Life histories maternal energetics and brain size

Cinderalla Solution

Natural Metabolism Boosters

Get Instant Access

Perhaps differences in brain size reflect overall life history strategies or biological constraints (Sacher, 1959; Hofman, 1984; Shea, 1987; Harvey et al., 1987; Parker, 1990; Allman, McLaughlin and Hakeem, 1993; Martin, 1996), rather than ecologically related neural specialisation. One suggestion has been that brain size is linked to life span (Sacher, 1959; Hofman, 1984; Allman et al., 1993; see also Harvey and Read (1988) for a discussion). Sacher (1959) found that brain size and life span were more strongly correlated than either were with body size. Similarly, Allman et al. (1993) found that primate life spans and brain sizes were positively correlated even after the effects of body weight had been removed from each. Harvey et al. (1987) suggested that brain size is more directly related to age at maturity than to life span, because age at maturity reflects the amount of postnatal brain development and learning during the juvenile phase (see also Joffe, 1997). Others have suggested that energy constraints may play a key role in limiting brain size (Martin, 1981; Armstrong, 1983; Foley and Lee, 1992). Martin's ideas (1981, 1996) have been particularly influential and stimulating. He suggested that variation in adult brain size is linked to the metabolic turnover of the mother during gestation and lactation. The reason why frugivores have comparatively large brains might be simply that their energy-rich diets allow them to have higher basal metabolic rates (Martin, 1996: p. 153). As mentioned above, this idea originated in part from the apparent similarity of the allometric exponents for brain size and basal metabolic rate (Martin, 1981). Subsequently, Martin (1996) included gestation length with basal metabolic rate as components of maternal energy expenditure. He analysed the partial correlations between brain size, basal metabolic rate, gestation length and body weight in 53 species of placental mammals, finding significant correlations between brain weight and both metabolic rate and gestation length, with the effect of body weight partialled out.

The analyses described above suffer from two problems. The first is that they were carried out using means for species, or for higher taxonomic groupings such as genera, as independent data points, and thus are susceptible to the type of phylogenetic bias discussed above. The second is the 'Economos problem' (see Harvey and Krebs, 1990). Economos (1980) pointed out that body weight is an intrinsically 'messy' variable, subject to a high degree of intraspecific variability related to nutrition, disease and genetics. Life span and brain size, or metabolic rate and brain size, may be more highly correlated with each other than with body weight merely because they are more accurate indices of a species' body size than is weight. Using regression to remove the effects of body weight from both variables of interest (e.g. Allman et al., 1993) would only compound the problem, because this adds the same error variance in the body size measurements to both variables. If we then correlate the 'size-corrected' variables with each other, there is a relatively high chance of finding a spurious positive correlation, because of the correlated error added to each variable. To make this clearer, imagine the (admittedly unlikely) case in which the true correlations between brain size and body size and between life span and body size were perfect. Residual values for brain size and life span after regressing them on mass would be entirely determined by the measurement errors in each variable (because, in theory, all points should lie on the line, making their true residual values zero). If mass was substantially more error prone than brain size and life span, then the size-corrected values for each of the latter variables would be strongly and similarly affected by the error in body size, leading to positive correlations even in the absence of any biological relationship. A possible solution to mass errors is to estimate size from separate samples of individuals for each variable being investigated (Harvey and Krebs, 1990), or perhaps to use size measures other than weight (Mace and Eisenberg, 1982). The extent to which this theoretical problem is a practical reality has not yet been assessed, and therefore the results of studies in which two or more variables corrected for body size are correlated with each other should be treated with caution. In mitigation, Allman et al. (1993) showed that the size of organs other than the brain did not correlate with life span in the same way, which would have been expected if the original correlation was a statistical artefact.

If we ignore the theoretical Economos problem for the moment, what does phylogenetic re-analysis of the links between brain size and life histories reveal? Life-history variables such as age at first reproduction and life span are so closely linked to each other (e.g. Harvey et al., 1987; Charnov, 1991; Purvis and Harvey, 1995) that it may not make sense to treat them as different variables within a single comparative analysis. Thus, separate multiple regressions were carried out for each with 55 independent contrasts. Body weight was accounted for in these analyses by entering it into the multiple regressions (p < 0.001 in each case). There is no significant association between brain size and life span (t = 1.58, p = 0.12), nor between brain size and age at first reproduction (t =— 0.42, p = 0.67). These tests are, however, quite strongly affected by a few outlying contrasts calculated at younger nodes in the phylogeny. As noted earlier and in Chapter 3, error variance tends to be amplified in contrasts at younger nodes. Excluding nodes younger than 5 million years gives significant results (with 38 contrasts, t = 3.27, p = 0.002 for life span, t = 2.24, p = 0.03 for age at first reproduction).

