Ecological factors have been shown to be important for brain size evolution. In this comparative study among catarrhine primates, we examine two different ways in which seasonality may be related to brain size. First, seasonality may impose energetic constraints on the brain because it forces animals to deal with periods of food scarcity (Expensive Brain hypothesis). Second, seasonality may act as a selective pressure to increase brain size, as behavioral flexibility helps to overcome periods of food scarcity (Cognitive Buffer hypothesis). Controlling for phylogeny, we found a strong negative relationship between brain size (relative to body mass) and the degree of experienced seasonality, as estimated by the variation in net energy intake. However, we also found a significant positive relationship between relative brain size and the effect of so-called cognitive buffering, proxied by the difference between environmental seasonality and the seasonality in net energy intake actually experienced by the animals. These results show that both energetic constraints of seasonal habitats as well as cognitive buffering affect brain size evolution, leaving environmental seasonality uncorrelated to brain size. With this study we show the importance of simultaneously considering both costs and benefits in models of brain size evolution.

To explain the observed variation in primate brain sizes, several adaptive hypotheses have been proposed (reviewed in Healy and Rowe 2007). Most of these hypotheses have focused on relating the evolution of brain size to the selective benefits due to enhanced cognitive abilities (e.g., Dunbar 1998; Tomasello 1999; Deaner et al. 2007; Reader et al. 2011) which may be favored by certain ecological conditions (Reader and Laland 2002; Fish and Lockwood 2003; Shultz and Dunbar 2006). However, environmental conditions may also constrain the evolution of relatively large brains. In recent years, we have followed an approach of addressing brain size evolution from a purely energetic perspective, that is, applying the Expensive Brain framework (Isler and van Schaik 2009a). Brain tissue requires a high and continuous energy supply (Mink et al. 1981) and consequently serious starvation, especially during development, leads to permanent brain damage (Lukas and Campbell 2000). The costs of brain function cannot be temporarily reduced (except probably in deeply hibernating rodents, Krilowicz et al. 1988). Evolution of an increased brain size (relative to body size) is therefore only possible if either total energy throughput is increased, the energy allocation to other functions is reduced, or a combination of the two is achieved (see also Martin 1981; Aiello and Wheeler 1995; Isler and van Schaik 2009a). This energetic perspective predicts that brain size is reduced where animals experience periodic energy shortages in a seasonal habitat. Even if physiological buffers such as fallback foods, fat storage, reduced activity, or hibernation allow for survival during lean periods, the net energy availability is still reduced relative to the season in which food is abundant, and brains are expected to be relatively smaller in comparison with a similar species that does not experience such seasonal food shortages.

Therefore, the central prediction of the Expensive Brain framework is that, ceteris paribus, a species’ brain size is negatively related to the intensity of seasonality in net food intake, that is, the “experienced seasonality” (XPR) (Fig. 1A). Recently, we confirmed this prediction in African strepsirrhine primates (van Woerden et al. 2010). Additional support comes from a comparison of orangutans subspecies (Taylor and van Schaik 2007) and studies in various mammals showing that the dwarfing effects of unavoidable food scarcity on brain size exceed those on body size on small oceanic islands (Filin and Ziv 2004; Köhler and Moyà-Solà 2004; Lomolino 2005; Niven 2007; Weston and Lister 2009).

Figure 1.

The two hypotheses presented in this article with their predicted relationship between seasonality and brain size. Note that each hypothesis forms a prediction between relative brain size and a different, nonexclusive measure of seasonality.

