CLPEX Chat Board

On Thursday 10 March at 9 pm BBC Radio 4 is broadcasting an investigative programme, 'Fingerprints on Trial'. A number of experts at home and abroad have been interviewed and the programme should hopefully provide a balanced critique of the science and highlight some of the current issues that require to be addressed.

http://www.bbc.co.uk/programmes/b00z5zyc

To listen: http://www.bbc.co.uk/iplayer/episode/b0 ... _on_Trial/

This programme can now be listened to at:

http://www.bbc.co.uk/iplayer/episode/b0 ... _on_Trial/

Hello Iain

Thank you for the posting, what has been assembled in the programme should become a strong wake-up call for experts. Not that I agree with the investigative angle taken by the presenter or some of the interviewed people. Firstly all experts must be able to explain why fingerprints are unique, in a scientific manner not dogma. Secondly experts need to know the circumstances of each of the mentioned cases or research frames so that any particular slant can be corrected or at least balanced on the scales. Experts also need to be aware of the varied researchers and critics who are carving niches and building uncertainty/unreliability without themselves actually conducting fingerprint examinations [eg Cole]. I do agree that the science of fingerprinting deserves criticism for its lack of strategic global administration and governance but this is something unlikely to be corrected before I retire this year. The legacy I leave my Australian colleagues and those who have expressed an interest is an awareness of contemporary scientific literature on how fingerprints develop which is afterall the underpinning principle of fingerprints. The presenter mentioned that the Inquiry report was forthcoming, another wake up call. Cheers from oz. Les

It was an interesting programme but it did not discuss - and the Scottish Fingerprint Inquiry did not discuss - and the 2009 NAS report did not discuss - how criminal justice systems can safely handle forensic evidence that it might be wrong. Just because some fingerprint misidentifications have been found does not mean that the technique is unreliable or that IDs have no probative value. That is the type of black and white thinking that the adversarial system and words like ‘infallible’ encourage. What we need is to find a way to make police officers, criminal justice professionals and juries REALLY understand what a forensic result with a small false positive rate means in the context of the case in hand. Look at this quote from the summary of the NAS report:

"Nuclear DNA analysis enjoys a pre-eminent position not only because the chances of a false positive are minuscule, but also because the likelihood of such errors is quantifiable, the report notes. Studies have been conducted on the amount of genetic variation among individuals, so an examiner can state in numerical terms the chances that a declared match is wrong."

That statement is dangerously misleading and I hope nothing like it is ever heard in a court of law. Firstly an examiner can’t state in numerical terms the chances that the particular declared match in the case is wrong. They can only state the theoretical chance of a match being wrong for matches selected at random. But identifications that are of interest to the police are never selected at random. They are selected because they are suspicious so misIDs, even if they are very rare, will tend to be picked out and used in prosecutions.

Secondly the only thing that is known about DNA identification is the theoretical chances of someone who did not deposit the sample having DNA that is indistinguishable with the technique used. This does not take into account the possibility of contamination of the sample and human errors in the lab. Nothing is known about the likelihood of these errors which will almost certainly swamp the theoretical RMP, the figure given to the court when DNA evidence is used.

I don’t think it is at all intuitive that a very small general possibility of forensic error can, if there is little else to link the suspect to the crime, translate into very little certainty in that case. I don’t see much point in trying to quantify the general likelihood of misID very precisely since there is no way it can be interpreted precisely. Perhaps ballpark figures might be useful as long as people really understand what they mean but the language of the academics is not very helpful here. “The chances of the evidence given that the suspect is innocent” vs. “the chances that the suspect is innocent given the evidence” just sends people’s heads spinning.

I have corresponded about this and discussed it with many people including police officers, lawyers and criminal justice professionals since the Scottish Parliament Inquiry into the McKie case in 2006. None of the professionals (including a lawyer acting for the McKie side) could see that there was anything abnormal or risky about accusing someone of lying because they challenged a fingerprint identification that had no connection with a crime. Nobody pointed out the abnormality to the jury at Ms McKie’s perjury trial and no guidance was given on how it might affect the probative value of the evidence.

I wonder how many people have read the chapter on the individuality of fingerprints from the book "Finger Prints, Palms and Soles", by Harold Cummins PH.D. and Charles Midlo, M.D.

Here is part of that chapter.

