Behaviour is a non-trivial concept that can admit a variety of definitions depending on the discipline, the theoretical approach and even the biological model of predilection (Levitis et al., 2009). Explicit and truly operational definitions are lacking even in the very ethological literature. Here we propose to define behaviour as observable movements and postures that allow individuals to act on their environment and therefore maintain an adequate internal state for survival, growth, and reproduction. When behaviour is often presented as the “motor output” triggered by the coordinated analysis of stimuli present in the environment (see for example Hogan, 2021), we believe that it is of utmost importance to stress the agency property of biological organisms (Gomez-Marin and Ghazanfar, 2019). From an operational point of view, agentivity is a crucial parameter to distinguish between behaviour, which is what an animal does, and any other movement or posture of the animal that is merely due to the action of the environment. We can agree with the definition proposed by Levitis et al. (2009): “behaviour is the internally coordinated responses (actions or inactions) of whole living organisms (individuals or groups) to internal and/or external stimuli, excluding responses more easily understood as developmental changes”. We however stress that the notion of “collective behaviour” is a metaphor referring to the emergence and maintenance of collective patterns structuring (generally by means of self-organised process) the behaviour of individuals within the group. We propose that the determinants of behaviour are rooted in the interaction between the individual organism and its environment.

Behaviour should not be seen as a mere reaction of organisms to pre-existing stimuli in the environment, but rather as the way an organism actively regulates its interactions with its environment at the organismic level, including how it controls its perception of relevant features of its surroundings (Pellis and Pellis, 2021). Then behaviour can be described as an emergent property of the functional loops linking action to perception (Varela et al., 1991; Thompson, 2010).

If we endeavour to propose an integrated approach to animal behaviour, we must first define how we envisage the four questions in this review. The causes encompass all mechanisms responsible for the expression of behaviour here-and-now, i.e., the internal state and processes of the animal, and the environmental context that triggers, support and/or orientate this behavioural expression. This includes the sensory, neural, physiological and morphological constituents of behaviour within the organism (for example, the mechanisms approach can be taken by analysing how pheromones affect foraging behaviours; Ma et al., 2019). Some of these mechanisms, such as genetic and epigenetic factors, are of utmost importance since they can act as supports of inheritance for behavioural traits. Other processes are associated with potential modifications of the behavioural response, such as sensory modification, learning and experienced-based mechanisms, and motivational processes (Lehrman, 1953). This means that the internal state of the animal is also an object of study. Although not a behaviour per se, it is affected by the interactions of the animal with its environment. Reciprocally, it affects all aspects of behaviour since it influences the animal’s point of view on its surroundings (d’Ettorre et al., 2017; Hogan, 2021). This stand sharply contrasts with the “anti-subjective” stances of Tinbergen in 1963 (e.g., p.413), which must be properly replaced in its historical context. Nowadays most academic textbooks in Ethology contain sections dedicated to Cognitive ethology, or at least referring to the link between cognition and behaviour (e.g. Shettleworth, 2010; Bolhuis et al., 2022). This also stands for monographies on insects, especially for Hymenoptera (e.g. Wehner, 2020; Chittka, 2022). Noteworthy, we propose that after almost fifty years of cognitive ethology (Griffin, 1976; Allen and Bekoff, 1997; Andrews, 2015; Bolhuis et al., 2022) even subjective states of animals can nowadays be scientifically studied and should be considered in explanations of behaviour when possible (for recent reviews see Birch et al., 2020; Paul et al., 2020). For example the standard method of cognitive judgement bias test for assessing emotion-like processes in non-human animals, where the behavioural reaction of an animal confronted to an ambiguous situation depends on its affective state (for a review see Roelofs et al., 2016), can be adapted for insects (Perry and Baciadonna, 2017). How conscious such internal states can be obviously remains a hard and vividly discussed question (e.g. Baracchi et al., 2017). It will undoubtedly benefit from close collaboration between cognitive scientists and philosophers (e.g. Veit, 2023).

