Cooperation includes a wide diversity of traits, for example predator avoidance by warning signals or nest defense, resource acquisition through harvesting, communication or cooperative hunting, or cooperative breeding through provisioning of reproductive females or offspring (Dugatkin, 1997; Davies et al., 2012). Understanding the evolution of sociality and cooperation is a long-standing challenge, not least because both immediate direct benefits and genetic indirect benefits of cooperation may depend on ecological context and therefore vary with different environmental and ecological conditions. For example, cooperative breeding in birds is favored when ecological conditions are harsh and individual reproduction constrained, whereas individuals disperse and attempt own reproduction under benign conditions (Emlen, 1982, 1994). Our understanding of these relationships are, further challenged by a complex set of interactions between life history and environment on the probability of group formation and degree of social organization (Hatchwell and Komdeur, 2000). Geographical patterns of variation in the level of sociality has been documented in a number of social lineages, suggesting a strong influence on ecology, and environment on the evolution and complexity of cooperation, e.g., in social insects (Wilson, 1975; Kaspari and Vargo, 1995; Wcislo, 1997; Schwarz et al., 2007; Gunnels et al., 2008; Kaspari and Weiser, 2011; Kocher et al., 2014), social spiders (Avilés et al., 2007; Purcell, 2011; Majer et al., 2013a), and cooperatively breeding birds (Jetz and Rubenstein, 2011). While these patterns show altitudinal or latitudinal and thereby ecological correlations with degree of sociality, nest size or other social traits, our understanding of the causal relationships between ecological factors and social traits is incomplete.

In a study incorporating variation in ecological conditions into social evolution theory, Van Dyken and Wade (2012) point out that various social traits differ in their ecological function depending on the type of benefits they confer (fecundity/survival/resource acquisition benefits). Given that different ecological roles of cooperative traits are likely to influence selection is different ways, they show how different modes of cooperation evolve under distinct ecological conditions. Traits that contribute to resource acquisition (e.g., provisioning, group hunting, communicating location of food) evolve more readily under strong local competition among social partners, for example when resources are limited, or fecundity and/or survival causes high local crowding. In contrast, altruistic traits that confer survival and/or fecundity benefits, i.e., nest defense or worker altruism, are favored when conditions are reversed and there are abundant resources and weak local competition. If different modes of cooperation evolve under distinct ecological conditions, this framework may provide a useful way of examining the evolution of cooperative traits in different ecological settings (Van Dyken and Wade, 2012).

In the present study, we aim to explore the ecological conditions under which sociality and cooperation in spiders may evolve. We do this by determining the relationship between environmental factors and global geographic distribution of cooperative spiders, and relating the ecological factors that determine the distribution of social spiders to hypotheses of social evolution (Lubin and Bilde, 2007). Permanent cooperation in spiders is rare and known in only approximately 25 of almost 46.000 extant described species (World Spider Catalog, 2015). Interestingly, among these there are at least 20 independent origins of sociality distributed across seven families, and most species show remarkable convergent evolution of a suite of traits associated with their social life style (Lubin and Bilde, 2007; Bilde and Lubin, 2011). Social spiders live in permanent groups and cooperate in web building, prey capture, and brood care. The transition to permanent group-living involves complete elimination of pre-mating dispersal, which results in mating among siblings (family groups) in an obligatory inbreeding mating system. Groups show a female biased primary sex ratio, and new colonies are founded by adult mated females and her offspring. Furthermore, all social species are geographically restricted to subtropical and tropical environments (Avilés, 1997; Agnarsson et al., 2006; Johannesen et al., 2007). The combination of this distribution pattern and the multiple origins of convergent traits suggests that, common ecological factors may underlie or facilitate permanent sociality in spiders.

Benefits of cooperation is spiders appear to be associated with resource acquisition in the form of insect prey (Nentwig, 1985; Ward, 1986; Guevara and Aviles, 2007; Majer et al., 2013b). Social spiders live in sedentary colonies, and as sit-and-wait predators they are entirely dependent on the arrival of insect prey in their capture webs to secure sufficient food to meet demands of higher local densities. This implies that a certain supply of prey in the environment must be present for groups to form (Parrish and Hamner, 1997; Jarvis et al., 1998). Empirical evidence suggests that local competition increases with group size, and that the balance between improved survival but decreased fecundity with increasing group size results in highest fitness (per capita lifetime reproductive success) at intermediate group size (Avilés and Tufiño, 1998; Bilde et al., 2007). Traits that increase capture success, for example cooperative web building and prey capture, or cooperative brood care that facilitates provisioning of young, may be favored under these conditions. The social spider system may therefore follow the predictions of a limited resources situation for the evolution of cooperative traits that contribute to resource-enhancement or resource efficiency (Van Dyken and Wade, 2012).