In the previous section it was shown that ecological variables (percentage frugivory and social group size) are correlated with brain size. How can this be reconciled with the results for life history variables? Are ecology and life histories confounded, such that only one of them is the true correlate of brain size? Multiple regressions, again on contrasts calculated at older nodes ( > 5 million years), suggest that both ecological and life history variables are separately correlated with brain size. Brain size is significantly positively correlated with age at first reproduction (t = 2.34, p = 0.03), group size (t = 2.61, p = 0.01) and percentage frugivory (t = 2.33, p = 0.03). In a separate multiple regression, brain size is positively related to life span (t = 2.79, p = 0.01), group size (t = 1.98, p = 0.05) and frugivory (t = 2.42, p = 0.02). These analyses therefore support both ecological specialisation and life histories as correlates of brain size.

Next, the maternal energy hypothesis (Martin, 1996) is re-examined, following Martin in considering the joint effects of two maternal energy variables, basal metabolic rate and gestation length. The problem of phylogenetic bias is very apparent here. For example, correlating basal metabolic rate with brain size across species gives a statistically significant result (multiple regression controlling for body weight, t = 2.7, df = 2,18, p = 0.01). However, this appears to be largely because, relative to their body sizes, haplorhines have significantly higher metabolic rates (t = 2.6, df = 2,19, p = 0.02) and brain sizes (t = 8.1, df = 2,116, p < 0.001) than do strepsirhines. When the original multiple regression is re-run, controlling for suborder membership by entering this as a dummy variable, brain size is uncorrelated with basal metabolic rate (t = 1.6, df = 3,17, p = 0.13). Similarly, there are no significant associations between brain size and maternal energy variables when the independent contrasts method is used (multiple regression on 18 sets of contrast values, controlling for body size: t = — 0.21, p = 0.84 for basal metabolic rate, and t = 1.19, p = 0.25 for gestation length). The contrasts were nearly all made at older nodes, and excluding the one node younger than 5 million years does not make the results significant (multiple regression on 17 contrasts, t = 0.47, p = 0.65 for basal metabolic rate, and t = 1.17, p = 0.26 for gestation length), nor does the additional removal of another contrast that appeared as a relatively large outlier in scatterplots (with 16 contrasts, t = 0.47, p = 0.65 for basal metabolic rate and t = 1.14, p = 0.27 for gestation length). Perhaps life history variables are confounding relationships with the maternal energy variables? To test this, the author incorporated age at first reproduction into the analysis. Because this makes the sample size rather small (n = 11) relative to the number of test variables, a stepwise regression was used to select the variables significantly associated with brain size, taking body size into account by entering it first. This showed a significant positive correlation for age at first reproduction (p = 0.02), but not for basal metabolic rate or gestation length (p > 0.1 in each case).

These results cast doubt on a link between maternal energy and brain size in primates. Results of other comparative studies also lend little support. Among mammals generally, neonatal brain size is not correlated with basal metabolic rate (Pagel and Harvey, 1988b). Although neonatal brain size is correlated with gestation length (Pagel and Harvey, 1988b), species with small brained neonates are not small-brained as adults, because they compensate by greater postnatal brain growth (Harvey and Read, 1988). While frugivores have large brains, they do not seem to have high basal metabolic rates (Elgar and Harvey, 1987; Ross, 1992), and carnivory appears to be the only dietary correlate of basal metabolic rate that is independent of phylogeny (Elgar and Harvey, 1987). Within the order Carnivora, the more frugivorous species actually have the lowest metabolic rates (McNab, 1995). Because several of these studies did not control for phylogenetic bias, further work is needed to establish the validity of their conclusions.

If substantiated by further work, the lack of correlations between brain size and metabolic rate does raise the question of how large brains are accommodated in metabolic terms. Because brains are metabolically expensive (Martin, 1981, 1996; Foley and Lee, 1992; Aiello and Wheeler, 1995), the extra energy for brain expansion must come from somewhere.

One possibility is a trade-off between the size of the brain and of other organs, such as the liver and gut (Aiello and Wheeler, 1995). A folivorous diet creates more demands on the digestive system than does a frugivorous diet, requring a larger, more metabolically active gut. Thus, frugivory may simultaneously select for large brains and for small guts, leading to a trade-off between the two (Aiello and Wheeler, 1995). Presumably, the implication of such a trade-off is that the need for large brains and the relaxation of the need for large guts is a happy coincidence.

Ontogenetic constraints on brain specialisation?

