The efficiency and accuracy of estrus detection using HeatWatch (DDx Inc., Denver, CO) or visual observation were compared in an autumn-calving Friesian herd (n = 48 per group) and a spring-calving Jersey herd (n = 50 per group) grazing on pasture. Cows in the group monitored by the HeatWatch system were fitted with a pressure-sensitive transmitter that signaled mounting activities associated with estrus. Visual observation was carried out for about 20 min before the morning and afternoon milkings and was aided by a strip of paint applied over the tailhead. Ovarian cyclicity was monitored with progesterone concentrations in milk samples collected twice a week. The efficiency and accuracy of estrus detection were, respectively, 98.4 and 97.6% for visual observation and 91.7 and 100% for HeatWatch detection. Autumn-calving herds differed from spring-calving herds in duration of estrus (9.7 vs. 7.3 h), number of mounts (13.6 vs. 8.5), total duration of mounts (36.8 vs. 19.9 s), and mean duration of a mount (2.6 vs. 2.3 s). There was no significant variation in the distribution of the time of onset of estrus or mounting activities at different hours of the day. Conception rate was similar for AI after estrus detection with HeatWatch (65.8%) or after visual observation (65.0%). The highest conception rate was obtained when AI was carried out between 12 and 18 h after the first mount. Both the HeatWatch system and visual observation plus tail painting can be used for estrus detection of dairy cows on pasture.