The use of creatinine as a check on the completeness of 24-h urine collections is based on the assumption that excretion per kg body mass is constant. In a study of eight volunteers maintained on uniform diets the within-subject coefficient of variation (CV) in daily 24-h creatinine excretion was only 4 per cent under conditions in which the completeness of collections was assured. Cooked meat contains creatine and creatinine and an addition of 260 g/d cooked meat to the diet of four volunteers on a metabolic diet increased urine creatinine excretion by 23 per cent. A further source of variation in urine creatinine is the completeness of the collection itself. A new marker, PABA, was used to check the completeness of 24-h urine collections in eight subjects studied for 28 d whilst eating their normal diets. The within-subject CV of 9 per cent for 24-h creatinine excretion was not significantly different from that of 13 per cent for overall urinary N. In 122 complete 24-h urine specimens collected from a randomly selected population, the CV in creatinine excretion was 18 per cent. Twenty-one urine samples were incomplete, of which only 3 would have been detected on the basis of low creatinine by conventional standards. Of 39 hospital patients asked to collect 24-h urine samples, 12 provided incomplete samples, of which only 4 had low creatinine excretion. It is concluded that PABA is a more sensitive and reliable verification of the completeness of 24-h urine collections than creatinine.