The aim of this study was to test if the pattern of human mood variation across the day is consistent with the hypothesis that self-reports of positive affect (PA) have a circadian component, and self-reports of negative affect (NA) do not. Data were collected under two protocols: normal ambulatory conditions of activity and rest and during a 27 h constant routine (CR) procedure. Mood data were collected every 3 h during the wake span of the ambulatory protocol and hourly during the 27 h CR. In both protocols, rectal temperature data were continuously recorded. In the ambulatory protocol, activity data were also collected to enable estimation of the unmasked (purified) temperature rhythm. Participants were 14 healthy females aged 18-25 yr in the follicular phase of the menstrual cycle. Under both protocols, PA exhibited significant 24 h temporal variation [CR: F(23, 161) = 2.12, p < 0.01; ambulatory: F(5,55) = 2.44, p < 0.05] with a significant sinusoidal component [CR: F(2, 21) = 7.51, p < 0.01; ambulatory: F(2,3) = 20.49, p < .05] of the same form as the circadian temperature rhythm. In contrast, NA exhibited an increasing linear trend over time under the ambulatory protocol [F(1, 11) = 5.74, p < 0.05] but nonsignificant temporal variation under the CR protocol. The findings support the hypothesis of a circadian component in PA variation.