Friendship across species borders: factors that facilitate and constrain heterospecific sociality

Abstract

Our understanding of animal sociality is based almost entirely on single-species sociality. Heterospecific sociality, although documented in numerous taxa and contexts, remains at the margins of sociality research and is rarely investigated in conjunction with single-species sociality. This could be because heterospecific and single-species sociality are thought to be based on fundamentally different mechanisms. However, our literature survey shows that heterospecific sociality based on mechanisms similar to single-species sociality is reported from many taxa, contexts and for various benefits. Therefore, we propose a conceptual framework to understand conspecific versus heterospecific social partner choice. Previous attempts, which are all in the context of social information, model partner choice as a trade-off between information benefit and competition cost, along a single phenotypic distance axis. Our framework of partner choice considers both direct grouping benefits and information benefits, allows heterospecific and conspecific partners to differ in degree and qualitatively, and uses a multi-dimensional trait space analysis of costs (competition and activity matching) and benefits (relevance of partner and quality of partner). We conclude that social partner choice is best-viewed as a continuum: some social benefits are obtainable only from conspecifics, some only from dissimilar heterospecifics, while many are potentially obtainable from conspecifics and heterospecifics.

This article is part of the theme issue ‘Collective movement ecology'.

1. Introduction

Our current view of animal sociality (italicized terms are defined in Glossary) comes, almost entirely, through the lens of single-species sociality, i.e. sociality involving individuals of the same species. Heterospecific sociality, i.e. sociality involving individuals of multiple species, although documented from numerous taxa and ecological contexts [1–5], has rarely been examined theoretically in conjunction with single-species sociality (but see [3,4]), and mostly remains at the margins of animal sociality research. For example, textbooks (e.g. [6,7]; but see [8,9]) and review papers (e.g. [10,11]) on animal sociality provide cursory mention of heterospecific sociality, and do not integrate it into conceptual frameworks used to understand animal sociality in general. Why is this so? We believe this is because single-species and heterospecific sociality research have tended to emphasize different aspects of social partners (e.g. [6,12]): the former has focused on the importance of similarity among social partners whereas the latter has tended to emphasize the differences. Not surprisingly therefore, heterospecific sociality is thought to be based on mechanisms fundamentally different from single-species sociality (e.g. [12]), despite a huge body of empirical evidence indicative of the contrary (see review below). The only exception is research on social information use, where a few studies have modelled the costs and benefits of social information from conspecifics compared with heterospecifics [3,13,14]. Our paper has two goals: to document the diversity of heterospecific sociality that is based on mechanisms similar to single-species sociality, and to provide a conceptual framework, applicable to both social information benefits and direct grouping benefits, to understand how individuals choose a conspecific or a heterospecific as their social partner.

In the first part of this paper, we systematically survey the published empirical literature on heterospecific sociality to document the diversity of heterospecific sociality that is likely to be based on mechanisms similar to single-species sociality. Our survey reveals numerous cases of this type of heterospecific sociality, in a variety of taxa and contexts, for many different benefits, both in the form of social information benefits as well as direct grouping benefits (from here on, we use social benefit as a generic phrase to indicate either social information benefit or direct grouping benefit). Therefore, there are likely to be numerous scenarios in which the same social benefit can potentially be obtained from either conspecifics or heterospecifics. Given that any organism is likely to be surrounded by conspecific and heterospecific individuals, how is a social partner chosen?

In the second part of the paper, we address this question by providing a conceptual framework to understand factors that influence choice of conspecific or heterospecific as social partner, either for social information or direct grouping benefits. Earlier such attempts have all been only in the context of social information use [3,13,14]; they model partner choice as a trade-off between information benefit and competition cost, as a function of phenotypic distance of the partner: the more similar a partner is, more beneficial is the information it provides, but greater are also the costs of competition. Consequently, depending on the phenotypic distance at which information use is optimal, the partner chosen will be either a conspecific or a heterospecific individual. In other words, the difference between heterospecific and conspecific partners is only one of degree. We, however, argue that a general model of social partner choice needs to account for the possibility that heterospecifics that are qualitatively different from conspecifics can, nevertheless, provide social benefits through the same mechanisms as conspecifics. Our conceptual framework modifies existing social information use models in ways that make it generally applicable to any situation involving social partner choice from among both conspecifics and heterospecifics.

