The present study investigated the influence of positive affect on clinical reasoning among practicing physicians using a simulated patient protocol. Forty-four internists were randomized to one of three groups; a control group, an affect-induction group, in which they received a small package of candy, or a group that was asked to read humanistic statements regarding the practice of medicine. Physicians "thought aloud" while they solved a case of a patient with liver disease. Two raters reviewed the transcripts to determine how soon the liver disease domain was considered, or at what point it was established, and the degree of anchoring displayed (distortion or inflexibility in thinking). The Affect group initially considered the diagnosis of liver disease significantly earlier in the protocol (20 vs 39%, or 19% earlier; 95% confidence interval; 6 to 32%,p= .008) and showed significantly less anchoring than did controls (means of 1.5 vs 3.9, or a difference of 2.4 in a scale of 1 to 10; 95% confidence interval; 0.3 to 4.5,p= .031). The Affect and Control groups established the diagnosis at similar points in the protocol. Ancillary analyses examined questions relevant to the amounts of heuristic processing, premature closure, jumping to a diagnosis without sufficient evidence, or other evidence of superficial or flawed processing. No other differences were found. Physicians in whom positive affect had been induced integrated information earlier (considered liver) and demonstrated less anchoring, but did not show evidence of premature closure or other superficial processing. © 1997 Academic Press.