The general scientific method used to develop the T-Model is described below.  It is the same method as described by Richard Feynman (click here).

1.  I make a guess, then 

2.  I compute the consequences of the guess to see what it would imply, then

3.  I compare the computational results of the guess to experiment to see of it works. 

4.  If the guess disagrees with experiment, then it's wrong

 

The T-Model uses results gathered by experiment to make best estimates for quantitative and qualitative values for fingerprint ridge feature types in position (see Validation Study).  It is the results gathered by experiment that are the real foundations of the T-Model. 

Experiments are never perfect or complete and as a result they are always ongoing.  Consequently, the T-Model will continue to develop based on results gathered by well-controlled, reproducible, honest experiment.   

Foundations for the T-Model

Basic foundations for the T-Model which have been previously supported by the FBI are as follows:

1. Utilizes the ridge unit approach to illustrate a ridge (as opposed to compound ridge formations), and

2. Considers relevant fingerprint population.

These basic foundations are supported by position statements published by the Federal Bureau of Investigation, U.S. Department of Justice, in the June 1972 FBI Law Enforcement Bulletin "An Analysis of Standards in Fingerprint Identification" [49] as follows:

1. “From the general working standpoint, the types of characteristics can be narrowed to the ridge ending, the bifurcation, and dot.  Two additional characteristics are commonly distinguished or mentioned because of the ease with which they are visually discernible as distinctive formations.  These are the short ridge and the enclosure or island.  Technically, however, unless these formations are extremely short length or extent, they can be used legitimately as two ridge endings instead of a short ridge and as two bifurcations instead of an enclosure.”

2. “From a theoretical standpoint any fingerprint, when it is identified, needs to be distinguished from every other fingerprint existent in the world.  A little reflection will show, however, that this greatly exceeds the practical aspects of identification, since in the average case a crime scene print is initially an investigative tool to identify a suspect who was in the same country, State, or locality at the time the crime was committed.”

In addition to using a ridge unit approach and relevant population to define discrimination values for individual amounts of ridge features and match probabilities for amounts of corresponding ridge features found in two impressions, the following additional variables serve to complete the foundations for the T-Model:

Pattern Force [Link]

Friction Ridge Skin Elasticity Threshold [Link]

Intervening Ridge Count to Nearest Neighbor (Level II) [Link]

Ridge Unit Clarity and Reliability [Link]

Ridge Unit Quality of Agreement [Link]

"A reasonable probability is the only certainty."

Edgar Watson Howe

The T-Model:  A Probability Model for Fingerprints

• First version published online August 2008.

• Open source fingerprint match probability model for fingerprints. 

• Estimates conservative discriminating values for individual fingerprint ridge feature types and clusters of fingerprint ridge feature types in position. (Note:  The discriminating values for the most common individual ridge feature types used by latent print examiners have been corroborated by independent experimentation.  Extended experimental testing to corroborate, refute or refine existing values for other fingerprint ridge feature types is ongoing.)

• Utilizes a ridge unit approach to describe a ridge.

• Considers Level I, II and III fingerprint ridge detail.

• Considers intervening ridge count, ridge path, discrepancies, and "red flags".

• Considers friction skin elasticity to define spatial relationship thresholds between fingerprint ridge features. 

• Can be applied to any quantity and quality of fingerprint ridge features.

• Can be applied to any population group. 

• Defines thresholds to infer identification or exclusion based on relevant population for the case at hand.

• Estimates conservative, upper-bound numbers of close matches (i.e., look-alikes) present in any population group.

• Defines qualitative metrics for reduced levels of ridge feature clarity, ridge feature reliability, and quality of agreement. 

• Applies objective quantitative and qualitative values for fingerprint ridge features based on subjective interpretation of ridge feature types in position and subjective assessment of ridge feature clarity, reliability and quality of agreement during the analysis and comparison phases of a latent v. exemplar examination. 

• Evaluation of inference for identification is objective.  It is the T-Model that establishes inference for identification, not the examiner.

• Establishes inference for identification when the estimated number of close matches is calculated to be less than 1 based on relevant population for the case at hand.

• Has been successfully tested against the largest and best fingerprint close matches ever presented during examiner training, published or otherwise recorded.

• Quantitative values and qualitative metrics are based on results from experiments and represent a "best guess" for what are the true (i.e., conservative, lower-bound) discriminating values for individual and clusters of fingerprint ridge features.  

• Considers diminishing areas in fingerprints, i.e., funnel patterns, due to pattern force.

• Offers an open source "T-Model Fingerprint Calculator" in Microsoft Excel format.

• Can only be used for distal segments of the fingers at this time (not medial or proximal phalanges and not palm prints).

• Has shown to be more accurate that human decision-making by expert latent print examiners to estimate numbers of close matches present in fixed population groups. 

• Has shown to be more accurate that human decision-making by expert latent print examiners to identify the largest and best look-alikes as insufficient to infer identification.

NEXT PAGE >>> 

The T-Model is intended to serve as a tool for latent fingerprint examiners to express criteria to assert sufficiency to infer identification in a more precise form than has been done in the past, and to reliably ferret out and identify look-alikes in order to help prevent erroneous identification. 

The T-Model is designed to provide quality control for examiner work performance in which examiner interpretation of ridge formation types, assignment of quantitative weights, and assessment of ridge formation clarity and reliability, and quality of agreement between pairs of corresponding ridge formations in two impressions can be inspected critically. 

The T-Model is also meant to be used as a way to scientifically validate problem, complex or borderline latent fingerprints as sufficient to infer positive identification and invalidate previous latent fingerprint identifications bearing amounts that are insufficient to infer positive identification with a degree of probability that borders on certainty, and as a result potentially exonerate innocent persons wrongly convicted.

The following high profile erroneous fingerprint identifications illustrate the need to minimize the likelihood for expert fingerprint examiners to make wrong conclusions:

Brandon Mayfield [22] [58]

Shirley McKie [22] [57]

Stephen Cowans [52]

As a result of these wrong conclusions and inability of expert  examiners to elaborate on criteria for sufficiency to establish positive identification, judges have ruled to limit and disallow fingerprint evidence in the courtroom, i.e. Maryland v. Bryan Rose [55] [64] and Maryland v. Lamont Anthony Johnson [56].

There is an ever increasing pool of researchers, scientists and fingerprint examiners who do not accept the current non-statistical method of “professional judgment” to establish valid basis for sufficiency to individualize [59]. 

The T-Model is the author’s best response at this time for a possible way to express criteria for sufficiency to individualize in a more precise manner than any non-statistical or fixed-numerical approach that is currently used throughout the fingerprint community.