2. Review of heterospecific sociality

(a). Methods

We used keyword searches in Google Scholar to find peer-reviewed publications on heterospecific sociality, i.e. spatial and/or temporal clumping of organisms as a result of one- or two-way social attraction between organisms. Each search combined two words: one word from among ‘heterospecific', ‘mixed-species', ‘interspecific' and ‘multi-species' and another from ‘group', ‘social*', ‘association', ‘aggregation' and ‘cooperation', resulting in 20 keyword combinations. Our choice of keyword combinations was based on terms commonly encountered in the heterospecific sociality literature. We scanned papers returned by our searches (980–1000 publications for each keyword combination) and identified those in which authors of the original paper had indicated occurrence of some type of heterospecific sociality (e.g. groups that are based on social attraction, use of social information, exhibition of cooperative behaviours) that conferred benefits on all or some parties involved, and is not known to negatively affect any parties involved. We did not include examples of aggregations that are caused purely due to clumped resources, but included social aggregations. Our interest here was only in heterospecific sociality that is based on mechanisms similar to single-species sociality, i.e. social benefit is in the same currency for all social parties involved (e.g. dilution of predation risk, communal mobbing, copying foraging locations, huddling for warmth, grouping to avoid desiccation, collective navigation, use of social information on resources, predators and environment). We excluded cases that are solely based on benefits in different currencies for different social partners (e.g. one partner gets food, other gets protection), cases of trophic interaction (e.g. plant–animal mutualisms such as plant-disperser and plant-pollinator) and instances of interspecific mimicry. However, we included cases involving multiple currencies if at least one of the currencies was shared by all social partners (e.g. both partners get protection, but in addition, one of the partners also gets food). For each reported case, we extracted information on type of heterospecific sociality seen, taxa involved, ecological context of sociality and likely benefit obtained. The aim of our literature survey was to examine the diversity of heterospecific sociality that is based on mechanisms similar to single-species sociality, and not to provide an exhaustive list of all documented cases. Therefore, from the information extracted from the literature survey, we identified unique combinations of taxa, ecological context and type of heterospecific sociality, which we present below. We did not include the nature of benefit in our categorization because, for most documented cases of heterospecific sociality to date, benefits are either unknown or only hypothesized, and require confirmation from further studies. For taxon information, we focused on ecologically distinct taxonomic groupings at the level of taxonomic order or above.

(b). Findings

Our survey revealed 203 unique combinations of taxa (at order level or above) and ecological contexts of heterospecific sociality that are likely to be based on mechanisms similar to single-species sociality (see electronic supplementary material, table S1, for a full list of 203 references). Of these, 21 are known commensalistic relationships (electronic supplementary material, table S1). The type of interaction in the remaining 182, i.e. whether mutualistic, cooperative, commensalistic, etc., needs to be confirmed by further research. Taxa range from bacteria and fungi, to a variety of marine and terrestrial invertebrate groups, as well as all major vertebrate groups (fish, amphibians, lizards, snakes, birds, mammals) on land, freshwater and marine habitats, and at different life cycle stages including eggs, young ones, juveniles and adults (electronic supplementary material, table S1). We also found reports of heterospecific synchrony (spatial and temporal) in plant phenology for pollinator attraction, and spatial clustering of plants for protection from herbivory, which seem to be based on principles similar to animal sociality (electronic supplementary material, table S1). Ecological contexts in which heterospecific sociality is seen include basking, courtship, foraging, parental care, resting/sleeping, travelling, etc. (table 1). In these contexts, heterospecific sociality was expressed in the form of moving social groups, social aggregations, social information use without group or aggregation formation, or the use of artefacts produced by heterospecifics (table 1). Reported benefits for heterospecific sociality include most of those reported for single-species sociality, such as food, protection, thermoregulation, avoiding desiccation and reducing energetic costs, but protection from predators is, by far, the most common (table 1). One conclusion that emerges from our literature survey is that there are numerous scenarios in which the same social benefit, can, in principle, be obtained both from conspecifics and heterospecifics. In §3, we propose a conceptual framework to understand the factors that are likely to influence social partner choice in such scenarios.

