ACE-V Methodology and the Brandon Mayfield Madrid Bombing Error

ACE-V stands for Analysis, Comparison, Evaluation, and Verification. The Scientific Working Group for Friction Ridge Analysis, Study and Technology (SWGFAST), which published the methodology's foundational documents before being succeeded by the OSAC Friction Ridge Subcommittee in 2015, defined each phase carefully.

Analysis is the first phase, and it is the one most often compressed or skipped under casework pressure. The examiner works from the latent print in isolation, before looking at any exemplar. The goal is to characterise everything the latent can tell you: quality, pressure distortion, the level of detail visible (Level 1 pattern type, Level 2 minutiae positions and types, Level 3 pore and edge-contour detail where resolution permits), and any artefacts that could be mistaken for ridge detail. The analyst records what they see before they have a suspect. In practice, many laboratories did not enforce this documentation requirement, and examiners moved directly from receiving the latent to comparing it against a named suspect's ten-print record.

Comparison is the second phase. The examiner places the latent and the exemplar in correspondence, aligning them to the same region, and systematically works through the points of similarity and difference. The comparison is supposed to be driven by what the analyst documented in Phase 1.

Evaluation is the third phase. The examiner reaches one of three conclusions: identification (the latent and the exemplar originated from the same source, to the practical exclusion of all others), exclusion (they did not originate from the same source), or inconclusive (the quality or quantity of corresponding detail is insufficient to reach either conclusion). Historically, the most contentious feature of this phase was the implicit claim attached to the word "identification": that the error rate was effectively zero.

Verification is the fourth phase. An independent examiner repeats the ACE process without knowing the conclusion reached by the original examiner and records their own conclusion before comparison with the first.

The latent print from the Madrid blue bag was of modest quality. It showed a fragmented loop pattern with a limited but workable number of minutiae visible at Level 2. The IAFIS candidate list returned several candidates; Mayfield appeared in the top twenty. The examiner who first reviewed the list was aware that Mayfield was a person of interest to the FBI's counterterrorism division before he began his comparison.

The FBI Office of Inspector General (OIG) report, published in November 2006, reconstructed the sequence in detail. The first examiner found what he judged to be fifteen points of correspondence. He ignored or explained away features that did not match. The print from the bag showed a ridge flow in one area that appeared inconsistent with Mayfield's ten-print record; the examiner attributed the discrepancy to pressure distortion on the latent. Two senior examiners conducted their own comparisons and reached the same identification conclusion.

The Spanish National Police had also run their own comparison and reached a different conclusion: the latent did not match Mayfield. They transmitted this finding to the FBI, who responded that they remained confident in their identification. The Spaniards ran additional comparison work and identified the actual source as Daoud, an Algerian national whose prints were on file from a prior investigation. Only then did the FBI withdraw its identification and release Mayfield.

The OIG report identified three primary causes. First, confirmation bias: each subsequent examiner knew that the first had reached an identification, which anchored their own analysis. The verification phase, which was supposed to be independent, was not operationally independent. Second, the examiners over-weighted similarities and under-weighted differences, explaining away discrepancies rather than treating them as evidence against the match. Third, contextual information about Mayfield's religion, his legal representation of a terrorism suspect, and his status as a counterterrorism subject of interest had contaminated the analysis before it began.

Mayfield sued the US government. The case was settled in 2006 for two million US dollars and a formal written apology from the FBI.

Twelve years before Mayfield, a Scottish police officer named Shirley McKie became the subject of a contested fingerprint identification that destroyed her career and her health. In 1997, four examiners at the Scottish Criminal Record Office (SCRO) concluded that a fingerprint found inside the murder scene of Marion Ross belonged to Shirley McKie, who denied ever having been in the house. McKie was charged with perjury for denying the print was hers at the murder trial of David Asbury. American fingerprint experts Fred Zain and Pat Wertman concluded the SCRO identification was wrong. The perjury charge was eventually dropped; the Scottish Executive paid McKie compensation of GBP 750,000 in 2006. The Crawford review, published in 2011, concluded the SCRO examiners had made a genuine error and that contextual bias had played a role.

In the United States, the National Academy of Sciences 2009 report "Strengthening Forensic Science in the United States" catalogued a range of documented fingerprint errors from US courts and concluded that the categorical claim of zero error rate attached to fingerprint identification was empirically unsupported. The Ulery et al. 2011 study, commissioned by the FBI in direct response to the NAS report, had 169 latent print examiners compare 744 latent-exemplar pairs. The false-positive rate (incorrectly calling an exclusion pair an identification) was 0.1%. The false-negative rate (incorrectly calling an identification pair an inconclusive or exclusion) was substantially higher. These were not zero.

In India, the CFSL and state FSL fingerprint divisions have operated under the provisions of the Indian Evidence Act section 45 (expert opinion) since 1872, and now under the Bharatiya Sakshya Adhiniyam 2023. Indian courts have periodically addressed contested fingerprint evidence, including in cases before the Madras High Court and the Supreme Court, where the quality and methodology of the comparison have been placed in issue. The Mahalakshmi 2018 Tamil Nadu case and Karnataka High Court precedents from 2019 both addressed the requirements for fingerprint evidence to satisfy BSA 2023 section 45's expert-opinion admissibility standard. These cases, like Mayfield and McKie in the US and UK contexts, all point to the same remedial need: structural protections built into the methodology before an opinion is formed, not after.

Sequential unmasking, articulated by cognitive psychologist Itiel Dror and adopted into laboratory protocols by the FBI and the UK Forensic Science Regulator by the mid-2010s, is a management protocol rather than a technical one. The principle is that information about a case should flow to the examiner only in the sequence in which it is needed, and only in the amount required for each phase.

