Published values for the number of neurons in the superior cervical ganglion of the adult rat range from 13 000 to 45 000. These studies have employed different methods for determining what unit to count (cell body, nucleus, nucleolus), how many sections to count, and how to correct the raw counts for split particles and for profiles that are too small to resolve. The purpose of this study was to examine the extent to which these parameters may influence the calculated value for the total number of neurons, using computer simulations of neuron populations. These simulations permitted us to determine the effects on neuron number of varying the diameter of the neuronal nucleus, the size of the smallest resolvable profile, and the thickness of the section. The data from the simulations were used to test the validity of several methods that are in common use for correction of neuron counts. Our results indicate that most of the methods that are in routine use are unsatisfactory. We propose the use of either one of two methods that consistently result in highly accurate estimates of neuron numbers. These are: (1) a modification of the method proposed by Hendry (1976), using computer analysis; or (2) a modification of the method proposed by Abercrombie (1946), which does not require the use of a computer.