Table 1.

Summary (ecological context, number of taxa (at the level of taxonomic order and above) and benefits obtained) of reported cases of heterospecific sociality in peer-reviewed literature that are likely to be based on mechanisms similar to those underlying single-species sociality. See electronic supplementary material, table S1 for taxa names and supporting references.

ecological context	no. taxa	potential benefits
foraginga	18	food, protection
multiple contextsa	9	allogrooming, food, mate finding, play, reducing energy costs
travelling (short- and long-distance)a	6	navigation; protection; reducing energy costs
aerial breathingb	1	protection
baskingb	2	protection
clay lickingb	1	protection
colony living (others)b	4	food, protection
colony living (social invertebrates)b	4	brood care
drinking waterb	1	protection
foragingb	17	food, protection, reducing host defence
lekkingb	1	protection
mud-puddlingb	1	protection
multiple contextsb	34	allopreening, brood care, protection
resting/sleepingb	27	allogrooming, avoiding desiccation, avoiding hypoxia, avoiding physical damage, protection, thermoregulation
singingb	2	resource defence, mate attraction
social aggregation of eggsb	7	brood defence, protection
social aggregation of larvae/pupae/young onesb	2	faster development, protection
social aggregation of nestsb	6	brood care, protection
spatial/temporal synchrony in floweringb	1	attracting mutualists
web buildingb	1	food, protection
cooperative hunting	11	cooperative hunting
defending territories	2	territory defence
mobbing	3	protection
food patch selectionc	11	food
habitat selectionc	4	habitat
breeding site selectionc	2	breeding site
safe site selectionc	24	protection
ovi/larviposition site selectionc	3	site for oviposition/larviposition
nest site selectionc	1	site for nest location
shelter selectionc	1	shelter
use of artefacts	8	food, nest, shelter

^aHeterospecific sociality takes the form of moving groups.

^bHeterospecific sociality takes the form of social aggregations.

^cHeterospecific sociality takes the form of social information use without group or aggregation formation.

3. A conceptual framework for social partner choice

We begin with a brief description of existing models of partner choice for social information use [3,13,14]. The broad common rationale underlying these models is as follows: organisms can obtain information about resources, predators or environment either on their own (personal information) or by observing decisions made (and/or their consequences) by other organisms (social information). Use of social information, however, entails a trade-off: the benefits of social information and competition cost decrease with phenotypic distance. The more similar a partner is (e.g. in resource use or shared predators), more beneficial is the information it provides, but the greater is also the competition cost it imposes. In addition to competition cost, organisms using social information will also incur fixed costs, in the form of neural processing of information, social learning, etc. [3]. Therefore, the net benefit obtained through social information use will be the difference between the benefit of the information obtained and the total costs of competition and social information use. Organisms will use social information whenever this net benefit is greater than the net benefit obtained using personal information. Depending on how information benefit and competition cost decrease with phenotypic distance, the optimal partner will be either a conspecific or a heterospecific individual.

Although models of social information use [3,13,14] capture the central trade-off underlying social partner choice, they only apply to scenarios where the difference between conspecifics and heterospecifics is a matter of degree. We, however, argue that a general model of social partner choice needs also to account for the possibility of heterospecifics and conspecifics being different qualitatively. Starting with existing social partner choice models, we describe two ways in which expectations regarding benefits (points (i) and (ii) below) and two ways in which expectations regarding costs (points (iii) and (iv) below) need to be modified. We argue that a model of social partner choice must account for trade-offs along four axes of phenotypic distance that influence cost of competition, cost of activity matching, relevance of partner and quality of partner. We propose a unified and generic single framework for social partner choice that is applicable both to direct grouping benefits and social information benefits.

