By use of an interview, return tablet count, and a pharmacologic indicator (low-dose phenobarbital), we compared compliance with tablets prescribed to be taken once, twice, or three times daily. One hundred seventy-nine patients with type II diabetes were randomly allocated to take one 2 mg phenobarbital tablet once, twice, or three times daily for 28 days. Phenobarbital level/dose ratios indicated that compliance was similar with once- and twice-daily regimens, and both were better than thrice-daily dosing. Mean return tablet counts suggested that compliance was best with the once-daily regimen; both twice- and thrice-daily regimens were similarly inferior. This difference between the techniques may be explained by the inadequacies of the residual tablet count, which identified only 13% of cases identified by phenobarbital. We conclude that compliance with the once-daily regimen was best, but that compliance with a twice-daily regimen was very similar, and both were superior to dosing three times a day.