Using self-report to assess everyday memory in alcoholics presents challenges given the presence of both memory and metamemory deficits. Accordingly, evaluation of the reliability and validity of proxy ratings as well as the frequency of these memory lapses are of clinical importance. In the present study, 180 patient-proxy dyads completed the Prospective and Retrospective Memory Questionnaire (PRMQ). 31.7% of proxy-rated versus 2.8% of patient-rated prospective memory scores fell in the impaired to below average range. 15% of proxy-rated retrospective memory scores were below average, whereas none of the patients reported problems in this regard. Longer delays between intention formation and action yielded better prospective memory performance, while the opposite was true for retrospective memory. Agreement between patients and proxies was generally poor to fair across severity levels and the magnitude of observed differences was large (standardized response mean > 0.8). For all PRMQ items, exact agreement occurred in 45.3% of the cases. Larger patient-proxy discrepancy was associated with older age, less education and greater disease severity. Proxy ratings were internally consistent, significantly correlated with objective memory performance, and were sensitive to differences in overall PRMQ performance between severity groups. Caution should be used in the interpretations of patients' reports.