SPEECH IS THE ACOUSTIC REALIZATION OF A NUMBER OF SIMULTANEOUS MESSAGES INTERWOVEN IN A COMPLEX OF WORDS, PHRASES, MOODS, AND INDIVIDUAL VOICE CHARACTERISTICS. ALTHOUGH THESE MESSAGES ARE HIGHLY INTERDEPENDENT AND DIFFICULT TO DISENTANGLE, HUMAN OBSERVERS CAN, TO A LIMITED EXTENT, IDENTIFY VOICES AURALLY OR BY VISUAL EXAMINATION OF THE ACOUSTIC PATTERNS OF SPEECH. THE ACOUSTIC SPEECH SIGNAL CAN BE ANALYZED BY FREQUENCY, ENERGY, AND TIME, AND RECORDED GRAPHICALLY ON A SPECTROGRAM. HUMAN OBSERVERS EXAMINE THE SPECTROGRAMS TO DECIDE SUBJECTIVELY THE IDENTITIES OF THE SPEAKERS IN QUESTION. SIMILARITIES AND DIFFERENCES AMONG SPECTROGRAPHIC PATTERNS ARE AMBIGUOUS AND MAY BE MISLEADING. FOR EXAMPLE, DIFFERENCES IN PATTERN, WHEN THE WORDS ARE THE SAME, MAY INDICATE DIFFERENT SPEAKERS OR ONLY NORMAL VARIATIONS IN THE UTTERANCES OF A SINGLE SPEAKER. SPEECH SPECTROGRAMS ARE NOT ANALOGOUS TO FINGERPRINTS. EXPERIMENTAL STUDIES HAVE FOUND FALSE IDENTIFICATION RATES RANGING FROM 0 TO 63 PERCENT IN VISUAL INTERPRETATION OF SPECTROGRAMS. ALTHOUGH ANECDOTAL EVIDENCE IS CITED IN SUPPORT OF THE SPEECH SPECTROGRAM METHOD, THE EVIDENCE IS NOT SCIENTIFICALLY CONVINCING. PROCEDURES FOR EVALUATING THE RELIABILITY OF VOICE IDENTIFICATION METHODS EXIST AND SHOULD BE IMPLEMENTED. SUPPORTING MATERIALS ARE APPENDED. (AUTHOR ABSTRACT MODIFIED).