What is enough.

[Gerald Clough]

Gerald, can you clear up one small issue please a yes or no answer will suffice. If you had every set of fingerprints of every one living today would you expect to find a small area of ridged skin repeated on another person (say 1/2 an inch square)?

I see no point in including every person who has or has not yet lived as they are less likely to leave a fingerprint at a scene of crime today.

Well, I can't say. I guess I'd not be shocked to the point of disbelief to find that small area (assuming a collection of details within the "normal" range of number and complexity) that on L2 alone would be hard to differentiate with confidence. (I think that greatly changes the likelihood.) I'd be very surprised to find the same clearly impressed area that couldn't confidently be differentiated with good L3 detail throughout. But I'd expect to have to look at a very large number of folks with any reasonable hope of finding either. And I wouldn't be any more or less surprised if I didn't find anything. Numbers can't tell you what exists, just how often it happens as the set gets very large. The cards have no memory.

I can't say I'd expect to find your matching pair out of the whole of humanity. We don't know nearly enough to have any particular expectation either way. Plenty of ID's have been made with that size area, and if the examiners read the details correctly, apparently there is enough variation that thresholds produced by the usual training and experience have a track record sufficient to rely on for very high confidence. I guess the best way to put it is that I wouldn't refuse on principle to accept a clear demonstration, should indistinguishable impressions of two individuals turn up. But my confidence in the extreme likelihood of a properly done identification being correct would be as high, even if one did appear.

The thing about every person who ever was or will be was just applicable to a discussion of whether anything is truly unique.

There's been some really interesting thought about how the number of people who could reasonably be said to have been able to leave a latent affects likelihood of correct identification. I don't worry about it too much, because the variability seems to be quite high enough for very high likelihood, even with the whole population considered.

[16_Points]

Gerald, can you clear up one small issue please a yes or no answer will suffice. If you had every set of fingerprints of every one living today would you expect to find a small area of ridged skin repeated on another person (say 1/2 an inch square)?

Gerald, I'm surprised you didn't seize on this immediately.

I say, Yes! An emphatic, Yes!

In fact, over a 1/2 inch square, I'm certain I could find numerous examples of friction ridge impressions that repeat the same pattern. (The open field or parallel ridges for one . . . or are these open fields not unique?).

Don't forget to check out the "Captain Close" section of CLPEX. While you won't find the "missing link", the comparisons are quite interesting and suggest to that with a large enough sample size, you'll find the "smoking gun" (i.e. a friction ridge arrangement from two different sources that are indistinguishable).

[Gerald Clough]

I didn't seize on it, because I just can't say all that much about sample sizes and expectations. There's more than enough guessing and assuming in this business for me to be confident of demonstrating something for a specific set. Actually, the question was asked a little broadly. One could find plenty of instances of 1/2-inch squares with sparse details or none that matched. I took it to imply the kind of density of detail typical that, for want of anything better, you find in the interesting portions of the some of the Close Calls.

I agree that the similarities shown in the Close Calls suggests something striking might turn up in a large population. I can't at all say how large it would have to be to make it very likely. And while the Close Calls really don't represent the results of any great efforts, the fact that no one has demonstrated two differently sourced, detail-rich areas in that kind of size range that cannot be differentiated leads me to be leery of being too confident of guessing the sample size required for finding an example.

We're really talking about two impressions, each sufficient to identify, that are of different sources but cannot be differentiated through the common processes of latent print examination. Let's imagine that such a thing were discovered today. I wouldn't be amazed, just mildly surprised that I got to see it. But it wouldn't change a thing about the meaning of or confidence in a conclusion of identification. It would just be a demonstration of something I already knew was likely to occur, even though I didn't know the probability. If you think about it, extraordinarily unlikely events happen constantly. It's guaranteed that something will happen. Any one of them is unlikely on account of the great variability among possibilities. Finding two "matches" is not an event of finding two special entities. We could as well be looking for two perfect mirror image impressions. Or any other precisely defined pair we could describe.

