It’s an open-and-shut case. The fingerprints found at the scene of a crime match those of the man the police have just arrested. So he’s guilty, yes? Not so fast . . . No one has ever proved that every finger possesses a unique pattern of ridges and grooves, or that a single smudgy print can be matched to just one person, says historian of science Simon Cole. In his new book, he argues that contrary to public opinion, there’s scarcely any science to back it up. Indeed, what little evidence thereis suggests that fingerprint examiners may match prints to the wrong person as much as 20 per cent of the time. Adrian Cho asked Cole if the cops and the courts are likely to come clean.
Are you arguing that fingerprinting doesn’t work or that no one has ever scientifically proved that it works?
I’m arguing that nobody has ever scientifically measured how well it works.
Advertisement
But surely a century of experience proves that fingerprinting is extremely reliable?
That’s not really a very scientific argument, is it? People used to believe that the planets revolved around the Earth, and that’s not considered scientific evidence. Remember, a trial is not an experiment because you have no way of validating your results. If you falsely match a fingerprint, then the guy is convicted and goes to jail. Who’s going to believe he’s really innocent unless there’s some extraordinary circumstance, like DNA comes along or someone else confesses? There’s no way you would ever know.
But if fingerprint examiners routinely implicated innocent people, wouldn’t defence lawyers have shouted about this years ago?
The question is to what extent defence lawyers have internalised the idea that fingerprint evidence is scientific and unchallengeable, so that when they see it, they use their scarce resources elsewhere. One good example of this is the scandal surrounding the New York State Police, when it turned out that state troopers had been fabricating extremely crude, transparently fake fingerprint evidence and using it to secure numerous murder convictions. In no case did the defence lawyers question the evidence or hire experts to look at it. Also, the idea that there have been no false convictions is a myth. There was an inadvertent one in the early 1980s in Minnesota, which was a clear misidentification. In my book, I mention another one back in the 1930s. And there may be lots more that we don’t know about.
How did fingerprinting gain this reputation for infallibility?
In the early 20th century, fingerprinting went from being a new thing to being accepted as virtually infallible. The reasoning was that no two fingerprints are alike, so fingerprint identification must be certain. Which seems to follow, but doesn’t. Establishing that all fingerprints are unique doesn’t establish how reliable fingerprint evidence is. In the earliest trials, I don’t think anybody appreciated that, including fingerprint examiners themselves. Experts also testified in terms of absolutes. They’d be asked: “Is it your opinion that this print and that print were made by the same person?” And they would say: “It’s not my opinion, it’s a fact.” Even today they don’t use probabilistic language.
In your book, you say that using a full set of 10 well-inked prints is a more trustworthy than lifting a single smudged fingerprint from a doorknob. So you wouldn’t argue against using the sets of 10 prints to catalogue convicts in prison?
No, I don’t think I would. Then they’re on much more solid ground. Assuming the prints are of good quality, you have a lot more detail to work with. Again, it isn’t necessarily proven that no two sets are the same, but it seems a lot more likely.
Can’t the FBI or Scotland Yard test the similarity of fingerprints by comparing prints taken from their huge archives?
The FBI did that with 50,000 prints. But they undermined the test by taking a fingerprint compared to itself as the standard for a match. They measured the similarity score that that generated. Then they ran that print against all the other 49,999 and generated similarity scores. And the similarity score for the print run against itself was higher. So they said: “Well, this proves that this print is more similar to itself than to any of the other 49,999.” The problem with that is in forensics you’re not comparing a print to itself. You’re comparing a print to another print made from the same finger. What I think they should have done is taken one print of a finger and then another print of the same finger, not the same print, and seen what that similarity score was.
So why doesn’t somebody simply do the experiments again that way?
I think the sociological reason is that fingerprint examiners have no incentive to do such a test, because by the 1990s, when such a test became technologically possible, fingerprinting had been wholly accepted by the courts. So why would they bother? Most professional fingerprint examiners aren’t the kind of scientists that would be equipped to carry out such an experiment anyway. For the statisticians and computer scientists who might be qualified, it’s not a very sexy research problem. Where’s the glamour in establishing the variability of fingerprint patterns, especially when everybody’s all excited about DNA?
How are fingerprints matched in the first place?
You run a “latent” print, a print taken from a crime scene, against your database and the computer spits out a list of candidate matches. Then the human examiner looks at each candidate on the screen to see whether points of detail are consistent with the latent print. The examiner looks at what are called ridge characteristics-basically, places where a ridge either ends abruptly or forks into two ridges. More recently, some examiners have also begun looking at smaller details such as bumps on ridge, pore location, and so on. Many countries require some fixed number of matching details to declare a match, but in North America and Britain there is no numerical standard. Sufficient detail is however much detail the examiner thinks is enough.