In practice, this means the latent print arrives at the examiner's desk without any candidate name, without any contextual information about the crime, and without any pre-existing conclusion from another examiner. The examiner documents their analysis of the latent, recording the features they can see and the quality assessment. Only after this documentation is completed and signed off does the examiner receive the candidate's exemplar for comparison.

This sequence addresses the first and third causes identified in the Mayfield OIG report: it prevents contextual contamination of the analysis phase, and it means the examiner is committed to a feature inventory before they know who they are comparing against.

Linear ACE-V, the related protocol developed by the FBI Laboratory and codified in the FBI's Quality Assurance Standards for Latent Print Examination, enforces the four ACE-V phases as discrete, documented steps that must be completed in order. An examiner cannot proceed to Comparison without a completed and documented Analysis. An examiner cannot proceed to Evaluation without a completed Comparison. Each phase has a documented record that can be reviewed, audited, and placed in evidence.

Blind verification addresses the second cause from the Mayfield report: the verifier should not know the original examiner's conclusion before completing their own ACE phases. The UK Forensic Science Regulator's Codes of Practice (FSR-C-128, 2020 edition) require blind verification as the default for identification conclusions. The ENFSI Fingerprint Working Group Best Practice Manual (2015, revised 2022) recommends the same. The FBI QAS for Latent Print Examination requires blind verification for all identification conclusions and for inconclusive conclusions that are to be reported.

In the United States, the primary standard documents are the FBI Quality Assurance Standards for Latent Print Examination (2022 edition), the OSAC Friction Ridge Subcommittee's standard for examination methodology (OSAC 2021-S-0007), and the broader Scientific Area Committee reports on error rate and contextual bias published between 2016 and 2022. Together, they require: documented sequential analysis before candidate disclosure, linear phase progression with records at each phase, blind verification for identification and inconclusive-for-reporting conclusions, and explicit prohibition on accessing prior identification conclusions before completing independent analysis.

In the United Kingdom, the Forensic Science Regulator's Codes of Practice and Conduct (FSR-C-128 for fingerprint examination) are the operative standard. The FSR issued a specific Fingerprint Standard as part of the Codes, requiring sequential unmasking, blind verification, and the use of the evaluative reporting framework (which expresses conclusions as a likelihood ratio or graduated verbal scale rather than categorical identification). The UK courts, through the Criminal Procedure Rules Part 19 and the regime for expert witnesses, apply Daubert-analogous scrutiny through the gatekeeping role of the trial judge under R v. Turner (1975) and R v. Reed (2009).

In Europe more broadly, the ENFSI Fingerprint Working Group publishes the Best Practice Manual for Fingerprint Examination, which covers the ACE-V methodology, evaluative reporting, and the specific requirements for blind verification. ENFSI member laboratories operate under national accreditation bodies (DAkkS in Germany, RvA in the Netherlands, COFRAC in France, UKAS in the UK) that conduct surveillance audits against ISO/IEC 17025, which requires documented procedures for all forensic examination phases.

In India, the CFSL fingerprint methodology aligns with the Directorate of Forensic Science Services (DFSS) standard operating procedures and NABL T-126 accreditation criteria. The NCRB NAFIS technical documentation references ACE-V as the examination standard for all latent-print comparisons involving NAFIS candidates. The BSA 2023 framework, like the earlier IEA section 45, requires that fingerprint expert opinion be grounded in a demonstrable, reproducible method, which post-Mayfield reform protocols address directly.

Itiel Dror's 2006 study, published in the Journal of Forensic Sciences, was the first controlled experiment to demonstrate that experienced fingerprint examiners would change their identification decisions when given misleading contextual information. Dror took five experienced examiners and showed them the same prints they had themselves previously examined and decided in a real case, but embedded in a new context: one group was told the prints had led to a misidentification (the Mayfield narrative), another was shown them in a neutral context. Four of the five examiners changed their conclusions. They were not aware they were looking at the same prints.

Dror and Charlton extended this in a 2006 follow-on study, and subsequent work by Dror and colleagues from 2008 through 2020 has consistently shown that fingerprint examination conclusions are susceptible to anchoring, framing, and authority effects. The FBI Laboratory's own black-box study (Ulery et al. 2011) confirmed that error rates were not zero. The PCAST report of 2016 (President's Council of Advisors on Science and Technology, "Forensic Science in Criminal Courts: Ensuring Scientific Validity of Feature-Comparison Methods") concluded that fingerprint examination had a valid scientific foundation but required further error-rate research and urged the establishment of robust sequential unmasking and blind verification procedures.

The response from the fingerprint community has been substantive. The FBI Laboratory adopted linear ACE-V and sequential unmasking protocols between 2006 and 2012. The UK FSR made blind verification a formal requirement in the 2014 Codes of Practice revision. The ENFSI Fingerprint Working Group embedded contextual-bias management into its 2015 Best Practice Manual. The International Association for Identification (IAI), which governs the Latent Print Certification programme, addressed bias management in its 2018 and 2022 certification requirements.

The open research questions now cluster around three areas. First, how much contextual information is unavoidable in casework, and how should that information be managed rather than simply prohibited? Second, what is the optimal blind-verification workflow for high-volume laboratories where resource constraints are real? Third, how should the documented error rate inform the strength of fingerprint evidence in court, and how should that strength be communicated to a jury that has been told for a century that fingerprint identification is infallible?

These questions are being addressed simultaneously in the US by the OSAC Friction Ridge Subcommittee, in the UK by the Forensic Science Regulator's ongoing Codes of Practice revision cycle, and in Europe by the ENFSI research programme. The Mayfield error was a catastrophic failure that produced a research and reform programme whose depth the discipline had never previously had to confront.