Collecting uncontaminated urine specimens from infants is difficult. Commonly, an adhesive urinecollecting bag is used, which is uncomfortable. This study determined bacterial contamination rates using three methods of urine collection sequentially on the same infant (without known urinary tract infection)-clean-catch, cotton wool (sanitary) pad and urine bag. The study was undertaken in children under 3 years of age in the Institute of Maternal and Child Health of Pernambuco (IMIP), Recife, Brazil. Urine samples were analysed using phase contrast microscopy and routine culture. Culture of bacteria at any level was interpreted as a contaminated urine specimen. Cultures with > 10(5) colony-forming units/ml of one species by all three collection methods were regarded as true urinary tract infection and these children were excluded. Altogether, 534 urine samples from 191 patients were analysed. Median age was 2 months (1 day-36 months) and 124 (65%) were boys. Twelve children (6.3%) were considered to have true urinary tract infection, three were indeterminate and in 16 one or more samples were missing and all were excluded from analysis. There were more missing samples using the clean-catch method (12%) than when using the bag (4%) or pad (4%). Seventy-six of 160 (47.5%) children had evidence of bacterial contamination. Clean-catch specimens showed the least contamination (14.7%) and rates were similar between pads (29%) and bags (26.6%) (kappa = 0.40). Urine contamination rates were similar for sanitary pads and urine bags and significantly higher than for clean-catch (p<0.01). However, pads were a simple, non-invasive and comfortable alternative to bags.