We showed that there is large variation in absolute brain volume, relative brain volume and relative volume of the telencephalon, optic tectum and cerebellum across wild nine-spined stickleback populations. Brain size patterns in the wild show habitat specificity both in absolute and relative scales: pond fish have larger brains than marine fish. Further, we found a marginally significant trend in the relative telencephalon size: marine fish tend to have larger telencephala than pond fish. The hypoallometric relationship between brain size and body size (slope = 0.5) is in accordance with a previous study on tropical fish . We also found that wild-caught pond fish have larger brains than laboratory-reared pond fish, whereas no differences were observed between wild-caught and laboratory-reared marine conspecifics. The relative sizes of all brain parts were smaller in common garden than in the wild in all populations. These findings indicate that even though large brain size and brain part size variation exist in the wild, both in absolute and relative terms, patterns in nature may differ from those gathered in a standardized common garden and in some cases even in a habitat-dependent way. This strongly suggests that environmental effects on brain development can obscure and confound evolutionary inference based on purely phenotypic data collected from the wild. Hence, our results under-line the importance of not basing evolutionary inference on phenotypic patterns of brain size variation unless the environmental sources of variation have been controlled for - a point reinforced by other studies focussing on differentiation in morphological and life history traits [38–40].
We found large habitat-specific population variation in absolute brain size: all marine populations (and a single lake population) had similarly sized brains that were nearly twofold smaller than those of pond fish. Within the pond habitat, there was large variation in average brain size. Although most studies have investigated only relative brain size variation (correcting brain size for variation in body size), absolute brain size can also account for differences in behaviour and/or habitat use. This is evident in comparisons of closely related species [see e.g. ], as the brains of these species tend to be more similar than those of distant taxa. Indeed, absolute brain size variation is routinely utilized in studies of primate and human evolution [e.g. ]. In general, increased brain size is attributable to an increase in neuron number, and not in neuron size . Further, increases in absolute brain size result in decreased proportional connectivity . Obviously, larger bodies need larger brains to be controlled at a similar level . Hence, it is not surprising that pond populations that have evolved to giants [34, 35] have also much larger brains than smaller sized marine or lake populations.
Previous findings have demonstrated a shallower hypoallometric slope at the intraspecific level and among closely related species than across broader taxonomic groups (mammals, intraspecific: 0.2-0.4, broad interspecific: 0.66, ; fishes, intraspecific: 0.44, intrafamiliar: 0.5, broad interspecific: 0.66, ). In accordance with these results we found a hypoallometric relationship between brain and body size with a slope = 0.5. In mammals, it has been demonstrated that on a broad evolutionary time scale, there has been greater net directional selection on brain size than on body size, while the short-term differentiation in brain vs. body size in closely related mammalian taxa has resulted from directional selection acting mostly on body size with changes in brain sizes being largely correlated responses . Further, Gonzalez-Voyer et al.  demonstrated in Tanganyikan cichlids that even strongly correlated traits, such as brain and body size, can evolve independently from each other, and that body size may be under stronger selection than brain size during adaptive radiation. In the case of the nine-spined stickleback system, habitat-dependent body size diversification has been demonstrated [34, 35], and body weight differences among these recently differentiated populations can be tenfold. Hence, it seems feasible to suggest that the observed brain size divergence might have been a correlated response to body-size divergence.
To assess body-size-independent brain size trends, we also investigated brain size differences relative to body size. Similarly to results for absolute brain size, and in contrast to our expectations, pond sticklebacks had relatively larger brains than marine sticklebacks. Intraspecific variation in relative brain size and brain architecture also appears to be strongly correlated with different ecological factors and/or life history traits. For example, environmental harshness has been shown to correlate positively with the size and neuron number of the brain region linked with memory storage (hippocampus) in the black-capped chickadee, Poecile atricapillus, a food caching species for which good memory can be essential for survival . Garamszegi & Eens  found a positive correlation between song length and repertoire size and both relative and absolute volumes of different song nuclei. By comparing two subspecies of the white-crowned sparrow, the migratory Zonotrichia leucophrys gambelii and non-migratory Z. l. nuttalli, Pravosudov et al.  found that migratory subspecies had larger hippocampus and more hippocampal neurons. Habitat-independent, genetically based intraspecific variation in brain architecture has also been found both in wild guppy (Poecilia reticulata) populations reared in common environments , and in laboratory lines of the medaka (Oryzias latipes; ).