The second of Tinbergen’s proximal questions, the ontogeny of behaviour, is intimately linked to the mechanisms of behaviour. Following Tinbergen (1963), one needs to consider the historicity of all living organisms at three intermingled timescales: (i) how evolution shaped organisms and led to the inheritance of behavioural traits (including both genetic and non-genetic mechanisms; Bonduriansky and Day, 2018; Jablonka and Lamb, 2020), (ii) how contingent individual experiences of the organism influenced its development (Spirov et al., 2021) and (iii) the actual expression of behaviour here-and-now. In this review, we will refer to this question as development, since the focus of Tinbergen on ontogeny did not consider all aspects of behavioural development, but rather insisted on the changes that occurred in the “behavioural machinery” during the animal’s development. Nowadays, researchers on behaviour consider this subject to be broader, including how experience and learning deeply modify the expression of the behaviours of individuals even after they have developed to their mature expression (Hogan and Bolhuis, 2005). Development is thus seen as an interaction process between the individual and the environment (Lehrman, 1953). This allows us to understand at a proximal level the great flexibility of behaviour and the decision process associated with it. Examples of studies that analyse behaviour at the developmental level are those that evaluate the effect of age and experience on the performance of foragers (Beshers and Fewell, 2001; Franklin et al., 2012).

The two distal, or ultimate questions of Tinbergen (1963), are related to the function and evolution of behaviour. The study of function, or current utility of behaviour (Bateson and Laland, 2013), aims at understanding its adaptive value, i.e., its consequences on individual fitness. It posits that behaviour is under the effect of natural selection as soon as: (i) its expression varies among individuals of a given population, (ii) it impacts individuals’ fitness and (iii) it can be transmitted through one of the various inheritance supports (Bonduriansky and Day, 2018; Jablonka and Lamb, 2020). This can be done by analysing the costs and benefits linked to a particular foraging strategy (Fewell et al., 1996). The study of evolution seeks understanding from a historical point of view how the different behaviours came to be selected in species and populations. Phylogenetic reconstructions of how foraging strategies emerged and diverged in the Formicidae group are an example of this type of analysis (Reeves and Moreau, 2019). Although it looks at evolutionary and genetic/epigenetic hereditary processes, several authors advocate that cultural processes and phenomena such as niche construction should be included in this question (Laland et al., 2016). This also stresses the fact that behaviour is not only a result of evolutionary processes but also plays a key role in evolutionary mechanisms. This circular causality (as for the feedback loop we underlined between behaviour and the perception of the world) has been put forward in recent advances in evolutionary sciences, starting with the field of EvoDevo (Laland et al., 2013). It is thus clear that the question of the evolution of behaviour is not independent of the proximal questions addressed above since it is both a cause and a consequence of the observed behaviours.

From the 1960s onwards, ethology branched out, and ethologists no longer saw themselves as researchers who dealt with the same problem using different approaches (Griffiths, 2007). Depending on the kind of study, explanations may be limited to proximate (in fields of behavioural analysis, cellular and molecular approaches to behaviour) or ultimate levels (typical in behavioural ecology) (MacDougall-Shackleton, 2011). The behavioural studies also got used to not specifying which approach would be utilised in many cases, despite several authors having highlighted potential problems with these blurred distinctions (Hogan, 2015). Thus, there seems to be a mismatch between the theoretical foundation and the research in ethology.

Foraging is a suite of behaviours that together allow an animal to obtain food resources. It is traditionally associated with the hunger motivational system, and it ultimately permits self-maintenance through the resources obtained. Foraging behaviours are not restricted to food consumption but also encompass all the aspects associated with the localization, exploitation, food processing and sometimes stocking of resources. Consequently, they include a wide diversity of behaviours that can be articulated in a variety of behavioural strategies. In social insects, which display a social division of labour, foraging is usually performed to provide food to individuals who do not forage by themselves. In this case, individual behaviours should then be explained both at the individual and collective levels (e.g., social regulations and auto-organising phenomena).

There are several reasons to study foraging, such as the obvious adaptive function of these behaviours, their enormous diversity (such as different foraging strategies) and complexity, and the ease of being analysed analytically and mathematically which generated numerous theories (Griffiths, 2007). Unsurprisingly, foraging is probably one of the most studied themes in ant species, whose behavioural repertoires are notably vast. Extremely remarkable phenomena can be observed in many ants, for example, the massive recruitment of thousands of workers to a resource (Reeves and Moreau, 2019). Also, ants occupy diverse ecological niches and present feeding habits that range from generalist predators and detritivores to fungi cultivator specialists (Dornhaus and Powell, 2010). Food acquisition, done by a small part of group members to feed the whole colony, requires the resolution of several problems that vary according to species, ecology, life history traits, and immediate context (Traniello, 1989). Generally, this activity consists in the discovery, retrieval, and transportation of the food resource to the colony, where most of the time, other individuals, mainly the immatures, consume it.