Individuality of Fingerprints[1]


The reasoning leading to the conclusion that two identical prints must originate from the same finger is now to be examined. Complete observational proof that prints from two fingers are never identical is unattainable. Even if it were possible to compare every finger pattern with every other one now available, the investigator could only arrive at a conclusion based upon experience with a fraction of the fingers that have existed and that are to come into being. It would be unnecessary, as a matter of fact, to consider such a task, since a significant experience is already accumulated. There are millions of finger-print cards on file, and thousands of experts working with them. Constantly it is necessary to compare the details of patterns, yet never has there been discovered an instance of duplication of different fingers. To make the case stronger, the prints of “identical” twins have been studied exhaustively, without encountering a single case of duplication. The possibility of duplication is put to severe test in these twins, for the two members of the pair have the same inheritance. In spite of the control by the same genetic factors, their finger-print characteristics are never identical, and at best there is merely a close resemblance.

It is a familiar observation that the structures of plants and animals are widely variable. Corresponding parts of the same species may seem to present little or no difference if the inspection is merely casual. But many unlikenesses become apparent if the objects are examined closely, and the number of differences increases as attention is directed to more and more minute characters. The philosopher Leibniz contended, as have many others before and after him, that “there are never in nature two beings which are exactly alike, and in which it is not possible to find a difference.” Thomas de Quincey relates that Leibniz was once explaining the matter to a royal personage; to give point he turned to a gentleman in attendance with a challenge to produce from any tree or shrub two leaves duplicating each other in venation. The challenge was accepted—but the duplicate leaves could not be found. As with leaves, so it is with finger prints. The London newspaper, News of the World, was quite safe when in 1939 it offered a prize of ₤1000 to the person having a finger print identical with any one of a serried of prints published for the contest.

The unique character of every biological aggregate—a single leaf, a finger print, ‘an ear of corn, the striped pattern of a zebra—has been recognized in the axiom, “Nature never repeats.” Without questioning the intended meaning of this axiom, the suggestion might be offered that it read “Nature never repeats exactly.” If nature did not repeat at all, there would be no multiples of the same class—trees and men, whorls and loops, and indeed no universe of fingers to bear whorls and loops. Such repetition, however, is confined to the general molds of things, and in the last analysis of detail “Nature does not repeat.”

In spite of diligent search, an instance of duplication of two finger prints never has been found. This is not unexpected in view of the operation of the law of simple probability, or chance. The occurrence of minutiae at specific points is governed at least in large part by developmental processes which yield random results. Accordingly, the presence and locations of forks, ends and other ridge details may be considered from the same mathematical approach which applies, for example, to the chance of throwing a particular face of a die, or the head of a coin. To some degree finger-print minutiae are subject to control through inheritance, but even the maximum possible “loading” by inheritance is insufficient to counteract the random production of these details.

Pearl[2] refers to the tossing of a coin as “a classical event” because this act has been so frequently used in the discussion of probability. Following tradition, the same example may be chosen here. Imagine first the random tossing of a penny. Because the coin is a thin disc, it is bound not to stand on edge after an ordinary toss. This much is certainty, but no one can be certain of throwing a head, or a tail. The face which lies upward after the throw may be either head or tail, and that each has an equal chance in the result can be determined by trial. If the coin is tossed many times, it will be found that equal numbers of heads and tails have appeared and that no toss can influence the result of any other. The chance of a head, or a tail, may be expressed as the fraction 1/2. “Head” has one in two chances and “tail” has one in two.

Having dealt with this problem in its simplest terms, with the toss of one penny, the chance involved with the use of two pennies is next to be considered. At the toss each coin has an equal chance of falling head up. What is the chance of heads for both pennies? Of the three possibilities – two heads, one head and one tail, and two tails – there is one chance in four that two heads will appear. The mathematical formula for determining this result is simple. Knowing the chance involved in each or the two events, the probability of their occurring together is the product of these two chances: 1/2 X 1/2 = 1/4. On this same principle, the chance of all heads in the toss of any number of pennies may be calculated. If there were 25 pennies the chance of falling all heads is 1/2 raised to the 25th power, or 1/35,554,432.