Finlay and Darlington (1995) proposed that mammalian brain evolution has been highly constrained by ontogenetic mechanisms limiting variability in the size of specific brain regions. They stated that 'the most likely brain alteration resulting from selection for any behavioural ability may be a coordinated enlargement of the entire non-olfactory brain'. If true, this means that individual brain regions have not been free to vary in size independently of the whole, and that selection has acted primarily on overall brain size. The idea fits most easily with theories of brain size that invoke general life history strategies or metabolic constraints, rather than adaptive specialisation in response to the sensory and cognitive demands of particular niches. Is it correct?

Finlay and Darlington's claims are based largely on the apparent predictability of the size of major brain structures from overall brain size across three orders of mammals. They showed that the size of each individual structure is highly correlated with overall brain size, explaining 96% of the variance in the size of any given structure. Only the olfactory bulb and associated structures deviated from this predictability to any extent, and a two-factor model comprising the effects of overall brain size and olfactory bulb size accounted for 99% of the variance in the size of individual brain structures. Brain structures that develop late in ontogeny, such as the neocortex, were found to increase relatively rapidly in size with interspecies increases in brain size, leading to the suggestion that evolutionary changes in brain size occur by simple developmental shifts, such as an increase in the period of brain growth. An increase in this period would result in all structures growing larger, but late-developing structures would be most affected.

Finlay and Darlington's results seem to provide strong support for their conclusion that individual brain systems have not, in general, evolved independently of the whole, and therefore that adaptive specialisation has

Fig. 7.2 Neocortex size, relative to the size of the rest of the brain, in primates (circles) and insectivores (crosses). It is clear that primates' larger brains are associated with relative expansion of the neocortex.

Log volume of rest of brain

Fig. 7.2 Neocortex size, relative to the size of the rest of the brain, in primates (circles) and insectivores (crosses). It is clear that primates' larger brains are associated with relative expansion of the neocortex.

been minimal. There are, however, a number of reasons to doubt this conclusion.

First, Finlay and Darlington's analysis used species values for primates, bats and insectivores as independent data points. Their analysis of the regression slopes of structure size on overall brain size depends on the validity of these slopes, but, as explained in the section on phylogeny and brain evolution above, the slopes are likely to have been biased by grade shifts between orders. For example, they pointed to the 'explosive nature of the change in neocortex size' with increasing brain size. The apparently explosive scaling of neocortex size is, however, partly a product of adaptive grade shifts between orders and suborders (Fig. 7.2). Primates have larger neocortices relative to brain size than do insectivores, and slope values calculated across orders will therefore be inflated by this grade shift. Additional grade shifts in neocortex size are present within orders (e.g. Barton, 1996). Other structures, such as the cerebellum, exhibit similar grade shifts (see Barton, in press). Although a re-analysis by Darlington (personal communication) suggests that between-order grade shifts have a small effect on slopes, the fact that they exist at all of course contradicts the idea that the size of each brain structure is strictly tied to overall brain size by a universal growth law. Instead, such grade shifts indicate adaptive specialisation in brains.

Second, the regression slopes relating the size of each structure to overall brain size are affected by autocorrelation. Brain structures that make up a relatively large proportion of total brain size (such as the neocortex and cerebellum) are inevitably more strongly correlated with it than are smaller structures. Large structures will therefore tend to exhibit higher slopes on total size than do small structures, even in the absence of any real biological effect, because least-squares regression slopes vary with the strength of the correlation (the higher the correlation, the greater the slope). Thus, the differences in regression slopes may indicate differences in the extent of autocorrelation as much as they do differences in ontogenetically driven scaling laws.

Third, the conclusion that only a trivial amount of variance in brain structure size remains unexplained by Finlay and Darlington's two-factor model depends on some questionable assumptions. Much of the variance they attributed to the overall brain size factor may, in fact, be associated with body size. They argued that the first factor in their principal components analysis was most usefully construed as simply overall brain size, rather than body size. One reason for this was that the correlation with body weight was relatively low. As explained above, however, this may be simply because of the greater error in body size data. A slightly different pattern is revealed by performing separate principal components analyses for each order, first removing the effects of body size, and using independent contrasts rather than species as data points.1 The first factor accounts for 49%, 62% and 67% of the variance among bats, insectivores and primates respectively, which are quite high, but substantially less than the 96% in Finlay and Darlington's analysis. As in Finlay and Darlington's analysis, the first factor was a 'global' factor, in that it loaded moderately to strongly on all structures except for, in primates and insectivores, the olfactory bulb. Ontogenetic constraints are, however, only one of a number of possible explanations for this global factor. Another is failure to remove all body size-related variance owing to the use of error-prone weight measurements. The second factor extracted in this author's analyses differed between orders: in primates and insectivores it loaded most highly on the olfactory bulb and another olfactory structure, the piriform lobe, in broad agreement with Finlay and Darlington's two-factor model. Again functional specialisation is suggested because olfactory systems are correlated with specific ecological factors (Barton, Purvis and Harvey, 1995). Bats showed a different pattern. A second and a third factor were extracted, the former loading on the cerebellum, medulla and mesencephalon, the latter on the striatum and neocortex (which have major functional and anatomical links - see Keverne, Martel and Nevison, 1996). These differen ces between orders are not predicted by strict ontogenetic constraints, but accord with the idea of selection affecting separate but functionally linked brain structures.

