errors that can misattribute or mis-timestamp data; and "widely used and expensive" is not "validated." This is the §4.4 validation demand and the Chapter 6 foundational-validity logic applied to software: a black-box tool's output should be a starting point to be verified by hand and documented (tool + version), with candid acknowledgment on the stand that part of the method cannot be independently inspected.

Exercise 25.33 †

• Cryptographic hashing — top of the spectrum. Pure, testable mathematics with the avalanche and collision-resistance properties; for the narrow claim "this data is unaltered," it is near-certain — in rigor it stands alongside DNA for what it proves.
• GPS-derived location log — strong (for the device). A direct positional fix accurate to meters; strong evidence of the phone's location, though not of who held it. Well up the spectrum, below hashing's certainty.
• Historical cell-site pinpointing — near the bottom. The core claim (a precise spot) outruns the data, which give only a coverage area; pinpoint testimony has caused wrongful convictions (Lisa Roberts). Sits with the overstated pattern methods.
• Bite-mark comparison — bottom/discredited (Chapter 16, previewed): no validated basis for its individualization claim; multiple exonerations. Why two "digital" methods sit at opposite ends: "digital" is a medium, not a validity rating. Hashing's claim is mathematically testable and modest; cell-site pinpointing's claim is physically unsupported and immodest. The label they share tells you nothing about where their claims sit — which is the section's whole point.

Exercise 25.35

• Integrity vs. truth: Sound — "the hash proves this file has not changed since seizure." Overstatement — "the hash proves this file is true / was written by the defendant." (Hashing covers the container, not the contents.)
• Device vs. person: Sound — "the phone was at the cabin at 11:40 (EXIF GPS)." Overstatement — "the suspect was at the cabin at 11:40." (The phone's location is not provably the owner's.)
• Area vs. point: Sound — "the cell records are consistent with the phone being in the coverage area that includes the cabin." Overstatement — "the cell records place the phone at the cabin." (Coverage area, not a spot.)

Exercise 25.37

"Keller's alibi breaks" and "Keller did it" are different sentences because the first is a statement about one thread (his account of where he was is contradicted by his phone's location records and his own deleted messages), while the second is a conclusion about the ultimate question that no single thread can support. Breaking an alibi removes a defense; it does not supply proof of the act — Keller might have been in the area for innocent reasons, and the phone's presence is not provably the man's. Converting "his alibi is gone" into "he is guilty" is precisely the overstatement the book warns against. The only place the whole file is finally weighed — every thread together, with exclusions, convergence, and the defense's counter-arguments — is Chapter 39 (the capstone).

Exercise 25.39

An honest examiner's response: "Counsel is right that the cell-site records cannot pinpoint the cabin — a tower sector covers a broad area, and a phone does not always use the nearest tower, so I cannot and do not say the phone was at the cabin. But that concession does not make the records worthless. They show Mr. Keller's phone connecting to towers whose coverage area includes the cabin during the relevant window — which is inconsistent with his stated alibi of being an hour away at the time. The records do not prove where he was; they contradict his account of where he was. Those are different things, and only the second is what I'm offering." This concedes the pinpoint (as §25.5 requires) while preserving the legitimate negative use — breaking the alibi.

Exercise 25.40 †

Model answer (150–200 words): Two pieces of digital evidence, two very different levels of confidence — and honesty means treating them differently. When a digital examiner says your data is "unaltered," that is about as close to certainty as forensic science gets, because it rests on a mathematical fingerprint called a hash: run the data through a fixed formula and you get a value that changes completely if even one bit changes. If the value still matches months later, the data did not change — provably, not as a matter of trust. But a hash only proves the data did not change; it cannot tell you the data is true. Cell-site evidence is the opposite trap. It feels precise — "the phone was near the scene" — but it only shows the phone was somewhere in a tower's coverage area, which can be miles across, and phones do not always use the nearest tower. It places a phone in an area, not a person at a spot. Both lessons — trust the math, distrust the pin — are the same discipline: claim exactly what the evidence supports, and not one inch more.

Chapter 26 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 26.1

Image forensics is the application of scientific methods to the analysis, measurement, and authentication of images and video for legal purposes. Its three broad jobs: (1) measurement — recovering real-world dimensions (e.g., a person's height) from images (photogrammetry); (2) authentication — determining whether an image/video is what it purports to be; and (3) reading content — establishing what footage actually shows about an event, time, place, and action. It is not the television job of conjuring detail that was never captured.

Exercise 26.3 †

The rule: enhancement makes information already present in an image easier to perceive; it can never add detail the sensor did not record. Why "zoom-and-enhance" is impossible: the detail in an image is fixed at capture by resolution, lens, focus, exposure, motion, and compression. If a license plate occupies six blurry pixels, those pixels hold a fixed, finite amount of information; no processing can manufacture the dozens of crisp pixels needed to read the plate, because that information was never recorded. You can only clarify what is there, not create what is not.

Exercise 26.5

ELA re-saves a JPEG at a known compression level and maps the difference between the original and the re-saved version; regions that were edited and re-saved may carry a different "error level" than regions saved only once, so a manipulated area can appear with a distinct brightness. Does a bright region prove manipulation? No. Innocent features — sharp edges, high-contrast boundaries, overlaid text, and fine texture — routinely produce high error and "light up." ELA is a screening hint toward areas to examine with other methods, never a standalone proof.

Exercise 26.7

Provenance is the documented origin and history of a digital file — its source device, capture time and place, and edit history. Metadata is data embedded in or attached to a file that describes the file rather than its visible content. The common image-metadata standard is EXIF (Exchangeable Image File Format), which can record camera make/model, date/time, exposure settings, and (often on phones) GPS coordinates.

Exercise 26.9 †

Three reasons for a range: (a) posture and footwear/headwear add or subtract centimeters the analyst cannot directly see; (b) stand-point uncertainty — the exact spot the person occupied is itself inferred from the footage, and perspective changes apparent size with distance; (c) image quality — blur/low resolution makes the head-top and floor line ambiguous by several pixels, which is centimeters at distance. A real-world factor that should widen the range: unknown shoe (or hat) thickness, since footwear of unknown height directly shifts apparent stature and cannot be measured from the frame.

Exercise 26.11

At "approximately 165–171 cm" vs. a 188 cm suspect, the honest conclusion is a clean exclusion: the figure is too short to be the suspect, and the applicable verb is exclude (the figure is not the suspect, barring an error in the reconstruction). Contrast an overlap case — e.g., the figure estimated at 180–186 cm and the suspect 183 cm: there the verb is at most consistent with, because the band contains a very large fraction of adult men; an overlap fails to exclude the suspect but does not identify him. (This is the §1.6 asymmetry: a mismatch can close a door; a match only narrows the hallway.)

Exercise 26.13

Three independent checks on a time stamp: (a) the recording system's clock settings (was it set correctly, in the right time zone, drifted?); (b) a known synchronized event captured on the same footage (a phone call, a transaction, another camera's verified clock) to cross-reference; (c) file/container metadata consistency (do the embedded timestamps agree with the displayed stamp?). The stamp is a claim because it reflects whatever time the device's clock was set to by a human or network — which can be wrong, drifted, or mis-zoned — not an independently verified fact about when the event occurred.

Exercise 26.15

Three metadata-independent physics/content checks: (a) shadow consistency — do all shadows fall coherently from a single light source, or does a pasted object cast its shadow the wrong way?; (b) reflection coherence — are reflections in eyes, windows, and water geometrically consistent with the scene?; (c) perspective/vanishing lines — do they agree across the image, or does an inserted object violate the scene's geometry? (Also: uniform noise vs. a region carrying a different sensor fingerprint.) Each is hard to fake because a forger must get every physical detail right and the real world is unforgiving — a single inconsistency betrays a composite, and the checks rest on physics rather than a fragile single statistic.

Exercise 26.16 †

Version (A) — brightness/contrast and noise reduction that leave two characters faintly readable — is admissible enhancement: it reveals the limited information actually present in the original pixels. Version (B) — a machine-learning "super-resolution" output showing a crisp, fully legible plate — is fabrication: the original six pixels could not contain that many characters of real detail, so the crisp characters are what a model trained on other plates generated to fill the gap, not what the camera recorded. The separating rule: enhancement reveals what the sensor captured and stops there; the moment an output contains more detail than the original resolution can support, it has crossed from revealing into inventing, and is an illustration of the software's guess, not evidence of the scene.

Exercise 26.17

The error: ELA bright regions do not prove editing. Two innocent features that routinely "light up": (a) sharp edges / high-contrast boundaries; (b) overlaid text (also fine texture). Honest version: "Error level analysis flags this region as compressing differently from its surroundings, which warrants closer examination by other methods; on its own it does not establish manipulation, and the brightness here is consistent with a high-contrast edge / overlaid text." (A competent cross would show the same glow on the image's unedited edges.)

Exercise 26.19

No — it does not prove authenticity. By the metadata asymmetry (§26.6), internal inconsistency can undercut a claim cleanly, but consistency only fails to exclude tampering. Metadata that all agrees and matches the device has merely passed one test among many; it could still have been edited carefully, fabricated wholesale (with consistent fake metadata), or stripped and rebuilt. Consistency is necessary but not sufficient: it removes one red flag but does not establish that the file is genuine, which requires corroboration (other authentication indicators, provenance, independent evidence).

Exercise 26.20 †

Two overstatements in "Enhanced surveillance imagery positively identifies the defendant as the man buying gas cans": 1. "Enhanced … imagery" implies enhancement added identifying detail. Enhancement can only reveal what was captured; it cannot create the resolution needed to identify a face that the sensor did not record. Presenting an enhanced image as having produced an identification overstates what enhancement can do. 2. "Positively identifies" overstates identification from footage. Appearance/height from surveillance footage supports "consistent with," not identification to the exclusion of all others of similar build; facial comparison from imagery is contested and error-prone. Honest rewrite: "Surveillance footage, processed only to make the captured detail perceptible, shows a person consistent with the defendant in height, build, and clothing purchasing gas cans. It does not, by itself, identify the defendant to the exclusion of others of similar appearance."

Exercise 26.21

Problem: an upscaler's output is generated, not captured — a neural network trained on many faces produces a plausible face by filling in detail from its training data, so calling the result "what the camera captured" misrepresents a model's guess as a recording of the scene. The crisp face reflects the training set, not the crime scene's photons. Defensible description: "This is a machine-generated reconstruction that estimates a face consistent with the low-resolution input; it is an illustration of what the model inferred, not a record of what the camera captured, and it must not be treated as evidence of the person's appearance."

Exercise 26.23

Missing cautions that should undercut "definitely not a deepfake": (a) novel-method generalization — detectors perform well on the fakes they were trained on and degrade sharply on new generation methods, so a high score on a benchmark says little about a fake made by a method released after training; (b) compression/re-sharing — social-media re-compression degrades the signals detectors rely on, lowering reliability on real-world footage; (c) error-rate validation — the field lacks the well-designed, independent error-rate studies (Daubert/PCAST) that would justify "definitely." A clean detector result supports "no detected synthetic signatures," not "definitely authentic."

Exercise 26.24 †

The ethical line: honest enhancement reveals information already in the image and is documented and reproducible; producing a misleading image means generating or sharpening detail beyond what the original resolution supports (interpolation/"super-resolution" presented as captured) so that characters appear legible that the sensor never recorded. The analyst must not manufacture a readable plate from unreadable pixels. What to document and say: record every processing step from the unaltered original (reproducibly), report exactly which characters are legible at the true resolution and which are not, and state plainly that the remaining characters cannot be recovered — "three of seven characters are legible and consistent with X, Y, Z; the remainder is not determinable from this footage." Refusing to fabricate is the service to the court, not yielding to the request.

Exercise 26.25

Not necessarily a scandal. ELA and similar screening tools are confounded (edges, text, texture all produce glows), so a disagreement about whether a glow indicates manipulation or a high-contrast edge is a predictable consequence of the tool's limits, not proof one analyst is incompetent. The analyst attributing the glow to the high-contrast edge is reasoning more defensibly, because they are accounting for ELA's known confounds rather than treating a bright region as a verdict. The responsible resolution is to test the region with other, independent methods (compression history, content/physics consistency, source-device noise) rather than to rest on the ELA map.

Exercise 26.27

Automated face recognition has been shown to vary in accuracy across demographic groups and with image quality, so deploying it as proof rather than a lead risks systematically higher error for some groups and a false sense of certainty (overstated individualization). Previewing Chapter 31, the danger compounds with confirmation bias: a "candidate match" can anchor the investigation and the analyst's subsequent judgments. The safeguard: treat any system output as an investigative lead requiring independent corroboration, document the system's known limitations and error variation, and never present a database "candidate" as an identification on its own.

Exercise 26.28 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → unsettled/contested): 1. Documented reverse-projection photogrammetry — SOUND/real, with honest error bars. Rests on real geometry, is empirical (it measures the actual projection via a known reference), and yields a height range with stated assumptions; strongest as an exclusion. Below DNA (its inputs are noisier and its output is a range), but well-grounded. 2. Well-understood image enhancement (brightness/contrast/noise reduction) — SOUND. Well-characterized, reproducible signal processing that reveals captured detail; valid when documented and not over-reaching. 3. Error level analysis as a standalone manipulation verdict — WEAK/contested. A heavily confounded screening hint, not a proof; fails on many formats and re-saved images. 4. Current deepfake detection — UNSETTLED frontier. Real and improving but fast-changing, lacking validated error rates on novel fakes and on compressed media; not yet at courtroom-conclusion standard. Justification turns on each method's measured reliability and the rigor behind its core claim — the PCAST question — and on whether the output is a quantified measurement (photogrammetry) or an unvalidated judgment (ELA verdict, deepfake conclusion).

Exercise 26.29

The metadata caution connects to Chapter 25 because both chapters teach that digital evidence's integrity and provenance must be established by method, not assumed: Chapter 25 introduces hash values and write blockers to prove a file has not changed and to image devices without altering them, and flags metadata and cell-site data as fragile. This chapter applies the same discipline to images — secure the native file, hash it, and treat embedded metadata as fragile and forgeable. The lesson appears twice because images and video are digital evidence: the same threats (alteration, stripped/forged metadata, copies degrading the original) apply, and the same safeguards (originals, hashing, corroboration, skepticism toward unverified metadata) are required.

Exercise 26.31 †

Cold-case evidence-log entry (model): - (a) Defensible inference: a person consistent with Keller (in height, build, and clothing) purchased two gas cans at the station in the relevant window before the fire. - (b) Honest verb: consistent with (corroboration of purchase and presence). - (c) Does NOT establish: that the person is Keller to the exclusion of others of similar appearance; when relative to the fire the cans were used; that Keller used them for the fire; that Keller killed Diallo. (Any three suffice.) - (d) Why not "identified": photogrammetric height and general appearance support membership in a large class of similar people, not individualization; facial comparison from such footage is contested and error-prone. Logging "identified Keller" would convert "consistent with" into proof the footage cannot bear — exactly the leap the crowdsourced Boston misidentification made (Case Study 26.1). The footage corroborates alongside other evidence (the confirmed gasoline accelerant, the digital trail, soil); it does not, alone, name the buyer.

Exercise 26.33

Two other evidence types already in the file that strengthen the Keller link when combined with the CCTV: (a) the digital trail — Keller's deleted messages and cell-site data placing him near the cabin against his alibi (Chapter 25); (b) soil on Keller's boots matching the cabin's distinctive soil (Chapter 24). (A third: the gasoline confirmed as the accelerant, Chapters 21–23, which makes a gas-can purchase relevant rather than incidental.) The footage alone cannot make the leap because surveillance imagery supports only "a person consistent with Keller," not identification, and says nothing about whether that person committed the crime; person-level and act-level links must come from the convergence of these independent threads — no single one proves it.

Exercise 26.35 †

"Enhancement" honesty (model answer, ~150–200 words). On television, "enhancement" conjures detail from nothing — a blur becomes a readable face on command. That is a lie, and an expensive one. A digital image's detail is fixed at the moment of capture by the sensor's resolution, the lens, focus, exposure, and compression; if the information was never recorded, no processing can create it. What real enhancement can do is make information already present easier to see: brighten shadows to reveal a captured face, reduce noise to expose a real edge, average many frames of a static scene to recover a genuinely cleaner image. Those operations are honest, documented, and reproducible. The danger is the modern "super-resolution" tool — a neural network that renders a crisp, confident face or plate from a few pixels. That output is more dangerous than an honest blur precisely because it looks real while being fabricated: it shows what the model guessed from its training data, not what the camera saw. An honest blur tells the truth about its limits; a fabricated sharp image tells a convincing lie.

Chapter 27 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 27.1

Forensic accounting is the application of accounting, auditing, and investigative skills to financial questions that a legal proceeding must answer. The feature that most distinguishes it from a routine external audit is its investigative, non-assurance mindset: where an audit presumes good faith and seeks reasonable assurance that statements are fairly presented, forensic accounting assumes someone may have deliberately made the books balance and asks how — it begins where the auditor's presumption of good faith ends.

Exercise 27.3 †

The fraud triangle holds that financial crime by otherwise law-abiding people tends to require three conditions at once: 1. Pressure — a (typically non-shareable) financial problem the person feels they cannot solve legitimately. Question: who needed money, and why couldn't they ask for help? 2. Opportunity — access to commit the act and a way to conceal it, with a low perceived chance of being caught. Question: who could both act and hide it (where is segregation of duties missing)? 3. Rationalization — a story that squares the act with the person's self-image as honest ("just borrowing it," "they owe me"). Question: what grievance or sense of entitlement is in the air? The triangle predicts where to look, not who offended; most people under pressure with opportunity never steal.

Exercise 27.5

The audit trail is the chronological, documented record of transactions and process steps, detailed enough to trace each entry back to its source and forward to its result. It lets an examiner answer not merely "what is the balance?" but "where did every dollar of it come from, when, authorized by whom, and supported by what document?" — i.e., it reconstructs the history and authorization of a balance, not just its total.

Exercise 27.7

Benford's law is the observation that, in many naturally occurring datasets spanning several orders of magnitude, the leading digits are not uniformly distributed but follow a smooth decreasing curve — the digit 1 leads about 30% of the time and the digit 9 only about 5% — rather than each digit appearing roughly one-ninth (~11%) of the time.

Exercise 27.9 †

With no segregation of duties, one partner initiates, approves, records, and reconciles every transaction — so this most directly opens the opportunity leg of the fraud triangle (the ability both to commit and to conceal). Two schemes the arrangement makes available: (a) a fictitious-vendor (or inflated-invoice) scheme — the same person can create a payee, approve the invoice, sign the check, and book it, with no one to catch the fabrication; and (b) skimming / check tampering — incoming funds or outgoing checks pass through one set of hands with no independent reconciliation, so diversions go unnoticed. The absence of an independent second person is the enabling condition for both.

Exercise 27.11

This describes the layering stage. Counterintuitively, all the movement makes the money easier to trace once an investigator has subpoena power because each transaction is recorded in an institution the launderer does not control — every wire, account, shell-company transfer, and offshore step leaves a record at a bank, processor, or registry. Layering aims to make following the trail impractical (long, multi-jurisdictional, tangled), but it does not make the records disappear; with legal access to those third-party records, the very complexity becomes a documented chain. Layering relocates the evidence; it does not erase it.

Exercise 27.13

Money laundering "relocates" rather than destroys evidence because every stage that disguises money also records it — in institutions the launderer does not own, which survive even when the launderer shreds his own copies. Three record types created by the act of laundering itself: (1) deposit and placement records (and the currency-transaction or suspicious-activity reports those deposits trigger); (2) transaction records in every account the money touches during layering (wire-transfer trails, inter-account transfers, shell-company ledgers, correspondent-bank records); and (3) integration documents — sham loan agreements, corporate filings, and asset titles that return the money to the economy. The launderer's effort to erase the trail is itself a trail.

Exercise 27.12 †

The subcontractor has skipped placement — he never introduced the cash into the financial system at all (no deposit, no instrument, no layering). Skipping placement makes a net-worth analysis of his lifestyle more revealing, not less: because the illicit money never entered any account, none of his legitimate, recorded income can account for the assets and spending the cash supports. When investigators measure how much his net worth grew, or line up his documented inflows against his outflows (source-and-application), the cash-funded house, vehicle, or spending appears as a gap with no lawful recorded source — the cleaner the avoidance of the banking system, the more starkly the unexplained wealth stands out against his on-the-books income. Cash that is never laundered is invisible as a transaction but glaringly visible as unexplained net worth.

Exercise 27.15

A financial indicator is a red flag that directs attention (a backdated entry, an out-of-hours transaction, a round-number adjustment, a gap in the trail); a financial finding is a defensible conclusion reached only after the indicator is investigated and innocent explanations are tested and excluded. Example that is both a legitimate red flag and has a common innocent explanation: a backdated entry in the accounting system — it can signal a fraudster manufacturing a record to fit a false date, or it can be an honest correction of a transaction that was genuinely missed when it occurred. The indicator justifies asking; only the answer justifies concluding.

Exercise 27.17

The chart shows too few values beginning with 1 and a conspicuous excess at 4 and 5. The chapter's threshold-gaming explanation: if claims of \$500 or more require a supervisor's sign-off, a manager submitting padded or fictitious claims while avoiding scrutiny will cluster invented amounts *just under* \$500 (e.g., \$420, \$465, \$490, \$455) — producing exactly the surplus of 4s and high 5s and the deficit of naturally common 1s. What the Benford screen did: it flagged a statistically anomalous subset and pointed the investigator at a specific stack of claims to examine by hand. What it did not do: prove fraud, or identify a culprit — the deviation is an indicator that earned a closer look, and the hands-on review of the actual claims is what would establish (or refute) padding.

Exercise 27.19

A fraudster who learns about Benford can fabricate amounts whose leading digits match the expected distribution, defeating the screen entirely. This tells you that a passed Benford test clears no one (conformity can be manufactured) and that the screen is only ever a triage tool: it can direct attention toward anomalous data, but its silence is not exoneration and its alarm is not proof.

Exercise 27.16 †

The formula for the expected proportion of leading digit $d$ (for $d = 1 \ldots 9$) is $$P(d) = \log_{10}\!\left(1 + \frac{1}{d}\right).$$ Reasoning from the chapter's Figure 27.1 (no calculator needed): the term $1/d$ shrinks as $d$ grows, so $1 + 1/d$ shrinks toward 1 and its logarithm shrinks toward 0 — the expected frequency therefore declines with each successive digit. It is a smooth decline, not a flat ~11% across all nine digits, because the differences $1+1/d$ are large for small $d$ (1 → $\log_{10}2 \approx 30\%$) and small for large $d$ (9 → ~4.6%). A flat ~11% would be the (wrong) human intuition that each digit is equally likely; Benford-obeying data instead favor small leading digits, steeply at first and then gently.

Exercise 27.20 †

A first-two-digits Benford test examines the distribution of the leading pair of digits (10 through 99) rather than just the first digit. It is especially good at catching transactions structured just under an approval threshold because threshold-gaming concentrates fabricated values in a narrow band — amounts like \$480, \$485, \$490, \$495 all share the first digit 4 but reveal themselves sharply as an excess at the two-digit values 48, 49 sitting right beneath a \$500 limit. The first-digit test smears that signal across the whole "4" bucket; the first-two-digit test localizes it to the exact pair of digits where the gaming clusters, making the spike beneath the threshold far more visible.

Exercise 27.21

Practitioners use a statistical goodness-of-fit measure rather than eyeballing because the eye cannot reliably judge whether an observed departure from the Benford curve is larger than would arise by chance for a dataset of that size. A formal test gives you a principled threshold: it quantifies how improbable the observed deviation is under the assumption that the data obey Benford, so you can distinguish a meaningful anomaly worth investigating from ordinary sampling noise — and you can defend that judgment, with a stated criterion, under cross-examination, rather than relying on "it looked off to me."

Exercise 27.23 †

The expert is not entitled to say "which proves he had his partner killed." That asserts the ultimate issue (Chapter 30) — the conclusion of guilt reserved for the jury — and it converts evidence of motive (a financial why) into a finding about who committed the act, which accounting cannot establish. Motive is shared by many innocent people; the ledgers depict a financial situation, never the crime itself. Defensible version: "Mr. Keller was the named beneficiary of a large life-insurance policy on the decedent, was carrying liabilities exceeding his liquid assets, and the renovation books contain entries that a Benford screen flagged and a hands-on review found consistent with inflated costs — facts establishing a strong financial motive." (State the motive; leave the verdict to the jury.)

Exercise 27.25

Real businesses contain irregular-looking entries that are not fraud for many reasons; two: (a) innocent messiness — genuine data-entry errors, legitimate cash transactions, lost receipts, late entries, and round-number estimates are routine, especially in small and informal firms; and (b) legitimate corrections — a backdated or reversing entry is often an honest fix of a missed or mistaken record, and after-hours entries can simply be a busy owner doing the books at night. The forensic accountant's obligation is to treat each anomaly as a question, pursue the innocent explanation, and report a manipulation only when the records, taken together and after alternatives are tested, will not yield to an honest account.

Exercise 27.27 †

Risk: told before analysis that "Keller is good for it — find the money," the analyst lets the chosen suspect anchor the reconstruction — ambiguous entries get read as confirmation, the investigation goes looking for a motive rather than letting the records surrender one, and a confirmation-bias spiral (Chapter 31) converts "fits the profile of someone who could have done it" into "the evidence shows he did." Safeguard: keep domain-irrelevant information (who the suspect "is," what the detective believes) away from the analyst until the financial reconstruction is fixed; then compare to the case facts. Have the reconstruction checked by someone who does not already "know" the answer. A motive the analyst went hunting for, having already chosen a suspect, is worth LESS than one the records yield on their own — because only the latter is uncontaminated by the expectation, and a motive that might be the product of bias cannot be distinguished from genuine independent evidence.

Exercise 27.29

Not necessarily a sign of dishonesty. A net-worth reconstruction from incomplete records depends on assumptions — which assets and liabilities are counted, how cash of uncertain origin is treated, what is presumed about unrecorded income — and reasonable experts can weigh those inputs differently in good faith, producing different estimates. This is the same structure as honest disagreement elsewhere in the book (e.g., two entomologists differing because the temperature reconstruction is uncertain): the disagreement lives in the reconstructed inputs, not in arithmetic. The defensible response is for each expert to state assumptions explicitly so the fact-finder can see where, and why, they diverge — and a competent cross-examination probes those input assumptions rather than alleging fraud.

Exercise 27.31 †

On the NAS 2009 / PCAST 2016 validity reasoning (strong → discredited): 1. Single-source nuclear DNA — STRONG. Quantified, rigorously validated, known error structure (Chapter 7); the gold standard. 2. Benford's-law screen used to direct an examination — SOUND, as a screen. Its mathematical basis is uncontroversial; used to triage where to look (like a presumptive color test, Chapter 21), it is excellent. It does not individualize anything; it earns a closer look. 3. A Benford "violation" offered as proof of fabrication — JUNK / overstatement. Same analysis, dishonest claim: a screening indicator presented as a conclusion, a non-sequitur dressed in mathematics. It commits the overstated-fingerprint-match error — treating an indicator as a finding. 4. Bite-mark "matching" — DISCREDITED. No validated basis for the specific claim; multiple exonerations (Chapter 16); at the bottom. The key point: items 2 and 3 are the same Benford analysis — its place on the spectrum depends entirely on the claim made for it (screen vs. proof), which is the chapter's central discipline.

Exercise 27.33

The independence is a strength at assembly (Chapter 39) because independent lines of evidence that agree corroborate one another, whereas lines that secretly share an input do not — they would be a single line wearing two hats, and a flaw in the shared input would propagate to both. The pathology finding (no soot in the airways → dead before the fire) rests entirely on the body and the autopsy; the financial motive rests entirely on the records. Because neither depends on the other, their agreement genuinely multiplies confidence. If the accountant's "motive" had secretly been inferred from the assumption of homicide (rather than from the records), it would add no real weight — it would just echo the pathologist. Genuine convergence requires genuinely independent threads.

Exercise 27.34 †

Cold-case evidence-log entry (model): - (a) Defensible finding: Roy Keller had a strong financial motive for Diallo's death — a death-contingent life-insurance policy naming Keller, severe non-shareable debts, and renovation books containing Benford-flagged entries found on hands-on review to be consistent with inflated/fabricated costs. - (b) Honest verb: the records establish a strong financial motive / strongly support the existence of a motive (resting on the documented policy, the documented debts, and the document review behind the Benford flag). - (c) Does NOT establish: that Keller killed Diallo; that he was at the cabin; when anything happened; that he acted on the motive at all. (Any three suffice.) Motive answers why, not who. - (d) Why no conversion to "who killed Diallo": the records show circumstance and incentive, never the violent act; many innocent people have debts and convenient policies, so "strongest motive" cannot, by itself, identify the actor without committing the prosecutor's-fallacy-style leap (Chapter 9). Anything stronger than "person of interest" must await convergence with opportunity and means at the capstone (Chapter 39).

Exercise 27.35

Two next steps to move from "statistically anomalous" to a defensible statement about the renovation books: (a) Pull and examine the flagged entries by hand against source documents — match each anomalous cost entry to an invoice, a supplier record, a bank payment, and (where possible) physical evidence of the work, to see whether the figure corresponds to a real, delivered cost or to a fabrication; and run a first-two-digits test to localize any threshold-gaming. (b) Reconcile the books against independent third-party records — supplier statements, bank and card records, and the audit log's metadata (who entered/modified each entry, and when) — since a fabricated cost will lack a genuine counterparty and may show backdated or out-of-hours creation. Each step replaces the indicator (the deviation) with a finding grounded in documents the suspect did not control.

Exercise 27.37

(Short-writing model.) "Following the money" is often the most durable evidence in a case because money has to move, and every movement is recorded redundantly in institutions the perpetrator does not control — banks, processors, insurers, title companies, tax agencies. A scene can be burned and a laptop wiped, but the bank still holds the payment and the insurer still holds the policy; the trail routinely outlives the physical evidence. Yet a strong financial motive is never proof of guilt: it establishes why a crime might have been committed, and many people with debts, heirs, and convenient policies commit no crime at all. To reason "he had the most to gain, therefore he did it" is to ignore the base rate of innocent people with motives — the same error, structurally, as the prosecutor's fallacy. Motive earns its real weight only in convergence with independent evidence of opportunity and means; alone, it is a reason to look harder, not a finding. (≈150 words.)