The chance of all heads in a toss of 25 pennies is small, but an enormous reduction of the chance of obtaining a prescribed result would be introduced by imposing specific restrictions. Assume that the coins are marked for identification, each of them with a different letter, and that the floor on which they are tossed is laid off in 25 squares correspondingly lettered. Postulating that mechanical provisions for the toss insure that one coin will lie in each of the squares, what is the chance that each coin will fall head up within the square corresponding to its identifying mark? The judgment of common sense is that the chance must be exceedingly small; it may be calculated by the formula previously used. The chance of one coin lying head up within its proper square is the product of the chances of these two independent events, namely 1/2 X 1/25 = 1/50. The chance that all 25 coins satisfy this requirement is 1/50 raised to the 25th power, or 1/2,980,232,238,769,531,250,000,000,000,000,000,000,000,000. The chance of the occurrence is therefore so infinitesimally small that from a practical view it may be completely disregarded.

Probability, or chance, is subject to experimental proof in applications such as coin tossing. Though this experimental proof is feasible only in the higher brackets of chance, the correspondence of computed expectation and actual result is a comforting sign that the same law holds when the chance is lowered through increase in the number of items that must be satisfied.

The accelerated diminution of chance, with progressive increase of the number of coins fulfilling the double requirement of “head” and “lying in proper square,” is hardly appreciated unless one actually sets down the numbers. There are good odds that one of the coins will conform to requirement, the chance being 1/50. The chance that two will conform sinks to 1/2,500; for each additional conforming coin the chance is only 1/50 of the preceding, thus; 3 coins, 1/125,000; 4 coins, 1/6,250,000; 5 coins, 1/312,500,000; 6 coins, 1/15,625,000,000, etc.

How does this apply to the individuality of a fingerprint? In brief, the concatenation of 25 specific ridge details existing in the finger-print example chosen may be likened to a successful result in the tossing of the 25 coins. In the finger print the result is already in existence, having been brought about during the period of differentiation of the skin ridges, several months before the person was born. The cogent question is whether an identical result ever might be realized in some other finger. The practical answer to this question is no.

The occurrence of a particular ridge detail in a particular place is not a strictly random event, but that the element of randomness plays the chief role in producing it is evidenced by the differences which occur in “identical” twins. Inheritance is the factor which may influence randomness, but even in two individuals having the same inheritance the combinations of details are widely different. For treatment of chance in reference to finger-print details it seems safe to apply the usual computation for the concurrence of random events, only remembering that in closely related individuals the chance is increased. The increase, however, can not be mathematically corrected. The only correction which is available is to set the chance of duplication of the single items at a figure which is undoubtedly much higher than actuality.

Only to a limited extent would it be possible to determine the actual frequencies of the finger-print characteristics. The pattern used in this discussion is an ulnar loop. Ulnar loops are common, as instanced by their 64% occurrence in the Scotland Yard series. Disregarding the varying frequencies of ulnar loops on different digits, there is thus a mathematical chance of 1/1.6± that two prints from different digits might both be ulnar loops. The pattern in question is an ulnar loop having a count of 11 ridges, and if this feature be also taken into account the chance of duplication of the two characters (pattern type and ridge count) is much smaller. Roscher’s data on ridge counts of 3000 ulnar loops show only 154 having counts of 11 ridges. Still ignoring unlikenesses among different digits, there is a chance of 1/19.5 that two randomly chosen ulnar loops would have counts of 11 ridges. Employing the usual mathematical formula, the chance of concurrence of this particular pattern type and ridge count in two fingers is

1/1.6 X 1/19.5 = 1/31

There are no data on the frequencies of specific minutiae occupying specific positions in patterns. Balthazard and others discuss the chance of duplication of two prints on the basis of a 1/4 probability of repetition of a single detail. This figure exaggerates the chance of coincidence. Wentworth and Wilder point out that the real probability would be closer to 1/50, or even 1/100. Avoiding both undue exaggeration of chance and the possibility of minimizing it through the use of too low a value, we may choose 1/50 as a working figure. If each of the 25 details indicated in (the above example) might be duplicated by chance in a pattern of another finger, the mathematical setup for the chance of duplicating the entire series of details is exactly that which applies to the problem of tossing the 25 coins onto 25 squares, the requirement being that each coin fall head up within the square having its own letter. That chance as shown above is expressed in a fraction in which the numerator is one and the denominator is a number having 43 places!

