Mechanisms of recognition in birds and social Hymenoptera: from detection to information processing

Abstract

In this opinion piece, we briefly review our knowledge of the mechanisms underlying auditory individual recognition in birds and chemical nest-mate recognition in social Hymenoptera. We argue that even though detection and perception of recognition cues are well studied in social Hymenoptera, the neural mechanisms remain a black box. We compare our knowledge of these insect systems with that of the well-studied avian ‘song control system’. We suggest that future studies on recognition should focus on the hypothesis of a distributed template instead of trying to locate the seat of the template as recent results do not seem to point in that direction.

This article is part of the theme issue ‘Signal detection theory in recognition systems: from evolving models to experimental tests’.

1. Introduction

Recognition behaviour is a critical aspect in the life history of organisms across all taxa. It is relevant for virtually all inter- and intra-specific interactions in which any form of discrimination occurs at different levels: species, kin, familiar and individual. Explanations of recognition systems [1] typically subdivide the behaviour into three main components: expression, perception and action [2–4]. The expression component refers to the development of recognition cues, the perception component refers to the detection and assessment of cues, and the action component refers to the resultant behaviour given the assessment of the cues. In this opinion piece, we will focus on the perception component of social recognition systems by reviewing our knowledge of the neural mechanisms underlying recognition.

We chose to focus on birds and social Hymenoptera because there is substantial literature on avian and social hymenopteran recognition systems, providing a strong foundation of information for comparing them. There is additional value in comparing vertebrate systems with invertebrate systems to see how evolution shaped them, given the long time elapsed since their common ancestor. Avian recognition systems are mostly based on visual and acoustic traits (although recent studies are beginning to tackle the chemical aspect of their recognition systems), while social Hymenoptera recognize each other chemically, which makes it harder for scientists to study given that chemical signals are less easy to manipulate. For this reason, we will compare our knowledge of the well-studied avian ‘song control system’ [5] to the chemical one of social Hymenoptera, the neural mechanisms of which remain a black box.

Individual recognition (i.e. the recognition of an individual according to its distinctive characteristics) plays a crucial role in a wide variety of contexts (mate selection, territorial defence, dominance, social competition, altruism) and is supposed to be a widespread characteristic of social species [6]. However, while birds show abundant evidence of precise individual recognition [7–11], there are only few examples of individual recognition in social Hymenoptera [12,13]. While it was long believed that this difference was owing to restrained cognitive abilities in social Hymenoptera [14], the explanation could instead lie in the selective pressures that shape the socio-ecological contexts of the different lineages. Individual recognition is supposed to occur when there are repeated interactions among multiple individuals with differing aims [15] and is, therefore, unlikely to occur in monogynous eusocial Hymenoptera in which the colony functions as one individual. Recognition of kin or group members is essential to the evolution of social behaviour, whether living in a small family group (like in most bird species) or in a society of thousands of individuals (as in honeybees and ants). However, kin and nest-mate recognition are equivalent in social Hymenoptera only when colonies are headed by a singly mated single queen and there is no queen turnover. Because monogynous and monandrous species represent only a fraction of social Hymenoptera, recognition studies usually deal with nest-mate rather than kin or individual recognition (e.g. [16]). We will, therefore, address the finest possible degree of recognition in social Hymenoptera and birds to understand the underlying neural mechanisms: nest-mate and individual recognition, respectively.

2. The signals and cues

Both signals and cues are traits that have evolved because they alter the behaviour of other individuals. However, while signals are plastic (they can be expressed in response to relevant cues in the environment) and might also exhibit some intentionality, cues are either permanently expressed by the individuals emitting them or expressed depending on specific conditions [17]. In birds, the recognition of partners depends on the life-history traits of the considered species. Studies have demonstrated the recognition of kin and social partners on the basis of vocal, visual and olfactory cues. In the vocal domain, it has been evidenced that parents and young of many bird species are able to recognize each other vocally [18]. Such recognition is at stake especially for species living in large colonies. For example, in the king penguin (Aptenodytes patagonicus), chicks identify the call of their parents in the continuous background noise of the colony [19]. In the black redstart (Phoenicurus ochruros), where the male feeds a subset of the offspring and the female the other subset (brood division), it has been demonstrated that both parents use acoustic signals in the begging call to discriminate between the different fledglings [20]. Individual vocal recognition can also develop later in a reproductive context. Both males and females of many species are able to recognize the vocal signature of their sexual partner [21]. Males of many territorial species are able to recognize the vocal signature of their neighbours as well: they often respond less aggressively to song playbacks of neighbours than to playbacks of strangers, which represent a higher threat to territory ownership, a phenomenon called ‘dear enemy effect’ [22]. There is also evidence in birds that learned signals can be used as vocal labels to address particular individuals in a social group. For example, spectacled parrotlets (Forpus conspicillatus) use contact calls to refer to a social companion and thus label or ‘name’ their conspecifics [11].

