Longitudinal information on human activity patterns is required to perform reliable assessments of chronic exposure to pesticides in nonoccupational settings. As part of a longitudinal, multimedia exposure survey in Maryland, USA, we collected 2521 person-days of information on the frequency of five activities that are potentially important to pesticide exposure in residential settings: preparation of pesticides; application of pesticides; contact with soil on the skin; contact with grass on the skin; and contact with carpet. Logistic regression was used to evaluate the data for variability by time of year, day of week, and several demographic factors. Eight percent of the population reported preparing pesticides on at least one occasion over the year, while 24% of the population applied pesticides at least once. The majority of the population reported contact with soil (68%), grass (72%), and carpet (64%) over the course of a year. Skin contact with soil and grass was found to occur approximately twice as frequently (p<0.001) in spring and summer periods than in fall and winter periods. Similarly, contact with soil and grass was about 50% (p<0.05) more likely to occur on weekend days than on weekdays. These results indicate that short-term measurements (e.g., 1-day or 1-week) of these activities for an individual may not be equivalent to long-term average patterns. We also found that selected demographic characteristics were significant predictors of several of these activities. Thus, some demographic characteristics may be useful in categorizing exposure although it is unlikely that a single short-term measurement will suffice to describe annual average activity or multiple short-term activity periods. Journal of Exposure Analysis and Environmental Epidemiology (2001) 11, 389–397. [ABSTRACT FROM AUTHOR]