The Ants Chapter 16

CHAPTER 16. THE ARMY ANTS

No spectacle of the tropical world is more exciting and mystifying than that of a colony of army ants on the march. In his book, Ants, Their Structure, Development and Behavior, Wheeler expressed its poetry in the following way: "The driver and legionary ants are the Huns and Tartars of the insect world.  Their vast armies of blind but exquisitely coöperating and highly polymorphic workers filled with an insatiable carnivorous appetite and a longing for perennial migrations, accompanied by a motley host of weird myrmecophilous camp-followers and concealing the nuptials of their strange, fertile castes, and the rearing of their young, in the inaccessible penetralia of the soil--all suggest to the observer who first comes upon these insects in some tropical thicket, the existence of a subtle, relentless and uncanny agency, directing and permeating all their activities."

The swarm raiders
Eciton burchellii is one of the best understood of the army ants. This big, conspicuous species is abundant in humid lowland forests from Brazil and Peru north to southern Mexico (Borgmeier, 1955). Its marauding workers, together with those of other species of Eciton, are well known to native peoples by such local names as padicours, tuocas, tepeguas, and soldados. In English they are called army ants, as well as foraging ants, legionary ants, soldier ants, and visiting ants. These insects have understandably been a prime target for study by naturalists for a long time, from Lund (1831) through Bates, Belt, von Ihering, Müller, and Sumichrast in the last century to Beebe, Wheeler, and many others in more recent times. But it was T. C. Schneirla (1933-1971) who, by conducting patient studies over virtually his entire career, first unraveled the complex behavior and life cycle of this and other species of Eciton. His results were confirmed and greatly extended in rich studies conducted by C. W. Rettenmeyer (1963a).

A day in the life of an Eciton burchellii colony seen through the eyes of Schneirla and Rettenmeyer begins at dawn, as the first light suffuses the heavily shaded forest floor. At this moment the colony is in “bivouac,” meaning that it is temporarily camped in a more or less exposed position. The sites most favored for bivouacs are the spaces between the buttresses of forest trees and beneath fallen tree trunks (see Figure 16-1 and Plate 20) or any sheltered spot along the trunks and main branches of standing trees to a height of twenty meters or more above the ground. Most of the shelter for the queen and immature forms is provided by the bodies of the workers themselves. As they gather to form the bivouac, they link their legs and bodies together with their strong tarsal claws (Figure 16-2), forming chains and nets of their own bodies that accumulate layer upon interlocking layer until finally the entire worker force comprises a solid cylindrical or ellipsoidal mass up to a meter across. For this reason Schneirla and others have spoken of the ant swarm itself as the “bivouac.” Between 150,000 and 700,000 workers are present. Toward the center of the mass are found thousands of immature forms, a single mother queen, and, for a brief interval in the dry season, a thousand or so males and several virgin queens. The entire dark-brown conglomerate exudes a musky, somewhat fetid odor.

When the light level around the ants exceeds about 0.5 lux, the bivouac begins to dissolve. The chains and clusters break up and tumble down into a churning mass on the ground. As the pressure builds, the mass flows outward in all directions. Then a raiding column emerges along the path of least resistance and grows away from the bivouac at a rate of up to 20 m an hour. No leaders take command of the raiding column. Instead, workers finding themselves in the van press forward for a few centimeters and then wheel back into the throng behind them, to be supplanted immediately by others who extend the march a little farther. As the workers run on to new ground, they lay down small quantities of chemical trail substances from the tips of their abdomens, originating in the hindgut and probably also in the pygidial gland (Hölldobler and Engel, 1978), guiding others forward. Workers encountering prey lay extra recruitment trails that draw nestmates differentially in that direction (Chadab and Rettenmeyer, 1975). A loose organization emerges in the columns, based on behavioral differences among the castes. The smaller and medium-sized workers race along the chemical trails and extend it at the point, while the larger, clumsier soldiers, unable to keep a secure footing among their nestmates, travel for the most part on either side. The location of the Eciton soldiers misled early observers into concluding that they are the leaders. As Thomas Belt put it, “Here and there one of the light-colored officers moves backwards and forwards directing the columns.” Actually the soldiers, with their large heads and exceptionally long, sickle-shaped mandibles, have relatively little control over their nestmates and serve instead almost exclusively as a defense force. The minimas and medias, bearing shorter, clamp-shaped mandibles, are the generalists. They capture and transport the prey, choose the bivouac sites, and care for the brood and queen. Workers often work in teams, with large medias serving as porters. These specialists initiate the transport of large prey items and are joined by workers of equal or smaller size. The teams accomplish their task with greater energetic efficiency than if they cut the prey into small pieces and carried them individually (Franks, 1986; see Figure 8-48).

