Interpreting Biological Evidence in Court

For stated propositions H1 and H2 and findings E, the likelihood ratio is P(E given H1) divided by P(E given H2). An LR above one supports H1 relative to H2, below one supports H2 relative to H1, and one provides no discrimination between them through those findings.

The propositions must be explicit and relevant. A contributor comparison might contrast the person of interest contributing with an unknown unrelated person contributing, alongside any specified other contributors. A close relative or a different contributor count may require a different comparison rather than an unchanged denominator.

In an idealised single-source model, a complete matching genotype can produce an LR approximately reciprocal to its genotype probability under the alternative. Real calculations require justified population-genetic and analytical assumptions; error, relatedness and other issues cannot always be ignored by assigning the numerator one.

Allele-frequency reference datasets differ from operational databases of offender or crime-scene profiles. CODIS and the UK National DNA Database are not interchangeable names for population-frequency tables. Database choice, population structure and applicable corrections should be justified, with relevant sensitivity analyses where appropriate.

Both “the chance of innocence is one in a billion” and “the chance someone else deposited this DNA is one in a billion” go beyond a random-match calculation. Posterior propositions require additional information and reasoning. The LR alone neither counts possible offenders nor supplies prior odds.

A DNA mixture contains contributions from more than one person. Peak overlap, stutter, degradation, imbalance and dropout can make interpretation difficult. A missing allele does not automatically exclude a person if the validated model allows dropout, while apparent agreement does not automatically establish contribution.

Probabilistic genotyping evaluates findings under genetic and analytical models. Depending on the system, it can model peak heights, stutter and other effects and calculate LRs for specified propositions. Systems do not necessarily consider every possible contributor count or every possible artefact automatically.

STRmix and TrueAllele are examples of probabilistic genotyping systems, but the product name is not proof that a particular laboratory application is valid. Software version, settings, training, validation range and the tested case conditions matter. A national adoption claim does not replace laboratory-specific evidence.

The analyst should explain contributor assumptions, conditioning on known people, population information and relevant limitations. Alternative plausible assumptions can affect the LR. Model diagnostics and repeated computational runs may help assess performance without proving that the assumed biological history is correct.

Disclosure and independent scrutiny follow the applicable process. Data, methods, validation and software-related information may be important, but there is no universal rule that every case automatically requires release of source code. The scientist should identify what was available, what was withheld and how limitations affect review.

A report should identify the items and methods, distinguish observations from interpretation, and state the propositions behind an evaluative conclusion. Include relevant controls, assumptions, limitations and departures. A reader should be able to tell whether a conclusion addresses a fluid, a contributor or an activity.

Numerical LRs and verbal descriptions can communicate evidential support, but verbal scales are conventions with defined ranges. There is no single universal ENFSI table governing every laboratory and court. If a scale is used, identify the applicable guidance and retain the proposition pair and numerical information where appropriate.

A verbal label must not manufacture a number that the examination cannot support. Conversely, an impressive number should not hide fragile assumptions or a question mismatch. Neutral results and findings favouring the alternative must receive equally clear treatment.

Quality frameworks and procedural rules are distinct. In the US, OSAC supports development and implementation of consensus standards; it does not itself legislate nationwide admissibility. ISO/IEC 17025 assesses laboratory competence within scope, while the court applies its legal criteria.

For England and Wales, the statutory Forensic Science Regulator framework applies to specified forensic activities and requirements. It should not be described as one undifferentiated rule for every UK legal system or every scientific activity. Indian and other courts apply their own rules; a quality badge is not an automatic admission certificate.

A sub-source DNA comparison asks about genetic contribution, while a source question may concern material such as blood or semen. An activity question concerns how material arrived. The same DNA result can be relevant to several questions without answering them with the same calculation.

A high contributor LR can remain relevant even when contribution is agreed, but may add little to the actual dispute about innocent contact or transfer. It is misleading either to treat the number as proof of the alleged activity or to say it has no meaning in every such case. State the level of the conclusion.

Activity-level evaluation can consider transfer, persistence, background and recovery using suitable data and specified circumstances. It requires more than listing secondary transfer as theoretically possible. Where reliable evaluation is unavailable, explain that limit rather than inventing a transfer probability or leaving a false impression that biology settles it.

A partial profile often contains less information than a comparable complete profile, but no fixed number of loci guarantees a particular LR. The alleles, mixture, analytical uncertainty, reference population and propositions matter. A degraded specimen is not automatically weak evidence, nor does a very large LR make degradation irrelevant.

DNA can be transferred indirectly, and failure to detect a person’s DNA does not establish that the person was absent. Collection, persistence and analytical sensitivity affect detection. Neither a positive nor negative result independently resolves consent, intent or guilt.

Contamination and continuity concerns need investigation and proportionate explanation. Good records improve assessability but do not eliminate every error pathway. The legal consequence of a gap varies with its nature and jurisdiction rather than following an automatic exclusion rule.

In US federal proceedings, Rule 702 addresses expert assistance, sufficient basis, reliable principles and their reliable application. The proponent must establish the rule’s admissibility conditions to the court’s required standard. State rules and case law can differ, so a federal description should not be presented as identical law in every US court.

In the UK, distinguish the relevant legal system. England and Wales use Criminal Procedure Rules Part 19 for criminal expert evidence, including objective, unbiased assistance within expertise and reporting duties. Civil Procedure Rules Part 35 concern civil work and should not replace the criminal rules in a crime-scene lesson.

In India, section 39 of the Bharatiya Sakshya Adhiniyam 2023 addresses relevant expert opinions. Applicable commencement and transitional arrangements matter when older proceedings are discussed. The provision does not turn every expert assertion into proven fact or require one universal LR method for all biological results.

Canada and Australia have their own admissibility law and procedural requirements, with variation across courts and types of proceedings. Elsewhere, check the current local framework before relying on a familiar common-law label. The common teaching principle is to identify competence, method, basis and limits; the precise obligations vary by jurisdiction.

Before testimony, review the actual data and records, not only the final conclusion. A scientist may explain validated biology while referring questions outside expertise to an appropriate specialist. If an opinion changes, disclose the change through the applicable process rather than silently defending an outdated statement.

Review should check the chain from observation to calculation to wording. Does the report describe the actual sample, distinguish mixtures and partial results, state appropriate alternatives and avoid equating agreement with unique identity? A technically correct LR can still be presented misleadingly.

Independent technical review and access to relevant records can identify problems before testimony. Quality systems do not guarantee that every error will be found. When a material problem emerges, the laboratory should assess affected work and follow appropriate correction and notification procedures.

The FBI and US Department of Justice review of microscopic hair comparison identified historical statements that exceeded the science. That documented example illustrates a general reporting risk without justifying an invented claim that every biological discipline or every country conducted the same closed-case review.

Clear explanations need not abandon precision. “The findings are more probable under these propositions” is different from “this person committed the offence.” Avoid analogies that silently change the question or import certainty from another discipline.

A useful final check asks what a reasonable reader might wrongly infer. Correct misunderstandings about timing, direct contact, contributor certainty and guilt explicitly where relevant. The goal is a conclusion the court can assess fairly, including evidence that does not support the instructing party’s preferred account.