A difficulty with interpreting the results of principal components analysis is that the cut-off between what constitutes interesting variance and what constitutes trivial residual variance is quite arbitrary; a variance proportion that superficially seems trivially small may nevertheless be evolutionarily and neurally highly significant. For example, when the author extracted a third factor for primates, this loaded on both schizocor-tex and hippocampus, in accord with the fact that these two structures have strong anatomical and functional connections and so may be expected to have co-evolved under appropriate selection pressures. The point is reinforced by examining partial correlations between all brain structures. This shows that many significant correlations are found when variation in other structures is partialled out, and these evolutionary correlations reflect known anatomical and functional links, such as between amygdala and piriform cortex (Barton, in press). Finlay and Darlington themselves made the point that even very small percentages of the total interspecies variance in brain size may represent functionally significant amounts of neural tissue, and they agree that scope for adaptive specialisations exists in spite of the apparent uniformity of brain proportions. Unfortunately, these important comments have been rather lost in some subsequent discussions (e.g. Quartz and Sjenowski, 1997). Finlay and Darlington are right that, in statistical terms, the residual variance unexplained by overall size is generally quite small. The question is, does any of that residual variance represent adaptive specialisation?

Brain specialisation

A number of comparative studies of mammals provide evidence for mosaic evolution of the brain, in which specific parts have evolved independently of the whole. Importantly, the correlations found between lifestyle and the size of specific brain regions accord with the functions of these structures. In primates, for example, activity period is correlated with the size of primary sensory structures (Barton et al., 1995): relative to the size of the rest of the brain, diurnal primates have larger visual cortices, whereas nocturnal primates have larger olfactory bulbs (Fig. 7.3). Clearly, selection has emphasised the sensory systems most appropriate to the ambient light levels associated with each niche. Diet also shows correlations with sensory brain structures in primates; among diurnal species, frugivores have a

Contrasts between nocturnal and diurnal lineages

Olfactory bulb Primary visual cortex

Fig. 7.3 Relative sizes of the olfactory bulb and primary visual cortex in nocturnal versus diurnal primates. The relative size of each structure was calculated as the residuals from the regression of structure size on the size of the rest of the brain (exluding all main sensory structures). The graph plots four independent contrasts in the relative size of each structure. Positive contrasts indicate larger structure size in nocturnal lineages; negative contrasts indicate larger structures in diurnal lineages. The compensatory differences in the size of the olfactory and visual structures may explain why overall brain size does not differ between nocturnal and diurnal lineages. Graph based on analysis in Barton et al. (1995).

larger primary visual cortex than do folivores, whereas among nocturnal species, frugivores have relatively enlarged olfactory bulbs and piriform lobes (Barton et al., 1995). The implication that different sensory systems have been favoured for locating fruit depending on activity period has recently found support in an experimental study by Bolen and Green (1997), showing that owl monkeys (Aotus nancymai) are more efficient at locating fruit by olfactory cues than are diurnal capuchin monkeys (Cebus apella). On the other hand, owl monkeys have lower retinal cone densities than their diurnal relatives, poorer colour vision, and lack a distinct fovea. Also in primates, relative neocortex size is correlated with social group size (Dunbar, 1992; Sawaguchi, 1992; Barton, 1996; Keverne, et al, 1996), and this has been taken as evidence for the 'social intelligence' hypothesis (see Byrne and Whiten, 1988).

Among mammals other than primates, olfactory structures are relatively small in aquatic taxa, within both insectivores (Fig. 7.4) and carnivores

Contrasts between nocturnal and diurnal lineages

Was this article helpful?

0 0
Boost Your Metabolism and Burn Fat

Boost Your Metabolism and Burn Fat

Metabolism. There isn’t perhaps a more frequently used word in the weight loss (and weight gain) vocabulary than this. Indeed, it’s not uncommon to overhear people talking about their struggles or triumphs over the holiday bulge or love handles in terms of whether their metabolism is working, or not.

Get My Free Ebook

Post a comment