It must be realized that numerous details are available in a finger-print comparison. Twenty-five are selected in the example discussed, but the number present in one print often reaches a much higher figure, 60, 80 or 100. Another circumstance deserving emphasis is that negative characteristics are not included in the enumeration; the lack of an interruption of fork in an extent of a ridge is a feature which is just as important in the mathematics of chance as the presence and positions of particular minutiae. The mathematical chance of duplication is therefore even smaller than the figure cited above. Even if only pattern type and ridge count are considered in addition to the 25 minutiae, the chance is reduced 31 times; each ridge detail added to the series would reduce the chance by 50 times. The chance of duplication of this finger print is therefore so extremely small that common sense rejects as fantastic the idea of an actual realization. The mathematical treatment is perforce used in evaluating the chance. It is unfortunate that this approach carries the implication that a complete correspondence of two patterns might occur, when as a matter of fact the mathematical reasoning merely supplements observations indicating that such duplication is beyond the range of possibility. Under the circumstances it is impossible to offer decisive proof that no two fingers bear identical patterns, but the facts in hand demonstrate the soundness of the working principle that prints from two different fingers never are identical.[3]

We are reminded in this connection of the distinguished scientist Carl Ernst von Baer, 1792 – 1876, who in his eightieth year declared his conviction that he might not die. The reasoning upon which he based that opinion was: “Thus far, all human beings eventually have died. The saying ‘All men must die’ goes too far; actually it should only claim ‘All men so far have died.’ Even so, the statement is based only upon an experience to which there might be exceptions.”[4] The claimant for actuality of duplication of patterns on two different fingers would take a position about as defensible as that of von Baer on exceptions to the law of mortality. To be sure, a defendant before a court of law might argue the possibility of duplication of finger prints. It might be claimed that an incriminating chance print, shown in expert testimony to be identical with one from a finger of the defendant, is in truth that of another man. The advocate for the defense hails the coincidence as the realization of an occurrence predicted by mathematics! Such a claim, instead of demonstrating that prints from two different fingers are duplicates, proves the weakness of a defense which must resort to patent misrepresentation of the attitude of science.

Workers familiar with finger-print minutiae all affirm that there are no two duplicate prints of different fingers. They recognize many qualities other than the mere occurrences of details. The minutiae, like total patterns, have individuality. The interruption between two ridge ends may be short or long, the ridges may or may not deviate in direction as they terminate; bifurcations exhibit varying spreads, and many similar individual distinctions of minutiae occur.

When all these finer qualities are appreciated, it is not surprising that identifications of individuals are possible when only partial prints are available. Some chance prints contain a limited number of ridge details, the impressions being fragmentary. Authorities agree that demonstration of 12 correspondences of minutiae (and of course not discordances) proves that two prints originate from the same finger. Others are willing to go further, holding that in some circumstances correspondence of six or eight points establishes a positive identification. In the routine of identification, there is naturally no question of the possibility of duplication, since prints of all ten fingers are available for comparison with a new finger-print set. The individual distinctiveness of the complete finger-print set is expressed in the combination of the various pattern types, ridge counts, pattern form and other conspicuous features as well as in the complex of details in each print.



--------------------------------------------------------------------------------

[1] This information is a direct quote from the book Finger Prints, Palms and Soles, by Harold Cummins PH.D. and Charles Midlo, M.D. (Dover Publications, Inc, New York 1961) p. 149-155

[2] Introduction to Medical Biometry and Statistics, 3rd edition. Philadelphia, W. B. Saunders Co., 1940

[3] There is an extensive literature on the philosophy of proof, as it relates to questions such as that here concerned with individuality of a finger print. The Problem of Proof, by Albert S. Osborn (2nd ed., Newark, The Essex Press, 1926) presents an excellent discussion of the canons of proof in legal applications; though the problems are illustrated especially by disputed documents, the general arguments (see especially his Chap. 25) and references to the literature are equally pertinent to the issue of finger-print proof. The tenets and history of the theory of probability are succinctly outlined by Florian Cajori in his A History of Mathematics (2nd ed., Now York, The Macmillan Co., 1931).

[4] From A. Ecker, 100 Jahre einer Freiburger Professorenfamilie, 1886—Quoted by E. Stemplinger, Von berühmien Ärzten, R. Piper & Co., 1938.

Hello Earle,

Thanks for the post re Cummins/Midlo, always nice to read this material, very much pioneering stuff from which fingerprints has taken advancement as well as the science of dermatoglyphics.

