The Ants Chapter 10

CHAPTER 10. FORAGING STRATEGIES, TERRITORY, AND POPULATION REGULATION

Ant colonies exploit the environment by wholly social means. They deploy aging workers into the field as expendable probes to blanket the terrain around the nest in a near continuous and instantaneous manner. In many species scouts that encounter a large food supply recruit nestmates in numbers appropriate to the size and richness of the discovery. When they meet enemies, they can recruit defenders or else withdraw, closing the nest entrance and lasting out the danger with the aid of food that has been conveniently stored in the crops and fat bodies of the younger workers.

In the formal framework of a theoretical ecologist an ant colony resembles a plant as much as it does a solitary animal. It is sessile, in other words rooted to a nest site. Its growth is indeterminate, so that under stress it can revert to the size and caste composition of younger colonies. As a consequence the size and age of colonies are only weakly correlated. As Rayner and Franks (1987) have pointed out, close parallels also exist between ant colonies and mycelia of fungi. “Both are collectives of genetically related or identical semi-autonomous units, consisting respectively of discrete multicellular individuals and hyphae.” The patterns created by the growing hyphae are remarkably similar in geometric design to the foraging columns and swarms of army ants and trunk trail systems of wood ants. In both cases the probes of genetically different cultures avoid one another, creating an interdigitating pattern that thoroughly exploits the available resources.

To continue the comparison, many ant colonies adjust to environmental change like plants more by flexibility of response than by movement from place to place. They attain this adaptability by a broad range of social behaviors, including adjustable alarm and recruitment, differential clustering, alterations of nest architecture, and shifts in foraging pattern.

Foraging theory
Foraging theory can be said to have begun, at least in mathematical metaphor, with Emlen (1966) and MacArthur and Pianka (1966). It has been extended greatly by Schoener (1971), Pyke et al. (1977), Orians and Pearson (1979), Stephens and Krebs (1986) and others. The special case of social insects has been theoretically treated by Oster and Wilson (1978) and advanced by many investigators through excellent field and laboratory studies.

Foraging theory is composed mostly of optimality models. The marginal value utilized, or “currency,” is usually energy. The modelers visualize the organism as seeking to maximize its net energetic yield. Evolution by natural selection, they reason, has modified behavior in one or the other of four basic ways so as to increase this currency: (1) choice of food items (optimum diet); (2) choice of food patch (optimum patch choice); (3) allocation of time invested in different patches (optimum time budget); and (4) regulation of the pattern and speed of movement. In the case of ants and other animals with a nest or other permanent retreat, the foraging patterns are analyzed in terms of sallies from a central location. This special case is the subject of “central-place foraging theory,” about which we will say more later.

Optimal foraging patterns are generally thought to be constrained by two forces, one external to the organism and the other internal. The external constraint is mortality due to accidents, disorientation, and attacks by predators and competitors. Many field studies have shown that worker populations of ant colonies are subject to terrible losses through predation and battles with competing colonies. It is also clear that the foraging patterns of individual species are profoundly influenced by these pressures. The internal constraints, on the other hand, are particular physiological constraints and the limited sensory and psychological capacities of organisms. Animals, and especially small-brained creatures like ants, can rarely if ever maximize their net energy yield by calculating exact costs, benefits, and mortality risks for each new situation in turn (of course, even human beings are hard put to accomplish that much). Instead, they are more likely to employ “rules of thumb,” which are quick decisions triggered by relatively simple stimuli or stimulus configurations. The rules of thumb work adequately for individual worker ants most of the time, and they may result in considerable precision when added up across multiple workers during the process of mass communication. Rules of thumb used by fire ants that we introduced earlier to illustrate colony homeostasis (Chapter 9) include the following: “continue hunting for a certain foodstuff if the present foraging load is accepted by nestmates” (it will probably satisfy the needs of the colony); “follow a trail if enough pheromone is present” (it will probably lead to food); and “feed the queen more if final-instar larvae are present” (this means that the colony developmental pipeline is in good working order).

In adapting foraging theory to the social insects, Oster and Wilson (1978) treated risk of worker mortality not as a constraint but as an energy cost that must be born to varying degrees by the colony. When a worker is killed in action, the game is far from over; the colony has merely lost a packet of energy that had been invested earlier through the processes of egg laying, larva rearing, and pupa care. The ultimate pay-off in colony-level selection is the summed production of reproductives over a colony generation. In general, ant colonies appear to act so as to build as large a population of workers as possible during the ergonomic phase, then “cash in” with a partial or full conversion to reproductives during the breeding season. Consequently, a reasonable measure of fitness for a colony in its ergonomic phase of growth is energy, with the growth rate of the adult biomass expressed as calories/unit-of-time and the full colony biomass achieved at the end of the ergonomic phase expressed in net caloric accumulation. The analytic advantage of employing energy as the basic currency is that it can be folded one way or the other into all colony processes. The key particulars are as follows:

• The biomass and biomass growth of the colony or any portion of it can be translated to energy units.

• Foraging success, brood care, and other nurturent activities can be translated into energy gained.

• Protective activities, including nest defense and construction, can be converted into calorie equivalents saved because of the reduction in mortality and impaired colony functions that such activity avoids.

• Metabolism and mortality can be converted directly into caloric loss.

It is of course an oversimplification to measure ergonomic success entirely by the energy equivalents of biomass. Reproductive success depends not just on colony size at the end of the ergonomic phase but also on the proportion of immature and mature forms as measured by the relative frequency of castes at this critical time. It also depends on the forms and kinds of “capital” at the disposal of the colony, particularly the amount of stored resources and the structure and location of the nest. Nevertheless, energy equivalence is probably the overriding determinant of reproductive success. The optimal caste frequency distribution and nest structure at the onset of the reproductive phase are unlikely to differ from those built up during the ergonomic phase, while few species store large amounts of food materials near the end of the ergonomic phase. In addition to such an audit of the total life span, analyses of particular components of foraging appear to be energy-efficient in the case of ants. As we shall see shortly, the time spent selecting food items and the degree of selectivity are consistent with the models of central place foraging theory, which are based on the postulate of net energy maximization. Also, in the case of leafcutting ants, net energy yield appears to be the key function among all those that might have influenced the evolutionary choice of leafcutting ant medias as foraging specialists over minors and majors (Wilson, 1980b).

Foraging behavior has two principal interlocking components: the search and retrieval of food items, often accompanied by the recruitment of nestmates, and the avoidance or defeat of enemies. It is feasible in theory at least to assess the relative importance of the two components by equating them in terms of energetic cost and gain.

The temperature-humidity envelope
Every ant species operates within ranges of temperature and humidity that can be depicted as a two-dimensional space; in other words, every species has a temperature-humidity envelope. Also, the tolerance of an isolated foraging worker is very different from that of an entire colony. The forager is more swiftly affected by the preexisting ambient temperature and humidity of the microenvironment, whereas the colony can control the microenvironment by shifting into deeper reaches of the nest or by the retention of metabolic heat and moisture through clustering. Thus no nuptial flights of the fire ant Solenopsis invicta occur when morning temperatures of the soil (from the surface to 10 cm deep) are below 18°C, and colony founding by newly inseminated queens occurs only if the soil temperatures at 5 to 7 cm below the surface are equal to or greater than 24°C. In addition, young colonies survive winter poorly even in southern Mississippi (Rhoades and Davies, 1967; Markin et al., 1973, 1974). As a consequence the imported fire ant appears already to have reached its geographic range limit in the northern part of the coastal states. To the west, the colonies are likely to be able to spread across desert areas only by colonizing stream beds, irrigated agricultural fields, and urban areas. Foraging fire ant workers, on the other hand, have even less tolerance for extreme cold, heat, and low relative humidity, and they are unable to survive in ambient conditions in which colonies as a whole do well (Francke and Cokendolpher, 1986). The main point to consider is that if fire ants were solitary insects, the geographic distribution of the species would be far more restricted than it is in fact.

A large portion of the available data on temperature tolerances by foraging workers is summarized in Table 10-1. Not surprisingly, a rough correlation exists between the tolerances and the environment, with desert species at the upper end and some (but not all) species from cold temperate forests at the lower end. The extreme thermophiles of the world fauna belong to the myrmicine genera Ocymyrmex of southern Africa and Cataglyphis of North Africa and Eurasia. The two are biogeographic vicariants of each other. In particular, both are specialized in similar manner for diurnal foraging on extremely hot desert terrain. The workers hunt singly for the corpses of insects and other arthropods that have succumbed from the heat (Harkness and Wehner, 1977; Marsh, 1985a,b; Wehner, 1987). Their behavior and even outward physical appearance are remarkably similar. Ocymyrmex barbiger forages in the sun at surface temperatures up to 67°C, which must be close to the record for insects generally. Cataglyphis fortis is a specialist of the extremely hot, dry, and food-impoverished terrain of the Saharan salt pans. At the opposite end, the cryophilic (cold-loving) species Camponotus vicinus and Prenolepis imparis start foraging at just above freezing and cease when the temperature reaches about 20°C. Similarly, the Australian species Nothomyrmecia macrops forages exclusively after dusk, and the workers seem to be more active at times when temperatures are low (5-10°C). Hölldobler and Taylor (1983) suggested that low temperatures hamper the escape of potential prey encountered in the tree tops increasing the hunting success of the Nothomyrmecia foragers (Plate 12).

Most diurnal desert ants pass through two periods in the day when their preferred temperatures occur, in the morning and late afternoon. As a consequence they have a bimodal distribution of foraging activity, with a period of decline around midday. In mesic temperature habitats, such as northern coniferous woodland, the reverse pattern occurs among diurnal species, with foraging peaking in the middle of the day. These patterns are probably determined by external environmental cues. Hunt (1974) was able to shift the desert bimodal pattern of the Chilean dolichoderine Dorymyrmex antarcticus to a mesic unimodal pattern merely by shading the nest at midday.

While the temperature tolerances of ant species are correlated with climate and major habitat, the relation is only loose. Added variance comes from two sources, microhabitat specialization and competition avoidance. Forest species, such as Aphaenogaster rudis and Paratrechina melanderi, have lower tolerance than other species, such as Monomorium minimum and Tetramorium caespitum, which are adapted to clearings in the same immediate area. It is also of advantage for ants to use unusual activity regimes to escape competitors. The thermophilic species of Cataglyphis and Ocymyrmex have the desert terrain virtually to themselves at midday. In the grassy meadows of Massachusetts, the little black ant Monomorium minimum recruits at higher temperatures than does its three closest competitors, Lasius neoniger, Myrmica americana, and Tetramorium caespitum. It combines this specialization with an effective form of venom dispersal to appropriate a substantial fraction of dead insects and other food finds (Adams and Traniello, 1981).

The subtle differences in microhabitat that characterize species are very well illustrated in Formica perpilosa and Trachymyrmex smithi, two ant species common in the Chihuahuan desert of southern New Mexico. As illustrated in Figure 10-1, Formica perpilosa is active in a much broader area around its nest than is T. smithi. The reason is that the [[Formica'' are more resistant to water loss, and they also tend to forage on mesquite and other forms of vegetation, where the relative humidity is higher (Schumacher and Whitford, 1974).

Quantitative studies of humidity preference by foraging ants of the kind conducted by Schumacher and Whitford are rare in the literature, but abundant anecdotal evidence exists to show that the higher the relative humidity, the greater the temperature tolerance. An increase in foraging activity when humidity rises at high temperatures has been observed in Pheidole militicida (Hölldobler and Möglich, 1980), Formica polyctena (Rosengren, 1977b), Prenolepis imparis (Talbot, 1943b, 1946), and a wide range of semi-desert ant genera in Australia (Briese and Macauley, 1980); and it is probably a general phenomenon in ants as a whole. Another source of differential foraging in the face of varying humidity is simple physiological resistance to desiccation, a phenomenon that has been observed in Tetramorium caespitum by Brian (1965b), Pogonomyrmex by Hansen (1978), and diurnal ant species in Australia by Briese and McCauley (1980).

On the other hand, rain halts most foraging in places where the drops pelt the ground and form small puddles and rivulets (Hodgson, 1955; Lewis et al., 1974b; Skinner, 1980a). Every collector is familiar with this phenomenon, to his frequent frustration and especially during tropical wet seasons. On the other hand, there are a few long-legged ants, including the species of the dolichoderine genus Leptomyrmex of Australia and New Guinea and the myrmicine “giraffe ants” Aphaenogaster phalangium of Central America, that use their unusual stature to navigate water films. As a result they are among the first insects to forage after the rain has ceased.

Daily cycles of activity
Each ant species has a distinctive daily foraging schedule. Some are active for only a few hours each day. To take an extreme example, there is a remarkable degree of precision in the changeover of ant species at dusk in the heath of southwestern Australia (Wilson, 1971). In midafternoon, the ground and the branches and leaves of the low bushes that dominate the vegetation contain hordes of workers, mostly brown, red, or black in color and with medium-sized compound eyes, belonging to ten or so species of Myrmecia, Rhytidoponera, Dacryon, Iridomyrmex, and other typically Australian genera. As dusk falls, first one species, then another, begins to pull back into their nests, while the nocturnal species--pale-colored, mostly large-eyed species in Colobostruma, Iridomyrmex, and Camponotus--make their appearance in a regularly staggered succession. So orderly is the changeover that approximately the same number of foraging workers remains on the bushes throughout.

Are such daily cycles based on circadian rhythms, or are they guided hour-by-hour by changes in temperature and other external stimuli? The answer appears to be that circadian rhythms affecting foraging behavior are widespread, but in many ant species they can be overridden--or at least frame-shifted--by colony hunger or certain environmental changes. McCluskey and Soong (1979), for example, established diel rhythms in four species that occur together in southern California. Pogonomyrmex californicus and Pogonomyrmex rugosus, which are harvesters, commence foraging at early midday and remain very active through the afternoon. Two other harvesters, Veromessor andrei and Veromessor pergandei, as well as Formica pilicornis and Myrmecocystus mimicus, which are general insectivores, have bimodal activity schedules, declining toward midday and picking up again from late afternoon through dusk. When colony fragments were placed in the laboratory at a constant temperature with alternating light and dark, they maintained their natural schedules to a significant degree. A similar result was obtained by Rosengren (1977b) with Finnish colonies of the wood ant Formica polyctena. Workers responded to a 12:12 light/dark cycle in the laboratory by a rise in activity toward the end of the dark period, as though anticipating daybreak. The ants could be entrained to new 12:12 cycles out of phase with that occurring in nature. When kept in complete darkness on a 12:12 warm-cool cycle, the workers grew most active in the middle of the warm period. Thus light alone is not crucial for the entrainment of diel cycles. In later work, Rosengren and Fortelius (1986a) extended this result to a variety of species in the exsecta and rufa groups of Formica. They observed a tendency of workers to increase activity before the artificial “dawn,” but there was also a great deal of variation in the overall activity pattern among colonies. Ambiguity is also the rule in the giant ponerine ant Paraponera clavata (Figure 10-2). McCluskey and Brown (1972) observed that workers on Barro Colorado Island emerged regularly at dusk and apparently foraged until dawn. When transferred to an artificial nest in the laboratory they displayed striking peaks of activity at dawn and dusk. But when placed in a dark room with alternating light and dark and constant temperature, they became diurnal rather than crepuscular and nocturnal. In Costa Rica over a period of two years, we observed considerable variation in the diel pattern of activity by Paraponera workers, although these giant ants appear to be primarily nocturnal. The same phenomenon was observed by Harrison and Breed (1987), who succeeded in training Paraponera workers to come to the sites of sugar baits at fixed times during both the day and night.

In fact, many ant species seem quite capable of shifting the time of peak foraging back and forth as an adjustment to vagaries of the environment. We have seen Atta cephalotes, a dominant leafcutter of the moist lowland forests in Central and South America, change from predominantly diurnal to nocturnal foraging over a period of a few days in the Brazilian Amazonian forest. Lewis et al. (1974a) found that in Trinidad the foraging columns emerge decisively and reach peak traffic within two hours. This event usually occurs at night and lasts for an average of 12 hours; when it unfolds during the day it persists for 7.5 hours on the average. Colonies in the same locality are often unsynchronized, and it seems likely that they are entrained in an independent manner by food discoveries rather than by some more pervasive cue from the physical environment.