Environmental seasonality does not lead to seasonality in food intake, if the decrease in food availability is fully compensated by an increase in foraging effort or a switch to other (hidden) high-quality food sources. The Cognitive Buffer hypothesis (see Allmann et al. 1993; Deaner et al. 2003; Sol 2009) predicts that larger brains provide the cognitive abilities that allow for increased behavioral flexibility, which among other things, facilitates the buffering of environmental seasonality. Thus, larger brained species are supposed to outperform smaller brained species in more seasonal habitats, which are more cognitively demanding because preferred food sources are more difficult to (re)locate in space or time (Klopfer and MacArthur 1960; Sol et al. 2005a; Sol et al. 2008). According to this hypothesis, selection is expected to favor species with relatively large brains in more seasonal habitats. Support comes mainly from studies in birds. Schuck-Paim et al. (2008) found a positive correlation between climatic variability and brain size in Neotropical parrots. In addition, migrating bird species have smaller brains than nonmigrating bird species (Winkler et al. 2004; Sol et al. 2005b), which can be reflecting a cognitive buffer effect in the residential species (Sol et al. 2005b) or a reduced selective advantage of enhanced cognitive performance in migratory species (Sol et al. 2010). In contrast, evidence of cognitive buffer effects in mammals is very limited (Sol et al. 2008). Reader and MacDonald (2003) reported that innovation rate or neocortex ratio (both closely related with overall brain size) are not correlated with climatic variability among African primates.

However, cognitive buffering and energetic constraints on brain size are not mutually exclusive. Indeed, we expect that cognitive buffering may only partially reduce the XPR of a species, still leaving the nonbuffered remnant of environmental seasonality to constrain brain size. To investigate the cognitive buffer effect, we must therefore not merely consider the environmental seasonality, but also the amount of buffering, that is, the difference between food availability and energy intake. Here, we assess the temporal variation in net energy intake, a measure of XPR by the monthly variation in consumption of major diet components (fruits, flowers, young and mature leaves, and insects) multiplied by their quality (as in van Woerden et al. 2010). If energetic constraints on brain size apply, we expect a negative correlation between brain size and this XPR. If cognitive buffering takes place to cope with harsh environmental conditions, the XPR is smaller than the variation in food availability, which is proxied by the seasonality of the environment (ENV). Animals that cognitively buffer will manage to keep their energy intake relatively constant despite dramatic environmental fluctuations. Therefore, the difference between the environmental (i.e., expected) seasonality and the seasonality that is experienced (ENV-XPR) tells us how much buffering is taking place. The Cognitive Buffer hypothesis then predicts brain size to be positively related to the amount of buffering (Fig. 1B).

Our recent study in Malagasy lemurs (van Woerden et al. 2010) was the first to look at both energetic constraints and cognitive buffering effects of seasonality on brain size evolution using this approach and found strong support for energetic constraints, but only a weak indication for a cognitive buffer effect. In the present study, we investigate whether cognitive buffering is more important in the catarrhine clade of nonhuman primates, consisting of Old World monkeys and apes. Catarrhine primates have greater cognitive abilities (reviewed in Fichtel and Kappeler 2010) and are generally larger brained relative to body mass than lemurs (Isler et al. 2008). We test (1) the predictions of the Expensive Brain framework by examining the relationship between XPR and brain size, (2) the predictions of the Cognitive Buffer hypothesis by looking at the relationship between the amount of buffering (ENV-XPR) and brain size, and (3) whether one of these effects prevails by looking at the relationship between environmental seasonality and brain size.



Our sample includes only female wild adult specimens (third molar present) of which the original provenance was known, to exclude a possible effect of captivity and a bias through sexual dimorphism (Plavcan and van Schaik 1992; Smith and Cheverud 2002). Female brain size of 70 species of nonhuman catarrhine primates were assessed through measuring endocranial volumes (ECVs) using glass beads in eight European and four American museums and added to the dataset of Isler et al. (2008) to yield a total of 1756 female skulls, 1576 of which with known origin from 1229 different locations. Conversion to brain mass is not needed as the two have been shown to correlate isometrically in primates (Isler et al. 2008). Because it has been documented that primates tend to have smaller body sizes in more seasonal habitats (Albrecht et al. 1990; Lehman et al. 2005; Plavcan et al. 2005), it is important to include body size as a covariate in the analyses. Female body masses from wild study populations were collected from literature sources, or if possible taken from the same museum specimens that ECV was measured from (Table S1). Species were included if more than five measurements were available to calculate an average female ECV value (Table S1).


