The first step in comparative analyses is to describe the patterns of distribution of the characteristics of interest across a chosen sample, in order to establish whether the conditions for one of the four methods exist. This involves investigating whether particular variants of one categorical variable are associated with particular variants of another, or whether continuous variables are correlated across the sample. Essentially, this is a similar process whether the variables concerned are categorical or continuous, and many features can be described either way. For example, dietary variation can be categorised according to the predominant food, as insectivory, frugivory etc., or measured according to the proportion of a particular food type in the diet, such as percentage fruit (e.g see Chapter 13). Where there is a choice of either a categorical or continuous measure, factors such as the nature and quality of available data, and the question under investigation are important. For example, duration of lactation is a measure used in both Lee and van Schaik et al.'s chapters (Chapters 5 and 8). Lee's study focuses on variation in the length of the lactation period itself and the degree of correlation with other continuous life history variables. However, for van Schaik et al., the feature of interest is categorical: whether the lactation period is longer or shorter than the gestation period. If it is shorter, post-partum oestrus, enabling reconception immediately following a birth, would be a viable evolutionary option, given that energetically the mother must wean one infant before having to feed a second. It should be borne in mind, however, that where it is possible to choose between categorical and continuous versions of a variable, this could affect the results, particularly levels of significance.
In comparing different species, variation in overall body size is commonly an important factor. The question of interest may be how a variable, say brain size, correlates with body size, or attention may be focused on residual variation from scaling relationships, such as relative brain size. Both types of investigation are utilised in this volume (e.g. Chapters 4 and 5). Allometric methods of analysis are commonly necessary in cross-species comparisons as many features do not scale to body size, or to each other, in a linear fashion. Rather, they are related through power functions. Thus, simple ratios, for example of neonate weight to maternal weight, should only be used with great care. The intention may be to 'remove size' from the comparison of species, but such ratios will not be 'size free' unless the variables concerned scale linearly with one another - that is, with the same exponent in relation to body size. For example, Charnov's recent life history theory (see Chapters 4 and 6) predicts that several life history characteristics will scale to body size with the same exponent (0.25). Hence, ratios of these variables are expected to be constant across species. However, instances of size-free ratios are rare in comparative biology.
The analyses presented by van Schaik et al. (Chapter 8) illustrate the centrality of investigating patterning to the comparative method. Data were collected for primate species for a wide range of features such as the incidence of infanticide, infant care styles, whether lactation is longer or shorter than gestation, mating patterns during pregnancy, the presence or absence of post-partum oestrus, the development of sex skin, and whether females produce calls related to mating. These data were examined to determine whether the pattern of distribution across species of variants of one feature is associated with that of another - the basic requirement if variables are causally or functionally related. For example, the species in which the mother alone carries the infant do not have post-partum oestrus, whereas most of those which park their infants, or in which carrying is shared with other individuals, do. In Ross and Jones' study (see Chapter 4), patterning is similarly fundamental, but here the variables are largely quantitative, such as maternal weight, age at first reproduction, mortality rates and interbirth intervals. The first step was to investigate the pattern of correlation between the variables. For example, across primates, taking variation in body weight into account, levels of mortality among infants and juveniles are correlated with birth rates, and species with higher pre-reproductive mortality reproduce faster. However, adult mortality rates are not correlated negatively with age of first reproduction for females, as was predicted; species with higher adult mortality do not start reproducing earlier (at least not in the small sample available).
Was this article helpful?