This form of adaptive plasticity has in fact been demonstrated in the case of Messor galla and Messor regalis, two granivores inhabiting the savanna of northeastern Ivory Coast (Lévieux, 1979). During the dry season, when seeds are abundant, the colonies forage mostly at night and in tight columns. During the rainy season, or in periods of the dry season with poor seed crops, the ants shift to diurnal foraging and the columns are less organized. It is very common for desert ants to switch from nocturnal foraging in the summer to diurnal foraging in the winter. The phenomenon has been observed in species of Aphaenogaster (Whitford and Ettershank, 1975), Atta (Mintzer, 1979b), Veromessor (Tevis, 1958), Pheidole (Hölldobler and Möglich, 1980), and Pogonomyrmex (Hölldobler, 1976a). Nearly the reverse pattern is displayed by Aphaenogaster rudis and Paratrechina melanderi in the hardwood forests of Maryland: peak activity is reached around midnight during the summer months, shifting to more nearly uniform round-the-clock activity during the spring and summer and quiescence during the winter (Lynch et al., 1980). Authors who have reported these patterns generally consider them to be adaptations to promote thermoregulation. Where rodents can remain strictly nocturnal year-round, at least in deserts, ants are more sensitive to ambient temperature changes and must alter their daily activity accordingly (Davidson, 1977b; Brown et al., 1979a,b).

Differences in foraging rhythms among sympatric species of ants can serve in temporal partitioning of significant resources. Such activity differences might be based proximately on different humidity and temperature ranges tolerated by the species, yet be an ultimate, evolutionary result of interspecific competition. Although there exist only a few studies which directly address this question, some of the observations reported by Talbot (1946), Greenslade (1972), Bernstein (1979b), Swain (1977), Klotz (1984), and other strongly suggest temporal partitioning of resources. Klotz reports that in his study area in Kansas both Camponotus pennsylvanicus and Formica subsericea utilize the same aphid honeydew sources, but in a temporally displaced manner. Camponotus is more active at cool temperatures and predominantly nocturnal, while Formica is active at higher temperatures and is primarily diurnal. We made similar observations on Camponotus socius and Camponotus floridanus in central Florida. Where both species coexisted and used the same honeydew sources, socius was primarily diurnal and floridanus primarily nocturnal.

The approximately 8,800 known ant species use a dazzling variety of procedures for the discovery and retrieval of food. Several authors have attempted classifications of this important category of behavior, including Oster and Wilson (1978), Passera (1984), and Moffett (1987b). We will provide a synthesis here of what we regard as the most useful elements of these schemes. The resulting classification breaks all of the phenomena into three categories: hunting (3 kinds), retrieving (4 kinds), and defense (4 kinds). The elements in each can be combined to form 3 x 4 x 4 = 48 possible three-state foraging techniques. This arrangement provides an informal and convenient framework for the description of the behavior of individual species.

Hunting: (1) by solitary workers; or (2) by solitary workers directed to specific trophophoric fields by trunk trails; or (3) by groups of workers searching in concert, in the manner of army ants.

Retrieving: (1) by solitary workers who return home on their own; or (2) by individual workers who return home along persistent trunk trails; or (3) by individual workers recruited to the food sites by scouts; or (4) by groups of workers who gang-carry the food items.

Defense: defense by guard workers during hunting, or an absence of such defense; and either defense by guard workers during harvesting and retrieval of food, or the absence of such defense (altogether, 4 combinations are possible).

Although no complete accounting has been attempted across all of the ants, it is our impression that virtually all of the 48 possibilities in this broad space of foraging techniques is employed by one species or another. At one extreme are the completely solitary foragers, including scavengers and specialized predators and in all subfamilies other than the army ants of the subfamilies Dorylinae and Ecitoninae. The workers hunt singly and retrieve food items entirely on their own. At the other extreme are the army and driver ants of the genera Eciton and Dorylus, which hunt in groups, often gang-retrieve large prey items, and deploy specialized guard workers along the perimeters of the foraging columns.

The total foraging strategy of an ant species often comprises two or more three-state techniques, each chosen according to the quality and nature of the food of the moment. For example, when the solitary foragers of Pogonomyrmex encounter a seed or dead insect small enough to be carried singly, they carry it home alone. When the object is too large to be transported by one ant, or there is a cluster of small food items, the worker lays a recruitment trail with secretions from its poison gland. When the food is persistent at a particular site, as is the case for a continuing seedfall, the ants deposit trunk trails with secretions that come at least in part from the Dufour's gland (Hölldobler and Wilson, 1970; Hölldobler, 1971b; Hölldobler, 1976a). The minor workers of the Amazonian myrmicine Pheidole embolopyx also search in solitary fashion and single-load the small insects they encounter. When the food item is too large to move, the scouts recruit other minor and major workers from the nest. The majors spend most of their time carving up the insect and defending it from intruders, while the minors concentrate on drinking the hemolymph and carrying fragments back to the nest (Wilson and Hölldobler, 1985). Minor workers of very small Pheidole dentata colonies spend more time feeding away from the nest and transferring the food to nestmates within the nest, as well as less time recruiting, than do minor workers in larger colonies. Evidently the effort of a single worker can more nearly satisfy the requirements of the small colony (Burkhardt, 1983).

Recruitment thus permits a multiple strategy of foraging and permits the exploitation of a wider range of food items. When workers of Monomorium, Myrmica, Formica, and Lasius forage singly, prey size is positively correlated with worker size. But when cooperative foraging based on recruitment is used, the correspondence breaks down. The ants are able to handle items far larger than their individual size (Traniello, 1987a). Whereas the workers in Traniello's study varied in fresh weight from 0.1 to 6.8 mg, their arthropod prey varied from 1 mg to over 2 grams, a range of 2000 times.

Solitary foragers
The approximately 8,800 known ant species use a dazzling variety of procedures for the discovery and retrieval of food. Several authors have attempted classifications of this important category of behavior, including Oster and Wilson (1978), Passera (1984), and Moffett (1987b). We will provide a synthesis here of what we regard as the most useful elements of these schemes. The resulting classification breaks all of the phenomena into three categories: hunting (3 kinds), retrieving (4 kinds), and defense (4 kinds). The elements in each can be combined to form 3 x 4 x 4 = 48 possible three-state foraging techniques. This arrangement provides an informal and convenient framework for the description of the behavior of individual species.

Hunting: (1) by solitary workers; or (2) by solitary workers directed to specific trophophoric fields by trunk trails; or (3) by groups of workers searching in concert, in the manner of army ants.

Retrieving: (1) by solitary workers who return home on their own; or (2) by individual workers who return home along persistent trunk trails; or (3) by individual workers recruited to the food sites by scouts; or (4) by groups of workers who gang-carry the food items.

Defense: defense by guard workers during hunting, or an absence of such defense; and either defense by guard workers during harvesting and retrieval of food, or the absence of such defense (altogether, 4 combinations are possible).

Although no complete accounting has been attempted across all of the ants, it is our impression that virtually all of the 48 possibilities in this broad space of foraging techniques is employed by one species or another. At one extreme are the completely solitary foragers, including scavengers and specialized predators and in all subfamilies other than the army ants of the subfamilies Dorylinae and Ecitoninae. The workers hunt singly and retrieve food items entirely on their own. At the other extreme are the army and driver ants of the genera Eciton and Dorylus, which hunt in groups, often gang-retrieve large prey items, and deploy specialized guard workers along the perimeters of the foraging columns.

The total foraging strategy of an ant species often comprises two or more three-state techniques, each chosen according to the quality and nature of the food of the moment. For example, when the solitary foragers of Pogonomyrmex encounter a seed or dead insect small enough to be carried singly, they carry it home alone. When the object is too large to be transported by one ant, or there is a cluster of small food items, the worker lays a recruitment trail with secretions from its poison gland. When the food is persistent at a particular site, as is the case for a continuing seedfall, the ants deposit trunk trails with secretions that come at least in part from the Dufour's gland (Hölldobler and Wilson, 1970; Hölldobler, 1971b; Hölldobler, 1976a). The minor workers of the Amazonian myrmicine Pheidole embolopyx also search in solitary fashion and single-load the small insects they encounter. When the food item is too large to move, the scouts recruit other minor and major workers from the nest. The majors spend most of their time carving up the insect and defending it from intruders, while the minors concentrate on drinking the hemolymph and carrying fragments back to the nest (Wilson and Hölldobler, 1985). Minor workers of very small Pheidole dentata colonies spend more time feeding away from the nest and transferring the food to nestmates within the nest, as well as less time recruiting, than do minor workers in larger colonies. Evidently the effort of a single worker can more nearly satisfy the requirements of the small colony (Burkhardt, 1983).

Recruitment thus permits a multiple strategy of foraging and permits the exploitation of a wider range of food items. When workers of Monomorium, Myrmica, Formica, and Lasius forage singly, prey size is positively correlated with worker size. But when cooperative foraging based on recruitment is used, the correspondence breaks down. The ants are able to handle items far larger than their individual size (Traniello, 1987a). Whereas the workers in Traniello's study varied in fresh weight from 0.1 to 6.8 mg, their arthropod prey varied from 1 mg to over 2 grams, a range of 2000 times.

Foraging strategies
The extreme category of solitary hunting combined with solitary retrieval is illustrated by the desert scavengers of the genera Cataglyphis and Ocymyrmex (Harkness and Wehner, 1977; Wehner et al., 1983; Schmid-Hempel, 1984, 1987; Schmid-Hempel and Schmid-Hempel, 1984; Wehner, 1987). The total foraging pattern of Cataglyphis bicolor is described in the following striking manner by Rüdiger Wehner in his 1987 review:

When the daily foraging activity of a desert ant's colony is compressed in both space and time by assuming a bird's eye point of view and using a quick-motion camera, an impressive picture emerges: a constant stream of particles radiates out from a centre and spreads evenly over a roughly circular area. Within this field of flow each (individually labelled) particle follows its own radial path, lingers at some distance from the start, and returns to the centre. Some time later, the same particle reappears, follows a similar path, disappears again in the centre, reappears, and continues to do so until at the end of the day the whole stream of particles ceases to flow. Now assume that the time window is expanded from one to several days and the time machine speeded up. One would then become aware of particles disappearing from the scene forever after having performed a number of moves, and new particles taking their place. It is as though one were observing a gigantic slime mold in which cells continually moved in and out from an aggregation centre, but in which the inward and outward movements of the cells were largely desynchronized. Scaling up the scene by about five orders of magnitude and returning to real time, the central-place foraging behaviour of desert ants unfolds in front of one's eyes, and a number of questions about the spatial and temporal organization of this behaviour immediately spring to mind.

The workers of Cataglyphis bicolor (Figure 10-3) make about five to ten forays each day, and an average of 290 foragers of a colony conduct 1500-3000 such expeditions. The entire effort is made during daytime. The ants retreat entirely into the nest at night and often close the entrance with soil. As morning approaches they dig out the entrance again. At sunrise, a worker comes out slowly and stands about for a minute or so within a few centimeters of the entrance. As Harkness and Wehner describe it, the first foragers then begin their runs:

Then one or two more come out and will set off away from the nest in more or less a straight line in some direction, which varies all round the compass with different individuals. Each individual, however, keeps to a constant foraging direction, that is maintained at least over a period of weeks. The frequency of ants leaving is low at first but eventually rises to 100-200 per hour (according to the size of the nest) later in the morning. In half-an-hour or more the first ants begin to come back, usually carrying the corpse or part of the corpse of a dead insect. The arrival of an ant with a burden seems often to be followed by the exit of a group of half-a-dozen or more ants. . . . After an ant returns from a foray, commonly it goes out again in a matter of minutes. By midday there is a constant traffic in and out. Although the frequency of exits and entries stays high for several hours, in general the activity of an individual ant is restricted to a limited period of time that is the same every day and lasts 2-3 hours. Towards sunset the traffic in and out of the nest falls away.

The solitary foragers are guided by sun-compass orientation and the location of bushes and other landmarks, with the latter being remembered as successive frames of configuration (Wehner, 1982, 1987). The foragers tend to persist in only one or a very few directions for their lifetime, if for no other reason than that travel outside the nest is very dangerous and life is short. Most of the workers are soon picked off by spiders and robber-flies, in spite of their ability to run at great speed, up to a meter per second. In southern Tunisia they enjoy a half-life time of only 4.2 days and a life expectancy of 6.1 days (Schmid-Hempel and Schmid-Hempel, 1984). Yet the system is so efficient that the average forager retrieves a food weight during its lifetime 15 to 20 times greater than its own body weight (Wehner et al., 1983).

When occasionally changing direction, both Cataglyphis and Ocymyrmex appear to follow a simple rule of thumb: continue to forage in the direction of the preceding foraging trip whenever the trip proved successful. Otherwise abandon that direction and select a new one at random, but decrease the probability of doing so as the number of previously successful runs increases. Following these rules, Wehner (1987) designed a model that closely described the directional choices actually made by individual foragers. The result was in accord with the observation that colonies exhibiting low foraging efficiencies simultaneously develop low sector fidelities. The overall effect of this method of foraging is that each colony blankets the area around the nest entrance so evenly that maps of foraging activity do not provide a clue to the location of the nest entrance (see Figures 10-4 and 10-5).

As Schmid-Hempel (1984) has pointed out, the rule of thumb used by Cataglyphis and other desert ants creates a kind of polymorphism in the population of foragers. When some of the ants encounter aggregated items, so that they are repeatedly rewarded, they concentrate their efforts. Other workers enter terrain with more scattered resources and disperse their efforts to a corresponding degree. In the ensuing total pattern, the entire foraging domain of the colony is covered efficiently, with a temporary focus on the most productive sites.

The efficiency of deployment is enhanced by “noise” in the repetitive journeys. That is, unpredictable deviations regularly occur away from paths taken on previous sallies. Such errors increase the chance that the workers will strike food items missed in earlier efforts. The same explanation applies to the scatter of fire ants around odor trails, a variance that sometimes allows the workers to locate moving prey (Wilson, 1962b, 1971).

Jacques Pasteels, Jean-Louis Deneubourg, and their co-workers have taken the next step of reasoning and developed a sophisticated theory of error adjustment to optimize net energetic yield during recruitment (Pasteels et al., 1982; Deneubourg et al., 1983; Verhaeghe and Deneubourg, 1983; Champagne et al., 1984; Pasteels et al., 1987). Each species concentrates on food items with a certain pattern of dispersion and size. When the items are extremely clumped and large, say as in the case of a termite colony, the recruitment needs to be very precise, as it is in the case of termitophages Neoponera laevigata and Megaponera foetans. When food is less aggregated and more easily handled by single workers, the error in recruitment is typically larger. Such seems to be the case, for example, in the Messor and Pogonomyrmex harvester ants. Furthermore, the error level may be adjusted within the repertories of individual species in ways that accord with the Pasteels-Deneubourg models. Food items that are small or of poor quality elicit less recruitment pheromone and tactile signaling and more error in the response, leading to a wider exploration of the environment--as in the single-forager strategy of the desert ants (Wilson, 1962b; Hangartner, 1969b; Hölldobler, 1976a; Verhaeghe, 1982; Crawford and Rissing, 1983). The same is true of food sources that are distant from the nest (Wilson, 1962a; Ayre, 1969; Hölldobler, 1976a; Taylor, 1977).

Ortstreue and majoring
Workers belonging to the same colony display a striking degree of specialization in their foraging activity. A growing body of research has revealed the widespread occurrence of individual persistence in foraging site, the phenomenon called Ortstreue. We have seen how Cataglyphis workers return repeatedly to the same approximate area for as long as they are rewarded with food. Similar learning and particularization of behavior has been documented in the ponerine ants Diacamma rugosum, and Neoponera apicalis, both of which are solitary hunting ants (Uezu, 1977; Fresneau, 1985); leafcutter ants of the genus Atta (Lewis et al., 1974a); harvester ants of the genera Chelaner, Messor, Pheidole, and Pogonomyrmex (Hölldobler, 1976a; Onoyama, 1982; Onoyama and Abe, 1982; Hölldobler and Möglich, 1980; Davison, 1982); the carton-building “shining black ant” Lasius fuliginosus of Europe (Dobrzanska, 1966; Hennaut-Riche et al., 1979); the North American species Lasius neoniger (Traniello and Levings, 1986); and ants of the Formica rufa group (Økland, 1930; Jander, 1957; Rosengren, 1971, 1977a; Herbers, 1977; Rosengren and Fortelius, 1986b).