Exercise 27.38 †

(Short-writing model.) A Benford's-law screen and a confirmatory laboratory test (e.g., GC-MS, Chapter 23) sit at opposite ends of the evidentiary chain. The Benford screen entitles you only to a closer look: it flags a subset of figures as statistically anomalous and tells you where to investigate, exactly like a presumptive color test in chemistry. It belongs at the start of an examination, as triage. A confirmatory test entitles you to a finding: GC-MS identifies a substance to a high degree of analytical certainty and belongs at the end of the chain, after a screen has narrowed the question. The cost of treating the screen as the confirmation is severe — you would present an indicator to a jury as a conclusion, the same overstatement as a fingerprint "match" with no error rate, and you would skip the only step (hands-on examination of the actual records, or the confirmatory instrument) that can actually establish the fact. The screen earns the look; the look proves the case. (≈170 words.)

Chapter 28 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 28.1

Forensic psychology is the application of psychological science and clinical practice to questions a legal system must answer. The three tasks that make up the bulk of the working field are competency to stand trial, criminal responsibility (the insanity defense), and violence risk assessment.

Exercise 28.2 †

Criminal profiling is the inference of an unknown offender's likely characteristics — age, sex, employment, personality, background, post-offense behavior — from the evidence of the crime and the crime scene. The premise on which the whole practice rests is that behavior reflects personality: that the way an offender behaves while committing a crime leaves a readable trace of who they are, so scene behavior can be "read back" into a portrait of the person. (The chapter's central critique is that this premise is overstated — behavior is heavily situation-driven and inconsistent across contexts — so the inference it licenses is unvalidated.)

Exercise 28.3

Competency to stand trial is present-tense: it asks whether the defendant, right now, can understand the proceedings and assist counsel — i.e., whether it is fair to try them at all. The insanity defense is past-tense: it concerns the defendant's mental state at the time of the offense and goes to criminal responsibility (guilt). Competency decides whether the trial may proceed; insanity decides whether the defendant is criminally responsible for the act. They are decided separately, by different evidence, sometimes years apart.

Exercise 28.5 †

The Barnum effect (Forer effect) is the tendency to accept vague, broadly applicable personality descriptions as highly accurate descriptions of oneself specifically, when the descriptions are general enough to fit almost anyone. Three everyday practices that exploit it: horoscopes/astrology, cold reading (by psychics/mediums/fortune-tellers), and generic personality "tests" or fortune-cookie-style readings that feel tailored but apply universally. (Any of these genres count; all manufacture felt specificity from statements that contain none.)

Exercise 28.7

Dusky v. United States (1960) holds, in plain terms, that a defendant is competent to stand trial only if they have sufficient present ability to consult with their lawyer with a reasonable degree of rational understanding and a rational as well as factual understanding of the proceedings against them. Mere orientation (knowing the date and place) is not enough; the test is about present, functional capacity to understand and participate.

Exercise 28.9 †

The portrait's features that make it feel specific while predicting almost nothing: - "a male in his twenties or thirties" — describes an enormous share of the plausible offender pool for many crimes; barely narrows anything. - "intelligent but underachieving" — flattering and near-universal; almost no one can be excluded by it, and it is hard to disconfirm. - "unresolved anger and a difficult family history" — extremely common; fits a large fraction of people and especially of offenders. - "likely follows the news coverage of his crime" — the unfalsifiable flourish: it can be confirmed by almost anyone and disconfirmed by almost no one; hedged with "likely."

Each is a Barnum-style statement: broadly applicable, hedged, hard to prove false. One claim rewritten as falsifiable: "The offender has a felony arson conviction in this county within the last five years" — a claim that can be checked against records and can turn out flatly false, which is exactly why a correct hit would actually narrow the field. (Other acceptable rewrites: a specific workplace, vehicle, handedness, or physical feature that a named suspect either has or does not.)

Exercise 28.11

Error: this is the prosecutor's fallacy (Chapter 9) applied to risk. A structured instrument estimates a probability for a group — the proportion of people in a risk category who reoffend over a period — not a certainty about this individual. Saying the test "proves he will be violent" converts a group base rate into an individual verdict, which the instrument cannot support. What it can honestly support: that the defendant falls in a category whose members reoffend at approximately a stated rate over a stated period, with the explicit caveat that the tool cannot say whether this particular person will be among those who do.

Exercise 28.12 †

The crimes are linked by behavior that is stable across the series and not required to commit the offense — i.e., a signature (a non-functional ritual repeated at each scene), as opposed to the modus operandi (the functional how-to of entry, control, and escape), which may vary as the offender adapts. So an analyst looks past the changing MO for a persistent, non-functional behavior common to all the scenes. Honest strength / verb: the conclusion is "these offenses are consistent with a common offender" — a linkage hypothesis, never "proves the same person committed all of them," because coincidence, copycat behavior, and analyst over-interpretation remain possible.

Exercise 28.13

Geographic profiling is more defensible because both its inputs and its output are objective and testable. Inputs: the locations of a linked series of crimes — geographic coordinates, not inferred personality. Output: a probability surface over a map (a prioritized search area for the offender's likely anchor point), which is falsifiable — it can be checked against where the offender actually turns out to live. It draws on well-studied regularities in how people travel and offend relative to home. It still produces a search area, not an address, and depends on the crimes truly being linked — but it traffics in measurable geography rather than an unfalsifiable portrait of a mind, which is why it sits higher on the spectrum than psychological profiling.

Exercise 28.15

The insanity defense reconstructs a past mental state — the defendant's mind at the moment of the offense — which cannot be examined directly and must be inferred from interview, records, and the circumstances of the crime. Competency assesses a present state that can be examined now and re-examined later, with structured instruments and a defined functional standard (Dusky). Because a past internal state is intrinsically harder to pin down than a present, observable capacity, qualified evaluators more often reach defensible disagreements about sanity — hence the "battles of the experts" cluster there rather than around competency.

Exercise 28.16 †

MO behaviors (functional, varying): the method of entry (pried door → cut screen → forced garage) and the means of control (on-site cord → zip ties → rope). Candidate signature (non-functional, constant): arranging the victim's shoes in a row — an act the burglary did not require. What the figure concludes: the varying MO plus the repeated, non-functional shoe ritual are consistent with a single offender across the three scenes — a linkage hypothesis at the strength "consistent with," not "proves." Three reasons the linkage is still not certain: (1) coincidence — the "constant" detail may recur by chance or be more common than assumed; (2) copycat — a later offender may imitate a publicized detail; (3) analyst over-interpretation — the detail may be read as more distinctive/stable than it really is. (Bonus: the behavior links behavior, not identity — it never reveals who the offender is.)

Exercise 28.17

The single distinguishing property is falsifiability — whether the statement risks being false. The specific-prediction column (handedness, a named workplace, a vehicle with a described dent, a felony record) makes claims that can be checked against a named suspect and can turn out wrong; the Barnum column makes claims so broad and hedged that they can scarcely be wrong. Why "cannot be wrong cannot be informative": information is the elimination of possibilities. A claim that fits virtually every candidate eliminates none of them, so confirming it tells you nothing new — no matter how accurate it feels. Only a claim that could have been false, and was checked and held, actually narrows the field.

Exercise 28.19

"Subject is likely to be employed in a position below his true intellectual capacity" is a classic Barnum claim because it is flattering, vague, and near-unfalsifiable: a great many people privately believe they are underemployed relative to their ability, so it resonates widely; "likely" hedges it; and "below his true capacity" is almost impossible to disconfirm, since the claimant's "true capacity" is undefined and unmeasurable. It feels insightful (it implies the profiler perceives a hidden, sympathetic truth about the offender) while excluding essentially no one and predicting nothing checkable.

Exercise 28.20 †

Overstatement: the witness moves from "the defendant fits the profile" to "in my opinion he committed this crime" — using a behavioral portrait as proof of identity. Profiling cannot support that leap: its core inference (scene behavior → offender characteristics) is unvalidated, its founding typology is empirically shaky, and it has no established error rate for the identity claim (§28.4). Why courts generally exclude it as proof (Chapter 5): under Daubert/FRE 702, a method offered to prove identity must show foundational validity — testability, a known/acceptable error rate, reproducibility — which profiling lacks; so it is generally not admitted to prove that a particular defendant committed the crime. Most a behavioral analyst could defensibly say: at most a linkage statement — e.g., "the behavior across these offenses is consistent with a common offender" — never a portrait offered to identify the defendant.

Exercise 28.21

Reasoning error: survivorship bias / selective storytelling — the documentary recounts only the hits (cases where a profile seemed to match) and silently omits the misses, so a curated anthology of successes stands in for an actual error rate. A method's validity is not established by its remembered triumphs. What would be needed instead: controlled studies that count the failures too — e.g., measuring profilers' predictions against known facts across many solved cases (hits and misses), comparing profilers to control groups, and reporting an error rate. When the failures are counted, the studies that exist find profiling weak — which a collection of vivid successes cannot outvote.

Exercise 28.23

A correct arrest at the end of an investigation does not show the profile worked as a method because the outcome is contaminated by everything that happened in between: the case may have been cracked by physical, digital, or financial evidence entirely unrelated to the profile, while the profile itself pointed the wrong way for weeks. A profile present at the end tells you nothing about whether it pointed away from the offender, or toward innocent people, in the middle. How a profile can do harm even in a "solved" case: by misdirecting attention and resources toward a phantom or an innocent (as with Richard Jewell), consuming time the investigation needed and, at worst, branding an innocent person — harm that a correct final arrest does not erase.

Exercise 28.24 †

Danger (tunnel vision): a "lone, opportunistic stranger" profile tells the investigation what kind of person to look for, and the search obligingly finds people of that kind — here, transients and outsiders — while the people the profile points away from (the victim's close associates) stop being scrutinized. Combined with confirmation bias (Chapter 31), facts that fit the profile get weighted and facts that don't get discounted, so a wrong profile becomes self-reinforcing and steers the case off course. How a profile should be treated instead: as one tentative, falsifiable hypothesis to be actively tested and disconfirmed — investigators should keep pursuing the evidence that does not fit it, weight verifiable physical/digital evidence over the behavioral narrative, and, when the profile feels most compelling, ask precisely what would prove it wrong. A profile believed is more dangerous than a profile doubted.

Exercise 28.25

The framing "find that the defendant is incompetent" asks the evaluator to reach a predetermined conclusion rather than to answer the Dusky question honestly — it inverts the proper order (assess, then conclude) and pressures the expert to serve the retaining side's litigation goal. The obligation is to conduct an impartial evaluation against the legal standard and report the honest finding whatever it is, owing a duty to the court and to professional ethics, not to the attorney's desired outcome. This is the hired-gun problem (previewed for Chapter 30): the safeguard is that the expert serves the question, the standard, and the record — and that an opposing evaluation and cross-examination can expose a result that was reached to order.

Exercise 28.27

Because the genuine products of forensic psychology (competency, sanity, risk) and the weak one (profiling) are different in validity, conflating them harms justice in both directions. If the validated assessments are tarred with profiling's weakness, courts and juries may wrongly discount competency and risk evidence that is, in fact, standards-governed and reliable — a real loss. Conversely, if profiling borrows the credibility of those validated assessments (riding on the profession's general prestige), juries may over-trust an unvalidated portrait as if it shared the rigor of a competency evaluation — exactly the over-selling this book warns against. Keeping the two distinct protects the credibility the real work has earned and denies the weak work a credibility it has not.

Exercise 28.28 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → discredited): 1. Single-source nuclear DNA — STRONG. Quantified, rigorously validated, known error structure; the field's gold standard (Chapter 7). 2. Structured violence risk assessment — real and validated, modest accuracy. Built on empirically validated risk factors and explicit rules; measurably outperforms unstructured judgment; yields a group probability with stated uncertainty. Below DNA because accuracy is modest and group-based, but genuine science. 3. Predictive criminal profiling — LOW / weak & unproven. Shaky founding typology, unvalidated behavior→biography inference, unimpressive accuracy studies, no established error rate for the identity claim; generally not admitted as proof of identity. 4. Bite-mark comparison — DISCREDITED. No validated basis for the specific claim that one set of teeth made a mark; multiple exonerations (Chapter 16). At the bottom. What profiling lacks that even contested physical-comparison methods attempt: a physical comparison that could, in principle, be validated. Firearms and bloodstain methods at least compare physical features against physical features and could be subjected to error-rate studies; predictive profiling rests on a psychological inference about an absent person's character, which the evidence does not support and which is far harder even to test. (Note: profiling is placed above bite marks here mainly because its cousins — linkage, geographic profiling — have traction and it is rarely offered as courtroom proof, so it has done less direct convicting harm.)

Exercise 28.29

This chapter is the purest example of the CSI effect (Theme 4) because no forensic product is more glamorized on television and less validated in fact than the criminal profiler: the activity the public most strongly associates with forensic psychology is precisely its least validated activity, and the Barnum effect is the psychological mechanism by which that over-sold product manufactures a feeling of accuracy it cannot earn. The second theme it most strongly advances is the validity spectrum (Theme 2): the chapter extends the book's yardstick into the behavioral sciences, separating the methods that examine testable subjects (competency, risk — higher) from the one that reads an absent mind (profiling — low). (One sentence of justification suffices; either ordering of the two themes is acceptable as long as both are named.)

Exercise 28.31 †

Cold-case workbook entry (model): - (a) Honest status: "Profiling misled; excluded as proof." The "lone drifter/stranger" profile is not evidence, proves nothing about who killed Marcus Diallo, and could not. - (b) Direction / whom it pointed away from: it pointed outward, toward transients and outsiders, and thereby away from the victim's close associates — specifically away from Roy Keller, the business partner with the closest financial entanglement to the property. - (c) Why it excludes/includes no one: a behavioral narrative is not proof; it generated suspicion, not evidence, so it neither rules anyone in nor rules anyone out — only physical, digital, financial, and biological evidence can do that. - (d) Methodological lesson: a profile is, at best, one tentative, falsifiable hypothesis to be actively tested and disconfirmed; here it pointed the wrong way and nearly cost the investigation time. When a profile feels most compelling, that is exactly the moment to ask what would prove it wrong and to keep following the evidence that does not fit it. (Do not state or hint at the case's solution.)

Exercise 28.33

Genuine psychological contributions that could still help Mill Creek (from §28.6): even though the profile was wrong and worthless, the validated parts of the field remain available — e.g., investigative interviewing done with non-leading technique (protecting the reliability of any statements taken from witnesses or persons of interest; Chapters 32–33), threat/risk assessment of any identified, examinable person where relevant, and behavioral linkage if the case were ever connected to other offenses (a hedged "consistent with" judgment, not a portrait). Why rejecting profiling ≠ rejecting the field: the failure was specific to predictive profiling of an absent stranger — an unvalidated inference — not to psychological science as such; the contributions that rest on examinable subjects and validated methods keep their value regardless of the profile's error. Throwing out the whole field because one weak product failed would be its own error.

Exercise 28.35 †

Competency vs. predictive profiling — validity foundations (model answer). Competency-to-stand-trial assessment rests on an explicit legal standard (Dusky: present rational/factual understanding of the proceedings and ability to assist counsel), a known, present, examinable defendant, structured assessment instruments designed for the specific question, and a correctable, re-testable answer (competency can be restored and re-evaluated). It is strong as bounded, standards-governed forensic science and is admitted in court routinely; it is most easily weakened only when an evaluator strays from the standard or yields to a retaining side's desired outcome (the hired-gun risk). Predictive criminal profiling rests on an overstated premise (behavior reliably reflects personality), an empirically shaky typology (organized/disorganized), an inference about an absent, unexaminable person, and no established error rate; its accuracy studies are unimpressive. Under Chapter 5's admissibility standards (Daubert/FRE 702), competency and sanity testimony — grounded in recognized standards and methods — is admitted, while profiling offered as proof of identity is generally excluded for lack of foundational validity. In short: competency examines a present person against a defined standard with validated tools and a feedback loop; profiling infers an absent mind with none of these — which is why they sit at opposite ends of the spectrum despite sharing a profession.

Chapter 29 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 29.1

Rapid DNA is the fully automated generation of an STR profile from a reference sample inside a single self-contained instrument, without a human analyst, in roughly 90 minutes to two hours. It is validated for clean, single-source reference samples — typically a buccal (cheek) swab from a known person.

Exercise 29.2 †

Microbial forensics is the application of the analysis of microbial communities (bacteria and other microorganisms) to legal questions. The necrobiome is the community of organisms that participates in decomposition and changes in a characteristic, time-dependent sequence after death. It is conceptually a cousin of insect succession (forensic entomology, Chapter 13): just as insects colonize remains in a predictable sequence read as a clock for the postmortem interval, the microbial communities on/in a body and in the underlying soil shift through a reproducible sequence that researchers propose to read as a PMI estimate — with the same kind of dependence on temperature, moisture, and local conditions that bounds the insect clock.

Exercise 29.3

Forensic isotope analysis is the measurement of stable-isotope ratios (of elements such as hydrogen, oxygen, carbon, nitrogen, sulfur, and strontium) in biological tissues to infer geographic origin, travel history, and diet. It answers from roughly where, and eating roughly what — a region and a life-history sketch. It cannot identify a person or pin a specific town or itinerary; many places share similar isotopic signatures and the reference maps are coarse.

Exercise 29.5

Familial searching is the deliberate search of a criminal DNA database (CODIS) for a partial match to a crime-scene profile, on the theory that a close relative of a known offender already in the database may be the true source. It rests on inheritance: a parent and child share half their DNA, and full siblings share on average half — so a close relative of the true offender produces not an exact match but a partial one, sharing many more alleles than an unrelated person but not all. Kinship software flags and ranks these near-misses.

Exercise 29.7

The Rapid DNA Act of 2017 is the U.S. statute (Public Law 115-50) that created a pathway for approved rapid-DNA instruments to generate profiles from arrestee reference swabs and, under defined conditions and approved processes, to search them against CODIS. It enabled the booking-station reference-typing use — the validated envelope of rapid DNA — not field analysis of crime-scene evidence.

Exercise 29.9 †

First use (legitimate). A clean, single-source buccal swab is inside rapid DNA's validated envelope: the instrument performs the same STR chemistry a laboratory would, on exactly the kind of sample it was validated for. Second use (not legitimate). A low-template, heat-degraded crime-scene swab is outside the envelope — it is precisely the kind of sample (touch DNA, mixtures; Chapter 8) that requires human interpretation and probabilistic reasoning (Chapter 9). The safeguard automation removed is the human analyst who would re-extract, adjust input DNA, inspect the data, and — critically — recognize that the sample is a mixture or too degraded to call and stop. The box returns a confidently formatted result either way; it does not know it is in trouble. Validity is a property of method-plus-sample, not of the instrument.

Exercise 29.11

Three distorting variables (any three): temperature (microbial succession rate is temperature-dependent), soil chemistry, moisture, the individual's own starting microbiome, and decomposition environment (buried vs. exposed vs. submerged). "Under given conditions" bounds the method because the reference data that translate a microbial community profile into a PMI estimate are condition-specific; outside the conditions for which reference data exist and match the scene, the estimate is an extrapolation — the same limitation that bounds insect succession (Chapter 13).

Exercise 29.12 †

Defensible: "The stable-isotope ratios in the tooth enamel, bone, and hair are consistent with an individual who spent childhood in one broad climatic/geological region, adult years in another, and who changed diet or location in the final months of life — an investigative lead to narrow a missing-persons search." Overstated: "The isotopes prove this person grew up in [specific town] and then moved to [specific city] on [a specific route]." The overstatement claims individual-level and specific-place/route resolution that isotope evidence — region-level, reference-map-limited — cannot support.

Exercise 29.13

A "96% similarity score" is the model's internal similarity metric — a number expressing how closely the system ranked the candidate against its database under its own training — not the probability that the match is correct. The probability of correctness depends on the system's measured false-match rate for this image quality and this population, on the size of the database searched, and on the base rate of the true source being present at all — none of which the bare score conveys.

Exercise 29.15

Three axes: 1. Genetic markers: familial searching uses the same ~20 STR loci as CODIS; IGG generates hundreds of thousands of SNPs. 2. Database searched: familial searching searches the criminal database (CODIS); IGG searches a consumer genealogy database (e.g., GEDmatch). 3. Degree of relatedness reached: familial searching reaches close relatives (parent, child, sibling) already in the criminal system; IGG reaches distant relatives (third–fourth cousins) via family-tree reconstruction.

Exercise 29.16 †

Five steps for the Golden State Killer identification: 1. Generate a SNP profile from the crime-scene DNA (not the ~20 STRs CODIS uses). (lead) 2. Upload to GEDmatch, a consumer genealogy database, returning distant relatives (3rd–4th cousins) ranked by shared DNA. (lead) 3. Reconstruct family trees from public records (censuses, obituaries, certificates), converging on common ancestors. (lead) 4. Triangulate downward and winnow by age, sex, and California geography to one candidate — Joseph James DeAngelo. (lead) 5. Confirm with a conventional STR comparison of abandoned DNA against the crime-scene profile. (identification) Steps 1–4 generate the lead; step 5 makes the identification. This division of labor is methodologically honest because the new, contested technique only points, while the old, validated gold-standard method (Chapter 7) does the identifying that reaches a jury — the technique never pretends to be the conclusion, unlike the self-confirming bite mark (Chapter 16).

Exercise 29.17

Automation bias is the human tendency to over-trust a machine's output because it is a machine and feels objective — an examiner waving through an algorithm's conclusion they would have scrutinized in a colleague. Bias laundering is a model trained on historically biased data learning the bias and re-emitting it wearing the costume of mathematical objectivity. Each is harder to challenge than the human bias it replaces: automation bias suppresses the very scrutiny that would catch the error ("the computer found it"), and laundered bias carries the false authority of "objective computation," which is more persuasive — and so harder to contest — than an obviously fallible human judgment.

Exercise 29.18

Forensic testimony rests on the premise that an expert's reasoning can be explained, scrutinized, and cross-examined (Chapter 30), and that a Daubert gatekeeper (Chapter 5) can assess the method's basis. An algorithm that cannot "show its work" defeats all of this: the defense cannot probe the basis of the conclusion, the jury cannot weigh the reasoning, and the gatekeeper cannot evaluate a logic opaque even to its builders. (A weather forecast faces no such requirement because no one's liberty turns on cross-examining it.) A correct answer with no auditable reasoning is an oracle, and oracles have no place on a witness stand.

Exercise 29.19 †

For Figure 29.2: (a) What it shows — the system ranked this driver's-license photo as its top candidate against its database, by its own internal similarity metric, under its own (undisclosed) training. (b) What it does not establish — identity; "96%" is not a probability of correctness; "99% accuracy" is unverified marketing on an unknown test set; false-match rates may be far higher for this image quality and this subgroup; the reasoning is unauditable. (c) Strongest honest use — an investigative lead: a candidate to confirm by independent means (a human examiner with documented reasoning, or other corroborating evidence), exactly as IGG produces a lead that conventional DNA confirms — never an identification on the strength of the score.

Exercise 29.21

The error is scope creep / use outside the validated envelope (§29.1): rapid DNA is validated for clean single-source reference samples, not for evidentiary crime-scene samples analyzed in the field with no laboratory and no analyst. The ignored earlier-chapter problem is the interpretation of touch DNA and mixtures (Chapter 8): a field box cannot recognize a low-template mixture or reason about secondary transfer, and reporting its output as an identification compounds the error by skipping the probabilistic interpretation (Chapter 9) such samples require.

Exercise 29.23 †

Problems with "AI facial recognition has positively identified the defendant from surveillance footage": 1. "Positively identified" overstates a lead. A facial-recognition output is, at most, a similarity-ranked candidate to investigate, not a positive identification — especially from low-quality footage. 2. No validation / no error rate. The claim implies a reliability the tool has not been shown to have: no independent validation on this population, no measured false-match rate for this image quality and subgroup, and a similarity score that is not a probability of correctness. Honest rewrite: "A facial-recognition system returned the defendant as a candidate match to the surveillance image; this is an investigative lead to be confirmed by other evidence, not an identification, and the system's error rate for images of this quality and for this population has not been independently established."

Exercise 29.25

The rhetorical move is the appeal to sophistication — "it uses next-generation sequencing and deep learning, so it must be reliable." Sophistication is not validity: an unvalidated complex method is actually less trustworthy than a validated simple one, because its failures are harder to see and harder to audit. The only thing that establishes reliability is independent validation with a measured error rate — which complexity neither provides nor excuses.

Exercise 29.27

The detective has fallen into automation bias — treating the machine's output as confirmation because a machine produced it. What the tool actually produced was an investigative lead: a similarity-ranked candidate to investigate by independent means. "Flagged" is not "confirmed"; confirmation requires auditable, validated, independent evidence (a documented human re-examination, a conventional DNA match, corroborating facts), not deference to the algorithm.

Exercise 29.29

A post-upload terms-of-service change sharpens the informed-consent problem because users uploaded under different expectations — to find relatives, not necessarily to enable law-enforcement matching — and consent given for one purpose does not extend to a materially different one imposed later. The interests affected reach far beyond the uploader: because of consent at a distance, a single person's upload partially exposes the genomes of their relatives (siblings, parents, children, cousins) who never uploaded, never consented, and may not know the database exists — yet become findable by police through the changed terms.

Exercise 29.31

A celebrated success silences scrutiny: when a method works in a dramatic, public case, questioning its costs or its validation can feel churlish or contrarian, and institutional momentum (funding, adoption, prosecutorial enthusiasm) builds around it before the boring validation and the cost-accounting are done. That is exactly when scrutiny matters most, because a method adopted on the strength of one triumph — without measured error rates or attention to who bears its costs — repeats the field's oldest mistake (admitting a method on impressive demonstration rather than validation), now with the added authority of a famous win. The Golden State Killer case is honest progress because its costs were examined even in victory.

Exercise 29.32 †

On the validity spectrum (strong → emerging/unvalidated): 1. Rapid DNA on a clean reference swab — STRONG. Inherits the validated STR method, automated, inside its envelope; near the top. 2. The STR confirmation following an IGG lead — STRONG. The gold-standard method of Chapter 7 doing the actual identifying; top of the spectrum. 3. Forensic isotope analysis as a geographic lead — MIDDLE. Rigorous analytical chemistry, but the inference from measurement to geography is coarse and reference-data-limited; strong only for the region-level investigative claim it honestly makes. 4. A microbial-PMI estimate today — EMERGING / BELOW. Real underlying science, but incomplete forensic validation and no established error rate; not yet courtroom-ready. 5. An opaque, vendor-validated facial-recognition match offered as proof of identity — BOTTOM. Unauditable reasoning, no independent validation, marketing accuracy claim; sits with the discredited methods regardless of architectural sophistication. Justification turns on the measured error rate and the auditability of each method's core claim — the PCAST question, applied prospectively.

Exercise 29.33

Rapid DNA can be near the top of the spectrum for a clean single-source reference swab and effectively off it for a degraded crime-scene mixture. The single deciding variable is the sample (and the question asked of it): inside the validated envelope the instrument does validated STR work; outside it, the instrument is being misapplied and the automation has removed the analyst who would have caught the problem. This proves that validity is a property of method-plus-sample-plus-question, not of the instrument alone — the same box is excellent or worthless depending entirely on what it is asked to do.

Exercise 29.35 †

Cold-case evidence-log entry (model): - (a) Defensible inference: the minor contributor to the gas-can DNA mixture is not an unknown stranger from outside the established field of persons of interest; the "random intruder" theory is closed. - (b) Honest verb: excludes (the stranger theory). This is an exclusion, not an identification. - (c) Does NOT establish: the identity of the minor contributor; how or when those cells reached the handle (secondary transfer remains live for a low-template minor component); that the contributor set the fire or committed any crime. (Any three suffice.) - (d) Why stop at an exclusion: rapid DNA and IGG here narrowed the field by closing a door — they reached a lead, not a courtroom identification, and the gold-standard STR interpretation (the likelihood ratio for any named person) is the work of Chapter 9, not this beat. To name the contributor now would claim an inclusion the science has not earned. The book's first theme: forensic science excludes far more reliably than it proves — claim exactly the exclusion and not one inch more.

Exercise 29.37

The detective is wrong because the genealogy generated an investigative lead — a name to investigate — not a courtroom identification. Even when decisive in narrowing the field (here, excluding the unknown-stranger theory), the genealogy points; it does not prove. For a name to be honestly spoken, the lead must be confirmed by the validated, direct method: a conventional STR comparison between the crime-scene profile and a sample known to come from the candidate, interpreted with its likelihood ratio (Chapters 7, 9). Until that confirmation, the honest output is a narrowed field and an exclusion — not an identification.

Exercise 29.39 †

Isotopes vs. an AI black box — validity foundations (model answer). Forensic isotope analysis rests on rigorous analytical chemistry: stable-isotope ratios are measured precisely and reflect real, lawful relationships between tissue, water, diet, and geology. Its honest claim is a region and a life-history sketch — an investigative lead — and it is most easily overstated when a region-level "match" is dressed as identification of a person or a specific town/itinerary, a precision the coarse, reference-map-limited inference cannot bear. An unvalidated AI black box rests on nothing a court can examine: it learned patterns from undisclosed data, produces an output with no auditable reasoning, and carries a vendor accuracy claim that is not a validation study. Its honest claim is, at most, a lead to confirm; it is most easily overstated when its output is presented as an identification and its similarity score mistaken for a probability of correctness. In short: isotopes have a sound foundation but a modest, coarse inference (overstated about place); the black box may have no examinable foundation at all (overstated about identity), and the latter is the more dangerous because its opacity hides the very thing that should be scrutinized.