Experienced seasonality

We measured XPR by using temporal variation in the time spent feeding on diet components (and thus their estimated consumption) weighted for their nutritional value. Monthly dietary data were collected by literature research. Only studies that reported consumption of different dietary items over 12 consecutive months were included. In total, dietary data were available for 63 populations of 36 catarrhine species (Table S1). The coefficient of variation (CV) in consumption of dietary components over a year was estimated as follows. From the monthly means of the time spent feeding on the following food items (insects, fruit/seeds, flowers, young leaves, and mature leaves), we calculated the energy gained per month by multiplying the sum of each item by its energetic quality (eight for insects, five for fruits, seeds, and flowers, three for young leaves, and one for mature leaves, as calculated from gram of crude fiber per kilogram of dry matter by Langer 2003). Fiber content is commonly used as a measure of digestibility and thus energy gained per unit time (McNab 2002). The CV among months in this measure yielded the CV of net energy intake. The standard deviation (SD) in this measure was larger between species (0.07) than between populations (0.04), demonstrating that the variation is mostly between species. We assumed an equal energy expenditure throughout the year, because the variation of energy expenditure of wild populations has been reported for only few species (e.g., Tsuji et al. 2008). Daily travel distance does vary seasonally, but as costs of travel per day are only a minor portion of the daily energy expenditure in primates (Altmann 1998), we assume that we are allowed to ignore variation in energy expenditure for our purpose.

Ideally, we would use diet variability, brain size, and body mass of the same population for each primate species. However, diet composition has generally been studied in different populations than the specimens available in museums from which brain sizes were measured. Therefore, we also compiled values of brain size and body mass sampled within a 100 km radius of the population in which diet composition was studied. Because the results from this reduced, conservative dataset (N = 26, see Table S2) did not differ from those derived from a larger dataset containing the species averages of brain size and body mass (N = 36), all of the results presented here are based on the larger dataset.

Cognitive buffering (ENV-XPR) and environmental seasonality (ENV)

As a proxy for food availability, we extracted several measures of environmental variables from remote sensing databases. Precipitation and temperature seasonality were extracted from the WorldClim database (Hijmans et al. 2005) using ArcGIS 9.1 and a more direct measure of seasonality in plant productivity, the Normalized Difference Vegetation Index (NDVI, see Myneni et al. 2005), from the GIMMS database (Tucker et al. 2005). First, from monthly precipitation means, we calculated the CV (CV = SD/mean); the mean vector length (r), that estimates the concentration of precipitation over the year (Batschelet 1981); and P2T, a measure of the length of the dry season, where a dry month has a total precipitation (mm) that is less than two times the mean temperature (˚C) (Walter 1971). Second, we calculated SD among monthly mean temperatures. And finally we calculated the CV among months in the NDVI as a more direct measure of seasonality in plant productivity (Pettorelli et al. 2005) than the climatic variables.

To estimate the extent of cognitive buffering, we calculated the difference between the seasonality of the environment and the experienced seasonality (ENV-XPR). The measures of environmental seasonality were extracted from the locations of the study populations for which dietary data were reported. Animals that buffer more will have a large difference (ENV-XPR), because they are able to keep their energy intake (XPR = experienced seasonality) more constant than expected on the basis of the seasonality in their habitat (ENV = environmental seasonality). To calculate this difference, we subtracted CV in diet (XPR) from either CV in plant productivity (NDVI) (ENV1) or from CV in precipitation (ENV2).

Climatic data and plant productivity were additionally compiled (see above for details) for all the species for which we had a measure of female brain size and body mass (N = 70 species, see Table S1). Locations from which these environmental data were taken from were matched to the locations where the specimens’ ECV originated from.


We controlled for phylogenetic relatedness using least-squared regressions (PGLS) analyses in R (R Development Core Team 2010) with the CAIC package (Orme et al. 2009). Phylogeny was based on version 2 of the consensus tree from the 10K Trees Project (Arnold et al. 2010) with Colobus satanas, Procolobus verus, Piliocolobus kirkii, Cercopithecus stuhlmanni, C. pogonias, and Semnopithecus priam added according to Bininda-Emonds et al. (2007), Presbytis spp. according to Meijaard and Groves (2004), Trachypithecus vetulus according to Osterholz et al. (2008), and Gorilla beringei according to Jensen-Seaman et al. (2003) (Fig. S1). All continuous variables were log-transformed before analysis, and statistical tests were parametric least-squares regressions, using R. In all multiple regressions, body mass was included as a covariate. Degree of folivory, group size, home range size, geographical range, gestation length, and mating system have been shown to correlate with brain size in primates (e.g., reviewed in Healy and Rowe 2007). Hence, we built models including these variables to eliminate their possible confounding effects. We estimated degree of folivory as yearly average percentage of leaves in the diet and group size as the average for the population concerned (values and sources are listed in the Supplementary Information). To choose the best fit from a set of models, we followed the standard approach (e.g., Richards 2005) of comparing the AIC (Information Theory Criterion, Akaike 1974) of different models. A lower AIC value indicates a better fit of the model to the data.