In the visual domain, Lorenz's work on imprinting [23] has shown that in precocious birds such as ducks and geese, the offspring will follow the first moving object that they encounter, which under natural conditions is the mother. Several studies have shown that a bird is able to recognize a familiar individual based on visual criteria in the absence of other signals [24,25]. For instance, budgerigar males (Melopsittacus undulates) are able to discriminate between pictures of two familiar conspecific females in the same way that they do with live birds [26]. Some social insect species are also able to recognize nest-mates and individuals visually, particularly in wasps [12,27]. However, most insects use olfactory cues to discriminate each other as the behaviour of insects is controlled to a great extent by olfactory stimuli [28].

In social Hymenoptera, nest-mate recognition is thought to be based on the permanent expression of cuticular hydrocarbons (CHCs). Insect cuticles are covered by waxy substances (mostly long-chain hydrocarbons from 20 to 35 carbon atoms) that evolved originally as a barrier against desiccation and pathogens and were later co-opted as recognition chemical cues [29]. Social Hymenoptera show a complex pattern of CHCs, which varies in quality among species and quantity (relative amount) within species, thus representing an ideal multi-component signal with the level of polymorphism required for recognition to be effective. These substances can be both genetically and environmentally determined and are permanently mixed through trophallaxis and grooming to form a uniform blend [30]. In birds, several species of marine seabirds are able to recognize their own odour [31] as well as that of their mates [32–34]. More recently, it has been demonstrated that young zebra finches (Taeniopygia guttata) are able to distinguish between kin and non-kin based on olfactory cues alone [35]. Moreover, hatchlings are able to recognize the body odour of their parents, both mother and father, already a few hours after hatching [36]. The main source of odours used in olfactory communication in birds is most likely the preen gland and its secretion [37]. The major histocompatibility complex, a set of cell surface proteins that have their main function in the recognition of self and non-self-molecules, has also been argued as a potential component involved in olfactory recognition of individuals [38].

3. The templates

Templates are neural representations of the characteristics of desirable or undesirable recipients. Recognition occurs when the phenotype of encountered recipients matches these internal representations closely enough [39].

Templates can emerge early during ontogeny and depend on a genetic program modulated by environmental influences from both the biotic and abiotic milieu (e.g. [40]). Templates that exhibit a strong genetic determinism are particularly present in the frame of sexual selection where there is a genetic correlation between females' templates and males’ traits (reviewed by [41]). When template learning might increase mortality or when recipients are not reliably present for template formation, genetically determined templates can be favoured as well. This is the case for naive Turquoise-browed motmot birds (Eumomota superciliosa), which are instantly repulsed by models resembling venomous coral snakes [42]. In the olfactory domain, juvenile zebra finches are able to recognize their genetic mother without prior associative learning [36]. In ancestral social Hymenoptera, members of the same colony were closely related since they were typically monogynous and monandrous [43], so that kin recognition equalled nest-mate recognition. Ozaki & Hefetz [44] hypothesized that given this high relatedness between colony members and the possibility that colony odour in ants was primarily genetically determined [45], templates may have been genetically encoded as well. With the evolution of insect societies and the introduction of genetic variability, the role of the environmental influences in template shaping became more important.

Indeed, templates must be learned when the characteristics of recipients vary over space or time. In the vocal domain, this template can be acquired well before hatching in precocial birds whose auditory system is functional embryonically [46]. Such prenatal acquisition of individual auditory recognition of the parent has been demonstrated in domestic chicks [47,48], guillemots [49], gulls [50] and pheasants [51]. In the superb fairy-wren (Malurus cyaneus), a species that undergoes brood parasitism by a cuckoo species, females call to their eggs during the late stage of development and, upon hatching, nestlings produce begging calls with key elements from their mother's incubation call. This strategy could allow both parents to detect foreign cuckoo nestlings [52]. In white-browed scrubwrens (Sericornis frontalis), nestlings learn the characteristics of predator contexts by listening to parents' alarm calls [53].

