If sociality involves processing and storing more information, certain brain structures might be enlarged in social insects compared to solitary ones. Early observations suggest that this is the case in Hymenoptera (Dujardin, 1850; von Alten, 1910; Howse, 1974). However, Farris and Schulmeister (2011) recently challenged these observations by comparing the mushroom bodies of 22 hymenopteran species. Their study demonstrate that large mushroom bodies with elaborated calyces, receiving inputs from visual and olfactory neuropils, are equally common in solitary, presocial, and eusocial Hymenoptera. In fact, the study suggests that large mushroom bodies were acquired 90Myr before the evolution of sociality in the group and coincided with the transition from phytophagy to parasitoidism in solitary wasps (Figure 1B). Presumably, the novel challenges of central place foraging, finding, and overwhelming prey to provision larvae, and especially the need to develop spatial memories for their nesting sites, placed much higher cognitive demands on these parasitoids than their vagabond herbivorous ancestors. A similar approach in non-hymenopteran insects confirms that the evolution of large mushroom bodies is primarily associated with complex foraging behavior rather than with sociality. Examples include generalist scarab beetles that must discriminate among many potential resources to balance their diet in heterogeneous nutritional environments (Farris and Roberts, 2005), or pollinating butterflies that exploit flowers using habitual foraging routes (Sivinski, 1989). Indeed, many of the most impressive cognitive feats in insects have been identified in the context of individual foraging and are unrelated to sociality (e.g., attention-like processes, Spaethe et al., 2006; interval timing, Boisvert and Sherry, 2006; numerosity, Chittka and Geiger, 1995; rule learning, Giurfa et al., 2001; and route optimization Lihoreau et al., 2012). Recent studies have begun to focus on other brain components, and suggest that the size of antennal lobes (Figure 1A) varies with social organization in wasps (Molina et al., 2009) and ants (Riveros et al., 2012). It will be important to explore which differences in internal structure mediate these changes; for example, a larger volume of antennal lobe glomeruli might indicate higher sensitivity to certain odorants, whereas a larger number of glomeruli might mediate a sensitivity to a higher diversity of chemicals such as pheromones.

Some adult insects exhibit dramatic structural plasticity in their brains in the form of dendritic outgrowth (Withers et al., 1993) or neurogenesis (Ott and Rogers, 2010), which also provides a potential test of the social brain hypothesis. If sociality requires added circuitry, brain size should increase in individuals that change from being solitary to social at different stages of their life cycle. In the facultatively social bee Megalopta genalis, where cooperatively breeding females establish a dominance hierarchy to determine their contribution to reproduction, the mushroom bodies of dominant reproductives are much larger than those of subordinate workers and solitary reproductives (Smith et al., 2010). Similar enlargement of the mushroom bodies of reproductives is found in multiple obligatorily social wasp species (Molina and O’Donnell, 2007, 2008; O’Donnell et al., 2011), suggesting that the demands of competing over reproduction or maintaining dominance (and thus remembering others’ identity, assessing their status, and displaying aggressive behaviors) promote mushroom body growth (but see Ehmer et al., 2001). Thus while in this case there is a link between social behavior and mushroom body volume, it does not equally affect all members of the colony.

Another striking example of neuroplasticity in the adult brain is the continuous axonal pruning and dendritic outgrowth in the calyces of the mushroom bodies of social Hymenoptera. In honeybee workers (Apis mellifera), the mushroom bodies enlarge before the transition to becoming foragers, while individuals still remain inside the hive and engage in brood care duties (Fahrbach et al., 1998). This “experience-expectant” plasticity prepares bees to store memories related to outdoor navigation, which involves exploring multiple foraging options, learning flower locations, their sensory signals, and handling procedures. Actual foraging experience also triggers a further “experience-dependent” increase in mushroom body size (Withers et al., 1993). This plasticity is not specific to honeybees but has been observed in ants (Gronenberg et al., 1996; Kuhn-Buhlmann and Wehner, 2006), wasps (O’Donnell et al., 2004), and solitary bees (Whiters et al., 2007), thus emphasizing the tight relationship between foraging and the development of mushroom bodies in these insects.

