One Discrepancy Rule

[Bill Leo]

Hi Michele,
The topic of the one dissimilarity rule is an old one that we were addressing back in the early 1990s in California after a murder trail was lost because of the failure of a latent print examiner to properly explain what is a dissimilarity. Please see an article that I wrote on the subject at that time at http://fingerprintidentification.net/_wsn/page2.html

[Michele]

Bill,

I loved the article and it did clear some things up for me but I still have two of my first questions unanswered. First, how do we know the difference between a discrepancy and distortion? Sometimes there's clear evidence that gives us the answers but in the prints posted in this weeks Detail, it seems like a different conclusion (of whether it's distortion or a dissimilarity) could be arrived at depending on how much of the print you're viewing. It seems to be that whether it's distortion or a dissimilarity is determined only after we've decided if it's an identification or not. If this is true, then (my second question) how can we have a "one discrepancy rule" from which we exclude someone as the donor?

You're right, it seems like this issue has been around a long time. I'm surprised after so long that the subject is still going on. One of the problems (as you stated in your paper) is oversimplification of definitions. Even though this topic has been going on for so long, I still think it's worth discussing because as you showed in your article, this issue can make or break a court case and it's also an issue that brings someone's level of expertise into question.

Thanks for bringing this article to my attention. At first I thought it was going to be really long, but the link is actually for 2 articles. This paper was really only 3 pages and well worth reading.

I also wanted to say that I thought 'Dogma' brought up some good points but I haven't had time to reply yet (hopefully tomorrow).

Thanks,

[Les Bush]

Hi Michele

I'm not a guru with AFIS technology and the advice received regarding your inked images seems to shed some light on how the extra ridge appeared. The sulci distance for old generation afis/live scan machines had a small tolerance with regard to the dimensions of the ridges and spacing. Newer machines record the impression with a greater number of photo scans which improves the quality of the detail. In the older machine where a larger than average sulci was present the program would try and compensate for the interval by generating a new ridge. This was apparently true for ridges at the edges of impressions and not normal for the center of the pattern. The position of your extra ridge does fall into the outer zone of the impression and it is in the region of larger than normal sulci. These might be a couple of things to check with regard to how the images were obtained. I'd still like to see a couple more examples of the same print before discussing whether you have an actual discrepancy in the friction ridges. Cheers from oz. Les

[Pat A. Wertheim]

Regarding the two photographs in the Detail, I thought I recognized the fingerprints but I wanted to find my copies before I posted. Just today while digging deeply into dust covered boxes I came across one of them, which was sufficient for me to be certain.

I was given a set of these photographs by Des Benton, supervisor of fingerprints at Garda Headquarters in Dublin, Ireland, a few years ago. The Garda had an extensive arrest file on the individual whose fingerprint appears in the photographs and the puzzling discrepancy showed up late in the series of fingerprint cards in his file. As they began to look more closely at the whole series, there was a very telling inked card between the early fingerprints and the later ones in which the extra ridge and point occurs. In that intervening photograph, the inked print of that finger actually has a band aid that shows up covering that part of the finger.

Des did not know exactly what kind of injury lay beneath the band aid, but one might speculate there was a razor-edge cut running down the middle of the ridge that split it into two and healed without the scar itself visible in the inked print. At least, that's my guess. Any others?

Anybody from the Garda following this discussion board with additional information?

[RL Tavernaro]

A razor split of the ridge and subsequent clean healing was my best guess also (so far), pending seeing the other prints in the series.

Pat, do you know if all prints in the series are inked prints, or are one or more livescan?

[Michele]

These are all inked impression (which I did get from Pat Wertheim, I didn't mention this earlier because I hadn't talked to Pat about using his name, but he did tell me I could share these images when he gave them to me).

I've scanned them all into my computer (6 scans - one with a bandage) but the images I have are all enlarged images (about 6 inches by 8 inches) so the files are fairly large. They're scanned in at 300 ppi gray scale. The file sizes are just under 6 mg each.

If anyone would like me to email me these images, just let me know. I can probably even figure out how to change the file size if needed (where's Steve Everist when I need him??)

michele.triplett@metrokc.gov

[cchampod]

The distinction between dissimilarity and discrepancy is useful. Discrepanccy expresses the view that the difference cannot be explained if truly the mark has been left by this finger. The dissimilarity covers the grey area and is related the assessment of distortion and clarity. I think that there is an urgent need to study distortion systematically in order to define with precision what is meant by "within tolerance". As soon as we do this using objective measures, probabilities will come into play. Indeed there are dissimilarities that are more likely than others. The very fact that a "possible" explanation does not alleviate the need to take the difference into consideration during the comparison process. Only a probabilistic approach allows to handle conflicting information (similarities and dissimilarities) in a coherent manner. Except in the clear cut scenarios where we can safely define a discrepancy, all the other cases can only be resolved properly using a probabilistic approach.

