Indoor volatile organic compounds (VOCs) have been associated with asthma, but there is little epidemiologic work on ambient exposures, and no data on relationships between respiratory health and exhaled breath VOCs, which is a biomarker of VOC exposure. We recruited 26 Hispanic children with mild asthma in a Los Angeles community with high VOC levels near major freeways and trucking routes. Two dropped out, three had invalid peak expiratory flow (PEF) or breath VOC data, leaving 21. Children filled out symptom diaries and performed PEF maneuvers daily, November 1999-January 2000. We aimed to collect breath VOC samples on asthma episode and baseline symptom-free days, but six subjects only gave samples on symptom-free days. We analyzed 106 breath samples by GC-MS. Eight VOCs were quantifiable in >75% of breath samples (benzene, methylene chloride, styrene, tetrachloroethylene, toluene, m,p-xylene, o-xylene, and p-dichlorobenzene). Generalized estimating equation and mixed linear regression models for VOC exposure-response relationships controlled for temperature and respiratory infections. We found marginally positive associations between bothersome or more severe asthma symptoms and same day breath concentrations of benzene [odds ratio (OR) 2.03, 95% confidence interval (CI) 0.80, 5.11] but not other breath VOCs. Ambient petroleum-related VOCs measured on the same person-days as breath VOCs showed notably stronger associations with symptoms, including toluene, m,p-xylene, o-xylene, and benzene (OR 5.93, 95% CI 1.64, 21.4). On breath sample days, symptoms were also associated with 1-h ambient NO(2), OR 8.13 (1.52, 43.4), and SO(2), OR 2.36 (1.16, 4.81). Consistent inverse relationships were found between evening PEF and the same ambient VOCs, NO(2), and SO(2). There were no associations with O(3). Given the high traffic density of the region, stronger associations for ambient than for breath VOCs suggest that ambient VOC measurements were better markers for daily exposure to combustion-related compounds thought to be causally related to acute asthma. Alternatively, the low sample size of symptom responses (15-21 responses per 108 breath samples) may have led to the nonsignificant results for breath VOCs.