It is suggested that the task of drawing conclusions from data on hand about other data is a problem of statistical inference. The scientist seldom proves anything with certainty or beyond a doubt. The difference between the scientific and legal situations is that the scientist has learned to calculate the magnitude of the doubt. This fact is the contribution of statistics. Thus, statistical proof underlies all scientific investigations. Any opinion of identification, as in handwriting analysis, derives from statistical inference and is an expression of probability. Qualified opinions should be thought of in terms consistent with their statistical basis; their translation into probability values serves to avoid confusion, ambiguity, and misunderstanding. Statements that a given individual probably executed a given writing are technically improper and should be avoided. The examiner should be prepared to rephrase a statement in a technically acceptable form, such as 'the probability is something less than 0.001.' Qualified opinions constitute evidence regardless of the nature of the legal forum. Their admissibility should be governed by the normal rules of evidence. In addition, it should be remembered that document examination may produce evidence of either identification or elimination, a fact which further justifies the requirement of precision in expert opinions. Two footnotes and 14 references are included in the article. (Author abstract modified).