Chapter 30 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 30.1

An expert witness is a witness permitted, because of specialized knowledge, skill, experience, training, or education, to offer opinions and conclusions (not merely firsthand observations) to help the jury understand the evidence. The single most important difference from a lay witness: a lay witness may testify only to what they perceived (what they saw, heard, did), while the expert is the one category of witness the law allows to say what the evidence means — to interpret, not merely report.

Exercise 30.2 †

Voir dire of an expert is the preliminary examination, conducted before the witness gives substantive testimony, to decide whether they may testify as an expert at all (if they survive it, the court "qualifies" them in a stated field). It is fought on two fronts, each tied to Chapter 5: 1. Qualification — is this person, by training and experience, competent in the specific field of the specific opinion offered? This is the Kumho Tire point: the court must ask about the particular opinion, not the witness's general impressiveness (a toxicologist may be qualified for toxicology and not for bloodstain geometry). 2. Reliability of the method — even a qualified witness may not present an opinion built on a method that fails the Daubert/FRE 702 reliability test. A shrewd attorney folds a Daubert challenge into voir dire to exclude the testimony before the jury can be impressed by it.

Exercise 30.3

(a) Who conducts each: direct examination is conducted by the attorney who called the witness; cross-examination is conducted by the opposing attorney (the side that did not call them). (b) Kind of question: direct uses largely open-ended questions ("What did you do? What did you find? What does that mean?") that let the witness explain; cross uses largely leading questions that suggest their own answer and confine the witness toward "yes"/"no," letting the examiner control the witness and expose what direct smoothed over.

Exercise 30.5

Communicating uncertainty is the skill of conveying a finding's true strength and its limits to a non-specialist audience as a strength of evidence or probability rather than a certainty — being exactly as confident as the science warrants. The two opposite failure modes it threads between are overstatement (rounding a probability up into a certainty to satisfy the room) and uselessly vague hedging (burying the finding in so many qualifications that the jury, unable to extract any usable strength, defaults to ignoring it).

Exercise 30.7

Adversarial allegiance is the well-documented tendency for experts to reach conclusions favorable to whichever side engaged (and pays) them, even when they sincerely believe themselves objective. "I am an objective scientist" is a symptom, not a defense because the bias is unconscious — the same expectation-driven mechanism that warped the Mayfield comparison (Chapter 14), now operating through the structure of who retained you. An expert who is most confident they are immune is often the most exposed, because confidence in one's own objectivity is precisely the state in which the drift goes undetected.

Exercise 30.9 †

Being "qualified as an expert in forensic odontology" is a statement about the witness's standing — that the court has accepted this person as competent to opine in the field. It says almost nothing about whether this opinion — that a bite mark "matches" the defendant — is sound, because that depends on the validity of the method, which is a separate question. Bite-mark comparison (Chapter 16) has no validated basis for the specific claim that one set of teeth made a particular mark; it is among the discredited pattern disciplines, with multiple exonerations. So an analyst can be readily qualified (the legal bar for qualification is often low and courts are generous) and still deliver testimony the discipline cannot support. Qualification asks "is this person an accepted expert?"; the validity spectrum asks "can the method do what the opinion claims?" — and only the second tells you whether to believe the bite-mark "match." (The ceremony of the judge "accepting" the witness primes the jury to conflate the two; making them hear the distinction is a central task of cross-examination.)

Exercise 30.11

Each selling sentence rewritten as an honest teaching sentence: - (a) "I calculated that this DNA result is about a million times more probable if the defendant contributed than if an unknown, unrelated person did. That is very strong support for the first possibility over the second — it does not, by itself, tell you he is guilty, which depends on all the other evidence." (The original transposes the conditional — the prosecutor's fallacy of Chapter 9; the rewrite keeps the comparison attached and the verb on the evidence.) - (b) "I found agreement in the features I documented, with no unexplained differences, supporting the conclusion that the latent and the defendant's exemplar came from the same source. I cannot exclude every other person on Earth, so I do not state this as a certainty." (The original asserts identity-as-proof; the rewrite states strength of support and concedes the limit.) - (c) "The origin-and-cause analysis and the laboratory-confirmed ignitable-liquid residue support the conclusion that the fire was incendiary. No single burn indicator proves arson; the conclusion rests on the convergence of valid origin analysis and confirmed accelerant." ("A reasonable degree of scientific certainty" is an empty incantation, §30.4; the rewrite states what the conclusion actually rests on.)

Exercise 30.12 †

The six cross-examination seams (§30.3) and the upstream chapter where each vulnerability is created: 1. Certainty — over-claiming on the stand itself; the Mayfield trap (Chapter 14, the fingerprint "100% match"). 2. Assumptions and inputs — e.g., the number of contributors assumed for a likelihood ratio, or a reconstructed temperature; the error lives in the inputs (modeled by the dueling entomologists of Chapter 13; the LR's inputs, Chapter 9). 3. Case-specific application — the gap between "the method is valid in general" and "it was performed correctly here," the reliable-application prong of FRE 702 (Chapter 5). 4. Contamination and chain of custody — upstream handling errors that taint everything downstream (Chapters 2 and 4). 5. Bias and what you were told — contextual bias from knowing the "wanted" answer before analyzing (Chapter 31). 6. Error rate and the literature — the absence of a measured false-positive rate; "no method can claim zero" (Chapters 5 and 6). The unifying point: the seams are created upstream, not in the science the cross-examiner cannot defeat head-on.

Exercise 30.13

"The best defense against a withering cross is a candid direct" because you cannot be cross-examined out of a concession you already made honestly. Mechanically: a cross-examiner's productive questions ask the witness to concede the method's limits (you can't exclude everyone; the error rate is non-zero). If those limits were already stated on direct, the questions strike conclusions the witness volunteered, and there is nothing to extract. Figure 30.1 makes it concrete: Witness A claimed certainty on direct, so each forced "no" ("No, I didn't examine everyone on Earth") looks like a discovered weakness, and credibility bleeds out one concession at a time. Witness B stated the identical limitations first, as features of honest science, so the same questions land harmlessly. Same evidence, same likelihood ratio; opposite outcomes — driven entirely by the words chosen on direct.

Exercise 30.15

A competent cross-examiner avoids attacking validated science head-on because they cannot win that fight — they will not out-argue a population geneticist about allele frequencies or persuade a jury that GC-MS does not identify gasoline, and trying makes the lawyer look foolish and the expert look strong. Instead they attack the seams that surround the science: the witness's certainty, the assumptions and inputs, the case-specific application, chain of custody, bias, and the error rate. This is more effective because those are the places where an honest method can be honestly limited (or a dishonest witness exposed) without disputing the underlying chemistry — and where a jury primed for certainty feels the ground shift when the witness concedes a limit.

Exercise 30.16 †

(a) The only thing that differs between the two boxes is the words each witness chose on direct examination — certainty ("it's a match," "essentially zero error") versus strength of support ("strongly supports," "non-zero and measured"). The laboratory result, the likelihood ratio, and the actual strength of the evidence are identical. (b) Witness A's every "no" looks like a discovered weakness because the witness claimed certainty, and each honest limitation of the method (you can't exclude everyone; the error rate isn't zero) contradicts that claim — so conceding it reads as a retreat from a position the witness took, even though every limitation was always true. The cross-examiner discovers nothing; they simply collect the difference between what was claimed and what is defensible. (c) The figure's one-sentence lesson: you cannot be cross-examined out of a concession you already made honestly — the best defense against a withering cross is a candid direct.

Exercise 30.17

The sentence ("I found agreement in the minutiae I documented, with no unexplained differences, and in my opinion the latent and the exemplar came from the same source; I cannot and will not put a number like 'certainty' on that") does at least three things well: 1. It states the basis for the opinion (agreement in documented minutiae, no unexplained differences) rather than asserting a bare conclusion — the jury can see why. 2. It keeps the verb on the evidence/source comparison ("came from the same source"), not on the defendant's guilt — it stays inside the ultimate-issue rule. 3. It refuses certainty explicitly ("I cannot and will not put a number like 'certainty' on that"), conceding the method's limit on direct, which both is honest and disarms the cross. (Bonus: it avoids the empty incantation "reasonable degree of scientific certainty.")

Exercise 30.19

A witness who gives nothing but qualifications fails the court because the purpose of expert testimony is to give the jury something to weigh; "it's complicated, I can't really say" offers no usable strength of evidence, so the jury defaults to ignoring it — exactly as harmful, in its way, as overstatement, because the court learns nothing. The skill is the narrow path that is both fully hedged and usable. Rewrite: "This evidence strongly supports the proposition that he contributed — about a million times more probable than if an unknown person did. I want to be clear about the limits: it is a mixture interpreted within its validated range, it tells you his DNA is consistent with being present and not that he is guilty, and it does not tell you how or when his DNA got there. With those limits stated, the strength of support is real and substantial." (Honest about its limits, still informative about its strength.)

Exercise 30.20 †

The prosecutor's closing — "The expert told you it's a match — a match — so you know the defendant did it" — commits two distinct errors: 1. About the science: it treats a "match" as identity to the exclusion of all others, when most pattern methods can support only "consistent with"/"strongly supports." A match is a strength of evidence, not proof of unique identity (overstatement, §30.4). 2. About the expert's role / the verdict: it converts the expert's finding directly into guilt ("so you know the defendant did it"), performing the jury's step — weighing all the evidence against the burden of proof — that no expert opinion can supply (the ultimate-issue confusion). How honest direct testimony makes this argument impossible: if the analyst had testified "the evidence is consistent with / strongly supports the defendant as a contributor — it does not tell you he is guilty, which depends on everything else you will weigh," the prosecutor would have no "match" to inflate and would be visibly contradicting the witness's own stated limits. Candor on direct removes the raw material of the overstated closing.

Exercise 30.21

A phrase that sounds rigorous is dangerous precisely because it borrows the authority of science without its substance. "To a reasonable degree of scientific certainty" is a legal formula, not a scientific one: it has no defined numerical meaning, and for a method with no measured error rate it asserts a confidence that was never quantified. It launders the absence of validation into the appearance of certainty, and a jury hears reassurance where there is none. The one question to ask (from Chapter 5) whenever it is attached to a pattern-comparison conclusion: "certainty measured how, against what error rate?" For most methods to which the phrase is attached, the honest answer is that the certainty was never measured at all.

Exercise 30.23

The overstatement is "match … to the exclusion of all other firearms" — a claim of individualization (one source, no other on Earth) that firearms/toolmark comparison cannot support. What PCAST 2016 concluded (Chapter 15): firearms identification had not been established as foundationally valid in the way DNA has — at best it is supported by limited studies, and any claim must acknowledge a non-trivial, non-zero error rate; it cannot deliver certain individualization. Defensible version: "I found agreement in the toolmarks consistent with these striations having been produced by the defendant's firearm; this is support for a common source, not an identification to the exclusion of all other firearms, and the method has a measured, non-zero error rate."

Exercise 30.24 †

What the analyst must do: report the exclusion — the result that helps the defense — with exactly the same care, completeness, and promptness as they would report an inclusion. The analyst must not bury it, soften it, delay it, or shade the interpretation because it disappoints the side that paid for the test. Why the identity of the payer is irrelevant: the expert's overriding duty is to the court and the truth, not to the party who retains and pays them (§30.6). The chain of custody and the electropherogram "do not know who is paying"; the duty runs to the result. An analyst who reports inclusions faithfully but suppresses or weakens exclusions has become a hired gun — an advocate for a side rather than a neutral witness — and has abandoned the role the law grants the expert. (Concretely: the prosecution's own analyst reporting an exclusion that frees the defendant is the clearest possible demonstration that the duty runs to the truth.)

Exercise 30.25

The duty to the court requires the expert to decline — they may not testify that a validly performed analysis is "junk science," because that is false, and the obligation to the truth outranks the obligation to the client. The expert may legitimately probe the analysis, point out genuine limits, and decline to overstate its strength against the defendant — but they may not assert a falsehood (that sound science is junk) merely because it helps the side that hired them. What it costs: the expert may lose the engagement, the fee, and future referrals from a defense bar that wanted a more obliging witness — the same career cost a prosecution-aligned expert bears for reporting an exclusion. Bearing that cost is the difference between a forensic scientist and a hired gun.

Exercise 30.27

Restated in plain terms: the honest expert serves no side's interest — not the prosecution's, not the defense's — but only the accurate reporting of what the evidence shows; in that sense they belong to everyone in the courtroom and, above all, to the truth. The concrete career cost: an expert who reports exclusions that hurt the prosecution, declines to call sound science "junk" for the defense, and concedes limits that wound whichever side called them will, over time, be passed over for future engagements by attorneys who want a more obliging witness. The willingness to bear that cost — to be nobody's reliable ally — is precisely what makes the testimony trustworthy.

Exercise 30.29 †

The ultimate-issue rule is "the same rule Chapter 9 gave for DNA, restated in statistical language" because of the Bayesian decomposition posterior odds = likelihood ratio × prior odds. The expert owns the likelihood ratio — the weight of the evidence, the multiplier — and may state it ("the evidence is about a million times more probable if he contributed than if an unknown person did"). The prior — everything else known about the case — and therefore the posterior (the probability of guilt) belong to the jury, who supply the prior and perform the multiplication that yields a verdict. An expert who states a "probability of guilt" has reached into the jury's box and grabbed the prior, performing a calculation that is not theirs, with information they do not have. Guilt is a posterior; the scientist must claim only the LR and never the posterior.

Exercise 30.31

Which themes and the quotes about language. This paragraph should name at least two of the four themes and show how each appears in the chapter's argument about words: - Exclusion over proof (Theme 1) — in court it "becomes a claim about language, because the entire difference between an honest expert and a dangerous one lives in whether they say 'consistent with' or 'proves.'" The cross-examiner "punish[es] 'proves' and leave[s] 'strongly supports' untouched." - The CSI effect cuts both ways (Theme 4) — it is "no longer about distant jurors; it is about the person on the stand," who must communicate uncertainty to "an audience that television has trained to hear hedging as weakness and confidence as truth"; the discipline is "making the truth too plain to mishear." - (Also available: the validity spectrum — overstatement as a low-validity method dressed in high-validity certainty; cognitive bias — adversarial allegiance as "the same mechanism that warped the Mayfield comparison, now operating through the structure of who hired you.") A complete answer names the themes explicitly and quotes the specific language move for each.

Exercise 30.32 †

Cold-case model answer. The state's DNA analyst's direct-examination answer about the gas-can mixture:

"I compared the minor contributor's profile in the mixture against Mr. Keller's reference. Stating my propositions first: under one, Mr. Keller is the minor contributor; under the other, an unknown, unrelated person is. The evidence is approximately a million times more probable if Mr. Keller contributed than if an unknown person did — very strong support for the first proposition over the second. Three limits matter. First, this is a mixture, heat-degraded and interpreted by validated software within its range — not a clean single-source match. Second, it tells you Mr. Keller's DNA is consistent with being present; it does not tell you he is guilty. Third, it says nothing about how or when his DNA reached that can — and because Mr. Keller is a co-owner of this property, an innocent explanation for his DNA on an item stored there exists and must be weighed."

The one sentence the analyst must refuse to say, no matter how the prosecutor phrases the question: "This is Mr. Keller's DNA" (or any equivalent — "it's him," "the probability he's innocent is one in a million," "this proves he did it"). Each converts a strength of support into a certainty or a probability of guilt, transposing the conditional (Chapter 9) and usurping the jury (§30.4).

Exercise 30.33

The honest fire investigator's answer to "Does any single burn indicator prove this fire was deliberately set?":

"No. No individual burn pattern proves arson — flashover can mimic the old so-called 'arson signatures,' and indicators like 'crazed glass' and 'pour patterns' are exactly the discredited folklore that helped convict Cameron Todd Willingham. My conclusion does not rest on any single indicator. It rests on the convergence of a valid origin-and-cause analysis — multiple origins inconsistent with a single accidental source — and laboratory-confirmed ignitable-liquid residue, gasoline, identified by gas chromatography–mass spectrometry, not by a scene investigator's eye." The conclusion rests on (Chapters 22, 23): valid origin-and-cause analysis consistent with the modern fire-science standard plus instrumentally confirmed accelerant. Because it never depended on folklore, conceding that no single indicator proves arson costs the conclusion nothing — it survives the cross.

Exercise 30.35 †

(150–200 words; model.) When an expert tells you "I am giving you a strength of support, not a certainty," they are being more scientific, not less, because real forensic methods measure how strongly evidence favors one explanation over another — they do not deliver certainties, and a method's honesty is measured by whether it reports its own limits and error rate. A witness who offers you certainty is offering you something the science does not contain: no method (apart from the narrow case of a clean single-source DNA profile) can exclude every other person or every other explanation on Earth, and every method has a non-zero error rate. So "100% certain" or "zero error rate" is not a stronger result; it is a false one. The expert who says "strongly supports, not proves" is teaching you what real evidence is — and protecting you against the next witness, in some other case, who offers a confidence the science cannot back. Hearing the difference between strength of support and certainty is the most important thing a juror can learn.

Exercise 30.36 †

(150–200 words; model.) The advocate and the witness have different jobs, and the difference is the whole of it. The attorney is an advocate: retained to argue one side as forcefully as the rules allow, presenting the evidence in the light most favorable to the client. That is their proper role — the adversarial system assigns it to them. The expert is a witness: retained to tell the truth about the science regardless of which side it helps. The attorney may shade and emphasize; the expert may not. What goes wrong when an expert starts arguing the case instead of reporting the evidence is twofold. For the truth: the jury loses its one disinterested source of what the evidence actually shows, and an overstatement enters the record as if it were science. For the side that hired them: a witness who can be shown, on cross, to be advocating rather than reporting is destroyed — every concession looks like a caught partisan, and the testimony's credibility collapses, taking the side's case with it. The expert who becomes an advocate helps no one, least of all the client.

Chapter 31 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 31.1

Cognitive bias is the systematic departure of human judgment from what the evidence dictates, produced by normal mental shortcuts, expectations, and motivation rather than by conscious intent. The word systematic adds direction: bias does not scatter judgments randomly around the truth (the way a slightly miscalibrated scale reads a little high and a little low at random), it pushes them consistently toward the expected answer. That is why it is not the same as random error: random error averages out if you take more measurements, but systematic bias does not, because more measurements made by the same expectant mind are all pushed the same way. You can beat random error with repetition; you cannot beat bias with repetition.

Exercise 31.2 †

Confirmation bias is the tendency to seek, notice, weight, and remember information that confirms an existing belief while discounting or rationalizing information that contradicts it. Contextual bias is the distortion of a judgment by domain-irrelevant information from outside the evidence (the confession, the record, the case theory). The distinction the chapter draws: contextual bias is usually the trigger; confirmation bias is usually the engine. The external context supplies the expectation (it tells the examiner what answer the case "wants"); confirmation bias is the mechanism by which that expectation then warps the handling of the evidence — making agreements salient, discrepancies forgivable, the conclusion self-sealing. In practice this matters because the two have different entry points and different fixes: you block contextual bias structurally by keeping the irrelevant context away from the bench (context management, §31.5); confirmation bias you constrain by locking judgments in writing before the next layer of information can reach them (documentation before comparison, sequential unmasking). You can have confirmation bias without external context (an examiner's own early tentative call can seed it), so naming both lets you diagnose and fix what goes wrong.

Exercise 31.3

Domain-irrelevant information is any information not necessary for the specific technical task an examiner is performing and that could bias the result — as opposed to domain-relevant (task-relevant) information, which the examiner genuinely needs. The counterfactual test: would knowing this fact help me perform the technical comparison if I had no idea who the suspect was — or does it only tell me what answer the case wants? If only the latter, it is domain-irrelevant, and it is contamination. Classifying for a latent-print comparison: - (a) the ridge detail of the latent — domain-relevant (you cannot do the comparison without it). - (b) the suspect's confession — domain-irrelevant (tells you nothing about the ridges; only what answer is wanted). - (c) the substrate the print was lifted from — domain-relevant (it explains expected distortions, a genuine task need). - (d) the suspect's prior record — domain-irrelevant (irrelevant to ridge correspondence; supplies an expectation). - (e) an eyewitness identification — domain-irrelevant (not part of comparing two patterns of ridges).

Exercise 31.5

The bias cascade is the propagation of an expectation or irrelevant context introduced at one point in an investigation through subsequent analyses, verifications, and re-examinations, so that conclusions appearing to independently corroborate one another are in fact one biased judgment amplified. The single most important consequence for interpreting several examiners "agreeing": unanimity is not corroboration, and agreement is not independence. If the later examiners knew the first conclusion (or shared the contaminating context) before forming their own, there were never several independent judgments — there was one judgment, formed under bias, and several echoes of it, no more reliable than its single biased source.

Exercise 31.7

Before the reference sample is revealed, the examiner must analyze the evidence sample on its own and document that interpretation in writing (Stage 1 of sequential unmasking): the features present, their quality, what the sample can and cannot support — with no exemplar in view, no suspect in mind, no case theory. The written, time-stamped form is essential because it cannot be silently revised to fit the exemplar later. If the examiner's reading of the latent shifts after the reference appears (the exact failure the Mayfield review found — the exemplar shaping the reading of the latent), the record shows the shift, making it visible and auditable instead of invisible and deniable. Bias thrives on the freedom to reinterpret without a trace; the written Stage-1 record takes that freedom away.

Exercise 31.9 †

• (a) The domain-irrelevant information present: the assertion "this is the guy," and the fact of the confession — neither has any bearing on whether the two patterns of ridges correspond; both only signal the wanted answer. (The exemplar itself is task-relevant; the story attached to it is the contamination.)
• (b) The specific word the chapter calls "an alarm": "confirm." It has already told the examiner the answer and recast the job from compare two impressions and report what you find to ratify a conclusion already reached — a different and more dangerous cognitive posture.
• (c) Two distinct ways this setup biases the comparison: (1) Contextual bias supplies an expectation — the examiner now approaches an ambiguous latent already "knowing" the answer the case needs, so ambiguous ridge detail is resolved toward the suspect. (2) Confirmation bias steers the handling of evidence — points of agreement feel significant and get tallied, while discrepancies (a non-matching ridge ending) get rationalized as distortion rather than counted as strikes against the conclusion. (A third acceptable answer: the request to confirm induces deference/compliance pressure, biasing beyond pure confirmation.)

Exercise 31.11

Under genuine blind verification, the second examiner independently repeats the analysis without knowing the first examiner's conclusion (or the contaminating case context). The verification is worthless as a check on bias when it is not blind — when the "verifier" is told "Examiner 1 already called this an identification, please confirm," so they bring the same expectation as Examiner 1 plus deference to a respected colleague's stated conclusion. That is not an independent test; it is the first echo of the cascade (this is exactly what happened in Mayfield, §31.4). It becomes genuine corroboration only when the second examiner reaches their conclusion blind to the first examiner's result and free of the domain-irrelevant context that started the cascade — at which point agreement is at last what the cascade only pretended to be.

Exercise 31.12 †

The detective's claim confuses the issue because it treats bias as a failure of effort or integrity that a "true professional" can overcome by trying — and §31.1 shows that is precisely backward. Perception and interpretation are a single process: the brain does not first perceive the evidence neutrally and then, as a separate step, add interpretation; it reaches for available context to resolve ambiguity as it perceives. So by the time the sincere examiner is "being objective" about an ambiguous latent, the case facts have already shaped what they saw — the contamination happens upstream of the will. An examiner's sincerity is not a safeguard for two reasons: first, you cannot consciously "ignore" information that has already structured your perception (you cannot un-see the smudge as a "B" once told it is a letter); second, the belief that one is personally immune is itself the bias blind spot — one of the best-documented biases in psychology — and it is lethal here, because the confident examiner will reject the very safeguards (blinding, sequential unmasking) that would actually protect them. The fix is structural, not attitudinal: withhold the contaminating context, do not exhort people to resist it.

Exercise 31.13

Three-stage sequential unmasking for a firearms/toolmark comparison (questioned bullet vs. test-fired bullet from a suspect's gun): - Stage 1 — analyze the questioned bullet alone, and document it in writing. Characterize the class characteristics (caliber, rifling: lands and grooves, twist) and the quality and individual striation detail that are actually present, before any test-fire is in view and with no suspect, no gun, no case theory. Lock this in writing. - Stage 2 — reveal the reference (the test-fired bullet from the suspect's gun); compare; record the comparison conclusion before any case context. The comparison is made and documented while the examiner is still naïve about what answer the case wants. - Stage 3 — reveal limited task-relevant context only if genuinely needed (e.g., the firearm's condition or a known mechanical issue that bears on expected mark reproducibility). - Most-biasing information that should never reach the bench: that the suspect confessed, the suspect's record, the case theory, that a fingerprint or DNA result already implicates this suspect, and the investigators' confidence — all handled by a case manager, kept away from the comparison. Verify blind: a second examiner repeats Stages 1–2 without being told the first conclusion.

Exercise 31.15

"Hire better people / train them to be objective" is the wrong fix because the problem is structural (contaminating information reaching the analyst's perception), not a deficiency of the individual. The chapter's evidence is decisive on this: the examiners in the worst cases (Mayfield) were among the best in the world, sincere, and following accepted procedure — and they failed anyway. You cannot train away an effect that operates upstream of conscious effort (perception reaches for context automatically), and the bias blind spot means the better and more confident the analyst, the more they will resist the safeguard. A structural solution controls what reaches the analyst, and when: withhold the domain-irrelevant context (context management), fix each judgment in writing before the next layer of information is unmasked (sequential unmasking), and verify blind. The fix engineers the workflow, not the willpower — denying bias its raw material rather than asking people to overpower it.

Exercise 31.16 †

Figure 31.2 claims the Mayfield-style error was a resolution of an ambiguous feature in the direction of the expectation supplied by domain-irrelevant context — an ambiguous ridge region, read by an examiner expecting a match, perceived as a corresponding minutia (and, read with no expectation, at least as readily perceived as distortion or noise); the ridges never changed, only the context did, amplified by non-blind verification (the cascade). It claims the error was not a sign that the examiner was incompetent, dishonest, or careless, and not evidence that the fingerprint method is invalid. The distinction is the whole point of the chapter because the two readings imply opposite fixes: if the error were incompetence or dishonesty, the remedy would be better people or discipline; because it is contextual contamination of an ambiguous comparison by normal cognition, the only thing that works is engineering the workflow (context management, blind verification) so the contaminating context never reaches the analyst. Misdiagnose the cause and you apply a fix that cannot work — which is precisely why most labs, blaming individuals rather than the workflow, still have not adopted the structural safeguard (§31.6).

Exercise 31.17

Trace "the suspect confessed" through Figure 31.1. Entry point: it enters with the evidence — the investigator submits not a bare latent and exemplar but a latent, an exemplar, and the story "this is our man, he confessed." Three places it contaminates: (1) Examiner 1, who now compares the ambiguous latent already knowing the wanted answer, so ambiguities resolve toward the suspect; (2) Examiner 2 ("verification"), who is told an identification was already declared by a respected colleague — the confession-fed expectation plus deference produces agreement that is recorded as independent corroboration but is not; (3) the other examinations (DNA, toolmarks), to which the "confirmed" match travels sideways ("the print already matched, so this hair probably will too"), biasing them in turn. Where the cascade feeds back: the "confirmed" match flows into the investigation — it hardens the case theory, discourages detectives from looking at other suspects, and justifies a hard interrogation of the "confirmed" suspect. Why that feedback is hard to unwind: each of those steps generates new context (a pressured confession, a witness firming up an ID) that flows back to the lab as further "corroboration" for the next analysis, so the system bootstraps itself from one ambiguous latent to an apparently overwhelming, mutually reinforcing case — with no individual ever doing anything they recognize as wrong, which is why no amount of individual diligence can stop it.

Exercise 31.19

Bias(es) at work: the reviewer is subject to contextual bias (the irrelevant fact of the prior conclusion sets the expectation) operating through confirmation bias and deference to authority (a respected colleague's stated identification), with the "confirm" alarm explicitly sounded — the instruction recasts review as ratification. This is a non-blind verification, i.e., part of the cascade, not a check on it. Rewrite for genuine blind verification: "Please independently examine this evidence sample and the reference and report your own conclusion. Do not be told, and do not infer, what conclusion any prior examiner reached; analyze and document the evidence sample first, then compare." (The defining change: the verifier must not know the first examiner's result before forming their own.)

Exercise 31.20 †

How the claim can be false even though peer review and verification are genuinely happening: because the chapter's distinction is not whether a second person reviews, but whether that review is blind. A laboratory can require "peer review," "technical review," and "verification" of every conclusion while leaving every reviewer fully informed of the conclusion they are checking — in which case the review is part of the bias cascade, not a cure for it. It manufactures the appearance of corroboration while adding none of the substance, and may make matters worse by lending false confidence. So a lab can adopt the entire vocabulary of quality and still have institutionalized bias rather than addressed it. The one-word property that distinguishes a real fix from a comforting ritual: blind (the verification must be blind). The right question is never "do you verify?" but "is the verification blind?"

Exercise 31.21

The overstatement: the experiments did not show that examiners "get the answer wrong most of the time." What they actually demonstrated, under controlled conditions, is that domain-irrelevant context can change the conclusions of qualified experts on the very same physical evidence — the conclusions that changed were a portion of cases, not all of them, and the finding does not rest on any single dramatic statistic (which is why this book invents none). What they did not demonstrate: that fingerprint examiners are usually wrong, that most comparisons are erroneous, or that the method is junk science (it is foundationally valid; the experiments concern the human judgment, not the friction-ridge science). Honest rewrite: "Controlled experiments have shown that domain-irrelevant context can change qualified examiners' conclusions on identical evidence — establishing that bias is a real, measurable threat to forensic comparison, and that safeguards like blinding are needed. They do not show that examiners are usually wrong."

Exercise 31.23

The trainer's sentence — "bias is only a risk for sloppy or dishonest analysts; if you do careful work, you're fine" — contains several errors against §31.1–31.2: 1. It locates bias in effort and integrity. Bias is a property of normal cognition operating on ambiguous data; it is not the residue of sloppiness or dishonesty, and the most careful, sincere examiners are fully subject to it (Mayfield). 2. It assumes care and objectivity are the same thing. Care cannot interrupt a process that happens upstream of effort — perception reaches for context automatically, so an examiner can be meticulous and still have an ambiguous feature shaped by an expectation before any "careful work" begins. 3. It expresses the bias blind spot. "If you do careful work, you're fine" is the very belief in one's own immunity that the chapter calls lethal, because it leads the examiner to reject the safeguards (blinding, sequential unmasking) that would actually protect them. 4. It implies a fix that cannot work ("just be careful"): the real fix is structural (control the context), not a matter of trying harder.

