A Review of the Office of Technology Assessment Report on Polygraph Validity

This article reviews the assessment of polygraphic lie detection produced by the Office of Technology Assessment (OTA) in November 1983. The review argues that use of the lambda statistic to summarize polygraph accuracy in the OTA report was inappropriate because the studies examined differ widely in base rates of guilt and innocence. Using Lykken's (1981) average accuracy statistic and avoiding overlapping data sets found in the OTA report, the review finds that field studies, analog studies, and guilty knowledge studies produce very similar average accuracy (82% to 88%). Guilty knowledge tests tend to be biased toward false negatives, whereas control question tests tend to be biased toward false positives. The striking similarity of results for field and analog studies using control question tests contradicts the common belief that results of analog testing are weak and not generalizable to field practice. It is argued that analog studies are in fact the only hope for improving the validity of polygraph testing.