Fingerprints, Part 2: A bunch of stuff no one wants to find out except criminal defendants and the wrongfully accused.
It’s been a while. I hope you all remember the last post on fingerprints because a certain amount of that background is necessary to know why there isn’t really any science backing up fingerprint evidence. Although there was a lot of movement in this area as of 2008ish so something more might have come out in the past few years. If someone knows they should speak up.
The SCIENCE of FINGERPRINTS!
There are a lot of “scientific studies” on damn near anything you could imagine. Merck had a whole bunch of studies that showed Vioxx was safe. Dalkon had some studies showing that the Dalkon Shield was safe. Those two guys from the University of Utah who discovered cold fusion had a study showing they were right. Some guy from the Enlightenment had a study showing how much the soul weighed. Now we know that all these things were wrong. The studies were poorly designed, poorly executed, cherry picked or in the case of Dalkon, outright faked.
The studies surrounding fingerprints suffer from the same difficulties with some compounding factors rendering it almost impossible to get a true comprehensive review of the claims of fingerprints. It is helpful to begin by listing the number of claims that must be true in order for fingerprint identification to work.
1. Fingerprints must be permanent and unchanging.
2. Fingerprints must be unique.
3. Fingerprints must be capable of transferring that uniqueness to another object.
4. The unique trace must be capable of accurate visualization.
5. The unique trace must be capable of accurate transfer to a third surface.
6. A fingerprint examiner must be able to match the twice-transferred trace accurately to the original subject’s fingerprint.
I could break #6 down into a few more points – can a person be trained to match prints accurately, how much training is required, etc – but a well designed study could just do a whole lot of these points at once.
As of now, to the best of my knowledge, there have been well designed studies and accepted studies on very few of these. Fingerprints appear to be permanent and unchanging. Someone’s fingerprints as a youth should be, barring some catastrophic injury, the same as when they are an adult or elderly. Injury, scars or amputations all tend to add uniqueness rather than render the fingerprints “changed” so don’t get any clever ideas.
Identical twins appear to have non-identical fingerprints lending some credibility to #2. However, just because twins fingerprints are different does not confirm their uniqueness, just that fingerprints aren’t based entirely on genetics. Beyond twin studies there haven’t been any real studies of #2. The evidence cited by experts and the government when this is pointed out is “well if they weren’t unique where are the false positives!?” The short answer is, there are a lot of false positives but other evidence tends to render the false ID obviously impossible. Then, once it is a known impossibility, the experts redo their analysis and determine there was operator error and the fingerprints were NOT EVER a match.
#3, #4, #5, and #6 are all part of the process of collecting fingerprints ‘in the wild’ so to speak. There haven’t really been any well designed comprehensive studies on the whole “analysis” aspect of fingerprints. This is for a lot of very good reasons. First, it would be expensive. A study of this type would require a large amount of samples, a large number of fingerprint examiners, and a whole unbelievable shitload of man hours. Good analysis by good examiners takes hours and hours for a single print. Second, it would require the cooperation of the FBI, and the national board that certifies fingerprint examiners. Because the only purpose of this study would be to cast doubt on the reliability of fingerprint evidence, they will never ever ever cooperate. In fact, a major and comprehensive study of fingerprint reliability was commissioned by the FBI years ago but canceled in a final stage, likely because the results were less than perfect. We’ll never know the actual results. Third, the only people who are interested in this information are people accused of crimes especially those based “cold” fingerprint hits, the occasional defense lawyer, certain scientific types who insist on proven accuracy and reliability, and the just plain ornery and disagreeable, like myself. To paraphrase some great author “the story was widely known and widely disbelieved.”
As far as I know here are the most recent “scientific” studies which done supporting fingerprint evidence.
The 50/50 Study
Before anyone asks the government named the “study,” apparently with no sense of irony. Technically the full name is the 50k by 50k study but in court everyone was tossing around 50/50 so….Heh.
This study involved the random selection of 50,000 fingerprints from white males bearing a “left-sloped whorl pattern” from the FBI computer database. Then all these prints were compared to each of the 50,000 total. The second part of the study was to artificially create a fake latent which was only about 20% of a full sized print. This is apparently the average size of a recovered print. So what did they find? The only false matches were “data entry” errors – either duplicates or prints overlapping onto other prints entry spaces. Based on this the FBI concluded that the odds of a full print being falsely matched to another full print was, brace yourself, 1 in 10 to the 86th power. For the partial print? 1 in 10 to the 16th. Fingerprints blow DNA out of the goddamn WATER. How could I doubt these well meaning and omniscient FBI computers? How could anyone?
