We previously reported a case study of a man with right frontal lobe damage, BG, who showed extraordinarily high false alarm rates on remember-know recognition tests (Schacter, D. L. et al., Neuropsychologia, 1996, Vol. 34, pp. 793-808). Experiment 1 extends his high false alarm rate to yes-no recognition tests. BG typically gives false 'remember' responses on remember-know tests, and this pattern was uninfluenced when he was asked to explain the basis for his 'remember' responses (Experiments 2 and 3). When BG was given a semantic encoding task, he stopped giving 'remember'-based false alarms (Experiment 4). Signal detection analyses revealed that BG had a discrimination deficit and an abnormally liberal response bias (especially for 'remember' responses) in most conditions. Overall, BG's high false alarm rate is interpreted as reflecting an over-reliance on the general similarity between a test item and the study episode.