How the work being done on statistical fingerprint models provides the basis for a much broader and greater impact affecting many areas within the criminal justice system

Abstract

In the process of developing and improving statistical models to address flaws in the examination and interpretation of highly selective fingermarks, the groundwork is being laid for a much broader and greater impact. This impact will arise from the use of these same improved statistical methods to exploit information from the examination of fingermarks with lower degrees of selectivity—those fingermarks traditionally considered to be devoid of evidentiary value. To the contrary, research has shown that fingermarks of lower selectivity have much to offer. They occur very frequently: much more often than those assessed to be sufficient for inclusion in existing fingerprint examination processes. In individual cases, they occur in locations and numbers that can provide important new information for investigators and additional routes to further investigation. As evidence contributing to proving a case, they can provide detailed activity-level information and new avenues to address the relevance and probative value of other direct and circumstantial evidence. The broader application of fingerprint models to these traditionally unused fingermarks of lower selectivity needs to be specifically developed and implemented to realize the contributions and to responsibly manage the risks and benefits.

1. Introduction

Fingerprints are a highly selective means of personal identification and their value as highly probative evidence of who made a fingermark is universally accepted. However, for many years, the way that fingerprints were examined, reported, and offered as evidence received little attention from the broader scientific community and fell short of good scientific practice in many respects. This included the use of non-standardized and poorly defined methods, often conducted without documentation of observations, and the offering of categorical conclusions of identification to a single individual. The categorical conclusions were a mandatory ethical requirement of professional practice: any opinion of possible or probable association was explicitly proscribed. This was an all-or-nothing approach: either a definitive identification of this individual to the exclusion of all others or, failing in this, the assertion that the fingermark has no value at all in support of association (IAI 2010; Smith et al. 2011; Triplett and Everist 2015).

This practice had three effects. First, it preserved the reputation of fingerprint evidence as an infallible means of identification: the evidence was only offered when it was deemed certain. Second, it placed strict control on individual fingerprint examiners: you needed to be absolutely sure of your identification, or you didn’t make the call. Third, it blocked the use of any fingermarks with lower degrees of selectivity.

When the broader scientific community began seriously considering fingerprint practices, assertions of infallibility and reporting identifications as fact became the focus of strong criticism and were soon acknowledged as unscientific and unacceptable (Cole 2008; Mnookin 2008; NRC 2009, 135-145). This led to a tempering of fingerprint conclusions, softening of the concept of identification and redefining the word itself to mean, for example, “the conclusion that the observations provide extremely strong support for the proposition that the impressions originated from the same source and extremely weak support for the proposition that the impressions originated from different sources” (OSAC 2018).

Additional substantial improvements in fingerprint examination practices have also been made, most notably more transparent practices, explicit documentation of observations in support of examiner conclusions, and measures to recognize and address sources of bias.

As discussed in this Special Issue, statistical models for fingerprint analysis now offer still greater means for substantial improvements in how, and how well, we conduct fingerprint examinations and comparisons: objectivity, transparency, efficiency, standardization, reviewability, and quality control. Overall, they change the paradigm from one where the accuracy and effectiveness of fingerprint evidence depends almost entirely on the subjective judgment of individual experts to one that includes and exploits objective measurements. These measurements provide the opportunity for scientific assessment and systematic improvement of how experts examine and interpret fingerprint evidence.

As important as these improvements are, there is a greater opportunity: one that substantially expands capabilities, rather than improves existing ones. Our work (Stoney et al. 2020; Stoney and Stoney 2021) has shown the potential for a distinct, revolutionary impact from the application of statistical fingerprint models: enabling the use of the fragmentary, partial fingermarks that are currently unused within the criminal justice process. The earlier practices of fingerprint examiners entirely blocked the use of these “non-identifiable fingermarks” as associative evidence, and the current improvements, working backward to temper assertions and strengthen scientific foundations, are focused only on very highly selective fingermarks.

