Case Study 1 — The Golden State Killer: How a Distant Cousin's DNA Closed a Forty-Year-Old Case

A real, publicly documented case. The facts below are drawn from the public record; where this study reconstructs a step for teaching, it is labeled. The point of the case is not the horror of the crimes — described here only as clinically as the analysis requires — but the method: how investigative genetic genealogy (IGG) inverted the database problem, and why it is this book's emblem of validated forensic progress rather than mere spectacle.

1. Background: a serial offender who left his DNA and was never caught

Between the mid-1970s and the mid-1980s, a single offender committed a long series of burglaries, rapes, and murders across California. At the time, the crimes were attributed to several seemingly separate offenders working in different regions — known variously as the East Area Rapist in the Sacramento area and the Original Night Stalker in Southern California, among other names. Only later did DNA reveal that these series were the work of one man, eventually called the Golden State Killer.

The crucial forensic fact for this chapter: the offender left biological evidence at crime scenes, and that evidence was preserved. Once DNA typing matured, investigators developed a crime-scene DNA profile (Chapter 7) and were able to link the otherwise-separate crime series to a single source — a powerful early demonstration of DNA's ability to connect cases. But linking the crimes did not identify the man. For that, you need to match the profile to a person, and the standard tool for doing so came up empty.

The honest framing. Treat the crimes with the clinical detachment of a report. The victims were real people; the analytical interest here is solely in how the evidence was eventually used, and what that teaches about a method's validity.

2. Why CODIS could not solve it

For years, the crime-scene profile was searched against CODIS, the national DNA database (Chapter 7). CODIS is, in essence, a database of profiles from convicted offenders and arrestees plus unsolved crime-scene profiles. Its searching logic asks one question: is this exact profile already in our criminal database?

For the Golden State Killer, the answer was no — repeatedly, for years. The offender had apparently never been convicted of a qualifying offense that would have entered his profile into CODIS. (He was later revealed to be a former police officer, which makes the absence even more striking.) This is the structural limit of CODIS that §8.5 of the chapter dwells on: a database of known offenders is silent on an offender who is not in it. A perfect crime-scene profile matched against an empty space yields nothing. The case stayed cold not because the DNA was weak — it was excellent — but because the answer wasn't in the database being searched.

This is the problem IGG was built to invert.

3. The method: how IGG was applied

In 2018, investigators (working with genealogists) applied investigative genetic genealogy to the case. The steps, mapped to the chapter's §8.5 framework:

  1. A different kind of profile. The crime-scene DNA was re-analyzed to produce a dense SNP profile — the same kind of genome-wide single-nucleotide-polymorphism data that consumer ancestry companies generate — rather than the ~20 STR loci CODIS uses. STR profiles are ideal for matching one profile to another but carry almost no genealogical information; SNP profiles can reveal relatedness.

  2. A genealogy database, not a criminal one. The SNP profile was uploaded to GEDmatch, a public site where consumer-DNA users voluntarily upload their results to find relatives. Instead of asking "is the offender here?", the search asked "are any of the offender's relatives here?" It returned a list of people who shared enough DNA with the crime-scene profile to be distant relatives — on the order of second to fourth cousins — ranked by shared DNA.

  3. Family trees, built backward. Genealogists took those distant-cousin matches and, using conventional records — censuses, birth, marriage, and death records, obituaries, newspaper archives — reconstructed the matches' family trees, working toward the common ancestors the offender must also share. Triangulating from matches on different branches narrowed the field downward to a manageable set of candidate descendants.

  4. Demographics, then a candidate. Known facts about the offender — approximate age, sex, and geographic ties from the case — winnowed the candidate descendants. This process pointed to Joseph James DeAngelo, a former police officer then in his early seventies.

  5. The indispensable confirmation. Investigators did not arrest on the strength of the genealogy. They obtained abandoned DNA — genetic material from items DeAngelo discarded — and performed a conventional STR comparison against the crime-scene profile. That comparison confirmed the match. DeAngelo was arrested in April 2018.

[Reconstruction note] The precise sequence and the exact databases and records used in the actual investigation are documented in public reporting and court records; the five-step rendering above follows the chapter's teaching framework and the well-established public outline of the case. Treat it as the shape of what was done, accurate in its method, not as a forensic transcript.

