The accuracy and precision of eight techniques used to measure a range of eight antiepileptic drugs in human serum were compared using data from 80 samples from the Heathcontrol External Quality Assurance scheme. The fluorescence polarization immunoassay was significantly more precise than other techniques for several analytes, producing the lowest number of measurements rejected as outliers and measurements with the lowest coefficient of variation. Other techniques had a significantly lower precision. Nephelometry (Neph) produced most outliers for phenytoin and carbamazepine and had the highest variability for phenytoin. Gas-liquid chromatography (GLC) with derivatization was most variable for phenobarbitone and primidone. Measurements by GLC with or without derivatization contained >10% of outliers and were least precise for carbamazepine. Differences between the majority of techniques were in many cases, however, not significant. The accuracy of techniques was assessed from differences between sample means and spiked drug concentrations. Neph was notable in producing overestimates at lower concentrations. Immunoassay methods had a 16–21% cross-reactivity with carbamazepine 10,11-epoxide when measuring carbamazepine. All techniques reported underestimates for valproic acid that were thought to result from the hygroscopic nature of the salt used for spiking.