(a). Modifications of existing social information use models

(i). Social benefit and competition cost can arise from different traits

Social information use models [3,13,14] examine the information benefit/competition cost trade-off along a single axis of phenotypic distance. They implicitly assume either that the same trait influences both information benefit and competition cost, or that different traits are so strongly correlated that it is okay to use a single composite measure. However, neither of these need be true. Sociality could be in response to a resource need, a threat or an abiotic variable. Only when the driver of sociality is a resource (e.g. copying foraging locations of another organism, using social information to select nesting sites and habitats) are traits important for social benefit and competition cost the same. For example, when a bird uses information on the location of another bird's nest as a cue for habitat quality, both social benefit and competition cost depend on overlap in nesting requirements [15]. On the other hand, when the driver of sociality is a predator or an abiotic variable, traits important for social benefit and competition cost need not be the same. Consider a forest bird grouping with another bird for risk dilution during foraging. For risk dilution to work, these birds have to share predators. From earlier work (e.g. [16]), we know that more similar two organisms are in body-size, the more likely are they to share predators (figure 1a). On the other hand, competition cost is likely to be related to foraging behaviour, i.e. the more similar two birds are in foraging behaviour, the higher is the competition cost (figure 1b). Therefore, in our example, traits underlying risk dilution benefit (similarity in body-size) and competition cost (similarity in foraging behaviour) are not the same. Although body-size and foraging behaviour are likely to be strongly correlated among conspecifics, the same is unlikely to be the case among a set of heterospecifics. Therefore, a bird looking for a social partner for risk dilution could pick a heterospecific that is similar in body-size but different in foraging behaviour.

Figure 1.

We propose that social partner choice must be examined as a function of phenotypic distance along four axes that determine relevance of partner, quality of partner, cost of competition and cost of activity matching, respectively (see Glossary for definitions). In general, we expect relevance of partner and cost of competition to decrease with phenotypic distance, and cost of activity matching to increase with phenotypic distance. Quality of partner can increase, decrease or change in a way unrelated to phenotypic distance. The shapes these relationships will take depend on the particular scenario in which sociality is required and the specific traits that underlie costs and benefits. (a)–(d) One set of example relationships for (a) relevance of partner, (b) cost of competition, (c) quality of partner and (d) cost of activity matching. In (b), red line depicts a situation where competition occurs only with conspecifics, and therefore cost of competition drops to zero at the transition from conspecific to heterospecific partners (indicated by vertical dotted line). Based on these relationships, one can calculate the net social benefit provided at each point in a multi-dimensional trait dissimilarity space. (e) An example net social benefit map built using only two axes: relevance of partner and cost of competition. Using the functional forms in a and b (black lines), net social benefit at each point in the space was calculated additively (i.e. net social benefit = relevance of partner – cost of competition). Circles represent conspecific individuals; squares and triangles represent heterospecific individuals of two different species. In this example, net social benefit provided by heterospecifics is represented by squares > conspecifics > heterospecifics represented by triangles.

Cost–benefit trade-offs, therefore, need to be modelled using separate axes of phenotypic distance for social benefit and competition cost in scenarios where a partner is being chosen from among conspecifics and heterospecifics. Other potential empirical examples of such scenarios are the numerous documented cases of social information use between birds and mammals for protection (e.g. jay–squirrel [17]; go-away bird—dik-dik [18]), where social partners, though ecologically very different, still share the same predators. In these cases, it will be worthwhile examining whether competition prevents these species from seeking information from conspecifics.

(ii). Social benefit is a function of both relevance of partner and quality of partner