Exercise 31.24 †

The honest tension, named with three real obstacles from §31.6 and at least one workable partial safeguard: - Obstacle 1 — the (institutional) bias blind spot. Examiners certain of their own objectivity experience context management as an insult, so the lab resists it culturally. - Obstacle 2 — structural/resource constraints. A small lab is backlogged and under-resourced; sequential unmasking adds documented steps and time, and in a two-person lab there may be no second examiner available to verify blind. - Obstacle 3 — lack of independence (police/prosecution adjacency). Embedded in a police department whose detectives expect the lab to "help make the case," the lab faces pressure that runs precisely opposite to blinding analysts from the case. - At least one partial safeguard achievable under these constraints: even without a second free examiner, the lab can require documentation of the evidence-sample analysis in writing before the reference is revealed (Stage 1 of sequential unmasking) for every comparison — this single, nearly free step prevents the most damaging failure (the exemplar reshaping the reading of the evidence) and makes any drift auditable. Other acceptable answers: designate whoever handles intake as a case manager who strips the confession/record/theory from the file before it reaches the bench; send contested comparisons to the state lab for blind verification; or have the examiner record their conclusion before learning the case facts even when a full blinded workflow is impossible.

Exercise 31.25

The honest thing to do is disclose it — record, in the case file and in any report or testimony, exactly what context you were exposed to and when (before or after you formed your conclusion). You should not pretend the exposure did not happen, and you should not quietly "try to ignore it," because by the time you are ignoring it, it may already have shaped what you saw. Concealing the exposure is worse than disclosing it for two reasons: (1) it removes the court's and the reviewer's ability to weight your conclusion appropriately — a conclusion formed after exposure to a confession is not the same evidentiary object as a blind one, and hiding the exposure passes off the former as the latter; (2) it converts an honest limitation into a misrepresentation, breaching the duty to the court (Chapter 30) and making the contamination invisible and therefore un-auditable. Transparency lets the system discount appropriately; concealment defrauds it.

Exercise 31.27

The chapter's voice would say no — it is not ethical to testify to a confident "identification" reached through a workflow you know was not blinded, without disclosing that fact. The condition of the judgment is a material fact about the testimony: a contaminated conclusion sounds exactly as confident as a clean one from the stand, so failing to disclose that the workflow was not blinded lets the jury treat a possibly bias-amplified conclusion as if it were independent and clean — exactly the error (unanimity-as-corroboration; confidence-as-accuracy) the chapter warns against. The expert's duty to the court (Chapter 30) is to communicate not just the conclusion but its limits and the conditions under which it was formed; the absence of blinding is one of those conditions. The honest expert states the conclusion and discloses that the analysis was not blind to the case context (and, on cross, answers the §31.4 questions truthfully). Withholding it is the "hired gun" posture — advocacy disguised as testimony — not honest science.

Exercise 31.28 †

Ranking by how much each needs context management (most → least), with justification: 1. Bite-mark comparison (Ch.16) — needs it MOST. A wholly subjective comparison of ambiguous patterns with no validated objective basis; the path from expectation to "match" is at its widest (and the method is independently discredited, compounding the danger). 2. Latent-print comparison (Ch.14) — high. A subjective comparison of often-ambiguous, partial impressions by judgment; the canonical site of the cascade (Mayfield). Foundationally valid, but the judgment is highly steerable. 3. Bloodstain-pattern interpretation (Ch.10) — moderate-to-high. Reconstructive interpretation of ambiguous patterns by judgment; substantial room for expectation to drive the reading. 4. Single-source DNA profile (Ch.7) — low. Largely instrumental with a quantified, well-characterized output; little interpretive latitude for expectation to exploit (its mixture interpretation, Chs. 8–9, would rank higher — that is the relevant contrast). 5. GC-MS drug identification (Ch.23) — needs it LEAST. A purely instrumental output; the machine produces a spectrum that is what it is regardless of the analyst's hopes. How this "bias-vulnerability" ranking relates to but is not identical to the NAS/PCAST validity-spectrum ranking: both tend to place the subjective pattern methods low and DNA/instrumental high, so they correlate. But they measure different axes. The validity spectrum asks is the method's core claim foundationally sound and tested (known error rate)?; the bias-vulnerability ranking asks how much is the result a subjective call on ambiguous input that an expectation could steer? They can come apart: latent prints are foundationally valid (high on validity) yet highly bias-vulnerable (high on this ranking), which is exactly the chapter's point — a valid method in a contaminated workflow can still produce a confident false positive. The property doing the work here is subjectivity, not validity.

Exercise 31.29

(Model answer — fingerprints chosen.) Chapter 14 taught fingerprint comparison as a foundationally valid discipline with a real, non-zero error rate, and it flagged Mayfield as proof that even certainty can be wrong. What that chapter could not fully convey at the time — because the reader did not yet have the vocabulary — is the mechanism and its reach: that the error was not a one-off misreading but a bias cascade, in which the exemplar reshaped the reading of the ambiguous latent (confirmation bias), a non-blind verification converted a single judgment into the appearance of independent corroboration, and the contamination reached even the examiner appointed for the defense, while confidence escalated to "100 percent" as agreements stacked. The new vulnerability exposed is therefore systemic, not individual: a valid method, in a contaminated workflow, can manufacture a unanimous, confident, escalating falsehood — so the reader must now reread the fingerprint chapter (and every comparison chapter) with one added question: what did the examiner know, and when? (Equally good answers exist for toolmarks, bloodstain analysis, or DNA mixtures — in each case the chapter adds the steerability of the judgment the earlier chapter treated as a reading.)

Exercise 31.30

(Model answer.) This chapter advances Theme 3 (cognitive bias is the chief threat) and Theme 2 (the validity spectrum), and it is the home chapter for Theme 3. It advances Theme 3 by doing what no earlier chapter could: naming the mechanism (perception-as-interpretation; confirmation vs. contextual bias; the cascade), measuring it (the Dror experiments, attributed honestly), and showing it lethal (Mayfield), then building the structural fix (context management, blind analysis, sequential unmasking). "Home chapter" is the right description because every other chapter only points at bias in passing, while this one supplies the full apparatus the rest of the book borrows — and because it explicitly reinterprets every method already taught by adding the question "what did the analyst know, and when?" It advances Theme 2 by insisting that foundational validity is necessary but not sufficient: a valid method (latent prints) in a contaminated workflow can still produce a confident false positive, so the bias-vulnerability of a result is a second axis layered on the validity spectrum — the surest conclusions are the ones an expectation could not steer. (A defensible alternative pairs Theme 3 with Theme 4, the CSI effect: a contaminated conclusion sounds exactly as confident as a clean one from the stand, so juror trust must shift from "is the expert sure?" to "could the expert have known the wanted answer before deciding?")

Exercise 31.31 †

Cold-case workbook entry (model) — a caution on the old evidence, not a new piece of evidence: - (a) The contaminating expectation: the "accidental fire" frame recorded by first responders on the morning of 18 October — an expectation set before any analysis — together with the later emergence of a favored suspect that the early analysts came to know. Both are domain-irrelevant context that had the power to steer the early work. - (b) How it could have propagated (the cascade): the frame caused the scene to be processed as an accident (Chapter 2) — documented less rigorously, searched less suspiciously, preserved less carefully (a contamination of the handling, driven by an expectation about the conclusion); and as a favored suspect emerged, each early examiner risked being told "the others already point to him," so an early non-blind agreement could have been recorded as independent corroboration when it was an echo — the cascade as a live hazard in our own file. - (c) Why it excludes and includes no one: the finding is about our confidence in the early work, not about any person — it identifies a risk of contamination, which neither implicates nor clears Keller, Salas, Dana, or Renner. What changes is how much weight the early conclusions can bear, not who did it. - (d) Which findings to trust most, and why: the autopsy ("no soot in the airways → dead before the fire"; the blunt-force skull fracture, Chapter 11) and the instrumental chemistry (GC-MS confirmation of gasoline, Chapter 23) — because no expectation could have steered them. They are instrumental or physiological facts whose outputs do not depend on what the analyst was told; the accidental-fire frame is powerless against soot in airways or a mass spectrum. Trust most the findings a wanted answer could not have shaped.

Exercise 31.33

(Model answer.) A three-stage sequential-unmasking protocol for the cold case's gas-can latent (the workflow that should produce the good "inconclusive" outcome of Chapter 14): - Stage 1 — the bench examiner analyzes the latent alone and documents it in writing. Given to the examiner: only the latent itself and, if genuinely needed, task-relevant facts about the substrate and development (it was a partial print on a gas-can handle, recovered from a fire scene — so expected heat/handling distortion can be anticipated). The examiner records the quality and the features actually present, and whether the latent is even of value, before any exemplar is in view. - Stage 2 — reveal the exemplar; compare; record the conclusion before any case context. The reference print (e.g., the AFIS candidate's exemplar) is shown, the comparison is made, and the conclusion — here, inconclusive — is documented while the examiner is still naïve about the case theory. - Stage 3 — reveal limited task-relevant context only if needed; verify blind (a second examiner repeats Stages 1–2 without knowing the first result). - What the case manager withholds from the bench: that investigators favor this AFIS candidate, the suspect's identity and any record, the "accidental fire" theory, any confession (e.g., Renner's), and all other forensic results. Those are domain-irrelevant to comparing two patterns of ridges and are exactly the inputs that, in Mayfield, turned a poor latent into a "100 percent" match. By withholding them, the protocol lets the examiner reach the honest, bias-resistant inconclusive — which is precisely the courageous answer the chapter celebrates.

Exercise 31.35 †

The strongest honest case for sequential unmasking, to a skeptical veteran (model, 150–200 words):

"You've testified four hundred times, and I believe every one of those calls was made in good faith. But sincerity was never the question. Perception and interpretation are a single process: on an ambiguous latent, the case facts you happen to know shape what you see before any conscious 'objectivity' kicks in — so by the time you're being careful, the contamination is already upstream. That isn't an insult to your skill; it's true of every human nervous system, including the best examiners in the world, who in Mayfield reached a unanimous '100 percent' identification that was 100 percent wrong. The conviction that you are the exception is itself the best-documented bias there is.

Here's what working blind gives you, not just what it costs. It makes your conclusion unattackable on the one cross-examination question that unmakes a contaminated expert — 'what did you know about the suspect before you decided?' A documented, blind analysis lets you answer: 'Nothing. I read the ridges, then the case.' That is the strongest position a witness can hold. Blinding doesn't doubt your judgment. It protects it — and proves it."

Chapter 32 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 32.1

Eyewitness identification is a witness's report that a particular person is the one they saw commit a crime (or be present at it). The single fact the DNA exonerations revealed about it: mistaken eyewitness identification was present in a large majority of the wrongful convictions later overturned by DNA — more than any single discredited forensic method, more than false confessions, more than informants. The most trusted evidence in the courtroom is, measured against DNA ground truth, among the least reliable.

Exercise 32.3 †

The video-recorder model holds that the eye records the event, the brain stores the recording intact, and remembering is pressing "play" on a faithful (if faded) copy. The reconstructive model — the correct one — holds that we store only fragments (a few details, a gist, an emotional tone) and rebuild the event each time we remember, filling gaps with inference, expectation, and information acquired after the event. The reconstructive model is right.

Why the wrong model makes jurors over-trust a confident witness: on the video-recorder model, confidence is a reasonable proxy for the clarity of the "recording," so "certain" implies "accurate." That intuition ignores the retrieval stage — that memory is rebuilt (and can be contaminated and inflated) at each recall — so the jury credits a confident trial identification as if it were a clear original recording, when it is in fact the latest reconstruction, polished by feedback and retelling.

Exercise 32.5

A lineup is a formal identification procedure in which a witness views the suspect together with several known-innocent fillers (live, or as a photo array) and is asked whether they recognize the perpetrator. A showup presents a single suspect — no fillers — to the witness for a yes/no identification, typically shortly after a crime and near the scene.

A showup is inherently more suggestive because there is only one choice and the witness knows the police have detained this particular person: the procedure itself implies "we think this is the one." There are no fillers to test whether the witness can actually pick the perpetrator out, so a showup measures little about the memory and strongly invites confirmation.

Exercise 32.7

Double-blind administration is a lineup procedure in which neither the witness nor the administrator knows which member is the suspect. The witness's blindness is automatic (that is the point of a lineup); the administrator's blindness is the reform.

The channel of contamination it severs: the path by which the administrator's knowledge of the right answer reaches the witness — steering (tone, a glance, a pause, "look carefully at that one"), confirming feedback after a choice, and any unconscious cueing. An administrator who does not know who the suspect is cannot steer toward the suspect or confirm the "right" pick, consciously or not. (This is the Chapter 31 blind-testing logic applied to a witness's memory.)

Exercise 32.9

What the chapter actually says about the confidence–accuracy relationship (not the folk version): confidence is not a general guide to accuracy. The single condition under which confidence carries real information is a pristine confidence statement — taken immediately, at the first identification, from a fair (double-blind, fairly composed) lineup, before any feedback. Under those clean conditions, high confidence is meaningfully associated with higher accuracy (and low with lower) — though still bounded by the witnessing conditions. After that moment, every contamination (feedback, repeated retrieval, time) decouples confidence from accuracy, so trial confidence is close to worthless.

Exercise 32.11 †

Estimator variables present (the witnessing conditions, §32.2): (1) brief exposure — about two seconds; (2) darkness / poor lighting — at night; (3) distance — across a parking lot; (4) weapon focus / high stress — a gun pointed at the witness, which narrows attention onto the weapon and away from the face and floods the witness with fear that impairs encoding; (5) the cross-race effect — the perpetrator was of a different race than the witness.

Why no lineup procedure, however perfect, can rescue this identification: these are estimator variables, and they operate at encoding — the moment of the event. They set a ceiling on how much reliable information ever entered memory. A perfect lineup is a system-variable fix: it avoids adding error, but it cannot subtract error already imposed by the conditions, and it cannot retrieve information that was never encoded in the first place. You cannot recover what was never stored; here, very little was. (The conditions also tend to raise the witness's felt certainty even as they destroy accuracy — the mismatch at the heart of §32.5.)

Exercise 32.13

The post-identification feedback effect: when an administrator confirms a witness's choice — "good, that's our guy," "you picked the one we suspected," or even a warm "thank you, that's very helpful" — the witness's confidence in the identification surges, and so do their later retrospective reports of how good a view they had, how much attention they paid, and how clearly they saw the face. A tentative "maybe number 3" becomes, over time, "absolutely certain."

Why this makes trial confidence misleading: the trial certainty the jury finds overwhelming was manufactured after the identification, by feedback and the case's history — not by any improvement in the underlying memory. The confidence measures the history of the case, not the quality of the original encoding, so it is a poor guide to accuracy precisely when the jury most relies on it.

Exercise 32.14 †

A simultaneous lineup (all members shown at once) invites a relative judgment because the witness can compare the members to one another and pick whoever looks most like their memory of the perpetrator — relative to the others present. With every face visible at once, "which of these is closest?" is the natural strategy.

Why this is dangerous specifically when the lineup contains an innocent suspect and not the true perpetrator: someone always looks most like the memory, even when the actual perpetrator is absent. If the true perpetrator is not in the array but an innocent suspect who happens to resemble the witness's memory is, the relative-judgment strategy steers the witness to that innocent person — the closest available face — and produces a confident mistaken identification. (A sequential lineup is designed to push toward an absolute judgment — each face vs. the memory — to blunt this, though the format matters less than the other safeguards; §32.4.)

Exercise 32.15

Recording confidence immediately and verbatim at the first identification is essential because that is the only measurement of confidence with evidentiary value: it captures the witness's certainty before feedback, repeated retrieval, and the case's history can inflate it (the post-identification feedback effect, §32.3/32.5). A pristine first statement — e.g., "maybe number 4, I only saw him for a second" — is the honest index of how strong the memory actually was.

What is lost if the only confidence on record is the trial statement: the informative number is gone, and what remains ("absolutely no doubt") is the inflated figure, decoupled from accuracy. A court can then no longer tell whether the witness was tentative or certain at the moment it mattered — and an undocumented procedure cannot be audited at all, which is itself an evidentiary problem (see Exercise 32.21).

Exercise 32.17

"Fillers should match the witness's description of the perpetrator" is more defensible than "fillers should resemble the suspect" because the lineup is a test of the witness's memory, and the proper benchmark is what the witness reported remembering. If fillers match the description, every member is a plausible candidate against the memory, and a witness who picks the suspect has demonstrated something.

If fillers are instead chosen to resemble the suspect, you can inadvertently build an array around the suspect's specific features (including features the witness never mentioned), which can make the suspect either blend in artificially or — more often in practice — stand out on the very features the witness actually described. Matching the description keeps the test honest and prevents the suspect from being the obvious answer to "which one fits what I told you?"

Exercise 32.19 †

Figure 32.1 — the four suggestive system variables in Version A and the Version B safeguard that neutralizes each:

Note that safeguard 1 (blind administration) does double duty: it structurally prevents both steering (3) and feedback (4), because an administrator who doesn't know the suspect cannot do either.

Exercise 32.21

The absence of documentation is itself an evidentiary problem because the one measurement that carries information — the witness's confidence and exact words at the first identification, before feedback (§32.5) — was never preserved, and the procedure can no longer be audited. Specifically, you can no longer assess: whether a proper instruction ("may or may not be present") was given; whether the administrator was blind or steered the witness; how the array was composed (fair fillers, one suspect); whether confirming feedback was given; and, above all, how confident and how hedged the witness actually was at first. All that remains is the inflated trial confidence, which is close to worthless. The missing record converts an answerable question ("was this a clean test?") into an unanswerable one — and the burden of that ambiguity should weigh against the reliability of the identification.

Exercise 32.23 †

The prosecutor's argument — "the witness was certain on the stand, so you can be certain too" — commits the chapter's central error: it treats trial confidence as a guide to accuracy. Trial confidence has been shaped by feedback, repeated retrieval, and the case's history; it is inflated by the reconstructive process (§32.1) and is decoupled from accuracy (§32.5). The mistaken-identification exonerations are full of witnesses who were utterly certain and flatly wrong, so courtroom certainty is close to evidentially worthless. (The argument also implicitly assumes confidence = honesty = accuracy; the witness can be perfectly honest and confidently mistaken.)

Honest rewrite: "The witness is certain today, and I have no doubt she is sincere. But certainty expressed at trial is not a reliable guide to accuracy — it grows with feedback and retelling regardless of whether the memory is correct. The measurement that would actually inform you is how confident she was at the first identification, recorded before anyone told her anything — and you should weigh the identification by the conditions under which she saw the perpetrator, not by how sure she is now."

Exercise 32.25

The detective's reasoning — "I've never led a witness, so blind administration is unnecessary" — does not establish that his lineups are unbiased because the steering the reform guards against happens unconsciously (§32.3, §32.6; the Chapter 31 point about the objective-examiner myth). A sincere, well-intentioned administrator cannot detect the subtle tone, glance, pause, or emphasis by which he cues a witness toward the suspect, precisely because it is below his awareness — so his confidence that he "never leads" is exactly the confidence the research says is unreliable. As in Chapter 31, good intentions are not a safeguard against a bias one cannot perceive; only the procedure (a blind administrator who cannot know which member is the suspect) can neutralize it. His certainty is evidence of nothing but his sincerity.

Exercise 32.27 †

Calling a sincerely mistaken witness a "second victim" means that the witness, too, has been harmed by the crime and the flawed process: sincerely trying to do justice, they were subjected to suggestive procedures they did not design and could not detect, and are left — when DNA finally reveals the error — to carry the knowledge that their certainty helped imprison an innocent person (the Cotton/Thompson case, §32.6, is the model).

Why treating eyewitness error as a moral failing of the witness misunderstands it: the error is not dishonesty or carelessness. It is the predictable output of reconstructive memory (§32.1), degraded by estimator variables the witness cannot assess (§32.2) and inflated by system variables the witness cannot see (§32.3). Jennifer Thompson did everything a witness is supposed to do and was still wrong.

What the reframing implies about reform: reform should be directed at procedure and understanding — the controllable system variables (blind administration, proper instructions, fair fillers, immediate confidence statements) and jury education — not at exhorting witnesses to "be more careful" or blaming them after the fact. You cannot fix the witness; you can fix the procedure.

Exercise 32.29

Six best-practice safeguards (from §32.4/§32.6) and the specific failure each prevents:

1. Fair lineup composition — fillers match the witness's description, the suspect doesn't stand out, only one suspect per lineup. Prevents: the "multiple-choice with one obvious answer" failure, where a witness picks the standout suspect and demonstrates nothing about memory.
2. Double-blind (or blinded) administration — the administrator doesn't know who the suspect is. Prevents: conscious or unconscious steering and confirming feedback.
3. Proper instruction — "the perpetrator may or may not be present; you need not choose anyone." Prevents: the "pick the closest" relative-judgment trap that selects an innocent look-alike.
4. Immediate, verbatim confidence statement at the first identification. Prevents: loss of the one informative confidence measurement (and provides a baseline against later inflation).
5. No confirming feedback. Prevents: the post-identification feedback effect — the inflation of confidence and of retrospective reports of view quality.
6. Complete documentation (ideally video) of the whole procedure. Prevents: an unauditable record — so a court can verify the safeguards actually held. (Also worth mandating: limits on suggestive showups and on repeated viewings of the same suspect, which breed unconscious transference.)

Exercise 32.31 †

Eyewitness identification advances (at least) two of the book's four themes:

• Theme 1 — Exclusion over proof. A confident identification can only ever include a person ("this looks like the one I saw"); inclusion is the weak direction and cannot establish that this person and no other is the source. The Cotton/Thompson and Marvin Anderson cases were overturned by DNA exclusions — the strong direction — a single mismatch defeating an identification a jury found overwhelming (Chapter 1, §1.6). Eyewitness evidence constrains and narrows; it does not prove.
• Theme 2 — The validity spectrum. The core claim — "this specific person, to the exclusion of others" — sits low on the NAS/PCAST-style spectrum, alongside the confident pattern-comparison disciplines and well below DNA, for the same structural reason: confident individualization asserted without a measured basis for the confidence. Human face memory is genuinely useful under good conditions, but the courtroom claim built on it outruns what the faculty can deliver.

(Also defensible: Theme 3 — cognitive bias, since double-blind lineup administration is the Chapter 31 blind-testing reform applied to a witness's memory; and Theme 4 — the CSI effect cutting both ways, since juries over-trust the confident in-court identification that the science says is least reliable. Naming any two, with the mechanism, earns full credit.)

Exercise 32.33

A single DNA mismatch could overturn the Cotton identification because of the evidentiary asymmetry (Chapter 1, §1.6): exclusion is the strong direction. If the perpetrator's biological evidence does not match the convicted man, then — barring laboratory error — he is not the source, and that conclusion is near-categorical; it takes only one genuine inconsistency to refute. Inclusion is the weak direction: a confident identification (or even a DNA inclusion) says "consistent with," and consistency is shared by others, so it narrows the field without ever proving a unique source. The Cotton/Thompson case is this asymmetry in action — the most persuasive evidence (a certain witness) included Cotton, and was overturned by the most reliable evidence (a DNA exclusion) that no appeal had managed.

Exercise 32.34 †

Cold-case evidence-log entry (model) — the neighbor's "tall stranger's truck":

• (a) Honest status of the lead: an eyewitness lead, discounted on witnessing-conditions grounds — logged in the workbook (Appendix I), not erased, but explicitly down-weighted and not used to include or exclude anyone. No person of interest is implicated or cleared by it.
• (b) At least three estimator variables that degrade it (§32.2): (1) darkness — it was night; (2) distance — the observation was from roughly a quarter-mile away, across that distance and the dark, yielding at most a silhouette; (3) brief exposure — the truck and driver were glimpsed in passing, not studied; (4) hindsight stress / post-event coloring — the account was given after the neighbor learned a neighbor had died violently nearby, which lends a remembered fragment a vividness and significance it may not have carried at the time. (Any three.)
• (c) Why "a tall stranger's truck" feels like specific evidence but contains almost none: at that distance and in that light, "tall" is an impression of a silhouette, not a measured feature; "stranger" is merely the absence of recognition — which a poorly seen familiar person would also produce; and "a truck" is a class of vehicle, not an identified one (no plate, make, or distinguishing mark). The phrase sounds like a description but encodes essentially no identifying information.
• (d) Why building (or excluding) a theory on it would repeat the Cotton/Thompson error in miniature: it would treat confident testimony produced under poor witnessing conditions as if it were reliable identification — exactly the mistake that convicted Ronald Cotton. To go hunting for "a tall stranger," or to clear a suspect because he is not especially tall, would let a confident-but-uninformative memory drive the investigation, when the conditions cap its reliability so severely that it can identify, and exclude, no one.

Exercise 32.35

Unlike the system variables of a lineup (which are controllable choices made after the event — composition, instruction, administrator, feedback — and so can be done well next time), the neighbor's observation is governed by estimator variables: the night, the quarter-mile distance, the brevity, and the hindsight coloring were fixed at the moment she saw the truck. They operate at encoding; the information either entered her memory then or it did not, and nothing done afterward can re-run the event under better conditions or add detail that was never stored.

What a careful re-interview could recover: a more complete and verbatim account of what she actually remembers; whether any detail is original versus acquired afterward (from rumor, news, or leading questions); and an honest statement of her certainty and the conditions — useful for weighting the lead. What it could not do: improve the underlying perception, turn "tall silhouette" into a reliable physical description, or convert the account into identification evidence. At most a re-interview clarifies how little the memory can bear, and guards against contamination adding false detail; it cannot manufacture reliability the conditions never permitted.

Exercise 32.37

(Model short answer, ~180 words.) A confident, sincere, weeping witness who says "I'll never forget his face" can be completely wrong because human memory does not work the way it feels like it works. It is not a recording played back; it is a reconstruction, rebuilt each time from fragments and filled in with expectation and with everything that happened after the crime — including the suggestions of a lineup and the reassurance that she "picked the right one." A face glimpsed briefly, in the dark, under terror, perhaps across racial lines, was barely encoded to begin with; then, at each retelling, the memory grows more vivid and more certain whether or not it is accurate. So the certainty you see is real — and it is generated by the same process whether the memory is true or false. This is not a claim that she is lying. She is doing her honest best; she simply cannot tell, from the inside, which parts of a vivid memory are original and which her own mind supplied later. Sincerity and certainty are not evidence of accuracy.

Exercise 32.38 †

(Model short answer, ~190 words.) The naive view — "confidence proves accuracy" — is wrong because the confidence a jury hears is trial confidence, and trial confidence has been inflated by confirming feedback, repeated retrieval, and the case's history. It measures what happened to the witness after the identification, not the quality of the original memory, which is why so many utterly certain witnesses in the DNA exonerations were flatly wrong. But the overcorrection — "confidence means nothing, ever" — is also wrong. When confidence is captured immediately, verbatim, at the first identification, from a fair (double-blind) lineup, before any feedback, a witness's high confidence is meaningfully associated with higher accuracy (though still bounded by how good the witnessing conditions were). What reconciles the two: confidence and accuracy are linked only at the instant of a clean first identification and are decoupled by every later contamination. The relationship is real at the start and corrupted by the time of trial. So the discipline is to trust the first, clean confidence statement and distrust the courtroom one — which is exactly why best practice records confidence immediately and before any feedback.

Exercise 32.39

(Model short answer, ~190 words.) The Cotton/Thompson case illustrates every major section at once. Reconstruction (§32.1): by a later proceeding Thompson failed to recognize the real perpetrator and stayed certain it was Cotton — her memory had been overwritten, substituting the innocent face she kept being shown for the guilty one. Estimator variables (§32.2): a nighttime assault, extreme stress, and a cross-race identification — three powerful accuracy-destroyers stacked. System variables (§32.3): a photo array followed by a live lineup meant repeated viewing of the same face, breeding false familiarity, with none of today's safeguards. Confidence–accuracy (§32.5): Thompson was completely certain and completely wrong, and her certainty grew as the case proceeded. Reform (§32.6): DNA excluded Cotton and identified Bobby Poole. Why the friendship matters to the reform argument: by becoming advocates together, the wrongly accusing witness and the wrongly convicted man embodied the chapter's thesis — don't blame the witness; fix the procedure — and made it persuasive to police and legislators in a way a journal article never could, contributing to real adoption of double-blind lineups, better instructions, and recorded confidence statements.

Chapter 33 — Worked Solutions (daggered † and odd-numbered exercises)

Solutions to the daggered exercises (2, 7, 11, 15, 19, 24, 28, 32, 35, 39) and all odd-numbered exercises. Model answers; students may phrase differently as long as the mechanism and the evidence-strength honesty are right.

Exercise 33.1

A false confession is an admission of guilt, made to authorities, for a crime the person did not in fact commit (a full narrative account or a briefer acceptance of responsibility). Among the DNA exonerations, a false confession or admission was a contributing factor in roughly a quarter to a third of the cases — and this figure is best understood as a floor, since it counts only the narrow set of crimes that happened to leave testable DNA.

Exercise 33.2 †

• (a) Goal. An interrogation aims to obtain a confession from a person already presumed guilty; an interview aims to gather information with an open mind.
• (b) Presumption. An interrogation presumes guilt (the decision "this is the person" has already been made); an interview carries no presumption yet and tries to learn what happened.
• (c) Talking. In an interview the subject talks most (uninterrupted account); in an interrogation the interrogator talks most (asserting the conclusion, refusing denials).
• Why only one can find the wrong person: the interview's open stance lets a suspect's account change the investigator's mind — disconfirming evidence is treated as information. The guilt-presumptive interrogation reinterprets every denial as "resistance to be overcome," so it has no mechanism for recognizing innocence; it can only confirm the conclusion it began with.

