This paper establishes that the relative proportion of amobarbital metabolites in urine is highly variable from person to person and that observations of plasma half-life give no indication of this variability, but it shows that a valid estimate of a given person's metabolite pattern can be obtained by studying a single urine specimen in the postdistributive phase. The two metabolites which were measured in urine accounted on the average of 9 subjects for 80% +/- 3% of the dose with a range from 66% to 94%. The two metabolites were the well known 3'-hydroxyamobarbital (COH) as a product of side chain hydroxylation and N-beta-D-glucopyranosyl amobarbital (N-glu), a glucose conjugate which at some earlier time had been mistaken for an N-hydroxylation product. Among 129 volunteer subjects, the metabolite ratio N-glu/COH showed a median value of about 0.5 with a range from 0 to 2.8. A virtual absence of N-glu was observed in one of the 129 subjects and confirmed by a second administration of amobarbital 3 mo later. Of the 14 subjects with predominant N-glu excretion 4 were of Chinese origin, while there were 6 Chinese among the 115 other subjects (p less than 0.02).