Unlike the FBI, who obviously hoped to wow the judge in the evidentiary hearing with insane crazy person numbers, anyone with any kind of statistical or scientific background would instantly balk at them. So what are the problems with the study? Well firstly, I’ve already discussed that computer matches aren’t matches at all but instead statistical modeling and collation of fingerprint type and appearance and then specific returning of that information to the degree and accuracy that are at the discretion of the person operating the search. Secondly, the “study” was just the computer matching identical pictures. Computers are very good at this! If you take an exact file it can tell you if another file is also that exact file! This is all that study shows. Is picture X the exact same file as picture Y? No one doubts computers can do this. Finally, no one attempts to submit this as evidence. The only evidence ever attempted to be admitted at trial is examination by human beings matching fingerprints. Why is this? Because computers can’t match “field” prints in any way. Its fucking complicated. The “simulated latent” was just a piece of the one of the images cropped to a smaller size, also a relatively trivial task for the computer. Although the computer did not do nearly as well at this task as you can tell from the reduction in the power of ten from the 80’s to the teens. If AFIS, the FBI system, only ever returned one print when you fed in a piece of data and that print was a match to some known accuracy level then something AFIS does might mean something. But it never did, never does, and who the fuck knows the error rate!
Next, the study presupposes its own conclusions. It takes prints of known identity and matches them to prints of known identity. The system is supposed to be used to take prints of unknown identity to those of known identity. This study didn’t even try to do something like that by, for example, shielding the system from knowing the print when it was searching. This study inherently assumes the accuracy and reliability of fingerprint ID, something it is supposed to be proving. Whoops! Turns out this study doesn’t really show anything at all! It’s meaningless, yet often cited for its statistics! ARGH!
Maybe if they had taken a second set of prints, taken under “field” ordinary conditions, then submitted them to AFIS, then shown perfect matches this study might mean something. But they didn’t do that. It is very unlikely the system is able to do this given its design – its not supposed to try and match prints!
The Meagher Survey – The Fix is in!
The survey involved 3 questions that were basically all just truisms of fingerprint ID. Do you use fingerprints, can they be used to ID, are there false positives? Every agency answered the way they were supposed to. After all, if they didn’t why are they in the business of using fingerprints to ID people? It is also accepted fact that no two fingerprints can match unless they were left by the same person
This survey also sent out a couple of latent prints and a 10-print card from an actual case being appealed to a bunch of fingerprint analysis agencies around the country. Around 35 or so responded. 7 responded there was no match between the first latent and the 10 card. 5 responded there was no match between the second latent and the 10 card. The FBI responded by sending enlarged photos of the latent prints and 10 card with all the points of matching comparison outlined for the agencies along with the conclusion that they matched. Included was a letter to review their findings in accordance with the FBI analysis. All the “wrong” agencies changed their minds! What a shock!
What are the problems with this? It was an actual case! We have no idea of the “true” source of the latent fingerprints. The suspect maintained his innocence and that they weren’t his fingerprints. That’s a lot of disagreement about the match for a field with a 0% error rate. The court in this case said that there being no false positives was very telling. Too bad for the defendant he was asserting that the ID in his trial was a false positive. But fuck that guy right he’s a criminal.
If you wanted a real study here’s what you’d do – COLLECT THE FUCKING STATISTICS ON FINGERPRINT UNIQUENESS. COLLECT THE GODDAMN STATS ON HOW MUCH UNIQUENESS IS LOST WITH ALL THE TRANSFERS.
Oh and probably the most important thing:
DO A FOR REAL DOUBLE BLIND STUDY WITH KNOWN SOURCES ON “REAL CIRCUMSTANCES” LATENTS FOR CHRIST’S SAKE!
It would take one, ONE GODDAMN STUDY, to know if fingerprints were potentially fraudulent or not. Once you knew it was possible to ID people you could test a shitload of other variables. I’m just gonna copy paste from some real expert’s analysis of the variables. Its from here: Itiel E. Dror, Contextual information renders experts vulnerable to making erroneous identifications, 156 Forensic Sci. Int’l 74 (2006).
“Each one of these differences serves to diminish, obscure, distort, or eliminate information necessary to the comparison process.
