Few studies have quantitatively examined Mallard behaviour in North America during the breeding season. We estimated diurnal time budgets of unmarked Mallard males, females, and broods from over 1,200 hours of observations at two study areas in western Minnesota and south-central North Dakota during 1988-91. Paired males spent less time feeding and more time alert than did females. Both pair members were engaged in the same behaviour about 67% of the time; the male was always most likely to be doing the same thing as the female, but when the male was resting on water or alert, the female was most likely to be feeding. Females with broods spent less time feeding and more time alert and in locomotion than did females without broods. Behaviour of brood females did not differ with brood age or size. Females temporarily left their broods alone 45 times - about once for each 11 hours of observation. Female absences ranged from 2 to >80 minutes (x>27 min); length of absence was not related to brood age or size. Broods of all ages (a few days old to near fledging) and sizes (1-10 ducklings) were left alone on land and water; broods mostly rested and fed during female absences. Brood females spent less time feeding and more time alert than did broods. Females and their broods were engaged in the same behaviour 6267% of the time; the female was always most likely to be doing the same behaviour as her brood, but when the female was resting on water, the brood was most likely to be feeding, and when the female was alert, the brood was most likely to be feeding (North Dakota site) or resting on land (Minnesota site). Daily activity patterns varied between sites for both pairs and broods. Feeding and resting behaviour showed opposite daily patterns, suggesting that time allocated to feeding constrained time spent resting. Differences between sites and years in time spent feeding by pairs and broods probably reflected varying water conditions and food availability. In light of these differences, we caution researchers against extrapolating from time budget data that are temporally or geographically limited.