Eciton burchellii has an unusual mode of hunting even for an army ant. It is a “swarm raider,” which means that the foraging workers spread out into a fan-shaped swarm with a broad front. Most other army ant species are “column raiders,” pressing outward along narrow dendritic odor trails in the pattern exemplified in Figure 16-3. Schneirla (1956b) has described a typical raid as follows:

For an Eciton burchellii raid nearing the height of its development in swarming, picture a rectangular body of 15 meters or more in width and 1 to 2 meters in depth, made up of many tens of thousands of scurrying reddish-black individuals, which as a mass manages to move broadside ahead in a fairly direct path. When it starts to develop at dawn, the foray at first has no particular direction, but in the course of time one section acquires a direction through a more rapid advance of its members and soon drains in the other radial expansions. Thereafter this growing mass holds its initial direction in an approximate manner through the pressure of ants arriving in rear columns from the direction of the bivouac. The steady advance in a principal direction, usually with not more than 15° deviation to either side, indicates a considerable degree of internal organization, notwithstanding the chaos and confusion that seem to prevail within the advancing mass. But organization does exist, indicated not only by the maintenance of a general direction but also by the occurrence of flanking movements of limited scope, alternately to right and left, at intervals of 5 to 20 minutes depending on the size of the swarm.

The huge sorties of burchellii in particular bring disaster to practically all animal life that lies in their path and fails to escape. Their normal bag includes tarantulas, scorpions, beetles, roaches, grasshoppers, and the adults and broods of other ants and many forest insects; few evade the dragnet. I have seen snakes, lizards, and nestling birds killed on various occasions; undoubtedly a larger vertebrate which, because of injury or for some other reason, could not run off, would be killed by stinging or asphyxiation. But lacking a cutting or shearing edge on their mandibles, unlike their African relatives the “driver ants,” these tropical American swarmers cannot tear down their occasional vertebrate victims. Arthropods, such as ticks, escape through their excitatory secretions, stick insects through repellent chemicals, as tests show, as well as through tonic immobility. The swarmers react to movement in particular as well as to the scent of their booty, and a motionless insect has some chance of escaping them. Common exceptions, which may enjoy almost a community invulnerability in many cases, include termites and Azteca ants in their bulb nests in trees, army ants of their own and other species both on raiding parties and in their bivouacs, and leaf-cutter ants in the larger mound communities; in various ways these often manage to fight off or somehow repel the swarmers.

The approach of the massive burchellii attack is heralded by three types of sound effect from very different sources. There is a kind of foundation noise from the rattling and rustling of leaves and vegetation as the ants seethe along and a screen of agitated small life is flushed out. This fuses with related sounds such as an irregular staccato produced in the random movements of jumping insects knocking against leaves and wood. This noise, more or less continuous, beats on the ears of an observer until it acquires a distinctive meaning almost as the collective death rattle of the countless victims. When this composite sound is muffled after a rain, as the swarm moves through soaked and heavily dripping vegetation, there is an uncanny effect of inappropriate silence.

Another characteristic accompaniment of the swarm raid is the loud and variable buzzing of the scattered crowd of flies of various species, some types hovering, circling, or darting just ahead of the advancing fringe of the swarm, others over the swarm itself or over the fan of columns behind. To the general hum are added irregular short notes of higher pitch as individuals or small groups of flies swoop down suddenly here or there upon some probable victim of the ants which has suddenly burst into view. . . No part of the prosaic clatter, but impressive solo effects, are the occasional calls of antbirds. One first catches from a distance the beautiful crescendo of the bicolored antbird, then closer to the scene of action the characteristic low twittering notes of the antwren and other common frequenters of the raid.

If you wish to find a colony of swarm raiders in Central or South America, the quickest way is to walk quietly through a tropical forest in the middle of the morning, listening. For long intervals the only birds you might hear are in the distance and mostly in the canopy. Then, as Johnson (1954) has expressed it, comes a “chirring, twittering, and piping” of antbirds close to the ground. Mingled in is the murmur or hissing caused by the frantic movements of countless insects trying to escape the raiders, and the buzzing of parasitic flies. Very soon you will see the ants themselves marching in a broad front, hundreds of thousands streaming forward as though drawn toward some goal just out of sight in the forest shadows. Also present may be ithomiine butterflies, which fly over the leading edge of the swarm. First noticed by Drummond (1976), the butterflies are now known to feed on the droppings of the antbirds (Ray and Andrews, 1980; Andrews, 1983).

On Barro Colorado Island, Panama, where Schneirla conducted most of his studies, the antbirds normally follow only the raids of Eciton burchellii and those of another common swarm-raider, Labidus praedator. They pay no attention to the less conspicuous forays of Eciton hamatum, Eciton dulcium, Eciton vagans, and other column-raiding army ants. There are at least ten species of antbirds on Barro Colorado Island, all members of the family Formicariidae. They feed principally on the insects and other arthropods flushed by the approaching burchellii swarms (Johnson, 1954; Willis, 1967). Although a specimen of Neomorphus geoffroyi has been recorded with its stomach stuffed with burchellii workers, most species appear to avoid the ants completely or at most consume them by accident while swallowing other food.

As one might anticipate from these accounts, the burchellii colonies and their efficient camp followers have a profound effect on the faunas of those particular parts of the forest over which the swarms pass. E. C. Williams (1941), for example, noted a sharp depletion of the arthropods at spots on the forest floor where a swarm had struck the previous day. On Barro Colorado Island, which has an area of approximately 17 km2, there exist only about 50 burchellii colonies at any one time. Since each colony travels at most 100 to 200 m every day (and not at all on about half the days), the collective population of burchellii colonies raids only a minute fraction of the island's surface in the course of one day, or even in the course of one week. But the strikes are probably frequent enough during the course of months to have a significant effect on the composition and age structure of the colonies of ants and social wasps.