Depending on the studies, the same term “foraging behaviour” can refer to a variety of behavioural items, studied at distinct levels of organisation. We here focus on behaviours concerning three distinct moments of foraging activity: the decision to start foraging, searching behaviours, and food recovery strategies/behaviours. In ants, many variables can alter the internal state of individual foragers and trigger foraging initiation ( Figure 1A ). These variables can be internal (e.g., response threshold) or external (e.g., social environment). After that, individuals use a variety of search strategies, information sources and cues to find food. For example, desert ants (Cataglyphis sp.) use sun light, idiothetic and visual odometers, path integration, learning of visual scenes and other strategies based on multiple sensory modalities to retrieve food resources (Wehner, 2020). Finally, food items are rescued and brought to the nest in different ways. Individual or solitary foraging refers to a situation where no cooperation or communication occurs during the discovery, capture, and transporting of food resources. Each forager then leaves the nest, searches for food, and transports it back independently (Jaffe, 1984). In collective or group foraging, usually, one ant performs the scouting role and, after finding food, goes back to the nest to recruit nestmates, which involves some kind of communication (e.g., using pheromone or tandem communication; Beckers et al., 1989).

The most common types of recruitment are tandem running, group and mass recruitment by pheromone trails (Jaffe, 1984; Beckers et al., 1989). Tandem running involves mechanical as well as chemical signals (Hölldobler et al., 1974). During a tandem run, one inexperienced ant (follower) follows an experienced ant (leader) closely, touching its gaster with its antennae. If this contact is interrupted, the ant leader stays in place and waits for the follower to catch up (Richardson et al., 2007). In group recruitment, one individual can recruit a small group of nestmates that keep physical contact with it (Lanan, 2014). Mass recruitment, instead, happens when pheromones are deposited from the nest to a target place; thus, several ants can be indirectly recruited to the destination (Lanan, 2014). Foraging by swarm or column happens in army ants without scouts and consists of many individuals going out together (forming a raid) in search of food (Lanan, 2014).

The quest for understanding how behaviours operate is old and can already be found in Aristotle’s writings (Bateson and Laland, 2013). In modern ethology, however, the mechanism evaluation lost momentum by not being able to satisfactorily explain the adaptive value of behaviours as investigated by behavioural ecology (Griffiths, 2007). Here, we consider known and hypothesised causes in different moments, levels of biological organisation to present a broad approach with many examples of mechanisms that initiate, operate, and finish foraging behaviours in ants ( Figure 1A ). It is worth mentioning that, here, causal mechanisms are taken on separately from each other and from the three other levels of explanation (development, function and evolution) for the sake of clarity. However, foraging depends on a set of factors which present intersections, do not act like separate or exclusionary and has intertwinement with ultimate factors (view introduction session). We organise this section following biological organization levels from infra-organism molecules to supra-organism ecosystems.

A hallmark of insect societies is division of labour, also known as task allocation among colony members. One subset of the main models of work self-organisation in social insects — built on both theoretical and empirical studies — proposes that nestmates belonging to different behavioural groups (e.g., nurses or foragers) show different response thresholds when faced with tasks-related stimuli (e.g., larvae or food odours; Beshers and Fewell, 2001; Duarte et al., 2012). As a result, workers collectively exposed to a variety of stimuli will show varying probability to responding with a given behaviour (Bonabeau et al., 1998; Duarte et al., 2012; Leitner and Dornhaus, 2019). As an illustration, experimental studies have shown that foragers present low response thresholds and therefore respond first to stimuli related to food and external environment, like the response to sucrose or light (Ben-Shahar, 2005; Perez et al., 2013; Detrain and Prieur, 2014).

In social insects, behavioural differences can also be traced at a colonial level. Colonies can indeed differ consistently from others in task performances and activity regulation (Webster and Ward, 2011). Such variations of behavioural responses that are consistent across time and contexts are referred to as personality traits (Stamps, 2016). Gordon et al. (2011) for instance showed that, in Pogonomyrmex barbatus (Smith, 1858), colonies can vary in the scouts rate and the number of foragers leaving the nest. These inter-colony differences persist each year (Gordon et al., 2013), thus validating the characterization as colonial personality across time. Colonies of Pogonomyrmex occidentalis (Cresson, 1865) show consistent variation in the temporal pattern and thermal tolerance during foraging (Cole et al., 2010). Colonies of Temnothorax rugatulus (Emery, 1895) vary consistently in foraging efforts (Bengston and Dornhaus, 2014). In Lasius niger (Linnaeus, 1758), colonies with greater exploratory levels exploit sources faster (Pasquier and Grüter, 2016). Despite being able to quantify traits to the colonial level, these results arise from a complex interaction of factors, one of them being the characteristics of the individuals that make up the group. Therefore, one weakness of these studies comes from the lack of hypotheses and tests of the mechanisms associated with such consistent differences and their permanence in natural environments. Another flaw is that most analyse personality traits over time but not across different contexts.