We collected presence/absence data on a world-wide set of 21 social spider species to explore environmental factors that restrict the distribution of social spider lineages to the subtropics and tropics and identify ecological determinates of their limited distribution ranges. In addition, we performed spatial analysis of species diversity of social species. Comparative spatial analyses of closely related social and solitary sister species in the two most species rich genera, Stegodyphus and Anelosimus (Majer et al., 2013a,b), suggest that social species are confined to lower latitudinal gradients relative to their solitary and subsocial sister species. To corroborate this finding, we also performed spatial analysis of the distribution (presence/absence) on a data set of seven solitary sister species to social species of Stegodyphus and Anelosimus (Supplementary Material). We formulated three not mutually exclusive hypotheses relating to habitat productivity (prey abundance) or seasonality (prey constancy or constant environment), to identify environmental predictors of the distribution of social species.

Our study formally establishes the global distribution of permanently-social spiders in subtropical and tropical environments (Figure 1). Out of the 335 botanical regions, 50 had one or more social spider species occurring across the territory. Twenty four countries had one social species, while 26 countries had two or more social species. The maximum number of social species (n = 6; or 5 when excluding the “atypical” social Tapillinus spp.) was found in Ecuador (Figure 1).

The only environmental predictor for the global presence-absence of social spiders with strong support was GVI (Table 2). This implies that social spiders globally occur in more productive environments (Figure 2). However, results differ between the Old and New Worlds, in the Old World, the savannah and not the rainforest is the most species rich with the social species confined to the most productive savannah biome. In the New World, the occurrence of social spiders are in rainforests as the most productive biome. This finding was supported by both the non-spatial and the spatial models (Table 2). Notably, not just the GVI main effect, but also the GVI × Region interaction received strong support. GVI was best supported in both sets of models, and had a much stronger effect on the probability of occurrence of social spiders in the Old World (best modeled with as a quadratic function of GVI, Table 4). This effect of GVI could be attributed to the differences in the proportional occurrences of different species of social spiders across the Old and New World habitats (Figure 3). This effect was much stronger in the non-spatial models without the auto-covariate (Table 4). We cannot rule out that our results could be partly biased by differences in sampling effort between regions/countries, i.e., whether the higher species richness in Ecuador might reflect more intense sampling in this region. Therefore, we repeated the species richness modeling excluding the data from the most species rich region, Ecuador, this analyses reproduced and therefore corroborates our main results (Table S8 in the Supplementary Material).

In addition, we looked at global predictors of occurrence of solitary living sister species to the New World Anelosimus and Old World Stegodypus spiders (Table S5 in the Supplementary Material). Both sets of models, spatial and non-spatial, showed that subsocial spiders in the genera are constrained to more productive areas (Tables S6, S7 in the Supplementary Material), with productivity being the only supported predictor of their occurrences. However, these effects were much weaker than the predictive power of productivity for the occurrence of social spiders (Table 4), suggesting that resource availability (productivity) indeed is a main global driver of occurrence and distribution patterns of social lineages. Notably, there was no differential effect of productivity as a predictor of subsocial spiders' occurrences in the Old vs. New World. Unlike social Anelosimus spiders, that occur only in the Neotropical rain forests, American subsocial species in the genus occur more north and in much drier environments (e.g., in Mexico).

We analyzed the global patterns of occurrence and species richness of social spiders in relation to the hypotheses of habitat productivity and climatic stability as ecological predictors of social species' distribution ranges. We found strong support for the effect of habitat productivity on the distribution of social spiders. Within the overall range of social spiders, precipitation seasonality was the best supported predictor of species richness (two or more species), i.e., there was a tendency for higher species richness in areas of lower precipitation seasonality. This means that permanently social spiders occur in areas of high habitat productivity, with relatively high species richness concentrated within this range where precipitation is less seasonal. The permanently social species are derived from solitary subsocial sister species (Johannesen et al., 2007), but occupy more limited distribution ranges at lower latitudes and relatively more productive habitats than their subsocial sister species (Majer et al., 2013a,b). Our analyses of seven subsocial sister species confirms this pattern, which suggests that relative to the distribution ranges of close relatives with a solitary life style, the permanently social and derived species are confined to the most productive areas. More productive habitats support higher abundance of prey (Clarke and Gaston, 2006; Majer et al., 2013b) and/or higher abundance of larger prey size (Borer et al., 2012). Food availability is an important ecological driver of group formation (Davies et al., 2012), and has been hypothesized as a main factor restricting social spider species to tropical regions (Nentwig, 1985). Our results support the productivity hypothesis suggesting that food resources is an important driver for the evolution and maintenance of permanent group living.