The food supply is quickly and drastically reduced in the immediate vicinity of each colony. Early writers, especially Müller (1886) and Vosseler (1905), jumped to the reasonable conclusion that army ant colonies change their bivouac sites whenever the surrounding food supply is exhausted. At an early stage of his work, however, Schneirla (1933b, 1938) discovered that the emigrations are subject to an endogenous, precisely rhythmic control unconnected to the immediate food supply. He proceeded to demonstrate that each Eciton colony alternates between a statary phase, in which it remains at the same bivouac site for as long as two to three weeks, and a nomadic phase, in which it moves to a new bivouac site at the close of each day, also for a period of two to three weeks. (The nomadic phase is better called the migratory phase, since army ants are migratory hunters rather than nomads in the strict sense. That is, they move periodically to areas of fresh prey, rather than guide herds to fresh pastures in the manner of true nomads.  True nomadism is known among ants only in certain Malaysian species of Hypoclinea [see Chapter 13].)  The basic Eciton cycle is summarized in Figure 16-4. Its key feature is the correlation between the reproductive cycle, in which broods of workers are reared in periodic batches, and the behavior cycle, consisting of the alternation of the statary and migratory phases. The single most important feature of Eciton biology to bear in mind in trying to grasp this rather complex relation is the remarkable degree to which development is synchronized within each successive brood. The ovaries of the queen begin developing rapidly when the colony enters the statary phase, and within a week her abdomen is greatly swollen by 55,000 to 66,000 eggs (Figure 16-5). Then, in a burst of prodigious labor lasting for several days in the middle of the statary period, the queen lays from 100,000 to 300,000 eggs. By the end of the third and final week of the statary period, larvae hatch, again all within a few days of each other. A few days later the “callow” workers (so called because they are at first weak and lightly pigmented) emerge from the cocoons. The sudden appearance of tens of thousands of new adult workers has a galvanic effect on their older sisters. The general level of activity increases, the size and intensity of the swarm raids grow, and the colony starts emigrating at the end of each day's marauding. In short, the colony enters the migratory phase. The migratory phase itself continues as long as the brood initiated during the previous statary period remains in the larval stage. As soon as the larvae pupate, however, the intensity of the raids diminishes, the emigrations cease, and the colony (by definition) passes into the next statary phase.

The emigration is a dramatic event requiring sudden complex behavioral changes on the part of all adult members of the Eciton colony. At dusk or slightly before workers stop carrying food into the old bivouac and start carrying it, along with their own larvae, in an outward direction to some new bivouac site along the pheromone-impregnated trails (Figure 16-6). Eventually, usually after most of the larvae have been transported to the site, the queen herself makes the journey. This event usually transpires between 8:00 and 10:00 p.m., well after nightfall. Just before the queen emerges from the bivouac, the workers on the trail nearby become distinctly more excited, and the column of running workers thickens beyond its usual width of 2 to 3 cm, soon widening to as much as 15 cm. Suddenly the queen appears in the thickest part. As she runs along she is crowded in by the “retinue,” a shifting mob consisting of an unusual number of soldiers and darkly colored, unladen smaller workers. The members of the retinue jostle her, press in underfoot, climb up on her back, and at times literally envelop her body in a solid mass. But, even with this encumbrance, the queen moves along easily to the new bivouac site. She is guided by the odor trail and can follow it all by herself even if the surrounding workers are taken away. After passage the emigration tapers off, and it is usually finished by midnight. If the column is disturbed near the queen, she stops and is swiftly covered by a blanket of protecting workers. All New World army ants employ retinues during emigrations ready to react this way. The largest are formed by Eciton burchellii and other species that travel aboveground and hence are most exposed to predators (Rettenmeyer et al., 1978).

The activity cycle of Eciton colonies is truly endogenous. It is not linked to any known astronomical rhythm or weather event. It continues at an even tempo month after month, in both wet and dry seasons throughout the entire year. Propelled by the daily emigrations of the migratory phase, the colony drifts perpetually back and forth over the forest floor (Figure 16-7). The results of experiments performed by Schneirla indicate that the phases of the activity cycle are determined by the stages of development of the brood and their effect on worker behavior. When he deprived Eciton colonies in the early migratory phase of their callow workers, they lapsed into the relatively lethargic state characteristic of the statary phase, and emigrations ceased. Migratory behavior was not resumed until the larvae present at the start of the experiments had grown much larger and more active. In order to test further the role of larvae in the activation of the workers, Schneirla divided colony fragments into two parts of equal size, one part with larvae and the other without. Those workers left with larvae showed much greater continuous activity.

