Frontotemporal dementia is a term used to characterize diverse neuropathological conditions that can present with the same clinical phenotype. Five different neuropathologies underlie this disorder. However, consistent frontal and/or temporal neuronal loss and gliosis characterize all cases, the majority having no obvious pathological inclusions. Because neuronal loss and gliosis are consistent features across all cases, the present study aimed to determine the relationship between neuronal loss, gliosis and, for cases with abnormal tau inclusions, intracellular tau deposition. Formalin-fixed brain specimens from sporadic cases with frontotemporal dementia (eight with tau-positive Pick bodies, five with frontotemporal lobar degeneration without inclusions) were compared with those from non-diseased controls (n = 5). Brain specimens were cut into 3 mm coronal slices for evaluation and tissue samples from the superior frontal gyrus were taken for microscopic analysis. Immuno histochemistry for glia-specific proteins (astrocytic glial fibrillary acidic protein and microglial major histocompatibility complex II) and different tau epitopes was performed on 50 microm free-floating sections. Gross patterns of brain atrophy were analysed and upper and lower layer pyramidal neurons and glial cell numbers were quantified. A disease severity scheme was devised using the degree of gross macroscopic frontal and temporal atrophy to establish the relationship between the gliosis and neurodegeneration. In this small sample, the patterns of gross atrophy could be grouped reliably into four stages of severity. These stages were the same across disease groups and correlated with volume- corrected pyramidal neuron densities. In cases with Pick bodies, disease stage also correlated with duration, providing further evidence that these stages represent the progression of degeneration in this limited sample. Whereas there were, on average, many more reactive astrocytes in the cases with Pick bodies than in those with frontotemporal lobar atrophy, there was significant overlap between cases in the degree of astrocytosis. However, a large proportion of the astrocytes in Pick's disease displayed phosphorylated tau immunoreactivity, whereas no tau-positive astrocytes were found in frontotemporal lobar degeneration. The pattern and degree of microglia activation were similar in all the dementia cases analysed, with considerably more activated microglia accumulating in white matter. In this small sample, the abundance of white matter microglia at early disease stages suggests a prominent role for this cell type in the neurodegenerative process. In frontotemporal lobar degeneration, a significant proportion of the activated white matter microglia were tau-2-immunoreactive, suggesting direct involvement in axonal degeneration, possibly via immune processes.