Experiments on Cataglyphis, Formica, and Pogonomyrmex suggest that the memories of location in site fidelity are primarily visual in nature. In extreme cases the learning persists for weeks or months. Rosengren and Fortelius were able to show that in Formica aquilonia at least, olfactory orientation is used to the same purpose in the dark. Specifically, the workers appear to follow scent marks used in home range marking. Thus the Formica evidently depend on a hierarchy of cues of the kind demonstrated in homing Pogonomyrmex workers by Hölldobler (1971b, 1976a). A cue in one sensory modality (such as vision) is used predominantly until it fails, then a cue in a second modality (such as olfaction) is employed, and so on. Pheidole militicida, which forages extensively during the night, also employs primarily olfactory cues in home range orientation and Ortstreue (Hölldobler and Möglich, 1980).

It is reasonable to expect that memory and fidelity to particular sites improve, through natural selection, as food sources become richer and more persistent. Such appears to have been the case in herds of honeydew-producing aphids and other homopterans (Pasteels et al., 1987; Goss et al., 1988). When mistakes are made, so that the ants arrive at the wrong tree, the result can be a substantial loss of time and energy. Just such an unproductive leakage has been documented in Formica lugubris by Sudd (1983), who found that a small percentage of workers wander from a species of pine supporting large populations of aphids to another species with few or no aphids.

Another way to divide the environment among foragers belonging to the same colony is by majoring, or specialization on different kinds of food. It is well known that individual honeybees and bumblebees persist in visiting one species of flower for days at a time, even when the favored blossoms are intermingled with those of other flower species (Heinrich, 1979; Seeley, 1985). A similar phenomenon occurs in ants, although it has been documented less extensively. In field experiments by Rissing (1981a), for example, workers of Veromessor pergandei and Pogonomyrmex rugosus continued to harvest one species of grass seed mixed with two other species, even when larger seeds in the medley were available and being collected by nestmates. The specialization sometimes lasted for several days but occasionally shifted rapidly. The choices made were not just a function of body size, with larger ants selecting larger seeds. This variable accounted for only 4 percent of the variance in seed size. It is not known how the initial selections are made and why preferences change, but the process is far from a random first choice: Messor and Pogonomyrmex workers examine as many as 60 seeds before carrying one home (Hölldobler, 1976a; Rissing, 1981a).

Central place foraging
When a forager is bound to a nest or sleeping site, it faces a different set of problems in harvesting energy than if it merely rested at intervals while conducting a search. How these problems can be solved so as to maximize the net energy yield is the subject of the special set of models that make up “central place foraging theory” (Orians and Pearson, 1979; Schoener, 1979; Stephens and Krebs, 1986). The basic reasoning must be modified to accommodate the social insects, which commit expendable “energy packets” in the form of nonreproductive workers (Oster and Wilson, 1978; Seeley, 1985). The number of workers and hence the energy costs can be fitted through mass communication to the spatial distribution of food items and thus the yield of energy moment by moment. Also, the workers can individually specialize on particular sectors of the terrain and on different kinds of food items, allowing a nearly simultaneous coverage of a wide area.

Most of the predictions made from the foraging models are intuitively clear and possess a double heuristic value in the study of ant ecology. First, they allow a test of the basic proposition that natural selection shapes behavior and, in the case of social insects, does so by acting at the level of colonies. Second, by framing research in terms of these models, the ecologist asks questions and searches for predicted phenomena that might otherwise be easily missed.

Let us now consider some of the predictions and see how they have fared so far in empirical studies. The key overall assumption is that the greater the energy expenditure to get to a patch of food items, the more choosy the animal should be in selecting the item out of the patch to carry home. In other words, the more energy you spend to get there, the larger the energy package you try to bring home. This general conception leads to a series of explicit predictions, as follows:

The greater the distance the ant travels from the nest to a food patch, the longer it should take to select the food item. This prediction was confirmed by Schmid-Hempel (1984) for workers of Cataglyphis bicolor, which spend more time searching around a productive site (where dead insects were found) when the site is farther from the nest. It was also confirmed by the finding of Rissing and Pollock (1984) that workers of Veromessor pergandei take longer to search through more distant patches of seeds once they have arrived on the site. It is not supported by data of Shepherd (1982) on the leafcutter Atta colombica, the workers of which fail to increase their search time on leaf baits when they have traveled farther to get there. However, the Atta data are ambiguous because the large variance in individual times obscured possible real differences in a one-way analysis of variance conducted by Shepherd. For ants generally, then, this test of the basic proposition of energy conservation appears to be supported by the available data, but more experimental studies with different species are needed to make the test convincing.

The greater the distance the ant travels to a food patch, the more selective the ant should be in choosing a food item. The expected relation was obtained by Davidson (1979), who found that the farther from the nest Pogonomyrmex rugosus foragers journeyed, the narrower the range of barley seeds they accepted in a patch, with smaller seeds being consistently more ignored. In similar experiments on Pogonomyrmex occidentalis, Pollock (1977) detected the same trend, but his data were not strong enough to be statistically significant. (Davidson points out that Taylor's tests were conducted over shorter distances than her own.) Rissing and Pollock (1984) discovered that while Veromessor pergandei workers generally take larger seeds from patches, the size of the seed is not correlated with the time spent selecting it and hence the time spent in traveling to the patch. Possible nutritive cues other than size could not be discounted. Overall, the available data appear to favor the energy model, but more studies are clearly needed.

The higher the temperature, the more selective the forager should be in food patches. To our knowledge no experiment has been performed to test this expected correlation. However, at varying temperatures Traniello et al. (1984) have examined the behavior of Formica schaufussi in the pursuit of scattered prey, as opposed to variable prey aggregated in patches. As the temperature rose, the ants became less selective, accepting smaller and less profitable prey. Ants, like other insects, increase their metabolism and hence pay higher energy costs at higher temperatures (Nielsen, 1986). We would therefore expect that if the ants encounter food items in groups, so that they are able to pick and choose, the ants will become more selective at higher temperatures. But if they encounter prey items singly, they will be inclined to be less selective, because “a bird in hand is worth two in the bush.” This is the result obtained by Traniello and his co-workers.

Energy maximizers versus time minimizers
Schoener (1971) made an important distinction between animals that maximize their net energy yield and those that have a fixed quota of energy required for body maintenance and reproduction. The first class, the energy maximizers increase their fitness by reducing the amount of time required to harvest the needed quota of energy, leaving them with time to hide from predators, build nests, search for mates, and other tasks extraneous to energy. Probably only bacteria and other very simple organisms are pure energy maximizers, but even within the ranks of the more complex animals some species appear to follow an energy maximization program more closely than others. A model of conversion of energy to genetic fitness in social insects has been offered by Oster and Wilson (1978). Their key result can be stated as follows: so long as the production of new queens remains a linear function of the amount of energy harvested, ant colonies can be expected to maximize their net energy yield. This inference, along with the thrust of information from natural history, indicates that ant foraging procedures are based predominantly on energy maximization. Indeed, the celebrated industriousness of ant colonies stems largely from their seemingly incessant foraging activity and rearing of young. And there is a direct connection to genetic fitness. Numerous studies have shown (see Chapter 3) that the larger the colony, the more likely it is to extirpate neighboring colonies and to extend its foraging range. Consequently, the larger the number of queens it produces.

Nevertheless, the ant colony is far from a mere growth machine. Prudence would seem to dictate that the colony commit only a small fraction of the worker force at any one time to foraging. Workers expend up to seven times more energy while running than while resting (Nielsen et al., 1982; Lighton et al., 1987), so that a point of diminishing return in energy harvesting is quickly reached. To this can be added the high construction costs of replacing lost foragers, along with the need to have a substantial force available to defend the colony against attack at any moment. The trade-offs among those countervailing selection forces remain to be quantified in both theory and experiment.

Special stratagems that improve harvesting
One of the most distinctive devices evolved by ants in foraging is group retrieval, in which two or more workers cooperate to bring home a food object too large to be managed by a single individual. The behavior occurs in elaborate form in the Asiatic marauder ant Pheidologeton diversus (Moffett, 1987b). When individual workers carry prey or other food items, they lift them from the ground in their mandibles and hold them forward while walking back toward the nest. Group-retrieving ants carry burdens differently. They lay one or both forelegs on the burden, an action that appears to assist in lifting it. The ants also open their mandibles and press them against the object but seldom attempt to grip it in the manner of solitary Pheidologeton foragers carrying smaller burdens. Workers utilize different movement patterns corresponding to their positions around the perimeter of the object and the direction of transport. Those at the forward margin walk backward, pulling the burden, while those along the trailing margin walk forward and push it. The ants along the sides of the burden shuffle their legs more or less laterally in the direction of transport (Figure 10-6).

This group-retrieval technique permits the Pheidologeton to carry food at many times the weight per worker than is possible through retrieval by solitary workers. One of the heaviest burdens recorded by Moffett was an earthworm 10 cm long with a dry weight of 0.55 grams, or over 5000 times as much as a single minor worker. The worm was borne by about 100 ants, which transported it at about 0.4 cm per second on level ground. Comparisons with transport of smaller objects showed that the workers carrying the worm handled at least ten times more weight per ant than did solitary ants carrying a summed equivalent in small loads, with a loss of only a little more than half the velocity. In another extreme case, a large piece of cereal was carried by 14 ants. If the ants had first gnawed the chunk into pieces just small enough to be handled by single ants, at least 60 workers would have been needed to retrieve the same amount of food. In other words, group transport reduced the required worker force by seventy-five percent.

Group transport has evolved many times independently among most of the subfamilies of ants (see Table 10-2). It is limited to species that practice recruitment to food sources. Among various phyletic groups it is developed to variable degrees of skill, with the greatest sophistication and efficiency occurring in Pheidologeton, Aphaenogaster, Eciton, and Oecophylla (Plate 13). At least three genera, Camponotus, Leptogenys, and Pheidole, contain some species that utilize it conspicuously and others that do not practice it at all. Franks (1986) has shown that workers of the swarm-raiding army ant Eciton burchellii actually form teams to carry large burdens, with medias specialized for transport getting the process started. Medias also frequently join the transport gangs of Pheidologeton diversus minor workers, but their participation is much less important than is true of Eciton. Finally, the species of Pheidologeton, Eciton, and Neivamyrmex use groups to transport food, immature stages, and (in the case of Pheidologeton) large pieces of nest material and refuse.

A second highly evolved technique that enhances productivity, namely the frequent rotation of foraging direction, has been discovered in two species of ants that possess large colonies and send out columns or swarms of foragers. During the nomadic phase of the Eciton burchellii 35-day cycle, when colonies emigrate to a new site following the daily raid, the ants run little risk of depleting the arthropod populations on which they depend. However, during the statary phase, when the colonies remain in the same site for an extended period of time, the arthropod populations are decimated and require as much as a week to recover. Other ant and wasp species, which are favored targets, take much longer. Franks and Fletcher (1983) showed that the Eciton burchellii solve the problem by rotating their central swarm direction around the central bivouac site. Each day they shift an average of 126°, which is significantly greater than the 90° that would be expected if raids were oriented at random. Thus the ants increase their harvesting efficiency by a design similar to the spiral leaf arrangement used by many plant species, to minimize self-shading. The raiding fronts are in fact analogous to very long, thin leaves, in which a spiral phytotamy can be less than exact and yet still avoid overlap.

A similar foraging rotation has been claimed in the desert ant Veromessor pergandei, which sends out tens of thousands of foragers in lengthy columns to collect seeds. Went et al. (1972) reported the average change between consecutive foraging periods to be 15°, while Ruth Bernstein (in Carroll and Janzen, 1973) gave it as about 20°. However, Rissing and Wheeler (1976) found the shifts to be very irregular in magnitude and direction, sometimes proceeding consistently clockwise, sometimes counterclockwise, and sometimes reversing to head in the opposite direction. Whereas the Eciton “phylletaxy” is relatively precise and endogenous, the Messor pattern is evidently more dependent on the day-to-day vagaries of seedfall.

Yet another method that improves foraging efficiency is to deploy workers in outposts, shortening the length of their sallies and allowing them to store food temporarily at way stations or even to feed larvae carried there. The behavior is displayed by the termite-hunting workers of the Malaysian Eurhopalothrix heliscata, which tend to settle in groups and rest for prolonged periods of time away from the brood chambers. The pattern enhances the effectiveness of their predation. Eurhopalothrix, like other members of the tribe Basicerotini, do not rely on mass raiding to overcome termites as do some species of Leptogenys, Megaponera, and Pachycondyla. They are able to recruit nestmates, but the communication is relatively short-range and imprecise. It therefore seems to be of advantage for colonies with these traits to possess staging areas from which forays against termite colonies can be conducted over relatively short distances whenever workers discover foraging columns of the insects or holes in the termite nests. The prey items can then be relayed at a more leisurely pace back to the brood chambers (Wilson and Brown, 1984). A similar system of outposts is practiced by the strange little myrmicine Proatta butteli of tropical Asia. The workers prey on termites, isopods, and other arthropods, some of which are quite large and can be subdued only by rapidly recruited workers (Moffett, 1986d).

The outpost phenomenon probably accounts for many cases of polydomy, defined as the dispersal of the colony into multiple nest sites. Polydomy may occur in colonies with either single or multiple queens. Traniello and Levings (1986) found that in Lasius neoniger, a monogynous formicine that nests in the soil in fields and other open habitats, polydomy follows an annual cycle and is close to optimal for foraging efficiency. During early and middle summer, workers construct increasing numbers of little crater nests, expanding the foraging territory of the colony as a whole. The nests at the Massachusetts study site were more uniformly distributed than predicted from chance alone. Workers emerging from a given nest entrance were most successful at retrieving prey items over distances of less than 20 centimeters. The average distance between nest entrances was 38 centimeters, in other words about twice the most effective foraging range, as expected. In late summer, when the season grew to a close, the number of nests declined and the colony contracted toward a central core.

Population regulation
The population of workers in an ant colony grows in a roughly sigmoidal (S-shaped) pattern. It accelerates at first, then decelerates as some of its resources are diverted to the production of the virgin queens and males. During the dry season or winter, during which local conditions are harshest, the worker population declines still further, because mortality is not adequately offset by larval growth and new worker production. The reproductive forms leave the nest following the growth season, so that the workers are free to increase their own numbers during the first part of the ensuing growth season. As a consequence the population of workers in a mature colony fluctuates annually around a mean. It roughly approximates a stable limit cycle (Brian 1965b, 1983; Wilson, 1971).

Superimposed on the growth of individual colonies is the increase in the population of colonies. The standing density of colonies is determined by the rate at which queens start new colonies, balanced by their mortality rate and that of the colonies they found. At the next lower level, the density of workers belonging to a given species in a particular area is the summed product of the number of colonies in various growth stages and the workers present in each growth stage.

It is convenient at this stage of our knowledge of ant ecology to make a rough distinction between density-independent effects and density-dependent effects in the determination of the local numbers of colonies and workers. Independent effects are those that alter birth and death rates in a manner not related to the density of the colonies or workers. Some features of the physical environment, for example, limit numbers by restricting the replacement rate of the colonies and workers regardless of the density attained in the growth curves. Such is the case for flooding and severe cold waves. In contrast, density-dependent effects either lower the birth rate or raise the death rate as population density increases. In time, the breaking effect slows the growth to zero. Depending on the initial conditions and the lag time in which the density-dependent effects take place, the population either stabilizes at a more or less constant level (the “carrying capacity” of that particular environment), or enters a stable limit cycle around a constant level, or enters a chaotic regime in which the density fluctuates in an unpredictable manner. The totality of these effects constitute the population regulation.

Put briefly, the existing evidence implicates territorial aggression as an important and possibly premier mode of population regulation in ants. Other factors, including climate, predation, food, and availability of nest sites, also clearly affect the density of ant colonies, but so far they have been only sparsely documented. No experiments have been performed that permit an evaluation of all of the controls together. However, it is possible to describe the effects of separate factors separately, at least in qualitative terms.

Starting then with density-independent factors, the role of temperature is most clear-cut in the biology of the ants of cold temperate zones. In northern Europe the distribution of Myrmica species, for example, is highly correlated with the number of hours of bright sunshine per day (Baroni Urbani and Collingwood, 1977). Barrett (1979) found that portions of England lacking Myrmica sabuleti average only 4.6 hours of bright sunshine per day, those with sparse records of the species have 5.2 hours per day, and those with abundant records average 5.8 hours per day. In general Myrmica colonies of various species receiving the most solar heat acquire the largest worker populations, and those with the most workers produce a disproportionate share of the virgin queens and males (Brian and Brian, 1951; Brian and Elmes, 1974; Elmes and Wardlaw, 1982). Elmes and Wardlaw fitted the known correlates to a model in which a chain of annual events determine the production of these reproductive forms; the key factors they intuited include the number and size of larvae held over winter and the amount of solar heat acquired in the following spring.

