An efficient method is presented for obtaining, in under 4 h, an unbiased estimate of the total number of neurons in the human neocortex, with a coefficient of error on the estimate of approximately 5%. The novel sampling scheme used in this study is unbiased and was designed so that only a small amount of neocortical grey matter had to be removed. Hence, the majority of the cerebral grey matter and all the internal grey matter was left intact for further resampling and analysis. Each cerebral hemisphere was subdivided into the four major neocortical regions, sliced coronally at 7-mm intervals and the volume of the neocortex determined using Cavalieri's principle. Uniform sampling of neocortex was performed in the hemisphere followed by regional subsampling with a varying sampling fraction being taken from each region. Neuronal density estimates were made in thick plastic sections using optical disectors. Shrinkage estimates were made in parallel with the number estimates and found to be negligible. The total number of neocortical neurons in the right hemisphere of five normal 80-year-old men was found to be 13.7 x 10(9) with an inter-individual coefficient of variation of 12%.