Finding the "smoking gun" would not imply that the event was more or less likely. It would say nothing about how likely it is that the next pair you looked at would be another example. An event might have a one in 100 billion likelihood. Or it might have a one in 300 trillion likelihood. Or it might be a little as one in ten million. Searching ten million and finding one match wouldn't reveal the true probability. I talk about the prospect of one day gaining some solid insight into probabilities, but I'm not at all sure it's doable. We might one day generate some likelihoods of finding pairs to add to the Close Calls, but I suspect coding for difference being indiscernible by an examiner-observer would be very hard to do or impossible.

[16_Points]

Fair enough, Gerald. But I suspect that the subjectivity of latent print comparisons makes this something of an impossible task. Why? There's that pesky issue of sufficiency again. It raises the spectre of circular reasoning. If you find two impressions from different sources that are indistinguishable, then the latent must not be "sufficient" to identify or perhaps it may have marginal comparative value. Or people tell me, "Well, you can exclude based on the 3rd level detail." Really? I rarely encounter comparisons in which I have clear Level 3 detail in BOTH the latent and the known.

So, I would propose the opposite. Without any solid statistical data or comprehensive population studies, you MUST assume that these coincidental matches exist in the population. If one has never been found, then it simply suggests that the incidence is very low (but not zero).

[Gerald Clough]

That's my point when I say that we need to keep the argument from being centered around immature science and solidly centered on demonstrated value. That means not allowing any perception of a claim to absolute scientific certainty when we are simply demonstrating a high degree of likelihood of having identified the actual source. DNA evidence rests upon a much more mature scientific validity, and its conclusions are stated in terms of state-of-the-science knowledge. But that's something of a rarity among commonly applied trace-to-source forensics.

In my opinion, one of our best arguments, and one that is often not adequately articulated, is the fact that there have been errors. At first, it seems odd to invoke proven errors, but those errors are really our best demonstration that latent print identification has one feature that sets it apart from otherwise similar comparative disciplines. For every known erroneous identification, it requires nothing more than the same instrument, an examiner, to detect the error and demonstrate the nature of the error. None of the known error cases have presented the situation of a known erroneous identification AND a sufficient latent that cannot be discriminated as to source. No error has ever been demonstrated to be a mystery as to why it's erroneous. If I were judging the value of a forensic opinion field, I would be very leery of one with no demonstrated errors. That tells me that field is too weak to detect error and therefore too weak to rely upon.

That is the ultimate test of value in evidence. Every standard of evidence revolves around potential for challenge, be it confrontation or adversarial opinion. Therefore, we must actively resist and correct attempts to cast any examiner as an ultimate authority. We cannot allow any pretense that our conclusions must have the reliability of scientific fact, that an examiner should be viewed as a sort of flawed almanac whose pronouncements are taken as gospel without scientific authority. It is absolutely wrong to accept an LPE conclusion are absolute fact. It isn't meant to be taken as such and can never be validated as such. It weakens the image of the discipline but is no weakness in the discipline that conclusions are often unchallenged. That error lies strictly with the litigators who fail to properly approach expert evidence. Latent print examination conclusions are perfectly amenable to challenge. If you want to question the examiner's opinion of high likelihood, seek another examination. The whole of experience suggests that, if the opinion is erroneous, another examiner or examiners will detect it and can explain why it's incorrect. We can't let the fact that we use processes like verification imply that parties in litigation are bound to accept every conclusion that makes it out of a unit. You can argue frank error in interpretation. You can argue sufficiency. What matters is that you can argue. You can, if you can justify it, argue a counter-conclusion. We can ask, "If you believe my opinion of very high likelihood of identification is not justified, show me why you believe I'm wrong to hold that opinion, because I can show you why I formed it." That, and a long history of demonstrated value, is all that's required of expert evidence. We cannot allow ourselves to be held accountable for misuse of our product, nor can we accept responsibility for slack litigation.