Earlier models [3,13,14] assume that social benefit is only a function of relevance of partner: a potential partner at shorter phenotypic distances (i.e. higher ecological overlap) produces more relevant, and therefore more beneficial, social information. We, however, argue that, in addition to partner relevance, social benefit (whether in the form of social information benefit or direct group benefit) provided by a partner is also influenced by the quality of a partner, i.e. a partner's efficacy in fulfilling its role in relation to the specific mechanism through which social benefit is obtained. Equally relevant partners can vary in their efficacy in providing social benefit. For example, let's again consider the case of a forest bird seeking risk dilution while foraging. We already saw that the trait important for partner relevance in this case is body-size, because similar body-sized birds are more likely to share predators [16]. However, even among the set of equally relevant potential partners (i.e. same level of predator overlap with focal bird), a bird should choose to group with those that are most phenotypically similar to itself, to avoid the oddity effect, i.e. an individual that stands out in a group is more likely to be singled out by a predator [19]. The measure of phenotypic distance underlying quality of partner in this case could be the same measure that underlies relevance, e.g. both relevance and quality decrease with body-size dissimilarity (e.g. [20]) (figure 1c), or a different measure, e.g. relevance depends on body-size similarity, but quality depends on colour similarity (e.g. [21]).

In other cases, the quality of partner could be unrelated to its phenotypic distance from the focal individual. For example, all else being equal, an animal seeking shared vigilance will be best served by grouping with the most vigilant individual, irrespective of its own vigilance level [22]. Finally, the quality of a partner might even increase with phenotypic distance in some cases. In the case of cooperative hunting, an animal might be best served by grouping with a partner that is different from it in hunting strategy [23]. In summary, given that the relevance and quality of a partner could be determined by different phenotypic traits, these need to be included as separate axes in a model of partner choice.

(iii). Cost of activity matching will increase with phenotypic distance

Previous models [3,13,14] assume that the cost of social activity is fixed, i.e. it does not vary with phenotypic distance. Although this might be true in a few specific contexts of social information use, the cost of activity matching, in general, is likely to increase with phenotypic distance. For example, for a forest bird joining a moving foraging group for risk dilution, the important trait for activity matching cost is likely to be foraging behaviour (e.g. [24]). A bird will need to make fewer adjustments in its foraging behaviour, and therefore will experience fewer lost foraging opportunities, if it chooses to group with individuals showing the same foraging behaviour (figure 1d). Similarly, in other ecological contexts too, activity matching costs will be related to the adjustments that animals have to make to spatially and temporally match their social partners in ecological activity, which, in general, are likely to increase with phenotypic distance (table 2). In the case of social information use, activity matching costs could also include neural processing of information and social learning, both of which are likely to increase with the phylogenetic distance of social partner [25,26]. In general, the trait that determines cost of activity matching need not be the same as those determining costs of competition or social benefit. Therefore, we need a separate axis of phenotypic distance to model cost of activity matching.

Table 2.

Traits that are likely to underlie relevance of partner, quality of partner, cost of competition and cost of activity matching for hypothetical examples of heterospecific sociality. n.a., not applicable, i.e. relevance does not vary across individuals.

example number	1	2	3	4	5
taxon	marine fish	bats	frogs	cetaceans	birds
type of sociality	moving group	social aggregation	social aggregation	moving group	social information use without group formation
context	foraging	sleeping	singing to attract mates	cooperative hunting	nesting
benefit	dilution of risk	thermoregulation	dilution of risk	food	information on nest site quality
trait important for:
relevance of partner	body-size similarity	n.a.	body-size similarity	diet similarity	nesting substrate similarity
quality of partner	body colour similarity	body temperature	vocalization similarity	hunting behaviour dissimilarity	knowledge of habitat
cost of activity matching	foraging speed dissimilarity	sleeping location + time dissimilarity	singing location + time dissimilarity	foraging location + time dissimilarity	nesting cycle dissimilarity
cost of competition	diet similarity	sleeping site similarity	mate choice similarity	diet similarity	nesting site similarity
Is competition only intraspecific?	no	no	yes	no	yes

(iv). In some scenarios, competition costs might be incurred only from conspecific social partners

Earlier models [3,13,14] assume that competition cost is determined only by phenotypic distance, irrespective of whether the partner is a conspecific or heterospecific. This need not always be the case. For instance, in scenarios where competition is linked to breeding, heterospecific social partners might not impose any competition costs. Prairie chickens and grouse, for example, form mixed-species leks, potentially to obtain the protection benefits of grouping but without incurring the cost of mate competition that they would each entail if they grouped with conspecifics [27]. Similarly, male frogs that form mixed-species choruses obtain the protection of a group, without incurring the costs of mate competition that would come from chorusing with conspecifics [28]. A third example is animals that maintain territories from which they exclude only conspecifics [29]; if such animals need to group for any reason, they can do so only with heterospecifics, because they are the only individuals available within their territories. In summary, while, in general, it is reasonable to model competition cost as reducing with phenotypic distance, our framework needs to account for the possibility of competition changing in an all-or-none fashion across the species boundary.

