What is enough.

[Gerald Clough]

Yes. I agree that the theoretical likelihood of being a match pair (I trust we can dispense with long explanations about what that means at this point) depends on what I might call the "richness" of the features in the area in question. Of course, we're talking about a minimum sufficient quantity of detail for identification, whatever physical area is involved. I don't know that it has much to do with L1 patterns, since, assuming two portions of skin could impress alike, it could happen without regard to the L1 pattern. Or maybe it's just that I'm trying to view this in the most basic way, considering features arrayed within an area. But I don't doubt that, if an adequate study could be designed, that some sorts of feature arrays would occur less frequently and would therefore be less likely to be duplicated. But I think that's looking farther down the road than necessary initially.

The essential first step is the hardest. In fact, I think that once it's done, the rest will fall out in analysis. That first step, though, I think will be really hard. That would be to devise a coding that would represent an area of details in such a way that two indistinguishable areas would code alike, or would at least be coded so similarly that too many false positives of close-but-observably-different wouldn't be generated. If we ignore the issue of an area being sufficient for identification, we can imagine a very simple case of two objects, each bearing one visually identical detail, two differently sources areas that are not observably different. When we add a second detail, we then have to work out how to code the relationship between the two, and it has to consider and be tolerant of distortion effects. At that point, I have no good notion of how that can be defined. I mean, give me a precise definition of when an examiner will deem two details to be in the same relationship in two impressions.

So I think the only tool available is logic, and that won't produce numbers but allows a couple of valid statements. The fact that in any area of an impression, there can only be some finite number of details arranged in a finite number of observable ways tells us the number of observable variations is also finite. The is a number greater than 0.0, which make it possible but tells us nothing about whether a match pair actual exists or ever has. The number of details required for identification, variable though it may be, tells us the number of possible variations is very large. And we know from a great deal of experience (and would guess from what we know of how the ridge formations occur) that there are more and less usual or unusual sorts of arrangements. That means that there are classes of arrangements that are more and less probable.

So when I said the neighbor was as likely to be a match as was the Zimbabwean, I could better have said that the likelihood of any random pair of individuals matching was the same as any other random pair matching. Clearly, it's quite possible for the portion of impression from the Zimbabwean or the neighbor to be a more rare type in common with the other possible member of the pair.

I don't know that we will ever have well refined numbers, and the argument will, for the foreseeable future, remain one of plain demonstrated value. And you're right. We know very little about the inverse of the question about what it takes for identification: what it takes to not identify. That's the kind of knowledge you would need to construct a virtual examiner you could trust, since the examiner function can as reasonably be stated as recognizing what isn't a match. Very likely, until you could do such a thing, you couldn't begin to come up with some credible probabilities.

[16_Points]

Gerald, the short answer is that I agree. However, we can study and quantify this phenomena (of a near match between two different sources). It's neither simple nor easy but entirely doable. The question is, who has the time and resources to accomplish it?

[Gerald Clough]

It's probably more in the realm of academics than what's being done within the discipline. There are so many more concrete study possibilities with readily applicable results that there are more ideas than there is time and funding. And there's a whole lot to be done that requires creative thinking before we can really get to validating such things as how to code what's observable. We can collect a lot of near matches, but at some point, to crunch some numbers, we have to work out how to describe what we're seeing. That may well fall more into the category of artificial intelligence and pattern recognition. I've got a feeling that the way in may be either from the direction of the virtual examiner I mentioned or from the direction of emulating ridge development and distortion modes. But there might be something to the notion of working from particular classes of really close call to try to develop some rules defining when it's not a match. Have to think about that one.

[Neville]

HI 16

I agree very interesting, it is a shame there are not more opinions being voiced.
What on earth is a doppelgangers. Is that a scientific term or something?

You seem a bit fixated about pattern types, of cause pattern types repeat, I think your terminology may be a bit off putting there, surely we are not discussing pattern types and a person with all arches is inherently MORE likely to have his/her friction ridge PATTERN duplicated, that is if they are plain arches, any old school fingerprint expert having done plain arch comparisons knows that. With regard to your comment about AFIS as far as I know most Henry collections were dumped on to NEC AFIS in bulk by NEC in the California. I do not know about other providers but I imagine that for Morfo to be in the running they would have had to have offered a similar deal. My point being that 20 years ago they were very large collections.

Now Gerald on to DNA, every one knows that Identical Twins have the same DNA and seeing I can't guarantee I do not have a twin who was adopted out at birth I must assume that there maybe a dup of my DNA out there. Also I understand if I had a bone marrow transplant I would take on the donors DNA and it may also be that if I have a Blood transfusion I may take on the Donors DNA for a short time. What studies have the scientific sector done on this I wonder? Yet DNA is supposed to be the new benchmark in identification! Sorry, but I for one do not buy into that.