In milder portions of the temperate zones, population densities are more likely to be reduced not only by cold stress in the winter but also by heat and drought stress in the summer. In fact, studies of the meat ant Iridomyrmex purpureus in Australia across terrain of variable drainage showed that summer stress affects the growth of worker populations within individual colonies, while winter stress affects the survival of colonies as a whole (Greenslade, 1975a,b). A similar bracketing control through the annual cycle has limited the spread of the imported fire ant Solenopsis invicta in the southern United States (Francke and Cokendolpher, 1986).

For some kinds of ants a limitation in the number of suitable nest sites is important in population control. This is especially true in regions with marginal climatic conditions, including northern Europe where exposure to the sun is a critical limiting factor (Brian, 1952b, 1956c; Sudd et al., 1977). It also appears to be the case in the lowland rain forests of New Guinea, where most of the ant species are specialized for living in rotting tree branches and other small fragments of wood lying on the ground, and most such nest sites are fully occupied by colonies of either ants or termites (Wilson, 1959a,c). Hence, the rotting-wood specialists appear to have largely filled the space available to them. A majority of species studied by Room (1971) in a cocoa farm in Ghana occupy the same kind of nest sites. In Sri Lankan forests, a similar near-saturation occurs, but with termites preempting a larger percentage of the nest sites. Daceton armigerum, a large predatory ant of South America, is specialized for life in preformed cavities in the canopy of moist tropical forests. The relative scarcity of such retreats appears to be an important factor in determining the density of Daceton colonies, but this hypothesis has not been subject to verification due to the relative inaccessibility of the habitat (Wilson, 1962a). A similar specialization appears to restrict the European “shining black ant” Lasius fuliginosus, which nests preferentially in large preformed cavities in tree trunks (Maschwitz and Hölldobler, 1970).

On the other hand, persuasive experimental confirmation has been provided by Herbers (1986b) in the case of Leptothorax longispinosus, an ant that favors preformed cavities in small pieces of rotting wood on the floor of North American deciduous forests. When Herbers added additional nest sites in the form of hollow dowels, the density of colony fragments increased, and the total worker population appeared to grow as well. In an independent study, Brian (1956c) has provided evidence that in the west of Scotland the density of colonies is controlled by the density of available nest sites. In England as a whole, which has only a marginal environment for ants, “nest sites need to have high insolation, to be moist but not too wet, to be soft enough for excavation but mechanically stable, and they need to be within reach of plants to supply sugar and water through Homoptera and protein through the many small insects that feed on the plants and their litter.” The constraining condition of having to nest in a place with few plants and yet still be near places with many plants is offset by the extreme abundance of insect food during the spring and summer. In Brian's study area in western Scotland, much of the land is covered by higher vegetation of one form or another. Consequently nest sites are in short supply and the ant colonies compete heavily for them. When Brian laid out slabs of rocks, which make ideal nest sites, in an open, food-rich area populated by Myrmica ruginodis, the number of colonies increased. In a parallel study, he found that the gradual growth of trees in a glade, which increased shading and reduced the number of warm, dry nest sites on the ground, resulted in a reduction in the number of ant colonies. The same general conclusion was independently drawn by Gösswald (1951b) on the basis of his long-term studies of the same ant genera in Germany. He pointed, for example, to the curious fact that dense ant populations characteristically occur around quarries, where fragments of stones laying on the ground provide an unusual number of nest sites. Yet in spite of such evidence, nest-site limitation is probably far from universal. For example, it appears not to be a significant factor in Puerto Rican forests (Torres, 1984a,b), and it is obviously of limited importance for the soil dwelling ants of deserts and grasslands.

The food supply of colonies has been implicated in colony regulation in at least two studies. According to Davison (1982), colonies of Chelaner in arid New South Wales are sensitive to fluctuations in the supply of seeds on which they depend. In times of scarcity, the number of larvae declines. Because the adult workers depend on larval secretions for food (the larvae consume and metabolize the raw seeds for the colony), the worker population also declines. By excluding rodents that compete with seed-eating ants and counting the seeds available in the soil, Brown et al. (1979a) obtained evidence that the colony density of the granivorous ants in Arizona is limited by food.

As noted earlier, swarm-raiding army ants (Eciton burchellii) devastate the arthropod fauna on the ground and low vegetation over which they conduct their daily forays. On Barro Colorado Island in Panama, approximately 50 colonies harvest the island in a relatively efficient manner, seldom if ever colliding with one another (Franks and Bossert, 1983). Nearby Orchid Island has only one-eightieth the area of Barro Colorado, and for many years no Eciton burchellii occurred there. Since both islands were created by the rise of Gatun Lake, Franks and Fletcher (1983) concluded that the army ants had previously occurred on Orchid Island but disappeared due to an inadequate food supply. Put another way, Orchid Island has less area than the average portion of land shared by each Eciton colony on Barro Colorado Island.

In a reverse direction, predation can also depress colony and worker density of the ant species taken as prey. Franks and Fletcher found that the ant colonies on Orchid Island were denser and more mature than those on Barro Colorado. When these investigators introduced an Eciton burchellii colony onto the small island, it preyed much more heavily on ants during the first few weeks than did the Eciton colonies remaining on Barro Colorado and as a result reduced their densities substantially.

Offense and defense by foragers
The ant species studied possess an enormous variety of offensive and defensive techniques. They also employ them in almost every imaginable circumstance in the search for food and protection of the nests. The existing information on this very eclectic category of behavior is briefly summarized in Table 10-3.

Hunt (1983) correctly noted that predators, especially visually searching vertebrates, play a key role in the shaping of foraging behavior, perhaps rivaling the maximization of net energetic yield as a natural selection factor. Separate species have “chosen” whether to commit only a few foragers or a large force, whether to hunt stealthily or in the open, whether to recruit soldiers to the food site or not, and so forth. All of these options are important in the evolution of foraging strategies. In spite of their potential significance, studies of the phenomenon have been curiously incomplete to the present time. A great deal of information has been acquired on the vertebrates, spiders, assassin bugs, and other animals specialized to prey on ants, as reviewed for example by Wilson (1971), Hunt (1983), Oliveira and Sazima (1984, 1985), and Redford (1987). Yet there are virtually no systematic audits of the entire set of predators of individual ant species, nor have field experiments been performed to determine the long-term effects of predation on population growth.

Competition. To summarize our presentation on foraging strategies so far, each ant species lives within a particular temperature-humidity envelope. Its workers employ foraging procedures that evidently represent trade-offs between the optimal food-gathering strategy and the need to avoid being eaten by predators. Population density is constrained by the kinds of food and nest sites utilized by the species, as well as by predation. Each of these factors is known to limit geographic ranges and local densities in particular species. Some of the published accounts documenting their effect have been detailed and definitive, others anecdotal and less persuasive. In either case, studies of population control in ants have been for the most part unifactorial. Assessments have not been made of the relative importance of multiple factors and their interactive, second-order effects. Long-term studies remain very scarce.

Yet in spite of the inchoate, not to say incoherent state of population regulation studies in ants, one process has asserted itself with undeniable force. Competition, by its ubiquity and the ease with which it is detected, has become the dominant theme in studies of ant ecology. The trend cannot be written off as fashion or an artifact of the procedures of field study. It is clearly a very important phenomenon in the biology of ants.

Before laying out the results of research on the subject to date, it will be worthwhile to provide a brief theoretical background. The role of competition has been challenged on largely methodological and statistical grounds in recent years (for example Connell, 1975; and Simberloff, 1983, 1984), but the large amount of evidence has stood up well under critical examination, while additional field studies have supported it further (Schoener, 1983, 1986; articles in Diamond and Case, 1986). It has been convenient to divide competition into two modes: interference competition, in which individuals exclude one another by threats, fighting, or poisons; and exploitative competition, in which individuals use resources and hence deprive others of their use, but without direct forms of aggression. The distinction is made more vivid by the occasional use of the expressions “contest” for interference competition and “scramble” for exploitative competition. An analogy would be the comparison of small boys running a race to see who wins a pile of coins (contest competition) and the same small boys racing to pick up as many coins as possible thrown in front of them (scramble competition).

The intensity of competition both within and between species differs greatly according to habitat and the position of the species in the food web. In 1960 Hairston, Smith, and Slobodkin presented a model that postulates major differences among species as a function of their trophic level. They proposed that carnivores, at or near the top of the food web, should actively compete, since their principal density-dependent control would be the exhaustible food supply of herbivores on which they prey. The same should be true of primary producers (plants), which have a limited supply of incident solar energy, and scavengers and other decomposers, which are limited by the finite amount of dead tissue and waste material made available by the other trophic levels. In contrast, herbivores should not compete as much because, occupying the intermediate position in the food web, they are more likely to be held down by predators. This trophic-level rule of competition has held up well in numerous field studies of competition during the past twenty-five years. It is well marked in terrestrial and fresh-water systems, where producers, as well as granivores, nectarivores, carnivores, and scavengers taken together, display more competition than herbivores and filter feeders. However, in marine systems no trend is yet detectable in one direction or the other.

In two important reviews of the subject, Schoener (1983, 1986) argued that not only do the trophic levels matter, but also other properties of particular species within each trophic level. Perhaps most importantly, we should expect competition to be most intense in large animals, since they are closest to the top of the food chain as well as long-lived and resistant to other predators. Such organisms are most likely to saturate the environment in a way that reduces food or fills nest sites until these requisites are in short supply and competition becomes a principal density-limiting force. Large size and long life are of course most familiarly displayed by carnivorous vertebrates. But they are also hallmarks of ant colonies. An ant colony, the replicating unit of these social insects, is a large “organism.” It is long-lived, in some species the record holders of all insects and longer-lived than most vertebrates. Beyond its founding stage, the colony is also typically sheltered and heavily defended in specialized nests, so that the queen is usually immune to predators. Finally, the ant colony weighs heavily on the surrounding terrain, regardless of whether it preys on insects and other organisms, scavenges, or gathers seeds. In a nutshell, ants have all of the traits expected to generate competition within and between species, and that is what we find.

The evidence for the pervasive role of competition in ants is complex and multilayered. It can be arranged into four categories as follows:

(1) Local communities of ant species are often overdispersed with reference to body size, food type, or weight of food item.

(2) Nest sites and foraging columns of the same or closely related species are overdispersed.

(3) Experimental removal of some colonies (of the same or different species) causes increased growth in other colonies.

(4) Displacement of individual foragers or colonies by other foragers or colonies is commonly observed.

Overdispersion of biological traits. The single most prominent pattern on which animal ecologists fasten their attention is the diversification of closely related species occupying the same locality. Bernstein (1974), for example, noted that three dominant seed-eating ants in the Mojave Desert of California, Veromessor pergandei and two species of Pogonomyrmex, forage at different surface temperatures, which in turn are correlated with the time of day at which foraging occurs. In arid environments of southwestern France, the granivore Messor capitatus is limited to dry, calcareous grassland slopes, while Messor structor occupies a wide range of surrounding habitats (Delage, 1968). In the savanna of the Ivory Coast, Crematogaster heliophila builds carton nests in the canopy and defends arboreal territories that extend through several trees. Crematogaster impressa occupies a complementary position, nesting mostly in hollow twigs in bushes near the ground (Delage-Darchen, 1971). In a wholly different dimension studied by Chew and Chew (1980), ant species occupying an evergreen woodland in Arizona and belonging to the same feeding guild (granivores or fluid-feeding) were most different in body size. The differences between a given pair of species were less marked when one of the species was uncommon, and they were smallest of all when the species occupied different guilds. Comparable results were obtained by Whitford (1978b) in a study of the granivore and omnivore guilds of the New Mexican desert, and by Chew and De Vita (1980) in granivore and insectivore guilds of the Arizona desert.

The divergence of species along many axes of the potential niche open to ants is of course to be expected if species are competing with each other. In order to coexist--at least in the classical view--species must differ from one another to a sufficient degree in their utilization of limiting resources such as food and nest sites. This minimal critical divergence should be reflected in the phenotypic traits that make it possible, whether body size and hence size of food items, the time of foraging, and so forth. However, the mere existence of the pattern is not proof in itself of competitive displacement. Such is the fate of correlative studies in general: suggestive but not definitive. There is always the possibility of some third, still undisclosed class of causative factors. Another complication is the fact that ants of very different sizes often fight and exclude one another locally, or steal food from each other in a way that tends to neutralize the phenotype differences that might otherwise circumvent competition. It is therefore desirable to seek other, more rigorous forms of verification, and, happily, these are easily obtained.

Overdispersion of colonies. As a rule, mature colonies of social insects belonging to the same or closely related species are overdispersed, that is, spaced so that the distances between them are too nearly uniform to have been randomly set. Most entomologists who have examined local distributions of ant colonies have arrived at this conclusion including, among earlier researchers, Elton (1932), Talbot (1943a, 1954), Brian (1956a, 1965b), Wilson (1959a,c), and Yasuno (1964, 1965b). In an important review, Levings and Traniello (1981) statistically analyzed 80 data sets gathered by themselves and previous authors. In many cases the distributional data had been independently assembled by different authors for the same species, especially in the genera Formica, Lasius, and Myrmica. The studies covered both terrestrial and arboreal nest sites in a wide range of habitats in both temperate regions and the tropics. For the 80 cases that Levings and Traniello were able to test statistically, 67 showed overdispersed nest distributions or tended toward such a pattern. In another 80 cases that could not be tested statistically, all appeared nevertheless to be overdispersed. Thus it is safe to say that a majority of species so far studied tend to have regular nest arrays. This pattern holds across the principal ant subfamilies with persistent nest sites (Ponerinae, Pseudomyrmecinae, Myrmicinae, Dolichoderinae, and Formicinae), as well as across many habitats and food types, not to mention investigators. One group that could not be assessed in this manner, due to inadequate data, are the ants that defend only their nests. Many, including the collembolan-feeding members of Strumigenys and other Dacetini, are cryptic and difficult to census in the field.

In a second study, Levings and Franks (1982) examined the nest distribution of 16 ecologically similar species that live on the floor of the moist tropical forest on Barro Colorado Island, Panama. When all the species were grouped into a single sample, the nests were found to be overdispersed from each other. Each of the more abundant species treated separately was also overdispersed. Levings and Franks concluded that the aversive interaction, whatever its nature, is stronger among the colonies belonging to the same species than among colonies belonging to different species, but both types are potent enough to create a relatively easily detected effect in the fauna as a whole.

What of army ants, which have no fixed nest site but wander from one place to another? It might seem that their paths would be randomly distributed, but this appears not to be entirely the case, at least not in the swarm raider Eciton burchellii. Franks and Bossert (1983) developed a computer simulation model incorporating all of the known aspects of Eciton burchellii foraging behavior. With a density of model colonies similar to that actually occurring on Barro Colorado Island, collisions occurred in the computer at a frequency of approximately once per colony per 250 days. Yet during the thousands of raids observed since 1929 by Schneirla (1933b), Willis (1967), and Franks (in Franks and Fletcher, 1983), not a single collision has been recorded in nature. A likely explanation for this non-event is that colonies reject areas still contaminated with the trail pheromones of other Eciton burchellii colonies. The hypothesis is supported by Willis' observation that one colony withdrew its raid from an area recently visited by another colony. Accordingly, Franks and Bossert programmed their computer colonies to stop raiding when they encountered trail systems that had been laid during the past 20 days--in accordance with field observations of trail durability. With this and several other realistic alterations in raiding frequency and direction, the model colonies collided only once per colony per 600 days.

Overdispersion is most simply explained as the outcome of some form of territorial behavior. The mutual exclusion can come about by an active aversion in which the colonies emigrate until they are spaced on all sides from their nearest competing neighbors. It can also result from colonies mutually annihilating one another until only one remains in the minimal defensible space. Finally, it can arise through preemption, in which the first colony to become established destroys incoming foundress queens before they are able to establish strong colonies. All of these processes occur widely in the ant, as we shall see, making the simplest unitary explanation of overdispersion also the most plausible.