[16_Points]

Indeed, Gerald, I see the impetus to adopt a "conservative" approach which says that until uniqueness or it's equivalent is mathematically/scientifically proven we should not claim that no two people have the same fingerprint pattern. Rather the inherent variability of friction ridge patterns based on embryonic growth stresses is "sufficient" for identification.

Who cares if some 57 year old paraplegic in Zimbabwe has the the same fingerprint pattern as the defendant in a commercial armed robbery from Juneau, Alaska? Does the presence of this person really diminish the weight of evidence against the accused? I mean, really? It's just time to move on.

[Gerald Clough]

The changes over the last several decades actually make me more confident in one regard. Prior to AFIS, if I assume a countable probability that a sufficient latent I look at might be matched to two individuals, I would have little chance of ever discovering an ambiguous match with my population limited to, at best, the previously booked criminals in my jurisdiction. (Realistically, unless I was amazingly fast and patient, it would really be limited to those who came under suspicion or who were associated with the case somehow.) My chances got a whole lot better with AFIS, but it didn't happen. We are daily engaged in a rather intense exercise that effectively hunts for ambiguous matches with about as much rigor as is possible today. And every day, the power of correctly applying the community standards to correctly identify source individuals is more strongly demonstrated.

But from a detective's point of view, I now might have another consideration. If I get a report that a latent from my burglary in Small Town, Texas has been matched to an individual in North Dakota, I'm not going to blindly assume I have my crook. And I don't think many prosecutors are going to run with a case where I have a latent-identified suspect who's long been planted in one spot at a great distance without a conspicuous gap in documented whereabouts and no reasonable theory of motive and opportunity. We sometimes talk about a "fingerprint only" case. But that's not correct. There's always more, usually factual motive or opportunity that drastically limits the reasonable suspect population.

[kevin]

I see the impetus to adopt a "conservative" approach which says that until uniqueness or it's equivalent is mathematically/scientifically proven we should not claim that no two people have the same fingerprint pattern. Rather the inherent variability of friction ridge patterns based on embryonic growth stresses is "sufficient" for identification.

Who cares if some 57 year old paraplegic in Zimbabwe has the the same fingerprint pattern as the defendant in a commercial armed robbery from Juneau, Alaska? Does the presence of this person really diminish the weight of evidence against the accused? I mean, really? It's just time to move on.

16,

Are you saying that you believe there are two individuals with the same fingerprints out there somewhere?? And lets use your name as the standard for arguement's sake - two individuals with 16 L2 details in agreement?? Thats a good round about number for a 1/2 inch by 1/2 inch area. Do you think there are two individuals out there with that number of characteristics in agreement?

[Gerald Clough]

16 can answer for himself, but think about the question and how you might interpret any answer. "Believe" can mean anything from opinion held to no particular greater or lesser degree of confidence to dogmatic faith to truth based on demonstrated fact. What would it take to hold a belief in the actual fact of two living individuals meeting your test? It could only be on account of producing the individuals. (And, for our purposes, a consensus that the impressions could not be differentiated.) Since we can reasonably assume that no one qualified to recognize such a thing has discovered these two people, we cannot hold the belief to that standard.

Even if we had sufficiently reliable and valid probability, we can never infer from probability that any event is actually occurring at this time. (Or indeed that it will occur in any finite time period. It just becomes more of less likely.) And we would be poor analysts if we formed an absolute belief on faith.

We could be asked if we believe it possible that those two individuals exist. However we answer that question, the answer must be supported by at least rational argument. Without again belaboring the analysis, I find it logically impossible to reject the possibility. As to the opinion belief, I think any examiner might answer either way with many different degrees of confidence. It's a guess, since we don't have the knowledge to be more precise. My only opinion is that it is likely that no two living individuals exist who meet your test. It's not an opinion of absolute confidence, but it's more than enough for me to hold an opinion of very high likelihood of accurate identification.