(b). Examining partner choice based on multi-dimensional trait dissimilarity

Based on the factors discussed above, we propose that it is useful to examine social partner choice using a multi-dimensional trait space (figure 1). Traits underlying cost of competition, relevance of partner, quality of partner and cost of activity matching (see table 2 for examples of such traits in different taxa and contexts) form the axes of this space. Each potential partner's position in this trait space is determined by its phenotypic distance, along these trait axes, to the focal individual. Conspecific individuals will, typically, be found near the origin of the trait space, whereas heterospecifics can, in principle, occur anywhere. For a given scenario, based on how social benefit (relevance and quality of partner) and costs (competition and activity matching) vary along relevant axes of phenotypic distance (e.g. figure 1a–d), we can delineate the portion of trait space within which being social will result in higher fitness than being solitary for our focal individual, as well as points in the trait space that are optimal for partner choice.

According to earlier social information use models [3,12,14], partner choice is determined by only two factors: decrease of social benefit (only in terms of information relevance) and competition cost, in relation to phenotypic distance. Depending on the shapes of these two curves, the optimal partner for social information use could be a conspecific or a heterospecific. In other words, only one scenario, corresponding to the optimum in the above trade-off, leads to a heterospecific being chosen as a partner. The key insight from our model is that multiple scenarios could lead to heterospecifics being chosen as social partners, with the simple social benefit–competition trade-off being only one of them. Net social benefit could be maximized in different parts of the multi-dimensional trait space depending on how social benefit and costs vary along four axes (electronic supplementary material, figure S1). Based on our model, we make three qualitative predictions about scenarios in which the likelihood of heterospecific sociality is particularly high (discussed in more detail in electronic supplementary material, Appendix S1):

(i). Traits underlying competition costs and partner relevance are decoupled

In a scenario where partner relevance and competition costs are determined by different traits, the best partner will be one that is highly similar in the trait(s) important for relevance but highly dissimilar in the trait(s) important for competition. Such a partner is more likely to be found among heterospecifics than conspecifics because different traits are, usually, strongly correlated in the latter.

(ii). Only heterospecifics can provide the required social benefit

In some scenarios, heterospecifics and conspecifics might both be equally relevant partners but only the former might be able to provide a certain social benefit. For example, individuals of migrant bird species are known to use cues from resident heterospecifics to choose habitats in their wintering areas [30]. Even if conspecifics and heterospecifics share the same level of overlap in habitat requirements with the focal bird (i.e. equal relevance), only the latter has the experience of the area to be able to provide reliable information on habitat quality. Therefore, in such scenarios, we expect heterospecific sociality.

(iii). Costs of competition are only intraspecific

As discussed earlier, in scenarios where competition is linked to breeding, competition costs will be applicable only for conspecific social partners. In such scenarios, heterospecifics, even if less relevant and of poorer quality as social partners than conspecifics, might still get chosen because partnering with them brings no competition costs.

In addition to the factors described above, social partner choice is also likely to be influenced by factors unrelated to the trait space, such as availability, dispersion and familiarity of potential partners (discussed in electronic supplementary material, Appendix S1).

