Preliminary evidence for the effectiveness of cognitive rehabilitation interventions based on errorless learning principles in early-stage Alzheimer's disease (AD) was provided by Clare et al. (1999, 2000, 2001). The present study extends these findings in a controlled trial. Twelve participants meeting criteria for probable AD, with Mini-Mental State Examination scores of 18 or above, were trained in face-name associations using an errorless learning paradigm. Training produced a significant group improvement in recall of trained, but not control, items. Gains were largely maintained 6 months later, in the absence of practice. There were differences in individual response to intervention. Results did not differ according to medication status, and the intervention had no adverse effects on self-reported well-being, but participants who were more aware of their memory difficulties achieved better outcomes.