Preamble

During the course of routine casework fingerprint examiners regularly find non-matches containing minimal amounts of corresponding quantitative-qualitative ridge formations. Areas displaying small look-alike amounts are not uncommon, while as a result of decades of experience by thousands of fingerprint examiners around the world, non-matches displaying unusually large amounts have been documented, presented during examiner training, and published.  Some of the most prominent and experienced fingerprint experts in the world have singled out and set apart the largest and best look-alikes ever seen in a non-match, and based on the direct experience and observations by these experts, the largest and best amounts of corresponding friction ridge formations in a non-match, to date, have been identified (see below).

Although look-alikes exists in fingerprints, false associations to look-alikes have been recorded, published and presented during examiner training.  One of the most notable erroneous fingerprint individualization ever recorded to a look-alike was made by 3 FBI latent print examiners and 1 independent latent print examiner to Brandon Mayfield, a Portland Oregon attorney, during the investigation of the terrorist train bombing that occurred in Madrid, Spain on March 11, 2004.  Based on conservative interpretation of the amounts of quantitative and qualitative corresponding ridge formations found in the Madrid latent fingerprint impression versus the Mayfield exemplar, and based on an conservative upper bound IAFIS fingerprint population of 530 million, which represents the approximate size of the FBI criminal fingerprint database [32] which produced the Mayfield exmplar, the total value, e.g. T-Value, for the amount of ridge formations “in agreement” between these two impressions clearly failed to establish that the number of estimated look-alikes would be less than or equal to 1.  In addition, the T-Value also clearly failed to exceed that for the largest and best amounts of corresponding ridge formation ever recorded in a non-match, i.e. The Chesapeake IAFIS and Clark non-matches. This can be easily demonstrated by applying T-Model formulae to the evidence print belonging to Daoud Ouhnane that was falsely associated to the Brandon Mayfield exemplar by the FBI in the Madrid train bombing (see Madrid Error).  Fingerprint examiners are encouraged to test this out for themselves using The Formulae.

It is significant to note here that the need to consider amounts of corresponding ridge formations previously found in non-matches is supported by position statements from prominent latent print examiners.  With regards to how much is enough to establish valid basis for sufficiency to infer identification, the following statements have been made:

David Ashbaugh states it should:

“exceed amount personally ever found in non-matches and exceed amount anyone ever found in non-matches [2]”. 

John “Dusty” Clark writes:

“Sufficiency is the amount of detail present in each unknown impression, and how this amount of detail compares to the closest non-match ever found with concurrent equal levels of detail [10].”

Christophe Champod writes:

"In a nutshell, the individualization will be reached when the examiner observes an extent of agreement (across the three levels of legible features) that exceeds the largest extent of correspondence he observed through his/her training and experience in comparisons involving non-matching entities. [102]"

In addition, the theory of individualization, which requires corresponding details in two objects to exceed the largest and best amounts ever seen in a non-match, is not new to forensic science. The Association of Firearms and Tool Mark Examiners (ATFE) apply this theory as it relates to tool marks as follows [11]:

1.         The theory of identification as it pertains to the comparison of tool marks enables opinions of common origin to be made when the unique surface contours of two impressions are in "sufficient agreement".

2.         This "sufficient agreement" is related to the significant duplication of random tool marks as evidenced by a correspondence of a pattern or combination of patterns of surface contours. Significance is determined by the comparative examination of two or more sets of surface contour patterns comprised of individual peaks, ridges and furrows. Specifically, the relative height or depth, width, curvature and spatial relationship of the individual peaks, ridges and furrows within one set of surface contours are defined and compared to the corresponding features in the second set of surface contours. Agreement is significant when it exceeds the best agreement demonstrated between tool marks known to have been produced by different tools and is consistent with agreement demonstrated by tool marks known to have been produced by the same tool. The statement that "sufficient agreement" exists between two tool marks means that the likelihood another tool could have made the mark is so remote as to be considered a practical impossibility.

Consistent with the theory of identification as it relates to tool marks, any amount of corresponding ridge formations that are found in two impressions which exceeds the largest and best look-alike ever recorded [based on relevant population] may be considered significant enough that the likelihood for duplication by a different source is so remote as to be considered a practical impossibility.  

Error Rate of the T-Model 

Based on the National Academy of Sciences (NAS) report Strengthening Forensic Science in the United States:  A Path Forward, in the case of all analyses leading to an identification or exclusion conclusions, the analysis process must be subjected to performance and validation studies in which appropriate error rates can be defined and estimated. In terms of latent fingerprint analysis, the aim is to address whether or not a particular fingerprint fragment is associated with one particular source.  Fingerprint analysis leads to “individualization” and “exclusion” conclusions—for example, whether or not a particular fingerprint fragment belongs or does not belong to individual X.  Although conclusions of absolute certainty made by latent print examiners have been deemed unjustified by the NAS, the paradigm of yes/no conclusions has been found useful for describing and quantifying the accuracy with which fingerprint analysis can provide answers.   In such situations, results from analyses for which the truth is known can be classified in a two-way table as follows [91]:

The accuracy of the T-Model to make correct conclusions, e.g., correct positive identifications and correct exclusions, was subjected to rigorous performance and validation studies as follows: 

The T-Model was tested against a random sampling of known matches and known non-matches which included a sampling of 5 largest and best amounts of corresponding ridge features ever seen in a non-match (e.g., fingerprint fragment “look-alike”) and a sampling of 5 largest and best amounts of non-corresponding ridge features ever seen in a match.   The study comprised of 200 fingerprint samples as follows: 

1) 100 samples of known matches as follows: 

• 95 samples of flat fingerprints from different criminal arrest records from the same individual were used.  The 95 match samples were selected at random from criminal ten-print records at the San Jose Police Department Central Identification Unit.

