from the known to the unknown

[Michele]

I had a conversation last week with an examiner from another lab. This person was adamant that when we do our comparisons, we must go from the unknown to the known. I didn’t think much about it at the time but now I’m starting to wonder why.

Does anyone know when this was first recommended and the reason it was suggested?

[Ann Horsman]

When this started- I don't know. It's my understanding that if you start at the known you may end up biasing (if that is a word) yourself since typically the minutia are clearer on the exemplar. We don't want to 'see things' that are not there, so-to-speak. As we know, latents are often just a portion of a fingerprint area and contain (most of the time) less information than the exemplars.

[Pat A. Wertheim]

Ann is correct, and I think the idea has been with us since early in the practice of fingerprint comparison. The concept here is that if you study the clear print first (usually, the inked) you can form a bias, or an expectation of what you should see in the unclear print (usually, the latent). There is no danger of an erroneous identification if both prints or clear, but if the latent is very unclear, then "points" should be chosen first without any expectation based on prior examination of the inked print. It is the concept that is important, not a strict one way latent to inked comparison. In some cases, the latent print is clear and undistorted with a large number of points, but the inked print is smudged and unclear. In those cases, it might be more advisable to work in the opposite direction, first finding your "points" in the inked print to avoid a backwards bias.

[Boyd Baumgartner]

Ashbaugh mentions it in "QQFRA" page 137, saying the same thing Pat said about clarity being the driving factor. He also mentions the fact that the clarity of the latent is sometimes better than the exemplar and in those cases it's beneficial to go from the exemplar to the latent.

Either way, isn't it in your best interest to go in both directions once you've developed your "hypothesis of identification" and try to predicitively prove and disprove your hypothesis?

[Steve Everist]

This is something that comes around once in a while, and I always wondered why there were those so adamant that it MUST be done this way. I understand the idea of using a cleaner print first could 'bias' the comparison. But how can we truly determine source if we aren't looking for shared features as opposed to "this is in the latent, can I find it in the known?"

I think it's a two-way street for determining if features are present or absent in both marks as opposed to finding what is in one is also in the other, and only going in that one direction.

[Dogma]

How about the common sense approach? Latent prints are chance impressions and as such were generally considered to contain less ridge detail than their fully recorded inked counterparts. So it was always more likely that all the features seen in the latent were also seen in the inked, but not vice versa. So why waste time looking for features in the inked which may not have developed in the latent?

Secondly, we are forgetting that AFIS wasn’t always there handing you candidates to examine. You had to make a full evaluation of the latent in order to search your files or even the stack that some detective tossed on you desk.

Is there anyone out there who took their certification or proficiency test by first looking for details in the known prints and then searching for a latent to fit?

[Steve Everist]

How about the common sense approach? Latent prints are chance impressions and as such were generally considered to contain less ridge detail than their fully recorded inked counterparts. So it was always more likely that all the features seen in the latent were also seen in the inked, but not vice versa. So why waste time looking for features in the inked which may not have developed in the latent?

I'm trying to figure out why this is the "common sense" approach. Regardless of quality, I'm trying to determine same source or exclude from being same source. This involves information contained in both the known exemplar and the latent print. If you only allow yourself one direction, then it seems to follow that if something shows up in the known that doesn't show up in the latent, you'd have to ignore it as a result of the lower quality in the latent must not have made it apparent. Or in your words "So why waste time looking for features in the inked which may not have developed in the latent?" I guess I could answer, because they may not be from the same source and you'd be ignoring information that would lead you to the correct conclusion.

I had a print last year that I was struggling with simply because one area of the known displayed something I didn't see in the latent. If this specific area of the latent didn't exist, there'd be enough that I feel most would have individualized and moved on. But because it was there, I had to consider it. Anybody who believes in the One Dissimilarity/Discrepancy Rule would have HAD to exclude. I spent a lot of time analyzing both the known and the latent to come to grips with what I was seeing, but I didn't know the answer. I remember telling Michele Triplett about this, since I had presented a lecture on Exclusions in San Diego. I went back to the archives and found that there was a new card taken when the person had been arrested for the crime that I was working on - I had been provided with an earlier card. Sure enough, once I got the new print, I found that there were no issues and this area of the known was represented the same as in the latent, but not the same as in the earlier known. My best guess was that there was some sort of damage that only included a few ridges, but it wasn't permanent. I wish I could find this latent again as it would be a perfect example for an updated version of my Exclusion lecture that I'll be giving in a couple of months. Unfortunately I haven't been able to do so.