1. Size: The latent print is partial, not complete. Typical latent prints contain only about one fifth of the finger surface contained in an inked or scanned print.
2. Location: Some parts of a finger’s surface are more informative than others, and the latent print’s smaller area may contain little useful detail.
3. The latent print may be deposited on a dirty surface, which obscured critical features.
4. The latent print may be smudged or smeared.
5. The latent print may be overlaid or underlaid by other prints.
6. The latent print may be deposited on a noisy surface, such as wood, which itself consists of ridges and grooves that may be confused with those contributed from a finger.
7. The medium in which the latent print was deposited, such as sweat, water, blood, or oil, may interfere with its definition.
8. The amount of pressure and the direction of the pressure between finger and surface produce distortion in the latent print.
9. A curved or irregular surface distorts the patterns of the finger.
10. The procedure used in lifting the latent print usually causes some loss in print detail.”
As of right now, the best challenge to fingerprint evidence has been defeated many times over in court, despite the total lack of the necessary standards. In fact, I think the Supreme Court has outright said that “certain types” of evidence, when specifically referring to fingerprints, were so old and entrenched that they would be admissible, despite any scientific flaws or disputes, merely because of tradition. WE ARE THROUGH THE GODDAMN LOOKING GLASS HERE.
The Current State of Fingerprints – Here and Aroooound the Glooooooobe (Edited to add the actual # of points people use)
They are used pretty much everywhere. However, standards vary WIDELY from country to country. Would it surprise you to know that the USA has no real standards for how many points of comparison have to match in order to determine a match? Given this forum, probably not! The government has gotten convictions on as few as 5, yes that’s right 5, points. That was a bit of an outlier and now the expected, but not enforced or well documented, standard is a very healthy 8 points! England follows this “system” of “making up whatever we want” as well. Sometimes cited is that they use a 16 point match but that is not a hard and fast rule and they will make ID’s on the same flimsy basis as US authorities. Further, even with their more rigorous standard there have been a lot of false positives coming from England recently. Continental Europe tends to follow a system requiring points numbering in the teens or twenties but I don’t have the numbers right in front of me. Ed: I do now! France requires 24 generally! Brazil has a pretty high standard as well I think. Ed: Brazil and Argentina generally require 30 (30!) points of similarity.
There is no required fingerprint examiner certification in the USA, just as a final word.
Funny Stories Time – Not that funny edition
In 2004 some terrorists blew up a train in Madrid. You may have heard of it. It was called the Madrid Train Bombing. It was a pretty big deal at the time. 9/11 of Europe some were saying. The Spanish authorities recovered parts of the bomb used in the attack. They had a fingerprint which they sent out to everyone. The FBI, miraculously, matched the print to a US native. ISLAMS AT HOME! ALERT! It was some Oregon lefty lawyer, a liberal and a LIKELY cover! The guy was harassed, labeled a terrorist, and had to fight, at length, for his civil rights. Problem is, the guy didn’t have a passport. And the more the FBI investigated the more it appeared he hadn’t even left the country in the last 10 years! But the FBI was certain. The print was a match and anyone who said otherwise JUST DIDN’T UNDERSTAND THEY WERE EXPERTS! Then the Spanish, after hearing the FBI’s ridiculous conclusion began to do a more indepth and careful analysis of the print. Soon they matched it to a guy who actually lived in Spain and was eventually found to be a part of the group who blew up the train. STILL the FBI was certain. They really fought with the Spanish authorities to keep our poor Oregon lawyer under suspicion. Eventually the FBI came around, but somehow it wasn’t their fault or the fault of fingerprints. I don’t know. Shit was crazy.
Stephan Cowans was convicted of murder of a cop based largely on fingerprint evidence. Later he was totally exonerated when DNA proved it couldn’t have been him. Whoops!
Ricki Jackson was convicted of murder based primarily on fingerprint evidence. The FBI actually exonerated him and came out to say the fingerprint ID was wrong. Whoops again!
I don’t have much more info on these two bros because my access to academic databases long ago dried up. But if anyone has subscriptions to newspaper archives or academic databases just search for these names and you’ll laugh and laugh!
Addendum – Other “Prints”
Footprints, ear prints, palm prints, lip prints (I knew I was forgetting one!) all that shit IS FAKE AND MADE UP (as of right now, I don’t want to rule it out I guess!). There is no evidence supporting that shit is unique and identifiable. CSI folks! CSI!