Reference Standards, Controls and Comparison Logic

Every forensic comparison begins with a categorical distinction: which sample has a known origin, and which is the sample whose origin is under investigation? The known sample, called K in many traditions, is the anchor. It might be a buccal swab taken from a named individual under caution, a cutting from a specific garment seized from a suspect, a paint standard reference from the manufacturer's database, or soil taken from a GPS-mapped location. The origin of K is documented and the documentation is part of the exhibit record.

The questioned sample, Q, is recovered from a context whose significance is not yet established. The blood on a door handle, the fibre on a victim's coat, the glass fragment in a suspect's hair: these are Q samples. Comparison answers the question of whether Q and K are consistent with sharing a common source, and if so, how strongly the agreement supports that conclusion.

Suppose a forensic glass examiner measures the refractive index of a fragment recovered from a suspect's jacket and reports it as 1.5183. That number only becomes evidence when it can be compared to the index measured on the broken window at the scene. For the comparison to be valid, both measurements must be on the same scale: they must be traceable to the same calibration standard.

Certified Reference Materials are the mechanism that provides this traceability. A CRM is a material whose property has been measured by a national metrology institute (such as NIST in the United States or NPL in the United Kingdom) and certified with a stated value and uncertainty. Forensic laboratories use CRMs to calibrate their instruments. When Instrument A and Instrument B in two different cities are both calibrated against the same CRM, a measurement made on one can be meaningfully compared to a measurement made on the other.

For many forensic applications, purpose-built CRMs exist. The NIST Standard Reference Materials (SRM) programme offers glass standards with certified refractive indices (SRM 1822a), DNA profiling standards with certified allele designations and genotypes (SRM 2391 series), and ethanol solutions for blood alcohol calibration with certified concentrations. The UK Forensic Science Regulator's statutory Code of Practice, which came into force on 2 October 2023 under the Forensic Science Regulator Act 2021, requires that all methods used in regulated UK forensic laboratories be linked to traceable reference standards. Without this linkage, measurements cannot be challenged, corrected, or compared across time.

Every batch of casework samples is accompanied by a set of controls that run through the entire analytical process, from sample preparation through to instrument output, under identical conditions. Controls are not procedural add-ons; they are the primary check that the batch result is valid.

The positive control failure and the negative control failure are the two critical failures in a batch. A failed positive means the method missed the target: perhaps a reagent degraded, an instrument went out of specification, or a preparation step was omitted. Every case result in that batch must be treated as potentially false-negative until the cause is found and re-testing confirms the results. A failed negative means something contaminated the batch: a reagent, a surface, analyst carry-over from a previous high-concentration sample. Every case result must be treated as potentially false-positive.

When a K and a Q sample agree on a set of measured characteristics, the analyst must reason about what that agreement actually means. The critical variable is discrimination power: how many independent sources would also agree on these characteristics, and how is the relevant population defined?

Most trace evidence carries class characteristics: properties shared by all items made the same way. A blue polyester fibre has a colour and cross-section shared by all fibre from the same production run, potentially millions of garments worldwide. Agreement on class characteristics supports a common source but does not exclude other sources in the same class. Individual characteristics are features that, by theory or empirical study, are unique to one item: the ridge detail of a fingerprint, the random striation pattern on a bullet from a specific barrel, the full 20-locus STR profile in a large enough reference database.

• Class characteristic match: narrows the population of possible sources. Reports should state what class was established and how common that class is, not imply individual identification.
• Combination of class characteristics: each additional agreeing characteristic multiplies the discrimination if they are independent. Three independent class characteristics that each occur in 1 in 100 items together occur in roughly 1 in a million, but only if genuinely independent.
• Individual characteristic match: supports a very strong probabilistic link to a single source, but even here the question of how large the reference population is and whether the characteristic is truly unique must be addressed.

A blank is a control with no target material, processed through the entire analytical workflow. It probes a specific question: did any target material enter the system from sources other than the case samples? The sources of contamination in a forensic analytical workflow are numerous: reagents that were manufactured using the same cell lines as a reference standard, surfaces and tools that were not fully decontaminated between samples, analyst carry-over from handling a high-concentration sample earlier in the session, and environmental DNA from biological aerosols in the laboratory space.

The blank's passage through the full workflow, including extraction, amplification, and detection, is what makes it informative. A blank that is only subjected to the final detection step tells you nothing about whether contamination was introduced during extraction. A full-process blank mimics the path the case sample takes and can catch problems at any point along it.

In DNA laboratories, contamination management is particularly rigorous because PCR amplification can turn a few cells' worth of contamination into a strong signal. The UK National DNA Database, for example, holds elimination profiles for all laboratory staff, so that if an analyst's DNA profile appears in a case result it can be identified and reported as a contamination event rather than being mistaken for a true case finding. Many laboratories also hold profiles for cleaners, maintenance staff, and anyone who regularly enters the analysis area.

When a comparison shows agreement between K and Q, the analyst must answer three questions before reaching a conclusion. First: how many characteristics agree, and are they class or individual? Second: how common are those characteristics in the relevant population? Third: what is the probability of the agreement under the prosecution proposition (the K and Q are from the same source) versus the defence proposition (K and Q are from different sources but happen to agree by coincidence)?

These three questions are the structure of the likelihood ratio approach. The LR does not produce a finding of guilt or innocence; it quantifies how much the scientific evidence should shift a rational person's belief about which proposition is more probable. An LR of 10,000 for a fibre comparison means the observed agreement is 10,000 times more probable if the fibres came from the same source than if they came from different sources in the relevant population. Communicating that number accurately, and its uncertainty, is the analyst's job.

Controls and reference standards are what give the LR calculation its validity. Without a certified reference material anchoring the refractive index measurement, the numerical agreement between K and Q is meaningless. Without a negative control confirming no contamination, the agreement might be an artefact. Without a positive control confirming the method's sensitivity, a non-agreement cannot be reported as a meaningful exclusion. Every element of the comparison logic rests on the validity of the measurement infrastructure beneath it.