Probability is the mathematical language for expressing the strength of forensic evidence. The three Kolmogorov axioms establish that every event has a probability between 0 and 1, the certain event has probability 1, and mutually exclusive events combine by addition.
From these axioms follow the addition rule, the multiplication rule, and the complement rule, the four building blocks used in every area of forensic statistics from DNA profiling to glass comparison. A practitioner who cannot apply these rules cannot correctly compute a random match probability, evaluate a likelihood ratio, or understand the reports written by colleagues in other disciplines.
Forensic scientists regularly work with very small or very large probabilities. A nine-locus STR profile probability may be 1 in several trillion. A database search over a large population may produce many false positives even when individual match probabilities are low. Both situations require clear reasoning about probability rules, not intuition. Courts in the UK, US, India, Germany, Australia, and elsewhere have repeatedly received evidence in which probability calculations were performed incorrectly because practitioners misunderstood which rule applied.
This topic builds the vocabulary and mechanics needed for every subsequent module in the subject. The rules covered here are applied directly when computing random match probabilities and when constructing the conditional probability statements that underlie likelihood ratios and Bayesian reasoning. Mastering them now prevents systematic errors in all that follows.
By the end of this topic you will be able to:
- State the three Kolmogorov axioms and explain what each means for a forensic probability statement.
- Apply the addition rule to events that are mutually exclusive and to events that are not, and identify which form is appropriate in a given forensic scenario.
- Apply the multiplication rule under independence and under conditional dependence, and recognise when independence cannot be assumed.
- Calculate a complementary probability and explain why the complement framing can change how a jury perceives the same statistic.
- Distinguish mutually exclusive events from independent events and avoid the common error of treating them as equivalent.
- Sample space
- The set of all possible outcomes of an experiment. In forensic genetics, the sample space for a single allele call is the set of all alleles observed in the relevant population database. The probability of the entire sample space equals 1 by the second Kolmogorov axiom.
- Event
- Any subset of the sample space to which a probability is assigned. An event may be a single outcome (this particular allele) or a collection of outcomes (any allele with frequency above 0.05). Forensic probability statements are always about events, not individual measurements.
- Mutually exclusive events
- Two events that cannot both occur in the same trial. If a fibre is classified as cotton, it cannot simultaneously be classified as polyester under the same classification scheme. Mutually exclusive events obey the simple addition rule: P(A or B) = P(A) + P(B).
- Independent events
- Two events where the occurrence of one provides no information about the other: P(A and B) = P(A) x P(B). In forensic STR profiling, loci on different chromosomes are treated as independent, permitting allele frequencies to be multiplied across loci to compute a profile probability.
- Complement
- The complement of event A is the event that A does not occur. Its probability is 1 - P(A). In court, practitioners frequently present the complement of a match probability to communicate how rare a coincidental match would be: if P(match) = 0.0001, then P(no match) = 0.9999.
- Conditional probability
- The probability of event A given that event B has already occurred, written P(A|B). Conditional probability is the basis of the multiplication rule in its general form and the foundation of the likelihood ratio framework used in evaluative forensic reporting. It is treated in depth in the next topic.
The Kolmogorov Axioms
Andrei Kolmogorov published his axiomatic treatment of probability in 1933. Before that, probability had been used widely but defined inconsistently. The three axioms he proposed are brief, but every rule in this topic and every formula in forensic statistics follows from them.
Scroll horizontally if needed to see the full table.
These axioms are not definitions of what probability means physically, they say nothing about whether probability represents a long-run frequency, a subjective degree of belief, or something else. That philosophical question matters for how forensic scientists interpret probability statements in court, but the mathematics is the same regardless of interpretation. A Bayesian forensic scientist and a frequentist forensic scientist both use these three axioms.
The Addition Rule
The addition rule calculates the probability that at least one of two events occurs. It has two forms depending on whether the events can overlap.
For mutually exclusive events: P(A or B) = P(A) + P(B). For any two events: P(A or B) = P(A) + P(B) - P(A and B). The subtraction corrects for double-counting: if both events can occur together, the joint probability P(A and B) is counted once in P(A) and once in P(B), so it must be subtracted once.
Forensic example: a population database shows that 12% of individuals carry the variant at locus X, and 8% carry the variant at locus Y. Of these, 2% carry both. The probability that a randomly selected person carries at least one of the two variants is 0.12 + 0.08 - 0.02 = 0.18, or 18%.
If the analyst assumed the events were mutually exclusive and simply added 0.12 + 0.08 = 0.20, the answer would be wrong by 2 percentage points, a material error in a forensic calculation.
