Objectives: to describe stability and change in levels of customary physical activity assessed in recall-based questionnaire surveys of older people conducted in 1985, 1989 and 1993.

Design: longitudinal study.

Subjects: 1042 people originally aged 65 and over randomly sampled from general practitioner lists in Nottingham, UK.

Methods: logistic and multiple regression analyses, intraclass correlation coefficients.

Main outcome measures: self-reported time spent per day walking and shopping; self-reported time spent per week in other indoor, outdoor and leisure activities; frequency of performance of strength and flexibility activities.

Results: among survivors, activity levels at baseline tended to be higher than those of their non-surviving peers. Overall, 8-year change between 1985 and 1993 was characterized by progressively declining activity levels. Nevertheless, in both trajectories and stability profiles, differences did emerge among the seven activity categories studied. At least one in four respondents increased the time they spent walking, and approximately one in three respondents increased the time they spent shopping between 1985 and 1993.

Conclusions: these findings suggest that, while some activity variables show levels of stability consistent with trait-like constructs, others are clearly more labile. While the present data cannot offer a definitive explanation for these differences, it seems reasonable that within each activity the influence of ability, opportunity and need interact to determine levels of participation.