1. The seasonal variations in time of daily onset and end of locomotor activity are described for 3 species of mammals and 5 species of birds kept in captivity at the arctic circle and at lower latitude. These variations are most pronounced at high latitude. 2. The duration of daily activity plotted versus the photoperiod can be described as an S-curve in all species studied so far, both in nature and in captivity. In both male and female fringillid birds activity times were longer before the summer solstice (spring) than after the summer solstice at equal photoperiods. 3. The seasonal changes in activity time result from roughly mirror-image changes in the times of onset and end of activity relative to sunrise and sunset, cancelling out each other. Therefore the midpoint of activity stays relatively stable; remaining minor changes in the midpoint of activity do not produce a general seasonal pattern. 4. At high latitude, a large seasonal fluctuation in the day-to-day variability (or precision) of activity timing is detected. These patterns of precision of the rhythm cannot be attributed to a single Zeitgeber property without complex assumptions. Onset and ends of activity become more precise when occurring during the civil twilight, i.e. at times of day with most rapid changes in light intensity. This may reflect direct action of light on the rhythm rather than a property of the entrainment mechanism. 5. The data do not give compelling evidence for any formal model of the oscillations driving the activity rhythms. Predictions concerning the relation between phase and activity time derived from a single oscillator model are not matched by the data. On the other hand, the general seasonal patterns can be easily described in terms of a two-oscillator model. 6. Seasonal variations in duration of activity are larger in birds than in mammals. Day-to-day variations in timing are larger in mammals than in birds. The implications for photoperiodic time measurement are discussed.