Expert Witness Testimony and Cognitive Bias Mitigation

Before 1993, US federal courts and most state courts applied the Frye standard, derived from a 1923 DC Circuit decision in Frye v. United States, which asked whether the methodology underlying an expert opinion had "general acceptance" in the relevant scientific community. For document examination, this meant that testimony from ASQDE or ABFDE members, drawing on SWGDOC-standard methodology, was generally admitted because those bodies represented the professional consensus.

The Supreme Court's 1993 decision in Daubert v. Merrell Dow Pharmaceuticals displaced the Frye standard in federal courts. Daubert held that the Federal Rules of Evidence, specifically Rule 702, required the trial judge to serve as a gatekeeper who assessed whether the expert's methodology was scientifically valid and reliably applied. The Court identified four non-exclusive factors for this assessment: whether the theory or technique had been tested; whether it had been subjected to peer review and publication; whether it had a known or potential error rate; and whether it had been generally accepted in the relevant scientific community.

The Daubert trilogy (completed by General Electric v. Joiner in 1997 and Kumho Tire v. Carmichael in 1999) extended this gatekeeping function to all expert testimony, not just scientific experts, and confirmed that judges must assess the methodology itself, not just the conclusions. Rule 702 was subsequently amended in 2000 and most recently in 2023 to codify these requirements, including an explicit requirement that the expert's opinion "reflects a reliable application of the principles and methods to the facts of the case."

For document examiners, the post-Daubert period produced a wave of challenges, particularly in federal courts following Kumho Tire. Several district courts limited handwriting testimony to "probability" and "strong probability" conclusions, declining to permit "identification" on the grounds that the error rate for that level of conclusion was not established. The PCAST 2016 report, discussed separately in the conclusion-scales topic, reinforced this concern. US document examiners now routinely expect to defend: their training and experience; the SWGDOC standards their methodology follows; the literature on examiner accuracy (Srihari 2002, the Galbraith 2018 studies); the basis for their specific conclusion; and any limitations in the material examined.

About half of US states retain the Frye standard. In those jurisdictions, general acceptance remains the primary admissibility criterion, and the focus is on whether the examiner's methodology aligns with SWGDOC and ABFDE standards rather than on the specific error rate.

The United Kingdom's approach to expert-witness testimony in criminal proceedings is governed principally by the Criminal Procedure Rules Part 19 (CrimPR Part 19), which applies to proceedings in the Crown Court and magistrates' courts in England and Wales. Part 19 requires that an expert's report state the substance of all material instructions (written or oral), identify any range of opinion on the matter and give reasons for the expert's own position within that range, and include a statement that the expert understands their duty to the court and has complied with it.

The duty-to-the-court obligation is fundamental to the UK frame. The expert is not an advocate for the party who instructs and pays them. Where the evidence does not support the instructing party's position, the expert is obligated to say so, and to say so clearly. Document examiners who have written reports that overstated the strength of their conclusions, omitted contradictory observations, or failed to address alternative interpretations have faced professional sanctions from the Forensic Science Regulator and, in some cases, criminal referrals.

The Cairns checklist, named after the expert-witness guidance developed within the UK forensic-science community and subsequently formalised through professional guidelines, is a practical tool for report self-review before submission. The checklist asks whether the report states: the qualifications and experience of the expert; the instructions received; the materials examined; the methodology applied; any limitations on the examination; the findings; the opinion in appropriately hedged language; any range of opinion that exists on the issue; and whether the expert has complied with their duty to the court. A document examiner who works through the Cairns checklist before submitting a report has substantially reduced the risk of a successful challenge on foundational grounds.

The UK Forensic Science Regulator's Codes of Practice (2021, updated through 2023, now having statutory authority under the Forensic Science Regulator Act 2021) overlay these procedural requirements with substantive accreditation criteria. Forensic science providers in England and Wales who work on criminal cases are required to be accredited against ISO/IEC 17025, with UKAS as the accrediting body, and the Regulator has power to review and investigate providers who fall below the standard. The accreditation criteria include method validation requirements and proficiency testing obligations that directly bear on the reliability of the expert's opinion.