I think of going only in one direction similar to seeing ACE as a linear model that once you've analyzed your latent and have begun the comparison, you can't go back and reconsider your analysis at any time.

[Dogma]

When I referred to the “common sense” approach, I was referring to the historical reason why this trend started, long before we became scientists and analysts. The fact is once you pick a key feature in the latent and reference it to the known, you have begun the back and forth process of comparison. Then it becomes a “chicken or egg” thing.

Whenever we are presented with anything that we will be expected to make note of, we will, either consciously or subconsciously, begin to observe and record even before the formal process begins. If at first glance you spot a rare feature in the known, by all means go for it.

I don’t believe that seeing the known first will necessarily introduce bias. I feel that bias is a threshold phenomenon: when the two prints have some similarities but no disqualifying features and the similarities are mundane, there is always the urge to go on the hunt for something that will tip the scales. That could be an implied feature such as when two ridges diverge around an unseen but predictable ridge event or external information relating to the investigation.

I suppose that you could say that looking at the known first is like studying for an exam. Its always good to have a head full of details and facts before the question is asked.

[Steve Everist]

OK, thanks for clearing that up - you were responding to the initial question of when this was first recommended and the reason it was suggested? Without quotes, I assumed your response was to mine which was before yours.

I agree - how can you not somehow take note of nearby details in each as your searching back and forth. Having come from a 10-Print background before latents, I began doing comparisons with two magnifiers side-by-side going back and forth between two different known cards. When comparing manually in front of me, I still compare in a similar fashion.

I don't see a problem with using them both in a somewhat circular fashion, although I don't think a clear known should be used as a crutch by which an examiner comes to grips with a poor latent either. If you can only work from the known to the latent, something isn't right.

[Gerald Clough]

I don't think we have to react so strongly to critics who talk of bias that we have to represent and defend so much of the process in terms of it. It may be that I'm splitting hairs, but I see the issue as something more fundamental and not so arguable as a matter of bias. The question, the purpose of the examination is one of discovering if the evidence associated with the crime is evidence of a particular actor, one individual out of many. The two questions about the latent are if it is sufficient for identification and then, of course, if its source is among one or more candidates. All of this effort depends upon determinations of features of the latent that might represent characteristics that may be found in a record print and then which of those features can be confidently identified as specific, defined characteristics. I would hold that this determination must be made for each feature, standing alone, with certainty. Any uncertainty, any feature that cannot be positively described, cannot then be arbitrarily assigned a value from among two or more possible alternatives. Anything else admits the possibility that it has been misinterpreted and therefore may as easily be said to exclude as to contribute to identification.

This is not to say that a feature must absolutely be identified as a bifurcation as opposed to an ending ridge that hooks slightly toward the adjacent ridge. But the character of the two features as representing the same ridge flows must be certain. Obviously, the quality of the latent impression has much to say about this. But if one cannot describe the essential character of the latent feature in concrete terms, without qualification, one cannot utilize it, and the characteristic in the record cannot enter the comparison process. If we allow the record characteristic to settle the description of a latent feature, we are repeating the errors we have looked so hard at in the past in which an array of latent features were interpreted to fit the record. (They may not have literally been made to fit, but, since we only see the ones in which they do fit, the effect is the same.) Sometimes these latent features were ambiguous and could be assigned one of two ways. Sometimes, they are frankly identifiable as having other than the character of the record.

I suppose this may legitimately be called bias, but I prefer to think of it as such a fundamental requirement that it is better described as a violation of first principles, simply a failure to perform.

It is, in the normal way of working, usually impossible to avoid exposure to the record print. One cannot avoid wondering if the feature noted but not yet analyzed in the latent might represent a characteristic noted in the record. The potential for making a biased decision is always present. But it only becomes a biased decision when the analysis of the the latent feature results in an admission of two possible contradictory interpretations, and then one of them is selected based on any consideration of the record. I didn't have to say it was selected based upon the characteristic in the record. I could have just said selected, meaning for any reason, there being no proper means of selecting between two contradictory alternatives outside of cold analysis of the sort that is done for the latent feature standing alone as if seen as an illustration.