In humans, societies are held together by the ability of individuals to recognize generalized features indicating group membership. This allows them to identify society members without needing to memorize the identity of every single individual in the group. Social insects, likewise, often recognize members of their own colony, allowing them to defend resources and brood against competitors. It is important to keep in mind that with the advent of the generalized “labels” defining members of a society (be they tribe-specific clothing in humans or chemical profiles in insects) the size of a society essentially becomes infinite. Identifying group members does not require any added cognitive demand; individuals need only remember a set of defining features. In ants, nestmate recognition is mediated by colony-specific chemosensory cues on the cuticle of colony-members (van Zweden and D’Ettorre, 2010). Ants recognize “friends” and “foes” by comparing the odor carried by encountered individuals to an internal representation of their own colony odor, which requires fine discrimination of multicomponent variable cues (Guerrieri et al., 2009). Neuroethological studies suggest that such recognition can be achieved through simple cognitive operations that do not involve information-processing in the higher integration centers of the brain or the acquisition of a long-term memory (Bos and D’Ettorre, 2012). For instance, electrophysiological recordings of the antennae of Camponotus japonicus have shown that cuticular hydrocarbon-sensitive chemosensilla are less responsive to non-nestmate odor upon prolonged exposure, suggesting that receptors adapt to the chemical environment and act as a “sensory filter” mediating recognition (Ozaki et al., 2005). Decreased responses to familiar odors could also result from non-associative learning (such as habituation) at the level of the antennal lobe, as suggested by the activity patterns of antennal lobes subsequent to exposure with colony odors in C. floridanus (Brandstaetter et al., 2011) and the absence of information transfer between brain hemispheres (through the mushroom bodies) during recognition in C. aethiops (Stroeymeyt et al., 2010). Although it is still not clear precisely where colony odors are stored and processed in the ant brain, the template is decentralized in the olfactory system and recognition is achieved through simple processes such as sensory adaptation or habituation.

In some small colonies of primitively social ants (D’Ettorre and Heinze, 2005) and wasps (Tibbetts, 2002), females establish a dominance hierarchy and develop a long-term memory of colony-members’ identity, which helps stabilize social interactions. In Polistes wasps, individual recognition is mediated through conspicuous facial markings (Sheehan and Tibbetts, 2008; Figure 2A). These wasps are naturally specialized “experts” in face recognition since they are less efficient at recognizing arbitrary patterns, or face pictures that lack antennae or contain scrambled features (Sheehan and Tibbetts, 2011). Comparisons of brains from wasp species that learn faces and closely related species that lack face recognition revealed no discernable difference in the size of the visual neuropils (Gronenberg et al., 2008). Presumably, minor modifications of the pre-existing neural circuitry used for recognition of landmarks or prey in the common ancestors of these wasps have sufficed to evolve individual recognition. Artificial neural networks confirm that basic circuitry for face recognition requires only a few hundred neurons (Aitkenhead and McDonald, 2003) and could thus easily be accommodated in the insect brain. Similar specializations of the visual system could potentially underpin a variety of social recognition processes, as for example the ability of some stingless bees to recognize nest identities using visual marks (Chittka et al., 1997; Figure 2B).

The symbolic “dance language” of honeybees is perhaps the most impressive form of communication found in insects (von Frisch, 1967). Upon its return to the nest, a bee that has located a new food resource performs the “waggle dance,” a figure-of-eight-shaped circuit conveying information about the distance and the direction of the resource to its nestmates (Figure 2C). In an even more impressive social use of the dance language, bees indicate suitable nest site locations to a swarm in search of a new home. The swarm builds a consensus from multiple “opinions” expressed by scouts with different information, to finally agree on a single destination to which the swarm relocates (Seeley, 2010). At least five sensory systems (motion sensitivity, sun compass, gravity, mechanosensitivity, and acoustic perception) are required to acquire, transmit, and decode the information of the dance language. Therefore, one might expect this unique behavioral innovation to have easily detectable correlates in terms of neuroanatomy. However, examination of the neural pathways mediating these sensory systems have revealed no “dance specific” projections in the brain of honeybee workers (A. mellifera) when compared with queens, drones, and workers of species that lack dance communication (Brockmann and Robinson, 2007). This suggests that small tweakings of existing circuitry, such as novel connections between sensory pathways, central circuits, and motor output may have sufficed to produce this behavior. It is easy to imagine, for example, how circuitry in place for forward walking locomotion can be supplemented by a few control neurons to recruit the leg muscles in the order required to generate a figure-eight pattern.

Many of the collective behaviors of insects, such as consensus building during nest site selection by honeybees (Seeley, 2010) and ants (Sasaki and Pratt, 2012), foraging decisions by cockroaches (Lihoreau et al., 2010) or nest construction by termites (Bonabeau et al., 1998; Figure 2D), emerge through self-organization, based on local interactions between partially informed individuals (Jeanson et al., 2012). Through communication, insects can overcome individual limitations in acquiring or processing information, allowing groups to make faster and more accurate decisions than they might as individuals (Couzin, 2009). Seeley (2010) makes a convincing case that workers in a honeybee swarm can essentially be viewed as sensory units of a “collective brain”. The mechanisms behind this “swarm intelligence” have been particularly well described in the context of path optimization in ants (Goss et al., 1989). Argentine ants (Linepithema humile), for instance, develop networks connecting the multiple nests of their colonies (Aron et al., 1990). Trails arise through the action of many ants, each depositing droplets of an attractive pheromone at regular intervals as they explore their environment. The initial network is complex, containing many redundant connections. But pheromones evaporate, so longer, more circuitous routes (which take longer to traverse and so get fewer passes) evaporate first, while shorter, more direct routes are reinforced more often and last longer. The end result of this positive feedback loop is a network that connects nests via the shortest path (Latty et al., 2011; Figure 2E). In such decentralized systems, pheromone trails serve as an “externalized memory” freeing individual ants from the need to store location information in their brains, whilst allowing colonies to reach network optimization performances not accessible to isolated individuals.