One could, of course, believe that every possible half-inch square portion of skin whose impression reveals 16 L2 details can be differentiated by observation from every impression of every other possible portion of skin. In that case, the answer would be that, of course, no two individuals, living or dead or yet to be born will be found who meet the test. But that must fall in the simple faith category and cannot be demonstrated or logically argued.

[kevin]

I understand where you are coming from Gerald - uniqueness as a scientific certainty is not proveable and my view does involve a degree of belief as it relates to this arguement. But it is only because uniqueness (in this case) is confined to the limits of the demonstrative value we have regarding the scientific theory at this time but it doesn't change the picture it presents to us. If this was 1984 and big brother had every inch of FR skin to look at I don't think we would find anything different - there isn't going to be a smoking gun. It would be false to assume only criminal's fingerprints are unique because that is our sample size...instead the science has proceeded logically based on the evidence we have. We are never going to have all of the evidence in this case nor can we maintain FR identification as scientific law but I don't think its unreasonable to confer uniqueness based on studies, empirical data, and scientific theory. Two people having the same print though?? I think that view is based in belief alone.

Sufficiency, statistical models, more research, updating terminology, etc - I'm on board with all of it but not if it involves throwing the baby out with the bath water. I ultimately think this is disagreement based on our viewpoints - we all seem to agree that FR skin is sufficient for identification whether its based on uniqueness or sufficiency. We seem to differ as to the why and the how and even the when. I think we have three areas of the FR identification spectrum here - conservative, moderate and Obama.

[Gerald Clough]

I try to make the clear distinction between unsupportable assumptions and plain demonstrated value, because it does matter how we play into attacks that falsely pretend we cannot produce valid conclusions without scientific certainty. We have not done a very good job so far. We often answer with arguments that clearly fail in the realm the attacker would like to limit us to. And we often fail to go on the offensive and call them out for misrepresentation. Why this is I can only put down to inept forensic (in the classical debate sense) skills.

When you view latent print examination in its proper category of expert evidence:

- Our century of experience isn't an invalid scientific proof. It's a powerful demonstration of the sufficiency of the process and self-imposed thresholds.

- The technical adjunct of AFIS doesn't make undetectable error more likely, because no real case depends solely on a fingerprint. It builds a more powerful demonstration than pre-AFIS experience that our standard is sufficient for high confidence.

- Our revealed errors aren't evidence that the discipline is unreliable. It's evidence that human examiners can detect human errors.

- The natures of every known erroneous identification aren't evidence that the discipline is flawed on account of human operators. They demonstrate that none of those errors are mysterious or undetectable.

- The lack of consistent opposing review isn't evidence that examiners are over-confident. It's evidence that litigators aren't doing their jobs.

- Differences of opinion on sufficiency and other issues isn't evidence the discipline is fatally vague. It's evidence that it is sufficiently well-ordered that expert opinions can be argued, rather than taken on faith, an essential element of value to a fact-finder.


16 is right. It should not make the slightest bit of difference if tomorrow someone presented us with an example of two distinct individuals whose clear impressions of sufficient detail to identify could not be differentiated by an examiner. We should work on representing the discipline in its proper perspective, which properly would find that unlikely pair event to be interesting but nothing more.

[kevin]

Differences of opinion on sufficiency and other issues isn't evidence the discipline is fatally vague. It's evidence that it is sufficiently well-ordered that expert opinions can be argued, rather than taken on faith, an essential element of value to a fact-finder.

Well said. We're just going to have to agree to disagree on some of the other issues. Good debate on the subject,

cheers

[Neville]

16 said 'Who cares if some 57 year old paraplegic in Zimbabwe has the same fingerprint pattern as the defendant in a commercial armed robbery from Juneau, Alaska? Does the presence of this person really diminish the weight of evidence against the accused? I mean, really? It's just time to move on.'