When templates are learned, the objects or individuals that provide information about the characteristics of desirable or undesirable recipients are called referents. Sometimes, actors (=discriminators) serve as their own referent. In several species of oscine songbirds, it has been proposed that the Bird's Own Song (BOS) could be used as a reference against other males' songs [54]. Such an ‘autogenous reference’ could be also used in non-vocal learner species as a template through self-referent phenotype matching. Self-referent phenotype matching has been proposed as a mechanism that could be used in individual recognition [40]. Parasitic brown-headed cowbird juveniles (Molothrus ater) seem to use themselves as referent since they approach faster and associate preferentially with individuals that are experimentally coloured similarly to themselves [55]. In social Hymenoptera, it is very difficult to experimentally distinguish between self-referent phenotype matching and recognition alleles ([56] but see [57]).

The timing of template learning depends on when the most informative referents are available and when discrimination is first adaptive. In some species, templates are learned during a short sensitive period early in life (imprinting period), presumably because the referents are predictably present and their characteristics reflect those of desirable recipients throughout an actor's life [58]. Learning the phenotypes of predictable referents would be more advantageous than innate preferences in rapidly evolving groups of animals because it could be refined from one generation to the next [59]. Imprinting is widespread in birds [60] and was first described in precocious birds such as ducks and geese by Lorenz [23]. Lorenz defined imprinting according to five criteria [61]: (i) learning does not require reinforcement and then cannot be considered as associative learning; (ii) learning takes place during a critical (or sensitive) period soon after birth or hatching; (iii) learning is an irreversible process with a mnesic trace; (iv) the characteristics of the mother are then generalized to other members of the species; (v) imprinting has long-lasting effects that occur later in life, especially in the case of mate choice [62]. There is also a sensitive period for nest-mate recognition to occur in social Hymenoptera (e.g. in bees [63], wasps [64] and ants [65,66]).

Recognition templates might last for long periods of time. For example, long-term recognition of songs produced by territorial neighbours has been described in migratory birds such as in male hooded warblers (Setophaga citrina) [7] and dusky warblers (Phylloscopus fuscatus) [67]. But recognition templates may be updated when there is a lengthy association between individuals whose characteristics may change through time. For example, the template of parents must be updated for them to recognize their grown offspring, the phenotype of which changes between their juvenile period and their adult period [1]. In social Hymenoptera, the template for nest-mate recognition is updated to match dynamic changes in the colony-specific CHC pattern [68,69]. Such continuous template reformation is analogous to the rewriting of learned memory [70]. The report that some ant species are able to learn to associate CHCs with food is consistent with template acquisition through standard associative learning experiment [71,72]. Ozaki & Hefetz [44] suggest that similar associative learning of nest-mate CHCs by callow workers while being groomed or fed by trophallaxis may be involved in the formation of the primary template for nest-mate recognition. Trophallaxis among mature nest-mate workers following mutual antennation may also largely contribute to associative learning of the nest-mates' CHC odour, thereby both updating the template and consolidating it.

4. The neuroanatomy of recognition

(a). The case of social Hymenoptera

Nest-mate recognition in social Hymenoptera is generally attained following physical antennal contact with the body surface of the encountering individual, or from a close distance [73]. Basiconic sensilla are supposed to mediate CHC detection and are female-specific [74]. They host numerous sensory neurons (e.g. 200 in Camponotus japonicus [75]) supposed to correspond to the highly complex cuticular profiles that have to be resolved for nest-mate recognition. Basiconic sensilla are oriented perpendicular to the antennal surface, which facilitates chemical inspection of encountered objects. They also bear numerous pores, an indication of their olfactory function [76].

Odorant molecules penetrate the cavity of the sensillum through their tiny pores, where they are trapped by olfactory receptor neuron (ORN)-specific receptor molecules located on the membrane of the ORN dendrites. The highly hydrophobic CHCs are carried through the lymph bathing the ORNs through the mediation of odorant-binding proteins (OBPs) [77,78], chemosensory proteins (CSPs) [79] and the Niemann–Pick type C2 protein (NPC2) [80].