The competition hypothesis of colony overdispersion gains added credence from the observation by Waloff and Blackith (1962) that nests of Lasius flavus, a species known from direct behavioral observations to be territorial, are randomly distributed in areas of low population density but overdispersed in areas of high population density. A comparable result has been reported by Byron et al.(1980), who found that during periods of food shortage the desert ants Aphaenogaster cockerelli and Veromessor pergandei broadened their diets, increased their foraging space, and overdispersed their nests with reference to each other. Ryti and Case (1984, 1986) found that not only are colonies of Veromessor pergandei and Pogonomyrmex californicus overdispersed, but large colonies are separated by greater distances than small colonies. By careful quantitative studies they were able to eliminate virtually every conceivable cause of overdispersion except intraspecific competition. A study of Myrmica lemani in Poland led Petal (1980) independently to a similar conclusion. During a reduction of food over two years, the ants enlarged the spectrum of the food items they collected while foraging farther from their nests. The dispersal of the nests simultaneously shifted from a clumped to a uniform pattern.

Experiments on competition. A series of field and laboratory experiments have left no doubt of the powerful role of competition in determining the structure of ant populations and community structure. One of the simplest but most persuasive was Brian's 1952 analysis of habitat selection by Scottish ants. He noticed that a rank order exists among species in the appropriation of favored nest sites by competing queens. In the cool, moist woodland of western Scotland, ant colonies are limited to sunny places where higher temperatures persist long enough to permit the rearing of brood. As newly mated ant queens enter rotting pine stumps, they move to the south side of the vertical surface just beneath the bark. When individuals belonging to the same species encounter one another, they group together or space out at very short intervals. When queens of different species meet, however, they space out at much greater distances. Under such conditions, Formica fusca occupies the southern face of the stump, which is the warmest area, while Myrmica ruginodis (= M. “rubra”) moves, for the most part, to the east face, which is the second warmest. Myrmica scabrinodis takes what is left. The tiny Leptothorax acervorum avoids conflict altogether by occupying galleries in the core of the stump too narrow to admit the other species.

Brian was then able to demonstrate through a laboratory experiment that the segregation of the species is due, at least in large part, to the repulsion of the Myrmica by the Formica from the favored southerly position. He placed newly inseminated queens in glass vessels heated to 30°C on one side. When joined with members of their own species alone, both Formica fusca queens and Myrmica ruginodis queens invariably clumped together in the warmest part of the chambers. When the two species were mixed together, however, the Myrmica ruginodis were displaced to the cooler part. The segregation matches that observed in dead stumps under natural conditions. In later stages of their population growth, the colonies of the different species occupy different positions, and in some cases they remain there into maturity.

Every ant researcher has observed the displacement of one ant species by another at baits, and the simplicity of the phenomenon invites its use in experimental tests of competition. In the forests of the eastern United States, for example, the formicine Prenolepis imparis is dominant over the myrmicine Aphaenogaster rudis in this respect. When Lynch et al. (1980) experimentally removed Prenolepis foragers from baits made of small pieces of tuna, the number of Aphaenogaster foragers at the baits dramatically increased. They were also able to harvest more food, as expected.

An alternative experimental procedure is to remove foragers of a colony entirely from the field and observe the response of its presumptive competitors. All such procedures known to us have resulted in an expansion of the released colonies, as expected if competition is occurring. The following examples are among the most clear-cut.

• When Kugler (1984) removed a colony of the Colombian harvester ant Pogonomyrmex mayri, the nearest Pogonomyrmex mayri colony nearly doubled its foraging area within four days, sending scouts deep within the territory of its former neighbor. In another locality two Pogonomyrmex mayri nests were surrounded by the multiple nest entrances of an apparent competitor species, the abundant ponerine Ectatomma ruidum. When Kugler removed the Ectatomma from the field by plugging up their entrances, both Pogonomyrmex colonies expanded the area of their foraging activity. When their fields met, however, one colony retreated as the other advanced, with the result that the areas occupied by the two colonies remained mutually exclusive.

• In Ghana, as elsewhere in tropical Africa, the weaver ant Oecophylla longinoda forms huge colonies that suppress many other arboricolous ant species. When Majer (1976a-c) removed the ant from local areas in a cocoa farm, Crematogaster striatula moved in. When Crematogaster striatula was also deleted, Crematogaster depressa moved in and Tetramorium aculeatum spread its nests into the thinner vegetation of the kind earlier preempted by the dominant species.

• On the Siberian steppes, Formica pratensis dominates Formica cunicularia. When Reznikova (1982) isolated nests of pratensis by means of oil-smeared obstacles, so that the workers were no longer able to forage, the weight of prey collected by the newly liberated cunicularia workers increased threefold and the average weight of the prey increased from 1.4 mg to 3.1 mg.

• Only four species of ants occupy small mangrove islets in the Florida Keys with areas of 12 square meters or less. When Cole (1983a) introduced colonies of Pseudomyrmex elongatus and Zacryptocerus varians onto empty islets, they persisted. When placed on similar islets containing Crematogaster ashmeadi or Xenomyrmex floridanus, they soon disappeared. (This experiment, which is important for the understanding of the assembly of island faunas, will be discussed in fuller detail in Chapter 11.)

• Three sympatric species in the desert of the southwestern United States, Pogonomyrmex rugosus, Pogonomyrmex desertorum, and Pheidole xerophila, feed on seeds of different but partially overlapping size ranges. Pogonomyrmex rugosus and Pogonomyrmex desertorum resemble each other substantially in this regard, as do Pogonomyrmex desertorum and the much smaller Pheidole xerophila, but there is little overlap between Pogonomyrmex rugosus and the Pheidole. If the three species are competing for seeds in the zones of overlap to an extent affecting their population size, general theory (see Levins, 1976) predicts that Pogonomyrmex rugosus actually aids Pheidole xerophila by impacting Pogonomyrmex desertorum, their common competitor. If Pogonomyrmex rugosus were absent, Pogonomyrmex desertorum should increase and Pheidole xerophila should decline. When Davidson (1985) destroyed Pogonomyrmex rugosus colonies with insecticides, she got precisely the predicted effect in the two surviving species.

• In the deserts of the American Southwest, ants and rodents are the principal consumers of seeds (Davidson et al., 1980). When Brown et al. (1979a) excluded rodents from small experimental plots near Tucson, Arizona, the number of ant colonies increased by 71 percent. In the reverse experiment, where ants were excluded, the numbers of rodents increased 20 percent and their biomass rose by 29 percent. A second series of experiments near Portal, Arizona, yielded a similar result for one species of ant, Pheidole xerophila, but not for another, Pogonomyrmex desertorum, or for rodents (Davidson and Samson, 1985).

The converse of the exclusion experiments just cited are packing experiments, in which species are added to localities to observe their effects on population structure and growth. Those performed to date have strongly implicated competition.

• The overdispersion of colonies due to mutual repulsion of different species was demonstrated in experiments by Bradley (1972). In the jack pine stands of Manitoba, Canada, mature colonies of Formica obscuripes and Hypoclinea taschenbergi reach very large size, with an estimated minimum of 40,000 adult workers; and they appear to be spaced at a minimum distance from each other. Bradley transplanted colonies into previously unoccupied pine plantations. When nest sites of the two species were alternated in position and spaced at intervals of 20 meters, few of the colonies emigrated. But when the sites were spaced at only 5 meters, a large percentage of the colonies emigrated, often repeatedly, and in sufficiently divergent directions to create greater spacing.

• Formica truncorum is an aggressive ant species that forms huge polydomous colonies in the drier habitats of northern Europe. Detailed biogeographic data collected in Finland suggests that it competes successfully with other ants and is most effective in suppressing populations of the smaller formicine Lasius niger (Rosengren et al., 1985, 1986b; Rosengren, 1986). Lasius niger colonies in turn compete with those of the more subterranean Lasius flavus (see Pontin, 1960, 1961, 1963), although less intensely than with one another. Formica truncorum and Lasius flavus, it appears, interact to a lesser degree. Rosengren and his associates introduced colonies of Formica truncorum to tiny islands in the Gulf of Finland already occupied by the Lasius species but lacking the Formica. The introduced colonies prospered as the [[Lasius niger colonies declined.  On the other hand, Lasius flavus colonies increased their populations.  Nearby “control” islets, on which no introduction of Formica truncorum were attempted, experienced little change in the Lasius'' populations.

Direct observation of competition. One of the behaviors of ants most commonly observed in nature is overt aggression, often resulting in injury, death, and displacement of one colony by another. Auguste Forel was correct in observing that “The greatest enemies of ants are other ants, just as the greatest enemies of men are other men.”

Interference competition among ants can be demonstrated within hours or even minutes almost anywhere in the world merely by placing food baits on the ground. This ultra simple method has become a standard tool for analysis (for example, DeVroey, 1978; Lynch et al., 1980). The baits most commonly used by entomologists are sugar water and pieces of tuna. In the faunas of the West Indies, which have relatively few species, consistent interactions follow easily perceived rules (Wilson, 1971; Levins et al., 1973). Paratrechina longicornis, the swift-running hormiga loca (crazy ant) are “opportunists.” The workers are very adept at locating food and often are the first to arrive at newly placed baits. They fill their crops rapidly and hurry to recruit nestmates with odor trails laid from the rectal sac of the hindgut. But they are also very timid in the presence of competitors. As soon as more aggressive species begin to arrive in force, the Paratrechina draw back and run excitedly in search of new, unoccupied baits. The species appears to survive by arriving and recruiting early enough to appropriate a share of the food, without however exposing themselves to any great risk in battle. A second class of species, the “extirpators,” recruit by odor trails and fight it out with competitor species. Examples include some species of Pheidole, Crematogaster steinheili, the fire ant Solenopsis geminata, the “little fire ant” Wasmannia auropunctata, and Camponotus sexguttatus. Some of the species have a well-developed soldier caste which, in the case of Pheidole and Solenopsis at least, plays a key role in the fighting. Injury and death are commonplace, and one species (or one colony in competition with other colonies of the same species) usually comes to dominate the bait. Preemption is the rule: the colony whose foragers arrive first typically wins, because the foragers recruit nestmates, who surround the bait and prevent scouts of competing colonies from sampling the food and recruiting on their own. When a few scouts encounter large numbers of a competitor colony they are more timid and easily repulsed. Large colonies also enjoy a local advantage over small colonies.

Finally, a third class of species, the “insinuators,” are exemplified by Tetramorium simillimum and species of Cardiocondyla. The colonies are small, lack a soldier caste, and evidently rely on small body size and stealthy behavior to reach the baits. When a Cardiocondyla scout worker discovers food, it recruits only one nestmate at a time by means of tandem running. The small numbers of workers reaching the baits by this means are usually able to ease their way to the edge of the bait through crowds of “extirpator” workers without provoking an aggressive response.

Territorial fighting is very general in the ants, sometimes widespread over the foraging areas and sometimes confined to the immediate vicinity of the nest. Among the most dramatic battles witnessed among colonies of the same species are those conducted by the common pavement ant Tetramorium caespitum. First described by the Reverend Henry C. McCook (1880), these “wars” occur in abundance on sidewalks and lawns in towns and cities of the eastern United States throughout the warm season. Masses of hundreds or thousands of the small dark brown workers lock in combat for hours at a time, tumbling, biting, and pulling each other, while new recruits are guided to the melee along freshly laid odor trails.

When major workers from two laboratory colonies of the African weaver ant Oecophylla longinoda are permitted to enter a new arena simultaneously, they quickly join in a territorial battle (Hölldobler and Wilson, 1978). As individual pairs meet, they rear up by straightening their legs and dodge around each other like prancing, stiff-legged dancers, then thrust back and forth in an attempt to seize each other. When one succeeds in grasping a mandible, antenna, leg, or other part of the adversary's body to secure a good grip, she bites down and pulls back on her legs (Figure 10-7). Sometimes the mandibles cut directly through, lopping off one of the appendages or slashing open the abdominal wall. Often nestmates recruited to the site join in the attack, spread-eagling the enemy by pulling its legs in all directions and finally dispatching her with further mandibular cuts. Encounters with alien members of the same species are exceptionally exciting to the Oecophylla workers. Some of those not immediately locked in combat return to the nest laying odor trails by dragging their rectal glands over the ground. As they encounter nestmates they antennate them vigorously while jerking their bodies back and forth in swift, exaggerated movements. As a result, substantial forces of new workers are recruited, and a veritable war between the two colonies ensues. The colony that wins is the one able to assemble the heaviest battalions of major workers, and this is typically the one that meets the enemy on ground previously marked with its own territorial pheromone. To use the celebrated phrase of General Nathan Bedford Forrest during the American Civil War, the winners arrive at the battle scene “fustest with the mostest.” The colony with the highest density of workers present is also victorious in a majority of local duels, since small groups are able to gang up more frequently on single opponents. As the combat d'usure proceeds and fatalities mount, the more weakly represented colony is gradually eliminated from the arena. The same pattern of combat and exclusion has been observed under natural conditions in Oecophylla longinoda in East Africa and in its sister species Oecophylla smaragdina in Australia (Hölldobler, 1979, 1983; Hölldobler and Lumsden, 1980). It also occurs regularly when Oecophylla meets other species of ants, including members of Anoplolepis, Iridomyrmex, and Pheidole (Way, 1953; Steyn, 1954; E. S. Brown, 1959). This leads to a mosaic distribution of territories of co-dominant ant species (Leston, 1973a-c; Majer, 1976a; Jackson, 1984). Similar interspecific territorial patterns have been reported from the Neotropical ant fauna (Leston, 1978) and from Europe (Pisarski, 1982), and will be treated in more detail in Chapter 11.

Ant wars are in fact a commonplace among species with large colonies. They have been especially well documented in the mound-building ants Formica lugubris (Cherix and Gris, 1980) and Formica polyctena (Mabelis, 1979b; Driessen et al., 1984). In [[Formica polyctena at least, they are cannibal wars.  The bodies of the enemy dead are carried back to the nest and eaten; battles are especially intense during times of general food shortage.  Formica polyctena has also been directly observed extirpating colonies of Lasius, Myrmica, and Formica fusca for use as food (Vepsäläinen and Pisarski, 1982).  In the case of the supercolonial Formica yessensis in Japan, the combination of warfare and local population saturation results in the virtual elimination of many other species within the local range of the dominant species (Higashi and Yamauchi, 1979).  Another case is the spread of the unicolonial Wasmannia auropunctata'' on Santa Cruz Island in the Galapagos, where it was introduced by human commerce. By 1976 the ant had expanded across approximately the southern half of the island, drastically reducing or even extinguishing the populations of every other ant species it encountered, including two endemics of the Galapagos (Clark et al., 1982).

The most extreme episodes of species exclusion, in which colonies can be observed extirpating colonies, occur when species possessing large colonies are introduced by human commerce into previously undisturbed native ant faunas. Among the best known cases involve, in addition to Wasmannia, the notorious Pheidole megacephala, a principal insect pest and destroyer of the Hawaiian native insect fauna, the fire ant Solenopsis invicta, which has invaded almost the entire southern United States, and the so-called Argentine ant Iridomyrmex humilis (Illingworth, 1917; M. R. Smith, 1936a; Haskins, 1939; Wilson and Brown, 1958a; Haskins and Haskins, 1965; Wilson and Taylor, 1967b; Ward, 1987). The process of extirpation has been observed by many authors. Fluker and Beardsley (1970) found that in Hawaii an Iridomyrmex humilis colony takes about 10 to 14 days to eliminate all of the workers from a Solenopsis geminata nest. When they invade Pheidole megacephala nests, the Pheidole abscond and move to a new, presumably less desirable site.

It is entirely possible that given enough time, evolutionary or ecological change will result in the “taming” of such species and their increasingly harmonious assimilation into the indigenous faunas. Wheeler (annotations to Réaumur, 1926) reminds us that in the early 1500s a stinging ant appeared in such huge numbers as to cause the near abandonment of the early Spanish settlements on Hispaniola and Jamaica. The colonists of Hispaniola called on their patron saint, St. Saturnin, to protect them from the ant and conducted religious processions through the village streets to exorcise the pest. What was evidently the same species, which came to be known by the Linnaean name Formica omnivora, appeared in plague proportions in Barbados, Grenada, and Martinique in the 1760s and 1770s. The legislature of Grenada offered a reward of £20,000 for anyone who could devise a way of exterminating the pest, but to no avail. It now appears that Formica omnivora was none other than the familiar “native” fire ant, Solenopsis geminata, which at the present time is a peaceable and only moderately abundant member of the West Indies fauna.

