Past studies of bird migration times have shown great variation in migratory responses to climate change. We used 33 years of bird capture data (1970–2002) from Manomet, Massachusetts to examine variation in spring migration times for 32 species of North American passerines. We found that changes in first arrival dates – the unit of observation used in most studies of bird migration times – often differ dramatically from changes in the mean arrival date of the migration cohort as a whole. In our study, the earliest recorded springtime arrival date for each species occurred 0.20 days later each decade. In contrast, the mean arrival dates for birds of each species occurred 0.78 days earlier each decade. The difference in the two trends was largely explained by declining migration cohort sizes, a factor not examined in many previous studies. We found that changes in migration cohort or population sizes may account for a substantial amount of the variation in previously documented changes in migration times. After controlling for changes in migration cohort size, we found that climate variables, migration distance, and date of migration explained portions of the variation in migratory changes over time. In particular, short-distance migrants appeared to respond to changes in temperature, while mid-distance migrants responded particularly strongly to changes in the Southern Oscillation Index. The migration times of long-distance migrants tended not to change over time. Our findings suggest that previously reported changes in migration times may need to be reinterpreted to incorporate changes in migration cohort sizes.