IN TODAY'S PRACTICE, AN EXAMINER LISTENS TO RECORDED VOICE SOUNDS AND LOOKS AT VOICEGRAMS, WHICH GRAPHICALLY REPRESENT CERTAIN FEATURES OF VOICE SOUNDS, IN AN EFFORT TO MATCH VOICE SAMPLES FROM AN UNIDENTIFIED PERSON WITH THOSE OF ONE OR MORE IDENTIFIED INDIVIDUALS. BECAUSE TECHNOLOGY USED IN TRANSMITTING, RECORDING, REPRODUCING, AND ANALYZING SOUND WAS DEVELOPED FOR PURPOSES OTHER THAN VOICE IDENTIFICATION, PRESENT PRACTICE IS BASED ON LIMITED KNOWLEDGE ABOUT THE PROPERTIES OF VOICE SOUNDS AND ANALYSIS IS CONDUCTED LARGELY AS AN EMPIRICAL ART IN WHICH THE EXAMINER ACQUIRES SKILL THROUGH EXTENSIVE TRAINING AND EXPERIENCE. SCIENTIFIC RESEARCH IN PHONETICS AND ACOUSTICS HAS PRODUCED CONSIDERABLE INFORMATION ABOUT SPEECH SOUNDS AS RELATED TO THE SPEECH MESSAGE BUT RELATIVELY LITTLE INFORMATION ABOUT THE SOUNDS AS RELATED TO THE IDENTITY OF THE SPEAKER. THE PRACTICE OF VOICE IDENTIFICATION RESTS ON THE ASSUMPTION THAT INTRASPEAKER VARIABILITY IS LESS THAN OR DIFFERENT FROM INTERSPEAKER VARIABILITY. HOWEVER, THIS ASSUMPTION IS NOT WIDELY SUPPORTED BY SCIENTIFIC DATA, AND THEORY AND AVAILABLE VIEWPOINTS ABOUT PROBABLE ERRORS IN IDENTIFICATION DECISION RESULT MAINLY FROM VARIOUS PROFESSIONAL JUDGMENTS AND FRAGMENTARY EXPERIMENTAL RESULTS RATHER THAN FROM OBJECTIVE DATA REPRESENTATIVE OF FORENSIC APPLICATIONS. TO SOME EXTENT, THE VARIOUS LEGAL VIEWPOINTS REFLECT THIS TECHNICAL UNCERTAINTY. ALTHOUGH THE FIRST TESTIMONY ON VOICE IDENTIFICATION BY THE AURAL-VISUAL METHOD APPEARED IN A 1966 COURT CASE, NO MORE THAN 100 SUBSEQUENT CASES HAVE ARISEN. THE TECHNICAL UNCERTAINTIES CONCERNING THE PRACTICE OF VOICE IDENTIFICATION ARE SO GREAT THAT FORENSIC APPLICATIONS SHOULD CONTINUE TO BE APPROACHED WITH MUCH CAUTION. GRAPHIC DATA AND REFERENCES ARE INCLUDED. (KBL)