Brain biopsy may be performed to make a definitive diagnosis in patients with rapidly progressive dementia. To assess the value of this procedure, we previously studied 90 consecutive cerebral biopsies performed in the tertiary referral centre of the National Hospital for Neurology and Neurosurgery, Queen Square between 1989 and 2003 (6 biopsies/year). Fifty-seven percent of all biopsies were diagnostic with Alzheimer's disease (18%), Creutzfeldt-Jakob disease (CJD) (12%) and inflammatory disorders (9%) being the most frequent. In the non-diagnostic group and for the series as a whole non-specific gliosis was the commonest diagnosis (37%). Treatment was altered because of information obtained from neuropathological findings in 11% of cases. To identify changes in practice that may have occurred due to recent advances in clinical assessment and improved histopathological techniques, we performed a follow-up study of 19 brain biopsies (approximately 3 cases/year) carried out for a dementing illness in the same centre between 2004 and 2009. These data suggest that brain biopsy may be less frequently used to help clinical diagnosis whilst its diagnostic yield increased from 57 to 74%. The commonest diagnosis was CJD, mostly suspected during life. Amongst the diagnoses, there were two cases of vasculitis and two cases of primary neurodegenerative dementia. These data suggest that improved clinical selection criteria supported by advances in diagnostic testing may result in brain biopsy being less frequently required, although it may still provide useful diagnostic information in difficult cases. We propose algorithms to aid the clinician in selecting appropriate patients for a biopsy and the neuropathologist in assessing a biopsy specimen.