Iridomyrmex humilis and Pheidole megacephala are wholly incompatible species, with humilis the usual if not invariable winner in subtropical to warm temperate localities between 30° and 36° latitude north and south, and megacephala prevailing in the tropics. Thus humilis is dominant in disturbed habitats in California, the Mediterranean area, southwestern Australia, and the island of Madeira, but only above 1000 meters in Hawaii, the zone cool enough to favor it (Fluker and Beardsley, 1970). The Iridomyrmex populations penetrate new environments on foot, with raiding columns of workers clearing the way and pioneer communities of workers and queens, like Zulu bands, following into the freshly opened nest sites. New populations, on the other hand, spring from little groups of workers and queens that are carried inadvertently in human freight and baggage. The gradual replacement of Pheidole megacephala by Iridomyrmex humilis on Bermuda, which began with the establishment of the Iridomyrmex around 1953, has been recorded across a decade by Haskins and Haskins (1965), Crowell (1968), and Lieberburg et al. (1975). The most recent survey, by Lieberburg and his co-workers, suggests that the two contestants may be approaching a dynamic equilibrium. Between 1966 and 1973 the Iridomyrmex established two new local populations, while the Pheidole finished reclaiming three other sites.

As the red imported fire ant Solenopsis invicta progressed across the southern United States, it virtually eliminated the native fire ant Solenopsis xyloni and reduced the population densities of Solenopsis geminata and the harvesting ant Pogonomyrmex badius (Wilson and Eads, 1949; Wilson and Brown, 1958a). Throughout this newly conquered range, the Solenopsis invicta population spread principally by nuptial flights and the establishment of additional colonies from newly inseminated queens. As a result, the populations are patchy in local distribution and often form mosaics with those of the chief competing species (Tschinkel, 1988b). At the Brackenridge Field Laboratory of the University of Texas, in Austin, the local population has become unicolonial. In other words, it has abandoned its colony territorial boundaries and is now composed of huge numbers of workers and queens that emigrate over short distances. The continuity and high density of this peculiar population allows it to penetrate microhabitats not ordinarily available to the species, and it has been able to devastate most of the resident native species (Sanford D. Porter, personal communication).

Finally, De Vita (1979) has gone so far as to measure the amount of aggression among colonies in a natural population of the Mojave Desert harvester ant Pogonomyrmex californicus. Eighty-one percent of the encounters between foraging workers from different colonies resulted in some form of aggression, and 7 percent were fatal to one or the other participants. The cost of this interference competition was 0.06 deaths per ant foraging hour, a deficit that has a substantial impact on colony size and growth.

Territorial strategies
A territory is defined as an area occupied more or less exclusively by an animal or group of animals by means of repulsion through overt defense or advertisement. Any given territory can be absolute, in the sense that an established terrain is defended all the time, whether it contains food, nest sites, or any other resource at any particular time. Alternatively, the territory can be spatiotemporal, meaning that only limited sectors of the foraging area are defended at any moment, and there are usually places where resources are temporarily located. Thus the large colonies of fire ants, Oecophylla weaver ants, wood ants of the Formica rufa species group, and Australian meat ants of the species Iridomyrmex purpureus maintain absolute territories (Yasuno, 1965b; N. L. Wilson et al., 1971; Greenslade, 1975a,b; Mabelis, 1979b; Hölldobler, 1983). Spatiotemporal territories are defended by colonies of Prenolepis imparis, honeypot ants of the genus Myrmecocystus, and most species of the diverse and abundant genus Pheidole (Talbot, 1943a; Hölldobler, 1981a, 1986b; Wilson and Hölldobler, 1985).

Territorial behavior is in fact a very general if not universal phenomenon, ranging according to species and stage of colony growth from defense of large absolute foraging domains down to the immediate nest vicinity. Because of the division of labor between reproductive individuals and sterile worker castes, fatalities caused by territorial defense have a qualitatively different significance for ants and other social insects as compared to solitary animals. The death of worker ants represents no more than an energy and labor debit, rather than the destruction of a reproductive agent. The acts performed at risk of death can more than offset the cost of death by providing or maintaining resources and colony security. In this sense frequent death can even become a positive element in the colony's adaptive repertory, as suggested by Oster and Wilson (1978) and documented by Porter and Jorgensen (1981) for Pogonomyrmex and Schmid-Hempel and Schmid-Hempel (1984) for Cataglyphis. Hence strong territorial defense is more likely to have evolved in ants than in most solitary animals.

In more general terms, natural selection theory suggests that an animal should establish a territory only when its size and design should make it defensible in economic terms. That is, the territorial defense should gain more energy than it expends (J. L. Brown, 1964; J. L. Brown and Orians, 1970; Oster and Wilson, 1978; Davies and Houston, 1984).

It follows that the analysis of territorial strategies in ants must include the study of the design and spatiotemporal structure of the territory, as well as the social mechanisms through which the strategies are achieved. Important differences in the use of space exist among ant species. Species foraging on relatively stable resources dispersed over a wide area defend territories that are different in geometry from those of species exploiting stable but patchily distributed resources. Distinct strategies are also associated with food resources that are patchy, relatively stable, and predictable, or else frequently changing in location in an unpredictable manner. In the following sections we will briefly examine ant species that display much of this variation among themselves. Each of the species are ecologically dominant animals that have been studied extensively both in the laboratory and in their natural habitats. Adaptation to food sources that are relatively uniform in space and stable in time is exemplified by the African weaver ant Oecophylla longinoda. In contrast, certain species of harvester ants in the genus Pogonomyrmex utilize food sources that tend to be stable but are often distributed patchily in space. Finally, the honeypot ant Myrmecocystus mimicus harvests resources that for the most part occur randomly in space and time.

African weaver ants: absolute territories. For blackbirds (Euphagus cyanocephalus) and many other solitary animal species, resources are uniformly distributed and continuously renewing. It is advantageous for them to maintain a complete defense of whatever portions of the foraging area can be patrolled in a reasonably short period of time (Horn, 1968). A striking analogy exists in the absolute foraging territories of the African weaver ants (Oecophylla longinoda), although the mechanisms of establishing and maintaining the territories are very different.

Oecophylla longinoda is a dominant species in the forest canopy in a large part of tropical Africa. Workers bind leaves into tight nest compartments with silk spun by the final instar larvae. One colony usually builds hundreds of such leaf nests, which are distributed over several nest trees and concentrated in the peripheral canopy of the trees. Weaver ant colonies are monogynous; the single reproductive queen resides in one of the leaf nests, completely surrounded and protected by workers. Most of the other leaf nest compartments are filled with brood of all stages and hundreds of workers. The worker force of a mature Oecophylla longinoda colony sometimes consists of more than 500,000 individuals.

Weaver ants are strongly predaceous, using their cooperative ability to capture a wide range of large insect prey that ventures onto their territory (Figure 10-8). Although the foragers tend to remain on the trees and surrounding low vegetation, they also hunt extensively on the ground; their territories are three-dimensional. Oecophylla workers patrol every part of their territory, tolerating only very few ant species on the trees they occupy. But they exclude foreign colony members of their own species in aggressive interactions (Figures 10-9 and 10-10), so severely that they create narrow, unoccupied corridors, which are in effect “a no ant's land.”

In an area of the Shimba Hills Reserve in Kenya, Hölldobler (1979) studied individual territories sometimes covering an area of up to approximately 1600 square meters across the canopies of 17 large trees. Hölldobler was able to determine the territorial borders by repeatedly transplanting sets of 20 Oecophylla workers from one nest tree to another. So long as the trees belonged to the same territory, no conspicuous response was observed. But when the transplanted ants were released within the territory of a neighboring colony, a massive defensive recruitment response was elicited in the resident ants. Individual ants that encountered the foreign intruders ran back to some of the nearest leaf nest compartments in the tree canopy, laying a trail with secretion from the rectal gland. When they encountered nestmates, they performed a vigorous jerking display similar to the initial fighting behavior between two individual opponents in the manner described earlier in laboratory colonies (see Chapter 7). Nestmates approached by a jerking recruiter ant became very excited and ran along the trail toward the battle site. After 30 minutes, often many more than 100 ants had assembled in this area and remained there long after all the “intruder” ants had been killed.

The weaver ant colony finds considerable advantage in excluding competitor colonies of all but the smallest or most dietarily different species. Indeed, Oecophylla longinoda is highly aggressive not only to conspecific aliens but also to many other ant species. In the Shimba Hills, only a few ant species were found to coexist with Oecophylla on the same tree. Some other arboreal ant species (for example, Camponotus sp.) that were never found to coexist with Oecophylla on the same tree elicited as vehement a defensive recruitment response in Oecophylla as conspecific aliens when they were introduced into a weaver ant territory. It is interesting to note that the Camponotus also responded with a defensive recruitment of hundreds of nestmates when Oecophylla workers were released on a Camponotus tree. These results suggest that both Oecophylla and Camponotus react with a massive defensive response only to certain intruding ants, particularly the most serious competitors for essential resources (nesting sites and food), and predators including the army ant Dorylus and the formicine Anoplolepis longipes. Thus Oecophylla longinoda employs “enemy specification” in a selective form of absolute territory. Oecophylla diverges from the classic central place model in the mechanisms by which it maintains territorial control. Their nests are not aggregated in one central location but are effectively decentralized throughout much of the territory that they control. This large territory can be patrolled and exploited over much of its volume simultaneously without the colony's incurring the costs of transporting prey from distant capture points to a single central nest deep within the territory. These factors induced Hölldobler and Lumsden (1980) to formulate a different general model concerning the geometry of the “nest decentralization” and the economic defensibility of the Oecophylla territory.

A solitary animal, or even a social group with relatively few members, faces a challenge if a territory of considerable size is to be established and maintained: the defender cannot be everywhere at once. Invaders can often penetrate deep into the territory and even harvest resources before they are detected and expelled. This qualitative observation suggests that the solitary holder of a territory faces serious constraints in maintaining a large territory, especially when the density of the invader is high. In contrast, Oecophylla is virtually everywhere at once throughout its territory. Invaders are detected at or very close to a colony's boundary. Consequently, the Oecophylla territorial system allows harvesting benefits from throughout the interior of the territory while restricting the defensive costs almost purely to the boundary or margin.

It is interesting to note that the Nearctic species Lasius neoniger, although much smaller, appears to have similar decentralized territories. Like those of Oecophylla, societies of Lasius neoniger are monogynous and occupy multiple nests that are distributed throughout the colony's foraging area. Traniello and Levings (1986) hypothesize that this nest structure of Lasius neoniger partitions “the territory within a colony by spatial subdivision of its foragers, and thus may reduce loss of prey to competitors.”

The strategic implications of such a system are far-reaching. To explore some of the consequences, consider a circular territory of radius r with uniformly distributed, temporally stable food sources (see Figure 10-11). Similar conclusions hold for territories that approximate convex regions, as in the case of Oecophylla (see Figure 10-10). A cohort of foragers, dispersed throughout the interior of the territory at density rw, returns net foraging benefit B2 > 0 per unit area per unit time. Because resources and workers flow locally into many leaf nests dispersed throughout the territory and not back to one central place located deep inside the territory, B2, which includes the cost of food transport, does not decrease as r increases. Thus B2 can be approximated by a constant independent of r.

Since costs due to intercolony aggression occur primarily at the edge of the territory, let C2 be the cost per unit time per unit length of boundary. Under these circumstances a territory of radius r yields benefit B2πr2 per unit time to a colony and costs it 2C2πr. A net ergonomic profit rate πr(B2r - 2C2) is produced by this strategy. Thus we find economic defensibility threshold r* = 2C2/B2 beyond which the territory yields net positive return. Moreover, the net return is an increasing function of r. Once beyond r*, the net profit rate is positive and continues to increase (Figure 10-12a). The ratio of defense cost to net foraging benefit is just r*/r and thus decreases monotonically with r. Colonies that use such a strategy should prefer the largest possible territory. Eventually, sufficiently large r is reached such that the egg-laying capacity of the queen is reached and the colony, in addition, begins to control an unusable surplus of resource. Thus the increment to benefit eventually approaches zero, and optimal economic defensibility occurs at an intermediate r, say r0 (Figure 10-12b). The very large number of workers that mature Oecophylla colonies can maintain suggests a large laying capacity in the queen and the potential for very large territories.

The territories of Oecophylla are actually three-dimensional rather than two-dimensional. But so long as the territorial strategy is one of decentralized nests, with workers monopolizing the interior volume of the territory while confining costs of defense to the boundary, there are economic defensibility thresholds; and the ratio of cost to benefit continues to decline along the curve r*/r as the territory expands in size below saturation of the egg-laying capacity of the queen and the surplus limits of resources. Differences between the geometry of two and three dimensions and deviations from this idealized strategy impose quantitative differences in the details of territorial design.

There may also be territorial size limits beyond which such factors as transport of brood from the queen nest to other leaf nests in the colony becomes ergonomically impractical or the system of inter-nest coordination becomes too complex to maintain. The observed sizes of Oecophylla territories suggest, however, that the way stations offered by established leaf nests and the rather local nature of the nest-nest interactions within the territory make such limits relatively weak. It appears that there is a high selection pressure for expanding territorial space and thus increasing access to essential resources while simultaneously improving the relative economic defensibility of the territory. Indeed, the three-dimensional territories of Oecophylla longinoda are probably the largest territories known for all invertebrates. They are approached in structure and size by a few other arboreal ants, including some species of Crematogaster and Pseudomyrmex, which have evolved parallel social systems.

Harvester ants: trunk trail territories. Generally, the partitioning of space between ant colonies is effective in reducing aggression between individuals belonging to the same species but to different colonies. However, the pattern of space partitioning can be very different and depends largely on the foraging strategies of the species. In contrast to the African weaver ants, the harvester ants Pogonomyrmex badius, Pogonomyrmex barbatus, and Pogonomyrmex rugosus, which are among the most abundant species of ants in the southern United States, tend to exploit patchy food supplies and accordingly show a more complex partitioning structure (Hölldobler, 1974, 1976a; Davidson, 1977b; J. H. Brown et al., 1979a,b; Harrison and Gentry, 1981).

Foragers of the harvester species travel on well-established trunk trails before diverging onto individual excursions. After foraging, the workers return to these routes for homing. Such trunk trails sometimes extend for more than 40 meters; they are remarkably persistent over long periods of time and even survive heavy rainfalls. The trunk routes originate from recruitment trails laid to newly discovered seed falls. The recruitment pheromone, which is relatively short-lived, is discharged from the poison gland and deposited on the ground through the extruded sting of foraging ants. Because the seed patches are frequently quite stable, the ants continue to travel along the former recruitment trail to these foraging sites. Persistent chemical sign posts are also deposited along the trail and, together with the visual markers, serve as orientation cues long after the recruitment signal has vanished.

Trunk trails of intraspecific neighboring colonies are overdispersed and almost never cross (Figure 10-12). On the contrary, they usually diverge and channel the mass of foragers of hostile neighboring nests into very different directions. The trunk trails, together with the immediate surrounding of the nest entrance of mature colonies, can be considered the “core area” of the colony's territory. Although foraging areas of nearby colonies can overlap, aggression in the overlapping zone is usually limited to individual confrontations between two foragers. However, when two trunk trails of neighboring colonies are brought into contact, aggressive mass confrontations occur; and they continue until the trunk trails have diverged again (Figure

10-13). Although the major function of the trunk route foraging system seems to be to facilitate long distance orientation and the exploitation of patchily distributed and relatively stable food sources, the topographic design of the route system of each colony in turn depends greatly on the route maps of its neighboring colonies.

Up to this point we have considered only the partitioning of foraging grounds by the trunk route system between colonies belonging to the same species. Pogonomyrmex barbatus and Pogonomyrmex rugosus, however, seem to be identical with regard to food type, nesting site, and foraging activity period, as though they were ecologically the same species. Although populations of both species are largely separate geographically, zones of overlap exist (Hölldobler, 1976a; Davidson, 1977b). It is not surprising therefore that territoriality is strongly developed both within and between these two species, and that the foraging area is subdivided interspecifically by the trunk route system. These findings lead to the question of the cues used by the ants to discriminate between colonies of their own and other species (see Chapter 5). The foragers not only have to be able to recognize members of their own colony and those belonging to another conspecific colony, they also need to identify other species that are potential competitors for essential resources.

