Serum phenytoin monitoring is frequently used in the management of epileptic patients because phenytoin has a narrow therapeutic index and exhibits nonlinear pharmacokinetics. This study prospectively evaluated serum phenytoin monitoring in an acute care teaching hospital. Two sets of criteria were established a priori to define (a) appropriate selection of patients with regard to serum phenytoin monitoring, and (b) inappropriate serum phenytoin determinations (SPDs). Eighty patients receiving phenytoin were studied, of whom 58 (72.5%) were appropriately selected. These included 35 patients (43.8%) for whom monitoring was indicated and was performed, and 23 patients (28.7%) for whom monitoring was not indicated and was not performed. There were 39 patients with no indications for serum phenytoin monitoring; however, 16 (41%) of them were monitored. A total of 113 SPDs were performed, of which 83 (73.5%) were deemed to be inappropriate. Seventy percent of SPDs resulted in phenytoin concentrations outside the usual therapeutic range (10–20 μg/ml). Overall, physicians appropriately selected patients with regard to serum phenytoin monitoring; however, when inappropriate selection did occur, it tended to involve monitoring of patients who did not require it. The majority of SPDs performed were deemed to be inappropriate, since they were done too soon after admission or a change in therapy to reliably indicate steady-state serum concentrations.