Exercise 33.3

The three Reid steps emphasized in §33.2, and how each can move an innocent person toward confessing: 1. Confrontation — the interrogator asserts, with total confidence, that guilt is already established and only why remains. For an innocent person, this signals the system has decided and resistance may be futile (perceived hopelessness). 2. Refusal of denials — the interrogator interrupts and overrides "I didn't do it." For an innocent person this removes their one true statement from the table, teaching them the truth is not an available move; what's left is the interrogator's preferred story. 3. Minimization / themes — a morally softened version ("it got out of hand, didn't it?") makes saying "yes" feel survivable and carries an implied bargain (confess to this and things go better). An innocent person, squeezed, may take the offered exit.

Exercise 33.5

A voluntary false confession is one given without interrogation pressure — the person comes forward on their own initiative. Two motives (any two): a desire for notoriety/attention (high-profile crimes attract serial confessors); a wish to protect the real perpetrator (e.g., a parent shielding a child); an inability to distinguish fantasy from reality driven by mental illness; or a desire to expiate guilt over something unrelated. It is the least relevant to interrogation reform because the interrogation did not cause it — fixing how police question suspects cannot prevent a confession the police did nothing to produce. (It remains a reminder that the bare fact of a confession, without corroboration, proves nothing — see the Lindbergh example.)

Exercise 33.7 †

Confession contamination is the disclosure of nonpublic crime-scene facts to a suspect during interrogation, such that the suspect's later "knowledge" of those facts reflects the interrogation rather than the crime (it can be deliberate or careless — e.g., leading questions, visible photos, correcting "wrong" guesses). Matching "guilty-knowledge" details do not authenticate a confession by themselves because the same fact can reach the confession by two indistinguishable paths — CRIME → SUSPECT (genuine knowledge) or POLICE → SUSPECT (contamination) — so a detail the police already knew and could have conveyed proves nothing about whether the suspect was there.

Exercise 33.9

The constitutional test for admitting a confession is voluntariness under the totality of the circumstances — whether the confession was the product of a rational, free will rather than coercion. A confession can pass that test and still be false because voluntariness is a legal question, not an accuracy question: a confession can be obtained with no beatings, the Miranda warnings read, the suspect "agreeing" to talk — and still be produced by exhaustion, deception, and minimization acting on a vulnerable person. Legality is not truth.

Exercise 33.11 †

This is a coerced-compliant false confession. Its defining internal contradiction is that the suspect says "I did it" while privately knowing they did not — they do not come to believe they are guilty; they consciously comply with a demand to escape the interrogation (or gain a benefit), expecting that the truth will reassert itself later. (The quoted reasoning — "I just wanted it to stop, and I figured I'd straighten it out later" — is the signature of the type.)

Exercise 33.13

This is a coerced-internalized false confession: the suspect comes genuinely (if temporarily) to believe they committed the crime, persuaded by relentless, confident interrogation and fabricated evidence that they did it and simply blacked out / repressed it. The Chapter 32 idea that makes it possible is that memory is reconstruction, not a fixed recording — under sufficient pressure and suggestion, the same malleability that distorts an eyewitness's memory of a stranger's face can distort a suspect's memory of their own actions.

Exercise 33.15 †

Dispositional risk factors (who the suspect is): youth (the single highest-risk group); intellectual disability / cognitive impairment; mental illness and acute psychological state (grief, fear, distress). Situational risk factors (what the interrogation does): interrogation length; sleep deprivation / physiological depletion; deception and the false-evidence ploy; minimization and implied leniency. Why the combination is far more dangerous than any one alone: the factors compound. A frightened juvenile (already more suggestible, more deferential, more focused on immediate escape) who is also sleep-deprived (more suggestible, more impulsive) and subjected to the false-evidence ploy (told "proof" he cannot square with his own memory) and minimization (an "understandable" exit) is not facing a fair test of guilt but a near-optimal machine for producing a false confession: each factor lowers resistance and amplifies the others, so the combined risk is multiplicative, not additive. (This stacked profile is, very nearly, the Renner file.)

Exercise 33.17

The phrase is honest because each component refuses an overstatement. Minimization ("it got out of hand, didn't it?") reframes the crime as understandable, even forgivable, making "yes" psychologically easier and carrying the implicit bargain confess to this softer version and things go better. Maximization ("the evidence is overwhelming") exaggerates the strength of the evidence and the severity of what awaits continued "lying." Together they squeeze the suspect between an unbearable position (keep denying → face the worst) and an apparent escape (accept the minimized story → relief). The implied bargain: admitting the minimized account is the way to leniency and out of the room — an offer the room was engineered to make irresistible.

Exercise 33.19 †

Re-read of Figure 33.2 — the two paths a scene detail can take into a confession: 1. CRIME → SUSPECT (genuine guilty knowledge): the perpetrator was at the crime, acquired the knowledge there, and reports it. This is what makes a true confession probative. 2. POLICE → SUSPECT (contamination): the police know the fact and it flows from them into the suspect during the interrogation — by a leading question, a visible photo, "isn't it true that…?", or correcting a "wrong" guess — and the suspect echoes it back, never having been at the crime. Reading only the final confession, the two paths are indistinguishable: a statement that "the door was unlocked" looks identical whether the suspect learned it at the scene or from a detective's question. This is why content alone cannot authenticate a confession, and why the decisive question is not what does it contain? but where did each detail come from? — which only a complete recording can answer.

Exercise 33.21

• Four process features matching the §33.3–33.4 risk profile (from Figure 33.3): (1) youth and exhaustion — a ~11-hour overnight session; (2) the false-evidence ploy — told, untruthfully, that physical evidence tied him to the scene; (3) minimization plus an implied path home; (4) contamination potential — every incriminating detail is one the investigators already knew, with no volunteered-then-confirmed fact (and the first hours unrecorded).
• Two physical findings that contradict the confession's content: the autopsy timeline (Chapter 11) — no soot in the airways and a blunt-force skull fracture show Diallo was dead before the fire, not killed by a fire set around a living man (the act Renner described); and the cell-site records (Chapter 25) — Renner's phone connected to towers far from Mill Creek Road across the relevant window.

Exercise 33.23

What you cannot determine from a beautifully matching but unrecorded confession: where each incriminating detail came from — i.e., whether the suspect volunteered facts (genuine guilty knowledge) or echoed facts the interrogators supplied (contamination). You also cannot reconstruct the length, the suspect's state, the tactics used, or any promises/threats. The recording gap is decisive, not incidental, because the matching detail is the entire basis of the confession's persuasive power, and contamination is a live, unexcludable explanation for all of it: without the recording, the one test that authenticates a confession (volunteered + previously unknown + later confirmed) can never be applied. A black-box interrogation produces a confession a court cannot properly evaluate.

Exercise 33.24 †

A prosecutor's "He confessed, and he knew the door was unlocked — only the arsonist could know that. Case closed." Two distinct errors: 1. The contamination error (§33.5). "Only the arsonist could know that" assumes the detail reached the confession by CRIME → SUSPECT. But if the police already knew the door was unlocked and could have conveyed it (POLICE → SUSPECT), the detail proves nothing. Without a recording showing the suspect volunteered it as a previously unknown fact, contamination is unexcluded. 2. The "confession ends the inquiry" error (Theme 1 / §33.1). "Case closed" treats the confession as a fact that settles guilt, rather than a claim to be tested against the physical record. It also smuggles in the "no innocent person would confess" intuition the chapter dismantles. Honest rewrite: "He confessed and his statement says the door was unlocked. Before this counts for anything, we need to know whether he volunteered that fact, that the police did not already have, and that was later confirmed — and we need to test the whole confession against the physical evidence. Unless those check out, the confession's detail is consistent with contamination and is not proof of guilt."

Exercise 33.25

The embedded assumption is that a detective can tell a true confession from a false one by experience or by how the suspect seemed ("no innocent person has ever confessed to me"). It is not reassurance because: (a) there is no validated way to distinguish a true from a false confession by its content or demeanor (§33.4 Junk-Science Alert) — the detective has no feedback signal telling him which of his confessions were false, since false confessors are typically convicted, not exonerated, so they never come back labeled; (b) the claim is exactly the fundamental attribution error at the professional level — confidence that his suspects were guilty because they confessed; and (c) the documented exoneration record shows experienced detectives have obtained false confessions, often without ever learning it. Absence of known false confessions is not absence of false confessions.

Exercise 33.27

Misleading in two ways (using §33.1–33.2): 1. Speed. Real interrogations that produce false confessions are characteristically very long (often many hours, past exhaustion); the twenty-minute extraction is fiction. Length is one of the strongest situational predictors (§33.4), and the brevity hides the very pressure that produces confessions. 2. A confession treated as proof of guilt. The show presents the confession as self-authenticating — the truth coming out — when the chapter's whole point is that a confession is a claim to be tested (§33.1, §33.5), and a detailed, tearful one can be entirely false. (Also acceptable: the TV detective's certainty models the guilt-presumptive interrogation that cannot recognize innocence, §33.2.)

Exercise 33.28 †

Argue both sides of using the false-evidence ploy (legal under Frazier v. Cupp): - For: it is legal and routine; with a genuinely guilty suspect it can break a lie and elicit a true confession; investigators have limited tools and confronting a suspect with (claimed) evidence is sometimes how a guilty person stops denying. - Against: it is the tactic most strongly tied to false confessions in both laboratory and case studies, and it is the engine of the coerced-internalized type — an innocent person told "we have your fingerprints" may conclude they must have done it and forgotten. Its risk falls hardest on the most vulnerable (juveniles, the intellectually disabled). "Legal" is not "safe" or "accurate" (§33.2). - Rule I would adopt (model): Do not use the false-evidence ploy at all — and certainly never on a juvenile or a person with intellectual disability. The benefit (faster admissions from the guilty) is outweighed by a documented, recurring harm (admissions from the innocent) that I cannot detect at the time, and the inquiry-based approach (§33.6 PEACE) obtains reliable information without it. Rationale: I cannot tell, in the room, whether my suspect is the rare guilty one or an innocent one the ploy will break — and the cost of being wrong is a wrongful conviction and a guilty person left free.

Exercise 33.29

Risk to the rest of the investigation (the bias cascade, §33.5 Cognitive-Bias Watch): once a confession is in the file, every subsequent analyst works in its shadow — the DNA examiner reads an ambiguous mixture toward the confessor, the fire investigator looks for the pattern the confession describes, and disconfirming evidence is explained away rather than followed. The confession doesn't add one wrong fact; it biases the interpretation of every other fact. Safeguard: keep the confession (and other domain-irrelevant context) away from analysts until their independent results are fixed (context management / sequential unmasking, Chapter 31); the confession does not get to set the conclusion the evidence must then confirm — the evidence is allowed to exclude the confessor. Why more dangerous than one ambiguous data point: a single ambiguous result is weighed alongside others; a confession reframes all the others, propagating one error through the whole file — and it does so while carrying the maximum persuasive weight, so it is the least likely to be questioned.

Exercise 33.31

Strongest argument for admitting the defense psychologist's testimony: it counters the overwhelming juror intuition that "no innocent person would confess" — the fundamental attribution error that otherwise decides the case before deliberation. The expert explains that false confessions are real, why they happen, and which risk factors (youth, intellectual disability, length, sleep deprivation, the false-evidence ploy) and contamination markers this interrogation displayed, giving the jury the knowledge it lacks to weigh the confession. The limit: the expert may not tell the jury the confession was false (or that the defendant is innocent) — that is the jury's call (and runs into the ultimate-issue concern, Chapter 30). The expert testifies to the science and the process, equipping the jury to evaluate how the confession was produced, not delivering the verdict.

Exercise 33.32 †

This chapter advances Theme 1 (exclusion over proof) and Theme 2 (the validity spectrum), with Theme 3 (cognitive bias) and Theme 4 (CSI effect) also touched. - Theme 1 — exclusion over proof, made personal. In the cold case the science does not merely fail to support Renner's confession; the autopsy timeline (Ch. 11) and cell-site data (Ch. 25) exclude him, overriding a signed admission that would otherwise have convicted an innocent man. Forensic evidence excludes — here it excludes a confessor. - Theme 2 — the validity spectrum, inverted. A confession is the "gold standard" in juries' eyes yet is an unrecorded human judgment produced under pressure, with no measured error rate, never validated for accuracy; its one feature that feels conclusive (corroborating detail) is the most easily manufactured (contamination). That is the spectrum stood on its head — maximum confidence on a "method" never tested. - (Theme 3: a confession biases every other analyst — the bias cascade. Theme 4: juries over-trust the confession and under-weight the physical evidence that contradicts it.) A student naming any two, with the exact mechanism for each, has answered fully.

Exercise 33.33

What all three share: each is evidence juries over-trust because it feels certain and intuitive, while its actual reliability is much weaker than its persuasive force — and each is dangerous when its strength is overstated. A microscopic hair "match" (Ch. 19) sounds individualizing but is, at best, weak class-level association (and was a leading exoneration cause); cell-site pinpointing (Ch. 25) sounds like a dot on a map but only places a phone in a broad coverage area; a confession sounds like the truth from the source's own mouth. Why the confession is the most dangerous of the three: (1) it carries the greatest persuasive force — jurors discount even DNA against it; (2) it has no error rate at all — at least hair and cell-site have a literature to argue limits, whereas a confession's accuracy was never measured; and (3) it biases the interpretation of all the other evidence (the bias cascade), so it doesn't just mislead on its own point — it corrupts the whole file.

Exercise 33.35 †

Cold-case exclusion entry for the Mill Creek workbook (model): - (a) Honest status of the confession, at its true strength: Renner's confession is false. (Not "a confession that needs checking"; on this evidence it is affirmatively false — a coerced-compliant false confession whose detail is contaminated, not crime-derived.) - (b) The two independent physical findings that exclude him, and their chapters: the autopsy timeline (Chapter 11) — no soot in the airways + a blunt-force skull fracture → Diallo was dead before the fire, not killed by a fire set around a living man; and the cell-site records (Chapter 25) — Renner's phone connected to towers far from Mill Creek Road across the relevant window. - (c) Why his confession's matching details prove nothing: every incriminating detail is a fact the investigators already knew and could have conveyed; not one is a volunteered, previously unknown fact later confirmed. By the only test that authenticates a confession, the corroboration is consistent with contamination (Figure 33.2) end to end — and the first hours were unrecorded, so it cannot even be examined. - (d) The one thing I decline to conclude, and why: I do not conclude who actually set the fire. This chapter's discipline is narrower and just as important — removing a false answer the system had already written down. Naming the perpetrator is the capstone's job (Chapter 39), and to do it now would claim more than this chapter's evidence has earned. - Log line: Renner — confession false; excluded by autopsy timeline (Ch. 11) and cell-site records (Ch. 25).

Exercise 33.37

Answer to "shouldn't the confession count for something against him?": In this case, no — it counts for nothing as evidence of guilt, and saying so is honest, not soft. Apply the contamination test (§33.5): a confession's incriminating power rests on facts the suspect volunteered, that police lacked, and that were later confirmed. Renner's statement contains none; every detail traces to what interrogators already knew, so contamination explains all of it and the confession adds no genuine guilty knowledge. Then apply the physical contradictions (Chapters 11, 25): the confession is not merely uncorroborated, it is contradicted — Diallo was dead before the fire (so the act Renner "confessed" to did not occur as described), and Renner's phone was far from the cabin. A claim that is both unauthenticated and refuted by independent physical evidence has zero probative value for guilt. Calling it "false" is the evidence-strength-honest conclusion (Theme 1); pretending it still "counts for something" would be the opposite of rigor — it would let the persuasive feel of a confession override the science that excludes him, which is exactly how innocent people are convicted.

Exercise 33.39 †

Reid technique vs. PEACE model (150–200 words, model): The two methods invert each other at every step. The Reid technique is an interrogation: its purpose is to obtain a confession from a suspect already presumed guilty. It confronts the suspect with confident assertions of guilt, refuses denials, and offers minimizing themes, and it permits deception about the evidence. Because it presumes guilt, it treats denials as resistance to overcome — so it has no mechanism for discovering it has the wrong person, which is precisely why it is so effective at producing confessions and so poor at producing truth. The PEACE model is an interview: its purpose is accurate information, not an admission. It does not presume guilt, does not lie about evidence, and does not try to extract a confession; instead it obtains a full, uninterrupted account and tests it against the known facts through open, non-leading questions. Only PEACE can recognize the wrong person — an innocent account, left intact, can exonerate. That difference is decisive for false confessions: a method that lets the subject talk keeps its best tool for telling the guilty from the innocent, while the accusatory model throws it away.

Chapter 34 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 34.1

Contributing factors to wrongful conviction are the recurring, identifiable causes that the systematic study of exonerations shows appear repeatedly in cases where an innocent person was convicted (mistaken eyewitness identification, false confession, flawed or misapplied forensic science, incentivized informants, official misconduct). The phrase is plural and uses contributing rather than cause because a wrongful conviction is almost never the work of one factor acting alone; it is a convergence of several factors that combine and reinforce one another. "Cause" implies a single sufficient origin; "contributing factors" names a set whose members each add to, and borrow credibility from, the others — which is exactly the mechanism (the cascade) the chapter is built to explain.

Exercise 34.3 †

Post-conviction review is the set of legal and investigative processes by which a conviction that is already final — its ordinary appeals exhausted — is reexamined for error. An innocent person typically cannot get relief through the ordinary appeals process because that process is built to test the fairness of the process, not the factual accuracy of the outcome: it asks whether the rules were followed (evidence properly admitted, jury properly instructed, counsel adequate), and a trial can pass every one of those tests and still have convicted the wrong person on mistaken evidence. Because factual innocence is one of the hardest things to raise on direct appeal, the error this chapter is about must travel the separate, narrower, and heavily obstructed channel of post-conviction review.

Exercise 34.5

The five leading contributing factors (§34.2), with the chapter that treats each in depth: 1. Mistaken eyewitness identification — Chapter 32. 2. Flawed or misapplied forensic science — this chapter's §34.3 (and the home disciplines: bite marks Ch. 16, hair Ch. 19, arson Ch. 22, prints Ch. 14). 3. False confession — Chapter 33. 4. Incentivized informants ("jailhouse snitches") — treated here in §34.2 (no single owning chapter; cross-cuts the courtroom material, Ch. 30). 5. Official misconduct (incl. Brady violations) — treated here in §34.2 and §34.4 (and the ethics/reform chapter, Ch. 38).

Exercise 34.7

The two anchor cases and the "sin" each represents: - Cameron Todd Willingham — the first sin (a method that was never valid): the arson "indicators" (crazed glass, pour patterns) that condemned him are fire-investigation folklore with no scientific basis. Not overstatement — folklore presented as science, with a body count. - Brandon Mayfield — the second sin (a valid method overstated): latent-print comparison is foundationally valid, but bias plus overstatement produced a 100%-confident, 100%-wrong identification.

Exercise 34.9 †

The wrongful-conviction cascade, beginning from a mistaken eyewitness identification: - Step 1 → 2 (eyewitness error). A frightened witness, shown a suggestive lineup, identifies the wrong person with growing confidence (Ch. 32). This is the original error and the spark. - Step 3 (tunnel vision). The detective now "knows" the suspect; the investigation narrows to confirm rather than test, and disconfirming leads quietly die. Borrows credibility from step 2: the confident ID is treated as solid ground, so pursuing it feels rational, not biased. - Step 4 (false confession). The suspect, interrogated for hours, is fed scene details and confesses; the "non-public" facts he repeats came from the interrogators (Ch. 33). Borrows credibility from step 3: the detectives, already certain, press hard and interpret compliance as guilt. - Step 5 (forensic drift). The examiner, told this is the man, reads an ambiguous hair or print toward the expected answer and testifies to a "match" with unearned certainty (Ch. 14, 19, 31). Borrows credibility from steps 2–4: the surrounding "answer" supplies the contaminating context that confirmation bias acts on. - Step 6 (conviction). The jury sees a witness, a confession, and forensic science all agreeing — apparently overwhelming, mutually corroborating proof. Borrows credibility from everything before it, while in fact every thread traces back to the error in step 2.

Exercise 34.11

The two statements are compatible because the cascade is a bias mechanism, not a conspiracy. The contamination flows in one direction — from the early "answer" to each later interpretation — and it flows invisibly: each actor experiences their own contribution as an independent judgment. The examiner does not feel biased; he feels the print simply matches. The detective does not feel he has tunnel vision; he feels he has the right man. So no dishonesty is required for an innocent person to be convicted. What that compatibility tells you is where the fix must be applied: not to people's intentions (everyone already feels objective, so "be objective" changes nothing), but to the structure — by withholding the contaminating information until each judgment is fixed (blind, context-managed analysis, Ch. 31). You cannot exhort away a bias no one can feel.

Exercise 34.12 †

The phenomenon is tunnel vision (a form of confirmation bias, Ch. 31). It sits at step 3 of the cascade, immediately after the initial identification or lead. It converts an honest investigation into one that manufactures a wrongful conviction because, once the detective treats the suspect as settled, the investigation stops being a search for the truth and becomes a search for confirmation: leads pointing away from the suspect are discounted or never pursued, ambiguous facts are read as inculpatory, and the later stages (the interrogation, the forensic comparison) are fed a single "answer" to confirm. The investigation thereby removes its own ability to be refuted — the precise opposite of what evidence is supposed to do — so each subsequent step compounds the original error instead of catching it.

Exercise 34.13

A false confession comes to contain "details only the killer would know" because the suspect learned those details from the interrogators, not from guilty knowledge. Over a long interrogation (Ch. 33), detectives — often believing the suspect is guilty — discuss, display, or imply facts of the crime scene (how a body was positioned, what was taken, where an injury was). A suspect under exhaustion and pressure, especially a young or vulnerable one, absorbs these "non-public" facts and repeats them back in the confession. The statement then appears devastating precisely because it contains knowledge "only the killer could have" — when in fact the knowledge entered the suspect's account across the table. This is why an unrecorded interrogation is so dangerous: without a record, no one can later show that the incriminating detail originated with the police rather than the suspect.

Exercise 34.15 †

The two sins of forensic failure: - First sin — a method that was never valid. The discipline has no validated basis for the specific claim it makes. Example: bite-mark comparison (Willingham's arson "indicators" and microscopic hair "matching" also qualify). The method itself is hollow, so even a sincere, careful examiner misleads the jury. Fix: exclusion. A method without foundational validity should not be in the courtroom at all, no matter how skilled or sincere the examiner (PCAST's position). You cannot fix bite marks with better-trained bite-mark examiners. - Second sin — a valid method overstated. The method may be real and high on the validity spectrum, but the testimony outruns the evidence — the wrong verb ("is" for "is consistent with"), claimed individualization, or a "zero error rate." Example: Brandon Mayfield (a foundationally valid latent-print comparison reported with false certainty); overstated serology (Earl Washington) qualifies too. Fix: calibrated language and measured error rates. The method stays; the testimony must report only what validation supports and must state the known error rate. A reform that addresses only the second sin (better testimony standards) while leaving junk methods admissible has fixed the smaller problem and left the catastrophic one in place — which is why telling the two apart matters.

Exercise 34.17

The Willingham forensic evidence was not overstated — it was something worse: folklore presented as science (the first sin in its most absolute form). The "indicators" of an intentionally set, accelerant-fueled fire — "crazed glass," "pour patterns," "alligatoring" — that the investigators read as proof of arson are not a valid method used too confidently; they are invalid for the claim made. Modern fire science (Ch. 22) has shown that these patterns are ordinary products of an accidental fire that reaches flashover, not signatures of arson. Independent fire scientists reviewing the case concluded there was no valid scientific basis to call the fire arson at all. So the failure was not "a real method pushed too far" but "a hollow method with no foundation," which is why the fix for it is exclusion, not calibration.

Exercise 34.19

Mayfield is the right anchor for a chapter on wrongful convictions even though he was a near-miss (never convicted) because his case isolates the mechanism that convicts the innocent, cleanly. A foundationally valid method (latent prints) plus context bias (a Muslim convert, an apparent terrorism link) plus overstatement (100% confidence) produced an identification that was 100% wrong — and the only reason it did not end in a wrongful conviction is that the Spanish authorities found the true source first. The same mechanism that nearly convicted Mayfield is the one that did convict the exonerees. The parallel to the cold case is exact: Cody Renner is the cold case's near-miss — a cascade (tunnel vision + a false confession) caught in time by the science (autopsy timeline, cell-site, DNA exclusion) rather than by luck. Both show that the difference between a wrongful conviction and a near-miss is often whether someone let the evidence refute the theory before it was too late.

Exercise 34.20 †

Using the validity spectrum (Ch. 1, 6): the methods that convicted several exonerees (bite-mark comparison, microscopic hair "matching," arson folklore) sit at the discredited end — they have no validated basis for their specific claims — while the method that freed them (single-source nuclear DNA) sits at the strong, quantified end. What separated them was the gap between the methods, not the competence of any individual examiner. A perfectly skilled, perfectly honest bite-mark examiner still misleads, because the method cannot do what it claims; a competent DNA analyst delivers a reliable exclusion, because the method rests on a rigorous, tested foundation. This is the chapter's hardest and most important point: improving the people cannot rescue a method whose core claim is unvalidated. The validity spectrum is therefore not an academic ranking — it is the difference between the science that imprisoned the innocent and the science that freed them.

Exercise 34.21

Figure 34.2 ("Two pillars, one source"): - Pillar (A), the "microscopically consistent" hair. Traces back to the eyewitness identification (the original error): the analyst was told the defendant was the suspect and had confessed, so a weak, "consistent-with" hair comparison was read by a context-contaminated examiner who expected this answer (forensic drift, step 5, contaminated by steps 2–4). - Pillar (B), the confession's "only the killer would know" detail. Also traces back to the eyewitness identification, via the interrogation: the "non-public" positioning detail entered the suspect's account from the interrogators (step 4), themselves acting on the certainty the ID produced (step 3). The jury experienced them as mutually corroborating because, on their face, they are two different disciplines — forensic hair analysis and a voluntary admission — pointing to the same man. The contamination is invisible from the jury box: nothing in the courtroom shows that both pillars grew from the same root. The honest reading is not "two pillars agree" but "one possibly mistaken identification, echoed twice."

Exercise 34.23

The testimony — "The defendant's print is the latent print; I identified it to him to the exclusion of all others, with a zero error rate" — contains at least three overstatements: 1. "is the latent print" ("is"). The verb claims identity where the science supports only is consistent with / cannot be excluded (Ch. 1, §1.4). A latent-print comparison is an association, not an identity statement. 2. "to the exclusion of all others." This claims individualization — that no other person on Earth could be the source — which no feature-comparison method has earned (Ch. 6). 3. "with a zero error rate." No human-judgment method has a zero error rate; latent-print comparison has a measured, non-zero error rate (the Mayfield case is the proof). Defensible rewrite: "The latent print shares features with the defendant's known print and, in my judgment, is consistent with him as a possible source; I cannot identify him to the exclusion of all others, and this method carries a known, non-zero error rate."

Exercise 34.25 †

What is inaccurate about "Forensic science is responsible for most wrongful convictions" is that it confuses common with most common. Flawed or misapplied forensic science is common — present in roughly half of the DNA exonerations — but it is not the single most common factor; mistaken eyewitness identification is, appearing in a large majority of the early DNA exonerations. "Responsible for most" also overstates causation: forensic science usually contributes within a cascade rather than solely causing the conviction. Accurate version: "Flawed or misapplied forensic science is one of the most common contributing factors to wrongful conviction — present in roughly half of the DNA exonerations — though mistaken eyewitness identification is more common still; and these factors typically combine rather than acting alone."

Exercise 34.27

Three questions to test whether a conviction integrity unit is genuine or cosmetic, with the answer that indicates a real unit: 1. "What has it actually done — how many convictions has it helped overturn, including ones the office was proud of?" Real: a track record of actual exonerations, not just reviews that always affirm. Cosmetic: exists on paper, has overturned nothing. 2. "Does it accept cases without DNA?" Real: yes — it will act on a recanted informant, a suppressed report, or a discredited method, not only on a clean DNA exclusion. Cosmetic: only "safe," DNA-driven cases. 3. "Is it independent, and does it cooperate with defense and innocence organizations?" Real: staffed by people with genuine authority to overturn the office's prior work, and treats innocence claims as questions to investigate rather than attacks to repel. Cosmetic: captured by the same institutional self-interest that resists every exoneration; treats the defense as the adversary. (Bonus: "Does it publish what it finds?" Transparency indicates a real unit.)

Exercise 34.29 †

The risk. Being told, before you examine the print, that the suspect has confessed and an eyewitness has identified him supplies exactly the domain-irrelevant context that drives confirmation/contextual bias (Ch. 31): an ambiguous latent print is now likely to be read toward the expected answer, and the contamination will feel like ordinary perception — you will "just see" the match. This is forensic drift (cascade step 5), and it is how a valid method produces a wrongful result. The safeguard that should have prevented it: context management / sequential unmasking — the lab's workflow should have withheld the confession, the identification, and the detective's belief from you until after your comparison was fixed, so your judgment could be a genuinely independent pillar rather than an echo of the others. What you should do now that you have learned it: acknowledge the contamination rather than pretend it away. Concretely: document that you received the context; if possible, have the comparison performed (or independently verified) by an examiner who is blind to it; render your own conclusion in calibrated language with its limits and the method's error rate stated; and never let your report convert "cannot be excluded" into "identifies." The honest move is to treat your now-contaminated judgment as weakened evidence, not to suppress the doubt.

Exercise 34.31

A wrongful conviction is both a defense issue and a public-safety issue. It is a defense issue in the obvious sense: an innocent person loses their liberty, and undoing that is the work of defense and innocence advocates. But §34.6 makes it equally a public-safety issue: when the system convicts the wrong person, the right person — the actual perpetrator — remains free, and in a sobering number of documented cases went on to commit further crimes (additional rapes, additional murders) while an innocent person sat in prison for the first one. The DNA that exonerates the innocent has, in many cases, identified the true perpetrator through a database hit, sometimes revealing offenses committed during the years the wrong man was incarcerated. So "getting the right answer" is the same project as "protecting the public": a forensic science that convicts the innocent is, in the same act, failing to catch the guilty and failing the future victims of crimes the real perpetrator was left free to commit. Exoneration is therefore not a "soft on crime" indulgence — it is a correction that serves public safety directly.

