The aim of this study was to establish methodological variability in the estimation of the total number of neurones using the optical disector. Variations in the 3 dimensions of the disector probe were analysed under uniform sampling conditions in 50-microns-thick frozen sections of the human mediodorsal thalamic nucleus. There was no significant difference between the estimated neuronal number using samples of variable height (fractionator vs. non-fractionator sampling). In addition, different methods of volume calculation (individual sample vs. an average) did not significantly change the estimated total neuronal number. Large variations in the estimated total neuronal number occurred when the x and y dimensions of the disector probe were altered. In this study, accurate and reproducible estimates were achieved when the disector probe was large enough to have a probability of sampling at least 2 cells per frame. We conclude that the variables in the x-y plane (the disector frame size as well as the sample interval) significantly contribute to differences in the estimated total neuronal number. Several practical measurements to estimate this probability and enhance experimental design are discussed.