Fifty-four per cent of 41 chronically institutionalized adult patients with epilepsy had ataxia of gait (wide mean stride width). None of the following correlated with stride width: serum phenytoin, previous phenytoin toxicity, seizure frequency, or status epilepticus. Seventeen of the 41 patients had computed tomographic head scans. Patients with radiological evidence of cerebellar atrophy had a wider mean stride width, later age of onset of seizures, greater peak serum concentrations of phenytoin than did those without cerebellar atrophy. Ataxia of gait was inconsistently associated with cerebellar atrophy. Elevated serum/plasma concentrations of phenytoin may be a risk factor for cerebellar atrophy, but seizure frequency or status epilepticus are not independently related to this complication.