4. The outcome

DeAngelo was charged with multiple murders and ultimately, in 2020, pleaded guilty to a series of murders and admitted to numerous additional crimes (many of the rapes were beyond the statute of limitations and were admitted rather than separately prosecuted). He was sentenced to life imprisonment. The case became the highest-profile early success of IGG and triggered a wave of similar cold-case resolutions using the same approach.

5. What the forensic evidence did — and did not — establish

This is where the case earns its place in this chapter rather than Chapter 7. Read the method honestly:

What IGG established: a lead — a name to investigate — derived from the genetic relatedness of crime-scene DNA to voluntarily uploaded consumer profiles, plus traditional genealogical research. That is a genuine and powerful investigative achievement.

What IGG did not establish: the identification itself. IGG produced no number for a jury and no match in the courtroom sense. The thing offered as evidence of identity was the conventional STR comparison between the crime-scene profile and DeAngelo's confirmed sample — the gold-standard method of Chapter 7, with its quantified random match probability. The genealogy pointed; the STR typing confirmed.

This division of labor is exactly why the chapter calls the case honest progress:

  • The method that is new and less validated (genealogical inference from SNP matches) is used only to generate a lead, where its error mode is "follow the wrong family branch" — a cost in wasted effort and intrusion on innocent relatives, not a wrongful conviction.
  • The method that makes the courtroom claim (STR comparison) is old, validated, and quantified.
  • The new tool never pretends to be the conclusion. It hands off to the strongest method in the field.

Contrast this with the trajectory of bite-mark analysis (Chapter 16), where a method's confidence outran its proof and the unvalidated comparison was itself offered to juries as identification. IGG did the opposite: it kept the unvalidated step in the role of lead-generation and reserved the courtroom claim for the validated step. That structure is what methodological honesty looks like.

On the validity spectrum. IGG is not best placed on the NAS/PCAST spectrum as an identification method, because it is not offered as one. As a lead-generation method confirmed by gold-standard STR typing, it rests on extraordinarily solid ground. The STR confirmation that follows it sits at the top of the spectrum (Chapter 1). Conflating the two — judging IGG as if the genealogy itself were the courtroom evidence — misreads the method.

6. The shadow side: why a triumph is still contested

The same case that demonstrates IGG's power also opened the ethical questions of §8.6. DeAngelo's relatives — distant cousins who had uploaded their DNA to find family — became, without their knowledge, the thread that led police to him. None of them consented to a criminal investigation; their voluntary, personal genealogy created a search capability over an entire extended family. The case is therefore simultaneously:

  • a triumph (a decades-cold series of the most serious crimes, solved),
  • a model of methodological honesty (lead vs. identification kept distinct), and
  • the opening of a surveillance capability that the law had not anticipated and only began, afterward, to regulate.

Holding all three judgments at once — and refusing to collapse into either uncritical celebration or reflexive alarm — is the calibrated posture the chapter asks you to develop. We take up the regulation and ethics fully in Chapter 38.

Discussion questions

  1. Explain, to someone who has never heard of CODIS, why a "perfect" crime-scene DNA profile could leave a case cold for forty years. What single fact about CODIS's contents is the whole answer?

  2. The chapter insists IGG "does not identify anyone." Defend that statement using the role of the abandoned-DNA STR comparison. What was offered to the court as proof of identity, and why does the distinction matter?

  3. Why is IGG's failure mode ("follow the wrong family branch") a fundamentally different kind of risk than the failure mode of bite-mark analysis ("convict on an unvalidated comparison")? Tie your answer to which step does the courtroom work.

  4. Construct the strongest argument that IGG should be permitted for serial murder cases like this one, and the strongest argument that permitting it here makes it hard to deny for lesser crimes. (§8.6 scope creep.)

  5. DeAngelo's relatives uploaded their DNA for personal genealogy. Identify every party whose privacy was implicated by that upload, and explain why "consent at a distance" is the hard ethical knot.

  6. A commentator says, "The Golden State Killer case proves DNA can solve anything now." Using this case and the limits from the chapter (touch DNA, mixtures, degradation), write a calibrated correction.

  7. Where would you place IGG on the validity spectrum, and why is your answer different from where you would place the STR confirmation that follows it? (This is the question the case is designed to teach.)