Is Fingerprint Identification A "Science"
By: Andre A. Moenssens
Date: April 19, 2005
About the Author: Andre Moenssens is the editor of Forensic-Evidence.com, an information center in forensic science, law, education and public policy for lawyers, forensic scientists, educators and public officials. He is emeritus professor of Criminal Law at the University of Missouri and has written many books on forensic science.
Humans and other primates are unique in having thickened, roughened, skin on their fingertips, toes, palms, and soles of the feet. These skin surfaces, when magnified, look like a plowed field, with ridges and furrows, and allow us the useful ability to grip things. Forensic scientists refer to this feature of skin as friction ridges. Skin is never completely dry or clean; grime, oil and sweat on the fingerpads create an impression of the ridge pattern, known as a finger-print, whenever someone touches something. Fingerprints have been recovered from all kinds of surfaces, although it has proved hard to detect them on human skin, which would be a very useful source of evidence if it could be done.
Fingerprints are said to be unique to the individual, as discussed in the article below, and have been admissible evidence in court for over one hundred years. The first person to be convicted on fingerprint evidence was Francesca Rojas, an Argentinian woman who killed her two children in 1892. In the United States, the first conviction on fingerprint evidence was of Thomas Jennings, in Chicago, who was executed for murder in 1912. In the early days, fingerprint investigators had to sort manually through records of prints to find a match. Now fingerprint matching is done by a computer program. The Federal Bureau of Investigation began to automate fingerprint analysis with the Automated Fingerprint Identification System (AFIS) in the 1960s. The AFIS computer scans and digitally encodes fingerprint records into a database. It can match a sample by searching through the database very rapidly—the speed currently is 500,000 prints a second. But even with its long tradition, coupled with new technology, fingerprinting is being questioned as a scientific technique.
Fingerprint identification has been around for a long time. It has nearly a century of court acceptance in the United States. Yet, in the aftermath of United States Supreme Court cases like Daubert v. Merrell Dow Pharmaceuticals (1993) and Kumho Tire v. Carmichael (1999), requiring courts to determine the reliability (validity) of underlying techniques before admitting expert opinion based on it, questions are bound to be raised about the scientific legitimacy of many of the techniques commonly used in crime laboratories, fingerprint identification among them.
Skilled examiners of fingerprint evidence agree that the process of comparing latent fingerprints of unknown origin with inked impressions of known origin is an "art," rather than a science. It requires an examiner to assess, on the basis of experience in dealing with thousands of fingerprints, what parts of an incomplete and partially blurred latent print show visible friction ridge detail that can be used for identification purposes. But whether fingerprint identification is "art" or "science" is clearly no longer relevant to a Daubert inquiry. What needs to be examined is whether the underlying premises upon which fingerprint identification are based have been empirically validated. And these premises are three-fold: (1) the friction ridge detail of the epidermis on the palm side of the hands remains unchanged during the lifetime of an individual, except for accidental or intentional scarification or alteration; (2) friction ridge pattern areas exhibit so much variety of detail that no two patterns are ever found to be exactly the same on the digits (or palms or soles of feet) of one individual or on the digits (or palms or soles of feet) of other individuals; (3) while these friction ridge patterns exhibit an infinite variety of detail, they nevertheless fall within certain broad classes or categories that permit police to store and retrieve millions of prints according to classification formulae.
Premise one has clearly stood the test of time and experience. It has been established in over 100 years of accumulated experience that friction ridge patterns remain unchanged naturally in their ridge detail during the lifetime of an individual. The ridge patterns begin to form during pre-natal life and are fully formed by the seventh month of fetal life.
Premise three has equally proved to be true by verification and experience. We have been able to deal with millions of accumulated fingerprint cards by devising classification formulae based on pattern types and subgroups until the advent of automated computer and retrieval systems, referred to generically as AFIS systems in "the trade," made such classification formulae unnecessary.
Premise number two, that all fingerprints are unique and never duplicated in the universe, is a premise that is harder to prove empirically, despite the fact that all finger-print examiners fervently believe in it. However, in all of the human experience with fingerprints world-wide, no two fingerprints from different digits have ever been found to match exactly. It has been argued that, since millions of sets of prints have been stored in fingerprint files as voluminous as, say, the FBI Identification Section and no exact duplication of friction skin detail has been encountered in these fingerprint repositories, individuality is clearly proved. The problem with this assertion is that it does not stand the test of reason. The millions of sets of prints were never compared against one another for possible duplication of friction ridge patterns. Filing and retrieving prints from such a massive file only results in an examination of a comparatively small number of sets of prints: those with a matching, or approximately matching, classification formula.