Upon perception, the generated neural signal reaches the brain via axonal projection from the ORNs to the multiple glomeruli of the antennal lobe. Functionally identical ORNs converge to the same glomerulus [81–83], enhancing sensitivity while retaining the idiosyncrasy of the information. The total number of glomeruli varies from one species to another (e.g. 460 in Camponotus floridanus [81], 480 in C. japonicus [74] and 170 in Apis mellifera [84]). In the glomeruli, the olfactory information can be further processed by the local interneurons, after which it is transferred via projection neurons to the higher centres of the insect brain, the mushroom bodies (MB) and the lateral horn (LH) [83] (figure 1a).

Figure 1.

The neuroanatomy of recognition in social Hymenoptera and birds. (a) Frontal view of an ant brain representing the brain regions involved in cuticular hydrocarbon processing. Abbreviations: AL, antennal lobe; LH, lateral horn; MB ca, mushroom body calyx; MB ped, mushroom body pedonculus; OL, optic lobe. Adapted from [85]. (b) Frontal view of the ‘song control system’. Green regions are involved in learning, blue ones in production and red ones in perception. Abbreviations: CMM, caudomedial mesopallium; DLM, nucleus dorsolateralis anterior thalamis, pars medialis; HVC, a letter-based name; L, field L; LMAN, nucleus lateralis magnocellularis, pars lateralis; MLd, nucleus mesencephalicus lateralis, pars dorsalis; NCM, caudomedial nidopallium; nXIIts, nXII pars tracheosyringealis; Ov, nucleus ovoidalis; RA, nucleus robustus arcopallii; X, area X. Adapted from [86].

(b). The case of birds

At the neural level, dedicated sensorimotor circuits play an essential role in vocal learning, production and perception in oscine songbirds. The so-called ‘song control system’ [5] comprises a network of both forebrain and brainstem nuclei that distinguishes the songbird's brain from the brains of other birds including those that do not exhibit vocal production learning [87]. In starlings (Sturnus vulgaris), bilateral lesions in one of this nuclei, the HVC (a letter-based name) resulted in deficits in the ability to recognize the songs of different conspecifics without affecting discrimination. This result suggests that HVC plays a role in the formation of associations between a song and a referent [88]. In many oscine songbirds, HVC also exhibits robust auditory responses to the broadcast of the BOS [89,90]. In the zebra finch, when an adult is exposed to his tutor song, there is increased neuronal activation in the caudomedial nidopallium (NCM), the songbird equivalent of the auditory association cortex. Following bilateral lesions of the NCM, tutor song recognition is impaired showing that NCM is an important part of the neural representation of song memory [91] (figure 1b).

Imprinting has been established as a model of choice to study the basis of memory formation. Neural and behavioural analyses have shown that the formation of filial imprinting in precocial birds involves at least two separate processes: (i) one process is an emerging predisposition to approach stimuli with the characteristics of the mother; (ii) the other learning process results in chicks preferentially approaching a stimulus to which they have been exposed and involves forming an association between the components of the exposed stimulus. The neural substrate for the predisposition is different from that underlying imprinting, and different regions of the chick brain are involved in distinct aspects of learning during imprinting [92].

5. The neural localization of the template

In social Hymenoptera, label–template matching could occur at different levels of the neuronal processing of cues, from periphery to higher brain centres. The pre-filter hypothesis suggests that, at the level of the antennae, CHC-specific receptor neurons, when chronically exposed to nest-mate CHCs, undergo sensory adaptation and do not fire. Upon exposure of non-nest-mate CHCs, adaptation is alleviated and firing occurs. It is suggested that this adaptation provides the first discriminatory step in nest-mate recognition. This hypothesis suggests that the template (that of a nest-mate) is localized in the sensilla [44].

At the antennal lobe level (i.e. the first olfactory relay of the insect brain), the habituation hypothesis posits that a similar process occurs, i.e. individuals within a colony would habituate to the colony odour so that they would only detect cues present on non-nest-mates [93,94]. The differences between the two hypotheses lie within the location and the mechanism of the matching, i.e. sensory adaptation occurs at the periphery level and is a physiological process, while habituation occurs at a higher level of neural processing and is a form of non-associative learning. Those hypotheses rely on the pattern that individuals within a colony are almost constantly encountering nest-mates and that having to process nest-mate cues all the time seems costly (see [93,95]).

