Within any biological population there is considerable variation in the physical characteristics of individual members, and the understanding and classification of such populations always depends on the interpretation of this variation. A major point of this paper is that groups of neurones can also be regarded as biological populations, and that at least three distinct types of variation can be found within any neural population: role-indicating variation, which enables different cells or groups of cells to perform different functions; systematic variation, which allows different cells (or sometimes the same cell) to perform a particular function under varying conditions; and residual variation, which is principally related to mechanisms of evolution and provides the population with its biological adaptability. Examples of these three types of variation are suggested for a number of properties of retinal ganglion cell populations.A second major point is that any functional classification of nerve cells should contain multiple taxonomic levels, corresponding to different levels of complexity and interaction within the nervous system. Thus, individual cells can belong to more than one group, each at a different taxonomic level, and these groups of cells can be viewed as interacting with each other rather than as operating in isolation. A multiple-level classification of cat retinal ganglion cells is presented with two broad groups, each subdivided into two lower-level groups, and an attempt is made to identify the categories of visual function to which these groups are related.