Studies examining levels of illumination in adult populations have demonstrated that the level and amount of light exposure are lower in the elderly compared with younger adults, particularly in institutionalized patients with dementia. Although insufficient light exposure has been implied as a cause of sleep fragmentation, evidence for such a relationship is scant. Sixty-six institutionalized elderly had their activity and light exposure monitored for a 3-day period. Mean and median light levels, minutes spent over 1000 and over 2000 lux, percent sleep and wake, and number of naps were computed for daytime intervals, defined as 07.00-18.59. Percentages of sleep and wake, number of awakenings and mean duration of wake periods were computed for night-time intervals, defined as 22.00-05.59. Mesor, amplitude and acrophase of activity and of light were determined by cosinor analysis. A mixed linear model was used to assess the effects of daytime Actillume measures on subsequent night-time measures, and vice versa. Spearman correlations were computed, and multiple regression analyses were carried out with light variables and dementia level as predictors and sleep-wake and activity measures as dependent variables. The median light level was 54 lux and a median of only 10.5 min were spent over 1000 lux. Higher light levels predicted fewer night-time awakenings, and severe dementia predicted more daytime sleep and lower mean activity. Increased bright light exposure predicted later activity acrophase. There was an association between the acrophases of light and of activity, with maximum illumination preceding peak activity. These results suggest that daytime light exposure has an impact on both night-time sleep consolidation and timing of peak activity level.