Regards

Christophe

[RL Tavernaro]

I am neither a statistician nor mathmetician, and my question(s) may border more on the philosophical. However, conceding a probabilistic approach may be best for resolution of conflicting information, are there not inherant difficulties in ascribing consistent values to what is often subjective information (description/explanation of dissimilarities) ? How would a permissable degree of latitude be determined and incorporated into a useful model (maintaining coherency)?

Regards, RLT

[cchampod]

Excellent comment. I think that we should aim at collecting hard statistical data on distortion. Hence statistical estimates can be provided for the probability of a X degree distortion for a given finger. However in some areas, we need to rely on subjective asignement of probabilities. These are called subjective probabilities and are known to be coherent in probabilistic terms. Calibration of such opinion is indeed an issue but not very different from the consistency issue of a verification step of fingerprint comparison.

Regards

Christophe

[Charles Parker]

Please forgive my tardiness in this posting.

Over the past 2 months I have heard 4 latent print examiners state that if presented with two fingerprints that shared 30 to 40 ridge characteristics but there was a single discrepancy between them “that could not be explained” then they would pass on the identification. These were not new examiners but some with a number of years of experience. On the other hand I would have no problem with making the identification because the weight of the matching characteristics overrides the anomaly. As Michele stated there is an “acceptable difference”. My thoughts have gone to why the difference in their beliefs with mine! Perhaps it is a difference in training, or the mental picture of such an event!

In Michele’s original posting she states the rule as “One discrepancy or unexplained difference excludes the person as a donor”. In going over the other posts and Professor Thornton’s article this theme is repeated. I believe that is only part of the rule that I was instructed in a number of years ago. The second part of the rule comes in part from Locard’s writings and has to do with 12 or more clear and concise points. The rule as I was taught goes something like this: “If there is one discrepancy between two prints then they cannot be from the same person, but if you have 12 or more clear and concise points between two prints there can be no discrepancy”. Now that may not be exactly what the old Sergeant told me, but it is pretty close.

Now I know that the above statement has a number attached and in today’s world that might be the mark of Cain. Can I provide any studies that Locard was right? No I cannot. But I have not seen any studies that refute it. Until such time as there is I will bow to his stated observations.

Also in Michele’s first post she states that some people may determine enough information and then come up with a reason for the difference. I believe she is 100% correct. In the example she provided I have shown that to a number of examiners and some say that it was a live scan problem, some said that it was ridge movement, some said an inking problem, but most said they did not have a clue. None initially said it was an injury as there is no evidence of that readily apparent. But when given the additional information that between the inked prints there was another one taken in which the subject had a bandage on that finger most agreed that it was probably some type of injury but not exactly sure what type.

This brought up the issue that for me is the crux of Michele’s post and I would have to say is the salt on my tail feathers. With a preponderance of matching characteristics do I have to know exactly what caused the anomaly? At this point I would have to say no. Mr. Tavernaro reinforced this for me when he stated that a difference that had several possible explanations is different than one that has no logical explanation. Now I know this is shaky ground considering the erroneous identifications that have occurred in the past few years. One must remember that it rests on preponderance of clear and concise information. One must not fall into the dark abyss of using distortion to dismiss discrepancies on poor or marginal prints.

This brings me to Craig’s post that this is a QA issue. I agree with that concept, but would like to say it should apply to all examiners and not just new examiners. Michele made a good point in that the rule as stated is too brief. I know we all like brevity as it provides fewer complications in our lives, but the way the “One Discrepancy Rule” has been stated is way to brief. The issue needs to be studied and a guideline, standard, policy, or resolution be printed to clarify this concept. I will leave that up to those that conduct such things to pick up the gauntlet.

With Dogma’s post I think he was a little tongue in cheek. Please correct me if I am wrong in this Dogma, as I do like to read your posts.

For such a set of prints that Michele has presented I do like the coined term “Galton’s Dilemma”. What we have here is a Galton Dilemma. However in actuality I do not think Galton was really concerned about it. He wrote about it in his book and talked about it at the Belper Commission. Galton could not explain it exactly but he had no problem in declaring them from the same person (maybe because he took both sets of prints). That has a nice ring to it though. A Galton Dilemma is where a feature is dissimilar between two prints of which an exact explanation cannot be described but the dissimilarity is not sufficient to override the preponderance of clear and concise agreement.

Another issue that Michele raised is if I was presented with version of her print and it had only 8 to 10 matching ridge characteristics would I call it a discrepancy and then if the whole image was opened up to show 80-100 matching characteristics would I call it dissimilarity. Yes I would. Why I do that is because the level of information present in both scenarios is different? One has a very limited amount of information while the other has a lot more. We already do that with a latent print we have called inconclusive and an additional processing technique develops additional information that moves it from inconclusive to individualization or exclusion. Which way to go is dependent upon the depth of the information available?