The neural template hypothesis [44] posits that upon perception of the CHCs, whether that of nest-mate or non-nest-mate, the perceiving sensilla send output nerve impulses to the brain where they form a temporary neural network that is compared to a pre-existing neural network stored in the memory that constitutes the template. The deciphering of recognition signals according to the neural template hypothesis comprises four steps of neural events: (i) the acquisition, via the sensory system, of qualitative and quantitative information regarding own colony CHC profile; (ii) the memory formation and storage of the colony CHC profile in a brain centre, which constitutes the template; (iii) the acquisition of analytical information of the CHC profile of an encountered ant; and (iv) its comparison with the preformed template at the higher brain centre, i.e. the LH or MB. A recent study by Neupert et al. [96] reported that ants were able to detect a lack of component in the CHC profile of encountered ants in C. floridanus, suggesting that multiple templates are formed in the MB based on a parallel pathway processing.

In some species, like C. japonicus, the pre-filter functions very well at the sensory level for discrimination to occur [75]. However, as evidenced in Formica yessensis, this pre-filtration might be incomplete [97]. The behaviour-switching hypothesis reconciles these results by combining the pre-filter hypothesis with a higher brain centre responsible for triggering aggressive behaviour, which can be activated only by the above-threshold inputs [97].

The approach of the template concept in birds is more similar to the last two hypotheses in social Hymenoptera in that the sensory processing seems to convey different streams of acoustic information in parallel pathways of the brain regions from low to high level. However, while research on recognition mechanisms in social Hymenoptera has so far been focused on locating the seat of the template, a large body of evidence now points towards a distributed template in birds [98,99]. This distributed template would be formed of synaptic processes mediating storage of song-related memory involving central auditory structures like the caudomedial mesopallium (CMM), the NCM and field L, the song system itself and brain structures dedicated to song learning but linked to both receptive and sensory-motor circuitry (e.g. HVC shelf and the robust nucleus of the arcopallium (RA) cup) (figure 1b) [98,99]. This distributed template concept might be an interesting hypothesis to investigate in social Hymenoptera that Esponda & Gordon [100] began to tackle by studying it at the colony level: nest-mate recognition by the colony is achieved through a distributed process, so that overall, ants from one colony distinguish those from another, but individuals do not always do so.

6. Conclusion

Although we only addressed in detail the case of olfactory and auditory processing in social Hymenoptera and birds, respectively, it is important to note that recognition is, in fact, multimodal and that animals might prioritize one modality over others at different times of the decision-making process to achieve a better balance between speed and accuracy (e.g. in wasps [27]).

Similarly to the use of various sensory modalities, both taxa could also use various types of recognition learning (i.e. direct and indirect). Indirect recognition involves recognizing individuals not by their phenotype but by their environment (e.g. spatial and social cues). In ants, newly eclosed individuals lack recognition cues on their body [101] and are recognized as nest-mates indirectly because they are in the nest (spatial cues). Another type of learning is social learning, i.e. the extraction of information from other animals [102]. To our knowledge, there is no evidence of social learning in the context of nest-mate recognition in social Hymenoptera, although there are numerous demonstrations of social learning during foraging (e.g. in bumblebees [103], honeybees [104] and ants [105]). This might, therefore, be a promising aspect to investigate.

Other future lines of investigation could tackle the effect of emotions/experience (positive or negative) and sleep on memory and learning during recognition. It is increasingly acknowledged that affective state may influence cognitive processing such as attention, memory and judgement in animals as in humans (i.e. ‘cognitive bias’ [106,107]). Starlings have been shown to exhibit a pessimistic bias when they had recently experienced a decline in environmental quality [108]. In honeybees, individuals were more likely to interpret an ambiguous stimulus as predicting a punishment after having experienced stress [109]. Recently, a similar case of judgement bias has been demonstrated in ants [110].

Eventually, as has been shown in humans and other animals, sleep facilitates learning and memory consolidation. In the zebra finch, patterns of neural activity elicited by the playback of the BOS during sleep matches activity during daytime (when the bird sings) in many brain nuclei of the song control system [111]. Moreover, sleep affects the developmental learning of birdsong [112] and facilitates the long-term consolidation of memory trace during filial imprinting [113]. In social Hymenoptera, there is no direct evidence of the effect of sleep on recognition but Hussaini et al. [114] found that sleep deprivation in honeybees significantly reduced extinction learning. As the most common view in social Hymenoptera is that cue learning is necessary for nest-mate recognition to occur [70,96,115], one might hypothesize that sleep also affects nest-mate recognition in social Hymenoptera.

Data accessibility

This article has no additional data.

Authors' contributions

N.R. drafted the manuscript; S.D. critically revised the manuscript and helped draft it. All authors gave final approval for publication and agree to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

We received no funding for this study.