Believe or not there are still people out there that still talk about DNA fingerprinting! So what is the benchmark, I think in public opinion fingerprints is still on top. Public opinion is what counts when it comes to a trial.

In the latest JFI Nicolas and Pierre have published a very interesting article RE Paul-Jean COULIER, something may have been lost in translation though I doubt it, seems to me there has been a fair old amount of that 'body of scientific study' done on fingerprints in the past it just hasn't been on the WEB. Thank you Pierre Margot for some good old fashioned research and not relying on Wikipedia, CSI or the equivalent.

[ER]

Edit

[Neville]

Thanks ER
WOW that looks just like me, oh dear I just checked out Wikipedia I'm about to die. But then a day is the same as a thousand years so whats the big deal.

[kevin]

Now this bus is going somewhere...

[Neville]

Kevin what are you trying to tell me, now I am worried. Plus my co worker tells me that this thing is obviously a woman which is not helping me any!!!!!!! bludie ell mate, I'm a gonna.

[kevin]

No, I don't think that is you Neville....Notice the fine, tapered fingertips? Perfect for identifying ridge shape and pores. Has anyone seen Daktari lately?

[Gerald Clough]

Now Gerald on to DNA, every one knows that Identical Twins have the same DNA and seeing I can't guarantee I do not have a twin who was adopted out at birth I must assume that there maybe a dup of my DNA out there. Also I understand if I had a bone marrow transplant I would take on the donors DNA and it may also be that if I have a Blood transfusion I may take on the Donors DNA for a short time. What studies have the scientific sector done on this I wonder? Yet DNA is supposed to be the new benchmark in identification! Sorry, but I for one do not buy into that.

People like to rate things. DNA appeared to be precise because it featured actual probabilities that can be very high. But there are some similarities with fingerprints. DNA probabilities are reasonably useful but not really exact. We don't know enough about local distributions to be more exact. But whatever the exact numbers might be or however we might choose to calculate, it's still quite high because we do know about how many variations are possible. While for fingerprints we depend on considerable experience and some efforts to calculate the number of possible variations in general, we are still confident that the number of possible variations is extremely high and are justified in presuming that the exact number, if it was known, is also high.

Even twin "DNA" is not identical. We can't differentiate between twins with the common criminal forensic methods, but various alterations occur from conception, and those accidental variations can be detected with methods beyond what we use. And, just as with fingerprints, even without a twin around, there's still a possibility of duplicate DNA within the parameters of the forensic method. And, while for the purpose of focusing the discussion, we've been talking about L2D, it's quite possible that any given latent may exhibit L3D that will easily differentiate two impressions that can't be discriminated by L2D. There's some rationale for thinking DNA is superior to fingerprints. It's sort of a rock-paper-scissors kind of thing, because DNA can resolve fingerprint errors, as it did in Cowans. But that's just an artifact as a result of fingerprint analysis depending on human observation. But it's not that hard to imagine a situation in which fingerprints might lead to a revelation of an error in collecting and labeling recovered DNA. Or, since there is a possibilities of two non-twin individuals having the same DNA profile, were that event to be observed, fingerprints would discriminate and could potentially identify the correct criminal actor. Both DNA and latent print identification look at a subset of the physical characteristics.

Just as a fun guessing exercise, you can think about whether we're more likely we would find two individuals with the same limited DNA profile than we that we would find two individuals with the same fingerprints. That points up a difference in the methods. If you found those two DNA-matched people at the common forensic methods level, you'd have to invoke different DNA methods to discriminate them by DNA. If you found two individuals who each had identical areas of L2D just sufficient to identify, you'd just need a bit more to the latent to tell them apart by exactly the same common examination method. On the other hand, you might get just that barely sufficient latent, while DNA recovered unknowns more commonly can be used to examine much more than the usual amount of data. The difference is the great variation in the amount of data available from latent fingerprints, from a tip to two whole palms. (Brings to mind my guy who was carrying a cardboard beer case from a burglary and ducked down to avoid being seen and pressed both palms onto the top of the case, producing better images than the record.)

But there's more to value than that. A great deal of the value of fingerprints is that it's low-tech. All that's needed to get an exemplar is an ink pad and paper. All that's needed to analyze is a magnifier. It's not that either one is inherently faster. They identified Saddam very quickly from DNA. But you can do an ID from fingerprints anywhere you have an examiner. And with two record sets, elimination can be pretty much instantaneous. And I think that as time goes by and the sensitivity of DNA methods increases, analysts will more often face a lot of DNA profiles from one recovery. I'm already seeing attorneys trying to ask courts for analysis of "DNA collectors" like carpets in some cases. Of course, they're hoping to find multiple individuals so them can imply other suspects. But it's true that there are fewer requirements for the deposit of usable DNA than for deposit of usable fingerprints. And DNA does have more potentials than fingerprints. DNA can produce and has produced familial associations in criminal cases and, with more elaborate methods, might one day provide a lot of information about an unknown source. The same is true of evolving methods using other body material to propose things like a geographical area where the source likely was raised.