Latitudinal gradients of decreasing species richness from tropical to extratropical areas is a well-established pattern in biogeography (Pianka, 1966; Willig et al., 2003; Hillebrand, 2004). Our results show increasing diversity of social spider species at lower latitudes and thus, corroborates this pattern, therefore we cannot rule out that higher species richness of social spiders at lower latitudes is a by-product of general higher diversity in these areas. An analyses including distribution ranges of a larger number of solitary species within the social genera would be needed for a more conclusive analysis of the bio-geographical patterns underlying the distribution of social spider species. Nevertheless, while our analyses provided support for the effect of productivity on the occurrence of social spiders on a global scale, there are differences between the New and Old World (Figure 3), and our finding that social species are restricted to lower latitudes relative to their solitary sister species provides confidence in our interpretation. The quadratic relationship in the Old World can be contributed to the occurrence of single social Agelena spp. taxon in central African rainforests (i.e., very productive habitats) vs. several social Stegodyphus spp. in the African dry savannah habitats (i.e., less productive habitats). The differences in the effect of productivity gradients between the Old and New World may reflect the major differences in the ecology of the two most species rich genera, New World Anelosimus and Old World Stegodyphus, where the former occur in the most productive continental biomes, tropical rainforests while the latter occupy drier savannah habitats (Figure 1). Nevertheless, within the distribution range of subsocial Stegodyphus sister species, we find the social derived species in lower latitudes and more productive areas (Majer et al., 2013b). A similar pattern applies to the New World Anelosimus species (Majer et al., 2013a), however, for Anelosimus spiders there was a strong combined effect of productivity with other environmental gradients, in particular precipitation seasonality that influenced the social species' distribution ranges. Therefore, to the extent that we can take the evolutionary relationships of social and subsocial species into account, our analyses confirm the importance of habitat productivity for maintaining sedentary groups of permanently social species.

Within the distribution of social spiders, species richness of social spiders is higher where precipitation seasonality is low, i.e., in aseasonal productive environments such as tropical rain forests. This finding is consistent with the hypothesis that permanent group living is favored not only where the prey supply is relatively high, but also where prey is continuous throughout the seasons. Precipitation patterns have been shown to influence the phenology of insect populations. A set of field studies in the tropical mainland and islands showed that insect populations increased in abundance where rain patterns were more constant (Janzen and Schoener, 1968; Janzen, 1973; Wolda, 1978). This implies that food supply for spiders is also more constant in such areas. The negative effects of precipitation seasonality on species richness inside the tropics might be driven by data on Anelosimus species across the Americas, which have overlapping generations and year-round colony productivity, and which are also the most abundant species in the dataset. Nevertheless, there was no significant effect of the interaction term of Old/New World categorical variable with precipitation. The pattern therefore holds regardless of the fact that the majority of species in the data set are social theridiids (Anelosimus and Parasteatoda spp., Table 1), which occur in the tropical rainforest areas of the Americas and Papua New Guinea.

The lowlands of Amazonian Ecuador harbour six social spiders in total, thus being the most species-rich area (Figure 1). The six species include three Anelosimus spp., 1 dictynid species (Table 1), while the 6th one is the “atypical” social oxyiopid Tapinillus sp. Notably, this is where most of the social theridiid species occur, namely those of the genus Anelosimus. There is a latitudinal pattern of sociality in this genus, with more social species at lower latitudes, which is also paralleled in the altitudinal gradient, as the social species rarely occur above 1500 m (except for A. gucamayos that occurs up to 2000 m) (Avilés et al., 2007). Interestingly, there is also an altitudinal gradient in colony size for two social Anelosimus species (Avilés et al., 2007). Unlike the ants, where colony size increases with latitude (Kaspari and Vargo, 1995), the opposite seems to appear in the spider genus with the most social species. Two patterns have been suggested for these distributional phenomena of social traits in the genus Anelosimus: absence of subsocial congeners in the lowland tropical rainforests, and absence of social congeners in the highlands (Avilés et al., 2007). The authors suggested the prey size and maternal survival contribute to the occurrence and survival of social lineages in the lowlands, where the prey is sufficiently large in size, and where heavy rains affect the spiders in such a way that the survival of offspring is higher in multi-female and cooperative nests. The macro-ecological correlates of sociality in the Anelosimus genus lend support for these biological hypotheses (Majer et al., 2013a), and are in agreement with the results supporting productivity as the most important ecological predictor of the global distribution of social spiders reported here.

We hypothesized that temperature seasonality should favor the occurrence of social spiders by providing a constant environment, but found no support for this effect, indicating that sociality in spiders overall does not correlate with a thermally constant environment. This seemingly lack of effect of temperature seasonality on the distribution pattern of social spiders is perhaps not so surprising as a subset of all the social spider lineages, the genus Stegodyphus, occur across a set of seasonal habitats, spanning the African continent and Indian subcontinent (Majer et al., 2013b). Social Stegodyphus show a remarkably wide distribution range in terms of temperature gradient, which for example on the Indian subcontinent spans from Himalayas to Sri Lanka (Settepani et al., 2014). Given the strongly inbreeding mating system and very limited genetic variation of social spiders, their wide distribution range in terms of temperature is intriguing and suggests either strong plasticity in habitat tolerance or local adaptation (Settepani et al., 2014).