Big-brained birds survive better in nature

Daniel Sol () 3 Tama s Sze kely 0 2 Andr as Liker 1 Louis Lefebvre 4 0 Present address: Museum of Comparative Zoology, Harvard University , 26 Oxford Street, Cambridge, MA 02138 , USA 1 Department of Limnology, Pannon University , Pf. 158, 8201 Veszpre m , Hungary 2 Department of Biology and Biochemistry, University of Bath , Bath BA2 7AY , UK 3 CREAF (Centre for Ecological Research and Forestry Applications), Autonomous University of Barcelona , 08193 Bellaterra, Catalonia , Spain 4 Department of Biology, McGill University , 1205 avenue Docteur Penfield, Montre al, Que bec H3A 1B1 , Canada Big brains are hypothesized to enhance survival of animals by facilitating flexible cognitive responses that buffer individuals against environmental stresses. Although this theory receives partial support from the finding that brain size limits the capacity of animals to behaviourally respond to environmental challenges, the hypothesis that large brains are associated with reduced mortality has never been empirically tested. Using extensive information on avian adult mortality from natural populations, we show here that species with larger brains, relative to their body size, experience lower mortality than species with smaller brains, supporting the general importance of the cognitive buffer hypothesis in the evolution of large brains. 1. INTRODUCTION The evolution of large brains, such as those of humans, presents evolutionary biologists with an unresolved problem: if growing an enlarged brain has a high cost of development and maintenance (Allman 2000; Iwaniuk & Nelson 2003), why do some animals have large brains relative to their body size? A classic answer to this question is that the costs are compensated for later in life by the benefits that a large brain provides to survive environmental challenges through flexible behaviours (Allman et al. 1993; Allman 2000; Deaner et al. 2002), a theory known as the cognitive buffer hypothesis. Supporting evidence for the theory comes from the finding that big-brained animals have a higher propensity to innovate and learn (Lefebvre et al. 1997; Reader & Laland 2002; Byrne & Corp 2004) and such behavioural flexibility helps them face challenges presented by new or altered environments (Shultz et al. 2005; Sol et al. 2005a). Nonetheless, the hypothesis that large brains are associated with reduced adult mortality has not been tested in any group of animals. We have investigated the relationship between brain size and mortality rate with a comparative analysis in birds, using information on avian adult mortality from more than 300 natural populations of 220 species from polar, temperate and tropical regions (Liker & Szekely 2005). Birds are ideally suited for such a test as they represent one of the handful taxa for which the relationship between large brains and enhanced behavioural response to ecological challenges is best understood (Lefebvre et al. 2004; Sol et al. 2005a). Moreover, the unparalleled amount of data available on mortality from wild populations makes it possible to conduct a general test of the cognitive buffer hypothesis. 2. MATERIAL AND METHODS We gathered information on adult annual mortality rates (sexes averaged) for 319 populations of 236 species from published studies by searching extensively in reference books, species monographs and electronic databases that included biological abstracts (BIOSIS, 19752002; Liker & Szekely 2005). Mortality rate is considered context dependent and subject to measurement error, and may thus vary between populations from the same species. However, we found a high repeatability in our mortality measures, with 80.5% of variation found among rather than within species (see Lessells & Boag 1987 for the method). Information on brain and body masses (in grams) from published sources (see Sol et al. (2005a) for details) was available for 184 of these species; for 40 additional species, we estimated their brain size by using the average brain mass of species from the same genus (Sol et al. 2005a), a taxonomic level that predicts 91% of the species level variance. This yields a total of 303 populations and 224 species from polar, temperate and tropical regions (Liker & Szekely 2005). Conclusions are qualitatively similar whether or not we include the 40 species for which brain size is inferred; for this reason, we only report the results with the larger dataset. Previous work has shown that it is not brain size per se, but the extent to which the brain is either larger or smaller than that expected for a given body size which indicates adaptation for enhanced neural processing ( Jerison 1973). Three general methods have been proposed to remove the allometric effect of body size on brain size (Deaner et al. 2000, 2002): (i) estimate the residuals of a loglog least-square linear regression of brain mass against body mass, (ii) calculate the fraction of the body mass that corresponds to brain mass, and (iii) include absolute brain size and body mass (both logtransformed) as covariates in a multivariate model. Since there is still no consensus on which is the most appropriate method (Reader & Laland 2002), we validated the cognitive buffer hypothesis using the three approaches. The three methods yielded qualitatively similar results, although the second method did not appropriately remove the effect of body size and the third created problems of colinearity due to the high correlation between brain and body masses (rZ0.99). Thus, we report in the text the results obtained using the method of residuals. We tested the cognitive buffer hypothesis using a hierarchical approach (Bennett & Owens 2002) in which mortality rate was modelled at the levels in which substantial variation in mortality rates exist: populations and families. The population-level analysis was used to test the relationship between brain size and mortality rate while controlling for the effect of the environment and clade-traits, whereas the family-level analysis served to validate the hypothesis at the taxonomic level where the most diversification in brain size has occurred. A difficulty in population level analyses is the need to deal with the autocorrelation that may exist in mortality measures belonging to the same species, higher taxonomic levels or regions. Since it is not possible to deal with several sources of autocorrelation using classical phylogenetic-based techniques (e.g. independent contrasts), and a phylogenetic hypothesis for the studied populations is not available, we used generalized linear mixed models (GLMMs) to model mortality rate as a function of relative brain size while controlling for differences in mortality among species, higher taxonomic levels and regions (Blackburn & Duncan 2001; Sol et al. 2005a). We modelled the likely non-independence of mortality rates due to taxonomic or regional affiliations using a variance components model, assuming a common positive correlation between mortality (...truncated)