Colonial success also leads to the selection of individuals with specific traits. In that regard, personality on an individual level is one of the elements responsible for characteristics at the group level. However, it is hard to predict and evaluate experimentally how such collective properties can emerge from the association of many individuals with different traits. Today, three hypotheses could explain how individual personality acts towards building colonial personality: 1) colonies differ consistently in the personality average of their workers; 2) workers’ personality distribution, not average, varies between colonies, and 3) colonial personality does not emerge from workers’ personality but from consistent external limitations (Pinter-Wollman, 2012). As an example, the ingested volume threshold in L. niger is specific for each individual, independent from its size, and stays constant even with successive trips to the food source. There are also differences in the propensity to deposit trail pheromone, with some foragers never laying trails (Mailleux et al., 2005). Learning ability can also be correlated with the exploratory behaviour of foragers. In Camponotus aethiops (Latreille, 1798), more active explorers were slower to learn an odour than less active ones when harnessed in a conditioning maxilla-labium-extension paradigm (MALEr, Udino et al., 2017).

All these variations in behavioural traits probably shape the colonial phenotype. For example, dividing foragers into scouts and recruits, trail followers or innovators, leaves the colony more flexible to respond to the environmental conditions. Little is known about the ontogeny and the mechanisms responsible for personality differences at an individual level (Horna-Lowell et al., 2021). These differences can hamper the flexibility of the colonies and thus explain the differences in group fitness that collectively show different responses to the environment, notably in foraging.

Individual foragers also base their decision to forage on the information received from nestmates (social information). Information transfer can occur directly, mostly through antennal interactions, or indirectly through pheromone trails (Kolay et al., 2020). In P. barbatus the combination of these two processes stimulates foraging (Greene et al., 2013). Ants interact with conspecifics through antennal contacts, during which they identify the nestmate’s chemical profile of cuticular hydrocarbons (Greene and Gordon, 2003). Encounters between foragers coming in and out of the nest convey information on the identity of the nestmate, the type of food found, and the food source abundance. The rate at which successful foragers return to the nest entrance reflects the food type available and the time it takes to find food (Schafer et al., 2006). Foragers’ returning rate is the local information use by workers to regulate their own foraging effort, without the need to get direct access to actual food availability.

Social facilitation has been observed in several ant species where the return of successful foragers in the nest triggers the initiation of foraging bouts by nestmates (Holcoponera moelleri (Forel, 1912), Cogni and Oliveira, 2004; Dinoponera quadriceps (Kempf, 1971), Nascimento et al., 2012; Odontomachus bauri (Emery, 1892), Oliveira and Hölldobler, 1989; P. barbatus, Schafer et al., 2006). Decision-making models reported that the odds of a forager going out for food depend on its interactions with returning foragers (Greene et al., 2013; Pinter-Wollman et al., 2013). Foragers inside the nest engage in a series of antennal contacts with returning foragers. When the frequency of these contacts is high, ants are more likely to go out of the nest to forage. This mechanism was modelled successfully with a stochastic accumulation of contact numbers (Davidson et al., 2016). In social insects’ colonies, individuals only have access to local information depending on their nest position and the presence of nestmates in the vicinity (Mersch, 2016). Foragers lack direct knowledge on the colony’s nutritional needs. Thus, social communication is critical to regulating foraging activity.

Among all colony members, larvae are the most energy-demanding individuals to fulfil their development, and they usually are the end consumers of the collected resources. They could therefore play a key role to stimulate foraging by adult nestmates (Ulrich et al., 2016). Larvae can indeed signal their hunger level to workers (Cassill and Tschinkel, 1995; Creemers et al., 2003). Both mechanical and chemical signalling have been evidenced in ant larvae (Schultner and Pulliainen, 2020). Larvae number, size, and nutritional needs influence the workers’ foraging activity. For instance, the frequency of trophallaxis — mouth-to-mouth food transfer between colony members — is correlated to the larvae/worker ants’ proportion (Cassill and Tschinkel, 1999). In addition, workers respond to larvae number by increasing foraging activity and inhibiting ovarian activation in a progressive manner, especially in smaller colonies (Ulrich et al., 2016).