Exercise 34.33 †

All four themes converging in a single typical wrongful conviction: - Theme 1 — exclusion over proof. Feature: the conviction rests on a weak, overstated inclusion ("consistent with," "matches," "that's him") presented to the jury as if it were proof. Counter-practice: reserve "proves" for the rare quantified claim, and let exclusion do its surer work — state findings at their true strength. - Theme 2 — the validity spectrum. Feature: a discredited or weak method (bite marks, hair) carried evidentiary weight it had not earned. Counter-practice: admit and weight a method by where it sits on the NAS/PCAST spectrum; exclude the junk methods, calibrate the valid ones. - Theme 3 — cognitive bias is the chief threat. Feature: the examiner, knowing the wanted answer, drifted toward it (and the detective's tunnel vision fed the whole cascade) — without anyone being dishonest. Counter-practice: blind, context-managed analysis and sequential unmasking, so contaminating information is withheld until judgments are fixed. - Theme 4 — the CSI effect cuts both ways. Feature: jurors over-trusted forensic testimony they should have interrogated, hearing "the expert said match" as certainty. Counter-practice: communicating uncertainty honestly (Ch. 30) — making the evidence's true strength, and its error rate, audible to the jury.

Exercise 34.35

A high-validity method (e.g., fingerprints) can still produce a wrongful result, explained through Chapter 6's three rules of the spectrum: 1. Position is a ceiling, not a guarantee. A method's place on the spectrum is the best it can do when used correctly; it does not promise that any given application achieved that best. A valid method, operated under bias and reported with overstatement (Mayfield), performs far below its ceiling. 2. The spectrum is question-specific. A method valid for one claim is not automatically valid for a stronger one. Latent-print comparison may support "consistent with" or even a careful identification under good conditions, but stretched to "to the exclusion of all others, zero error rate," it is being asked a question its validity does not cover. 3. The honest verb tracks the position. When the testimony's verb outruns the method's earned position — "is" instead of "is consistent with," "identifies" instead of "strongly associates" — the report, not the method, becomes the failure. So a high-validity method produces a wrongful result chiefly through the second sin: bias drives the interpretation and overstatement reports it, pushing a valid method past its question-specific ceiling.

Exercise 34.36 †

Cold-case workbook entry (Appendix I) for Cody Renner — model: - (a) Honest status, at full strength: The evidence excludes Cody Renner as the person who killed Marcus Diallo. This is an exclusion logged at full strength (not "probably innocent"), resting on the surest voice forensic science has (the §1.6 asymmetry: a reproducible non-association is, barring laboratory error, conclusive as to source). - (b) The earlier chapter establishing each of the three threads: - The autopsy timeline (Diallo killed by blunt-force trauma and dead before the fire — no soot in the airways) — Chapter 11. - The cell-site / location data (Renner's phone placed him elsewhere during the relevant window) — Chapter 25. - The DNA exclusion (Renner is not a contributor to the heat-degraded mixture from the gas-can handle) — Chapters 7–9. - (c) Why the confession does not override them: a confession is a claim to be tested against the physical evidence, not a fact that ends the inquiry. Renner's confession is contradicted by all three independent threads, and its "non-public" details came from the interrogators (Ch. 33), not from guilty knowledge. A false confession cannot outweigh an autopsy timeline, a system-recorded location, and a DNA exclusion — the physical evidence was simply never asked the right questions until the science was done properly. - (d) Why the chapter stops short of naming who the science points toward: the three threads exclude Renner; they do not, by themselves, convict anyone else. Naming the guilty party is a separate act of evidence — the exclusion matrix and convergence of the capstone (Chapter 39) — and to leap from "not Renner" to a name would claim more than the evidence has yet earned. (It is a capital mistake to theorize before one has data.)

Exercise 34.37

Mapping the early Mill Creek investigation onto the cascade (§34.2): - Tunnel vision (step 3): the first comfortable narrative was accidental fire; when that cracked, the investigation narrowed onto a convenient local (Renner) with a thin alibi, searching to confirm rather than to test. - The confession (step 4): the eleven-hour, largely unrecorded interrogation produced a confession containing scene details Renner could only have learned from his interrogators (Ch. 33) — the cascade's most persuasive-seeming pillar, manufactured rather than discovered. - Where the physical evidence could have broken the chain: had the cabin's small lab worked the physical evidence before (or instead of) closing on the confession, the autopsy timeline (dead before the fire, Ch. 11), the cell-site data (Renner elsewhere, Ch. 25), and the DNA exclusion (Ch. 7–9) would each have contradicted the confession early, snapping the chain before a wrongful conviction formed. - The single most plausible §34.5 reform: mandatory electronic recording of the entire custodial interrogation. A complete recording would have made the manufacture of the confession visible — showing that the "non-public" details were supplied across the table — and would most directly have prevented the false confession from carrying the case. (Blind, context-managed analysis is the strongest forensic reform, but for Renner's near-conviction specifically, the confession is the load-bearing failure, so recording is the most plausibly decisive fix.)

Exercise 34.39

(Model, ~180 words.) "The expert said match, the witness was certain, and the defendant even confessed" can look like overwhelming proof and still be a single error wearing three masks. Ask three things. First, how independent are these pieces? If the witness's identification is what made police focus on this man, then the interrogation that produced the confession and the examiner who read the forensic "match" both began from that one identification — so they do not independently confirm it; they echo it. Second, what kind of forensic evidence, and how strong? "Match" is not a magic word: was it a quantified DNA result, or a weak "consistent-with" hair comparison from a method with a real error rate? Third, how was the confession produced? A long, unrecorded interrogation can feed a suspect the very details that later make the confession sound like guilty knowledge. The careful juror does not ask "did the expert say match?" but "what evidence, how strong, how independent, and how do we know?" That is the difference between catching a cascade and joining it.

Exercise 34.40 †

(Model, ~190 words.) The case for blind, context-managed analysis as the single most important reform is that the wrongful-conviction cascade is, at bottom, a bias mechanism: the forensic examiner who reads an ambiguous print toward the wanted answer is not lying — he is contaminated by knowing that a witness identified the suspect, that the suspect confessed, that the detective is sure. Withhold that domain-irrelevant context until the comparison is fixed, and the forensic link becomes a genuinely independent pillar instead of an echo of the original error. What it costs is almost nothing in money or science — no new instruments, no new methods — only the discipline of restructuring lab workflow to deliver evidence without the surrounding "answer." Why it works is that it attacks the contamination at its source rather than exhorting people to be objective, which fails because biased examiners already feel objective. And the most damning finding of the chapter is that most labs still have not adopted it: the reform is known, validated, and cheap, and the gap between knowing the fix and applying it is exactly where the next innocent person is, right now, being convicted. The reform is known; the will is the missing reagent.

Chapter 35 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 35.1

Disaster victim identification (DVI) is the organized, standardized, multi-disciplinary forensic process for recovering, examining, and identifying large numbers of the dead from a mass-fatality event and returning them to their families. The oldest forensic question it exists to answer is the one that precedes every other (Chapter 17): who is this?

Exercise 35.3 †

Commingled remains are the remains of two or more individuals mixed together, so that body parts and fragments cannot be assumed to belong to one person and must be re-associated — sorted back to individuals — before or alongside identification. The assumption "one set of remains = one person" is so automatic in a single-body case that we never name it, but it fails in a high-energy disaster (aircraft impact, explosion, building collapse) because bodies are not merely killed but disrupted, fragmented, and intermixed. Before you can ask "whose arm is this?" you may first have to establish that the arm and the torso recovered nearby are even the same person — a question answered by anthropology, physical matching, and especially DNA re-association.

Exercise 35.5

The five Interpol DVI phases in order: (1) the scene — recovery, documentation, and containment of commingling; (2) postmortem (PM) — standardized examination of the dead producing the PM record (fingerprints, dental, DNA, autopsy, property); (3) antemortem (AM) — collecting records on the reported-missing (dental, prints, medical, DNA references from relatives/effects); (4) reconciliation — comparing PM against AM to match and confirm identifications (via an Identification Board); (5) debriefing — reviewing the response for lessons learned. Antemortem/postmortem reconciliation is the systematic, case-by-case comparison of the PM record (from the bodies) against the AM record (on the missing) to establish which body belongs to which missing person.

Exercise 35.7

Kinship matching compares a profile from the remains to profiles from the victim's biological relatives (parents, children, siblings) and asks whether the two are related in the way the family claims, computing a likelihood ratio (Chapter 9) for the proposed relationship. It differs from a direct match — in which a remains profile is compared to a known reference from the victim themselves (e.g., a toothbrush) — because in a disaster the victim is dead and a direct reference is often unavailable, so identification is indirect, via relatives. Kinship matching is therefore generally weaker and more interpretation-heavy than a direct match.

Exercise 35.9 †

The office cannot "work faster" because the bottleneck is not speed per case but total capacity: 300 simultaneous cases overwhelm a one-body-at-a-time workflow that was never built for parallel volume, and decomposition and families will not wait. The situation forces a change of kind, not degree — a shift from artisanal depth (exquisite individual attention to each body) to industrial system: every body passes through the same standardized stations (photography, fingerprints, odontology, autopsy/sampling, property) in the same order, generating the same forms, so the data are comparable across hundreds of cases. Two concrete features a single-body workflow lacks: (a) standardized examination lines / forms so cases are mutually comparable; and (b) parallel primary-identifier streams (fingerprint, dental, DNA run at once) with triage so cheap/fast methods clear the easy cases and slow/expensive DNA is reserved for those that need it. (Also acceptable: a reconciliation/Identification-Board structure; family liaison; re-association of commingled remains.)

Exercise 35.11 †

The physical reason (Chapter 17): the friction-ridge skin of the fingertips and the soft tissue that carries it are destroyed early by fire, while teeth — the body's most durable identifying tissue, mostly mineral and shielded by the lips, tongue, and jaws — survive heat that obliterates skin, and dental restorations endure as well. So in a fire disaster fingerprints fail for most victims exactly where dental identification still succeeds. The disaster-specific ceiling on dental identification is the availability of antemortem dental records: a victim with no dentist of record, or whose records cannot be obtained (e.g., across a border) in time, cannot be dentally identified — and in fragmented events, not every recovered fragment includes the jaws.

Exercise 35.13

Because the three primary identifiers have different failure modes — the disaster condition that destroys one often spares another (fire destroys prints but spares teeth; fragmentation may leave a printable finger but no jaws; an explosion may leave only tissue for DNA). Since no single identifier works for every victim, running all three in parallel covers a victim population that any one method alone would leave partly unidentified. Determining a single "best" identifier and using only it would systematically fail every victim whose remains or records happen to defeat that one method.

Exercise 35.15

Depth is artisanal: maximal individual attention to each body. System is industrial: a consistent, standardized process applied identically to every case so that hundreds of cases yield comparable data. Heroic individual attention is the wrong instinct in a DVI because, at scale, it cannot be delivered to everyone, so it produces uneven treatment — some cases beautifully worked, others neglected — whereas a consistent good-enough process applied identically to all produces more correct identifications overall. "The system is the method": at scale, comparability and consistency beat brilliance applied unevenly.

Exercise 35.16 †

From Figure 35.2 (the reconciliation matrix): - (a) Exclusion vs. match. A single unexplainable discrepancy (a non-matching fingerprint, a tooth recorded as extracted that is present, a DNA profile incompatible with the proposed relatives) excludes a candidate cleanly — it empties that cell with confidence, because an irreconcilable difference can come from only one place: the wrong person. A positive identification, by contrast, requires accumulating enough concordant, specific features to overcome the pool size; concordance is shared by an unknown number of similar candidates, so it builds confidence more slowly. (This is the Chapter 1 / Chapter 17 exclusion asymmetry.) - (b) Pool size. A single match in a pool of 180 is weaker than the same match in a two-person problem because, in a large enough pool, broadly similar mouths, common fingerprint patterns, or compatible kinship statistics occur by chance — a quiet prosecutor's fallacy (Chapter 9). The strength of a match is not absolute; it depends on the size of the candidate pool it is drawn from. - (c) Convergence. "Convergence" means the same body matching on two or more independent primary identifiers (e.g., both dental and DNA). Because the identifiers are independent, their discriminating powers multiply, far exceeding either alone — so convergence is the safeguard that lets a confident identification be made even from a large pool, and it is what the Identification Board looks for before confirming.

Exercise 35.17

The statement does three things well: (a) it rests confirmation on a primary identifier with a quantified strength — a direct STR match with a stated random match probability (the rigorous identifier of Chapters 7–9), not on effects or recognition; (b) it uses a secondary identifier (the surgical implant) properly — to corroborate, not to confirm (§35.3), exactly its legitimate role; and (c) it states the match's strength honestly and specifically ("on the order of one in many billions"), the language of a defensible identification rather than a bare assertion. (Bonus: it ties the implant to the antemortem medical record, i.e., a documented known, not a guess.)

Exercise 35.18

The error is confirming an identification on secondary identifiers alone — clothing and a wristwatch. The category being misused is the secondary identifier: effects can be separated from a body and commingled in the event, borrowed, or coincidentally similar, so they are class-level associations of an object, not an identification of a person (the Chapter 1 duct-tape lesson). Honest version: "PM-077 is wearing clothing and a watch consistent with, and supporting an association with, AM-044; this guides reconciliation toward AM-044 and would corroborate a primary-identifier match, but does not by itself confirm the identification — confirmation requires a fingerprint, dental, or DNA primary identifier."

Exercise 35.19

Structural reason: the reconciliation matrix can only contain the AM columns that Phase 3 managed to build, so a victim for whom no antemortem record exists — no dentist of record, no prints on file, no relatives located to give DNA, perhaps never even reported missing — may remain unidentified no matter how complete the PM record is, because there is nothing to reconcile against. It falls unevenly because marginalized people — undocumented migrants, the unhoused, victims in places with poor record-keeping — are precisely those least likely to have a retrievable antemortem record, and so are likeliest to remain among the unidentified dead. This is an equity problem the field must name (§35.6).

Exercise 35.20 †

Two reasons to be cautious about "all 180 victims positively identified within 48 hours": (1) Some identifications surely rest on slow methods that cannot finish in 48 hours — degraded/fragmentary remains require DNA, which (outside rapid-DNA settings, Chapter 29) takes far longer; a complete fast count implies either an unusually intact population or, more worryingly, reliance on quick secondary-identifier or visual calls that may be wrong (§35.3). (2) Mass-fatality identification is, done correctly, conservative and slow (§35.1, §35.6): the framework would rather leave cases open than make fast, uncertain calls, so a claim of complete identification of a large population in days is in tension with how rigorous DVI actually works. Realistically still open: victims whose remains are too degraded for any quick method, and victims for whom antemortem records (or relatives for kinship) have not yet been assembled.

Exercise 35.21

Overstatement: confirming identity "from his wedding ring … no doubt." The category is a secondary identifier. The specific error is treating a class-level association of an object as confirmation of a person: a ring can be separated from a body, commingled, borrowed, or coincidentally similar. Defensible role: the ring guides reconciliation toward the right candidate and corroborates a primary-identifier match — it is genuinely useful in that supporting role, but confirmation must rest on fingerprints, dental, or DNA.

Exercise 35.23

A kinship identification should be expressed with its likelihood ratio (Chapter 9) because it is probabilistic: it tests whether the remains and the references are related as claimed, and how strongly the data support that relationship over the alternative. Stating it as a flat certainty hides the strength of the inference and invites the prosecutor's fallacy. A distant relationship gives a weaker result because more distant relatives share less DNA on average, so the expected sharing pattern is less distinctive and the likelihood ratio is lower — close relatives (parent/child) yield far stronger kinship statistics than, say, a second cousin.

Exercise 35.24 †

DVI's conservatism is a duty to families because the alternative error — a misidentification — is a "double catastrophe": family A buries and grieves around the wrong remains (and may never learn the error), while family B's person is mislabeled and effectively lost in the system, leaving them waiting in vain; and when the error surfaces, it reopens A's grief and compounds B's. Faced with that asymmetry, leaving a body unidentified and telling a family "we are not sure yet" — painful as it is — is better than a fast call that might be wrong. The whole framework (primary identifiers, the Identification Board, convergence) is the institutional expression of one promise: we will not hand you the wrong body. That is an obligation owed to families, not bureaucratic caution.

Exercise 35.25

The tension (§35.5): the humanitarian goal wants the relative's remains returned now, so a family can grieve; the judicial goal may need the grave and the remains preserved as evidence for a tribunal (the arrangement of bodies, associated cartridge cases, and trauma can bear on how the killing happened and on command responsibility). The two compete for the same scarce resources and even the same remains, and an exhumation optimized for fast identification may not be optimized for proving responsibility. A practitioner weighs them by recognizing both are legitimate, documenting the grave thoroughly enough to serve the judicial goal before releasing remains where possible, sequencing the work so families are served without destroying evidentiary value, and being honest with all parties that the priorities can pull against each other. There is no formula; there is disciplined, transparent triage.

Exercise 35.27

It is not primarily a failure of the science — the identification methods work where there is something to reconcile against. It is a failure of record availability and social structure: undocumented migrants, the unhoused, and people from places with poor record-keeping lack the antemortem records (dentists, print files, locatable relatives) that identification requires (§35.4, §35.6). What could reduce the inequity: better and more inclusive record-keeping; proactive collection of antemortem data and relatives' references; resourcing identification efforts for marginalized populations rather than letting them fall to the bottom of the queue. It is an ethical concern a careful field must name because the burden of remaining unidentified falls hardest on those already most marginalized — the science's reach is shaped by social inequities it did not create but can either mitigate or perpetuate.

Exercise 35.28 †

Under disaster conditions, on the validity spectrum (strong → weak/improper): 1. Direct single-source STR DNA match — STRONGEST. Quantified, rigorously validated, known error structure; identifies even degraded fragments; the gold standard (Chapters 7–9). 2. Dental identification with good antemortem radiographs — STRONG/VALID. Robust to the fire and decomposition that defeat prints; objective radiographic overlay; valid for the question it answers, error dominated by human factors (Chapter 17). (Reasonable to rank near a kinship DNA result rather than a direct one.) 3. Fingerprint comparison by a confident examiner — VALID BUT HUMAN-JUDGMENT. A real, useful identifier if the person is on file, but a comparison made by a human under a real, non-zero error rate; confidence is not correctness (Mayfield, Chapters 14, 31). Belongs below a clean direct DNA match. 4. Identification on a tattoo alone — IMPROPER. A secondary identifier and a class characteristic; it can guide and corroborate but cannot confirm alone (§35.3). At the bottom for confirmation purposes. Justification turns on each method's measured error rate, the rigor of its core claim, and whether it can confirm alone or only support.

Exercise 35.29

Both express the principle: describe the evidence on its own terms before fitting it to a theory. "No soot in the airways" (Chapter 11) was a finding read from the body itself, which then overturned the assumed accident narrative rather than being bent to fit it. The Interpol discipline of building the postmortem record independently of the antemortem record (§35.2) is the same move institutionalized: the dead are examined and described on their own terms before anyone asks whose body it is, so the description is not quietly shaped to match the expected identity. In both, the safeguard against confirmation bias is the same — let the evidence speak first, compare to the theory second.

Exercise 35.31 †

Cold-case workbook note (model): - (a) Three simplifications that made the Diallo identification easy: (1) no commingling — one body, indisputably one person, so no re-association was needed; (2) the antemortem record already existed and pointed at one named individual — there was a presumptive identity from the start (Diallo was the man renovating the cabin), so his own dentist's file could simply be pulled; (3) the candidate pool was effectively one, so the comparison was "is this Diallo or not?" rather than a search. - (b) Why the prosecutor's-fallacy worry of §35.4 did not arise: that worry is about a match weakening as the candidate pool grows. With a pool of essentially one, a strong dental concordance was decisive on its own — there was no large pool of similar mouths in which a coincidental match could hide. - (c) What the aside adds to the case: nothing evidentiary. It is a teaching comparison that introduces no new facts about Diallo's death, no new evidence, and nothing about any person of interest. The identification stands exactly where Chapter 17 left it — a sound positive dental identification confirming the burned body is Diallo and closing the substituted-body theory. (Teaching aside; no new suspect data.)

Exercise 35.33

Confirming Diallo's identity from the wallet and truck alone would be a methodological error because those are secondary identifiers (§35.3): a wallet and a vehicle are objects associated with a person, and objects can be planted, separated from a body, borrowed, or — central to this case — used to stage a scene. The whole "substituted-body" fraud theory (Chapter 17) was precisely that the body might not be Diallo even though his property was present; confirming identity from the wallet would beg exactly the question at issue. What the wallet can legitimately contribute: it guides the identification (it gives the presumptive identity that tells investigators whose dental records to request) and it corroborates the primary dental identification once made. It guides and corroborates; it does not confirm.

Exercise 35.35 †

Humanitarian vs. judicial purposes (model answer). Mass-fatality and mass-grave forensics serve two purposes that usually align but can compete. The humanitarian purpose is to recover and identify the dead so that families — especially the families of the missing, for whom not knowing is its own specific torment — can learn their relatives' fate, recover the remains, and begin to grieve. The judicial purpose is to gather evidence: in a disaster, for inquests, insurance, and any litigation; in a mass grave, for the prosecution of atrocity crimes, where the arrangement of remains, the trauma, and the manner of disposal establish that crimes occurred. What identification gives a family that nothing else can is the single most important fact — whether their person is among the dead and which remains are theirs — an answer that, however devastating, allows grief to begin where uncertainty cannot. What the science is honest about being unable to deliver: it cannot always succeed (some remains are too destroyed, and some victims have no antemortem record to reconcile against); it cannot return a person, only remains, sometimes only a fragment; and, in the judicial domain, it can establish the physical facts of a death but not, from the remains alone, who bears ultimate responsibility for it.

Chapter 36 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 36.1

Wildlife forensics is the application of forensic science to crimes and legal questions involving wild animals and plants (poaching, illegal trade in protected species, conservation-law violations). Its most foundational task is species identification — determining what species a piece of biological evidence came from, and whether that species is protected.

Exercise 36.3 †

Forensic engineering is the application of engineering principles and methods to investigate the causes of failures of structures, materials, components, and products when those failures cause injury, death, or loss and the cause must be established for legal proceedings. Failure analysis is its core method: the systematic determination of how and why something failed by examining the physical evidence and reasoning backward to a cause. It is the scientific method (Chapter 5) applied to wreckage: the engineer observes the failed object, gathers data, forms hypotheses about the cause (design, manufacturing, overload/misuse, deterioration), and tests each hypothesis against all the physical evidence, discarding those the evidence refutes until one survives on affirmative evidence. Like fire investigation (Chapter 22), it must sometimes honestly conclude the cause cannot be determined.

Exercise 36.5

A species identification (e.g., "this is elephant") is a class determination — it places the evidence in a group, like a herringbone sole placing a shoe in a manufacturing class. An individual/population identification (e.g., "this tusk came from this poached carcass" or "from this population") reaches toward the individual end of the arrow, using increasingly rare shared features — the same logic that takes a human comparison from "consistent with" toward "strongly associated" (Chapter 1, §1.3).

Exercise 36.7

Cause arcing is electrical activity that ignited the fire — a fault that produced enough heat to start combustion. Victim arcing (arc-through-char / fire-caused electrical damage) is electrical damage produced by the fire — the fire destroyed the insulation and caused arcing/melting in wiring that was a victim of the fire, not its cause. Only cause arcing indicates an electrical cause of the fire.

Exercise 36.9 †

Geographic-origin DNA compares a tusk's DNA against a reference map of how elephant DNA varies across the continent — and because populations in different regions carry different allele frequencies, the comparison estimates where the elephant lived. This is the same population-genetics logic that underlies the random match probability (Chapter 7): patterned, quantified genetic variation can be read backward to a source. The honest verb is "supports an association with a region" — it is an association with a place-type, not a GPS coordinate. Its confidence depends on how well that region is represented in the reference map.

Exercise 36.11 †

The attribution to factory X could be wrong because the real culprit is the unsampled upgradient factory Y, which used the same common solvent. A "match" between the plume and factory X's solvent then means only that X used the same common product as the true source — not that X produced the plume. The two factors that most control the real strength of a chemical source attribution are: (1) the distinctiveness of the chemical signature (a rare/unusual signature or marker is far more probative than a generic one a hundred sources could produce); and (2) the completeness of the candidate-source set (an attribution is only as good as the assurance that the true source was among those sampled). This is the same distinctiveness-and-completeness logic that governs pollen (Chapter 13) and soil (Chapter 24), and the same exclusion-over-proof asymmetry of Chapter 1, §1.6.

Exercise 36.13

Skid-mark speed is a minimum because a skid captures only the speed bled off after the wheels locked; any speed lost before the wheels locked (or after the skid ended) is not recorded, so the vehicle was going at least that fast. It is a range rather than a single number because the friction value feeding the energy calculation is itself an estimate. This is the same floor-not-verdict logic as the entomologist's minimum postmortem interval (Chapter 13): the method honestly delivers "at least X," with the true value possibly higher, reported as a range with stated assumptions.

Exercise 36.15

"We examined the electrical system and affirmatively determined it did not cause this fire" rests on affirmative evidence — a demonstration (e.g., victim arcing, no competent ignition mechanism, circuits not energized at the origin) that the electrical system did not cause the fire. "We found no accidental cause, so it must be arson" rests only on failure to find, which is the negative-corpus fallacy (Chapter 22): the accidental cause may have been destroyed by the fire or simply missed, so failure-to-find establishes nothing. The honest classification when a cause cannot be affirmatively established is "undetermined," not a conclusion reached by elimination alone.

Exercise 36.17

The sentence is appropriately honest in at least three ways: (a) it ties the conclusion to affirmative physical evidence ("the progression marks on the fracture surface and confirmatory materials testing") rather than to the engineer's authority or to failure-to-find; (b) it localizes and names the mechanism ("fatigue initiating at the keyway"), a checkable, specific claim another engineer could verify; and (c) it states consistency with the documented loads ("consistent with the cyclic loads documented in service"), grounding the cause in the actual service conditions rather than assumption. (Bonus: "consistent with" is the honest verb — it does not overstate to "proves.")

Exercise 36.19

Melted, beaded wiring alone does not support an electrical-cause conclusion because virtually every structure fire damages the building's wiring regardless of what started it — the fire melts insulation and beads conductors. "Melted wires near the origin" is therefore present at nearly every fire, which makes it almost worthless as a cause indicator on its own (the same ubiquity problem that made "crazed glass" worthless for arson, Chapter 22, §22.3). The affirmative analysis required is to distinguish cause arcing from victim arcing — examining whether the circuits were energized at the relevant time/location, whether the pattern fits an ignition source at the established origin or damage inflicted by an already-burning fire, and whether a competent ignition mechanism (a real fault capable of igniting nearby fuel) existed. The conclusion must rest on which kind of arcing is present, not on the mere presence of melted wire.

Exercise 36.21

Overstatement: claiming a partial, degraded sequence is "definitely African elephant" when it could not be distinguished from a closely related species that may not even be in the reference database. A species ID is only as good as the database's completeness and the sequence's resolution; a degraded sample may reach only genus, or may match multiple close relatives. Defensible version: "The sequence recovered is consistent with African elephant and excludes the substitute species tested; given that it is partial and a closely related species was not represented in the reference database, I cannot exclude that species, so the identification is at the level the data support, not a definitive single-species call."

Exercise 36.23

The fallacy is negative corpus (Chapter 22) — here "the product is defective because I couldn't think of any other reason it would fail" — the same back-into-a-conclusion-by-failure-to-find error, in a hard hat. The report needed affirmative evidence of the defect: a measurable flaw, a fracture surface showing the failure mode, materials testing showing the part was below specification, or a calculation showing the design was inadequate for the loads. Eliminating other causes can support a hypothesis, but a cause must rest on affirmative evidence of its own.

Exercise 36.25

Not necessarily a sign of dishonesty — it is consistent with how reconstruction works. The final speed range is built from uncertain inputs: the friction value, the scene measurements, and modeling assumptions. Reasonable, qualified reconstructionists can weigh those inputs differently and report different — even non-overlapping — ranges in good faith, exactly as the two qualified entomologists did in the Westerfield case (Chapter 13, Case Study 13.2). Overlapping ranges in particular signal honest uncertainty, not fraud. The remedy on cross is to probe the inputs (which friction value, which measurements, which assumptions), because that is where the estimate is actually made and where competent people can differ — and where an EDR, being a recorded log, is harder to bias than ambiguous physical marks.

Exercise 36.27

A species DNA result and a pollution source attribution each report a match against a set — the reference database of species, or the set of sampled candidate sources. If that set is incomplete — the true species not in the database, the true source never sampled — the "match" can be to the wrong member of an incomplete set, or can miss the real source entirely. It is therefore an ethical as well as a technical issue because reporting the match as if the set were complete misleads the court about the strength of the inference: it presents a probabilistic, set-dependent result as if it were exhaustive. The honest analyst discloses the gap — states that the database/candidate set may not contain the true source — so the court can weigh the attribution at its real strength (the same disclosure discipline the PCAST report demands of error-rate and validation gaps, Chapters 5–6).

Exercise 36.29

The two findings point at the same conclusion from opposite directions. Chapter 22 established affirmatively that the fire was deliberately set — multiple, separate, unconnected points of origin plus an ignitable-liquid (gasoline) distribution pattern (confirmed by GC-MS, Chapter 23). This chapter affirmatively closes the leading accidental alternative — the electrical system is ruled out as a cause (victim arcing, no competent ignition mechanism at origin). One shows the fire was incendiary; the other shows it was not an electrical accident. Both are needed for an honest staged-homicide reading because an incendiary finding invites the defense's natural challenge ("could it have been an overlooked accidental cause?"), and the engineering exclusion answers that challenge affirmatively rather than by failure-to-find — so the staged-homicide reading is supported from multiple independent directions and does not rest on the negative-corpus error that executed Cameron Todd Willingham.

