THE ONLY LARGE-SCALE STUDY OF OBSERVER ACCURACY IN THE INTERPRETATION OF VOICEPRINTS WAS CONDUCTED AT MICHIGAN STATE UNIVERSITY. ON THE AVERAGE, STUDY OBSERVERS IDENTIFIED THE WRONG PERSON IN 6.4 PERCENT OF THE VOICEPRINT MATCH ATTEMPTS AND FAILED TO IDENTIFY ANYONE IN 11.8 PERCENT OF THE ATTEMPTS. ON THE BASIS OF THEIR FINDINGS, THE MICHIGAN STATE RESEARCHERS MADE PREDICTIONS CONCERNING ERROR RATES FOR PROFESSIONAL VOICEPRINT ANALYZERS IN FORENSIC SITUATIONS. SOON AFTER THE STUDY WAS COMPLETED IN 1970, A MICHIGAN DEPARTMENT OF STATE POLICE DETECTIVE LIEUTENANT BEGAN TESTIFYING AS A VOICEPRINT EXPERT THROUGHOUT THE COUNTRY. THE DETECTIVE'S TESTIMONY HAS BEEN ADMITTED IN OVER 50 COURTS. THE CONDITIONS AND ASSUMPTIONS UNDER WHICH THE VOICEPRINT METHOD WAS TESTED AT MICHIGAN STATE ARE COMPARED WITH THE CONDITIONS ENCOUNTERED BY THE DETECTIVE AS A PROFESSIONAL VOICEPRINT EXAMINER. THE COMPARISON INDICATES THAT CERTAIN PARAMETERS WHICH MAY ADVERSELY AFFECT THE ACCURACY OF THE PROFESSIONAL EXAMINER WORKING ON ACTUAL CASES HAVE NOT BEEN INVESTIGATED ADEQUATELY. OTHER PARAMETERS -- THE ALTERED PHYSIOLOGICAL STATE OF THE SPEAKER, DISGUISE OF THE SPEAKER'S VOICE, INTERSPEAKER VARIABILITIES WITHIN LINGUISTIC COMMUNITIES -- HAVE NOT BEEN INVESTIGATED AT ALL. IT IS CONCLUDED THAT, ALTHOUGH ERROR RATES ACHIEVED BY EXAMINERS IN THE LABORATORY ARE LOW, ERROR RATES UNDER FIELD CONDITIONS MAY BE PROHIBITIVELY HIGH.