Subjective and objective measures of poor sleep in alcoholic insomniacs predict relapse to drinking. Nonalcoholic insomniacs underestimate their total sleep time (TST) and overestimate their sleep onset latency (SOL) and wake time after sleep onset (WASO) compared with polysomnography (PSG). This study evaluated 3 hypotheses: (1) subjective SOL would predict frequency of future drinking; (2) participants would overestimate SOL and WASO and underestimate TST; and (3) higher amounts of over- and underestimates of sleep at baseline would predict worse drinking outcomes prospectively.Participants (N=18), mean age 44.6 years (+/-13.2), underwent an adaptation night and then 2 nights of PSG 3 weeks apart. They also provided morning estimates of SOL, WASO, TST, and sleep efficiency (SE). Following the baseline PSG, participants were followed over 12 weeks. A 2-way ANOVA (night x method of measuring sleep) compared results and regression analyses predicted drinking. Drinking outcomes were defined as number of days drinking (DD) and number of heavy-drinking days (HDD) during 2 consecutive 6-week follow-up periods.Most participants (72%) overestimated SOL by a mean of 21.3 (+/-36) minutes compared with PSG [F(1, 14)=7.1, p<0.03]. Unexpectedly, 89% underestimated WASO by a mean difference of 48.7 (+/-49) minutes [F(1, 14)=15.6, p<0.01]. Drinking during the first 6-week study period was predicted by both subjective estimates of WASO and their accuracy, whereas drinking during the second 6-week period was predicted by both subjective estimations of sleep and rapid eye movement sleep latency.Greater subjective accuracy of wakefulness at night provided by the patient predicted drinking during the study. Unlike nonalcoholic insomniacs, this alcoholic sample significantly underestimated WASO compared with PSG values. The predictive ability of sleep parameters depended on the selected measure of drinking outcomes and when outcomes were measured. Subjective sleep measures were better predictors of future drinking than corresponding PSG measures.