We determined four carnitine constituents (total and free carnitine and short- and long-chain fatty acid carnitine esters) in serum from 471 patients treated for convulsions with phenobarbital, valproic acid, phenytoin, and/or carbamazepine. The 471 patients were in eight treatment groups; four were treated with monotherapy and four with polytherapy. The means of all four carnitine constituents were significantly reduced in all treatment groups (except for free carnitine in four groups). Total carnitine was reduced by 23% to 48%, free carnitine by 9% to 45%, short-chain fatty acid carnitine by 46% to 64%, and long-chain fatty acid carnitine by 6% to 29%. Patient frequency of reduction for total carnitine was 20% of all patients (10% for free carnitine), 23% of patients receiving valproate (9% for free carnitine), 36% of those receiving phenobarbital (21% for free carnitine), 12% of those receiving phenytoin (8% for free carnitine), and 8% of those receiving carbamazepine (1% for free carnitine). Only for phenobarbital was there an inverse correlation between the serum concentration of the drug and that of carnitine concentration. One patient receiving carbamazepine had a 59% reduction in the total and a 65% reduction in the free carnitine concentration and a fivefold increase in long-chain fatty acid carnitine, values similar to those seen in neonatal lethal carnitine palmitoyl transferase II deficiency. It remains to be determined whether a reduction in serum carnitine values in patients receiving anticonvulsant therapy is of clinical consequence, whether the reduction is present in some patients before the start of therapy, when and by what mechanism carnitine levels may become reduced during therapy, and whether the reduction exists in the solid tissues of these patients.