Trust, but verify: the necessity of empirical verification in quantitative neurobiology.

The preferred procedure for estimating neuron number has been the subject of intense discussion. Sampling error, systematic bias, and the physical properties of the tissue examined all contribute to the possibility that estimates may not reflect actual neuron number. Even when restricting analysis to a particular, well-defined structure such as dorsal root ganglia, these factors act in a manner that varies with maturity of the animal, species, fixation, processing, and neuron size. These considerations reinforce the necessity of comparing estimates with the actual number of neurons as determined by three-dimensional reconstructions. Copyright 2002 Wiley-Liss, Inc.