Although not much is known about the mechanisms of species discrimination in ants, there is some analytical evidence of species specificity in the mixtures of hydrocarbons of the Dufour's gland in Pogonomyrmex. It is interesting that Pogonomyrmex rugosus and Pogonomyrmex barbatus, which exhibit strong interspecific territoriality, have almost identical patterns in the mixtures of the compounds from their Dufour's glands, whereas other Pogonomyrmex species, whose territories overlap with those of Pogonomyrmex rugosus and Pogonomyrmex barbatus, have very different patterns (Regnier et al., 1973; Regnier and Hölldobler, unpublished; see Hölldobler, 1986b, 1987).

When designing a mathematical model of trunk trails and territories, Hölldobler and Lumsden (1980) focused on species that utilize spatially patchy but long-lived resources, such as the Pogonomyrmex harvester ants. They visualized the foraging area as a system of sectors that converge in a polar coordinate system on the nest (see Figure 10-14). The foragers are deployed in such a way as to maximize net energetic yield, measured as the benefit in calories of seeds harvested minus the costs incurred due to the metabolism and mortality during seed retrieval. The resulting territorial design approximates a trunk trail system. The model incorporates such details as the efficiencies of food retrieval over narrow trails, the effects of nearby alien colonies, the presence of multiple patches in each sector, and time-varying costs. In simplest terms, the width of each sector is likely to be determined by these several biological factors. The colony should avoid empty sectors altogether except for occasional patrols by scouts. And whenever the cost-to-benefit ratio of a formerly productive sector falls below unity, the colony should close it down. Such would be the case when the patch is not sufficiently rich or close to the nest, or when a rival nest already occupies the patch and usurpation costs would be greater than the acquired benefit.

Experimental field data agree well with these economic considerations. New foraging sites, which have not yet been discovered by competitors, are rapidly explored and occupied with the aid of an effective chemical recruitment system. In fact, the significance of the recruitment communication in harvester ants becomes especially apparent in view of the strong intra- and interspecific competition for the same foraging areas. Hölldobler (1976a) found that the forager recruitment activity not only depends on a number of parameters of the food source--such as distance to the nest, density of the seed fall, and size of the grains--but also on the presence or absence of foreign foragers at the resource patch. Seed sites previously occupied by competing foragers were considerably less attractive than unoccupied seed sites.

Although most Pogonomyrmex barbatus and Pogonomyrmex rugosus colonies studied by Hölldobler (1976a) and Davidson (1977b) had extensive trunk trail systems, a number of nests, especially in areas with sparser vegetation, were without trunk trails. This is a general trait of Pogonomyrmex maricopa, which was never observed to produce long-lasting trunk trails. In these cases ants usually leave the nest on their individual foraging excursions and disperse in all directions. Yet as Hölldobler's experiments demonstrated, the foragers also show a high directional fidelity, even though all sectors appear to be more or less equally frequented by the worker force as a whole. This suggests that the seeds do not accumulate in patches but instead are randomly dispersed. In other words, the individual sectors are more or less equally productive. This foraging system obviously does not allow as subtle a partitioning of foraging grounds between neighboring colonies as does the trunk trail system. Such colonies invariably show a much wider spacing pattern (Hölldobler, 1976b).

In short, the trunk routes of Pogonomyrmex are partitioning devices which curtail aggressive confrontations between neighboring colonies, while at the same time enlarging the foraging area for patchy food supplies. A similar foraging and partitioning system has been reported for several species of Formica (Rosengren, 1971; Bruyn and Mabelis, 1972; Bruyn, 1978), Lasius (Traniello, 1980), Messor (Lévieux and Diomande, 1978a), and the leafcutter ants of the genus Atta (S. Hubbell, personal communication).

Other trunk trail territories. The trunk routes of the seed harvesting ant Pheidole militicida share many traits with those of Pogonomyrmex (Hölldobler and Möglich, 1980). They, too, originate from chemical recruitment trails and are stabilized by enduring chemical orientation cues and visual markers. Nevertheless, they are not as persistent as the foraging pathways of Pogonomyrmex. In fact, the route foraging system of Pheidole militicida seems to be intermediate between that of Pogonomyrmex and that of Veromessor pergandei. Messor employs a foraging strategy in which various sectors of the foraging area around the nest are successively exploited by the ants through rotational changes in direction (Rissing and Wheeler, 1976). Like Messor, Pheidole militicida workers shift the direction of the foraging pathway or establish a new route when the seed supplies in a foraging area diminish. However, the shifts do not occur as regularly as in Messor, and no geometric pattern in the change of the foraging columns has been detected. There is some recent evidence that the trunk trails of Messor also serve in territorial partitioning (Ryti and Case, 1986; Rissing, 1987), but this appears not to be the case in Pheidole militicida. Although little is known about the territorial behavior of these species, it seems that they have less rigid territorial boundaries and defend only those areas being heavily used by foragers.

Honeypot ants: spatiotemporal territories. When resident ants defend only those portions of their territory in which they happen to forage and encounter intruders at close range, as opposed to the trunk trails or entire foraging area, the territorial defense can be said to be truly spatiotemporal. We will now describe a spatiotemporal territorial strategy of honeypot ants (Myrmecocystus mimicus) that appears to have evolved as part of a foraging system designed to utilize patchily distributed but unpredictable or unstable food sources. Myrmecocystus mimicus is abundant in the mesquite-acacia community of the southwestern United States (Snelling, 1976). Like other members of its genus, it has a special honeypot caste, the members of which function as living storage containers.

One of the major food sources of Myrmecocystus mimicus is termites. When a scouting ant discovers a rich supply of termites, for example under a piece of dried cattle dung, it directs a group of nestmates to this food supply by means of special recruitment signals, including motor displays and odor trails laid with hindgut liquid. If another colony of Myrmecocystus mimicus is located near the food source and is detected by the foragers of the first colony, some of these individuals rush home and recruit an army of 200 or more workers to the foreign colony. They swarm over the nest and engage all of the workers emerging from the alien nest entrance in an elaborate display tournament, thus blocking this colony's access to the food supply (Figure 10-15). Frequently scouts leave the tournament to return to their colony in order to recruit reinforcements, while the other group of nestmates continues to retrieve the termite prey. Once the food source has been exhausted, and the foraging activity in the area declines, the tournament activity at the neighboring nest site also declines and the intruding army finally retreats to its own nest.

Although hundreds of ants are often involved in the territorial tournaments, almost no physical fights occur. Instead individual ants engage each other in highly stereotyped aggressive displays (Plate 14). Each group of displaying ants breaks up after 10 to 30 seconds, but the ants continue to run about on stilt-like legs. When they meet a nestmate they respond with a brief jerking display, but when they meet another opponent the whole aggressive display ceremony is repeated.

Field observations have shown how it is possible for scouting ants to venture very deeply into the foraging area of a neighboring colony. Although individual foragers of Myrmecocystus mimicus frequently disperse in all directions when leaving the nest, they tend to swarm out of the nest at intervals. Such departures are usually spaced irregularly in time, and an interval between two departures can last as long as several hours (Figure 10-16). Sometimes only very few foragers leave a nest during a whole day, while in contrast the foragers of a neighboring nest may be very active. Then within a few days the situation is reversed. However, if a scout of a relatively inactive colony discovers a rich food source, such as an access to a gallery of a termite nest, it can quickly galvanize the colony into foraging activity by recruiting nestmates to the food source (Figure 10-17). But the termites may have already been discovered by foreign scouts of the very active neighboring colony, which then sends an army of ants to the competing nest in order to interfere with the first colony's foraging activity.

Territorial tournaments also occur in the zone between two adjacent Myrmecocystus mimicus nests, especially when both colonies are active at the same time. Alien foragers are then blocked from the respective foraging areas of each colony (Figure 10-17b). These tournaments sometimes last for several days, being interrupted only at night when workers of this species are normally inactive.

When one colony is considerably stronger than the other, that is, when it can summon a much larger worker force, the tournaments end quickly and the weaker colony is raided. During these raids the queen is killed or driven off. The larvae, pupae, callow workers, and honeypot workers are carried or dragged to the nest of the raiders (Figure 10-18). Field observations and laboratory experiments have led to the discovery that the surviving workers as well as the honeypots and brood of the raided colony are incorporated to a large extent into the raiders' colony as full members.

Hölldobler analyzed a total of 34 raids conducted by Myrmecocystus mimicus on conspecific neighboring nests in the field. These episodes constitute only about 8 percent of all tournament interactions actually observed. A total of nine raiding events was observed from beginning to end, enabling Hölldobler to make a fairly accurate count of the number of larvae, pupae, honeypots, and workers abducted into the raiders' nest. From these data he estimated that the raiding colony is at least ten times larger than the raided colony. Thus raiding seems to be primarily directed against younger, still developing colonies in the neighborhood. On the other hand, Hölldobler recently observed a successful raid against a mature colony that had diluted its defensive forces by being simultaneously aggressively engaged with three other neighboring colonies. In this case the victor colony captured approximately 1000 immatures.

The territorial logic of Myrmecocystus colonies. The adaptive significance of this peculiar territorial strategy is of interest. As already mentioned, one of the major food sources of Myrmecocystus is termites. The temporal and spatial distributions of the termite colonies is highly unpredictable. Since there is little point in defending an area that is unlikely to provide adequate food in a given time, Myrmecocystus does not establish fixed territorial borders around its entire foraging range. Instead it defends only areas into which it is currently conducting intensive forays. This procedure obviously enables it to extend its foraging range considerably and leads to frequent incursions into potential foraging ranges of neighboring Myrmecocystus colonies. Since there are no well-established territorial borders, aggressive mass confrontations with conspecific competitors are much more common in Myrmecocystus mimicus than in Oecophylla or Pogonomyrmex. If these confrontations were as violent as physical combat in the latter species, they would result in a constant and heavy drain on the worker force. Thus, the tournaments of fighting display seem to be the most economical strategy in defense of the temporal territorial borders. Only when one colony is considerably weaker does it risk being overrun by the stronger colony, losing its queen, and thus being enslaved (Hölldobler, 1976c, 1981a).

The territorial logic of the Myrmecocystus colonies can be further analyzed with the aid of a geometric model. We consider a group of neighboring colonies located in an area A in which resources occur patchily in space and time in an essentially stochastic manner. A honeypot ant colony pursuing a rigid monopolization strategy would maintain exclusive use of an area a around the nest. If the cost of monopolizing a territory of area a costs the colony C(a) per unit time, then the total cost of defense for a given interval T is C(a)T.

During T, a number of resource patches (fractions of termite nests) will appear in the area A, patterned in a spatial distribution. In the simplest case, this distribution is random and each point in A is just as likely to receive a resource patch as is any other point. Let R be the total number of resource patches which, on the average, appear in A during the period T. Then the average density of resource patches is r = R/A patches per unit area, and the probability that a honeypot ant colony will receive n of these on its area is

P(n) = e-ar(ar)n/n! (1)

For clarity let each patch return benefit B0 to the colony. During T the colony acquires on the average a total of ar resource patches on its territory and thus harvests an average benefit B = B0ar. As resources become scarcer, r decreases and the size of territory required to harvest a given benefit B increases.

If the colony requires a minimum return B* on its territorial investment during the period T, then the fixed territory of area a is economically defensible on the average if

rB0 - B*/a > C(a)T/a 	(2)

The surplus yield per unit area must exceed the costs of defense per unit area during T. For environments with scarce resources, r will be small, and these conditions are unlikely to be met.

The expected return B characterizes the average benefit accrued over a very large number of repeats of the period T. For any single period T, the appearance of resource patches in a fixed territory is a sampling from P(n) and will lie close to B each period only if P(n) peaks sharply around its mean ar. For P(n) with the Poisson structure (Eq. 1), however, the variance s2 is as large as the mean itself

s2 = ar 	(3)

and the harvestable return to the colony can vary significantly from period to period. Similar conclusions follow for related distributions. In such an environment, holding fixed territories is a high-risk strategy in the sense that variance on return is of the same magnitude as the return itself. For many periods T, there will be either a resource surplus on the territory, which can exceed the colony's handling capacity, or a substantial resource deficit, which will make the defense effort a heavy drain. If the colony can handle surplus and ride out hard times (which is evidently possible due to the honeypot caste), the effects of this short-term stochasticity can be smoothed out somewhat and net benefit B predicted by Eq. 1 would be more closely realized. Economic problems of defensibility would then again be addressed by Eq. 2. But a colony locked into a fixed territory is at the mercy of the fluctuations in the resource. A colony struck by chance with a long chain of “hard time” periods would be eliminated if it is unable to search further afield for resources.

A general rule of thumb is that animals faced with chains of deficits and surpluses on fixed territories should dissolve such boundaries and let foraging ranges overlap. We have seen that this constraint applies in part for honeypot ants. Furthermore, for territorial risk-prone colonies there is a basic economic formula that will determine whether fixed territories pay off in the long run. For scarce, quickly exhausted resources, this appears unlikely. The alternative to fixed territories is a foraging system with floating, temporary territorial boundaries, as has been observed in honeypot ants. Although the colony will lose a fraction of termite clusters in its region to competitors, it will gain others in the region of the competitor colony. Since costs of defense are now much reduced, and even further reduced by the ritualized fighting display strategy, floating boundaries and overlapping ranges become a preferred option.

Intercolony communication and assessment. The dynamics of tournament interactions indicate that a kind of dominance order can exist among neighboring colonies. For a period of 3 weeks in July, when the rainy season had already started and the foraging activity of Myrmecocystus mimicus was very high, the interactions among three neighboring nests were observed. In the sequence depicted in Figure 10-19 (I), nest 3 was very active and interfered with nest 1 and nest 5, foraging in the immediate vicinity of these two nests. Several days later (II) nest 3 was still engaging nest 5 in a tournament directly at nest 5, but the tournament between nest 1 and nest 3 had shifted toward nest 3, and nest 1 was now actively foraging. Seven days later (III) nest 1 and nest 3 did not show any outside activity, but nest 5 was highly active for the first time since the three nests were first monitored. The pattern suggests that nest 3 was the most dominant colony, and frequently interfered aggressively with the foraging activity of its conspecific neighboring colonies. Nest 5 seemed to be the weakest of the three colonies, because its foraging activity was most frequently suppressed by neighboring colonies. This is a representative example of what Hölldobler observed many times in his study area during the peak foraging period. Although on occasion one colony previously dominant over another was later suppressed by the same colony (still later the situation could be reversed yet again), in most cases the direction of domination remained constant throughout the active season.

During tournaments scouting ants repeatedly return to their nest, where they recruit reinforcement of 100 or more workers to the tournament area. Usually the recruitment of a new army of major workers by one colony is countered by a similar action by the opposing party. The relative size of the opposing forces appears to reflect the relative size of the two colonies. If the colonies are markedly different in size, so that one has a significant majority of the ants engaged in tournament interactions, the tournament is shifted more and more toward the nest of the weaker colony. When one of the opposing colonies is still an incipient colony and therefore unable to summon a large defending worker force, the tournament ends quickly and the weaker colony is raided by the mature colony.

These field data suggest that the tournaments of Myrmecocystus provide the colonies with a means of assessing one another's strength. Lumsden and Hölldobler (1983) postulated that numerous threat displays between individual workers are integrated into a massive group display between opposing colonies. In parallel to solitary animals, the groups' “strategic decision” whether to retreat, to recruit, to continue to fight by display or convention, or to launch an escalated attack depends on information about the strength of the opposing colony, and this information is somehow obtained during the ritualized combats at the tournament site. During the tournament contest the ants walk on stilt-like legs while raising the gaster and head. When two hostile workers meet, they initially turn to confront each other head on. Subsequently they engage in a more prolonged lateral display, during which they raise the gaster even higher and bend it toward the opponent (Figure 10-20a,b). Simultaneously they drum intensively with their antennae on and around each other's abdomen, and frequently kick with their legs against the opponent. This is almost the only physical contact, although each ant seems to push sideways as if it were trying to dislodge the other. After several seconds one of the ants usually yields and the encounter ends. The ants continue to move on stilt-like legs. They soon meet other opponents, and the whole procedure is repeated. In a tournament situation encounters with nestmates last only one to two seconds and are terminated by a brief jerking movement of the body. The exchanges usually do not develop into a lateral display. One other important feature that appears to be important during the displaying tournament is the size of the individual ants. If a large and a small ant are matched in a displaying encounter, usually the smaller ant yields. Displaying ants not only walk with their legs in a stilt-like manner while simultaneously raising their gaster and heads, but in addition they sometimes appear to inflate their gasters, so that the tergites are raised and the whole gaster appears to be considerably larger. There is also a tendency for the tournament ants to mount little stones and pebbles and display down to their opponents (Figure 10-20c). In fact, the behavioral analysis of the displaying patterns suggest that during individual encounters the contestants appear to gauge each others' size, and that there is a tendency among the ants to bluff, that is, they pretend to be larger than they really are.