Fragmentary, partial fingermarks of low or even moderate selectivity¹ are currently unused within the criminal justice process, yet they have a potential impact in an order of magnitude more cases than those where fingerprint evidence is currently used. Furthermore, within each case, the use of more fingermarks increases the chances that reasonable inferences can be made about when an individual was at a location and what the individual was doing. These and additional benefits have been identified by practitioners as highly significant and valuable at several key points within the criminal justice process (Stoney 2020; Stoney and Stoney 2021).

Once appropriate and vetted statistical fingerprint models are in place, there remains only one significant barrier to the implementation of this capability: it is different. To appreciate, or to even recognize, the potential requires a change in perspective. The well-established role of fingerprint evidence as a highly selective, highly probative means of association makes this change particularly difficult.

The first part of this article discusses the terminology that is used, which we have found essential to adapt for different audiences. This is followed by a section which is intended to nurture a new perspective for fingermark evidence, going beyond the common (nearly exclusive) use of fingermarks as strong support for identifying an individual, to consider its character if used as a different type of circumstantial evidence, less dependent on this strong support. The next part of the article illustrates the new potentials and capabilities that follow from acknowledging, accepting, and using the full range of fingermarks: not only those with high selectivity, but those with moderate or low selectivity. The last portion of the article introduces and briefly considers issues of implementation, perception, and risk.

2. Terminology

In extensive discussions about the research and results presented in this article and Stoney (2020; Stoney and Stoney 2021), we have found that the term “fingerprint” carries with it a variety of meanings, some of which can make what we present here confusing. Alternative terms are necessary, and this section outlines the rationale and context for the terms that we use in the rest of the article.

When fingers touch things they often leave traces, which often have discernable details.

We refer to these traces as “fingermarks” to distinguish them from the term “fingerprints” which is used in various different ways and contexts to mean different things. By fingermark, we mean any trace resulting from a finger touching a surface, independent of its clarity, size, or how much detail it contains.

There is a long history and a widely, bordering on universally held view that when there is enough detail within a fingermark, it can be very strong evidence that a particular finger touched the surface and made the mark. Over the past 20 years, there has been a dramatic increase in discussion, research, and debate about fingerprints revolving around a number of important issues such as:

• What level of detail in a fingermark should be considered sufficient for an expert or a court to conclude it was made by a particular finger

• Whether in fact a fingermark can “match” a particular finger with certainty, and so whether comparisons should be viewed as degrees of correspondence rather than matches

• If so, how the degrees of correspondence should be determined, measured, and communicated

• How different methods and procedures for acquiring information from fingermarks result in greater or lesser reliability and greater or lesser rates of false associations and false exclusions

• How different statistical models can be used to evaluate the probability that information acquired from a fingermark can be used to determine the probability that the mark was made by a particular finger

• Whether these statistical models should replace or supplement the work of expert examiners

• How error rates should be calculated and evaluated vis-a-vis different statistical models, different expert examiners, and different procedures and methods used by examiners

These are important current issues, that have been debated and discussed extensively in other articles both in this issue and elsewhere, but all participants have shared the view that fingerprints are a highly selective means of personal identification. The value of fingerprints as highly probative evidence of who made a fingermark is universally accepted.

The present article does not involve or bear on these issues. It is about something else: consideration of the potential information that can be obtained from those fingermarks that all of the people, models, criteria, experts and researchers involved in the discussions above agree do not have enough detail to be highly probative evidence of who made the fingermark. The utility of these fingermarks has received very little attention. Long standing fingerprint examination practices have ignored them because there has been no established way to judge their significance (Stoney et al. 2012) and the current focus for changes in fingerprint examination practices is almost entirely on very highly selective fingermarks. Serious consideration of fingermarks of lower selectivity is new. Clear communication of the distinction from traditional fingerprint work is essential for constructive discussion of the potential benefits, costs, and risks of using these fingermarks.