You asked about marrow transplants. A marrow transplant host has posted that his marrow transplant means his blood and marrow permanently retain the donor DNA. What happens if, as has been done, in the event of a hand or face transplant, aside from the new fingerprints? I suspect you leave a DNA mix, one profile from skin cells and another from fluids. Do sudoriferous glands in a transplanted hand produce donor or host DNA in sweat? Does a liver produce the donor's DNA in something secreted that would produce a mix in a blood sample? The liver permanently remains donor tissue (hence the continuing need for anti-rejection drugs) and I would think anything it produces would be as if it were still in the donor. Kidneys, too, a much more common transplant. I know that in some of the fundamental experimental work with DNA transplants, they have to "wash" the host DNA from the cells so they'll accept the donor DNA.

Oh. And here's reference to a case:

IT SOUNDS like an open-and-shut case: a clear DNA match is made between semen from a serious sexual assault and a blood sample from a known criminal. Yet in a recent case from Alaska, the criminal in question was in jail when the assault took place. And forensic scientists had already matched the crime sample to the DNA profile of another person who was their prime suspect. It was only after careful detective work that the mystery was solved: the jailed man had received bone marrow from the suspect many years earlier.

Since marrow transplants are frequently between family members, it's not all that unlikely that both could be or become criminals. What a pain that would be if the question was whether the recipient alone or both together did the crime.

http://www.newscientist.com/article/mg18825234.600

[16_Points]

HI 16

You seem a bit fixated about pattern types, of cause pattern types repeat, I think your terminology may be a bit off putting there, surely we are not discussing pattern types and a person with all arches is inherently MORE likely to have his/her friction ridge PATTERN duplicated, that is if they are plain arches, any old school fingerprint expert having done plain arch comparisons knows that. With regard to your comment about AFIS as far as I know most Henry collections were dumped on to NEC AFIS in bulk by NEC in the California. I do not know about other providers but I imagine that for Morfo to be in the running they would have had to have offered a similar deal. My point being that 20 years ago they were very large collections.

No, I'm not fixated on pattern types. But consider the data we have suggests that duplication is statistically more likely in forced minutiae areas such as deltas. Since arches are lacking deltas, one may hypothesize prima facie that they are less prone to duplication. This may prove to be false. But you may recall that I considered the risk of duplication as a conditional probability. And, I submit that the pattern type is a relevant condition affecting that probability. Further, arches are a minor contributor to the population of pattern types (~10-15%). Thus there are fewer opportunities for arches or portions of an arch pattern to be duplicated in an absolute sense.

With respect to AFIS database size, IAFIS has around ~60 million records? How large was it 20 years ago? Is a database of 5 million records considered large? It's a relative question. And I propose that the risk of encountering a near match vs. AFIS database size is not linear but logarithmic or exponential. Hence, the database much reach some critical number before near matches become an issue. And lets not forget that AFIS vendors have been redesigning their search algorithms too. The combination of increasing database size and better matching algorithms is a potent recipe for error.

Going back to the original thread topic . . . how much is enough? No one seems to want a point standard. OK, but no one seems to want any kind of standard. It sounds like a toothpaste commercial, "Four out of five dentists prefer Crest with Tartar Control over Colgate Whitening toothpaste." Our rationale is, "The case examiner and the verifier both agree that this latent print is sufficient for identification." Yes, I know, training, experience, etc.

But how do we express this as the result of a data driven process such that any examiner trained to competency will reach a similar conclusion with a reasonable degree of error, say 5%? This is the question I'd like answered. The toothpaste example is just a foil to frame the discussion. It should not be taken seriously as a criticism of latent print standards.


**Edit**

To clarify the above statement on AFIS. I mean that the risk of a near match increases exponentially or logarithmically with database size. Or so I hypothesize. Meaning that the curve is essentially flat or rises slightly as database size increases. Once the critical number is reached, the slope then takes off (rising very quickly). Without any data (and a weak background in math), it's only a guess.

[Gerald Clough]

Going back to the original thread topic . . . how much is enough? No one seems to want a point standard. OK, but no one seems to want any kind of standard. It sounds like a toothpaste commercial, "Four out of five dentists prefer Crest with Tartar Control over Colgate Whitening toothpaste." Our rationale is, "The case examiner and the verifier both agree that this latent print is sufficient for identification." Yes, I know, training, experience, etc.