Before Tinbergen, Julian Huxley had defined three aspects of research in biology: causation, adaptive value, and evolution (Huxley, 1923; Tinbergen, 1963). He and other researchers received criticism from developmental biologists (e.g., Lehrman, 1953), and from that, Tinbergen added a fourth why to the ones known in the literature: ontogeny. Ethologists and psychobiologists since then developed a vast literature on attachment, parental and sexual imprinting, and song development in birds, among other themes, explicitly addressing the question of how behaviour emerges during individual development (Bateson and Laland, 2013). In a broader sense ethologists consider that complete explanation of any behaviour includes to shed light on its historicity at the individual timescale.

Ants, as holometabolous insects, undergo a complete transformation during their development (egg, larva, pupa, and adult). Despite foraging behaviour being expressed only in adult life, it can be affected by pre-imaginal events. Here, we review how morphological and age modifications, as well as individual experiences, affect foraging behaviours in ants both at the individual and collective level ( Figure 1B ).

Tinbergen (1963) praised Lorenz for stressing the role of behavioural patterns as adaptations, using his famous analogy with organs. Today, the functional aspect of behaviour is by far the most studied in ethology and has been the foundation for the rise and maintenance of all research in behavioural ecology (Bateson and Laland, 2013).

Behaviour is, by definition, the property of an organism, even if one can metaphorically speak of ‘collective behaviour’ to characterise group properties emerging from the interaction of individuals – like for a flock of geese or the structure of raid columns in army ants (Deneubourg et al., 1989; Gordon, 2019; Walsh et al., 2020). Survival and reproduction of fertile individuals are essential for each sterile worker’s inclusive fitness (Bourke, 2011), and the former depends on optimised functioning of the colony as a whole, which is frequently subsumed under the metaphor of the colony forming a superorganism (Hölldobler and Wilson, 2009; Figure 1C ). Natural selection then act at both individual and group levels by selecting colonies that present individuals with specific advantageous characteristics (Hölldobler and Wilson, 1990).

A variety of foraging strategies, integrating individual behaviours within the social context of the colony, have been selected for. Foraging is a risky task for a social insect, since it means leaving the shelter of the nest and enduring many costs. Besides energy and time expenditure to search for food, foragers face an increased risk of death by predation, desiccation and pathogens or simply by getting lost (Hölldobler and Wilson, 1990; Nonacs and Dill, 1990; Friedman et al., 2019). The classical evolutionary calculation of the costs and benefits balance must be envisaged at the group level, for which typical task allocation system is a good illustration. As described in Section 3.3.1, foragers in species displaying age-polyethism are the oldest and less fertile members of the colony. One can estimate that losing an old nestmate is less prejudicial than losing a young worker, since the colony can benefit from the latter’s workforce for a longer time (Beshers and Fewell, 2001). Here again, combining ultimate (costs and benefits calculation at the colony level) with proximate (development of individuals) explanations of behaviours provide a deeper understanding of foraging in ants.

General economic theories have been proposed to evaluate gains and losses associated with foraging in animals. The main theory is optimal foraging, which builds on the principle that behavioural sequences of animals have been selected for because they maximise energy acquisition by time unit (MacArthur and Pianka, 1966). At a cognitive level, literature usually refers to (non-conscious) decision mechanisms as a basis for foraging behaviours (but see Wystrach, 2021). Those ‘decisions’ affect behavioural transitions, such as the time spent searching for food in a patch before moving on to the next, as well as the fine-tuning of behaviour, such as the type of food that is searched for.

Another core concept commonly applied to ants and other social insects is the theorem of central place foraging, where animals travel through an area searching for resources and return with them to the initial and central place, the nest (Orians and Pearson, 1979). One of the theorem’s predictions is that the farther from the central point the search occurs, the better the food quality must be, so there is real energetic net gain (Orians and Pearson, 1979). In Pogonomyrmex californicus (Buckley, 1867), several predictions of optimal foraging theory and central place foraging were tested (Holder Bailey and Polis, 1987). Ants selected seeds in function of both size and handling time and success, thus maximising the energy/time ratio. Selectivity was also influenced by the relative abundance of the less preferred food, another prediction of optimal foraging theory (Hughes, 1979). Last, in contrast with central place foraging theory, the energy intake did not differ according to the distance of the patch (Holder Bailey and Polis, 1987). P. clavata forage in the canopy for nectar sources and insect prey (Breed et al., 1987; Fewell et al., 1992), adapting their recruitment strategies to energy intake rate, that is recruiting more frequently to insect prey than to nectar sources. Fewell et al. (1996) performed a comparison of individual and colony energy intake and expenditure between P. clavata and the seed harvesting ant P. occidentalis. In the first species, the benefit to cost ratio was much smaller than in the latter one. This suggested that workers should flexibly adapt their foraging strategies, which they did, and that colony growth was much more constrained by energy intake. As a result, those constraints, mainly associated with ecological parameters, probably also influence reproductive strategies. P. occidentalis can stock large quantities of reserves, resulting in massive seasonal production of alates, while P. clavata is limited in its resources, and produce few alates in a continuous manner (Fewell et al., 1996).

