TEN FIELD-TRAINED POLYGRAPH EXAMINERS MADE BLIND JUDGMENTS OF A STRATIFIED SAMPLE OF THE POLYGRAPH RECORDS OF 112 CRIMINAL SUSPECTS. CORRECT CALLS AVERAGED 63.1 PERCENT ON RECORDS WHERE GROUND TRUTH WAS ESTABLISHED BY CONFESSIONS (VERIFIED) AND 62.1 PERCENT ON RECORDS WHERE GROUND TRUTH WAS NOT KNOWN (UNVERIFIED) BUT WHERE THE CRITERION MEASURE WAS THE ORIGINAL TESTING EXAMINERS' JUDGMENT. EVALUATOR'S HIT RATES IN BOTH SITUATIONS WERE QUITE SIMILAR, AVERAGING ABOUT 77 PERCENT TRUE POSITIVES AND ABOUT 51 PERCENT TRUE NEGATIVES; THE VARIABLE HAVING THE GREATEST EFFECT ON EVALUATORS' ERRORS WAS THE TYPE OF INVESTIGATION FROM WHICH RECORDS WERE DRAWN. RELIABILITY COEFFICIENTS SHOWED HIGH INTEREVALUATOR AGREEMENT ON BOTH VERIFIED AND UNVERIFIED RECORDS, .89 AND .85, RESPECTIVELY. HOWEVER, BECAUSE THIS STUDY DID NOT ADDRESS THE BASIC QUESTION OF THE USEFULNESS OF POLYGRAPH EXAMINERS' JUDGMENTS IN INVESTIGATIVE DECISION MAKING, TRUE-NEGATIVE AND TRUE-POSITIVE HITS WERE ASSUMED TO BE EQUALLY VALUABLE; FALSE NEGATIVES AND FALSE POSITIVES WERE ASSUMED TO BE EQUALLY COSTLY. TABULAR DATA AND REFERENCES ACCOMPANY THE TEXT. (WJR)