From these observations Lumsden and Hölldobler (1983) developed two models of ways in which Myrmecocystus might assess one another's strength during the tournaments. Individual workers could “count heads” during the encounters to gain a rough measure of the enemy's strength. Alternatively, individuals might determine whether a low or high percentage of the opponents are major workers and use it to estimate the opposing colonies' strength, since a high percentage is a reliable index of large colony size (Figure 10-20d). Hölldobler showed that among tournamenting ants majors are more frequently represented than among groups of foragers. Among colonies reared in the laboratory from founding queens, those younger than four years have a disproportionately small group of majors in the worker population.

Recent experiments indicate that both assessment mechanisms are involved in intercolony communication, and the data suggest that in particular small, immature colonies rely on the “caste polling” technique, which enables them to quickly assess whether or not the opponent is a mature colony. When confronted with large workers small colonies immediately retreat into the nest and close the nest entrance. This tactic enables small colonies to prevent larger ones from mounting a raid (Hölldobler, unpublished data).

Finally, still another method of assessment might be queuing: a long wait in the workers before meeting an unengaged opponent means a small colony, while a short wait means a large colony. However it is accomplished, whether by head counting, assessment of major percentages, or queuing, such colony-to-colony communication amounts to a kind of “negotiation” short of violent warfare. It entails behavior no more complex than that already demonstrated for the assessment and regulation of mass communication of many ant species during recruitment to food finds and new nest sites.

In addition to intercolony communication by means of tournament interactions, there is another, more subtle aspect of intercolony communication. Often no major antagonistic interactions occur for weeks between neighboring Myrmecocystus colonies. However, what Hölldobler has called guard contingents are positioned daily at the same spots. The contingents consist of a few workers (usually not more than a dozen) that stand on stilt-like legs, often posing on top of little stones. Neighboring colonies also dispatch guard contingents to these posts, where they conduct “mini-tournaments” with the foreign guards. These mini-tournaments represent demonstrations of temporarily stabilized territorial borders. However, if the number of workers of one party at the guard-post suddenly increases, the other party responds by recruiting a large counterforce, and a full tournament often ensues.

Ritualized territorial behavior in other ant species. Ritualized aggression of the Myrmecocystus type or at least approaching it may occur more widely among ants than previously suspected. It occurs for example in Prenolepis imparis (Hölldobler and Traniello, unpublished observation). The display might have the same ecological function as in Myrmecocystus, because as Talbot (1943a) first noted, Prenolepis imparis defend spatiotemporal territories. Le Moli and his collaborators report that fighting among workers of different colonies of Formica rufa never leads to overt aggression (Le Moli and Parmigiani, 1982; Le Moli et al., 1982). These authors state that alien workers instead confront one another by lifting their bodies up on stilt-like legs and bringing the abdomen forward between the posterior legs in what appears to be a ritualized version of formic acid spraying. A similar behavior has been observed as a threat display in Camponotus americanus (Carlin and Hölldobler, 1983; Figure 10-21). Actual spraying of the acid during such exchanges has rarely been noticed. According to Le Moli and Mori (1986), lack of overt fighting with injury and death among colonies of the same species appears to be a common trait of the members of the Formica rufa group, including Formica rufa, Formica cunicularia, and Formica lugubris. In contrast, workers belonging to different species typically respond violently to one another and often fight to the death. However, such generalizations must be made very cautiously, because seasonal intraspecific ant wars with many deaths have been reported in Formica polyctena (Mabelis, 1979b) and Formica lugubris (Cherix and Gris, 1980). Furthermore, Pisarski (1982) demonstrated that the degree aggressivity in Formica exsecta depends on the social organization of the colony.

According to Ettershank and Ettershank (1982), a clear-cut form of ritualized territorial behavior leading to tournaments of the Myrmecocystus type occurs in the Australian meat ant Iridomyrmex purpureus. Ants from different colonies perform a series of acts more or less in sequence: they gape at one another with their mandibles, raise their bodies in stilt-like posture, “box” with their front legs, grab at one another with their mandibles, stand side by side, circle around one another while presenting the tips of their abdomens, and kick with their hind legs. The workers do not actually fight. Rather, they either part at some point of the display or else appease their opponents by grooming them. The winner in the exchanges remains in the stilt-like posture, and the loser lowers its body while often leaning away (Figure 10-22). A similar ritualized territorial display has been observed in Iridomyrmex pruinosum, a common species in southwestern North America and Mexico. Members of separated nest populations usually do not engage in open physical aggression but instead perform a stereotyped display. This behavior consists of a rapid shaking motion as the ants confront each other, followed by a brief lateral display during which both ants hold their gasters raised up with the tip pointed towards the head of the other (Hölldobler, 1982c).

Workers of both Formica and Iridomyrmex use their distinctive behaviors to defend absolute territories, rather than spatiotemporal territories as in the case of the Myrmecocystus honeypot ants. In fact, Greenslade (1975a,b) and Greenslade and Halliday (1983) have demonstrated that the territories of meat ants are quite stable. Nevertheless, real tournaments occur in Iridomyrmex purpureus. Several hundred ants occasionally display at the same time for up to several days, during which some workers return to the same spot repeatedly. In one case observed by the Ettershanks the larger colony pressed steadily toward the weaker colony and eventually destroyed it. The myrmecocide--to coin a term equivalent to genocide--was similar to that common in Myrmecocystus.

Another form of ritualized aggressive interactions with appeasement behavior might be intercolony trophallaxis. In fact, Ettershank and Ettershank (1982) point out that the origin of the agonistic display of the meat ant “is seemingly the more benign interaction of solicitation of food. The initial contact, in which each ant tries to get the other to fold its mandibles and submit is very similar to the begging of a solicitor from a donor. The ultimate appeasement gesture of the loser lowering the body resembles the attitude adopted by the donor in feeding; similarly stilting is an exaggeration of the superior posture adopted by a successful solicitor so that it can feed from the droplet.” Intercolonial and interspecific food exchange during aggressive confrontations staged in a laboratory arena have been observed among conspecific alien workers, and even among workers belonging to different genera and subfamilies (Gösswald and Kloft, 1963; Bhatkar, 1979a,b, 1983; Kloft, 1983, 1987). This behavior has also been interpreted as a form of appeasement behavior regulating territorial interactions among neighboring colonies. However, these observations are based primarily on laboratory experiments and the full understanding of its functional significance has to await more extensive field studies.

Territory, predation, and true slavery
In Chapter 7 we introduced the idea of parsimony in the evolution of ant communication. Species have often combined the same or very similar elements of behavior in different ways to create new messages. They have also enlarged their repertories by ritualization, in which behavioral acts that are functional in one context are modified for use as signals in another context. A similar evolutionary parsimony exists in the relation between territorial aggression and other major categories of behavior, particularly predation and parasitism.

Many ant species are known to carry enemies killed in territorial battles back to their nests for use as food. This is true for at least a few within-species conflicts, as in the cannibal wars of Formica polyctena (Mabelis, 1979b, 1984; Driessen et al., 1984). It is more commonly the case for between-species aggression. For example, the fire ant Solenopsis invicta consumes Pheidole colonies that it displaces (Wilson, 1976b). The weaver ant treats alien workers of conspecific colonies and of other species like prey to be killed and carried back to the nest, but the effect as territorial defense is just the same (Hölldobler, 1983; Figure 10-23). Formica polyctena invades the nests of Formica fusca and species of Lasius and Myrmica and carries the inhabitants back as prey. In the process they largely eliminate these species for distances of up to tens of meters from their nest entrances, an effect that could equally well be called overutilization of prey or territorial elimination of competitors. It is easy to see how specialized predators of ants might originate by the territorial route. Such specialists are the rule in the army ant genera Aenictus, Eciton, and Neivamyrmex, as well as in the ponerine tribe Cerapachyini. They also occur in derived species of Gnamptogenys and Myopias, which are ponerine genera whose members prey primarily on other kinds of arthropods.

Territorial raids also appear to have led to slavemaking in a few species. The colony conquests of Myrmecocystus mimicus, in which worker and brood of alien colonies are captured and adopted as nestmates, is true slavery--the colonies use the labor of captives belonging to the same species (Hölldobler, 1981a). In addition, Myrmecocystus utilizes all injured enemy ants as prey, especially the honeypots. The largest proportion of the pillaged honeypots do not survive the raid and are eaten by the victors. Thus, a successful raid is a true bonanza for the victorious colony. It not only gains a large labor force, without having to spend energy and time to raise the workers, but it also obtains highly valuable food, by robbing and preying on the living storage containers of the victim colony. A closely similar phenomenon occurs in the curvispinosus group of Leptothorax, colonies of which attack related species during territorial expansion. The workers consume most of the larvae and pupae they capture but allow some to mature and remain as nestmates (Wilson, 1975a; Alloway, 1980; Stuart and Alloway, 1982). In Finland, Rosengren and Pamilo (1983) observed a case in which workers from a Formica polyctena colony raided a smaller colony of the same species and carried away its young, semireplete workers used in food storage (Speichertiere). They killed the queens but did not harm the workers, which evidently were incorporated as slaves.

Other techniques of competition and exploitation
In competition for food sources and nest sites, ants employ a rich repertory of interference techniques, of which territoriality is only the most commonly observed form. Many ant species monopolize food discoveries by the mass recruitment of foragers that use powerful chemical repellents. Their competitors often counter these actions by employing chemical assaults of their own or relying on the rapid cooperative retrieval of prey. Still other ant species invade the immediate nest area of their competitors and use threats and chemical weapons to prevent them from foraging. Let us examine some of the better studied examples in these several categories.

The European thief ant Solenopsis fugax usually lives close to colonies of other species and preys on their brood. Scouting workers of the Solenopsis build an elaborate tunnel system leading into the neighbor's brood chambers. As soon as the construction of the tunnels is completed, the scouts lay chemical trails back to their own nest and recruit masses of nestmates who then mount an invasion. While invading the brood chambers and preying on the brood, the Solenopsis workers discharge a highly effective and long-lasting repellent from the poison gland. The principal component of this secretion is trans-2-butyl-5-heptyl-pyrroldine. The substance prevents the brood-tending ants from defending their own larvae and enables the Solenopsis to rob brood virtually unopposed (Hölldobler, 1973a; Blum et al., 1980).

Similar chemical techniques are used in interference competition in Solenopsis (Baroni Urbani and Kannowski, 1974; Hölldobler et al., 1978; see Figure 10-24a), Monomorium (Hölldobler, 1973a; Adams and Traniello, 1981; Jones et al., 1988), Meranoplus (Hölldobler, unpublished; see Figure 10-24b) and Iridomyrmex (Hölldobler, 1982c). The last case shows how effective the methods can be. In the Arizona desert, the dolichoderine ant Iridomyrmex pruinosum and the formicines Myrmecocystus depilis and Myrmecocystus mimicus overlap widely in their nesting and food habits. Although the individual Iridomyrmex workers are only one-fourth the length of Myrmecocystus workers, they usually succeed in displacing the larger ants from food bait. They also block the nest entrance of their competitors, preventing them from leaving the nest altogether. They are able to accomplish this feat by swiftly channeling large numbers of workers to food sources and nest entrances of the Myrmecocystus, using a pheromonal mass communication system, while chemically repelling their competitors with secretions from the pygidial gland (Hölldobler, 1982c; Figure 10-25).

A bizarre variation of the nest-blocking technique is used by workers of the dolichoderine Conomyrma bicolor, which interferes with sympatric species of Myrmecocystus. The Conomyrma workers surround the nests of their competitors, pick up pebbles and other small objects in their mandibles, and drop then down the nest entrances (Möglich and Alpert, 1979; see Figure 10-26). Although it is not known exactly how the stone-dropping alters the behavior of the workers in the target nests, the effect is to reduce their foraging by a significant amount. A closely similar behavior was observed by Gordon (1988b) in Aphaenogaster cockerelli, which blocks the nest entrances of Pogonomyrmex barbatus and thereby shortens the daily foraging period of this harvesting species by one to three hours.

Oster and Wilson (1978) inferred from elementary mathematical models that as prey size increases, so does the probability of interference competition. The prediction was confirmed in field studies on Monomorium by Adams and Traniello (1981) and on Lasius by Traniello (1983). In the deserts of the southwestern United States, the species of Novomessor partially solve this problem by their large size, which enables single foragers to retrieve large prey items without delay (Hölldobler et al., 1978). However, when the items are very large the Aphaenogaster are still in danger of losing them to colonies of Iridomyrmex, Monomorium, and Solenopsis, which can assemble large forces of workers by mass communication. The Aphaenogaster foragers do not fight back very effectively, at least not with chemical weapons of their own. Instead, they use a highly efficient system of recruitment and retrieval. When a lone scout finds a prey object too large for it to carry, it discharges a poison-gland pheromone into the air as a short-range recruitment signal. Other scouts are soon attracted from as far away as two meters. The effect is enhanced by vibrational signals that lead to a more efficient group retrieval of the prey item (Markl and Hölldobler, 1978). Whenever possible the prey is not dissected on the spot but lifted up by the gang of workers and carried immediately back to the nest. In this fashion, Aphaenogaster greatly reduces the number of confrontations with its more formidable, mass-communicating competitors.

A very widespread form of competitive exploitation in ants is food robbing or “cleptobiosis.” Although the phenomenon has been described by many authors over the past century (Wroughton, 1892; Forel, 1901; Wheeler, 1910a, 1936c; Weber, 1946; Abe, 1971; Maschwitz and Mühlenberg, 1973) relatively little attention has been given to its ecological significance. A particularly interesting case involving desert ants in the southwestern United States was recently analyzed in this regard by Hölldobler (1986a). Harvester ants of the genus Pogonomyrmex, which harvest mostly seeds and other vegetable material, at first seem unlikely competitors for honeypot ants of the genus Myrmecocystus, which are predators and scavengers that also collect floral nectar and honeydew from homopterans. However, Pogonomyrmex workers also occasionally retrieve dead insects, and they are especially prone to join other ant species in preying on termites when these insects appear on the surface of the desert soil in large numbers following rains.

Workers of Myrmecocystus mimicus waylay returning foragers of several Pogonomyrmex species and rob them of insect prey, especially termites. As the Pogonomyrmex workers return to the nest, they are often stopped by the Myrmecocystus workers and thoroughly inspected. In a typical sequence, a Myrmecocystus climbs on the harvester ant's back and nibbles its head, mandibles, and mouthparts. Sometimes two or even three Myrmecocystus gang up to harass one Pogonomyrmex worker in this manner. If the harvester ant does not carry insect prey, it normally stands still and tolerates the inspections, which last only a few seconds. But if it has a termite or any other insect prey in its mandibles, the Myrmecocystus workers attack ferociously, vigorously yanking the prey away from the Pogonomyrmex. At most, the Pogonomyrmex responds to its tormentors with open mandibles and an occasional forward lunge in an apparent attempt to bite the Myrmecocystus (Figure 10-27). The Myrmecocystus are easily able to avoid the attack by swift, evasive movements of their own. In a survey of 76 nests belonging to Pogonomyrmex, Hölldobler found between 5 and 39 percent are affected by the Myrmecocystus according to species. Thus it seems likely that food robbing has a substantial influence on the food intake of both the robbers and their victims.

Finally, it is possible for species to invade a nearly closed community of species by heavy immigration of propagules. Either a beachhead is finally attained and the population is able to sustain itself, or the permanent supply of immigrants maintains the presence of the species in spite of rapid turnover. A general mathematical model of the process has been provided by Loeschcke (1985). Documentation of the phenomenon in ants has been provided by Rosengren and Pamilo (1983), who call it the “Austerlitz effect,” after the 1805 battle in which Napoleon's forces vigorously attacked superior forces on both sides of the Pratzen Plateau. Formica truncorum, which is a “beachland specialist,” assails habitats surrounding its established colonies with numerous inseminated queens evidently capable of long-distance dispersal. It also spreads rapidly by encroachment into vacant nesting sites by portions of the colony that emigrate over the ground. The result is a patchy distribution of the species in the midst of the more nearly continuous distribution of Formica aquilonia and Formica polyctena populations.