We present findings from three waves of a population-representative, UK time-use diary survey conducted both pre- and in real time during full 'lockdown', and again following the easing of social restrictions. We used an innovative online diary instrument that has proved both reliable and quick-to-field. Combining diary information on activity, location, and co-presence to estimate infection risks associated with daily behavior, we show clear changes in risk-associated behavior between the pre, full-lockdown and post full-lockdown periods. We document a shift from more to less risky daily behavior patterns (combinations of activity/location/co-presence categories) between the pre-pandemic pattern and full lockdown in May/June 2020, followed by a reversion (although not a complete reversal) of those patterns in August 2020 following the end of the first lockdown. Because, in general, a populations' time use changes relatively slowly, the behavioral changes revealed may be interpreted as a consequence of the UK COVID-19 lockdown social restrictions and their subsequent relaxation.