As the parameter lambda was close to 1 in most best-fit models, indicating a strong phylogenetic component in the data, usage of a phylogenetic method is required (Pagel 1999). Bivariate plots of seasonality measures versus residuals of brain size against body mass are shown for illustration, both using species means and independent contrasts.


First, as predicted by the Expensive Brain hypothesis and controlling for the effect of body mass, brain size was negatively correlated with XPR, as measured by variation in dietary consumption (Figs. 2A and B; Table 1A). Controlling for various covariates had little influence on the significance of the effect and did not improve the fit of the model (Table S3). This result indicates that species that experienced greater seasonality in their dietary energy intake had smaller brains relative to their body mass than species that experienced less seasonality in their diet.

Figure 2.

The influence of seasonality on relative brain size in nonhuman catarrhine primates. Species-level values are shown in the top panel, independent contrasts in the lower panel.. As predicted by the Expensive Brain framework, experienced seasonality and brain size were negatively correlated (A and B). Additionally, there was a very strong cognitive buffering effect (C and D), as predicted by the Cognitive Buffer hypothesis. Consequently, no clear relationship was found between environmental seasonality and brain size (E and F), indicating that the effects tend to cancel each other. For statistical significance see Table 1.

Table 1.  Phylogenetic least-squared regressions (PGLS) testing for possible effects of seasonality on brain size. Each predictor variable was tested separately along with ln body mass (results not shown, P < 0.0001 in all cases). Relationships are shown between relative brain size and (A) variation in diet, (B) environmental variation relative to diet variation, and (C) environmental seasonality for 36 nonhuman catarrhine primates. Body mass was always included as a covariate. All lambdas are close to 1, which indicates that there was a strong phylogenetic component in the data and the necessity of applying a phylogenetic method (Pagel 1999).
 Phylogenetic signal (lambda)t-valueP-valueAIC model
(a) Expensive Brain framework    
Experienced seasonality (XPR): CV in diet0.99−3.390.002−60.9
(b) Cognitive Buffer hypothesis    
Buffer (ENV1-XPR): CV in plant productivity—CV in diet0.99 3.280.002−60.3
Buffer (ENV2-XPR): CV in precipitation—CV in diet0.99 2.020.051−54.4
(c) Environmental seasonality    
CV in plant productivity (ENV1)0.99 1.490.14−60.2
CV in precipitation (ENV2)0.99 0.630.53−58.3
r in precipitation0.99 0.450.66−58.1
SD in temperature0.99 0.240.81−57.9

Second, in concurrence with the Cognitive Buffer hypothesis, relative brain size was significantly positively correlated with the amount of cognitive buffering as measured by the difference between XPR and the seasonal variation in plant productivity (Figs. 2C and D; Table 1B), and nearly so when precipitation is used instead. Again, including possible confounding variables did not affect these results (Table S3). Species that exhibited less variation in their energy intake than in their environment had larger brains relative to their body mass.

Furthermore, we looked at the relationship between environmental seasonality and brain size. None of the environmental variables were correlated with relative brain size (Figs. 2E and F; Table 1C). The enlarged sample of the 70 species yielded very similar results (Table S4).

Finally, there was no relationship between environmental seasonality and XPR (XPR vs. CV in precipitation N = 36, r2= 0.01, P = 0.5; XPR vs. CV in NDVI, r2= 0.04, P = 0.3), showing that catarrhine primates do not follow the seasonality of their habitat in their energy intake.


