Non-compliance in epileptics is widespread problem and significantly interferes with optimal therapy of patients. In this study, estent of compliance and various factors influencing compliance were assessed. Epileptic outpatients at Hanyang University Hospital, aged over 15 years, who have been taking medication regularly at least for 6 months without any change of drug regimen at least two weeks prior to this study, entered the study. Antiepileptic serum levles were measured three times by Fluorescence Polarization Immunoassay at intervals of least two to four weeks apart, and their coefficient of variation was claculated. Patients whose coefficient of variation was more than 30% was considered as a non-compliant group. Because non-compliance is very complex, and multidimensional problem and it can't be sufficiently evaluated only by serum droug level, various behavioral patterns disturbing treatment, evaluated by both questionnaire and chart review, were used to define non-complince, too. One hundred sixty seven outpatients were entered this study. Eighty nine out of 167 patients were grouped into non-compliance(53%). Seven universal, five clinical, three pharmacologic, and eight patient's factors were analysed. Sex, age at survey, age of onset, seizure pattern, seizure free interval, seizure frequency, side effect of drug, and worry about health were the statistically significant factors. Minnesota Multiphasic personality Inventory (MMP) of 383 questions was performed in 60 and clinical scales were compared between non-compliance. In conclusion, it seems to be certain factors which must be taken into consideration in treating epilesy, such as sex, age at survey, age of onser, seizure pattern, seizure free interval, seizure frequency, side effect of durg, and worry about health which was founded to be statistically significant in this study.