• 5 samples of fingerprint fragments representing largest and best amounts of non-corresponding ridge features ever seen in a match were used.  The samples were obtained from examiner training, published material, and criminal casework.  The particular matches used were as follows (see attachment):

• Judith Miller Match
• Triplett Match
• Ashbaugh Match
• Maciel Match
• Templeman Match

2) 100 pairs of known non-matches as follows:

• 95 samples of flat fingerprints from different criminal ten-print records from a different individuals were used.  The 95 non-match samples were selected at random from criminal ten-print records at the San Jose Police Department Central Identification Unit.

• 5 samples of fingerprint fragments representing largest and best amounts of corresponding ridge features ever seen in a non-match were used.  The samples were obtained from examiner training, published material, and criminal casework.  Specifically, the matches used were as follows.  Included was the most notable erroneous fingerprint identification ever recorded in the history of fingerprints, e.g., Brandon Mayfield.  The particular non-matches used were as follows (see attachment):

• Chesapeake IAFIS Non-Match [Link]
• Clark CAL-ID Non-Match [Link]
  
• CLPEX Non-Match [Close Calls: Print 001]
  
• CLPEX Non-Match [Close Calls:  Print 003]
• Brandon Mayfield Non-Match [Link]

Note:  Only corresponding ridge features in each non-match were used.

The T-Model was used to determine, for each of the 200 samples, whether the amounts of corresponding and non-corresponding ridge features present in the impressions were or were not sufficient to infer positive identification or exclusion, and the true answers were considered known. The qualitative and quantitative values for the ridge features in each were assessed, and based on T-Model theory, the determinations to infer positive identification or exclusion were made.   The validation study returned the following results:

In the test, if the analysis indicated a “non-match” but the fingerprint actually came from the same person, then the analysis returned an incorrect result.  If the analysis indicated a “match” but the fingerprint actually came from a different person, then the analysis returned an incorrect result.  The study identifies as many sources of error as possible that can affect both the accuracy and precision of the application of the T-Model for fingerprint analysis.  The tests are reproducible and falsifiable.

The accuracy of the validation study can be assessed in different ways.  Four characterizations and their associated measures are given below.  Each one is useful in its own way: the first two emphasize the ability to detect an association; the last two emphasize the ability to predict an association:

• Among match samples, the fraction that is correctly identified by the test is called the “sensitivity” or the “true positive rate” (TPR) of the test. In this table, the sensitivity would be estimated as [100 / (100+0)] x 100 = 100 percent.

• Among non-match samples, the fraction that is correctly identified by the test as a correct exclusion or inconclusive decision (e.g., decision of insufficiency to infer positive identification) is called the “specificity” or the “true negative rate” (TNR) of the test. In this table, the specificity would be estimated as [100 / (0+100)] x 100 = 100 percent.

• Among match samples, the fraction that actually turns out to be a match is called the “positive predictive value (PPV)” of the test. In this table, the PPV would be estimated as [100 / (100+0)] x 100 = 100 percent.

• Among non-match samples, the fraction that actually turns out to be non-match is called the “negative predictive value (NPV)” of the test. In this table, the NPV would be estimated as [100 / (0+100)] x 100 = 100 percent.

The above measures emphasize the ability of the fingerprint analysis using the T-Model to make correct determinations.  Each estimate (of sensitivity, specificity, PPV, NPV) is associated with an interval that has a high probability of containing the true sensitivity, specificity, PPV, NPV. 

It is significant to note that the inclusion of samples of largest and best amounts of corresponding ridge features ever seen in a non-match and samples of largest and best amounts of non-corresponding ridge features ever seen in a match insures the performance test is rigorous.

“Error rates” are defined as proportions of cases in which the analysis led to a false conclusion. For example, the complement of sensitivity (100 percent minus the sensitivity) is the percent of false negative cases in which the sample was from a different source but the analysis reached the opposite conclusion. In the above table, this would be estimated as 0 percent. 

Similarly, the complement of specificity (100 percent minus the specificity) is the percent of false positive cases in which the sample was not from a different source but the T-Model concluded that it was. In the above table this would be estimated as 0 percent.

A global error rate could be defined as the percent of incorrectly identified cases among all those analyzed. In the above table this would be estimated as [(0+0)/200] x 100 = 0 percent.

The validation study supports the value of the T-Model in establishing inference for positive identification or exclusion with, at a minimum, a degree of probability that borders on certainty.  Based on the results, the T-Model has a high probability of containing the true value when attributing identity or exclusion of source for fingerprint evidence.

NEXT PAGE >>> 

In order to test the accuracy and reliability of the T-Model to ferret out look-alikes as insufficient infer positive identification, largest and best amounts of corresponding ridge formations ever seen in a non-match, presented during examiner training, published or otherwise recorded, including some of the most notable erroneous identifications ever recorded, were researched and assembled.  The largest and best amounts of quantitative-qualitative corresponding ridge formations ever seen in a non-match, including some of the most notable erroneous identifications ever recorded, i.e. Brandon Mayfield and Shirlie McKie, were taken from the following sources:

David Ashbaugh [19]
Pat Wertheim [20]
IAFIS Non-Match [21]
Ed German [22]
Kasey Wertheim [23]
John "Dusty" Clark [26]

Look-alikes are a part of everyday life.  They are found in people, animals, and also in fingerprints.  The best look-alike fingerprints ever seen by the author are the Chesapeake IAFIS and Clark non-matches.  They have the most and best amount of corresponding ridge formations from different individuals that the author has ever seen published, presented during examiner training, or otherwise recorded.

Look-alike people

Look-alike lion cubs

Look-alike fish