Species that experienced a higher degree of seasonality had a relatively smaller brain, suggesting that the energetic constraints due to seasonally induced food scarcity are an important factor in brain size evolution. This finding supports the Expensive Brain framework. However, relatively large-brained species exhibited more cognitive buffering, that is, showed less seasonality in their diet consumption (XPR) than expected on the basis of environmental seasonality. This implies that cognitive buffering of environmental seasonality also operates, enabling larger brained primates to live in these habitats despite the costs. This supports the Cognitive Buffer hypothesis, because exploiting varying food sources probably requires cognitive behavioral flexibility, such as switching to alternative food sources. Extractive foraging and tool use may be the most energetically rewarding behaviors used for cognitive buffering of environmental conditions, because they provide access to hidden and highly nutritional food items. Because what matters is the relative energetic costs of encephalization, it is irrelevant whether brains are larger relative to body mass or body mass is smaller relative to brain size. We did not test here whether seasonality in food intake is related to body mass alone, but of course this relationship is very likely also found in catarrhine primates (cf. Albrecht et al. 1990; Lehman et al. 2005). However, Isler and van Schaik (2009b) showed it is unlikely that the correlations between XPR and brain size, controlling for body mass, were in fact due to the “Economos-effect” (Economos 1980), that is, because brain size is a better proxy of body size than body mass itself. Overall, our results indicate that energetic constraints and cognitive buffer effects tend to cancel each other in catarrhine primates, because we find no relationship between environmental seasonality and brain size.

Our results support the notion that larger brained species may benefit from dealing with environmental change through behavioral flexibility. Thus, relatively large brains may have evolved to deal with novel ecological challenges, as is suggested in birds (Sol and Lefebvre 2000; Shultz et al. 2005; Sol et al. 2005a) and a broad range of mammals (Sol et al. 2008). However, here we show that to be able to benefit from these advantages, energetic costs need to be overcome to actually grow and maintain a larger brain.

In African strespsirrhine primates (lemurs and lorises), the energetic effect of seasonality on their relative brain size is very pronounced (van Woerden et al. 2010), whereas cognitive buffer effects are much weaker (PGLS: P = 0.14, λ= 1.00, Fig. S2) than within catarrhine primates. Energetic constraints prevail over cognitive buffering in the lemurs, as shown by consistent negative correlation between relative brain size and environmental seasonality (both climatic seasonality and plant productivity, see van Woerden et al. 2010). In other words, XPR more closely reflects habitat seasonality in strepsirrhines compared to the catarrhine primates. The different pattern of results for the two lineages might be due to differences in the distribution of energetic costs between small and large primates. Extant strepsirrhines devote a relatively larger percentage of basal metabolism to brain maintenance (11–12%) compared to larger primates (8–9% in cercopithecoid primates, calculated from other studies Mink et al. 1981; Isler et al. 2008, see Fig. 3). Therefore, the threshold for the effectiveness of cognitive buffers to overcome the energetic constraints of increasing brain size may be higher in strepsirrhines (and other small primates such as callitrichines) compared to the larger monkeys or apes. However, a difference in body mass is unlikely to be the only factor, because this does not explain why most of the much larger bodied extinct lemur species also had relatively small brains compared to catarrhines (Schwartz et al. 2005). An alternative explanation is that perhaps lemurs show more limited cognitive buffering because they more often face periods of unavoidable starvation, an idea supported by the many adaptations to cope with long periods of food scarcity. Thus, the only hibernating primates are found among lemurs (Schülke and Ostner 2007), and all lemurs are seasonal breeders, except the large-brained aye-aye (Sterling 1994), which is an extractive forager.

Figure 3.

The relative energetic cost of strepshirrhine brains exceeds that of catarrhines (ANOVA: F(1,13) = 5.23, P = 0.038), some Strepsirrhines allocate a similar proportion of their metabolism to their brain as humans do (dashed line, Holliday 1971). The percentage of basal metabolic rate (BMR) used for the brain was estimated from calculating brain metabolic rates per gram of brain tissue (Mink et al. 1981). BMR values of primate species were taken from McNab (2008). Metabolic consumption of the brain mass was then divided by basal metabolic consumption of the body.

