Case Study 14.2 — When a Fingerprint Convicts the Wrong Man: Stephan Cowans (and Shirley McKie)

Sourcing and tone. This case study draws on the public record of Commonwealth v. Stephan Cowans (Suffolk County, Massachusetts), a conviction later overturned, and on the publicly documented Scottish case of Shirley McKie. Stephan Cowans was exonerated; the State of Massachusetts and the Innocence Project record his case as one in which a person convicted in part on a fingerprint identification was later cleared, including by DNA. We treat the underlying crimes soberly and stay within documented public facts. The purpose is to teach a single point that complements Case Study 14.1: where Mayfield shows a false identification caught before trial, Cowans shows one that was not — a fingerprint misidentification that helped send an innocent man to prison, and what finally undid it. We note where a detail is summarized.

Background

In 1997, a police officer in Boston, Massachusetts, was shot and wounded during the pursuit of a suspect. The assailant fled, at one point passing through a residence and reportedly drinking from a glass mug and handling other items. Investigators recovered a latent fingerprint and biological material (saliva on the mug, a hat). Stephan Cowans, a young Black man from the area, was charged with the shooting.

At trial, the prosecution presented two kinds of evidence that bear directly on this chapter. First, eyewitness identifications (a topic this book takes up in full in Chapter 32, on the unreliability of confident witnesses). Second — and decisively for our purposes — a latent fingerprint identification: examiners testified that a print recovered from the scene was Cowans's. In 1998, Cowans was convicted and sentenced to a long prison term. The fingerprint, presented as the kind of objective, scientific certainty juries had been taught to trust, helped corroborate the eyewitnesses and seal the case.

The forensic evidence and its undoing

The fingerprint identification against Cowans was, the later review established, wrong. It was a latent-print misidentification of the same family as Mayfield's — an examiner's confident conclusion that two impressions shared a source when they did not. But unlike Mayfield, the error was not caught by a dissenting foreign police force within weeks. It survived a trial, a conviction, and years of imprisonment.

What ultimately undid it was the one forensic method in this book with a rigorous, quantified foundation: DNA (Chapters 7–9). The biological material from the scene — the saliva on the mug and other items the assailant had handled — was preserved. Years after the conviction, with the assistance of the New England Innocence Project, that material was subjected to DNA testing that had not been available, or not been done, at the time of trial.

The DNA excluded Stephan Cowans. The saliva and associated samples did not come from him. Confronted with the biological exclusion, authorities re-examined the fingerprint identification — and the latent print, too, was found not to be Cowans's. The "scientific" evidence that had corroborated the eyewitnesses collapsed. In 2004, Stephan Cowans was exonerated and released, having served roughly six years for a crime he did not commit. His case is recorded as the first in Massachusetts in which a person convicted with fingerprint evidence was later cleared by DNA — and one of the early national examples of a fingerprint misidentification exposed by the more rigorous method.

Read the Evidence

text FIGURE CS14.2 — "Two kinds of certainty" [after the Stephan Cowans case, public record] THE ITEM A latent fingerprint identified at trial as the defendant's, and biological material (saliva) from the same scene, tested years later. THE CONTEXT The fingerprint was presented as objective scientific proof and corroborated eyewitness identifications; the DNA was tested post-conviction, with Innocence Project involvement, under known ground-truth conditions. WHAT IT SHOWS The fingerprint identification — a structured human judgment — was wrong. The DNA — a quantified comparison with a characterized error structure — excluded the defendant and pointed away from him. WHAT IT DOESN'T The fingerprint's wrongness was invisible at trial; nothing about the confident testimony signaled the error. Without preserved biological evidence and post- conviction testing, the misidentification might never have been caught. THE INFERENCE Two methods, two places on the validity spectrum. The latent comparison produced a confident false positive; the DNA corrected it. This is the spectrum made visible inside a single case. THE LESSON A fingerprint identification is not self-certifying. When a more rigorous method contradicts it, the fingerprint is the one that yields — because its error rate, though low, is real, and its conclusion is an opinion, not a measurement.

A second lens: Shirley McKie (Scotland)

A complementary case from outside the United States shows the same failure mode reaching a different kind of victim — and is worth knowing because it widens the lesson beyond one country's lab.

Shirley McKie was a Scottish police detective. In 1997, during a murder investigation, a latent print found at the crime scene was identified by fingerprint examiners as hers — placing her, by their account, inside a location she said she had never entered. McKie maintained the print was not hers. The identification, made by experienced examiners within an established bureau, was nonetheless treated as authoritative, and McKie was prosecuted (for perjury, for denying she had been there). She was acquitted in 1999 after independent examiners disputed the identification, and the matter became a prolonged public controversy in the United Kingdom about the reliability of fingerprint evidence and the institutions that vouch for it. Years later she received a financial settlement from the government.

