Studies of pituitary-gonadal function in men with erectile disorders have provided conflicting findings. This study compares blood LH and testosterone during sleep in 17 physically healthy men with erectile impotence and 12 matched normal controls, and relates hormonal variations to stages of sleep and penile tumescence. Subjects, aged 23-36 were studied in a sleep laboratory for 3-6 nights with the last two nights devoted to sequential blood sampling every 20 minutes. Five men had never been able to achieve intercourse (primary impotent group) and 12 suffered from a life-long history of intermittent erectile failures (secondary-impotent group). There were no significant differences in sleep duration and REM time among the impotent groups and normal controls. Primary-impotent men showed half as many full tumescent episodes as secondary-impotent men and controls, and spent significantly less time above 80% of full tumescence. The secondary-impotent group did not differ in nocturnal penile tumescent measures from controls. There was a pattern of irregularly occurring fluctuations in plasma LH and testosterone with no differences among groups in frequency and amount of peak hormonal increases. Normal subjects had significantly higher mean testosterone concentrations during REM sleep than during other sleep stages, and during full tumescence than during partial and nontumescent periods. A direct causal relation between REM-related activity and circulating testosterone was not supported by the observation that the hormonal levels during REM sleep and during full tumescence did not differ statistically from the levels measured during adjacent time periods. As with normal subjects, the secondary-impotent group exhibited higher testosterone levels during REM sleep and full tumescence; by contrast, the primary-impotent group did not show significant hormonal differences across stages of sleep and tumescent/ nontumescent periods. These data are discussed with respect to the possible existence of subgroups of impotent men without known organic pathology than may be characterized by psychophysiologic and endocrine differences during sleep.