COLONY SIZE. COMMUNICATION AND ANT FORAGING STRATEGY

The foraging strategy of 98 ant species is examined in relation to their colony size. Six foraging stratégies are distinguished, namely individual, tandem, group/mass and mass recruitment, tnjnk trail, and army ant type, and are seen to be associated with increasing colony size. This supports the hypothesis that the larger the colony, the more the individual worker is integrated into a network of Chemical communication. Two extrême organisational blueprints are proposed. The first consists of small societies which rely on the capacity for learning of its members to exploit the foraging area efficiently. The second relies on the complex collective patterns that spontaneousiy émerge from chemically mediated recruitment processes interacting with the environment.


INTRODUCTION
Some 12,000 ant species are known by now, with colony sizes ranging from a few individuals to 20,000,000 individuals. What constraints does this vast range of colony sizes place on the Systems of organisation that they use? Alternatively, how does this range of colony sizes reflect the différent Systems of organisation used? We shall examine thèse questions in relation to ant foraging strategy, which as well as being the most visible aspect of their activity illustrâtes most clearly the rôles and limits of communication in their collective behavior.
This paper aims to verify a prédiction of the following hypothesis (Pasteels et al. 1985;Deneubourg et al. 1986). In theory, the organization of a small insect Society can rely on most individuals at any moment "knowing", principally by learning, what it must do, where it must go, etc., and the workers' behavior has a strong determinist component. In a large insect society organization by individual learning is harder to achieve . The workers' behavior is necessarily more random and their coordination becomes a major problem. To cope with this, a completely différent organisational System is added to that aiready in place. This supplementary System is based on the complex collective structures, patterns and décisions that spontaneously émerge from simple autocatalytic interactions between numerous individuals and with the environment, mediated by essentially chemical communication (see, e.g., to inactive foragers waiting in tlie nest. Thèse recruits can become recmiters in their turn. It should be noted that recruiting species rely to a large extent on individual foraging for the discovery and exploitation of small sources. Roughiy speaking, three recruitment types can be distinguished. With tandem recruitment, the scout guides one recruit to the food item, with or without trail laying (e.g. Leptothorax sp.). With mass recruitment, a trail laid by the recruiter while returning to the nest guides recruits to the food (e.g. Solenopsis invicta, Monomorium pharaonis). Invitation by the recruiter in the nest is often active. With group recruitment, the scout guides a group of nestmates, in some cases (if not al!) laying a pheromone trail to the nest (e.g. Tetramohum caespitum, Camponotus socius). However, as every species that we know uses group recruitment aiso uses a more or less efficient mass recruitment, we shall refer to group/mass recruitment. Note that some authors distinguish a fourth recruitment System, group raiding (type IV -Oster and Wilson 1978), which is characterised by a very strong invitation and recruitment trail that results in a large group of recnjits leaving the nest together in a njsh. We have included this System in group/mass recmitment.
With trunk trails, semi-permanent trails guide foragers to long-lasting food sources (e.g. many Formica sp.), and aIso serve as starting-off points for individual foragers, which may aIso recruit to the tnjnk-trails. Finally, group hunting foragers (sensu Moffet 1988, including army ants) leave the nest and forage collectively in a swarm along a well-defined trail System that is constructed as the swarm progresses (e.g. Dorylinae sp.).
Thèse descriptions are by no means meant to be définitive, and there are of course species whose foraging does not fall neatly into one or indeed any of thèse catégories. Nevertheless, as shall be discussed below, they represent a crescendo of the intégration of the individual forager into a network of communication. Other recruitment Systems are known to exist (such as shortdistance recruitment or non-directional recruitment), but for lack of data have been omitted. Similarly, the colony sizes given are average figures, obtained by différent techniques, and generally with rather small sample sizes. Polycalic societies pose a spécial problem. For thèse reasons, the values quoted must be considered only as first-order approximations. Table I présents the colony size and foraging System of 98 différent species. Fig. 1 présents the the foraging System as a function of the colony size.