It is painful at times to feel that if a latent feature were only represented more clearly, it would be described as consistent with a record characteristic, but the line must be held. That strong feeling simply cannot be allowed to become part of the analysis. This does not exclude all circularity in the examination. For one thing, analysis of the record impression may lead to the discovery of a simple error in analysis of the latent. We would not be so foolish as to hold that one could not correct a plain miscount of ridges or the moment of clarity that has come to us all from time to time when we looked again at the other impression. As pointed out, we may seek a better record. But the mindset and default position must be such that the features of the latent must be unambiguous on their face. Beyond that, we can only depend upon a professional dedication to that position. We are, I think, most at risk when we deal with some distortions. Those can really test our sense of balance.

Comment on The Hypothesis of Identification

Why not The Hypothesis of Exclusion? Indeed, why a hypothesis at all? This is one of those things I think are response to the urge to be "scientific." And I think it is one more difference between latent print examination and "doing science." (I have somewhere at home a book titled "Doing Science." It is by research scientists, and they clearly have a reasonably well defined idea of what that means, and what we do ain't it.) The scientist may hypothesize. He may frankly guess, in the interest of learning something by following the hunch. The scientist can afford to be wrong. He can afford to have a pet theory. He can present the barest experimental support for it. He can, in fact, display considerable hard-headed bias, without losing his standing in the scientific community. It's all part of the global collaboration in poking away at the frontiers.

But why would we ever have a hypothesis. Hypothesis may be variously defined as assumption, guess, or proposition. We need no hypothesis of either identification or exclusion. We do not deal in tentative declarations of fact intended to be challenged and retested and perhaps overturned or modified by other authorities. The genuine analysis of the evidence of a latent fingerprint impression and the record print or prints will reveal absolute stark truth which ends all speculation and excludes any revision. I think the use of "hypothesis" in the context of a latent print examination is tantamount to declaring a bias from the start. It's enough that we, as humans, have the natural drive to make an identification, without codifying that bias by declaring a hypothesis. Especially when we don't even mean it and only use the word as a scientific touchstone. Leave hypothesis to the lawyers.

[Pat A. Wertheim]

From the earlier thread on that article about the question of whether fingerprint examination is science, poor Charles closed it out with the comment,

I think this topic is pretty well exhausted and so am I.

Now Gerald has gone and done it and attacked the word "hypothesis." Sorry, Gerald, I'll risk exasperating Charles and answer you on this one. In the paper I refered to on that other thread, I proposed a variation of "the scientific process" in which I suggested that, in doing a comparison, we form a hypothesis/counter hypothesis and test them both. Arbitrarily, we could call the hypothesis "identification" and the counter hypothesis "exclusion." In order to reach a final conclusion, the examiner must prove one and disprove the other.

Michele Triplett teaches an excellent two day class on fingerprint identification as a form of "hypothesis testing." She makes a great case for fingerprint examination as science. As I think I stated on that other thread, it seems to me to be a matter of semantics and definitions. If we are talking pure science or exact science or theoretical science, fingerprint comparison doesn't even come close. But if we move into the realm of applied science and further, into forensic science, then I will defend fingerprint examination under Daubert as good science. And what better way than with Michele's teachings of hypothesis testing? Or, if it's bias we're trying to avoid here, giving equal weight to the hypothesis and the counter hypothesis, proving one while disproving the other?

[Boyd Baumgartner]

With regards to the statement 'hypothesis of identification', since I made it, allow me to clarify. It was the lack of a better phrase plain and simple. I could use the phrase 'hypothesis of identification', 'snowball of information', 'mounting evidence', 'weight of the data that tips the scales in favor of (insert conclusion here)etc etc. in the same manner. Semantics aside however, don't discount the method that followed, namely attempting to falsify your own conclusion, because it's important in arriving at a sound, valid conclusion.

I am a Practicioner, plain and simple and have no qualms about it. Allow me to defend the notion of hypothesis regardless. While you may give a negative connotation to the word tentative, I take it as the denotation of 'probationary'. Your initial conclusion is tentative, which is why our identification process is known as ACE V, emphasis on the V. Verification is the validating processes to methods and decisions used in arriving at the initial conclusion in the same way that a 2nd opinion gives legitimacy to medical diagnoses. This view is further underscored by the fact that many latent labs standard operating procedures include a section for conflict resolution for different conclusions by different Examiners.