Exercise 36.31 †

Cold-case evidence-log entry (model): - (a) Defensible finding: a forensic-engineering failure analysis of the cabin's electrical system affirmatively ruled it out as an accidental cause of the fire — the electrical damage present is fire-caused (victim arcing), with no competent ignition mechanism at the points of origin. - (b) Why an exclusion: the finding says not the electrical system — it eliminates a candidate cause. In the book's grammar (Chapter 1, §1.6) that is an exclusion, the cleanest and strongest kind of statement forensic science makes, not an inclusion of any particular cause or person. - (c) Relation to Chapter 22 (without duplication): Chapter 22 affirmatively established the fire was incendiary (multiple origins + ignitable-liquid pattern). This finding complements that from the opposite direction — it closes the leading accidental alternative — rather than re-proving the incendiary finding. The two are independent: one shows the fire was set, the other shows it was not an electrical accident. - (d) Why no name: "not an electrical accident" is a statement about the fire, not about a person — exactly as "the fire was incendiary" was. No accidental-cause exclusion identifies a hand that lit the fire; person-level linkage must come from the convergence of other evidence (soil, digital, DNA, financial) toward the capstone (Chapter 39).

Exercise 36.33

Two evidence types that, with the engineering and arson findings, begin to connect a person to the fire: e.g., soil on a suspect's boots matching the cabin's distinctive soil (Chapter 24) and cell-site/location and deleted-message data placing a suspect's phone near the cabin against his alibi (Chapter 25); gas-station CCTV of a suspect buying gas cans (Chapter 26) is a third. The engineering finding alone cannot make the leap because it is an exclusion of an accidental cause — a statement about the fire's mechanism, not about who was present or who set it. Establishing a person requires evidence that ties a person to the scene, the accelerant, or the act; the engineering exclusion only removes an innocent explanation for the fire.

Exercise 36.35 †

Wildlife DNA vs. environmental source attribution — validity foundations (model answer). Wildlife DNA species identification rests on the same rigorous, quantified molecular biology and population genetics as human DNA (Chapter 7): it reads sequence and compares it against a reference, yielding a class identification (species) — or, with population data, a probabilistic geographic-origin association. It is strong at its quantified core and is most easily overstated when a partial/degraded sample's identification is called "definite" despite an incomplete reference database or an indistinguishable close relative — i.e., its risk is false certainty about identity/origin given a gap in the set. Environmental source attribution rests on validated analytical chemistry (GC-MS, Chapter 23) for identifying what the compound is — strong — but the attribution to a source is a probabilistic comparison that depends on the signature's distinctiveness and the completeness of the candidate set; its risk is false individualization of a source when the "fingerprint" metaphor is allowed to promise a uniqueness a class characteristic cannot deliver, or when an unsampled source is the true culprit. In short: wildlife DNA's risk is overstating identity/origin past the database; environmental attribution's risk is overstating source past the candidate set — and the honest output in both is calibrated to the completeness of the set and the distinctiveness of the match.

Chapter 37 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 37.1

Forensic nursing is the practice of nursing applied to patients affected by violence, trauma, or other legal matters, providing health care while recognizing, documenting, collecting, and preserving evidence. The distinguishing feature is the simultaneity of two co-equal mandates — medical care and evidence collection — in a single deliberate encounter; bedside nursing has no evidentiary mandate, and the rest of the book's disciplines mostly work on the dead, who cannot consent, refuse, or be harmed.

Exercise 37.3 †

Evidence from the living is evidence recovered from a person who is alive, conscious, and entitled to direct what is done to their own body — so collection is subject to ongoing, informed, revocable consent and must minimize physical and psychological harm. It overturns the assumption, implicit for thirty-six chapters, that the source of evidence has no say in its collection: a bloodstain (Ch.10) or a skeleton (Ch.12) does not consent and cannot be re-traumatized, so the analyst's only constraints are technical. A living victim can refuse any step at any time and can be harmed by the process itself, so the encounter is governed by the patient's autonomy, not solely by what is technically optimal for the evidence.

Exercise 37.5

The sexual-assault evidence kit is a standardized package of swabs, envelopes, slides, and documentation forms used by a trained examiner to collect, label, and preserve biological and trace evidence from a sexual-assault patient for possible laboratory analysis. The one thing it is designed to do: collect and preserve evidence (chiefly biological material for DNA). The one thing it cannot do by itself: prove that a crime occurred — DNA can establish whose biological material is present, but it is largely silent on consent.

Exercise 37.7

A laboratory delay is a kit that was submitted but waits in an overloaded lab's queue — a resource shortfall (Ch.4). The backlog of untested kits is evidence that was collected but never submitted to a lab at all — a decision (or non-decision) by the holding agency. Conflating them lets responsible parties hide a discretionary failure to submit behind the more sympathetic story of an under-resourced lab; the two are different failures with different fixes (more lab capacity vs. mandatory-submission policy), and only naming the difference assigns responsibility correctly.

Exercise 37.9 †

"Medical needs first" is also the route to better evidence because, in the pre-SANE emergency-department model, poor care and poor evidence were the same failure. A rushed, untrained, uncomfortable examiner — working a sexual-assault exam infrequently and treating it as an interruption — produced both: long waits, no privacy, no advocate, and re-traumatization (the care failure) and inconsistent collection, broken chain of custody, missed or mislabeled samples, and documentation too thin to survive cross-examination (the evidence failure). The remedy for both is identical: a trained examiner doing careful, unhurried, respectful work. Because the same conditions that respect the patient (time, privacy, control, calm) are the conditions that produce thorough, well-documented, well-preserved evidence, prioritizing care does not cost evidence — it tends to improve it. (A patient who feels safe and in control can also better consent to and tolerate thorough collection.)

Exercise 37.11

The chain of custody is unusually fragile because the evidence is collected over a long, emotionally charged encounter, by a clinician in a hospital rather than an investigator at a scene, and then passes through more hands than most evidence (examiner → law enforcement → laboratory → storage). Every transfer is a point of failure, and a single gap can render the work inadmissible or contaminated. The specific living-victim feature creating the extra risk: collection happens in a clinical setting outside the usual evidence-handling chain, over hours, with consent steps interleaved — so the sealing, labeling, and signed transfers built into the kit are what make the chain auditable.

Exercise 37.13

The most important interpretive limitation: a kit-derived DNA match establishes whose biological material is present (contact), but not that a crime occurred, because the presence of another person's DNA is consistent with consensual contact as well as with assault. DNA is largely silent on consent — the forensic question it answers ("whose material?") is not the legal question that usually decides a sexual-assault case ("was there consent?").

Exercise 37.15

The freeze response and the absence of injury are "two faces of the same misconception" because both are read, wrongly, as evidence against the victim. The underlying false belief is that a "real" victim of sexual assault would fight back and be visibly injured — so the absence of resistance is taken for consent and the absence of injury for "nothing happened." The science refutes both: the freeze response (involuntary tonic immobility) is a hardwired threat response, not consent; and many assaults leave no documentable injury because the mechanism does not require marking force. Both absences are expected, not exculpatory.

Exercise 37.16 †

In Figure 37.2, the findings that make a strangulation reading defensible despite a near-unmarked neck: scattered petechiae on the lower eyelids (rising venous pressure), a hoarse voice and reported difficulty swallowing (airway involvement), and the patient's reported "things going gray" and a brief memory gap (transient cerebral hypoxia / possible loss of consciousness). The reported symptoms may matter more than the visible ones because strangulation's danger is largely internal and systemic — occlusion of vessels and airway, oxygen deprivation to the brain — so a reported loss of consciousness or incontinence is a direct marker of how compromised blood flow/oxygen was, which the skin may not show at all. The faint redness alone would be weak; the constellation plus reported symptoms carries the weight, and the near-normal neck does not weaken it.

Exercise 37.17

The four features of the account that are the documented signature of traumatic memory: (1) vivid, specific recall of certain sensory/central details (a sound, a phrase); (2) impaired sense of sequence and time; (3) fragmentation and a gap the patient cannot fill; (4) a peripheral detail recalled differently than in an initial brief statement. The figure says the fragmentation is consistent with a truthful memory of a traumatic event and that the inconsistencies/gaps are expected — but it explicitly does not prove the account is true (only that its form is consistent with genuine traumatic encoding) and does not, by itself, prove the assault occurred or identify the assailant.

Exercise 37.19

"2 cm bruise, left upper arm, bluish, non-tender" is an observation — it records only what is measurable/visible. "Defensive injury from being grabbed by the assailant" is an interpretation — it imports a mechanism, an actor, and a story not contained in the observation. The distinction protects the patient and the case because objective documentation will survive cross-examination, whereas an interpretation embedded in the record can be attacked (and can reveal examiner bias), and because separating the two lets a court weigh the observation independently of any narrative — the document-observations-not-conclusions safeguard (§37.2; Ch.31).

Exercise 37.20 †

"The genital injury proves the contact was non-consensual." Overstatement: claiming an injury establishes the mental state of consent. The research supports no clean injury "signature" of non-consent — some injuries occur in consensual contact and many assaults produce none — so a finding is at most one piece of consistency evidence. An injury can support that the history is consistent with the findings; it can never establish consent or its absence. Honest rewrite: "The documented injury is consistent with the history the patient described; it does not establish whether the contact was consensual, and the absence of injury would not have ruled out an assault."

Exercise 37.21

The inference "no injuries, therefore no assault (or consensual)" is invalid because the absence of injury is consistent with an assault: many assaults and strangulations leave no documentable injury (the mechanism may not require marking force; tissue heals; the freeze response produces no struggle; injury is simply variable). The chapter's symmetric statement is that the presence of certain injuries is consistent with an assault and their absence is also consistent with an assault — so absence is uninformative about consent, and treating it as exculpatory misunderstands the evidence.

Exercise 37.23

Two reasons the TV "kit solved by next scene" is false: (1) Time and process — a kit must be carefully collected, sealed, and transported, then analyzed in a lab with a real queue (Ch.4); results take far longer than an afternoon, and a CODIS hit is an investigative lead, not an instant identification. (2) The backlog — far from being tested immediately, hundreds of thousands of kits nationwide sat untested for years (§37.5; Detroit ≈11,000), and even when tested, the kit identifies whose material is present, not that a crime occurred. The fiction compresses time, skips the chain of custody and the lab, and overstates what a result means.

Exercise 37.24 †

Concern: that caring for and believing the patient will bias later testimony. Actual safeguard: the nurse documents observations, not conclusions (§37.2; Ch.31). "A 2 cm bruise on the left upper arm" is an observation; "defensive injury consistent with an assailant grabbing her" is an interpretation that imports a story. The trained examiner records the former and is cautious and explicit about uncertainty with the latter. "Believing the patient" and "documenting objectively" are not in tension: belief governs care (the nurse treats the patient as a patient, not a suspect), while objective documentation governs the record (which must survive being attacked). Conflating the two — letting belief shade the documentation toward a conclusion — is the error; keeping them separate serves the patient better, because objective documentation is what holds up in court.

Exercise 37.25

The misconceptions in §37.4/§37.6 ("real victims fight and are visibly injured"; "a truthful account is consistent and chronological") translate into thousands of discretionary non-submission decisions because an investigator holding those false beliefs reads ordinary, expected features of genuine victimization — no injury, no resistance, a fragmented account — as signs the report is weak or false. Case by case, such an investigator judges that this kit "isn't needed" or this case is "unfounded," and declines to submit it. Repeated across an agency and over years, those individually-bias-driven decisions accumulate into a backlog — cognitive bias (Theme 3) operating at institutional scale.

Exercise 37.27

You should decline to testify that an injury pattern "individualizes" the causing object because, as with bite marks (Ch.16) and bloodstain pattern analysis (Ch.10), the science supports class/consistency claims, not individualization. A pattern injury can be consistent with a kind of object or mechanism; it cannot, by visual comparison, be matched to one specific object to the exclusion of all others, and there is no validated method with a known error rate for that claim. What you can honestly say: "this injury is consistent with having been made by [a type of object/mechanism]," with the limits stated.

Exercise 37.28 †

On the NAS 2009 / PCAST 2016 validity spectrum (strong → overstated): 1. Single-source DNA typing of a kit swab — STRONG. Quantified, validated, known error structure; identifies whose material is present and cleanly excludes (Chs.7–9). Caveat: strong as identification, but silent on consent. 2. Documenting the size and location of an injury — STRONG (as observation). Solid, reproducible clinical observation; the defensible bedrock of injury evidence. 3. Inferring the specific object that made a pattern injury — WEAK/OVERSTATED. Like bite marks (Ch.16): class/consistency at best, not individualization; no validated method with a known error rate for the specific-object claim. 4. Inferring non-consent from an injury — NOT SUPPORTED. No clean injury "signature" of non-consent; the absence of injury is equally uninformative; the mental state of consent is not establishable from a finding. Justification turns on each activity's measured error rate and the rigor behind its core claim — the PCAST question — and on the gap between describing a finding (solid) and interpreting it into a cause, date, or mental state (overstated).

Exercise 37.29

The common underlying fact is that human memory is reconstructive and is altered by stress and suggestion, not a faithful recording. In this chapter (§37.6), traumatic memory is fragmented, non-chronological, vivid-then-gapped — so apparent inconsistencies are expected, not proof of lying. In Chapter 32, the same reconstructive, malleable memory explains why confident eyewitnesses are often wrong and why suggestive procedures corrupt identifications. In Chapter 33, it explains how interrogation pressure can lead innocent people to internalize and "remember" a crime they did not commit. One thread — memory is not a recording — runs through victim accounts, eyewitness identifications, and false confessions alike.

Exercise 37.31 †

Cold-case evidence-log entry (model): - (a) How to treat the report: as context, not accusation. A years-old domestic-violence report involving Dana Whitfield and Marcus Diallo is, at most, background; a prior report of victimization does not make Dana a suspect, and treating it as evidence of present guilt is exactly the bias failure §37.5 dissects. - (b) What the physical evidence establishes: the evidence excludes Dana — consistent with her verified alibi and the DNA work finding no association placing her at the scene. The evidence-from-the-living review of the old report does not disturb that exclusion. - (c) Honest running status: Dana excluded; scope clarified. The old report is not probative of the homicide. - (d) Methodological lesson: a careful forensic practice serves the person and follows the evidence; it does not convert a past report of victimization into a present presumption of guilt. To let the report reopen suspicion despite exclusionary evidence would substitute assumption for data — the error the whole book warns against.

Exercise 37.33

The evidence-from-the-living review changes nothing about Dana's exclusion because exclusion here rests on physical and verifiable evidence — a confirmed alibi and DNA non-association — which the existence of an old report does not touch. If anything, the review clarifies the scope of the inquiry: it confirms the old report is not probative of the homicide, removing a distraction rather than adding a suspect. Methodologically, letting the old report reopen suspicion despite the exclusionary evidence would mean privileging a narrative ("there was prior conflict") over data (alibi + DNA) — precisely the institutional-bias failure (§37.5) that, applied to a person, produces wrongful focus and, ultimately, wrongful conviction (Ch.34).

Exercise 37.35 †

Care and evidence are the same work (model answer). The chapter's central claim is that, done well, caring for a victim and collecting good evidence are not competing goals but the same activity. The pre-SANE emergency-department model proves it by negation: a rushed, untrained examiner produced both re-traumatizing care (long waits, no privacy, no advocate) and unreliable evidence (inconsistent collection, broken chain of custody, thin documentation) — one failure with two faces. The SANE model fixed both at once, because the conditions that respect a traumatized person are the conditions that yield good evidence. A trauma-informed principle makes this concrete: giving the patient control over pacing and the right to pause or refuse not only reduces harm but produces a fuller, more accurate account (traumatic memory documented in its real shape, §37.6) and better consent to thorough collection. The kind, unhurried examination is also the rigorous one — which is why care leads and evidence follows, and why both improve together.

Chapter 38 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 38.1

Forensic ethics is the set of moral obligations governing forensic practice — duties to scientific accuracy, impartiality, the people whose liberty depends on the result, and the legal system — that constrain what an analyst may do, claim, and testify to. Three recurring built-in tensions it exists to resolve (any three): thoroughness vs. the backlog; certainty vs. honesty; loyalty to the investigative team vs. fidelity to the data; what the evidence supports vs. what the retaining attorney wants the jury to hear.

Exercise 38.2 †

A code of ethics is the written, formal statement of forensic ethical obligations adopted by an organization or lab, used to guide conduct and measure misconduct. "Necessary but not sufficient" means a code is a real and needed thing — it articulates the standard and makes violation namable — but it cannot, by itself, do the heavy lifting: it is a piece of paper that asks the individual to resist, by will alone, pressures built into the structure of the work (the embedded lab, the backlog, the team's expectations). It assumes good faith, so it does nothing about the analyst who lies (Dookhan and Farak signed codes too); it is enforced weakly; and an analyst's strongest incentives come from their employer, not their professional association. A code describes what honesty looks like; it does little to produce honesty in a system structured to discourage it. The real fixes are structural and procedural (independence, blind verification, disclosure) — they make honesty checkable by someone else rather than relying on private conscience.

Exercise 38.3

Five core principles, each matched to a technical theme: Be objective = context management against cognitive bias (Ch.31, Theme 3). Stay within your competence = the validity spectrum / foundational validity; don't testify to individualization a method can't support (Ch.6, Theme 2). Represent the evidence honestly = the honest-verb discipline — excludes / consistent with / strongly supports, never proves (Ch.1, Theme 1). Disclose = the openness/confrontation principle that lets a result be checked (Ch.30, Melendez-Diaz). Serve the court, not a side = the answer to the hired-gun problem (Ch.30) and the independence problem (§38.2).

Exercise 38.5

A conflict of interest is a situation in which a position, financial stake, or loyalty creates a risk that professional judgment will be improperly influenced. The key word is risk, not corruption, because a conflict does not mean the judgment was corrupted — it means the situation is structured so that corruption (usually unconscious) is more likely. We regulate conflicts because the situation itself is dangerous, not because everyone in one succumbs; good systems remove dangerous situations rather than trusting everyone to resist them.

Exercise 38.6 †

OSAC is the Organization of Scientific Area Committees for Forensic Science. It does: coordinate the development of consensus-based scientific standards across the forensic disciplines and maintain a registry of standards it judges sound. It is administered by NIST (the National Institute of Standards and Technology). The single most important thing it cannot do: it cannot compel any laboratory to adopt its standards — compliance is voluntary. This is the chapter's point about OSAC: it is the weaker, recommend-not-mandate structure that replaced the NAS report's call for a strong national authority with real teeth.

Exercise 38.7

Forensic reform is the structural, scientific, and procedural agenda — independence, validation before admission, blind testing and context management, mandatory accreditation, honest standardized reporting, and active judicial gatekeeping — proposed to put forensic science on a sound scientific and ethical footing. The two landmark reports are the 2009 NAS report (Strengthening Forensic Science in the United States) and the 2016 PCAST report (on feature-comparison methods).

Exercise 38.9 †

The four mechanisms of structural tilt, with safeguards: 1. The analyst receives the case context (suspect's name, the wanted answer) → safeguard: context management / sequential unmasking (Ch.31) — give the evidence and the question, withhold the answer. 2. The detective is a colleague, not a neutral party (shared team, shared goal) → safeguard: structural independence (§38.2) — remove the lab from the agency so the "team" relationship is gone. 3. Success is defined by the prosecution's success (an inclusion feels like a win, an exclusion does not) → safeguard: a culture (and structure) that treats an exclusion as equally good science — best secured by independence plus blind verification, so the analyst doesn't know which result "helps." 4. The pressure is funded (budget/priorities depend on the parent agency) → safeguard: separate budget and chain of command (independence) so the lab does not depend on the prosecuting agency it might disappoint.

Exercise 38.11

The independence problem is "Theme 3 scaled up" because Theme 3 (cognitive bias is the chief threat) operates at the individual level — an analyst biased by knowing the wanted answer — while the independence problem is the same bias delivered structurally: an entire lab tilted toward one answer by being organizationally part of the side that wants it. They are "two halves of one reform" because they reinforce each other: institutional alignment shapes which results get scrutinized, which analysts get praised, what "a good day" feels like — and that culture then steers a thousand individual bench judgments. You cannot fully fix individual bias (with context management) inside an institution whose structure pushes the other way, and independence still leaves room for individual bias on ambiguous calls. Each fix needs the other.

Exercise 38.13 †

The four reasons independence remains least adopted: (1) Cost and inertia — new budgets, buildings, bureaucracies, and the dismantling of century-old arrangements. (2) Resistance from law enforcement — police/prosecutors are reluctant to give up the responsiveness and alignment of "their" labs (which are precisely the problem). (3) The absence of a crisis with a face — structural bias produces wrongful convictions diffusely, over years, with no single dramatic failure to force action. (4) The comfort of the status quo for the powerful — the people with authority to mandate independence are largely the people who benefit from non-independence. Which is hardest: a defensible answer is (4), because it is self-perpetuating — the obstacle is the very people who would have to act, so reform must come from outside the beneficiaries (courts, legislatures, public pressure). (Reason (3) is also defensible: without a galvanizing visible failure, the political will never forms — which is exactly why Houston, a visible failure, is the rare success.)

Exercise 38.15 †

The pattern: the reforms that ask the field to measure and report itself (validation research, accreditation, honest reporting language, bias scholarship) advanced — partially; the reforms that ask it to restructure power (lab independence, a strong national authority, active judicial exclusion) advanced least. It is "exactly the pattern you would expect if the obstacle were interest, not ignorance" because: if ignorance were the problem, progress would track difficulty or cost roughly evenly; instead it tracks whose power is threatened. Improving technique threatens no one's authority and so proceeds; surrendering the lab's alignment with the prosecution threatens the prosecuting institutions' control, and so stalls. The field proved more willing to improve its technique than to change who its scientists work for — a sorting by interest, not by knowledge.

Exercise 38.17

The National Commission on Forensic Science was a federal advisory body (convened by the Department of Justice with NIST) created after the NAS report; it did serious work on standards and ethics, including a model national code of professional responsibility. It was allowed to lapse in 2017 when its charter was not renewed. Its fate illustrates that the strong, independent National Institute of Forensic Science the NAS report asked for was never built — and that even the weaker, advisory version was dissolvable at will. The difference between "an authority with real power" and "a commission that can be allowed to expire" is exactly how seriously the deepest recommendation was taken: not very.

Exercise 38.19

A standard for performing a comparison specifies the validated procedure, controls, documentation, limits, and reporting language for how to do the analysis. Evidence that the comparison's central claim is valid is foundational-validity research — black-box studies showing the method can actually distinguish true matches from false ones, with a measured error rate. Bite marks make the gap concrete: one can write a detailed, consensus standard for how a forensic odontologist should perform a bite-mark comparison, and analysts can follow it precisely and consistently — and the central claim ("these teeth, and no others, made this mark") still has no validated basis (Ch.16). A perfect process for an unvalidated method yields consistent, well-documented, unsupported results.

Exercise 38.20 †

OSAC is treated as both because the apparent tension dissolves once you separate what it does from what it represents. As progress: OSAC created, for the first time, a coordinated, scientifically vetted, cross-disciplinary process for writing down evidence-based standards — a genuine, forward-looking improvement, and an analyst who follows sound standards practices better forensic science than one who does not. As an emblem of the thesis: OSAC can only recommend, not mandate — compliance is voluntary, and a standard in a registry changes nothing until a lab, an accreditor, or a court demands it. It is precisely the weaker structure that replaced the NAS report's call for an authority with teeth. So OSAC is what reform looks like when the field will develop better standards but won't create a body that can compel their use — progress and limitation in one institution.

Exercise 38.21

The overstatement: "reliable" is doing illegitimate work. What is true: the lab follows a published national standard and is presumably doing the procedure consistently and in a documented way. What is smuggled in: that following a procedure standard establishes the method's claim is valid. It does not (see 38.19). If the method lacks foundational validity, a beautifully standardized, well-documented process produces results that are consistent and defensible-looking but still scientifically unsupported. Honest version: "We follow the published standard, so our process is consistent and documented — but for whether the method itself is valid, you must ask separately whether foundational-validity studies exist."

Exercise 38.23 †

Conflicts of interest are especially dangerous in forensic interpretation because the work involves interpretive judgment under ambiguity — calls like "is this an unexplained difference or an explainable one?", "is this peak a real allele or noise?", "do these ridge details correspond?" An interest, especially an unconscious one, tips exactly those close calls. The same conflict would be harmless in a purely mechanical task (no financial incentive will make you misread a ruler) because there is no ambiguity for the bias to exploit. Combine a genuine conflict (the embedded analyst, the hired gun) with a genuinely ambiguous comparison and you get the precise conditions under which honest people reach biased conclusions — which is why the fixes remove the conflict or hide the answer rather than asking people to try harder.

Exercise 38.25

Non-adversarial science is a cooperative search for what is true, where you want to find your own errors. The adversarial legal system is a structured contest where each side advances its position and truth is meant to emerge from the clash. The phrase "the adversarial system reaching back into the laboratory and bending the science before it reaches the courtroom" describes the failure of keeping these separate: when the analyst adopts the adversarial posture (sees themselves as part of the prosecution team), the courtroom's contest-logic invades the bench, where the work should be disinterested. The reform ideal is to keep the science non-adversarial — independent, blind, honest, serving the court — and let the adversarial process test, afterward, a result produced disinterestedly.

Exercise 38.26 †

The director's statement conflates/overstates two things: (1) that accreditation (which certifies conformance to a quality system — Ch.4) equals correctness of a given result; and (2) that following a standard (a procedure) equals the method's claim being valid (§38.4). Both are the "necessary but not sufficient" error. (One could also note (3): it implies the individual result is reliable when accreditation audits the system on a sampling basis.) Honest rewrite: "Our lab is accredited and follows national standards, so our processes conform to recognized quality and procedure requirements — which raises the floor on reliability. It does not, by itself, establish that any particular result is correct or that the underlying method is foundationally valid; those must be checked separately."

Exercise 38.27

The overstatement: "a reasonable degree of scientific certainty" from a method PCAST found lacks established foundational validity — it dresses an unvalidated claim in the language of scientific confidence it has not earned. What PCAST said instead: for a method with foundational validity, an examiner may testify to a match only by reporting the relevant error rate from validation studies, never "certainty"; for a method lacking foundational validity, such testimony should not be admitted at all. So the honest options are: report the measured error rate (if the method is valid) or do not offer the comparison (if it is not). "Reasonable degree of scientific certainty" is exactly the phrase PCAST and NAS singled out as overstatement.

Exercise 38.29 †

Five personal reforms (any five), each with its theme/chapter: Refuse to read the case narrative before analysis; ask for the question, not the answer → context management against bias (Ch.31, §38.2). Report "consistent with" when that is what the evidence supports; refuse "identifies"/"proves" when unearned → honest verbs / exclusion over proof (Ch.1, Theme 1). State your method's error rate without being asked → foundational validity / validity spectrum (Ch.6, PCAST). Document everything so it can be independently checked → disclosure/confrontation (Ch.30, Melendez-Diaz). Decline to testify beyond your competence → the competence principle / validity-as-applied (§38.1, Ch.6). Treat an exclusion as just as good a day's work as an inclusion → resisting the prosecution-success culture (§38.2). Point: none requires institutional permission — they are available to any analyst today.

Exercise 38.31

The mechanism: the field overstates because the courtroom rewards overstatement and punishes honest uncertainty (the CSI effect, Theme 4) — a confident "match" persuades juries, while "consistent with, error rate ~1 in N" can read as weakness. A more sophisticated jury — one that understands the validity spectrum, asks for error rates, and treats "consistent with" as the meaningful honest statement it is — flips that incentive: now honesty persuades and overstatement invites challenge. That changed incentive propagates "back down the chain": attorneys stop shopping for experts who overstate, experts stop overstating to satisfy juries, and analysts at the bench find that careful, hedged reporting is what wins. The juror is thus part of the structure that determines whether the science tells the truth.

Exercise 38.32 †

Yes, it is ethical for a competent expert to accept payment from one side — experts are routinely and legitimately retained and paid by a party, and being paid does not by itself corrupt judgment. The hired-gun problem is distinct: it is when the financial/reputational relationship shapes the conclusion — when an expert reliably reaches whatever opinion pleases the paying side, and gets re-hired for it. Practices that keep acceptance-of-payment from becoming the hired-gun problem: forming the opinion from the evidence before and independent of what the retaining side needs; being willing to deliver (and to have disclosed) an opinion unfavorable to the payer; charging for time, not for a conclusion; full disclosure of the financial relationship; and holding the same opinion regardless of which side called you. The duty (§38.1, Ch.30) is to the court and the truth, not the client — payment is compatible with that duty only if the opinion would be identical no matter who paid.

Exercise 38.33 †

Running a result through the five-question checklist (Figure 38.2): Node 1 (valid method?) — a bite-mark "match" fails here: the method lacks foundational validity, and no careful application rescues it; the result should not be relied on at all. Node 4 (reported at true strength?) — an overstated DNA result fails here: the method is valid (passes 1), may be correctly applied (2) and unbiased (3), but the testimony claims more than the evidence supports ("identifies to the exclusion of all others," "zero error rate") instead of the honest match-with-error-rate. They fail at different nodes because the failures are different in kind: bite marks are a method failure (the thing itself doesn't work), while overstated DNA is a reporting failure (a valid result described dishonestly). This is exactly the Chapter 6 distinction between a method that lacks validity and a valid method overstated — the checklist locates each failure precisely.

Exercise 38.35

On the validity spectrum, highest-confidence node passed / node failed: Single-source DNA, reported honestly — passes all five; the model trustworthy result. Latent-print comparison testified to as "zero error rate" — the method is foundationally valid (passes node 1) and may pass 2–3, but it fails node 4 (honest reporting): "zero error rate" is the overstatement PCAST demolished; the honest form is the measured false-positive rate. Firearms/toolmark "identification" — fails (or is in serious doubt at) node 1: PCAST found foundational validity not established; at best it supports class-level association, not individualization. Bite-mark comparison — fails node 1 outright: no validated basis for the source claim. Justification: nodes 1–2 are about the method and its application (the NAS/PCAST validity question), node 4 is about testimony — DNA and prints clear node 1, so their danger is overstatement; firearms and bite marks fail at node 1, so no careful reporting saves them.