The Indian Evidence Act 1872 (IEA) section 45 provided the foundational legal basis for expert testimony in Indian courts for over 150 years. Section 45 allowed courts to consider "the opinions of persons specially skilled" in particular areas as relevant facts, covering science, art, handwriting, and foreign law. For forensic document examiners in India, this provision was the gateway through which CFSL and state FSL reports entered the evidentiary record.

The Bharatiya Sakshya Adhiniyam 2023 (BSA 2023), which came into force on 1 July 2024, replaces the Indian Evidence Act and carries forward the expert-witness provision in a revised form. Section 39 of the BSA 2023 addresses opinion evidence, retaining the essential structure of IEA s.45 while adding provisions that align more closely with modern forensic practice. The BSA 2023 does not specify a conclusion scale format for forensic reports, leaving the specific articulation of expert opinions to institutional SOPs and professional standards.

Indian courts, including the Supreme Court and various High Courts, have developed a body of case law on the weight to be given to document examiner opinions. In Fakhruddin v. State of Madhya Pradesh (1967), the Supreme Court held that handwriting expert opinions should be treated with caution and that judges are entitled to reach their own conclusions by comparing documents without relying solely on expert testimony. This caution reflects a judicial awareness, predating the Daubert era, that pattern-recognition expert testimony is not infallible. Subsequent cases including State of Maharashtra v. Sukhdeo Singh (1992) and Shashi Kumar Banerjee v. Subodh Kumar Banerjee (1964) elaborated on the conditions under which expert opinion would be given greater or lesser weight.

The practical consequence for document examiners testifying in Indian courts is that their conclusions are treated as potentially important but not determinative evidence, subject to independent judicial assessment of the underlying documents. This places a premium on reports that explain the reasoning transparently rather than asserting a conclusion baldly. An examiner who can walk the court through the specific features examined and why they support the conclusion is more likely to have their opinion accepted than one who states a conclusion without accessible explanation.

Itiel Dror and colleagues published a landmark study in Forensic Science International in 2006 showing that experienced latent fingerprint examiners, when presented with contextual information that implied a suspect was guilty, reached different conclusions about the same fingerprint comparison than they had previously reached without that contextual information. The prints compared in the 2006 study were the same prints those examiners had previously examined as part of real casework and had reached definitive conclusions about. With biasing contextual information, a significant proportion of those previous conclusions changed.

The 2006 study was followed by a series of further studies by Dror and others that extended the finding across different forensic disciplines, different types of contextual information, and different examiner populations. A 2010 study with radiologists; a 2011 study applying the framework to DNA mixture interpretation; studies on forensic anthropology; and, relevant to questioned document examination, work by Dror and Cole (2010) and Dror and Bucht (2011) that applied the contextual-bias framework explicitly to forensic pattern-recognition disciplines, including document examination.

The mechanism is not individual weakness or dishonesty. It is the way human pattern recognition works under conditions of ambiguity. When an examiner faces a comparison that is genuinely difficult (limited exemplars, degraded writing, unusual variation), the pattern-recognition system that resolves ambiguity draws on available information, including case context. If that context includes information that the suspect is guilty, has a criminal history, or is implicated by other evidence, the examiner's pattern-recognition system uses that information to resolve the ambiguous comparison in a direction consistent with the context. This happens without the examiner being aware of it.

For document examiners, the contextual information problem is acute. Examiners in operational forensic laboratories routinely receive case files that include investigative summaries, information about the suspect, and sometimes the conclusions of other forensic tests. The examiner tasked with comparing the questioned ransom note against the suspect's exemplars may know that the suspect has already been linked to the crime by other means. Under Dror's framework, this creates a bias risk that the examiner cannot neutralise by intending to be objective.

The primary structural response to contextual cognitive bias in forensic science has been the development of information sequencing protocols that control what the examiner knows at each stage of the examination.

Sequential unmasking, proposed by Dror and Hampikian (2011) and subsequently elaborated in a National Commission on Forensic Science white paper (2015), requires that the examiner receive case information in a controlled sequence: first the questioned material (the item to be examined), then the known standard (the comparison item), and only after completing the analysis and comparison do they receive contextual information about the case. The principle is that the analysis of the questioned writing should be completed and documented before any information about the suspect can influence the examiner's assessment of the comparison's outcome.