It is often assumed that solitary strategies would be ‘less efficient’ than collective strategies, especially mass recruitment (Hölldobler and Wilson, 1990). This and other kinds of recruitment would help colonies find and monopolise food sources (Detrain and Deneubourg, 2008). However, recent studies, experiments, and models stress that no strategy can be said the most efficient for all species. Unsurprisingly, the efficiency of any foraging strategy is highly context-specific and varies according to the food sources characteristics (type, size, distribution and quality), colony size and ecology (Beckers et al., 1989; Lanan, 2014). Solitary foraging appears very efficient in environments with dispersed distribution of unpredictable food sources, as exemplified by scavenger desert ants (Wehner, 2020).

In a wide variety of contexts, independent search for food brings more energetic gain than recruitment since it speeds up food-collection: other foragers do not need to wait for the return of successful individuals to start the search (Dechaume-Moncharmont et al., 2005). Additionally, mass recruitment through pheromone trails is very costly and considered less efficient for small ant colonies (Beckers et al., 1989). Some empirically validated models also predict that species which mass recruit to exploit a resource could difficultly switch to a new and more valuable food source (Sumpter and Beekman, 2003; Detrain and Deneubourg, 2008). This happens because the positive feedback established surmounts any new trail initiated to the better source (but see Czaczkes, 2014).

Lastly, social foraging strategy emerges from the expression and articulation of individual behaviours, and can show some flexibility within a given species. As an example, Pachycondyla striata (Smith, 1858) workers can switch from solitary foraging to recruitment by tandem running depending on the nature and distance of food (Silva et al., 2021). In species that mass recruit, workers can nevertheless choose not to follow the chemical trail and search for new food sources (Czaczkes et al., 2011), as they do in L. niger. In these cases, individual experience plays a key role in the emergence of an adaptive response of colonies to environmental conditions.

Mayr (1961), earlier than Tinbergen, defined ‘ultimate causes’ as evolutionary advantages attributed to biological characteristics. This definition, however, had two problems: first, it mixes adaptive function and evolutionary history of the trait in a single aspect, and second, it is confusing by the word ‘cause’, which was frequently used by researchers investigating mechanisms, a proximate approach (Bateson and Laland, 2013). Tinbergen’s (Tinbergen, 1963) contribution in this regard aimed to organise this level of analysis conceptually and temporally.

The phylogenetic reconstructions of the evolution of foraging strategies in ants are not easy, since this group presents more than fourteen thousand known species (Bolton, 2023; Figure 1D ). With a huge diversity of diets, foraging behaviours, and foraging strategies, seldom do we find similarities, nor can we trace back to where and when some particular patterns, like recruitment, came to exist. Here, we summarise the literature that debates the evolution of foraging in ants. The fact that many species recruit for new nests but not for food raises the hypothesis that recruitment was first selected for emigration, and later co-opted for communicating the location of food sources (Glaser, 2022). A similar evolutionary path has been suggested for the bee dance (Beekman et al., 2008; l’Anson Price and Grüter, 2015). Nest-related behaviours could indeed play a paramount role in the evolution of the social behavioural repertoire in social insects (Bos et al., 2011; Robinson, 2014). The value of the nest is so high for ant colonies’ survival (except for some exceptions like army ants) that there could have been a strong selection enabling fast and precise relocation in case of nest damage, even for individuals that had no experience of the outside world (Franks et al., 2003; Dornhaus et al., 2004). Recruitment for food sources, on the other hand, could be under weaker selection pressures since foragers can uncover new sources for themselves, especially when they are abundant and evenly distributed (Dechaume-Moncharmont et al., 2005; Dornhaus et al., 2006; l’Anson Price et al., 2019; Goy et al., 2021). Note that recruitment nevertheless can bring a real advantage over competitors, namely in interference competition (e.g., Cordonnier et al., 2020).