These results, while provocative, are not decisive and at best solve only half the problem. For if the activity cycle is controlled by the reproductive cycle, what controls the reproductive cycle? The logical place to look would seem to be the queen. By her astonishing capacity to lay all of her eggs in one brief burst, she creates the synchronization of brood development, which is the essential feature for the colonial control of the activity cycle. At first Schneirla (1944) concluded that this reproductive effort by the queen is the “pace-maker,” thus implying that the queen herself is the seat of an endogenous rhythm. Later, however, Schneirla (1949b, 1956b) modified his hypothesis by viewing the queen and her colony as reciprocally donating elements in an oscillating system:

When each successive brood approaches larval maturity, the social-stimulative effect upon workers nears its peak. The workers thus energize and carry out some of the greatest raids in the nomadic phase, with their by-product larger and larger quantities of booty in the bivouac. But our histological studies show that, at the same time, more and more of the larvae (the largest first of all) soon reduce their feeding to zero as they begin to spin their cocoons. Thus in the last few days of each nomadic phase a food surplus inevitably arises. At this time the queen apparently begins to feed voraciously. It is probable that the queen does not overfeed automatically in the presence of plenty, but that she is started and maintained in the process by an augmented stimulation from the greatly enlivened worker population. Within the last few days of each nomadic phase, the queen's gaster begins to swell increasingly, first of all from a recrudescence of the fat bodies, then from an accelerating maturation of eggs. The overfeeding evidently continues into the statary phase, when, with colony food consumption greatly reduced after enclosure of the brood, smaller raids evidently bring in sufficient food to support the processes until the queen becomes maximally physogastric. These occurrences, which are regular and precise events in every Eciton colony, are adequate to prepare the queen for the massive egg-laying operation which begins about one week after the nomadic phase has ended.

While this interpretation makes a pretty story, it is constructed with fragments of very circumstantial evidence. The crucial question is unanswered as to whether the queen really is stimulated to feed in excess by the greater abundance of food or at least by the higher intensity of worker activity associated with the food, as Schneirla posited, or whether her increased feeding is timed by some other, undetermined physiological event. Since work on Eciton physiology is still virtually nonexistent, and experimental evidence of any kind very sparse, one can do no more than reflect on these possibilities as competing hypotheses.

Another question of considerable interest, added to the inducement of queen oogenesis, is the stimulus that triggers the onset of the migratory phase. According to Schneirla's theory of brood stimulation (actually a hypothesis), the migratory phase is initiated when workers become excited by the near-simultaneous eclosion of new, callow workers from the pupae. Migration is sustained thereafter by stimulation from the growing larvae. As illustrated in Figure 16-4, the mass eclosion at the start of the migratory phase coincides with the hatching of the egg mass. We may then ask which event, adult eclosion from the pupa or egg hatching, triggers the migratory phase? In an ingenious experiment conducted on Neivamyrmex nigrescens, Topoff et al. (1980) removed the larval brood of an early migratory colony and replaced it with the pupal brood of an early statary colony. As a result, the pupae eclosed well before the next batch of eggs laid by the host colony hatched, with the result that for several days the host colony was occupied by newly eclosed, callow workers but no larvae. It commenced its next migratory phase nonetheless, demonstrating that newly emerged adults are sufficient by themselves to drive this segment of the army-ant cycle.

Very little is known concerning the actual communicative stimuli that mediate the activity cycles. In his voluminous theoretical writings on the subject, Schneirla often spoke of “trophallaxis” as the driving force of the cycles of army ants, but it is clear that he meant this term to be virtually synonymous with “communication” in the broadest sense. Apparently he had no clear ideas about the nature of the signals utilized. In earlier articles he attributed much of the stimulative effect of the larvae to their “squirming”; later he stressed the probable existence of pheromones as well. But these speculations were based almost entirely on observations of the more obvious outward signs of communication, a level of study usually inadequate to distinguish even the sensory modalities employed in communication with insect colonies and unlikely to identify the signals employed. Lappano (1958) discovered that the labial glands of burchellii larvae become fully functional on the eighth or ninth migratory day, about the time raiding activity reaches its peak. She concluded that the labial glands are “probably” producing a pheromone that excites the worker. But again, the only evidence available is the stated coincidence in time of the two events. Our lack of knowledge of the semiotic basis of the Eciton cycle is due simply to the lack of any serious attempt to obtain it. This interesting subject does not seem likely to resist sustained experimental study; any such effort in the future is likely to yield exceptionally interesting results.

In his classic writings Schneirla was inadvertently misled by his failure to distinguish consistently between the ultimate causation and the proximate causation of the army-ant cycle. It is possible and even likely that the adaptive value, hence the ultimate causation, is the additional food made available to the colony when it emigrates frequently. However, biological systems often evolve so as to rely on endogenous rhythms to make the needed changes, rather than on a close reading of the environment from day to day. Put another way, the flush of callow workers becomes the token signal to the workers to initiate daily emigrations. They are the proximate cause of the emigrations, but the ultimate cause--the advantage emigrations give to emigration-prone genotypes in the ant population--remains the improved food supply. Although Schneirla occasionally mentioned that food availability might have been an important factor in the evolution of emigrations (1944, 1957b), the idea played no important role in his theoretical interpretation. After he had demonstrated the endogenous nature of the cycle, and its control by synchronous brood development, he dismissed the role of food depletion. The emigrations, he repeatedly asserted, are caused by the appearance of callow workers and the older larvae; they are not caused by food shortage. He meant only that food shortages are not the proximate cause.