There is, however, respectable empirically established evidence of the uniqueness of fingerprint patterns. Studies done by many examiners have shown that the fingerprints of identical twins are different, as are the prints of triplets, quadruplets, and quintuplets. In that sense, fingerprint identification has been found to be even more discriminating than the vaunted DNA (deoxyribonucleic acid) "fingerprinting" method, which cannot distinguish, by today's technology, between the DNA of identical twins. Since inherited traits for similarity in patterns and sub-pattern types are the most common among people who are very closely related, the difference in the prints of such persons certainly can be taken as empirical evidence of fingerprint individuality. Might we not infer from that experience that all fingerprints of different digits are, indeed, different?
Persons skilled in fingerprint identification, who have literally viewed, scanned, and studied tens—if not hundreds—of thousands of individual patterns, do not doubt this. Clearly, if exact pattern duplication were to exist in the world, at least a single instance of this would have been discovered by now. While such claims have been made often, every case, when examined, has established that the prints of different digits that were allegedly "the same" exhibited indeed clearly visible differences that would not have lead an examiner to an erroneous identification. There simply was no duplication of individual ridge detail in prints from different digits.
At the time when fingerprint evidence was first admitted by courts, such empirical evidence or experience in dealing with millions of fingerprint records was not available. If the courts at that era which confronted fingerprint identification evidence first [in Argentina (1892), India (1897), France (1902), and subsequently in England and the United States] had been required to satisfy a Daubert-like decision, perhaps fingerprint identifica-tions would not have been deemed admissible in those early years. With the data that is available today, however, it would be rather ludicrous to argue that the premises underlying fingerprint identification have not been scientifically validated in the face of the accumulated experience of the millions of fingerprints that have been scrutinized by experts.
But there is, today, an opportunity to prove the underlying principle of individuality empirically in a manner that was not available in the past. The tremendous computer databases holding millions of individual finger impressions can today be searched to determine whether pattern duplication exists. While experience has dictated such research is not necessary, the purists (or skeptics) could be satisfied by a rather simple research program that asks our AFIS systems to search, say, a partial individual print of a known person and compare the print against the entire data base. Competent fingerprint examiners feel confident that when the "statistical matches" the computer is bound to throw up are visually examined by them for concordance of individual ridge detail, no two prints from different digits will be found to match.
Scientific validity of evidence that may convict a suspect is, of course, crucial. The fact that the majority of forensic scientists believe in fingerprints and that it has a long tradition is not enough, in itself, to establish their validity. Further experiments, such as the one involving AFIS as suggested by Professor Moenssens, would strengthen confidence in this kind of evidence. Fingerprint theory has hardly been examined using the standard blinded approach which is traditional in science. When an examiner gives a second opinion on a match found during an investigation, he or she is usually aware of the judgement given by the first examiner. This, in itself, introduces a bias, however impartial the second examiner believes themselves to be.
Although computer databases play a huge role in fingerprint analysis, the human element—both in protecting the evidence and examining it—remains indispensable. Recent challenges to fingerprint evidence suggest that the approach is really an applied science, with a artistic or subjective component to it, rather than a pure science like chemistry or mathemetics. For instance, Oregon attorney Brandon Mayfield was identified as a suspect in the 2004 terrorist bombings in Spain. Later, Spanish police arrested an Algerian suspect and Mayfield was released without charge. Investigation of the case showed a number of concerns over the identification of Mayfield's fingerprints on a bag of detonators linked to the attacks. One was the enormous size of the Integrated AFIS system which can apparently pull out confusingly similar candidate prints for further examination. The sheer power of the database can, it seems, work against the investigation process at times.
In another recent case, Stephan Cowans was sentenced to 35 to 50 years in prison for shooting a police officer in 1997 after a positive identification from a fingerprint on a drinking glass. But DNA evidence found at the scene later exonerated him. Poor training on the part of fingerprint analysts led to errors that were not discovered by either the verification examiner or Cowans' own defense team. These are just two cases—there may be many more and they demonstrate that finger-print evidence is not as infallible as the experts might have us believe. So far, there have been 42 Daubert challenges to fingerprints—demanding more scientific validation, as described in the article above. None has been successful, showing that it may be difficult to get the experts to accept a need for further research. But scientific investigation into fingerprint evidence may help tighten up procedures and show the best way of using fingerprint evidence, in the interests of justice being done to suspects and victims alike.
Jackson, Andrew, and Julie Jackson. Forensic Science. Harlow, England: Pearson, 2004.
James, Stuart, and Jon Nordby, eds. Forensic Science. Boca Raton: CRC Press, 2003.
Massachusetts Bar Association. "Lawyer's Journal: Fingerprint Identification and its Aura of Infallibility." 〈http://www.massbar.org/publications/lawyersjournal/article/〉 (accessed March 28, 2006).