Cognitive traits and their underlying neural substrates may be shaped by natural selection. Accordingly, specialized food-hoarding birds have long been hypothesized to have larger hippocampal (HC) volumes than non-hoarding species (review in Healy et al. 2005). However, studies investigating the relationship between HC volume and food-hoarding specialization provide conflicting results. For example, Brodin & Lundborg (2003) found no significant support for the hypothesis in either the Corvidae or the Paridae. On the other hand, Lucas et al. (2004) showed that the HC volume of hoarding species was larger than the HC volume of non-hoarders in both families, but only when a previously undescribed continent difference was statistically removed.

In a separate study, Garamszegi & Eens (2004) performed a phylogenetic analysis of 55 bird species from 13 families. With this larger dataset, they showed that the evolutionary change from non-hoarding to specialized hoarding has been accompanied by an evolutionary increase in HC volume. Thus the HC/hoarding relationship appears to be valid in a broad evolutionary context involving distantly related avian families, even outside the Corvidae and Paridae. However, only Lucas et al. (2004) addressed the potential for continent differences. Given the limited taxonomic scope of the Lucas et al. (2004) study, it is still unclear whether the continent effect is restricted to avian families characterized by hoarding specialization or whether the effect is more broadly distributed across other families.

In this paper we test the generality of the ‘continent effect’, relying on the extended dataset of Garamszegi & Eens (2004). We also examine whether the continental differences in HC volume are independent of food-hoarding specialization. We make an effort to identify the role of an ecological factor that may cause continental variation in HC volume. We assess the importance of temperatures that wintering species experience at their northernmost distributions that may affect food hoarding or HC volume. We hypothesize that in cold wintering conditions, animals may hoard more food, retrieve their caches more efficiently and perform better in spatial memory tasks than animals wintering in milder climates (see also Pravosudov & Clayton 2002). Therefore, species that are efficient at exploitation of food resources through specialized food hoarding (or other behavioural traits requiring a larger HC) may be more likely than non-hoarding species to disperse to colder climates. Hence, winter temperatures may shape continental effects, if food hoarding and wintering conditions vary consistently along the continents. If Eurasian species have to adapt to colder winters than North American species, then Eurasian birds are expected to specialize in food hoarding and evolve larger HC volumes than North American birds.

We characterized the extent of food hoarding by using the standard three-point scale from Garamszegi & Eens (2004). Species not known to hoard were considered non-hoarders (score of 1), whereas species that store small amounts of food for short time-intervals were categorized as non-specialized hoarders (score of 2). Species that exhibit typical hoarding behaviour by storing thousands of items in the autumn for later recovery were treated as specialized hoarders (score of 3). Since intermediate states are biologically meaningful, food hoarding was treated as a continuous variable. However, food hoarding can be viewed as a categorical variable. Therefore, by collapsing non-specialized and specialized hoarders, food hoarding was also used as a two-state categorical variable.

To obtain wintering temperatures, we first estimated winter distribution ranges from continent-specific field guides (Heinzel et al. 1997; National Geographic 2000). Then, we calculated latitudinal and longitudinal geographical coordinates at the maximal northernmost distribution by the use of a geographical map to the nearest 1 degree. These estimates are significantly repeatable between sources (e.g. Cramp & Simmons 1977–1994; Poole et al. 1993–2002) and between observers (latitude, R=0.928; longitude, R=0.975). We preferred using information from field guides because the resolution of distribution maps are similar in continent-specific field guides, whereas the differences in the quality of maps between Cramp & Simmons (1977–1994) and Poole et al. (1993–2002) may raise biases for our specific purposes here. Similar measures of distributions provided biologically meaningful measures (Gaston & Blackburn 1996). Second, based on the geographical coordinates, average January temperatures were estimated to the nearest degree 1°C by using electronic sources from the RI/LDEO Climate Data Library Web site (http://iridl.ldeo.columbia.edu). Throughout the study, we assumed that when longitudinal effects are also taken into account, the northernmost winter distribution reflects the coldest temperatures that wintering species experience. Given their migratory habits or their southern distribution, quantitative data on the winter distribution for seven species (Ficedula albicollis, Molothrus badius, Molothrus bonariensis, Molothrus rufoaxillaris, Phylloscopus trochilus, Sylvia borin and Urocissa erythrorhyncha) were unavailable.

