When theory does not establish a tight link from the physiological responses to the psychological states presumably tied to deception, and particularly when theory raises the possibility that states other than deception may generate physiological responses from which deception is inferred, inference faces a major logical problem.8 This problem is not obviated by advances in neural and physiological measurement, which is now often highly sophisticated and precise. The logical problem is generic to inferences about psychological states from physiological indicators.
Inference commonly follows the subtractive method, in which experimental and control or contrast conditions differ by one element, stage, or process (Strube, 1990; Cacioppo, Tassinary, and Berntson, 2000b). Outcome differences between the experimental and control conditions are then considered to reflect the effect of that single component. This method allows the construction of physiological indices of the psychological phenomena that have been varied in experiments, which are then used to develop concepts and test theories about those phenomena.
The subtractive method underlies the interpretation of the polygraph chart and of other indicators used for the psychophysiological detection of deception. If there are sufficiently more or stronger “arousal” responses to relevant than control questions, the polygraph chart is interpreted as “deception indicated” or as showing “significant response.” This approach does not allow a strong inference (Cacioppo and Tassinary, 1990a).9 The confidence in such an interpretation would be enhanced if the particular result (e.g., relatively large skin conductance responses) could be shown to arise consistently under a wide range of conditions of deception, and if the result could not be attributable to some other aspect of the stimulus or context (e.g., fear of being suspected or anxiety over trivial or irrelevant transgressions). Even then, however, the autonomic responses could not be used definitively to infer the presence of deception, as other antecedent conditions (e.g., emotional reactions) may yield the same result.10
In most polygraph research, a psychological factor (deception) serves as the independent variable and a physiological factor serves as the dependent variable. This format provides information about the likelihood of a physiological response given a person who is being deceptive. Such evidence is commonly offered to address the question of how good the polygraph test is as a diagnostic of lying. However, a polygraph test, like other diagnostic instruments, is actually used to make the reverse inference: about the likelihood of deception given the physiological response