Only a Sith Deals in Absolutes: A Jedi Uses Likelihood Ratios

Nature of absolutes in forensic statements

For decades, many forensic disciplines have relied on definitive statements when communicating their conclusions to the courts. This practice was especially pronounced in older forensic handwriting examination testimonies, where experts might have said, ‘The suspect wrote this document’ or ‘This signature is undeniably forged.’ While such declarative statements are easily understood by judges, jurors, attorneys, and laypeople, they may overlook the complexities that experts are aware of, but which are often omitted for the sake of simplicity in courtroom communication.

Such an absolutist language could imply a finality that does not leave room for further scrutiny and gives more power to the evidence than it should. In a handwriting comparison, for instance, concluding with unequivocal certainty that a questioned note was penned by a specific individual can inadvertently hide the subtle variations and statistical probabilities inherent in handwriting features. Even among well-trained examiners, the possibility of misinterpretation or an unusual coincidence can usually not be dismissed and occurs more often than we think (6). In fact, the only way to be able to speak with a definitive voice would be if the expert witnessed the crime firsthand—a rather odd scenario. The decision of the facts is not a job for the experts, but for the judge or jury only (7).

Adversarial nature of court proceedings

Definitive statements feed into the adversarial nature of courtroom proceedings, often framing the debate as a zero-sum game. Each side, prosecution or defense, seeks a portrayal of the forensic evidence that aligns best with their narrative. Experts, therefore, may feel pressured to testify in more absolute terms than the data truly supports, to align their testimony with the expectations of the party that retained them.

This practice is especially challenging in forensic handwriting examination, where overlapping handwriting characteristics, disguised writing, or advanced forgeries can complicate the examiner’s determination. Nevertheless, under a black-and-white approach, even a trace of complexity can be perceived as weakening the expert’s conclusion, prompting some to overstate the certainty of their findings to avoid appearing unconvincing in court. To address these limitations, the forensic community has increasingly turned to the LR framework, which offers a balanced and transparent approach to evidence interpretation.

Gradation of evidence strength

Likelihood Ratios embrace the notion that evidence can be more or less likely under competing hypotheses, rather than these hypotheses being simply true or false (8, 9). During a forensic examination, one can consider two primary hypotheses:

1. H₁: The suspect authored the questioned document or signature.

2. H₂: The suspect did not author the questioned document or signature.

By evaluating handwriting or other document features, an examiner can calculate (or, rather, estimate) how much more likely it is to observe those specific features if H₁ is true versus if H₂ is true. The resulting LR may be greater than 1, indicating that the observations are more likely under H₁ than under H₂ (i.e., relative support for H₁). An LR close to 1 indicates that the features are about equally likely under both propositions (i.e., no relative support for either), and an LR smaller than 1 suggests that the features are more likely under H₂ than H₁ (i.e., relative support for H₂).

For example, consider a suspicious signature that exhibits unusual pen lifts and incongruent letter slants compared to the suspect’s known signatures. Under H₁ (i.e., the suspect wrote it), the probability of observing discrepancies might be low in general, but cannot be dismissed entirely. Genuine writers can produce irregular or unexpected features due to a range of factors such as stress, particular writing conditions, illness, intoxication, etc. These naturally occurring variations contribute to the uncertainty under H₁. Under H₂ (i.e., the suspect did not write it), the likelihood of observing these irregularities is generally higher, particularly if a forger with a distinct writing style attempted to mimic the suspect’s signature. The interpretation of such discrepancies depends critically on contextual information, including the writing conditions, known variability in the suspect’s writing, and any external factors that may have influenced the writing process. In most cases, an LR in favor of H₂ would then emerge from such an examination, illustrating the strength of evidence that the suspect was not the author.

Integration of contradictory evidence

The LR framework is valuable precisely because it can accommodate contradictory pieces of evidence and still maintain a coherent overall assessment (10). Suppose that, in the same case, additional handwriting traits, such as letter connections and spacing, bear a strong resemblance to the suspect’s usual writing style. While some features weigh against the suspect, others weigh in favor. Rather than forcing a single definitive conclusion, the LR approach allows each feature to contribute to the evidential strength in a logical way (11).

In scenarios where multiple questioned handwritings are submitted, some might more strongly align with the suspect’s known writing habits, while others might appear more suspicious or expertly forged. The expert’s final opinion, expressed as a series of LRs or a composite LR, conveys a multi-layered view of the evidence. This holistic approach reduces the risk of key contradictions being lost in the pursuit of a singular, absolute statement (11, 12). In a court system where you need to have proof beyond reasonable doubt, it is not the expert’s single numerical estimate that should carry all the weight, but rather the combination of all evidence, including prior information and the circumstances of the crime (13). While the expert’s role is limited to evaluating the strength of the written evidence, the court may ultimately regard even weak but consistent support for one proposition as significant when viewed in the broader context. Ultimately, the weight and integration of all evidence, including handwriting examination, rest with the trier of fact. The expert’s role is to inform the process, not to decide the outcome (14).

