Forty outpatients with primary depression were randomly assigned on a double-blind basis to treatment with amitriptyline (a tricyclic antidepressant) or clovoxamine (a nontricyclic, experimental antidepressant). Memory and depression were assessed during a pretreatment baseline period and at the end of days 4, 7, and 28 of drug treatment. A signal detection recognition memory task and conventional memory measures (including the Benton Visual Retention, Wechsler Logical Memory, and verbal learning tests) were used to assess memory. Although both drugs led to comparable clinical improvement in depression, they affected memory performance differently. The signal detection recognition memory task detected an impairment in memory after chronic amitriptyline administration, as contrasted with an improvement in memory after chronic administration of clovoxamine. The memory impairment in the amitriptyline group and improvement in the clovoxamine group were the result of changes in sensitivity [P(A)]. No changes in response bias (B) were detected. Conventional memory tests failed to detect drug-related differences in memory between the two groups. On the Benton, errors decreased over time within both drug treatment groups, whereas correct reproductions increased within the amitriptyline group only. However, between-group differences on the Benton did not reach significance. Results from the signal detection task suggest an amitriptyline-associated memory impairment. However, this interpretation is tempered by the finding that conventional memory measures failed to detect differences in memory performance between the two groups. We discuss the limitations of traditional memory measures and the utility of a signal detection approach in studies of psychopharmacologic influences on memory.