Given the ongoing changes in how fingerprint evidence is conceptualized, assessed, and communicated, we have found two different terms to be useful when explaining this distinction to different audiences. For audiences long familiar with and generally accepting the idea that a fingerprint “match” can definitively identify a fingerprint as coming from a specific individual, we have used the term “Non-Identifiable Fingermark.” These audiences have included most lay persons, attorneys, judges, and older fingerprint experts. For audiences that reject the concept of absolute identifications based on expert judgment, and embrace the application of statistical modeling and measurement, we have adopted the term “Fingermarks of Low to Moderate Selectivity.” These audiences have included many scientific, legal, and evidentiary scholars, statisticians, and more recently trained fingerprint experts.

Figure 1 illustrates the historical practice of fingerprint identification and the meaning of the term Non-Identifiable Fingermark (NIFM). Moving upward on the vertical scale represents more detail in the fingermark. Expert assessment of the detail results in an assessment of the sufficiency to support an identification decision. All competent experts agree in sufficiency for fingermarks with abundant detail, as represented by the upper rectangular area that is shaded dark green. The trapezoidal area just below (shaded light green) represents levels of detail where, based on differences in expert training, experience or risk assessments (Biedermann et al. 2008) some experts would decide on sufficiency for an identification and others would not. The lower rectangular region (shaded dark gray) represents fingermarks where all competent experts would agree that there is insufficient detail to support an identification decision.

Figure 2 illustrates an alternative nomenclature. Moving upward on the vertical scale represents more detail in the fingermark. A continuum of fingermark selectivity is represented by this scale. Selectivity of a fingermark is conceptualized as the denominator of the frequency of occurrence of individuals that could have made the mark. A mark that could have been made by one in a million individuals would be more selective than one that could have been made by one in 100,000 individuals. The upper portions of the pyramid in Fig. 2, showing a continuum from the apex of darker to lighter shades of green, represent very highly selective fingermarks: conceptually, those that 30 years ago would have been judged to be identifiable by fingerprint experts. Patterns of detail in very highly selective fingermarks might actually occur less frequently than once in populations of a million or more individuals. Moving downward, marks with less detail and lower selectivity are shaded lighter green and blend into an area that would parallel the expert-defined “sufficiency for identification” level in Fig. 1. Well below this level are fingermarks of moderate and low selectivity. The meanings low and moderate selectivity are not rigorously defined. They represent a range where it would be expected that relatively small populations (hundreds to tens of thousands) might have more than one individual that could have made the fingermark.

There are other possible sets of nomenclature that can be used to describe or categorize the range of fingermarks based on the amount of detail and how it is determined. We are not offering these two terms as the best means to do so in other contexts. However, we have found them to be effective in presenting and discussing the research and findings about fingermarks with levels of detail so low that they have heretofore not been considered for evidentiary value.

3. NIFMs: an exceptional form of circumstantial evidence

Circumstantial evidence is indirect evidence that does not, on its face, prove a fact in issue but gives rise to a logical inference that the fact exists. Circumstantial evidence requires drawing additional reasonable inferences to support the claim (Wex 2022a). By its nature circumstantial evidence is non-definitive. However, for fingerprint evidence there is a long history of use and expectation that it is definitive—or nearly so—for personal identification. In fact, it is a long and enduring practice that those fingermarks that experts judge to be insufficient for identification are not used as associative evidence. Nonetheless, these fingermarks often have discernable characteristics shown by statistical fingerprint models to form the basis for lesser degrees of correspondence and probability of occurrence within a population (Stoney et al. 2020).

3.1 Measurable associative value

Using a statistical fingerprint model, NIFMs (a defined set of fingerprint minutiae within the context of friction ridge structure) can be tested for a measure of associative value based on how frequently the set of ridge characteristics is expected to occur within a population. In a research environment, score-based likelihood ratios have been adapted to this purpose (Stoney et al. 2020). This approach may or may not be part of an ultimate application, but it is clear (in parallel with the underlying assumptions of this Special Issue) that statistical models addressing the measurement problem will not be long in coming.

Score-based likelihood ratios are one means to quantify the weight of forensic pattern evidence (e.g., Helper et al. 2012; Reinders et al. 2020). They are not true likelihood ratios, meeting the rigorous statistical properties of a Bayes factor (Neumann 2020). However, it is appreciated that the use of scores may be the only viable method to reduce the complexity of some types of forensic evidence (Neumann 2020). “Useful,” explicitly, does not mean to imply suitability for legal evidentiary purposes, but these methods are an important part of the process of working toward the goal of practical statistical approaches to pattern evidence interpretations.

