This study examined the within-subject variability of urinary cotinine levels in young children (aged = 0.6-7.2 years) of smoking parents to determine the number of urine samples needed to provide accurate estimates of exposure to environmental tobacco smoke (ETS) for different time intervals. Secondary analyses were conducted of five independent studies (N = 376), in which multiple urinary cotinine measures had been collected over time periods up to 13 months. Over measurement periods of 4-15 days, the within-subject cotinine levels varied 3-5 times more than would be expected based on measurement error alone. Over 7-13 months, the within-subject variability was 10-20 times higher than would be expected based on the measurement error. Findings indicated that cotinine measures from single urine samples provided highly accurate estimates of only recent exposure (i.e., 2-3 days; rho = 0.99). To achieve similarly precise estimates of the mean cotinine level of an individual child over 4-15 days, up to nine urine samples may be necessary. Up to 12 urine samples may be required to achieve similarly precise estimates of ETS exposure over a 4- to 13-month period. Epidemiologic and clinical research on ETS exposure in children can benefit from multiple urine samples (a) to accurately measure average exposure at the level of the individual child, (b) to describe temporal patterns, (c) to detect incidences of peak exposure that would remain underrecognized if monitoring is limited to a single time point, and (d) to establish stable baseline levels and endpoints based on urine samples collected over clinically relevant time periods.