Principles of classification and nomenclature relevant to studies of nociceptive neurones
Many classifications of neurones will be used in this book. How are we to evaluate them? Why do different authors sometimes use apparently quite dissimilar classifications? This brief article offers some views on these and related matters. The emphasis is on basic ideas and specific citations are not given. A list of sources is given at the end.
Let us first consider why making classifications is so useful. Principally, classification gives us a way of handling the number problem. In simple organisms such as Caenorhabditis elegans or the leech there are only small numbers of neurones and it is possible to treat each one as an individual. In the vertebrates, however, this is not possible and we have to deal with populations of neurones. Clearly, finding the appropriate functional groupings is crucial to developing sensible hypotheses about how the system operates. This is really the key to good classifications. They lead to new insights into how the system functions and provide an efficient terminology for describing system function. For example, the discovery of the X, Y, and W classes of retinal ganglion cells led to major developments in our understanding of visual processing, particularly the importance of parallel processing by functionally distinct groups of neurones. A good classification should also lead to insights into how adult neurones develop. Consider the parallel with taxonomy. The classification of plants and animals suggested interrelationships between groups and stimulated thinking about how such groups were formed, eventually playing a major part in the development of the theory of evolution.
How can we tell a good classification from a bad one? A good classification will utilize many variables, probably including both functional and morphological measures. Classifications depending on just one or two ‘essential’ features are liable to be superseded if just one significant new factor is discovered. Numerical methods can help in assessing multiple variables. Once there are more than four variables to consider, visualizing the n-dimensional space in which the data needs to be placed becomes a serious challenge to the imagination! But be warned, numerical methods such as the various forms of cluster analysis do not in themselves tell you how many groups are present. They can only indicate the probable groupings and should be used as part of the hypothesis-building process. Obvious limitations of numerical methods are how they can handle different tyes of variable (nominal, numerical, etc.), whether all variables should be weighted equally, and how to handle conditional variables (for example, response to hair movement is only a variable for units from hairy skin, so a study covering hairy and glabrous skin has to deal with the fact that this parameter is only present for a subset of the sample).
A not uncommon problem occurs with classifications that are really only subdivisions. In science, as in other activities there are lumpers’ and ‘splitters’. A group of neurones (p.2) with a wide range of properties will commonly be split into subgroups, without any consideration of whether the subdivisions represent separate classes. To be a class, a group of neurones must show a distinct set of properties not shared by other members of the sample. It may occasionally be useful to subdivide a population into arbitrary groups in this way, but this process is not classification, merely dissection.
The importance of basing classifications on several variables has been stressed, but in real life the identification of the class of a neurone needs to be based on a small number of easily measured variables. It is important to distinguish the process of identification from the process of classification. Identifiers will be a subset of the classifiers that give a high probability of correct classification, and are also easy to measure. They may not be particularly important parameters for the main function of the neurone. For example, conduction velocity is an important identifier for many afferent neurones, but is often of much less importance for their function than measures such as receptive field properties or thresholds for adequate stimulation.
A final important distinction to maintain is between nomenclature and classification. Choosing names for classes is good fun, but, compared with defining the classes themselves, is essentially trivial. As an example, consider the large population of Cfibre cutaneous afferents that respond to pressure, heat, and irritant chemicals. These are called either polymodal or mechanoheat. I have argued that the polymodal designation is preferable, but in practice the important thing is to recognize this distinct class of neurones. The question of which term is the best name is quite secondary. Good class names are helpful, however, and a profusion of parallel terminology can be confusing. It has been argued that class names such as polymodal nociceptor are dangerous, since they presuppose the function. More neutral terminology (e.g. type I and type II mechanoreceptor) is supposedly less prone to embarrassing error. The literature is full of examples of inappropriate names Vasopressin, rather than antidiuretic hormone, is one of my favourites. However, where function is well established it does have the advantage of conveying information and, often, of being easier to remember. Where would we be without our nociceptors, mechanoreceptors, and thermoreceptors! In the end, though, nomenclature is secondary to classification. To quote: ‘What’s in a name? That which we call a rose by any other name would smell as sweet’ (Shakespeare, Romeo and Juliet).
This brief account has borrowed heavily from the following sources.
Gordon, A.D. (1981). Classification. Methods for the exploratory analysis of multivariate data. Chapman & Hall, London.
Hughes, A. (1979). A rose by any other name . . . On ‘Naming of neurones’ by Rowe and Stone. Brain Behay. Evol. 16, 52–64.
Lynn, B. (1994). The fibre composition of cutaneous nerves and the classification and properties of cutaneous afferents with particular reference to nociception. Pain Rev. 1, 172–83.
Rowe, M.H. and Stone, J. (1977). Naming of neurones. Classification and naming of cat retinal ganglion cells. Brain Behay. Evol. 14, 185–216.
Rowe, M.H. and Stone, J. (1979). The importance of knowing our own presuppositions. Brain Behay. Evol. 16, 65–80.
Tyner, C.F. (1975). The naming of neurones: applications of taxonomic theory to the study of cellular populations. Brain Behay. Evol. 12, 75–96.