For fingermark evidence examinations, there are two immediate, important applications of probabilistic assessments (both discussed in Abraham et al. 2013). These are (1) as a means to screen crime scene evidence collections for highly selective (“identifiable”) fingermarks (such as in Neumann et al.’s field study, 2011) and (2) as a measure of support for expert conclusions relating to comparisons of very highly selective (“identifiable”) fingermarks (Langenburg, Champod, and Genessay 2012; Langenburg et al. 2015; Swofford et al. 2018; Neumann 2020; Neumann and Ausdemore 2020). A third application (also discussed in Abraham et al. 2013), representing a radical change in the way fingermark evidence has been traditionally reported, is using probabilistic methods to assign a weight of evidential support for associations using fingermarks that would otherwise be declared as inconclusive. This third, radical, application was pursued by Stoney et al. (2020) to measure how often such fingermarks occur in casework and their range of associative value. In Stoney and Stoney (2021), these results were reviewed by police investigators, prosecutors, judges, defense attorneys, and forensic scientists to assess the potential utility of these fingermarks and the feasibility of their use.

3.2 Frequent occurrence of NIFMs in casework

In casework, NIFMs with lower selectivity than traditionally identifiable fingerprints, but still with high potential associative value, have been found to occur commonly and in abundance. When fingerprints of value for identification occur, there are almost always fingermarks of lower selectivity, and in greater abundance when collected as part of existing practices. It is likely that these fingermarks also occur in cases where fingerprints are not collected for one reason or another (Stoney et al. 2020).

3.3 Accepting the radical paradigm: NIFM associations are not the same as fingerprint associations

There are important, fundamental distinctions between associations that follow from NIFMs and those resulting from fingerprint comparisons. This is true regardless of whether the fingerprint comparisons employ long-standing practices resulting in subjective categorical identifications, current practices resulting in more moderated opinions, or future practices informed by the results of statistical models of fingerprint analysis. Fingerprint comparison evidence has as its explicit goal an extremely high threshold of associative value, with nearly negligible rates of false association when qualified fingerprint examiners apply appropriate examination procedures monitored by rigorous quality assurance procedures.

NIFM associations are explicitly not presumed to provide strong support for associations. A better appreciation of their character results when one rejects the notion that NIFMs are a variation of fingerprint identifications (even though both approaches use friction ridge detail as the basis for their associations). Rather, NIFMs are a type of circumstantial evidence with a range of associative value, as are most types of evidence.

Consider a parallel to circumstantial eyewitness testimony describing a person’s appearance. There is a potential contribution, but with an acceptance of a range of credibility that must be carefully weighed within the context of the broader set of facts under consideration.² Alternatively, consider a parallel to DNA evidence that is based on a small number of loci, limiting the selectivity of the DNA profile. NIFMs are within the broad class of circumstantial evidence that has the potential to contribute to an accumulation of evidence, thereby contributing to an investigation or resulting in a convincing case. This perspective avoids a focus that the level of association provided by NIFMs is deficient (based on a comparison to traditional fingerprint identifications) and allows for the evaluation of potential contributions.

3.4 NIFMs are an exceptional form of circumstantial evidence

Once viewed from within this perspective, NIFM associations are found to be an exceptional class of circumstantial evidence: they are already routinely and inexpensively collected, they can be objectively evaluated using statistical fingerprint models, they occur very frequently in cases, and they occur multiple times in each case. They are comparable to having many eyewitnesses in one case, testifying as to what was touched and by whom. However, unlike eyewitnesses, each has a statistically measurable credibility.