The McKie case teaches two things the Cowans case does not foreground. First, the misidentified person was herself a police officer — a reminder that the error is about the method under bias, not the character of the accused. Second, it shows a fingerprint bureau closing ranks around a contested identification, the institutional analog of non-blind verification: once an authoritative conclusion is on the record, the pressure runs toward defending it rather than re-testing it. That is the same dynamic §14.4 names inside a single lab, scaled up to an institution.

What these cases did — and didn't — establish

Read together, Cowans and McKie establish something narrow and important, and they do not establish the lazy overread.

They do establish that latent fingerprint identification — a foundationally valid method (§14.6) — produces real, consequential false positives, and that those errors can survive the very safeguards (verification, courtroom presentation, cross-examination) meant to catch them. They establish that the "objective science" framing under which fingerprints were sold to juries was itself part of the danger: a jury told a print is a certain scientific match has little reason to probe it, and the eyewitness errors riding alongside (Cowans) gain false corroboration.

They do not establish that fingerprint comparison is junk science on a par with bite marks (Chapter 16). This is the distinction the chapter insists on. Bite-mark comparison rests on an unsound premise — that human dentition is demonstrably unique and that skin reliably records it — that empirical testing has not supported. Fingerprint comparison rests on a sound biological premise (friction ridge detail really is highly variable and persistent) and a method that, in well-designed studies, performs the task at a measurable, usefully low, non-zero error rate. Cowans and McKie are instances of that non-zero rate doing its damage — and of the method's vulnerability to bias and overconfidence — not evidence that the method is baseless. The honest reading is the chapter's: a very good silver standard, capable on a bad day of being confidently and completely wrong.

The decisive contrast in Cowans is the one to carry: the fingerprint convicted; the DNA exonerated. That is the validity spectrum (theme two) rendered as a man's six lost years. When the rigorous, quantified method spoke, it overrode the structured human judgment — because only one of the two had earned the right to that authority.

The lesson

Three lessons, all central to this chapter:

1. A confident fingerprint identification can be wrong, and the courtroom is not a reliable filter. Mayfield's error was caught in weeks by a dissenting agency; Cowans's was caught after years, only because biological evidence happened to be preserved and post-conviction testing became possible. The system did not self-correct in the ordinary course — an outside, more rigorous method did. Do not assume that a print that survived trial was therefore correct.

2. "Scientific" framing is part of the risk, not a guarantee against it. Presented as objective certainty, a fingerprint identification discourages the scrutiny it needs and lends borrowed credibility to weaker evidence riding alongside it (the eyewitnesses in Cowans). The disciplined examiner's hedged language — "agreement in N minutiae, no unexplained differences, in my opinion same-source" — is not lawyerly throat-clearing; it is the honest representation of a method with a real error rate (§14.6).

3. The error is in the human process, and it crosses borders and institutions. Cowans (a U.S. lab) and McKie (a Scottish bureau) show the same failure — a confident latent-print misidentification, defended by the institution that made it — in two countries. As with Mayfield, the fix is not better ridges but better procedure: blind verification, context management, and a culture that treats "inconclusive" and "I was wrong" as professional, not shameful (Chapters 31, 38). This is exclusion over proof (theme 1) and the validity spectrum (theme 2) in a single, sobering pairing — and it is exactly why the cold case's gas-can latent was recorded as inconclusive rather than forced into a match (the Case File).

Discussion questions

1. In Cowans, both a fingerprint identification and eyewitness identifications pointed to the defendant, and DNA later excluded him. Explain how a wrong fingerprint identification can lend false corroboration to wrong eyewitness testimony — and why two erroneous pieces of evidence "agreeing" is not twice the proof (connect to Figure CS14.1 in Case Study 14.1).

2. The Cowans misidentification survived a trial and a conviction; it was undone only by post-conviction DNA. What does this tell you about the reliability of verification and cross-examination as safeguards against latent-print error? Tie your answer to §14.4 and to the role of preserved biological evidence (Chapters 7–9).

3. Using §14.6, explain why Cowans and McKie are evidence of a non-zero error rate in a foundationally valid method, not evidence that fingerprinting is junk science like bite marks (Chapter 16). What specifically distinguishes the two methods' foundations?

4. In the McKie case, the misidentified person was a police detective, and a fingerprint bureau defended the contested identification. Explain how institutional defense of a conclusion is the large-scale analog of non-blind verification, and what reform (Chapter 38) would counter it.

5. The decisive corrective in Cowans was DNA. Using the validity spectrum (Chapter 6; §14.6), explain why DNA was entitled to override the fingerprint identification rather than the reverse, and what that ordering teaches about how to weigh conflicting forensic results.

6. Compare the outcomes of Mayfield (Case Study 14.1) and Cowans. Both were innocent; one was cleared before trial, one after years in prison. What features of each case determined when the error was caught — and what does the comparison imply about how much we should trust a fingerprint identification simply because it was never challenged?