I want to thank you for reading my post.

[Dogma]

Hello all and Charlie,

While part of my post may have been “tongue in cheek,” it was intended, as always, to inspire thought especially about the often opposing “dogmatic” themes of our discipline. There are those who think that the comparison aspect of what we do should be completely blind to the extent that the only thing that matters is the appearance of the specimen before us. How it got there is nobody’s business. However, because things like uneven pressure, “double taps”, injuries, disease, media, surface characteristics and method of development can affect the appearance of a print, these factors all become an important part of analyzing the print. I believe that Mr. Zeelenberg has attributed some of SCRO’s points in the McKie case as wood grain.

If you were to walk into your doctor’s office complaining of a cough, would you expect a good doctor to simply prescribe a suppressant and send you on your way? Or, would you prefer that he/she listen to your lungs for congestion, find out if you are a smoker, check blood gases for oxygen levels, suggest x-rays or a host of other examination protocols relating to coughing? Of course, the doctor’s procedure will depend on his level of experience, knowledge and things like whether or not there’s “something going around.” The point being that the doctor would not base his/her analysis on a single symptom.

It would seem that many “forensic scientists” have forgotten (or never knew?) that a scientist is not just an analyst using some scientific apparatus to arrive at conclusions. A scientist is also an investigator, seeking all related information which may influence their conclusions.

If a print contains a certain number of clear Galton points, it undoubtedly contains additional information contributing to individualization: the size and flow of the ridges, the topography of the ridge structure, the appearance of pores, etc. Also, certain “points” are rarer than others, but are routinely given the same “count value.” All of these give weight to the conclusion of the examiner. Unfortunately, we have not as yet agreed on how to quantify such things.

My reticence toward a probability mode, was clearly stated in this week’s Detail with reference to the “Prosecutor’s Fallacy.” While practitioners may understand what is meant by the examiner’s conclusion, the jury may not or may understand it to mean something unintended.

In “Ridgeology” Dave Ashbaugh references an article by Chief Superintendent R.A. Huber of the RCMP, titled “The Philosophy of Identification.” In that article, Huber states: “When any two items contain a combination of corresponding or similar and specifically oriented characteristics of such number and significance as to preclude the possibility of the occurrence by mere coincidence, and there are no unaccounted for differences, it may be concluded that they are the same, or their characteristics attributed to the same cause.”

The Interpol European Expert Group on Fingerprint Identification (IEEGFI) issued a paper on the “Method of Fingerprint Identification,” which states: “Fingerprint evidence should only be stated as absolute and positive conclusions. There is no basis for likely or probable conclusions neither based on statistics nor upon personal judgment.” Regarding the comparison process it states: “Differences are studied and are weighed whether they fit in the margins of tolerances normally found between prints coming from the same donor. Explanations for the differences are checked. A minutia of which the location and direction are apparent in the latent and absent in the comparison print (or vice versa) in principle forbids the drawing of any positive conclusion on identity.”

And lest we forget, SWGFAST’s “Standards for Conclusions,” which states that individualization must satisfy the conditions that include: “Absent any discrepancy.” Under the section titled “Exclusion” we read that one of the basic principles for exclusion is: “The presence of one discrepancy is sufficient to exclude.”

Please note that none of these refer to an overwhelming presence of characteristics in agreement nullifying a discrepancy.

So, for those of you who are still awake, going forward with an individualization which contains unaccounted for differences, would seem to put you and your methodology in contrast with accepted practices as endorsed by the “appropriate scientific community.” Daubert, anyone?

[Kasey Wertheim]

And lest we forget, SWGFAST’s “Standards for Conclusions,” which states that individualization must satisfy the conditions that include: “Absent any discrepancy.” Under the section titled “Exclusion” we read that one of the basic principles for exclusion is: “The presence of one discrepancy is sufficient to exclude.”

Please note that none of these refer to an overwhelming presence of characteristics in agreement nullifying a discrepancy.

So, for those of you who are still awake, going forward with an individualization which contains unaccounted for differences, would seem to put you and your methodology in contrast with accepted practices as endorsed by the “appropriate scientific community.” Daubert, anyone?

Again we find ourselves playing semantic games. SWGFAST uses discrepancy to represent an actual difference in ridge structure, not distortion. You are right... as you state, "an overwhelming presence of characteristics in agreement" (which indicate 'same source') cannot "nullify" a discrepancy with this definition, because one simply would not exist. It is an impossible situation to have the same source leave a discrepancy. I learned that in the first module of my training, as did many other latent print examiners.