4. Fingermarks of lower selectivity as material evidence

Evidence is material if it relates to an element necessary for proving or disproving a case (Wex 2022b). It is well known that highly selective “identifiable” fingerprints can provide material evidence. They can be offered as part of evidentiary proof, in the form of extremely strong support, for the identity of a person touching a surface. This can extend to proof of presence at a particular location when fingerprints are, for example, on an unmovable surface or on an item otherwise proven to be at the location. This can also extend to support for involvement in a particular activity, or, when combined with additional evidence, to establish presence at the time of a criminal action. Fingerprints can also improve or diminish the credibility of other information, such as statements made by victims, witnesses, or suspects.

Fingermarks of lower selectivity can also provide material evidence in each of these ways. By their nature these fingermarks comprise considerably smaller areas of high quality friction ridge detail. This means that they can be found on surfaces presenting smaller areas to the touch, or when only a small portion of the finger does the touching (as on firearm triggers or touchscreens), or when motion during contact causes smearing of most of the mark. The result is that when fingerprint probability models are applied to utilize less selective fingermarks, the number of cases where fingermarks can provide materially useful information increases dramatically. This does not occur when only identifiable fingerprints are considered, and when multiple occurrences in one case are discounted (Neumann et al. 2011).

When NIFMs are included, each case has greater numbers of fingermarks providing information and an extended range of locations where they can be found. Multiple fingermarks associated with one individual can provide cumulative weight through their co-occurrence, and the finding of a single associated mark among many others can be recognized as of doubtful significance. Furthermore, under an assumption that an individual is present at a location (supported by the fingermarks or other available information), the larger set of fingermarks provides an excellent opportunity to provide information about what an individual was actually doing. This is known as activity value (Cook et al. 1998a, 1998b). The smaller sizes of fingermarks within the set means that they can be found on smaller surfaces, including areas that are clearly material, but where experience has shown that identifiable fingerprints are very rarely found, for example, on keys, automobile door handles, or ammunition. Finally, the greater abundance of marks increases the likelihood that the time element can be established (by one or more of the marks occurring under circumstances that narrow the possible time when the marks could have been made).

Stoney and Stoney (2021) provide specific examples of the utility and anticipated impact of using fingermarks of lower selectivity at different stages of the criminal justice process. Overall, using the full set fingermarks results in more complete evaluations of activities at crime scenes. Additional contributions include enhancing or diminishing the significance of other physical or circumstantial evidence and helping to evaluate the credibility of statements or testimony. New routes of investigation are enabled, including identifying possible linkages among cases, developing suspects through the creative use of fingerprint databases, and efficiently evaluating suspects once they have been developed. As a new source of information, less selective fingermarks are a fresh means to review uncharged cases and contribute to cold case investigations. As a new and previously unavailable source of information, they can also be used as an argument for post-conviction relief (ABA 1978).

5. Issues of implementation and risk

Given the substantial potential benefits from probabilistic fingermark associations with lower selectivity, what issues need to be addressed to bring this type of associative evidence to its full potential? Do the benefits outweigh anticipated costs and risks? These are important questions that require both additional research and careful consideration of results from many different perspectives. Some of the issues that merit examination are introduced in this section with the intent of providing our perspective and encouraging the broader discussion that is essential.

5.1 Revised crime scene collection and fingermark retention practices

Taking advantage of the potential of less selective fingermark associations requires modification of crime scene collection practices to ensure that the smaller, more abundant fingermarks are recovered from surfaces where useful information could result. This should include the collection of fingermarks from any material locations that might previously have been omitted because of the unlikely occurrence of highly selective (“identifiable”) fingerprints.

In anticipation of future analyses and use of less selective fingermark associations, it is also important that fingermarks collected as evidence be retained, even if they do not show marks that are considered useful or comparable by current practices. The importance of evidence retention for cold case and post-conviction investigations has been made very clear by the experience with changes in biological fluid analyses (TWGBEP 2015) and the situation is no different for less selective fingermarks.

Although no new technical methods are needed, changes in policies and procedures would need to be endorsed by those administrating crime scene collections and evidence retention. As with any procedural change, the impetus will need to come from recognition of possible benefits or the demonstration of actual benefits.

5.2 Methods development

In parallel with the underlying assumptions of this Special Issue, we accept that statistical models allowing the measurement of fingermark selectivity will not be long in coming. When properly vetted, these will be established on both theoretical and empirical grounds, using ground truth data sets. Robustness will be determined, with clearly defined limitations.

