The beta-endorphin (BE) hypothesis of premenstrual symptomatology was assessed in 15 healthy, drug-free normally cycling women. Cycle-phase assignment was aided by measurement of basal body temperature and plasma progesterone. For the entire group, peripheral plasma BE concentrations by radioimmunoassay (with substantial beta-lipotropin (BL) cross-reactivity) were generally unvarying across the menstrual cycle. High and low symptom subgroups were defined by retrospective and prospective daily self-reports of premenstrual changes. For the cycle studied, moderate to extreme symptom severity was defined according to premenstrual increases in impaired concentration, water retention and negative affect. No between-groups difference was observed for any cycle-phase for absolute values of beta-endorphin (BE/BL). When difference scores were examined, there was a positive time-lag correlation between follicular-to-periovulatory changes in BE/BL and the subsequent luteal increase in impaired concentration over the periovulatory baseline (rs = 0.82); considering the number of comparisons, however, the correlation fails to reach significance. The salience of various types of stressors on mood was assessed; there was a trend for "task-demands" to account for a greater percentage of negative attributions premenstrually compared to post-menstrually, for the entire group. The relative inconsistency in plasma BE/BL from cycle to cycle may be related in part, to the lack of stability in certain cycle characteristics from month to month. Future studies of BE/BL across the menstrual cycle may benefit from a demanding task in a provocative stress-paradigm.