But how do we express this as the result of a data driven process such that any examiner trained to competency will reach a similar conclusion with a reasonable degree of error, say 5%? This is the question I'd like answered. The toothpaste example is just a foil to frame the discussion. It should not be taken seriously as a criticism of latent print standards.

I think we really are in the "4 out of 5 dentists" position. The toothpaste example isn't a bad one, because it raises the question of why the four agree and the fifth doesn't. Is the fifth of the opinion that it doesn't matter what you use, so long as you brush thoroughly, or does he really think Colgate is better? Do the four like Crest or just dislike Colgate? Or did Crest just have cuter field reps? Are the four selecting Crest for the same reasons?

There are a lot of issues here. One real one is, I think, whether it's error to hold a conclusion contrary to another conclusion that is acceptable to some number of other examiners. It's very true that, when you get outside of a unit and into litigated cases, properly trained and experienced examiners do indeed produce different conclusions. It's a lot like the dentists. Dentist number five isn't tarred and feathered for differing. The issue is resolved by the finder of fact listening to both examiners and judging who is the most credible.

Is it legitimate to judge whether an examiner is "trained to competency" by observing if that examiner's conclusions are agreeable to other examiners? I have to say it's not. Two examiners with differing conclusions do indeed believe their own conclusion is correct and the other examiner's incorrect. But their judgment is only as good as the effect of their argument to the fact finder. It's probably true that, if you stack three examiners on one side against the dissenter, you are more likely to sway the fact finder. But how, other than that argument, can you show that the dissenter is, for instance, too conservative or too liberal on sufficiency?

I think that at this time, seeking a data-driven standard is chasing a phantom. We would love to have such a standard, but it's not going to happen any time soon, because it remains a matter of expert opinion that states high confidence in a high likelihood of the identification. That's exactly as it is with almost all expert witness fields. It might even be said that a lot of expert livings depend on vague standards. If you have a validated precise standard, is there anything left for the expert but acquiring data and concluding that the data was collected accurately? A point standard does have the virtue of being a standard, of establishing a measure of minimum sufficiency. And if it's high, it at least tends to mean that everyone will agree that it's sufficient, even those who think it too conservative.

[raybie]

Uniqueness is still in the dictionary last I saw...The US law that was designed to filter out pseudo science has now given pseudo science a nice big soap box to try and refute other established sciences by arguing there is no such thing as uniqueness or we just don't know or cannot determine uniqueness. All I can say is that if there is a one and zero in mathematical probability there is always going to be certainty.

Another refutation is that certainty is a circular argument doesn't prove anything. In reality making decisions and acting on those decisions is a certainty. "I think therefore I am " is replaced with "I think therefore I am not too sure ".

Let's not forget that direct observation (through our senses), which is the foundation of well established sciences like biology and the basis of fingerprints is unreliable and has not been tested...Lets take the argument that we are looking at small distorted reproduction of skin surface details. So what! That is why there is the "expert".

Oh that is right, there is that wonderful argument that there is bias! Jeepers every science that requires a microscope and telescope or magnifying glass needs to be re-researched because we have bias.

To those who do not know the difference. DNA is different to fingerprints (nothing like fingerprints).

Until someone provides direct observational proof that fingerprints are not unique I am not going to loose any sleep and am quite happy to say:

Identified, not identified and insufficient...

[raybie]

experiments to determine uniqueness are difficult to set up, however nature has provided excellent support for it in identical twins.

furthermore follow up observations and field tests (millions of comparisons) is strong and compelling evidence.

We can use mathematics to determine if space bends but i can see the bifurcation.

[16_Points]

To be unique or not unique, that is the question . . . well, not really. One of things that makes science science is the body of knowledge AND data that supports it. I agree that simple observation is an important and essential aspect of science. But if we go back to early scientific references, we'll see that it is RECORDED observations that are important. That is to say a faithfully documented body of evidence that supports the proposition (or conclusion as you would have it).

Where, may I ask, is this vast 100 year old body of knowledge that supports fingerprint identification? In the minds of latent print examiners long dead? How can I peer review what only exists in your mind?

Science is not a static enterprise. This type of experience based expertise was once accepted much like medical science. But the current trend in medicine and other fields is toward an evidence based approach. And by that I mean a published body of hard data. Is it so surprising that standards of scientific validity would change over time? I mean, really? Scientific standards and "established conclusions" must be open to revision and review.

There's no harm in coming back a hundred years later and saying, "Hey, maybe we should look at this question a little more closely." By that I don't mean the legal/court implications. In this sense, I am speaking strictly from the perspective of a disinterested scientist.