RESULTS AND DISCUSSION
Although the data base is small compared to the number of known ant species, a distinct trend is clear (note the logarithmic scale). The smaller societies rely on individual foragers that do not transmit their discoveries. The largest societies rely on permanent Chemical communication between the individuals. Between thèse two limits, one finds the différent types of recruitment. Again, the smaller recruiting societies rely on a slow, individual recruitment, where a recruiter interacts directiy with one or a few individuals. The larger recruiting societies rely on the faster mass recruitment, where one recruiter can interact via a chemical trail with a large number of potential recruits. The trail transmits both the position of the source and that of the nest to the recruits. Note that we have listed the species in Table I by alphabetical order for facility, and that the same tendancy appears in each sub-family.
Taking thèse results into considération we propose two extrême blueprints for the way in which ants organise their foraging.
The first blueprint consists of small societies which rely on the capacity for learning of its members to exploit the foraging area efficiently. Individual foragers, for example, develop fidelity to parts of their foraging area and can orient themselves over large distances (Fresneau 1985;Wehner et al. 1983). They do not interact directiy with each other, nor do they communicate their food discoveries, yet they are capable of dividing the foraging area amongst themselves ). The society may be considered to have placed its complexity at the level of the individual.
The second blueprint consists of large societies whose individual behavior may be considered as simple. They rely on a highiy developed network of Chemical communication based on permanent trail-laying behavior to coordinate the foragers' activity and to aid their orientation. Their capacity for individual orientation is limited not only because it is not so needed as the trails are there, but also because too much individuality could prevent collective foraging from functioning efficiently. It is surely no coïncidence that the largest and most chemically integrated societies, i.e. the différent army ants and termites, are practically blind. The colony size is large, not simply to ensure that their "inefficient" workers manage to perform the necessary tasks by sheer weight of numbers (Oster and Wilson 1978;Herbers 1981), but because they need a large reserve of individuals for the amplifying mechanisms (e.g. recnjitment) by which they structure their foraging to work (e.g. Pasteels et al. 1987). The society may be said to have placed its complexity more at the level of the interactions between individuals.
Between thèse two extrêmes, we find intermediate sized societies which rely on individual scouts to forage small food items and on recruitment to amplify the information relating to important food sources. The séquence tandem/groupmass/mass is characterised by an increasing number of individuals that react to the recruiters' signais, and is associated with an increasing colony size. In mass recruiting species there is a tendancy in the largest societies to lay trail pheromone not only when returning with food but aiso when leaving the nest and more or less continually in the foraging area (e.g. Pheidole militicida - Hôlldobler and Môglich 1980;Iridomyrmex humilis -Van Vorhis Key and Baker 1986;Aron et al. in press).
There is of course a large degree of overlapping between the différent catégories in fig. 1. This is to be expected whenever one tries to categorize nature, but is aIso the resuit of imprécision in our knowledge of colony size, which is anyway highiy variable for a given species. Furthermore, others factors such as the size, distribution and type of food exploited intervene, and ant foraging strategy and food type are obviousiy connected (e.g. Carrol and Janzen 1973;Traniello in press).
Other less précise data confirm the tendancy seen in fig. 1. For example we know that Crematogaster ashmeadi colonies are very large and that they use mass recruitment (Leuthold 1968a, b), whereas Bothroponera tesserinoda colonies are small and use tandem recruitment foraging.
The same overall tendancy as shown here for ant species, aIso noted by Buschinger (1980) for dulotic ants, is well known in the Apidae. Species with small colonies, such as bumblebees, use an individual foraging strategy. Those with large colonies, such as honeybees, melipones and trigones, use recruitment (Lindauer and Kerr 1958;Seeley 1985). The tendancy is aIso observed in terns (Erwin 1978).
We would like to end this paper with an appeal to readers to help us increase the size of our data base. We would welcome any information about colony size and foraging System, whether for species aiready in Table I or 1. Foraging strategy as a function of colony size for 100 ant species (see Table 1). The small vertical bars markthe 25, 50 (médian) and 75 percentiles.