Belief that an examiner's genuine analysis is axiomatic is not plausible (eg Mayfield). I have no problem stating that my analysis of a ridge ending may be due to the absence of deposition or naturally occuring, especially when there is moderate distortion. When it is an agreed upon phenomenon that a 3 dimensional object never replicates itself identically in a 2 dimensional medium, claiming absolute stark truth through analysis is not logically defendable. While I could go into a whole diatribe on the one discrepency rule here, I will skip ahead to the point and that is that Scientific Inquiry, Interpretation, Analysis, Synthesis, Ternary Predicate Logic and other disciplines all apply to what we do, borrowing from the lexicon of each for the sake of understanding and communication is completely valid.

[Gerald Clough]

I'll yield on the hypothesis issue, because I think what we do is far more important than what we call it.

But I wouldn't at all mind hearing the one discrepancy rule diatribe. I think sometimes that thrashing out what seems to be a bit of a collateral issues develops insight into the whole process. The issue of exclusion is one of those. It's easy enough to say that a single clear discrepancy excludes. (Whose rule would that be, anyway? Does it follow from the required agreement? Anyway, let's take it as a rule for the purpose of discussion.) Now I'm always interested when something is stated as a firm rule but where problems of what I might call weight are glossed over. So, if one discrepancy excludes, how would that be applied? How much agreement can be negated by that discrepancy? It's not that this is a question that pesters an examiner's work day in and day out work. It does, however, raise questions about what quantity of detail is sufficient for identification.

To illustrate, imagine an examination that finds that the available portion of skin represented in a latent is sufficient, perhaps more than merely sufficient, for identification and a companion record impression that agrees in all respects. As with most latents, the portion of skin represented by the latent is considerably less than that represented by the record. Imagine a second latent in all respects like the first, except that it represents a bit more skin, and in that portion revealed by the second latent is a clear discrepancy. The rule, if applied to the second, excludes.

Aside from the sticky issue of whether the exclusion from the second latent carries over to the first, what do these two latents mean in terms of how we justify sufficiency to identify? Since our decision on sufficiency is not a matter of any specific amount of detail, how do we answer the proposition that many latents which are concluded as representing a particular source might exclude that source, if only a bit more skin impression were made available? At what point might an adversary demonstrate a latent that excludes a source but that would be deemed an identification if the portion with the discrepancy were removed? How may one answer the challenge that a poorly impressed portion within the area used for identification might hide a discrepancy?

I think these are good questions to ponder, and I would like to hear some answers. If the one discrepancy rule is accepted on its face, it logically would apply, no matter what quantity of agreement detail was available. Or do we hold that there is some quantity of detail in agreement that would cause us to believe that an outlying discrepancy, no matter how clear, had been mistakenly interpreted as a genuine characteristic? And if that apparently clear characteristic is held to have been misinterpreted, why are the identifying characteristics taken as genuine?

So, rule or not, the principle is still capable of stinking up the room. Or is it just my imagination?

[Ernie Hamm]

Back to the original inquiry, "from the known to the unknown" or perhaps better as, "from the unknown to the known".

I believe "Dogma" was in the area when it was stated, "The fact is once you pick a key feature in the latent and reference it to the known,...". In the long ago time when I was performing examinations in the latent discipline, it was accepted that you locate target or reference points for orientation purposes for a comparative examination between an unknown and a known. The unknown (latent) is the primary objective in the comparison and usually has a limited area of friction ridge detail. I can not even think about looking for a reference/target area in a single or, a more frightening aspect, a stack of X fingerprint cards for locating a similar feature in a latent, be it a particular arrangement or as I stated once, "an ending ridge with attitude'. AFIS technology would not care which way the comparison would be conducted, but the human examiner with the handheld magnifier faces a different task.

Back to the tar pit.

[Charles Parker]

Pat Wertheim

From the earlier thread on that article about the question of whether fingerprint examination is science, poor Charles closed it out with the comment,

Charles Parker wrote:

I think this topic is pretty well exhausted and so am I.

Now Gerald has gone and done it and attacked the word "hypothesis." Sorry, Gerald, I'll risk exasperating Charles and answer you on this one.

Aaaah the Honorable and Esteemed Mr. P. Wertheim ASSUMED TOO MUCH

The exhaustion comment was on the topic of "older examiners afraid of the mantle of science" and that it was 12 midnight when I wrote it.

No the discussion of Analytical Methods vs The Scientific Method is still alive and well as far as I am concerned.

However this discussion of "From the Unknown to the Known" sounds pretty interesting.