We investigated response bias (defined as the decision rule subjects adopt when uncertain) in two experiments using a variant of Signal Detection Theory (SDT) with the discrimination measure d'L and the bias measure CL, under which it is possible to independently evaluate discrimination and response bias. In the first experiment, manics, depressed subjects, and matched psychiatrically normal controls were tested with a recognition memory task with easier and more difficult components before and after 1 month of appropriate pharmacological treatment. This experiment showed that abnormally conservative bias was characteristic of depression and liberal (yea-saying) bias was found in mania regardless of severity of illness; discrimination deficits were found only when symptoms were severe. After treatment, aspects of discrimination worsened in both hypomanic and depressed nonresponders whereas response bias remained unchanged in these patients. In both groups of responders, improvements in response bias were more dramatic than improvements in discrimination. In the second experiment, psychiatrically normal hypertensives were tested with a Sternberg short-term memory scanning task on and off treatment with the centrally active beta-blocker propranolol. This experiment showed that treatment with propranolol modeled the bias deficit of depression; that is, bias became more conservative. Both sets of results suggest that disorders of decision may be modulated by beta-adrenergic function.