Error scores and response times from a computer-administered, forced-choice recognition test of symptom validity were evaluated for efficiency in detecting feigned memory deficits. Participants included controls (n = 95), experimental malingerers (n = 43), compensation-seeking patients (n = 206), and patients not seeking financial compensation (n = 32). Adopting a three-level cut-score system that classified participant performance as malingered, questionable, or valid greatly improved sensitivity with relatively little impact on specificity. For error scores, convergent validity was found to be adequate and divergent validity was found to be excellent. Although response times showed promise for assisting in the detection of feigned impairment, divergent and convergent validity were weaker, suggesting somewhat less utility than error scores.