Measurement methods for less selective fingermarks will require three distinct applications (Stoney and Stoney 2021). The first is to measure the expected associative value of a fingermark, standing alone, in the absence of a putative source. This application would be used in casework to recognize which fingermarks could result in useful associative (or exclusionary) information in a given case. The second application is the measurement of the associative value of an actual comparison between a fingermark and a corresponding reference fingerprint from a known individual. The third application is the measurement of exclusionary value: support that the fingermark was not made by a particular individual.

5.3 Close integration with investigative units and associated potential for misuse

As discussed earlier in this article, lower selectivity associations have the potential to make extraordinary contributions to criminal investigations. Many investigators feel that this contribution would be most helpful if the results of fingermark comparisons were obtainable quickly, allowing evaluations of the significance of the associations to be made within the context of the investigation (Stoney and Stoney 2021).

This is not a trivial requirement and necessarily confronts one of the biggest current areas of debate regarding the delivery of forensic science services. On the one hand is a view, founded on legitimate concerns regarding historical abuses and bias, that argues for more separation between crime laboratories and investigative units (NRC 2009; Giannelli 2010; Thompson 2015; Beety 2016). On the other hand is a view that there should be more integration, founded on legitimate needs for efficient and effective delivery of information (Ribaux and Talbot-Wright 2014; Baechler et al. 2020; Garrett 2020; Roux et al. 2022). Related issues have been discussed at greater length by Cole (2013). The potential benefits and drawbacks of close integration of fingermarks of lower selectivity with investigative units needs to be determined in the context of this broader discussion.

With respect to fingermarks of lower selectivity, it may appear a high risk to introduce a new method, producing information of low selectivity, closely integrated with criminal investigations and incorporating, rather than avoiding, contextual information. A view may be held that the potential for misuse during investigations, arrests, and charging is very high. Several aspects of fingermark associations mitigate against this view. As conceived (and as assumed under the premise of this Special Issue), the associations result from the application of a vetted statistical fingerprint model on specific, well-defined details within fingermarks. As such, the process is transparent and reviewable. The calculations can be repeated, and weaknesses can be explored. These elements, by their nature, counteract risks of misuse and, in fact, provide discrete information of measured value from a well-defined source.

5.4 Costs and resource allocation

Laboratory examination and reporting of findings from many more fingermarks will necessarily require allocation of resources. Responsible case management is essential for balancing the additional effort with the utility of the information in the context of individual cases. However, this is no different from the application of any forensic science capability. The analysis of currently unused fingermarks of lower selectivity has the strong potential to provide significant additional, useful, information. The capability can be applied with discretion, when the information is likely to help solve or prove a given case. Much of the work is likely to be automated, requiring minimal commitment of expert resources, and the benefits are expected to be high relative to the additional costs.

5.5 Education

Educational requirements will be particularly important for fingermark associations of lower selectivity. Although the basic concept is straightforward, the public and all criminal justice participants, have pre-existing experience with fingermarks that involves their appearance as evidence only when experts make decisions regarding either identifications or exclusions. Retraining of judges, attorneys, and juries will be necessary to address this prejudice and ensure that less selective fingermarks are evaluated properly as circumstantial evidence. Necessarily included as part of this task is the difficult, more general forensic science issue of how probabilistic measurements of the weight of associative evidence should be presented in court (which forms a major portion of the discussions in the accompanying papers in this Special Issue).

5.6 Acceptance by practitioners

The application of statistical models for fingerprint analysis, as an improvement to existing practices, will necessarily be accepted by practitioners and is expected to progress naturally and steadily—guided by a substantial interest and pressing evolution as discussed broadly in this Special Issue. Extending the application of these models to enable the use fingermarks of lower selectivity is more conceptually challenging, but upon reflection, it can hardly be avoided. Once we change the paradigm—allowing quantitative measurements of associative value for fingermarks—there is no defined barrier. The continuum of associative value for fingermarks will necessarily follow—from highly selective, to moderately selective, to weakly or mildly selective. We expect changes to come about through the application in individual cases, where the information is most needed, and the priorities are highest. Examples are major investigations, cold cases, and application to post-conviction and post-exoneration evidentiary reviews. As experience is gained in the actual use of these fingermarks, their observed value and impact will guide appropriate levels of acceptance and the development of policies for their use.

