The Record/Entry 3.04/One technique, and what it can and cannot carry
Cleared of the method
Several early techniques were later withdrawn or restricted after review, which is how a forensic discipline is supposed to work.
When the science fails the review, the record matters — and several early forensic DNA techniques did not survive scrutiny intact.
What the reviews found
DNA evidence arrived in court with enormous authority. The early multilocus probe results were striking images — banded patterns that looked as definitive as a fingerprint — and juries found them compelling. Yet even within the first decade, problems emerged that were technical rather than procedural: questions about band-matching tolerances, about the minimum signal needed to call a result, about whether visual comparison of autoradiographs met any defensible statistical standard. The technique was not discredited outright, but it was gradually displaced as capillary electrophoresis and digital peak calling replaced eye and film.
The harder reckoning came with mixed and low-template samples. When profiles contain more than one contributor, interpretation requires choices — how many contributors, which peaks belong to whom, whether a faint peak is signal or stochastic noise. For years those choices were made by examiners following internal guidelines that varied between laboratories, and the outputs were reported as conclusions rather than probability estimates. A 2009 National Academy of Sciences report examined the state of forensic science across disciplines and found that DNA was better validated than most, but that mixture interpretation and the statistical frameworks surrounding it remained inconsistent and sometimes poorly documented. The President's Council of Advisors on Science and Technology reached similar conclusions in a 2016 report that assessed the scientific validity ↗ of pattern-evidence methods across the field.
The restricted and the withdrawn
Bite-mark evidence, hair microscopy and certain serology methods were found to rest on validation that could not bear examination and were restricted or abandoned in a number of jurisdictions. DNA is not in that category, but specific protocols within it have been. Low-copy-number (LCN) amplification — pushing PCR beyond the standard cycle count to extract a profile from very few cells — was suspended in England and Wales in 2007 after concerns about a specific case and then reinstated with tighter interpretive rules. The episode illustrated the mechanism by which a technique should self-correct: a documented problem triggers a formal review, the technique either earns qualified reinstatement or it does not, and the decision is published.
The UK's Forensic Science Regulator ↗ now holds a statutory code, introduced under the Forensic Science Regulator Act 2021, requiring laboratories to demonstrate that their methods are fit for purpose before reporting results in criminal proceedings. SWGDAM in the United States has issued successive rounds of interpretation guidelines — the 2010 and 2017 versions each narrowed what an examiner may assert without probabilistic modelling to support it. These are not cosmetic revisions; they reflect a discipline acknowledging that earlier standards were insufficient.
From the register
Key events in the correction record
- 1989Castro case, New York: defence challenge to RFLP band-matching standards, DNA evidence excluded after a pretrial admissibility hearing
- 2007LCN amplification suspended in England and Wales pending review, then reinstated with new interpretive limits
- 2009National Academy of Sciences report identifies inconsistent mixture interpretation across US forensic laboratories
- 2016PCAST report assesses scientific validity of forensic feature-comparison methods; DNA mixture interpretation flagged as needing reform
- 2021Forensic Science Regulator Act gives the UK Regulator statutory powers to enforce laboratory standards
Why the record holds
The value of the correction record is not that it undermines DNA evidence — it is that it shows the mechanism working. Techniques that could not survive review were removed or constrained. The Innocence Project's post-conviction cases, now numbering more than three hundred and seventy-five DNA exonerations in the United States, include a proportion in which the original conviction relied partly on a forensic method — hair comparison, bite marks — that has since been rejected. DNA was the instrument of correction, not the source of the error.
What the review process has not fully resolved is the interpretation layer: who decides when a mixture is too complex to report, what statistical model governs the result, and whether software-generated likelihood ratios are explained to courts in terms that are genuinely understood. Those questions are live. NIST has published reference data and proficiency materials; ENFSI runs collaborative exercises ↗ so that European laboratories can be compared with one another. The infrastructure for accountability exists. Whether it is consistently used is a different question, and one the discipline continues to answer in public.