What we are really discussing is that anything less than a "discrepancy" is not valid to effect an exclusion. There has to be a sufficient quality and quantity of "discrepancy" for an examiner to state for certain that two prints did not come from the same source. If an examiner mistakenly interprets a feature as a "discrepancy" when in fact it was caused by an explainable event (matrix, substrate, etc.) and place their entire conclusion on that feature, then they are as close to erroneous exclusion as an examiner who places too much weight on just a few similar characteristics and becomes close to an erroneous individualization.

To specifically address your concerns, dogma, I would say that if enough detail exists for individualization, then a "discrepancy" as we have defined it becomes impossible - it must be distortion that appears dissimilar.

The real issue in this post is whether the examiner actually has to make a determination as to the cause of the dissimilar feature when they (correctly) know the prints match. Or is it sufficient to simply know it isn't a discrepancy because of the overwhelming similarity, and at the same time not project causes about which the examiner isn't certain. In short, the relevant question is, "Can a dissimilar feature be explained by citing overwhelming similarity for individualization?"

[Dogma]

Hello Kasey,

I think its more than a matter of semantics. With due deference to Prof. Champod (“The distinction between dissimilarity and discrepancy is useful. Discrepanccy expresses the view that the difference cannot be explained if truly the mark has been left by this finger. The dissimilarity covers the grey area and is related the assessment of distortion and clarity”) the SWGFAST glossary defines Discrepancy as: “A difference in two friction ridge impressions due to different sources of the impressions (exclusion).”

Our helpful friends at SWG go on further to define Dissimilarity as: “See Discrepancy.”

I trust that you will address this at the next meeting?? (I realize that you are one among many, but hey, ya gotta start somewhere!)

In offering my opinion regarding your question: "Can a dissimilar feature be explained by citing overwhelming similarity for individualization?" I have to say, no. Without deciding, concluding or investigating the reason for the “dissimilar feature” we are merely inferring its lack of significance. We also don't have a means to concretely quantify concepts of similarity. I don’t like asking a jury to simply “take my word for it,” even though that is what all expert witnesses are doing. My own intellectual integrity demands that I discover the reason for the difference or at least understand how it could have credibly occurred. After all, we still don’t have the luxury of the probability mode and are counseled, as indicated in my previous post, to speak in absolutes.

[Kasey Wertheim]

we are merely inferring its lack of significance. ...we still don’t have the luxury of the probability mode and are counseled, as indicated in my previous post, to speak in absolutes.

I'm not alluding to a probability... I'm saying if you have so much agreement that you know beyond any shadow of a doubt it is a match, given that existing sufficient threshold for absolute identification, if you had a dissimilar feature, the cause of which you could not figure out, would you not infer it's lack of significance and report the match? Would you be happy with "it must have been (insert favorite distortion here)"? If not, what would you report?

[Dogma]

Kasey,
The point I am trying to make is that without being able to assign clear, quantifiable and tangible values to concepts like sufficiency and thresholds, which change from examiner to examiner and without clear definitions and distinctions being made in the literature and guidelines regarding terms like discrepancy and dissimilarity, the examiner runs the risk of having their conclusions called in to question if they can’t account for the differences.

SWGFAST Formula:
Discrepancy = Exclusion
Dissimilarity = Discrepancy
Therefore:
Dissimilarity = Exclusion

This is where the problem lies. The problem is very similar to the simultaneous issue which popped up in Boston last year. There are examiners who believe in the practice and others who shy away from it. I believe that the court ruled in favor of the underlying science of fingerprints but would not accept that particular practice because there was not enough research done. Likewise, the current practice for identification requires an absence of a discrepancy without exception. I understand your understanding of dissimilarity and agree with it. But, the SWG guidelines force us in another direction.

You tend to view these issues strictly from the scientific and philosophical aspects. While I consider those issues, I also consider how it will play in court. No matter how good the product, if you can’t sell it, you’re on the loosing end.

Q: Mr. Jones, I have looked at the fingerprint charts that you have prepared for the court and notice a dissimilarity between the two prints in the area of 2 o’clock. Would you agree that the dissimilarity exists?

A: Yes.

Q: Can you account for this dissimilarity?

A: No.

Q: Are you telling the court that you have no idea how this difference came to be?

A. Yes.

Q: Mr. Jones, have you ever heard of SWGFAST?

I think you get the picture.

The reality is that the dissimilarity that you suggest would not be reported in any form by most examiners. Most would, I suspect, dismiss the difference to some form of distortion though not specifically stating so. It would probably not come up unless the identification was challenged. We can no doubt test for the theory that an overwhelming (or specified) amount of detail in agreement between two prints nullifies a dissimilarity (as defined by you and not SWG). But, until that testing is done, we are saddled with existing dogma.