"[i]In spite of diligent search, an instance of duplication of two finger prints never has been found. This is not unexpected in view of the operation of the law of simple probability, or chance. The occurrence of minutiae at specific points is governed at least in large part by developmental processes which yield random results. Accordingly, the presence and locations of forks, ends and other ridge details may be considered from the same mathematical approach which applies, for example, to the chance of throwing a particular face of a die, or the head of a coin. To some degree finger-print minutiae are subject to control through inheritance, but even the maximum possible “loading” by inheritance is insufficient to counteract the random production of these details[/i]."

Cummins et al wrote this material in the 1940's so in context we can now compare the scientific material they had available with what is currently known. The third sentence of this quote is significant. At no time have the critics and probability theorists put their finger on how the individual fingerprint features occur. If they did then they would know the variables of biological development cannot be governed by the science of mathematics. As Cummins rightly highlights back in the 1940's our experience in observing fingerprint patterns places enormous confidence in the principle of fingeprint pattern uniqueness. How much further confidence can we rightly claim in the year 2011 following the successful introduction of computerisation and measuring the reliability of non-duplication on a massive scale. The Law of Large numbers cannot and must not be avoided. The critics and probability theorists need to take account of both what Cummins was proposing and what supporters of fingerprint science are still advocating, fingerprints are unique. That uniqueness is detectable in ten print examinations and in latent. Regards from oz. Les

But then of course the mathematicians will say the Harold Cummins PH.D. and Charles Midlo, M.D. are not real scientists?

As we all know the tossing of the coin is only a binary approach, however with fingerprint detail it is not just if there is an occurance there, but what kind of occurance and what direction it points. Even the throwing of a dice is not a good example.

My tupence worth.

In my posting I was making the distinction between the theoretical random match probability for DNA identifications and the likelihood that a specific identification selected for use in a prosecution case is wrong, taking into account rmp, the possibility human error and the other facts of the case. If the totality of a prosecution case is that the identified person denies depositing the sample (no other reason to connect the subject to a crime), then a finder of fact should conclude that the likelihood of error in that case is not only much much higher than the theoretical rmp, it is also much much higher than the general likelihood of human error during the identification and verification process (whatever that might be). My other point was that this is poorly understood.

The chapter from the Cummins and Midlo book is making an argument that the theoretical rmp of clear ten prints is zero.

Hi Steve
I read somewhere that a famous person in history visited a prison where only one person behind bars said 'yes, I am guilty I deserve to be here', he ordered the warden to release this prisoner immediately as he may infect the other prisoners.

You should be careful you may infect the other mathematicians!

I love statistics and probability. Ponder this, the probability of achieving a royal straight flush is 1 in 2,598,960 possible hands. So how may hands does it take to get a royal straight flush? One! To believe in probabilities, one must first accept that the possibility exists.

How funny that someone called Gamber is talking about Poker. (Or is it???)

Curiously, when I worked out the probability of 5 cards forming a Royal straight flush I got 1 in 649,740.
This is uncoincidentally 4 times the probability you quoted.
Did you perhaps quote the proability of a Royal Straight flush in Hearts (or any other specified suit).

David,

Not a poker player? Yes, a flush indicates the straight is all in one suit. (Edit... I wasn't thinking that David's question was about a particular suit. I'm an idiot and wasn't thinking clearly when I posted this... sorry)

The probability of any royal flush (royal straight flush is redundant) is 1 in 649,740. The royal flush is 10, Jack, Queen, King, Ace, all in the same suit. So, to be dealt a royal flush you first have to receive any 20 out of 52 cards in the deck, the next card can be 4 of the 51 remaining, the next card would have to be 3 out of the remaining 50, the next card 2 out of the remaining 49, and finally only 1 card out of the remaining 48 will satisfy the royal flush.

Okay, but what about Texas Hold 'Em? What are the odds of a Royal Flush in that game? And does that beat the odds of two people having eight "matching" points in a fingerprint of each that would be indistinguishable in severely distorted latent prints?

OK, so you have a boat in the middle of the Pacific Ocean with 52 people on board 1 has a knife sticking out of the heart, even with odds of 1 in 649,740 I suppose a jury could be sucked in to thinking that 1 person is guilty. But would they? How often does a royal flush come up, I am no gambler I have no idea.
I would prefer to hear the fact that this is the guilty person myself.