In his culminating synthesis, however, Schneirla (1971) provided a more balanced view of evolution and physiological mechanisms. Toward the end of his life, he also learned that food shortages are in fact among the proximate controls, at least to a minor degree in some army ant species. In his last field study, on the small Asian army ant Aenictus, he discovered that short-term variation in colony activity depends on the “alimentary condition prevalent in the brood” (Schneirla and Reyes, 1969). To be specific, the ants appear to emigrate more often when their food supply runs low. Topoff and Mirenda (1980a,b) later tested the impact of food supply on emigrations in Neivamyrmex nigrescens by a series of experiments. This was made possible by their feat of culturing army ant colonies in the laboratory long enough to follow the brood cycle under controlled conditions. Two colonies overfed with prey emigrated on only 28 percent of the days during the migratory phase of the cycle, while two underfed colonies emigrated on 62 percent of the migratory days. It is now clear that although Neivamyrmex colonies follow an endogenous cycle of roughly the Eciton type (see Table 16-2), the tempo of their migratory phase is strongly affected by the amount of food they harvest. The complex interactions mediating raiding and emigrations in ecitonine army ants have been reviewed by Topoff (1984).

To summarize very briefly to this point, the colonies of Eciton and at least some other army ant species follow an endogenous cycle as Schneirla described it. The migratory phase is triggered at least in part by the emergence of new, callow adult workers. It is quite possible that the cycle was put in place genetically and is kept there through natural selection by the advantage it gives in overcoming food shortages. The frequency of raids is also fine-tuned in at least one species (Neivamyrmex nigrescens) by the day-to-day availability of food.

Colony multiplication in Eciton, first elucidated by Schneirla and R. Z. Brown (1950, 1952), is a highly specialized and ponderous operation. Through most of the year the mother queen is the paramount center of attraction for the workers, even when she is in competition with the mature worker larvae toward the close of each migratory phase. By serving as the focal point of the aggregating workers, she literally holds the colony together. The situation changes drastically, however, when the annual sexual brood is produced early in the dry season. This kind of brood contains no workers, but, in Eciton hamatum at least, it consists of about 1,500 males and 6 new queens. Even when the sexual larvae are still very young, a large fraction of the worker force becomes affiliated with the brood as opposed to the mother queen. By the time the larvae are nearly mature, the bivouac can be found to consist of two approximately equal zones: a brood-free zone containing the queen and her affiliated workers, and a zone in which the rest of the workers hold the sexual brood. The colony has not yet split in any overt manner, but important behavioral differences between the two sections do exist. For example, if the queen is removed for a few hours at a time, she is readily accepted back into the brood-free zone from which she originated, but she is rejected by workers belonging to the other zone. Also, there is evidence that workers from the queen zone cannibalize brood from the other zone when they contact them.

The young queens are the first members of the sexual brood to emerge from the cocoons. The workers cluster excitedly over them, paying closest attention to the first one or two to appear (see Figure 16-8). Several days later the new adult males emerge from their cocoons. This event energizes the colony, sets off a maximum raid followed by emigration, and at last splits the bivouac. Raids are conducted along two radial trails from the old bivouac site. As they intensify during the day, the young queens and their nuclei of workers move out along one of the trails, while the old queen with her nucleus moves out along the other. When the derivative swarm begins to cluster at the new bivouac site, only one of the virgin queens is able to make the journey to it. The others are held back by the clinging and clustering of small groups of workers. They are, to use Schneirla's expression, “sealed off” from the rest of the daughter colony. Eventually they are abandoned and left to die. Now there exist two colonies: one containing the old queen; the other, the successful virgin, daughter queen. In a minority of cases the old queen is also superseded. That is, the old queen herself falls victim to the sealing-off operation, leaving both of the two daughter colonies with new virgin queens. This presumably happens most often when the health and attractive power of the old queen begin to fail prior to colony fission. The maximum age of the Eciton queen is not known, but is believed to be relatively great for an insect; a marked queen of Eciton burchellii, for example, was recovered by Rettenmeyer after a period of four and a half years. The males, in contrast, enjoy only one to three weeks of adult existence. Shortly after their emergence they depart on flights away from the home bivouac in search of other colonies. Their bodies are heavily laden with exocrine glands resembling those of the queens. The new queens are fecundated within a few days of their emergence, and almost all of the males disappear within three weeks after that. Rettenmeyer (1963a) has described an actual mating, and he has presented evidence that a queen sometimes mates more than once in her lifetime and may even mate annually. Two other matings were observed by Schneirla (1971) after the ants had been removed for laboratory observation. The copulations lasted two and ten hours respectively.