The two hypotheses presented in this article, the Expensive Brain framework and the Cognitive Buffer hypothesis, are nonexclusive and both turn out to be crucial to disentangle the relationship between seasonality and brain size. They both affect how relative brain size responds to environmental seasonality, and can therefore be integrated as follows (Fig. 4). If the energetic constraints predominate, a negative correlation is found (dashed line, dark grey area, Fig. 4). If the cognitive buffer effect predominates, a positive correlation is found (dotted line, light grey area, Fig. 4). If both effects tend to cancel each other, there is no clear correlation between environmental seasonality and brain size (black area, Fig. 4).

Figure 4.

The predictions of the two hypotheses, the Expensive Brain framework and the Cognitive Buffer hypothesis, presented in this article can be integrated into one graphical representation between environmental seasonality and brain size. A negative correlation will be found if energetic constraints prevail (dark grey area), whereas if cognitive buffer effects are most important, a positive correlation will be found (light grey area). If these effects are equally strong, there will be no correlation between environmental seasonality and brain size (black area).

We expect the effects of energetic constraints to prevail if animals cannot move into other habitats, or if a dietary switch to explore hidden high-quality food sources is somehow prevented (e.g., extractive foraging). A high-energy consumption of the brain relative to total metabolism and a high extrinsic (unavoidable) mortality further reduce the possible benefits of a cognitive buffer. Thus, we expect cognitive buffer effects to be most apparent in the following categories of animals: (1) animals that can fly or swim and thus easily sample other habitats or move into other regions, such as birds, bats, and some classes of marine mammals; (2) animals that can more easily cope with minor reductions in food availability because their brains usurp only a relatively modest portion of the energy budget, in particular due to large body size, such as large carnivores; and (3) animals that rely on extractive foraging, food caching, or that exploit dispersed food patches. These predictions are in accordance with previous findings on birds. In temperate Palearctic temperate birds, cognitive buffer effects prevail over energetic constraints (Sol et al. 2005b; Sol et al. 2007). Furthermore, in South American parrots, there is a direct positive relationship between environmental seasonality and relative brain size (Schuck-Paim et al. 2008), also hinting at a prevalence of cognitive buffer effects over energetic constraints. In lineages lacking these features, especially smaller nonvolant mammals, the expensive brain effects should predominate, leading to a negative correlation between both environmental and XPR and brain size.

Both the cost and the benefits perspectives concern energy acquisition, that is, are explicitly ecological. They explain a reasonable amount of variation in brain size and thus support ecological approaches to brain size evolution (Byrne 1997), although they must be tested in more detail using direct measures of food availability instead of using environmental seasonality as a proxy. It is not clear whether the social benefits of brain size increases (Dunbar 1998) will account for additional variation in brain size once these ecological effects are factored in. Future studies should try to integrate all perspectives to assess their relative importance.

Overall, this study shows the importance of incorporating both costs and benefit perspectives in models on brain size evolution. In catarrhine primates, cognitive buffers just manage to level out the energetic constraints of the environmental seasonality. For any species, we must carefully consider the magnitude of these effects separately. The evolution of early hominins may be an example of how cognitive buffering can surmount energetic constraints. On the other hand, cognitive buffering may not be an option in a restricted island habitat like the one of Homo floresiensis or if severe nonperiodic droughts, such as El-Niño effect in the East Borneo for Pongo pygmaeus morio, lead to unavoidable periods of starvation.

Associate Editor: C. Farmer


We thank M. Genoud, C. Grueter, J. Pastorini, N. Ménard, and T. O’Brien for providing data. Furthermore, we thank C. Nunn for organizing the AnthroTree workshop, A. Bissonnette and M. Hamard for constructive comments and fruitful discussions, and two anonymous reviewers for their constructive comments on a previous draft of the manuscript. ECV were collected at the following museums: Zoological Museum, Zurich; Museum für Naturkunde, Berlin; Naturalis, Leiden; Zoological Museum, Amsterdam; Royal Belgian Institute for Natural Sciences, Brussels; Natural History Museum, Bern; Natural History Museum, Basel; Natural History Museum, Geneva; American Museum of Natural History, New York; Field Museum, Chicago; Museum of Comparative Zoology, Boston; and National Museum of Natural History, Washington, DC. This study was funded by the Swiss National Science Foundation grant 3100A0 –117789, with additional support from Synthesis and the A. H. Schultz Foundation.