6. Discussion of associations with low selectivity

A more general view can be taken that associations of low selectivity can, by their very nature, be more readily abused, that they are more frequently in error, and that when in error they constitute misleading evidence. These perceptions can be used to argue that associations with low selectivity (such as the fingermarks considered here) are a problematic type of evidence that is too risky to use.

Information of low selectivity is a commonly encountered and important component in criminal investigations and prosecutions. When this information addresses a material aspect of the case it provides circumstantial evidence, contributing to the overall accumulation of evidence that can lead to a convincing case. There is nothing wrong or inappropriate regarding this information, if its reliability is established and its probative value is not substantially outweighed by a confusing or misleading nature, or a prejudicial effect. Each of these is important.

In this discussion, we consider four distinct concepts relating to error

• An Erroneous Test Result

• A Valid, but Confusing Test Result

• A Valid, but Inaccurately Presented Test Result

• A Valid Test Result that Points the Wrong Way

These are by no means the only possible sources of error, and our treatment does not extend to all of the ways that evidence could be misused, either unintentionally or intentionally. Our intent is to make distinctions among these four sources and offer our observations on these, as part of the discussion, from a forensic scientist’s perspective.

6.1 An erroneous test result

The most primary source of error is an Erroneous Test Result: a mistake was made when test was performed, and the method gave the wrong result. For example, a test result could show that a fingermark, found to correspond to a particular individual, had an estimated chance of occurrence in the random population as 1 in 50,000, whereas the correct value was 1 in 500. Such errors would indicate an unreliable method. The cause of such an error might be attributable to a defective procedure, or to biased or incompetent observation, or some other cause.

Forensic science needs to provide appropriate and vetted statistical fingerprint models, with known limitations, applied with appropriate controls, to ensure that the associative value assigned to fingermark comparisons is accurate. Forensic science also needs to address related issues of communication and education so that the meaning of the associative value is accurately understood.

The probative value of information from a fingermark comes from a combination of its selectivity (the associative value shown by vetted probability models) and the circumstances of the case (where the fingermark was found and what elements of the criminal event it addresses). Judges and juries are assigned the responsibility for assessing probative value. Forensic scientists contribute the reliable measurement (with known limitations) and may be able to provide additional information (such as how an item was touched or how to best present numerical evidence) but is not the role of forensic scientists to make an assessment of probative value (as, e.g., in the presentation of opinions as fact) or to withhold information based on their own perceptions of its value (as, e.g., by ignoring fingermark associations of low selectivity).

6.2 A valid, but confusing test result

A second source of error is a Valid, but Confusing Test Result: a test was performed, and the (valid) test result was confusing, resulting in an inappropriate weight being assigned by the trier of fact. For example, if a fingermark was found to correspond to a particular individual, with an estimated chance of occurrence in the random population as 1 in 500, the jury may have confused the finding to mean that the chances were 500 to 1 that the fingermark was made by the particular individual. The possibility of this type of error is a well-recognized, important consideration whenever quantitative associative evidence is interpreted. It is already the subject of extensive discussion, in this Special Issue and elsewhere. Quantitative associative evidence arising from fingermarks of lower selectivity are directly subject to these concerns.

6.3 A valid, but inaccurately presented test result

A third source of error is a Valid, but Inaccurately Presented Test Result: A test was performed, and the (valid) test result was presented in a misleading way and misinterpreted by the trier of fact. For example (using the same results as in the preceding paragraph), the jury may have been told by the prosecutor that it actually meant the chances were 500 to 1 that the fingermark was made by the particular individual.