Recently Franks and Hölldobler (1987) proposed that in army ants an unusual form of sexual selection exists, in which the workers play a major role in selecting their sister's mate. Outbreeding appears to be the rule in army ants; Schneirla (1971) has suggested that Eciton males must fly before they can mate. Thus, males have to enter alien colonies and break the worker barrier to get to the females. Under these circumstances workers are in a position to choose both the mother and father of their future nestmates. Franks and Hölldobler argued that the reason why workers should be involved in choosing mates for their queen is that they will later invest in the progeny of these males. Therefore the principle of sexual selection and female choice should apply to their preferences. In order to maximize their own inclusive fitness, the workers should actively choose the males that are most fertile. Sexual selection theory suggests that males may demonstrate that they will be donors of highly viable gene combinations, for example by being themselves large, robust and vigorous. It is possible that worker involvement in sexual selection has favored males that are superficially similar in size and shape to their conspecific queen. Such is in fact the case in the polyphyletic lineages of army ants. The unusual robustness of army ant males is well known. In Africa they are called sausage flies, and elsewhere they are commonly mistaken for wasps. It is possible that during evolutionary history, the gaster of the male became elongated and enlarged to house the massive sperm vesicles (Gotwald and Burdette, 1981) together with the ever increasing glandular equipment needed to impress the workers. Recently, Hölldobler and Engel-Siegel (1982) discovered that ecitonine males are unusually well endowed with abdominal glands, especially between the tergites of the gaster. Such glands, which do not occur in workers, closely resemble those of the queen (Whelden, 1963). Fierce sexual competition is made likely by the highly skewed numerical sex ratio in army ants. Where only a few males succeed, the resulting selection has probably led to the evolution of the rich glandular equipment in males, because only the most attractive males were successful. In fact males have retinues of workers that accompany them like the entourage of queens.

There is one further feature of the army ant syndrome that can be explained in terms of this unusual form of sexual selection and sexual competition. In many army ant species males lose their wings upon entering an alien colony (Figure 16-9). Dealate males have been found in the Old World army ants, including Dorylus rubellus (Savage, 1847, cited in Gotwald, 1982), as well as in the New World Labidus praedator (Rettenmeyer, 1963a) and species of Eciton (Schneirla, 1971).

The phylogeny of the doryline and ecitonine army ants
The genus Eciton represents one of the furthest extensions of an evolutionary trend that began independently within each of many different groups of ants. According to Gotwald (1982), the “true” army ants of the subfamilies Dorylinae and Ecitoninae are triphyletic, or triple-origined. Specifically, the combination of army-ant behavior and the dichthadiigyne queen has arisen separately in the genera Aenictus and Dorylus, which compose the subfamily Dorylinae, and a third time in Eciton and its relatives (Cheliomyrmex, Labidus, Neivamyrmex, and Nomamyrmex), which compose the subfamily Ecitoninae. The division between the two subfamilies is supported by a number of strong morphological differences, including even the Dufour's gland epithelium (Billen, 1985a).

On the basis of a phenetic study of Dorylus, Barr et al. (1985) recognized three groupings within that genus: Dorylus (including the “subgenera” Anomma, Dichthadia, and Typhlopone), Alaopone, and Rhogmus. These authors (and Gotwald, 1985) took the conservative position of placing all of the species under the single genus Dorylus, noting that “whether to recognize these groupings at the generic or subgeneric level must await the broader consideration which will be accorded the group in a complete revision.” The higher taxonomic status of Aenictus is in similar limbo. If the genus is truly independent, having arisen out of a stock entirely separate from those of Dorylus and the ecitonines, it should logically be placed either in the same subfamily as the ancestral stock or accorded a subfamily of its own. But again, this decision awaits a full analysis of the Dorylinae and candidate ancestral groups, particularly the Ponerinae. Meanwhile, Gotwald has provided sound biogeographic data to support his belief (1979) that Dorylus arose in Africa and spread into Tropical Asia, while Aenictus arose in Asia and spread to Africa and Australia.

In the New World, Cheliomyrmex is the phylogenetically most interesting genus. It is distinguished by a single waist segment, as opposed to the two waist segments possessed by all the other New World army ants. (Among Old World forms, Dorylus also has a single segment.) Cheliomyrmex is otherwise anatomically close to the remainder of the New World army ants (Gotwald, 1971). Wheeler, and later Gotwald and Kupiec (1975) and Gotwald (1978), considered the single-jointed waist primitive, and thus Cheliomyrmex itself primitive. Cheliomyrmex workers also have a median tooth on the tarsal claws and a simple pygidium (as opposed to a longitudinally impressed pygidium with terminal spines). Both of these traits are probably primitive (see Figure 16-10). However, Cheliomyrmex remains a rarely seen and enigmatic genus and needs to be examined in much more detail before its phylogenetic position can be ascertained with confidence. Its biology and life cycle might also shed light on its position within the Ecitoninae.

The phylogenetic origins of the three army ant groups (Aenictini, Dorylini, Ecitonini plus Cheliomyrmecini) are more generally very poorly understood. A favorite view, originating with Emery (1895) and entertained more recently by Brown (1975) and Gotwald (1982), is that the army ants arose from the tribe Cerapachyini of the subfamily Ponerinae. One of the cerapachyine traits favorable to this hypothesis is the widespread habit of raiding other species of ants. But again, both the anatomy and the behavior of the cerapachyines could be merely convergent to those of the doryline and ecitonine army ants.