Linear ACE-V is a specific implementation of the analysis-comparison-evaluation-verification methodology that enforces one-way information flow through the examination stages. In the analysis phase, the examiner examines the questioned writing only, documenting observed features and their characteristics before any comparison material is introduced. The comparison phase then introduces the known standards. The evaluation phase assesses the significance of agreements and differences. Verification by a second independent examiner is conducted blind to the first examiner's conclusion. Linear ACE-V prevents the backward contamination that occurs when an examiner forms a preliminary conclusion early in the comparison and then unconsciously interprets subsequent observations to be consistent with that preliminary conclusion.

Blind verification requires that the verifying examiner, who independently examines the same materials, does not know the conclusion reached by the first examiner. Non-blind verification (in which the second examiner knows the first examiner's conclusion before reviewing the materials) is structurally incapable of detecting the bias that linear ACE-V is designed to prevent, because the second examiner has the same contextual information that may have shaped the first examiner's conclusion.

The UK Forensic Science Regulator's guidance on cognitive bias (2015 and updated 2020) describes information management protocols that align with sequential unmasking principles. The Regulator requires accredited forensic science providers to document their approach to bias management, including information management procedures and verification protocols. ISO 17025 does not mandate sequential unmasking specifically, but its requirements for method validation and personnel competence create conditions under which accredited laboratories must at least address the bias question.

Defence cross-examination of forensic document examiners follows recognisable patterns developed through decades of adversarial practice in US, UK, and Indian courts.

The first pattern attacks qualifications and training: Is the examiner a member of a professional organisation such as ASQDE or ABFDE? How many examinations have they conducted? Have they been tested through blind proficiency studies? Has their testimony previously been criticised by a court? For examiners at CFSL or state FSLs in India, questions about the accreditation status of their laboratory and the proficiency testing they have undergone are increasingly common in the upper courts.

The second pattern attacks the methodology: Is ACE-V a validated method? Was verification conducted blind? Did the examiner have access to information about the case that could have introduced bias? What literature supports the conclusion that the features identified are individually characteristic? This pattern directly targets the weaknesses identified by PCAST 2016 and Dror's bias research.

The third pattern attacks the conclusion itself: How certain is the examiner? What is the error rate for this type of conclusion? How many other writers might have produced the same writing? Has the examiner considered alternative hypotheses? This pattern exploits the gap between the certitude implied by verbal conclusion language and the actual probabilistic nature of the finding.

The fourth pattern attacks the report: Did the report omit observations that were inconsistent with the conclusion? Did the report address all significant differences, or only the agreements? Was the limitation section sufficient?

A report that explicitly addresses each of these areas before cross-examination leaves less attack surface and, as a practical consequence, produces the methodological transparency that makes the underlying examination more reliable.

1. Receive and log the instruction
   Document the instructing party, the question posed, the materials received (with reference numbers and condition), and the date received. Establish the chain of custody record for the examination materials.
2. Information sequencing decision
   Determine what contextual information is necessary for the examination and what can be withheld until after the analysis and comparison phases. Do not read the case file before completing the analysis of the questioned writing.
3. Analyse the questioned writing in isolation
   Examine and document the questioned document's features before introducing the known standards. Record: writing instrument, general writing style, letter forms, connecting strokes, spacing, slant, baseline, pen pressure, and any specific features of note.
4. Introduce and compare the known standards
   Review the provenance and adequacy of the exemplars. Examine the known writings using the same feature taxonomy. Compare against the questioned writing feature by feature. Document agreements and differences.
5. Evaluate and reach a conclusion
   Assess the significance of agreements and differences against the relevant population of writers. Apply the conclusion scale. State the conclusion and the basis for it. Address alternative hypotheses.
6. Blind verification
   Arrange for an independent examiner to review the same materials without being told the first examiner's conclusion. Document the verification outcome. Where conclusions differ, investigate and document the resolution process.
7. Write the report
   Structure the report to address: instruction; materials; methodology; findings (agreements and differences); conclusion with scale position named; qualifications and limitations; and the expert's duty statement as required by the relevant procedural rules.