It is believed that the evolution of foraging strategies has followed a succession model where, throughout evolutionary times, one strategy would replace the other (Glaser, 2022). Nowadays, two putative evolutionary paths are proposed. The first one considers solitary foraging as a basal strategy. Tandem running could have emerged in some groups, then evolving to group foraging and lastly, mass foraging as the most derived strategy (revised by Traniello, 1989). Recent phylogenetic analyses support this hypothesis, showing that the Formicidae group first displayed solitary foraging a basal state and then evolved ‘more complex’ strategies (Reeves and Moreau, 2019; Figure 1D ). However, this study is flawed by analysing strategies at the genus level. The authors indeed attributed to each genus the strategy that was the most frequently observed between the congeneric species, therefore masking the diversity existing within each genus. This is all the more important that genera themselves varies tremendously in their diversity. For example Camponotus harbours 1504 species, Pheidole 1294 while Pachycondyla counts only 17 species (Bolton, 2023).

An alternative hypothesis considers that the first ants displayed medium-sized colonies of around few hundreds of individuals (Burchill and Moreau, 2016) and performed mass foraging (Glaser, 2022). As the group differentiated, there was recruitment loss in some taxa, leading to the emergence of group foraging and later of solitary foraging, which in this scenario is the most derived strategy (Glaser, 2022). This hypothesis considers foraging strategies at the species level but is flawed by recruitment being considered both for foraging and nest emigration. Besides, it is based on the reconstitution of the common ancestor made by Burchill and Moreau (2016), where information used is not sufficient to firmly infer the foraging strategy — since species with medium colony size can use different strategies and use solitary foraging. Last, it is not consistent with the previously cited reconstitution (Reeves and Moreau, 2019), using a different database and level of analysis.

It is essential to point out that actually there are not enough data to robustly support any of these hypotheses, mainly because the social organisation of foraging is still unknown or registered anecdotally for many ant species. Moreover, it is noticeable that recruitment evolved independently multiple times in different subfamilies, so a simple succession and substitution mechanism cannot explain the range of existing strategies and their distribution in the clade of living ants (Reeves and Moreau, 2019). For instance, the species Neoponera marginata (Roger, 1861b) belongs to the Ponerinae subfamily, originating from a generalist and solitary hunting ancestor, which is also the most frequent phenotype of extant Neoponera species. N. marginata nevertheless displays swarm foraging, like attack columns of army ants, and a termite-specialised diet, both characteristics considered derived in Formicidae (Leal and Oliveira, 1995). Last, one can also deplore a research bias towards ant species that perform group or mass recruitment, neglecting solitary foraging ant species.Few recent studies use comparative methods to study the diversification of foraging behaviour in ants and its relationship with environmental conditions, especially in groups that perform solitary foraging (Reeves and Moreau, 2019). There is also a need for studies that tackle intraspecific behavioural differences between different ecosystems or biomes, a very interesting tool to understand flexibility, plasticity and potential niche, “cultural” and/or genetic populational differentiation. Previous speculations on how foraging evolved and differentiated suggest no correlation between phylogeny and the foraging method (Jaffe, 1984; Urbani, 1989). However, these studies are based on insufficient data and outdated taxonomy.

Ant behaviour is a complex subject that can be approached through various lenses. In the last sections, we brought examples of studies that address ant foraging behaviour focusing on each of Tinbergen’s four questions. Figure 1 serves as a visual synthesis, capturing these complementary approaches. The major known mechanisms, that triggers and shape the expression of behaviours associated with foraging in ants are diverse ( Figure 1A ). For the sake of clarity, we chose to present here a simplified version of the actual network of the factors involved in foraging behaviours. These factors can be internal or external to the individual and usually they interact with each other (not all interactions are shown, again for the sake of clarity). Thus, it can be challenging to isolate and measure their specific effect on foraging decision-making. We propose that the focus should rather be on the elucidation of the network of factors and their interactions, since the expression of foraging arises from the global functioning of this network. These factors and processes are modulated throughout the ants’ development and affect the propensity to forage, and the way behaviours associated with foraging are expressed ( Figure 1B ). Again, while some changes are generated by internal factors (genes, hormones), others are activated mainly by external variables, such as different feeding regimes in larval stages that lead genetically identical individuals to present different adult phenotypes, both for morphology and behaviour. Developmental influences are not confined to the immature stages; factors such as foraging experience in adulthood, age, and colony needs continually shape foraging activity across an individual’s lifespan.