The time of the origin of the three doryline and ecitonine lines is also unknown. A single species, Neivamyrmex ectopus, has been described from the Dominican Republic amber, which is late Oligocene or early Miocene in age (Wilson, 1985d). The two type workers are quite modern in aspect and closest in overall anatomy to species now living in Mexico and the southern United States than to the much richer Central and South American faunas. It is of special interest that army ants are today completely absent from the Dominican Republic and the remainder of the Greater Antilles. In other words they have retreated from this part of the world since no later than Oligocene times.

What is an army ant?
Behavior patterns characteristic of the doryline and ecitonine army ants also occur, at least to a limited extent, in the subfamilies Leptanillinae, Ponerinae, and Myrmicinae. In fact, the best definition of the expression “army ant” may well be a functional one, of the sort offered in the second edition of Webster's New International Dictionary: Any species of ant that goes out in search of food in companies, particularly the driver and legionary ants. It should be added here, in order to clarify the vernacular nomenclature, that there has been a tendency by recent authors to use the terms “army ant” and “legionary ant” interchangeably. According to the same source, the term “legionary ant” refers first to the New World army ants of the genus Eciton, and second to the “driver ants” of the genus Anomma. Most English-speaking authors of the past fifty years have designated Eciton and its relatives as legionary ants, while employing “army ant” to cover all of the dorylines and ecitonines. We prefer to use the two terms synonymously in the broader functional sense, as one of us recommended earlier (Wilson, 1971).

Actually, the definition just quoted is incomplete. Upon closer examination of the subject one finds that there are really two discrete features that can be considered fundamental in army ant (legionary) behavior. These diagnostic features have been distinguished under the concepts of migration and group-predation (Wilson, 1958e). Migration, or nomadism as Schneirla and Wilson called it, is defined as relatively frequent colony emigration. Most, if not all, ant species shift their nest site if the environment of the nest area becomes unfavorable, and some, for example the dolichoderine Tapinoma sessile, are exceptionally restless and may emigrate many times during the course of a single season. But none has been found that undertakes emigration so frequently or accomplishes it in such an orderly fashion to cover so much new territory as do the species of Eciton and the other, better-known dorylines. Group predation includes group raiding and, usually, group transport in the process of hunting living prey. These two processes must be carefully distinguished from each other since they involve quite different innate behavior patterns and are not invariably linked. Many ant species engage in the group transport of prey, meaning that two or more workers carry a single prey item back to the nest in a cooperative and efficient manner. Members of at least 27 genera representing most of the subfamilies use group transport (see Chapter 10). A few ponerines that specialize on termites also group-raid, including a few Leptogenys, as well as Onychomyrmex, a highly modified member of the ponerine tribe Amblyoponini found in tropical Australian forests, which group-raids to capture large arthropod prey (Figure 16-11). Other group-raiders include the tiny ants of the genus Leptanilla (subfamily Leptanillinae) and species of Pheidologeton (subfamily Myrmicinae) that swarm out to harvest a wide range of prey. Some species of the ponerine tribe Cerapachyini, which feed primarily or exclusively on other ants, may group-raid (Wilson, 1958e). However, recent laboratory studies of Cerapachys indicate that at least some of these ants employ conventional solitary scouting and group recruitment to organize their raids (Hölldobler, 1982b). The same is true of at least Megaponera foetans (Longhurst et al., 1979a), Pachycondyla (= Termitopone) laevigata (Hölldobler and Traniello, 1980b), and at least some Leptogenys species (Maschwitz and Mühlenberg, 1975; Maschwitz and Schönegge, 1983). The doryline and ecitonine “true” army ants display an extreme form of group foraging and raiding in which individual workers move exclusively in groups, venturing to run only short distances beyond the advancing edge before they turn back in. Their timidity is so strong that when masses of workers are dumped onto a clean flat surface or are cut off from the rest of the colony by rain, they commence “circular milling.” In this bizarre formation, workers go forward and inward with the crowd but not outward in a centrifugal direction, so that the whole mass continues to circle round and round until all the ants are dead (Figure 16-12).

How much are migratory behavior and group-predation associated in ants? Only a very incomplete answer can be supplied to this question since we are still handicapped by a scarcity of information on emigration of any sort in most ant groups. Yet it can at least be established that the association of frequent emigration and group raiding, constituting the most general characteristics of “army ant” behavior, does exist in groups other than the Dorylinae and Ecitoninae. We have listed the known cases in Table 16-1. An additional characteristic of most of these species is the independent evolutionary development of the peculiar queen form known as the dichthadiigyne. As exemplified in Figure 16-12, the dichthadiigyne is a permanently wingless form with greatly reduced eyes, massive pedicel and abdomen, and strong legs. Its aberrant morphology clearly contributes to two of its adaptations to migratory life: its capacity to deliver large quantities of eggs during a short span of time, and its ability to run under its own power from one bivouac site to another.

