Centrally mediated or cognitive variables have received considerable attention in clinical research. With the establishment of the effectiveness of such specific treatment techniques as systematic desensitization (Paul. 1969a. 1969b). a question arises as to the influence of cognitions on process and outcome variables in such learning based treatments (Lang. 1971). Research investigating the influence of demand characteristics and subject expectancies has demonstrated such centrally mediated variables can significantly influence overt, behavioral and self-report measures of fear, stress, or anxiety (cf. Borkovec. 1972: Marcia. Rubin and Efran. 1969; McGlynn. Maelia and Nawas. 1969: McGlynn. Reynolds and Linder. 1971: Oliveau. Agras. Leitenberg and Wright. 1969; Rappaport. 1972: Rosen. 1974). However. Borkovec (1972) and Rappaport (1972) failed to obtain expectancy effects on physiological measures of anxiety. Neither investigation provided for an independent assessment of subjects' actual expectancies, or belief in the instructions, as recommended by Davison and Wilson (1973). Borkovec administered clinical procedures over four sessions and varied the information communicated about the ‘purpose’ of the treatments (physiological versus therapeutic instructions). Differential rationales for differential expectancies of physiological reactivity were not provided. After each session, subjects reviewed the same false physiological records depicting reductions in fear responses. Although these changes were explained either physiologically or therapeutically. the fact that such reductions were emphasized should have attenuated differences between groups. Rappaport (1972) exposed subjects to a fearful stimulus and varied the suggestion and rationale for the pseudotreatment: no expectancy (stress research): therapeutic expectancy (fear reduction); and negative expectancy (fear increase). The lack of an independent assessment of the expectancy manipulation, as well as the exclusion of a specific therapeutic procedure, make it impossible to interpret the results in relation to expectancy effects on physiological fear responses during the administration of therapeutic procedures. There is other evidence, however, that suggestion and instructional set can significantly influence somatic and autonomic response systems (Barber. 1961. 1965; Sternbach. 1964. 1965). The primary purpose of this study was to evaluate the effects of instructional set on physiological responses to stressful imagery. This was an early step in a research program investigating the effects of instructional and informational variables on physiological responses of clinical interest. One half of the subjects received abbreviated live relaxation training ( R) as described m a manual by Bernstein and Borkovec (1972) and based upon the original work of Jacobson (1938). The other half received an inert placebo pill and undertook a target detection task ( P). The latter procedure was similar to that used by Paul (1966). Although both procedures had previously been shown to have relaxing effects, the P task was included as an additional control for any potential effects specific to the tension-release procedure of R. In order to evaluate the effects of instructional set, one half of each of the above two groups received instructions designed to lead to an expectancy of response inhibition to stressful imagery by the respective procedure (Decrease Stress Response. DSR). while the other half received instructions designed to lead to an expectancy of an increased response to stressful imagery by the respective procedure (Increase Stress Response. ISR). Each subject visualized three individually specified items that were the most frightening scenes imaginable prior to training and four times after training visualized the scene producing the largest response. Considerable effort was devoted to eliciting from each subject the most frightening items possible for her. and it was clear from subject comments that the items did elicit negative emotional responses. Since previous research (Barber. 1961. 1965; Sternbach. 1964. 1965) indicated that instructional set can influence physiological reactivity, it was hypothesized the DSR) subjects should exhibit a greater reduction in emotional response from pre- to posttraining than ISR subjects.