Considering that evolution ultimately acts both at the individual and colonial level, Figure 1C illustrates how foraging optimization in foragers, a caste of sterile workers, leads to increased colonial reproductive success and increased individual and group fitness. Finally, the diversity of foraging strategies and how they have diversified within the ant clade cannot be explained using one or a few biological or ecological parameters ( Figure 1D ). Ants foraging diversity also allows little room for generalization about how foraging evolved in these insects. Within the same subfamily, there is a wide range of different foraging strategies (and consequently, behaviours associated with them). This wide variety is also observed within ants of the same genus and in notable cases, even within the same species, with ant species showing polymorphism in foraging strategies. Thus, although a general picture of the evolution of the different strategies is difficult to draw, the repeated evolution and reversion of the different strategies across many taxa allows us to study in detail these events. This also suggests that the mechanisms and the behavioural machinery associated with the expression of foraging associated behaviours may be similar across species, but that specific ecological and selective pressures may lead to small modifications, for example during development, that lead to drastic changes of the behaviour at the collective level.

One of the best examples of integrative research in behaviour is the investigation of foraging in Pogonomyrmex species, carried out mainly by Deborah Gordon’s group. For the last decades, researchers made a great effort to evaluate most of the aspects influencing the foraging of those seed-harvesting ants, making this group one of the most studied. Several studies were developed addressing the four different questions and those were integrated and discussed together. For this reason, we discuss how integrative efforts can be made using those articles as examples.

In seed-harvesting ants, many factors can simultaneously alter behavioural activity and gene expressions, such as the circadian rhythm (Ingram et al., 2016), seasonal climatic effects, and species (Das and Gordon, 2023). Desiccation is a powerful selection pressure in desert ants. Friedman and collaborators (Friedman et al., 2019) showed that individual workers adapted their foraging activity to both their hydration state and the environmental risk of water loss. They then investigated at a genomic level the individual variation in tolerance to desiccation (Friedman et al., 2020). Gene expression was linked to sensitivity of foraging activity to humidity and was associated with dopamine to serotonin brain ratio (Friedman et al., 2020). Dopamine titre plays a key role in the initiation of foraging bouts, especially in colonies that were the most sensitive to humidity (Friedman et al., 2018). On a collective level, several works pointed out the importance of interactions among workers at the nest entrance in individual foraging decisions. Foragers display distinct cuticular hydrocarbon profiles, and encountering these profiles motivated workers to start foraging (Greene et al., 2013). More specifically, the amount of interactions with returning workers dynamically and positively influenced foraging decisions. Besides, workers who interacted less also moved further away from the nest entrance, thus being less likely to be activated (Pinter-Wollman et al., 2013). Transitions from inside tasks to foraging showed ample variability, and outside workers could change tasks rapidly when the context changed (Gordon, 1989; Gordon et al., 2005).

Long-term studies were also conducted to stress distal behavioural questions, such as function and evolution. Through observations over 30 years in the same populations, sensitivity to humidity of workers proved to be an important characteristic positively affecting reproductive success of the colony (Gordon, 2013). In the competition context, colony age and colony size influence foraging strategies (Gordon, 1992). Adler and Gordon (2003) modelled the competitive interactions of Pogonomyrmex colonies using optimal foraging theory. The results showed that the adaptive response of individuals and colonies depended on the existence of competition with adjacent colonies. They suggested that individual optimization of foraging behaviour explained well empirical data, even if in the collective model colonies could allocate a larger than optimal number of foragers to boundaries with neighbours, suggesting that both levels probably influence collective response (Adler and Gordon, 2003). Even though seed-eating is common in myrmicine ants (genera Messor, Aphenogaster, Pheidole, etc.), seed-harvesting behaviour and preference for xeric environment is derived in the tribe Pogonomyrmicini (Ward et al., 2015). This makes comparing species extremely relevant to understand evolutionary ecological patterns (Johnson, 2000). Several Pogonomyrmex spp. Occur in sympatry in the Sonoran Desert. Their foraging behaviour is also determined by this co-occurrence of species with the same diet. Some species are dominant and defend food resources and territory while others, with smaller workforces, tend to avoid other species when foraging (Gordon, 1984).

In conclusion, the set of research carried out on seed-harvesting Pogonomyrmex gives an unprecedented understanding of mechanistic, ontogenetic, ecological and evolutionary aspects of foraging. We advocate that similar research efforts should be carried out on more species, in an attempt to cover the range of foraging strategies (from solitary to mass recruitment). The fact that this research spans over forty years shows that integrating all these aspects is a time-consuming process that needs perennial and experienced research teams. Unfortunately, most research groups can face limitations while trying to integrate proximate and ultimate causes. Not all ant species can be kept in laboratories or have a fixed population in the same region throughout the years. Laboratories usually have a higher rotation of members, who have short-duration projects and restricted budgets.