4. Synthesis

Although the existence of heterospecific sociality has been known for a long time, it has not been integrated into conceptual frameworks used to understand animal sociality in general. This, we believe, is because heterospecific sociality and single-species sociality are thought to be based on fundamentally different mechanisms [6,12]. However, our literature review reveals a high diversity of cases of heterospecific sociality, in numerous taxa and ecological contexts, and for multiple social benefits that are likely to be based on mechanisms similar to single-species sociality

Based on the findings of our survey, we suggest that animal sociality is best-viewed along a continuum from conspecifics to heterospecifics. At one end of this continuum are social benefits obtainable only from conspecifics (e.g. lekking among males of same species to attract females for mating). At the other end are cases where benefits are in different currencies for different partners, which is possible only between dissimilar heterospecifics (e.g. one bird species gets food and the other gets protection in a mixed-species bird flock). Between these extremes, straddling the species boundary, are a range of benefits that can, in principle, be obtained either from conspecifics or heterospecifics. A non-exhaustive list of these benefits would include protection from predators, finding food, conserving heat and water, and reducing energetic costs of movement [6]. Given that, in nature, any animal is likely to be surrounded by both conspecifics and heterospecifics, how is a social partner chosen? We propose a conceptual framework for understanding social partner choice and highlight a range of different scenarios that might result in heterospecifics being chosen over conspecifics as social partners.

Given the evidence we have presented, both for the documented prevalence of heterospecific sociality and the numerous scenarios through which it can potentially happen, why is single-species sociality, then, so much more common in the natural world? For example, if one made a global list of species that are known to forage in groups, the list for single-species groups will be much longer than heterospecific groups. This will be the case even if we considered only non-kin social interactions [10,11], in which inclusive fitness benefits are not a biasing factor in favour of single-species sociality. What then might constrain the occurrence of heterospecific sociality versus single-species sociality?

Part of the answer might lie in cost of activity matching. Earlier social information use models [3,13,14] assumed that the cost of activity matching remains the same, irrespective of phenotypic distance between social partners. However, as we have discussed above, cost of activity matching is likely to increase with phenotypic distance, and therefore be higher, usually, for heterospecifics than conspecifics. How much higher will depend on the type of social interaction. In all types of social interaction, social partners will experience the costs that arise due to the need to spatially and temporally match ecological activity. For example, woodpeckers, which forage up and down the trunks of trees, will experience lower costs of activity matching when grouping with other trunk-foragers than with gleaning birds, which move horizontally through an area when foraging. However, in the case of moving social groups, social partners will also incur the additional cost arising out of the need to coordinate movement. Therefore, the likelihood of heterospecific sociality is likely to be relatively lower in moving social groups owing to the high costs of activity matching. The results of our literature survey also seem to suggest this is the case: moving groups of heterospecifics are much less prevalent than both stationary social aggregations and instances of social information use without group formation. Joining mobile animal groups, however, can offer benefits that are absent at an individual level ([31,32]; see electronic supplementary material, Appendix S1 for more details). Studies targeted at identifying and quantifying the cost of activity matching in different types of heterospecific sociality, versus benefits associated with emergent properties of group dynamics, might help better understand the importance of this constraint (see box 1 for key areas for future research).

Box 1. Key areas for future research.

(a) Prevalence of heterospecific versus single-species sociality in nature

Use standard methodology to quantify the prevalence of single-species and heterospecific sociality in different taxa, ecological contexts, geographical areas, seasons and for different social benefits. This information can form the raw material for comparative approaches to understanding social partner choice.

(b) Traits underlying costs and benefits of sociality

Identify the traits underlying cost and benefit axes of social partner choice, examine trait correlations (see (c) below) and understand how costs and benefit curves vary in relation to traits, in different sociality scenarios.

(c) Influence of trait correlations on partner choice

In our conceptual framework, the traits underlying the cost and benefit axes of social partner choice can be the same or different. When they are different, they can range from being strongly correlated to completely uncorrelated. We need to understand, theoretically, how variation in trait correlations influences the likelihood of heterospecific versus conspecific social partner choice, in different sociality scenarios.

(d) Influence of group size and composition on partner choice

Our framework is built from the point of view of an individual organism choosing a single social partner. Although this is a useful starting point to understand partner choice conceptually, real-world scenarios will also involve joining larger groups. Joining mobile animal groups can offer benefits that are absent at an individual level ([31,32]; see electronic supplementary material, Appendix S1 for more details). Furthermore, dynamics of merge and split among groups can result in nontrivial variations in group composition [33]. Such emergent properties of groups could be an additional factor in influencing the relative likelihood of heterospecific and conspecific partner choice. Agent-based spatially explicit evolutionary models (e.g. [34]) can enable theoretical understanding of how cost and benefit curves of partner choice, as a function of phenotypic distance, change in relation to group composition and size. Finally, our framework needs to be integrated with frameworks developed to understand group heterogeneity, in phenotype (e.g. [35]) and roles (e.g. leadership [36]), in single-species contexts.

