The long-term outcome of chronic epilepsy remains largely unknown, despite a long historical experience. We report the lifelong course of epilepsy of an historical cohort of 235 subjects who were in residential care at the Chalfont Centre for Epilepsy: 122 had comprehensive post-mortem examination. The populations admitted as resident to the centre over time followed the evolution of society's perception of epilepsy. 'Early residents' (before 1972) were admitted for sheltered employment, escaping stigmatization, whereas 'later' residents with more severe epilepsies were admitted for care. Subjects admitted before 1972 were similar to subjects followed nowadays as outpatients, whereas patients admitted later with a higher burden of disabilities are often those in residential care. This long follow-up allowed exploration of a wide spectrum of epilepsies, affecting both subjects who were otherwise healthy and those with co-morbidities. Age at death showed a bimodal distribution with an early peak of mortality between 45-50 years old, whilst the remainder had life expectancy comparable to the general population. As a group, subjects who had post-mortem examination were not significantly different from patients who did not have post-mortem examination, but post-mortem examination provided data that were otherwise unavailable. For those who had post-mortem examination, sudden unexpected death in epilepsy (SUDEP, 18% of all deaths) did not fully explain the early mortality, to which co-morbidities contributed. High seizure frequency was a significant independent predictor of early death even after excluding SUDEP (e.g. reduction in years of life for those who had >4 seizures/month compared with those who had <1 seizure/month: 13 years; 95% confidence interval: 6-19; overall P = 0.0006). Those who survived to older age increasingly went into spontaneous remission lasting until death (in the whole cohort, 38/166, 23% of those who died in or after sixth decade). In subjects who had post-mortem examination, older age (odds ratio = 1.13; 95% confidence interval: 1.06-1.20) and presence of neuropathologically confirmed degenerative changes (that were not the cause of epilepsy) (odds ratio 7.14; 1.95-26.2) were independent predictors of terminal remission. Epilepsy may cause premature death indirectly through co-morbid conditions. Terminal remission occurs even without prior remissions; ageing may improve epilepsy drug responsiveness although unknown factors related to the natural history may also play a role.