I have been thinking on this one a while and I guess that is where we differ; surely if that is the case it is just as likely my neighbour will have the same print as the Zimbabwean. I think we must not loose sight of the fact that this has never in the 100+ years found to be the case and since the advent of private and public fingerprint data bases all around the world, AFIS and boarder controls imposed in the US, there are some massive national and international fingerprint systems out there, USSR, RSA, FBI, NSY heaps I have never heard of and still this smoking gun as you put has not appeared.

[Gerald Clough]

I have been thinking on this one a while and I guess that is where we differ; surely if that is the case it is just as likely my neighbor will have the same print as the Zimbabwean. I think we must not loose sight of the fact that this has never in the 100+ years found to be the case and since the advent of private and public fingerprint data bases all around the world, AFIS and boarder controls imposed in the US, there are some massive national and international fingerprint systems out there, USSR, RSA, FBI, NSY heaps I have never heard of and still this smoking gun as you put has not appeared.

Exactly. There is the same probability of your neighbor being the second match as for the Zimbabwean. (If we consider them matching a random print and ignore the geographical realities for a moment.) Finding a matching Zimbabwean wouldn't make it any more or less likely that your neighbor would also match.* And the vast experience suggests that that probability of either matching is very, very low. We can't put a number on it, but that doesn't matter. What matters is that in the way fingerprint records and examinations are used, and with pretty much everyone involved in looking at them being very aware of how momentous such a discovery would be, none has been revealed. Very small probabilities don't tell us whether an event has happened or not happened. The very small probabilities in DNA don't tell us whether or not a second person actually exists as a match. That a LOT of looking hasn't turned up the fingerprint event doesn't tell us a match pair exists or not. But both express the justification for very high confidence in an identification.

It's really such an extraordinary thing that humans, the animals who have the ability to care about identification by contact traces and the lamentable urgent need for such identification, happen to bear these features on the skin of exactly the body parts that carry out their criminal actions. We tend to take for granted knowledge about how friction ridge skin is formed and about its persistence. But the fact that its features are expressed in particular and limited ways is what makes it possible to use such things as AFIS. Less variation in feature elements, and it wouldn't be a fine enough filter. If there was a lot more variation, it would likely be impractical code and search. Think of how difficult it would be to even conceive of a similar system for something like the random manufacturing flaws in shoe soles. Those features are far too variable in form. But we have this wonderful tool that constantly heaps on greater weight to confidence in the likelihood of our conclusions.

* There would indeed be a far lesser likelihood of three individuals matching, but observing a rare event doesn't change its probability. It doesn't imply such events suddenly became more likely, just because one happened. Probability tells us about expectations, not the actual outcome of a single opportunity to observe an outcome.

[16_Points]

Wow, folks, interesting discussion. Allow me to be brief . . .

I hypothesize (rather than believe) that two different individuals may have indistinguishable L2 detail up and even exceeding 16 "points" (externally verifiable phenomena?). Furthermore, we can test this hypothesis.

No, any two people do NOT have the same likelihood of being friction ridge doppelgangers. I would propose that certain types of friction ridge arrangements are more susceptible to duplication, deltas and the outflows of loops for example. Hence, a person with all arches is inherently less likely to have his/her friction ridge pattern duplicated. This is not a simple proposition by any means. Rather, any result will take the form of a conditional probability based on pattern type, ridge count etc.

Finally, don't be over confident regarding the 30-40 year history of AFIS. AFIS databases were quite small in the beginning and have only recently expanded in size. Plus, examiners typically have a stronger standard for searching latent prints in AFIS (known orientation, core placement, etc.). And examiners do not record their "near matches". There are some, such as Captain Close on this website, but otherwise no one records or publishes the difficult exclusions. If you intend to use AFIS as proof that friction patterns do not duplicate, then someone needs to record and consolidate this data. Otherwise, you only have weak anecdotal evidence.