Analysis of response consistency on neuropsychological test performance, both within and across testing sessions, can be an important method of detecting malingering. Little systematic research, however, has examined how suspected malingerers perform across repeat evaluations, a common forensic occurrence. To address this issue, we examined performance across a 3-week interval in an analogue malingering design on the California Verbal Learning Test (CVLT), the Rey Complex Figure, the Controlled Oral Word Association Test, and the Ruff Figural Fluency Test. Malingering simulators (n = 21) performed more poorly on all measures than the controls (n = 21) and demonstrated practice effects on the nonverbal, but not the verbal, tests. Controls demonstrated practice effects on all measures across time. Contrary to hypotheses, malingering simulators demonstrated high and similar levels of between and within time consistency as controls when assessed via a series of correlations. Despite this consistency, when qualitative performance patterns were assessed on the CVLT, simulators were less likely to consistently recall the same word across successive learning trials. The following issues are discussed: (a) the differential pattern of practice effects on verbal and nonverbal tasks, (b) qualitative and quantitative differences in assessment of consistency, and (c) how future research should study consistency/inconsistency.