Marine nine-spined sticklebacks are members of a diverse fish fauna, and as such, are faced with numerous predators and interspecific competitors. In contrast, pond fish communities are much simpler. Structural heterogeneity in the pond environment is also much lower than that found in marine environments. These environmental factors are all known to be important in shaping brain evolution. For instance, predation pressure has been shown to affect brain size evolution in Mallorcan bovids , diet affected brain size evolution of bats , both environmental complexity and social features sculpt the brain architecture of cichlid fish , while living in large and socially complex groups is the most accepted hypothesis for the evolution of the extremely large brain of humans . Hence, we expected selective pressures stemming from predation, interspecific competition, and habitat complexity to result in relatively larger brains in marine populations. Moreover, assuming that body size divergence (pond fish > marine/lake fish [34, 35]) preceded correlated brain size divergence, we also expected pond fish to have similar or smaller brains, in relative terms, than marine or lake fish. Our previous common garden experiment based on a subset of the populations used here revealed no habitat-dependence in relative brain size . Therefore, the pattern found in the current study (pond fish > marine fish) is highly unlikely to be a result of selection on brain size itself. Further, while we found no habitat-dependence in the common garden setting, strong population differentiation in relative brain size in a habitat-independent way was detected (selective force unknown; ). Therefore, the plasticity resulting in the habitat-dependent wild vs. common garden difference cannot be habitat-specific itself. In a controlled laboratory experiment we found that group rearing had a negative effect on brain development in pond but not in marine fish . Hence, the hypothesis that the aggressive, bold and antisocial pond fish [47, 48] have larger relative brain sizes due to ontogenetic phenotypic plasticity as a response to fierce intraspecific competition must be rejected. Another possible explanation for larger relative brains in pond than in marine populations can be found from differences in ontogenetic allometry: pond fish living under negligible predation can become twice as old as marine fish , and an ontogenetic change in body vs. brain growth might explain this pattern. However, this issue requires further investigation.
Not only absolute and relative brain size, but also the relative size of different brain parts of nine-spined sticklebacks varied in the wild. Significant population differences were found in the relative sizes of the telencephalon, optic tectum and cerebellum. Further, we found marginally significant (P < 0.08) habitat-specificity in the relative size of the telencephalon, with marine fish tending to develop larger telencephala than pond fish. This is in accordance with results from our previous common garden study . The telencephalon is larger in monogamous species, and shows a trend towards a positive correlation with rock size in the habitats in Tanganyikan lake cichlids , suggesting that both social and environmental heterogeneity may select for larger telencephalon. However, quite surprisingly, generalist limnetic populations of three-spined sticklebacks (Gasterosteus aculeatus) that use plankton as a main food source have larger telencephala than benthic-foraging populations of the same species as measured in samples from the wild . The optic tectum is relatively larger in fish that prey on fish or other fast-moving prey, and clear water fishes develop larger optic tecta than species inhabiting turbid waters . Cerebellum size correlates positively with the number of sympatric species in a fish community, and hypothalamus size is larger in monogamous than polygamous cichlids . However, we did not find habitat specificity in the relative size of the optic tectum or the cerebellum, neither in the present, nor in the previous common garden study .
There are some incongruence between the present study and our previous work . Here we found habitat-specific brain size divergence and population divergence in relative optic tectum size that was not seen in the common garden study. Only the habitat-dependence of relative telencephalon size found in the common garden study could be detected in the data from wild fish. A direct comparison between common garden and wild brains from the same populations revealed a habitat-dependent effect: pond (but not marine) fish had relatively larger brains in the wild than in the common garden. Further, the relative size of all brain parts was smaller in the laboratory than in the wild, perhaps due to a stimulus-poor environment during brain development. The most plausible explanation for the differences among common garden and wild data resides on phenotypic/ontogenetic plasticity in brain architecture. The potential for plastic responses to environmental heterogeneity is very high in fish [49–51]. Neurogenesis persists long into adulthood in fish [51–53] and contributes to lifelong growth of the brain. Hence, the fact that pond fish can live nearly twice as long as marine or lake fish may result in bias originating from plain ontogenetic plasticity or allometry. Furthermore, local random environmental variation may induce plasticity that could conceal genetic trends. Therefore, common garden studies seem to be of particular importance in studies of brain evolution. For instance, in this study system erroneous evolutionary conclusions could be drawn from the habitat-specificity (implying local adaptation) of relative brain size in the data from the wild given that observed differences cannot be reproduced under common garden conditions (showing that the differences are environmentally induced).
Finally, we showed that relative brain size and brain architecture are different between wild-caught and common garden sticklebacks from the same populations. The negative effect of domestication on brain size is well known both as a result of genetic adaptation and phenotypic plasticity [54, 55]. In a recent paper, Burns et al.  demonstrated that laboratory rearing caused a significant decrease in the relative brain size of guppies (P. reticulata). Interestingly, we found that laboratory rearing had a negative effect on brain size in pond but not in marine nine-spined sticklebacks. The reason for this difference is unknown and warrants further investigations. We also found that all brain parts (corrected for both body and brain size) were smaller in common garden than in nature, a pattern congruent with general expectations. The reason for this can be a phenotypically plastic response to the comparatively stimulus poor laboratory environment.