(e) Role of extrinsic factors

Our conceptual framework focuses only on understanding how intrinsic traits of potential partners affects partner choice. Factors unrelated to the trait space, such as availability, dispersion and familiarity of potential partners (electronic supplementary material, Appendix S1) and variation in the strength of sociality drivers (e.g. resources, predators, abiotic environment), make partner value dynamic and context-dependent. Theoretical and empirical investigations of such extrinsic factors form another direction for future research.

A greater number of documented cases of single-species sociality might also reflect biases in our research priorities. Single-species sociality has historically attracted more research attention and therefore is much better documented than heterospecific sociality. This is especially important in the case of social information benefits where, without careful study, it is highly unlikely that such an interaction will be revealed [3]. Even in the case of direct grouping benefits, groups containing heterospecifics are more likely to go unnoticed without focused attention because they, on average, tend to be less cohesive, more temporary and more variable in composition than single-species groups. This could, in turn, lead to an underestimation of the importance of heterospecific sociality. Nevertheless, cumulatively, across multiple such short-lived groups, heterospecific sociality might have considerable fitness consequences for species' life histories. It is likely, therefore, that our view on the relative commonness of single-species and heterospecific sociality, as well as their relative importance for species, might change if more research attention is focused on the latter (box 1).

Data accessibility

This article has no additional data.

Authors' contributions

H.S. conceptualized the study, carried out the literature review and proposed the conceptual framework. H.S. and V.G. refined the framework and wrote the paper.

Competing interests

We have no competing interests.

Funding

H.S. was supported by fellowships from the Indian National Science Academy and Wissenschaftskolleg zu Berlin. V.G. acknowledges support from DBT-IISc partnership program.

Glossary

Sociality

Spatial and/or temporal clumping of organisms as a result of one- or two-way social attraction, either for social information benefit or direct grouping benefit (see below). Our definition includes interactions categorized as mutualism, cooperation and commensalism. Single-species sociality and heterospecific sociality refer to sociality that involves individuals of the same species and multiple species, respectively.

Social information benefit

Fitness benefits are in the form of information (about food, predators, suitable nesting sites, etc.) produced by social partners.

Direct grouping benefit

Social partners are the direct source of fitness benefits, unmediated by information (when a social partner provides warmth, dilutes risk of predation, reduces wind resistance in flight, etc.).

Social benefit

Generic term used to indicate either social information or direct grouping benefit. Social benefit provided by a social partner is a function of both relevance and quality of partner (see below)

Cost of competition

Fitness cost due to resource use (food, nesting sites, etc.) by social partner.

Cost of activity matching

Fitness cost due to the need to match activities spatially and temporally with social partner.

Phenotypic distance

Generic term used to indicate dissimilarity between organisms in morphological, behavioural or ecological traits.

Moving social group

Social group whose members move together in a coordinated fashion, often for long periods of time.

Social aggregation

Aggregations of individuals maintained by social attraction, often for extended periods of time, but which do not move together in a coordinated fashion. Although clumping of individuals could reflect resource patchiness (e.g. basking sites, fruiting trees, mud-puddles), our interest is in the non-random temporal overlap of individuals at such patches, which is a result of social attraction.

Relevance of partner

The extent to which partner overlaps with focal individual in the driver of sociality (e.g. food, predator, nesting site). The greater the overlap, the more relevant is the partner.

Quality of partner

The efficacy with which a social partner fulfils its role in relation to the specific mechanism underlying social benefit. Equally relevant partners (i.e. equal overlap with focal individual in driver of sociality) can vary in their efficacy as social partners. Social benefit is, therefore, a function of both relevance and quality of partner.