The Multiplication Rule
The multiplication rule calculates the probability that two events both occur. In its general form: P(A and B) = P(A) x P(B|A), where P(B|A) is the probability of B given that A has already occurred. When A and B are independent, P(B|A) = P(B), so the rule simplifies to P(A and B) = P(A) x P(B).
The product rule for STR profiling is the most familiar forensic application. If a profile comprises six loci and the allele pair frequencies at each locus are 0.10, 0.08, 0.12, 0.06, 0.09, and 0.11, then under independence the profile probability is 0.10 x 0.08 x 0.12 x 0.06 x 0.09 x 0.11, which is approximately 5.7 in a billion.
The independence assumption is validated by testing for linkage disequilibrium across loci in the population database; loci on separate chromosomes are generally independent, while loci on the same chromosome may not be.
Trace evidence provides a case where independence may not hold. Suppose a perpetrator transfers both glass fragments and textile fibres. The probability of finding both types of trace on a suspect depends on whether the two transfer events are independent.
If both came from the same physical contact, the probability of the joint transfer is not simply the product of the individual transfer probabilities. The multiplication rule must be applied in its conditional form, or the model must account for the correlation between transfers.
Complementary Events
Every event A has a complement, written A' or A-complement, which is the event that A does not occur. By the normalization axiom: P(A) + P(A') = 1, so P(A') = 1 - P(A). This is the complement rule.
The complement rule is computationally useful when it is easier to calculate P(A') than P(A). In a DNA database search across N profiles, the probability of obtaining at least one adventitious match by chance is easier to compute as 1 - P(no match at all).
If the per-profile match probability is p, and the N profiles are independent, then P(no match at all) = (1 - p)^N, and the probability of at least one match is 1 - (1 - p)^N.
The complement also shapes how evidence is communicated. Research by cognitive psychologists including Gerd Gigerenzer has shown that people consistently interpret probability statements differently depending on framing. A statement that 'the probability of a coincidental match is 1 in 100,000' produces different intuitive responses from 'in a search of a database of 100,000 unrelated people, about one would be expected to match by chance'.
The second version describes the complement event and a base rate simultaneously. In England and Wales, the Court of Appeal in R v Adams [1996] addressed these communication issues, and the topic remains active in comparative studies from US courts, the Netherlands, and Australia.
Mutually Exclusive vs. Independent: A Critical Distinction
Mutually exclusive and independent are different properties. Mutually exclusive means the events cannot both occur. Independent means the occurrence of one does not change the probability of the other. The two concepts are not equivalent, and confusing them produces calculation errors that have appeared in published forensic science reports.
Scroll horizontally if needed to see the full table.
A practical illustration: blood type A and blood type B are mutually exclusive (a person cannot have both under the ABO classification in its simple form). They are not independent: knowing a person is type A changes the probability they are type B from its prior value to zero. By contrast, the ABO blood type of a person and whether they carry a particular Y-chromosome haplotype are approximately independent, because the two traits are governed by unrelated genetic mechanisms.
In forensic practice, the distinction matters most when combining evidence from multiple tests or measurements on the same sample. Two tests that target correlated biological features are not independent, even if the practitioner applies them sequentially and records separate results. An expert who multiplies the probabilities from two correlated tests has applied the multiplication rule under an invalid independence assumption and will produce a combined probability that is too small.
Probability Rules in Evaluative Reporting
The probability rules in this topic are not abstract exercises. They are the operational tools that forensic scientists use to move from a laboratory measurement to a number that is meaningful in a court. The history of statistical evidence in courts shows how errors in applying these rules have contributed to miscarriages of justice.
The Sally Clark case in England (1999) involved the incorrect application of the multiplication rule to non-independent events. The Dreyfus affair in France (1894) involved contested probability calculations about handwriting. Both occurred because probability rules were applied by experts who understood them imperfectly.
Modern evaluative reporting frameworks, including the ENFSI Guideline for Evaluative Reporting (2015), the UK Forensic Science Regulator's Codes of Practice, and guidance from the National Institute of Standards and Technology (NIST) in the US, all require forensic experts to state explicitly the probabilistic model underlying their conclusions, which rules they applied, and which assumptions they made.
The same requirements appear in the Bharatiya Sakshya Adhiniyam 2023 (sections governing expert opinion evidence) as interpreted by Indian appellate courts, and in the comparable provisions of the US Federal Rules of Evidence Rule 702 as shaped by Daubert v. Merrell Dow Pharmaceuticals (1993) and the 2023 amendments.
Forensic statistics as a field exists partly because these rules are easy to state and genuinely difficult to apply correctly when sample sizes are small, populations are structured, or evidence types are combined. Understanding the rules at the mathematical level is the prerequisite for understanding where they can fail. The subsequent topics in this subject treat those failure modes in detail, starting with the role of statistics in evidence evaluation.