Figure 1 (a) Relative hippocampal (HC) volume (corrected for telecephalon) and food-hoarding specialization in Eurasian and North American birds. (b) Relative HC volume and continental distribution when hoarding families (Corvidae, Paridae, Picidae and Sittidae) (more ...)

Our expanded dataset supports the conclusion of Lucas et al. (2004) that Eurasian birds have larger HC volumes than North American birds. However, this trend could be caused by continental differences in HC volumes expressed only in the caching birds. To evaluate the extent of the continent effect, we deleted all families known to hoard and tested for the continent effect in the rest of our dataset. The effect remained significant, even when phylogeny was controlled for (F_1,20=15.756, p<0.001, p_phyl=0.039; figure 1b). Thus, the continent effect is broadly distributed across the avifauna of these two continents.

Figure 2 The relationship between January temperatures at the northernmost wintering distribution of species and relative HC volume. Regression lines and sample sizes are given.

Our phylogenetic analyses indicate that the continent effect on relative HC volume found by Lucas et al. (2004) in corvids and parids is a general phenomenon in birds. It seems that relative HC volume is larger in birds inhabiting the Eurasian subcontinent than in birds from North America. We also found a significant relationship between continental distribution and food hoarding. This result may suggest that differences in food-hoarding specialization between the continents promote differences in relative HC volume. However, when food hoarding is statistically controlled for, the association between continental distribution and relative HC volume is still significant. In addition, the relative volume of the HC is significantly smaller in North America than in Europe when involving typically non-hoarding avian families only (figure 1b). Therefore, continental differences in food-hoarding specialization cannot lead to the phenomenon that Eurasian species have larger HC volumes. This pattern seems to be a general phenomenon among birds and it is not an attribute of food-hoarding families.

Lucas et al. (2004) suggested that systematic methodological differences between different laboratories in how brain measurements were taken may result in differences in HC volume between continents. However, a laboratory effect may only play a minor role because different laboratories are collaborating; analysing data from different laboratories separately gives similar trends for the HC/hoarding relationship (Lucas et al. 2004). In addition, the high between-studies repeatability of relative HC volume also argues against this explanation (Garamszegi & Eens 2004).

Lucas et al. (2004) also proposed that the phylogenetic distribution of birds may be responsible for the observed differences in HC volume between continents. However, the phylogenetic analyses presented here and in Garamszegi & Eens (2004) show that phylogenetic effects are unimportant confounding factors.

As a third possibility, Lucas et al. (2004) put forward the hypothesis that there truly is some continent-wide difference in an ecological factor or a life history trait that mediates differences in the scaling of brains. However, the role of these potential factors remained undetermined. Here we hypothesized that the wintering conditions of species may be different between Europe and North America and that these differences may select for different behavioural adaptations that are ultimately reflected by differences in relative HC volumes. We found a significantly negative relationship between winter temperatures and relative HC volumes at the raw species level, which was non-significant when we controlled for phylogenetic associations. This may indicate that the effect of wintering temperatures on relative HC volume, if it exists at all, is small and, thus, goes undetected in our small sample when controlling for common ancestry. It is also possible that January temperatures at the measured locations do not reflect the coldest temperatures that wintering species experience. However, we conclude that winter temperatures cannot be the primary basis of the continent effects because wintering temperatures did not differ between the continents. We suggest that differences in the degree of food hoarding and relative HC size in Europe and North America should be sought among other ecological factors and life-history traits that are not mirrored by winter temperatures. A comprehensive analysis of the role of biological factors should include a broad range of ecological or life history traits that can be related to HC volume and continental distribution and, thus, potentially shape the continent effect. Ideally, if the mediator variable is found, the continental effects should drop out of significance in the model.

The continent effect on relative HC volume, which is independent of food hoarding and winter temperatures, implies that other behavioural tasks requiring space in this neural substrate also allow species to adapt to environmental factors that vary across continents. Thus, food hoarding seems to be just a part of a complex cognitive adaptation process governed in part by the HC. The evolutionary role played by the HC may be even more important than previously thought.

K. Bledsoe, A. Boyko, M. Eens, B. Fanson, C. Jonason, T. Garland, K. Henry, J. Lanier, L. Lohmann, Mark Nolen and R. Van Damme provided useful comments. L.Z.G. was supported by a postdoctoral grant from the FWO-Flanders (Belgium). Part of the work was funded by an NIMH grant to J.R.L., Nicky Clayton and Hubert Schwabl.