Evidence that is confusing or misleading is a serious problem and has been found to be a major source of unsafe court rulings (Smit, Morgan, and Lagnado 2018). Forensic science must bring to judges’ attention the related and extremely important issues of communication and education, so that these risks can be meaningfully assessed. However, apart from the communication of probabilistic information itself, we have found that judges consider it unlikely that issues associated with fingermarks of lower selectivity will be seen as particularly challenging to the courts (Stoney and Stoney 2021).

6.4 A valid test result that points the wrong way

Our fourth conceptualization relating to error is a Valid Test Result that Points the Wrong Way: a test was performed, and the (valid) test result supports an association that is not, in fact, true. For example, if a fingermark was found to correspond to a particular individual, with an estimated chance of occurrence in the random population as 1 in 500, this information may have been used as evidence to support a case against a defendant, when in fact the fingermark actually came from someone else. As unfortunate (and unlikely) as this use may be, it is not an error in itself. In fact, under elementary assumptions it would be expected to occur once in about every 500 times a fingermark of that associative value were to be used. (If it didn’t occur with an average frequency close to 1 in 500, that would indicate an error in the frequency estimate.)

6.5 Summary discussion of associations with low selectivity

We share the view that the use of lower selectivity associations as evidence, or other presentations of probabilistic measurements of evidential weight, could be easily misunderstood. We also share the view that there is the potential for misuse of non-definitive associations by law-enforcement or prosecutors during investigations, arrests, and charging. There is also the potential for distortion of these associations by defense attorneys during argument or cross-examination. Juries can and do reach conclusions that are in error. However, these are issues with our criminal justice process as a whole and are shared by many other types of evidence; they are not peculiar to fingermarks, nor are they solely within the responsibility or purview of forensic scientists. Forensic scientists have the obligation to provide the legal process with information and the education necessary to minimize the potential for such errors. However, they cannot take the view that because fingermark associations of lower selectivity share these potentials, the information they can provide is too “risky” and must remain undeveloped and withheld from the criminal justice process. It is indefensible to consciously seek less information regarding material facts in a case as a response to the possibilities that reliable information might be confusing or misused.

7. Conclusions

Quantitative methods for the assessment of associative value, once properly vetted, will lead to increased objectivity and scientific reliability for both highly selective fingermark associations (replacing “identifications”) and those resulting from fingermarks of lower selectivity. Although in this article we have stressed the need for separate consideration of these less selective marks, clearly there is no fundamental distinction. The selectivity of the friction ridge patterns within fingermarks necessarily lies on a continuum. What is distinct is how we have, until now, elected to use fingermark evidence only when the marks are very highly selective. For decades, under the subjective expert-driven system, this meant we required a threshold of expert belief in the ultimate of selectivity: categorical identity. Today, for the most part, we have recognized the limitations of that system, and in response we have revised how we conceptualize and report fingermark associations; and imminently, we will have reasonable statistical processes to measure this selectivity. Yet we will fall short as scientists if we retain the essence of our previous ignorance: the notion that the use of fingermarks for association is to be restricted to those marks that are very highly selective. In practice, this means we ignore and discard most of the readily available fingermarks, along with a demonstrably rich source of information that as evidence can be material and highly probative, with measurable associative value. We can choose (at least initially) to ignore this source of information, but it will come, and it would be much better to carefully plan for it, helping to ensure a controlled and responsible implementation.

Footnotes

Acknowledgements

We acknowledge and greatly appreciate those who have engaged with us in the discussions of the utility of fingermarks of lower selectivity over the past several years. By their participation in these discussions, they make no representations or endorsements of our views. Rather, they have shown their willingness to engage in frank and constructive discussions of their views, ranging from excitement for the prospect of exploiting an obvious and long-neglected form of evidence, to assessments of scientific challenges regarding the measurement processes, to challenges relating to practical matters of education and implementation, to concerns over the possibilities of misuse within criminal justice processes.

Conflict of interest statement. None declared.

Funding

This work was supported by Awards Nos. 2016-R2-CX-0060 and 15PNIJ-21-GG-04192-RESS awarded by the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and conclusions or recommendations expressed in this presentation are those of the authors and do not necessarily reflect those of the Department of Justice.

References