Background: Phenytoin was the first non-sedative anticonvulsant introduced and is still the anticonvulsant most widely used worldwide in neurology. Given the efficacy of the anticonvulsant lamotrigine in the depressed phase of bipolar disorder, a critical theoretical question is whether other anticonvulsants used in treating bipolar disorder might be similarly effective. We therefore undertook a controlled trial of phenytoin versus fluoxetine in major depressive disorder. Method: Data were collected from July 2001 to July 2003. Thirty-three subjects entered the study. All patients met DSM-IV criteria for major depressive disorder and scored a minimum of 18 on the 24-item Hamilton Rating Scale for Depression (HAM-D) at baseline. After a 3-day washout of any previous medications, patients were randomly assigned to fluoxetine or phenytoin in identical capsules. Each capsule contained phenytoin 100 mg or fluoxetine 7 mg plus cornstarch. Patients started with 1 tablet daily and increased every other day until they were taking 1 tablet 3 times daily with meals. Blood phenytoin levels were taken after 1 week, 3 weeks, and 6 weeks, and dosage was adjusted to achieve blood levels of 10 to 20 μg/mL, to a maximum dose of 4 capsules per day or a minimum dose of 2 capsules per day. Fluoxetine patients were assigned dummy blood phenytoin levels by the control psychiatrist such that the treating physician would raise the number of capsules to at least 3 per day (20 mg of fluoxetine). Results: Thirty-three patients entered the study, and 28 (N = 14 in each treatment group) completed at least 3 weeks and were included in the data analysis. Patients who dropped out after week 3 (3 patients) were included in the study as last value carried forward. There was no difference between treatment groups in overall rate of response or speed of response. Conclusion: The absence of a placebo arm in our study allows for the possibility that neither treatment was more effective than placebo. However, the exclusion of past fluoxetine nonresponders and the minimum HAM-D score at baseline of 18 make this possibility unlikely.