The strength of association between urinary cotinine and questionnaire data on passive smoking among 542 adolescents was evaluated. There were 103 individuals with urinary cotinine higher than 30 ng/ml; they were compared with all other subjects. The single variable that best predicted the urinary cotinine level was maternal smoking. A strong effect was made by house size and, consequently, house crowding. Maternal smoking and house crowding had a synergic effect. The subject's perception of passive smoking at home also was an independent indication of a high cotinine level. The final model had a fairly good sensitivity, whereas the specificity was somewhat lower. The results suggest that maternal smoking, house crowding, and subject's perception of a smoky environment could be surrogate indices of high passive smoking exposure in this age group and could be useful for epidemiologic studies.