Exercise 38.36 †

Cold-case Process Critique entry (model): - (a) Structural problem present: Mill Creek was worked by a small, non-independent county lab embedded with the investigating agency, under an early "accidental fire" frame and, later, a confession already obtained — the independence problem (§38.2) and a contaminating context flowing to the analysts (the bias cascade, Ch.6; contextual bias, Ch.31). - (b) Reforms that would have produced a cleaner result: lab independence (an analyst outside the agency would not inherit the accident frame or the confession's pressure); context management / sequential unmasking (the DNA mixture and the partial print interpreted from the evidence and the question, not the suspect's name); blind verification (key comparisons — mixture, arson origin-and-cause, soil/pollen — independently re-examined by a second analyst blind to the first's conclusion and the case context); accreditation and sound standards (tighter documentation, controls, and reporting language). - (c) Honest status: process critique completed. - (d) Why this is not a change to the conclusion: the critique is of the machinery, not the answer. It establishes risk (the conditions for bias were present), not proof of harm (that any specific result is wrong). The case appears to have reached a sound result despite its process; assembling the evidence and naming who is responsible is the capstone's work (Ch.39), and this checkpoint deliberately does not pre-empt it. ("Process critique completed" ≠ "the conclusion is overturned.")

Exercise 38.37

Take the heat-degraded DNA mixture (Chs.8–9) as the example. (i) Lab independence: an analyst outside the prosecuting agency interprets the ambiguous mixture without the team's expectation that it should implicate the suspect — so a borderline contributor call isn't nudged toward the wanted answer. (ii) Context management / sequential unmasking: the analyst is given the mixture data and the reference profile to compare, but not the suspect's name, the confession, or the investigators' theory — so the interpretation of which peaks are real alleles vs. noise is made before any contaminating context arrives. (iii) Blind verification: a second qualified analyst independently interprets the same mixture and computes the likelihood ratio without knowing the first analyst's conclusion or the case context — so agreement means genuine corroboration, not shared expectation. What an opposing expert could currently argue, in their absence: that the mixture interpretation was made by an analyst who knew the wanted answer, in a lab aligned with the prosecution, without an independent blind check — so an ambiguous, heat-degraded mixture may have been read toward the suspect, and the LR's weight should be discounted accordingly. (The same logic applies to the arson origin-and-cause call and the soil/pollen associations.)

Exercise 38.39

Model answer (closing of an argument to a police chief): Making the crime lab independent does not weaken the department — it strengthens the evidence the department depends on, because evidence produced by a lab structurally aligned with the prosecution is vulnerable to attack on exactly that ground (§38.2; Melendez-Diaz). An independent lab's results carry more weight in court precisely because they cannot be dismissed as one-sided; an exclusion from an independent lab is more credible, and so is an inclusion. The objection "we'd lose control of our lab" mistakes control for reliability: the responsiveness and alignment the department values are the very features that create structural bias and give the defense its opening. Independence trades a kind of control the department should not want (the power to shape interpretation) for a kind of credibility it badly needs (results a jury and an appellate court will trust). The department wins more cases, and loses fewer on appeal, with evidence that was produced disinterestedly. (See Houston, Case Study 38.2: a department that lost control of a broken lab and gained a credible independent one.)

Exercise 38.40 †

Model closing argument of the book: The crisis in forensic science is no longer primarily a scientific problem because the science of what to fix and how is, to a remarkable degree, settled — the NAS and PCAST reports named the reforms more than a decade ago: validation before admission, blind testing and context management, structural independence, mandatory accreditation, honest reporting. What is missing is not knowledge but will — and that makes the crisis an ethical and structural one. Why is this a reason for hope rather than despair? Because a problem of will is, in principle, solvable by choosing differently — and because the reforms divide into two kinds. The reforms that restructure power (independence, a national authority, judicial exclusion) require institutional will the field has lacked, and they will come slowly, through courts, legislatures, and public pressure. But the reforms that matter most at the bench — refusing the case narrative, reporting only what the evidence supports, stating the error rate, treating an exclusion as a good day's work — require only the will of the individual scientist, and are available today regardless of who signs the paycheck. A field full of analysts who practice the personal reforms is already most of the way to the honest future, and is the constituency that will eventually demand the structural ones. The choice — to be a scientist first, to claim only what the evidence supports, to want to find your own errors — is the whole of forensic ethics and the seed of all the reform. It is now yours to make.

Chapter 39 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 39.1

Forensic case assembly is the systematic gathering and combined evaluation of every item of evidence in a case, each stated at its true evidentiary strength, to reach the most defensible overall conclusion and to mark that conclusion's limits. The first rule of honest assembly: a thread is carried forward at exactly the strength it earned and not one degree stronger — a "consistent with" finding does not become "proves" because it now has company. (The sheer number of threads creates a feeling of certainty no single thread supports.)

Exercise 39.3 †

Convergence of evidence is the agreement of multiple independent lines of evidence, each established on its own, all pointing the same way, which together support a conclusion no single line could support alone. The load-bearing word is independent. It matters because the power of convergence comes from multiplying independent improbabilities: if truly independent, the chance that several unlikely findings all coincidentally point at an innocent person is small. If the threads are not independent — if one shared cause produced them all — you cannot multiply, and apparent agreement means nothing (the Beatrice Six). Number is not the measure; independence is.

Exercise 39.5

The three excluded persons and the evidence that excluded each: - Victor Salas — DNA non-match (biological) + verified alibi (temporal). Two independent exclusions. - Dana Whitfield — DNA non-association + alibi. (An old DV report is a reason to look, not to conclude; exclusion stands.) - Cody Renner — DNA exclusion + cell-site placing him elsewhere + the dissection of his false confession against the physical timeline. Three independent exclusions overrode a confession. Plus the unknown-stranger theory, closed by IGG (the minor mixture contributor is not a stranger).

Exercise 39.7

The circumstantial triad: motive, opportunity, means. - Motive — the insurance naming Keller + crushing debt (Ch.27) and the altered beneficiary (Ch.18). - Opportunity — broken alibi + cell-site in the area (Ch.25), CCTV gas-can purchase (Ch.26), soil on boots (Ch.24), pollen placing a vehicle (Ch.13). - Means — confirmed gasoline accelerant (Ch.21–23), pry-bar class match = forced entry (Ch.16), access to the sedative (Ch.20).

Exercise 39.9 †

Renner's exclusion. Three independent findings: (1) DNA excluded him as a contributor; (2) cell-site records placed him elsewhere during the window; (3) the false-confession analysis showed his account contradicted the established physical timeline. His confession did not end the investigation because a confession is a claim, not a verified fact, and is subject to exclusion like any other claim — and the science did exclude it. The row teaches that physical and digital evidence can be more reliable than a confession: the single most persuasive evidence to a jury (an admission in the suspect's own words) was false, and only the patient, unglamorous science prevented a wrongful conviction. Confessions are not exempt from exclusion.

Exercise 39.11

Independent improbabilities multiply because for independent events $P(A \text{ and } B) = P(A)\times P(B)$; stacking several small independent probabilities yields a very small joint probability (the §39.3 illustration: $\frac{1}{10}\times\frac{1}{20}\times\frac{1}{30}\times\frac{1}{5}=\frac{1}{30{,}000}$). Non-independent threads cannot be multiplied: if a shared cause (e.g., bias) produced them, they are not separate improbabilities but one improbability counted many times, so multiplying massively overstates the strength. Presenting a multiplied figure to a jury is dangerous even when the threads are independent because (a) the input probabilities are usually illustrative/unmeasured, (b) it invites the prosecutor's fallacy (Ch.9), and (c) a precise-looking number launders uncertain assumptions into false certainty. The structure (independence → multiply) is the lesson; the number must not go to a jury as if measured.

Exercise 39.13

Motive leg: Forensic accounting (Ch.27) established insurance on the property and on Diallo's life naming Keller, severe non-shareable debt, and Benford anomalies consistent with manipulated renovation books; questioned-document analysis (Ch.18) indicated an altered beneficiary. Together: a strong financial motive. Why "strong motive" ≠ "guilt": motive answers why a crime might have been committed, not who committed it. Most people with debts and convenient policies harm no one; other people in the case had financial friction with Diallo. Motive becomes weighty only when corroborated by the other independent legs — not on its own.

Exercise 39.15

Means leg: confirmed gasoline accelerant (presumptive Ch.21 → GC-MS confirmation Ch.23, with the incendiary origin-and-cause finding Ch.22); pry-bar class match establishing forced entry (Ch.16); and a sedative at an incapacitating level in the victim's blood (Ch.20), which required access to the sedative. The single opportunity item that also reinforces means is the CCTV gas-can purchase (Ch.26): it ties Keller specifically to the material (gasoline/containers) used to commit the arson.

Exercise 39.17

Dental identification (Ch.17) sits higher than the DNA mixture because it is, in effect, a clean comparison of an antemortem dental record to postmortem teeth — a well-validated, low-ambiguity individual identification of the body — whereas the gas-can DNA is a heat-degraded mixture of at least two contributors, whose interpretation as a likelihood ratio involves deconvolution choices and (potentially) probabilistic-genotyping software with a black-box dimension (Ch.9). Same biology label, different interpretive load: the dental ID answers "whose body is this?" cleanly; the mixture answers "did Keller contribute?" only at a stated ratio, not a certainty.

Exercise 39.19

A negative GSR result is "consistent with" the case rather than evidence for Keller because it does not point at any person; it simply agrees with the already-established manner of death — blunt-force trauma, no gunshot wound (Ch.11). It corroborates the conclusion of Ch.15 that the stray cartridge case was a red herring, unrelated to the killing. (Recall §24.1: a negative GSR proves little alone; here it is not asked to carry weight alone — it merely corroborates a manner of death the body already established.) Misreading a negative as affirmative evidence of innocence or guilt would be the error.

Exercise 39.20 †

"The forensic evidence proves beyond doubt that Roy Keller murdered Marcus Diallo" overstates in at least four ways: 1. "Proves beyond doubt" — the forbidden verb plus metaphysical certainty; the evidence (a mixture, class toolmarks, area-level location, a co-owner's soil) sums to strong support, not certainty. 2. "Murdered" — a legal conclusion (the ultimate-issue problem, Ch.30); the expert testifies to the weight of evidence, leaving "guilty/murder" to the jury. 3. No statement of limits — an honest conclusion volunteers what it does not establish. 4. Treats threads as a single proof rather than an assembled, independently-converging body each at its own strength. Honest rewrite: "The totality of the forensic evidence strongly supports that Roy Keller killed Marcus Diallo and staged the fire; it does not establish this to a scientific certainty, and each finding carries the specific limits noted in the report."

Exercise 39.21

Defense (strongest form): the biological anchor is a heat-degraded mixture interpreted as a likelihood ratio, not a single-source match; mixtures involve interpretive/deconvolution choices, degraded mixtures more so, and probabilistic-genotyping software has a black-box dimension — so the DNA is consistent with Keller's contribution but does not identify him. Honest prosecution response: concede all of that, and state the DNA at its true strength — an LR that strongly favors inclusion over exclusion is genuine, quantified evidence and one independent thread among several — without upgrading "consistent with / supports by ratio X" into "is his." Neither dismiss the LR nor inflate it.

Exercise 39.23 †

A compromised scene (Ch.2: treated as an accidental fire, not held to a homicide standard in its first hours) widens the uncertainty around every physical thread collected afterward because it opens real questions about contamination, chain of custody, and whether a trace was deposited at the crime or at some innocent other time — each a legitimate alternative explanation a defense may raise. Does it erase the convergence? No. It bounds it. The strongest threads (DNA chemistry, GC-MS, cell-site records, bank documents) are largely produced by methods and records less vulnerable to scene-handling errors, and the independent agreement of many threads is not undone by doubt about any one. The compromised scene is a reason the conclusion stops at "strongly supports," not a reason to abandon it. (Honest answer must concede the real cost and deny that it dissolves an independent convergence.)

Exercise 39.25

Four further tests, each with its symmetry: 1. Resolve the mixture (Y-STR, repeat probabilistic genotyping) — could strengthen the LR for Keller or weaken it (or exclude him). 2. Tighten the location (any non-area source: a plate reader, a precisely-located transaction) — could move "in the area" toward "at the place," or could place the phone elsewhere. 3. Date the trace — could show the soil was deposited near the time of the crime (strengthen) or long before, innocently (weaken — the co-owner problem). 4. Blind re-analysis of the contested interpretations — could reproduce the threads (strengthen, by neutralizing the bias concession) or fail to reproduce one (the most important possible finding — weaken). The symmetry is the mark of honest science because a conclusion is falsifiable only if a real test could go against it; a test you would run only if you were sure of the result is a ritual, not a test.

Exercise 39.27 †

The commentator misunderstands Theme 1 (and the structure of a sound case). The bite-mark cases convicted innocent people on a single overstated inclusion from a discredited method, presented as proof. The Mill Creek conclusion is the opposite shape: an excluded field (Salas, Dana, Renner, the stranger theory cleanly ruled out) plus an independent convergence of multiple validated threads, stated at "strongly supports" with its limits volunteered and never as "proves." "Only 'strongly supports'" is not weakness disguised as strength; it is the strongest honest verb a properly assembled case earns, and it is categorically different from a lone weak thread inflated into certainty. The refutation turns on (a) excluded-field-plus-convergence vs. single inclusion, (b) validated vs. discredited methods, and (c) calibrated verb vs. overstatement.

Exercise 39.29

The two opposite CSI-effect errors about a "strongly supports" conclusion: - Over-credulity: hearing "strongly supports" as "proves" — convicting on the feeling of certainty the pile of threads creates, ignoring the named limits. - Over-skepticism: dismissing the case as "not enough" because there is no single "smoking gun" that "proves" it — acquitting despite a genuinely strong, independently-converged, excluded-field case. The calibrated middle: the evidence is genuinely strong (an excluded field plus independent convergence) and it does not reach certainty; a mature juror holds both at once and weighs the totality against the legal standard without collapsing into either credulity or cynicism.

Exercise 39.31 †

Model response to the detective: "I understand the jury wants certainty, but I can't testify that the evidence 'proves' Mr. Keller did it — and if I did, a competent cross-examination would expose it as overstatement and could discredit the entire case. What I can say, truthfully and durably, is that the totality of the forensic evidence strongly supports that he killed Mr. Diallo and set the fire, and I will explain each finding and its limits clearly. Whether that meets the legal standard for guilt is the jury's decision, not mine — that's the ultimate issue, and it isn't a scientific question (Ch.30). My duty is to the court and to the accuracy of the science, not to either side's preferred verdict. Honest, calibrated testimony is also the strongest testimony, because it survives cross."

Exercise 39.33

A blind re-analysis (Ch.38) that failed to reproduce a converging thread is "the most important finding of all" because it would mean that thread may have been an artifact of the original analysts' bias rather than an independent fact — and a convergence's whole power depends on the independence and reliability of its threads. What to do with the rest: re-weigh the assembly without that thread, re-examine whether other threads shared the same contaminating source, and restate the conclusion at whatever (possibly lower) strength the surviving independent threads jointly support. You do not discard the whole case for one failed thread, but you do not preserve the old verb either; you recompute honestly. (This is falsifiability in action.)

Exercise 39.35

On the stand, the expert MAY say: "In my opinion, the totality of the forensic evidence strongly supports that Mr. Keller killed Mr. Diallo and set the fire; I cannot and do not state it as a scientific certainty, and here are the limits of each finding." They must REFUSE to say: "The evidence proves Mr. Keller is guilty." The difference in one line: the first states the weight of the evidence (a scientific judgment, calibrated and bounded); the second asserts a legal conclusion of guilt to a certainty the science cannot deliver — the ultimate-issue overstep.

Exercise 39.37 †

Exclusion in Mill Creek vs. Bloodsworth (Ch.6). Same logic: in both, the surest, cleanest forensic move is exclusion — DNA (and, in Mill Creek, cell-site and timeline) ruling a person out, which a single conclusive mismatch can do (Ch.1, §1.6). Different direction: in Bloodsworth, exclusion ran post-conviction to free a man already wrongly convicted and sentenced to death — exclusion as exoneration. In Mill Creek, exclusion ran during the investigation to clear the innocent (Salas, Dana, Renner) before a wrongful conviction could occur — exclusion as prevention. Same power (the field's surest voice, "not this person"), aimed at different points in the timeline: undoing an injustice vs. preventing one.

Exercise 39.39

The whole assembly on the validity spectrum. Strongest-foundation method used: the instrumental chemistry — GC-MS confirmation of gasoline (Ch.23) — and the dental identification of the body (Ch.17), both high on the NAS/PCAST spectrum, near single-source DNA in rigor. Weakest: the criminal profile (Ch.28), at the bottom — and it was correctly given zero weight (it pointed away from Keller). In between, weighted accordingly: the DNA mixture (quantified but interpretive), the cell-site location (real but area-level), the class toolmarks and soil/pollen association. A case built by treating all "forensic" threads as equal would be junk because it would let a debunked profile or a class-level association carry the weight of validated, quantified evidence — exactly the error (overstated weak methods) behind the wrongful convictions in Chapters 6, 16, 19, and 34. Honest assembly = unequal weighting by validity.

Chapter 40 — Worked Solutions (daggered † and odd-numbered exercises)

Exercise 40.1

Forensic career paths are the distinct professional routes through forensic practice — bench analysts in separate disciplines, crime-scene investigators, forensic pathologists and medicolegal death investigators, digital examiners, consulting specialists, and laboratory directors/QA staff — each with its own education, daily work, and certifying body. "Forensic scientist" is an umbrella term because no single person does what the title implies on television; the work is divided among many narrow professions, often across several agencies and institutions that do not report to one another.

Exercise 40.2 †

Board certification is a credential awarded by an independent professional board attesting that an individual has met defined standards of education, training, and tested competence in a specific forensic discipline, and has agreed to a code of ethics and ongoing requirements. It does attest to: (1) that this person passed a defined competency examination, and (2) that they follow the discipline's accepted practices (and agreed to its ethics code). It does not establish the one thing people most often assume: that the discipline's central scientific claim is valid. A certified examiner skilled in an unvalidated method is still presenting an unvalidated result; the validity question is answered by the NAS/PCAST yardstick, not by the credential.

Exercise 40.3

AAFS is the major U.S. professional society for forensic science (founded 1948), organizing the disciplines into sections, holding an annual scientific meeting, and publishing a peer-reviewed journal. It differs from a certifying board (e.g., ABC) in that the AAFS does not issue competency credentials to individuals; it is an umbrella society for engagement, research, and standards. Two functions: (1) convening the disciplines (meetings/journal/sections) for research and debate; (2) promoting standards and education (and it is affiliated with FEPAC, which accredits degree programs).

Exercise 40.5

The boards named in §40.3: ABC (American Board of Criminalistics) — laboratory criminalistics practitioners (DNA, drug chemistry, fire debris, trace, etc.); ABFO (American Board of Forensic Odontology) — forensic dentists; ABMDI (American Board of Medicolegal Death Investigators) — medicolegal death investigators; American Board of Pathology — physicians in forensic pathology (a medical specialty board, distinct from the criminalistics boards).

Exercise 40.7

The forensic pathologist is a physician (MD or DO) with residency training in pathology and a fellowship in forensic pathology, who performs autopsies and certifies cause and manner of death. The medicolegal death investigator (MDI) investigates the circumstances of death — at the scene and through medical/contextual history — and gathers the information the pathologist needs. The pathologist is the physician; the MDI supplies the contextual investigation. They are a team.

Exercise 40.9 †

Considerations both ways. A forensic science degree can be excellent if the program is rigorous in laboratory science (and FEPAC accreditation is a signal of substantial natural-science content). But a forensic degree that is light on chemistry/biology can leave a graduate underqualified for the very bench jobs it advertises — many crime labs, especially for DNA, specify defined coursework (biochemistry, genetics, molecular biology, statistics) that flows from national DNA-personnel standards. A chemistry or biology degree builds exactly that underlying competence and is often more flexible. The deciding feature: the substance of the curriculum — does it deliver the real science (and the reasoning about contamination, controls, and statistics) the bench requires? — not the word "forensic" in the title. Choose the program by what it teaches, the same way the book teaches you to judge a method by what it can establish.

Exercise 40.11

"The bench analyst's job is a courtroom job" because producing a result is only half the work; the other half is defending it under oath, to a jury, while an opposing attorney probes every assumption. Roughly speaking, the two halves are comparable in importance — a perfectly produced result that is overstated (or hedged into uselessness) on the stand can do injustice. The principles of the second half were taught in Chapter 30 (expert testimony, communicating uncertainty, the duty to the court).

Exercise 40.13

Mapping each clause: "natural science + coursework" → §40.2 (the underlying science, not the label "forensic," is the gate; the DNA coursework is explicit). "competency evaluation prior to independent casework" → Figure 40.2, rung 3, and Chapter 4 (competence is tested, not assumed). "must testify" → §40.4 and Chapter 30 (the bench job is a courtroom job). "division of the State Police" → Chapter 38's independence problem (the embedded lab). "disturbing material" → §40.4's toll (trauma exposure named in the fine print).

Exercise 40.15

DNA generates backlogs because its very power invites ever more submissions: the more cases DNA can help with, the more evidence gets sent for DNA testing, so demand outpaces capacity even as capacity grows. The most painful public example from Chapter 37 is the untested sexual-assault kit backlog — kits collected from victims that sat untested for years.

Exercise 40.16 †

Reasoning error: treating a credential (and experience) as evidence of the method's reliability — conflating the competence of the person with the validity of the technique. This is the credential halo (and an ipse dixit-style appeal to authority; cf. Chapter 1). For bite marks specifically, no quantity of certifications or years changes the fact that the method's central claim — that one set of teeth, and no other, made a mark — lacks foundational validity (Chapter 16). What would actually establish reliability: well-designed validation studies measuring the method's error rate (the PCAST question), which for bite-mark comparison do not support the claim. The credential attests to the person; the validity is a separate question the credential cannot answer.

Exercise 40.17

Two overstatements. (1) "A forensic science degree guarantees a crime-lab job" — false; bench jobs are competitive, often require specific natural-science coursework, and a forensic degree light on lab science can underqualify a graduate. (2) "Doing what you see on TV" — false; the real job is narrower, slower, court-bound, and emotionally costly, and no one does the all-in-one televised role. Honest rewrite: "A rigorous science education (forensic or natural-science) can prepare you for a competitive crime-lab career — which is real science applied to legal questions, mostly slower and narrower than television, centered on careful analysis and honest testimony."

Exercise 40.19

Three questions a forensically literate journalist should ask before accepting "match": (1) Single-source or mixture? A mixture is interpreted with a likelihood ratio and supports a contributor; it does not "match" one person the way a clean single-source profile can (Chapters 8–9). (2) What is the random match probability / how was the statistic framed? — and beware the prosecutor's fallacy (Chapter 9). (3) Was the result independently verified, and was the analysis done blind to the suspect's identity? (contamination, bias, lab quality — Chapters 4, 31). (Also acceptable: was the database hit confirmed by a fresh comparison sample?)

Exercise 40.21

Ethical hazard: backlog pressure to "just clear" cases faster is, functionally, pressure to spend less time per case — and the corners cut under time pressure are exactly the controls, careful interpretation, and double-checks quality depends on (Chapter 4). Regardless of the pressure, the analyst owes the science an uncompromised process and owes the people affected (both the accused, who could be wrongly implicated, and any victim, who deserves a correct result) the refusal to let the queue degrade the work. The duty is to the truth, not to the detective's timeline.

Exercise 40.23

The advocate's mirror-image bias: a defense, exoneration, or reform advocate can read every ambiguous result as exculpatory because that is the answer they want — the same mechanism as the prosecution-leaning analyst, pointed the other way. The single discipline that applies identically to both: state what the evidence supports at its true strength, no more, regardless of which side that helps. (The strongest innocence cases are strong because the science is unambiguous — a clean exclusion — not because the advocacy is loud.)

Exercise 40.24 †

The trauma-exposure reframing matters because treating cumulative trauma as a personal weakness to be hidden drives practitioners to conceal it, go without support, and either burn out or let degraded well-being affect their work — whereas treating it as an occupational exposure to be managed (like a chemist's exposure to solvents) makes it a normal, addressable feature of the job. The reframing implies an honest laboratory should provide support: access to mental-health resources, peer support, workload management/rotation, and a culture that does not stigmatize seeking help. The practitioners who last are the ones who keep feeling something and keep working — which the supports make possible.

Exercise 40.25

Before accepting, you'd want to know (Chapters 31, 38): whether the lab follows any context-management / blind-verification practices at all; how case information (the suspect, the theory) reaches analysts; whether results are independently verified; the lab's accreditation and proficiency-testing record; and how it handles disagreements with investigators. Personal practices to protect independence within it: ask to be given the evidence and the question while withholding the answer the detective wants; document your reasoning before learning case context; insist on verification of close calls; and be willing to state "the evidence does not support that" to your own colleagues. Naming the structural pressure (Chapter 38) is the precondition: you cannot resist a pressure you refuse to see.

Exercise 40.26 †

Most → least warranted: 1. (a) ABC-certified DNA analyst, single-source match with a stated RMP — MOST warranted. The method is foundationally strong and quantified (Chapter 7); the credential and the validated, numeric method align. 2. (b) ABFO-certified odontologist, dental identification from records — well warranted. Dental identification is a genuinely valid use of odontology (Chapter 17); credential and valid method align. 3. (c) Certified examiner, firearms identification "therefore certain" — weakly warranted / overstated. Firearms/toolmark identification is contested on the validity spectrum (PCAST found foundational validity not established; Chapter 15), and "certain" overstates it regardless of the credential. 4. (d) ABFO-certified odontologist, bite-mark comparison "therefore certain" — LEAST warranted. Bite-mark comparison is discredited (Chapter 16); the identical ABFO credential cannot rescue an invalid method. The ordering turns on the method's validity, not the credential — the chapter's central point.

Exercise 40.27

"A credential attests to a person, never to a method" protects against wrongful conviction by forcing the court to evaluate the method's validity separately from the examiner's impressive qualifications — defeating the halo that lets unvalidated testimony convict. Example: in the bite-mark exonerations (Chapter 16, e.g., Ray Krone) or the microscopic-hair-comparison failures (Chapter 19), credentialed, confident experts gave testimony that DNA later proved wrong. Had the courts asked "is the method valid?" rather than "is the expert qualified?", the unvalidated evidence would have carried far less weight, and innocent people might not have been convicted on it.

Exercise 40.29

This chapter advances all four themes; name at least two. Exclusion over proof: the ethical practitioner's first duty is to claim only what the evidence supports, and the adjacent-paths material shows DNA's exclusionary power as the field's instrument of correcting wrongful convictions. The validity spectrum: the chapter's core careers lesson is that a credential attests to a person, not a method — the same ABFO credential spans valid dental ID and discredited bite marks, and only the NAS/PCAST yardstick distinguishes them. (Also: cognitive bias — the credential halo and the embedded-lab pressure; the CSI effect — replacing the televised "forensic scientist" with the real distributed profession.)

Exercise 40.31 †

Career-plan model answer (structure; specifics will vary by the path chosen — example: forensic DNA analyst). - (a) Underlying science: molecular biology/biochemistry, with genetics, statistics, and the reasoning about contamination and controls; obtained via a natural-science or rigorous (ideally FEPAC-accredited) forensic degree with the required coursework (§40.2). - (b) Education ladder (Figure 40.2): natural-science degree (+ often a master's) → hired into a lab (often after an internship) → in-house training + competency test → independent casework + testimony → board certification + continuing competency. - (c) Certifying body: the ABC (molecular biology). The credential would attest to my tested competence in DNA analysis; it would not establish the validity of any method I use — that is settled by validation studies, not the credential (§40.3). - (d) A daily reality to prepare for: the backlog (and its pressure on quality), and/or testimony under hostile cross — the willingness to disappoint the attorney who called me (§40.4). - (e) A theme I'd carry: exclusion over proof — reporting a mixture as a likelihood ratio supporting a contributor, never "identifying" one, and holding that line on the stand (§40.6; Chapters 1, 9). (A strong answer for any other path mirrors this structure with that path's science, ladder, board, reality, and theme.)

Exercise 40.33

Self-audit (model approach; personal). A good answer (i) identifies which of the four ethical-practitioner components would be hardest for the writer under pressure — e.g., admitting "I don't know" on the stand, refusing to overstate to please a colleague, resisting a credential's halo, or planning around one's own bias; (ii) explains why honestly; and (iii) names one concrete habit to build now — e.g., practicing saying "that is beyond what this method can establish" aloud; pre-committing to documenting reasoning before learning case context; or seeking blind verification by default. The point is honest self-knowledge plus a specific, actionable practice — not a "right" answer.

Exercise 40.35 †

Closing reflection (model answer). "Forensic science deals not in certainty but in honesty about uncertainty" has meant, across the book: that the field's surest voice is exclusion, not proof (Chapter 1; the DNA exonerations, Chapter 6); that methods sit on a validity spectrum and must be held to "how do we know, and how strong, really?" (DNA's quantified strength in Chapters 7–9 versus the discredited bite mark in Chapter 16); that the gravest errors come from overstatement — the confident arson "expert" whose folklore helped execute Cameron Todd Willingham (Chapter 22), the 100%-confident fingerprint "match" that was 100% wrong (Mayfield, Chapters 14, 31); and that communicating uncertainty honestly under courtroom pressure is itself the decisive forensic skill (Chapter 30). For me, in whatever role I take, it requires the disciplined refusal to claim more than the evidence can bear — saying "consistent with" when it is, "I cannot exclude" when I cannot, and "I don't know" when I don't — even when a more certain answer would be more welcome. The science will keep changing; that discipline will not. (A strong answer cites at least three chapters and connects the sentence to a concrete personal commitment.)