Exposure to environmental tobacco smoke is an important variable in many pediatric epidemiologic studies. We measured urinary cotinine, a specific metabolite of nicotine, in a population-based cohort of children every other month from birth through two years of age. Extensive data regarding exposure to smokers (people in the home, in home and away from home day care, home visitors, visits to smokers) were collected monthly by way of home interviews. We evaluated, with multiple cotinine measurements as the gold standard, other measures of exposure that are more feasible to obtain in large-scale studies. Comparing one cotinine to the average of multiple measurements, we found that 33.7% were in error in excess of 100 ng/mg, but 84% of the infants could be correctly classified into categories of low versus high. Parental smoking patterns had the highest predictive accuracy (fathers 67.0% and mothers 64.1%). Combining selected smoker categories (either parent, other home residents outside day care workers) resulted in improved accuracy of 79.3%.