What’s the evidence on the reliability of those examiners?
A new proficiency test, designed in conjunction with the International Association for Identification, was conducted by an independent testing service starting in 1995. And that first test yielded this horrible error rate of 22 per cent for false positives-matching a print to the wrong print. In the law enforcement context that could mean a false conviction. That got everyone upset in the fingerprint community. Subsequent tests revealed error rates ranging from 3 per cent to 15 per cent. You can always deconstruct proficiency tests and complain that they’re not a real representation of what’s going on. But they’re the best guess we have.
In 1993, the US Supreme Court said that scientific evidence had to be testable, subject to scrutiny by other scientists, accompanied by quantitative estimates of uncertainty, and generally accepted by the scientific community. Does fingerprinting meet these standards?
I don’t think it meets any of them. I don’t think fingerprint examiners have tested their claim-which is that they can reliably match a latent print to one and only one finger to the exclusion of all others in the world. They continually refuse to calculate an error rate for fingerprinting, believing that having a zero error rate will make them more scientific, even though among hard scientists a zero error rate would be cause for suspicion. Peer review and general acceptance are somewhat fuzzier criteria. The question is whether fingerprint examiners have created such a closed community that peer review is meaningless. There are ways in which astrology can meet the peer review requirement.
You have testified about the reliability of fingerprint evidence in court . . .
Yes. The trial was about a robbery in which two fingerprints taken from the alleged getaway car constituted very important evidence against the defendant. I testified at a pre-trial hearing called a Daubert hearing, in which a judge decides if a certain kind of evidence is scientific or not. The issue was both whether fingerprint evidence should be admitted as scientific evidence in this case and whether my testimony and that of the other defence experts should be admitted to challenge fingerprint evidence. The judge ruled that fingerprint evidence was reliable. As to whether it is scientific evidence or simply some kind of technical, experience-based evidence-such as an auto mechanic’s assessment of an engine-he did not decide. But our testimony was not permitted in the trial, essentially because we’re not fingerprint examiners.
What will it take to get courts to do a rethink on fingerprinting?
It may only take getting to an appellate court. There are reasons why an appellate court might be more receptive to the arguments that are being made than a trial judge might be. David Faigman of the Hastings College of the Law in San Francisco has predicted that a judge will rule fingerprints unscientific within a year. I wouldn’t make that prediction myself, because all judges certainly have compelling reasons not to be the one who overturns fingerprint evidence and open the way for a flood of appeals.
Are other types of forensic evidence more reliable?
In all other forensic sciences they use quantifiable uncertainties-that is, the evidence is given as probabilistic evidence. That’s what makes fingerprint evidence so problematic. It’s the only technique that doesn’t give evidence in terms of probabilities. But much of what we call forensic science-hair and fibre analysis, for example-turns out to have almost no science behind it at all. I think we’re going to see forensics getting a lot of attention in the next few years, prompted by DNA, whose relationship to mainstream science is much more solid than some of these other techniques.
Wouldn’t it be intolerable to give up on fingerprinting as a crime-solving technique?
I haven’t necessarily advocated throwing out fingerprinting altogether. What I have said is that it doesn’t meet the standards for science that the Supreme Court has laid down. And I have said that we should be careful about allowing fingerprint examiners to testify in terms of absolute certainty, rather than in terms of probabilities. We should be careful about letting them use the word “science” unless they really establish that fingerprinting is science. In general, we need more supervision of fingerprint identification. We can’t just take fingerprint examiners’ word for it in every case.
Suppose you were falsely accused of a crime based on fingerprint evidence. What would you do?
Believe it or not, I would call up some of my most implacable enemies, some of the leading fingerprint examiners in the country. They are very angry with me now and might like nothing better than to see me falsely convicted. Yet they are people of integrity. I’m certain that if they saw a false match, they would declare it a false match. That’s an important point, because I’m not accusing these people of being malevolent or acting in bad faith. I think the vast majority of them are acting entirely in good faith. On various occasions the best of them have gone out of their way and exposed themselves to all sorts of pressures and made people angry by exposing false matches. They’ve done so in the interest of truth and justice, and in the belief that their technique is not well served by getting erroneous convictions. If it were a bad match, I think they would exonerate me. But if even the top examiners didn’t detect it as a bad match, then I’d be in real trouble.