To date, no single technique of voice identification is completely reliable, i.e., one that could yield correct speaker identifications or eliminations in 100 percent of the cases, irrespective of the sample length, background noise, and recording system distortions. Typically, all types of aural examination of voices and visual examination of speech spectrograms are considered subjective methods, although the latter is closer to the objective part of the spectrum of methods than the former. Semiautomatic and automatic methods are considered objective. Whether subjective or objective, voice examinations for legal purposes always require the intervention of an examiner, at least to prepare the samples and to interpret the results from the computer. Tests of voice identification or elimination can be classified into three general groups according to the composition of known and unknown samples: (1) discrimination tests, where the examiner is provided with a known sample and an unknown sample; (2) open tests, where the examiner is given several known samples and an unknown sample; and (3) the closed test, where the examiner is told that the unknown speaker is one of the known speakers. Other topics considered are the receiver operating characteristic, speech production theory, aural examination, and legal cases. Graphic illustrations are provided.