Signature authenticity analysis

Imagine a series of checks allegedly signed by a company treasurer. The signatures on some checks match the known references closely, while on others, subtle distortions in letter formation are observed. Traditionally, a forensic handwriting examiner might be compelled to state: ‘All these checks are from the treasurer’ or ‘Some of these signatures are definitively forged.’ Under the LR model, however, an examiner could specify, for each check, the likelihood that the treasurer signed it versus the likelihood that someone else signed it while imitating the treasurer’s signature. Such a nuanced statement might read:

This range conveys a realistic picture of each signature’s authenticity, helping the judge and jury appreciate the complexity involved. Since the examiner may have no knowledge of the suspect’s neurological state, factors such as physical exertion, illness, or degenerative conditions could lead to atypical writing features. The same could be true for any degenerative disease. Rather than drawing a rigid conclusion, the LR framework allows the examiner to express varying degrees of support for competing hypotheses, based on the available evidence. This approach also avoids the risk of having to reverse a conclusion later, should new contextual information emerge. In this way, it promotes transparency and prepares the opinion to withstand critical scrutiny during cross-examination.

Questioned letters and handwriting comparison

Consider a situation where an anonymous threatening letter is mailed to a person of interest. The police collect handwritten samples from a suspect, and an examiner compares the morphological and stylistic features of the suspect’s writing to the threatening letter. Some letters or word formations strongly resemble the suspect’s known style, while others show significant divergences. Under rigid, definitive assumptions, an examiner might feel compelled to say either ‘Yes, it’s the suspect’ or ‘No, it’s not the suspect.’

However, employing the LR concept allows the examiner to quantify how each observed handwriting feature shifts the balance of probability. For instance:

• The shape of certain letter loops is quite distinctive and aligns well with the suspect’s normal writing; this elevates the LR in favor of the suspect having written the letter.

• Inconsistencies in pen pressure patterns might be less likely if the suspect was the author, slightly reducing the LR.

• The overall style of spacing between words might moderately align with the suspect’s usual spacing, further adjusting the LR upward.

By aggregating each of these features in a structured manner (often with the aid of statistical models or established empirical data), the examiner arrives at an LR that encapsulates the totality of the evidence. That final LR is then presented, often accompanied by a verbal scale to help non-experts in interpreting its magnitude.

Encouraging dialogue among forensic examiners

One of the most compelling advantages of the LR framework is its ability to foster transparent and structured discussions among experts. In forensic handwriting examination, where differing conclusions can arise, it is rarely due to disagreement over which features are present. Most competent examiners typically observe the same set of features, such as stroke direction, pen lifts, or letter formations. Rather, differences usually stem from how those features are evaluated, the propositions under consideration, the framework used, or the specific evidence available (15). By clearly defining these elements, examiners can more precisely identify the source of disagreement. Although discussions may involve individual features, the overall evaluation often relies on the gestalt: the holistic impression formed by the interplay of multiple features. This clarity not only supports constructive expert dialogue but also aligns well with practices like expert conferencing (or ‘hot-tubbing’), where examiners collaboratively discuss their reasoning in court (16, 17, 18).

Such an approach is more difficult under an absolutist paradigm, where disagreement may devolve into a direct clash of yes-or-no statements. With the LR framework, disagreements become relative and can often be explored through differences in interpretation rather than outright contradiction. Even when LRs are expressed qualitatively (such as describing a feature as offering limited, moderate, or strong evidence for H₁ over H₂), the approach still promotes transparency and reasoned dialogue (19). If one expert interprets a certain handwriting feature as offering strong evidence in favor of H₁ over H₂ (e.g., authorship versus not) while another views it as no evidence in favor of either hypothesis, they can revisit the case data or supporting literature to clarify the basis for their views. This process improves the reliability and credibility of the final opinion, regardless of whether the expression is numerical or descriptive.

An example of Canadian legislative provisions encouraging expert collaboration:

These provisions illustrate how Canadian law encourages a collaborative atmosphere, allowing experts to narrow issues before trial and promoting a more nuanced, LR-aligned approach. Comparable frameworks exist elsewhere: in England and Wales, concurrent expert evidence was introduced in the Civil Procedure Rules, Practice Direction 35, paragraph 11 (2013) (21); Australia formalized ‘hot-tubbing’ through the Federal Court Rules 2011 (Cth), r 23.15 (22); and New Zealand implemented expert conferencing and joint statements via the High Court Rules 2016 (rr. 9.43–9.51) (23). Despite their differences, these measures share the common goal of reducing duplication, clarifying expert reasoning, and focusing courtroom discussion on evidence rather than on discrediting experts simply because their conclusions differ.

Helping the legal community understand nuance

In many legal settings, judges and juries may initially find probabilistic statements intimidating or overly technical (24). However, LRs can be conveyed in straightforward language by relating them to everyday concepts of likelihood (25). For example, if you hear a loud siren outside, that observation is more likely if an ambulance is passing than if no emergency vehicle is nearby; the observation supports one explanation over the other without making it certain. In forensic handwriting examination, a lay audience may likewise better understand a statement such as: ‘The observed set of handwriting features is 10 times more likely if the suspect wrote the letter than if the suspect did not,’ as it expresses support for one proposition over another without making absolute claims. While traditional reporting scales in forensic handwriting examination often aimed to reflect posterior belief about authorship, the LR framework shifts the focus to the strength of the evidence itself under each hypothesis, clarifying the uncertainty and providing a more logically grounded expression of support. Despite this shift, both approaches share many structural similarities, particularly in how direction and weight are applied to the observed features (2).

Furthermore, the LR approach helps clarify that a forensic expert is not claiming certainty or omniscient knowledge; rather, they are expressing how the observed evidence supports one proposition over another based on their evaluation (1, 26). While the method can be framed mathematically, its foundation lies in the expert’s training, experience, and skill. The mathematics simply provides a structured way to communicate the strength of support but is usually still presented qualitatively or by using a verbal scale. This transparency allows for both qualitative and quantitative expression, depending on the context and the available data (19, 27, 28, 29).