We have studied the completeness of urine collections in 11 European centres. The completeness of collection was examined by questioning the participants, by calculating the ratio of observed to expected creatinine, and by measuring the recovery of p-aminobenzoic acid (PABA) in the urine after administration of a 240 mg dose. The ratio of observed to expected creatinine is a fairly insensitive measure of undercollection. People who report that their collection is incomplete are likely to have collected incompletely to a considerable degree. It was concluded that the use of PABA in epidemiological studies is still questionable; overcollection cannot be detected by using PABA, and it appeared that people sometimes forget or refuse to take the capsules. It is also suggested that differences in the meal-time patterns between countries may interfere with the PABA recovery test.