The driver ants
The African driver ants of the genus Dorylus differ from the ecitonines in several important details of the activity cycle, which are apparently caused by peculiarities in the queen caste (Raignier and van Boven, 1955). The queens are the largest of all ants (Figure 16-13). Those of Dorylus vary from 39 to 50 mm or more in total length and possess as many as 15,000 ovarioles capable of delivering 1 to 2 million eggs in a month. The abdomen is in a permanent state of moderate physogastry, and the queen lays eggs more or less continuously. Most of the eggs, however, are produced in bursts that come in approximately three-week intervals and last five or six days. The ensuing development of the larval brood appears to have little effect on the inducement of emigration. In fact, larvae are usually outnumbered by pupae in the brood of emigrating colonies. No clock-like alternation between statary and migratory phases of the Eciton type is displayed by the colonies of Dorylus. The emigrations, which take several days to complete, are separated by statary periods that vary in duration from six days to two or three months. The bivouacs of Dorylus are also much more stable affairs than those of Eciton and most other army ants. The colonies settle deeply into the soil at the end of the emigration, excavating labyrinthine systems of galleries and chambers to a depth of 1 to 4 m. The colonies each contain millions of workers, as many as 22 million in the case of Dorylus wilverthi, according to Raignier and van Boven. The colonies are surprisingly numerous in parts of Africa, considering their huge individual size. Leroux (1977) calculated that Dorylus nigricans colonies occur at a density of about one colony per 10 hectares in the Guinea savanna of the Ivory Coast and 3 colonies per 10 hectares in the nearby forest.

From these secure nests the Anomma send forth almost daily raids. The swarm pattern, illustrated in Figure 16-14, unfolds like a great pseudopodium. It engulves all of the ground and low vegetation in its path, and then, after a few hours, drains back to the bivouac site. The advance is leaderless. The excited workers rush back and forth at an average speed of 4 cm per second. Those in the van press forward for a short distance and then retreat back into the mass to give way to new advance runners. The columns resemble thick black ropes lying along the ground. A close examination shows them to be dozens or hundreds of workers wide. The ants are so dense that they pile on top of one another and run along on one another's backs, while some spill away from the column and form scattered crowds to either side, their antennae and mandibles pointed upward in threatening postures (Figure 16-15). The frontal swarm, which contains up to several millions of workers, advances at a rate of about 20 m per hour. In his field notes Gotwald (1984-85) described a march in Gabon as follows: “The advancing swarm of worker ants moved with the effortlessness of a rain-swollen river. It flowed across the forest floor with singleness of purpose, altered in its course only by the most obtrusive of natural barriers. . . As the swarm progressed over the forest litter, small stationary groups of workers formed, giving the moving mass of foragers the appearance of an island-choked delta.”

The ants sweep almost all forms of animal life before them, killing insects and larger creatures too sluggish to get out of the way. Thomas Savage's famous account of 1847 expresses the drama of the hunt:

They will soon kill the largest animal if confined. They attack lizards, guanas, snakes, etc. with complete success. We have lost several animals by them,--monkeys, pigs, fowls, etc. The severity of their bite, increased to great intensity by vast numbers, it is impossible to conceive. We may easily believe that it would prove fatal to almost any animal in confinement. They have been known to destroy the Python natalensis, our largest serpent. When gorged with prey it lies powerless for days; then, monster as it is, it easily becomes their victim. . . Their entrance into a house is soon known by the simultaneous and universal movement of rats, mice, lizards, Blapsidae, Blattidae and of the numerous vermin that infest our dwellings. Not being agreed, they cannot dwell together, which modifies in a good measure the severity of the Driver's habits, and renders their visits sometimes (though very seldom in my view) desirable. Their ascent into our beds we sometimes prevent by placing the feet of the bedsteads into a basin of vinegar, or some other uncongenial fluid; this will generally be successful if the rooms are ceiled, or the floors overhead tight, otherwise they will drop down upon us, bringing along with them their noxious prey in the very act of contending for victory. They move over the house with a good degree of order unless disturbed, occasionally spreading abroad, ransacking one point after another, till, either having found something desirable, they collect upon it, when they can be destroyed “en masse” by hot water; or, disappointed, they abandon the premises as a barren spot, and seek some other more promising for exploration. When they are fairly in we give up the house, and try to wait with patience their pleasure, thankful, indeed, if permitted to remain within the narrow limits of our beds or chairs. They are decidedly carnivorous in their propensities. Fresh meat of all kinds is their favourite food; fresh oils they also love, especially that of the Elais guiniensis, either in the fruit or expressed. Under my observation they pass by milk, sugar, and pastry of all kinds, also salt meat; the latter, when boiled, they have eaten, but not with the zest of fresh. It is an incorrect statement, often made, that “they devour everything eatable” by us in our houses; there are many articles which form an exception. If a heap of rubbish comes within their route, they invariably explore it when larvae and insects of all orders may be seen borne off in triumph,--especially the former.

The dominance of the driver ants in sub-Saharan Africa has earned them special names in different African cultures: siafu, ensanafu, kelelalu, bashikouay, and nkran, among others. Their offensive power lies not in their stings, which are rarely if ever inserted, but in the powerful bite and shearing action of their mandibles. Dorylus was given its vernacular name by Savage, who wrote in 1847 that the ant “drives everything before it capable of muscular motion, so formidable is it from its numbers and bite. . and, in distinction from other species of this country, may well take for its vulgar name that of Driver.” According to the nineteenth-century explorer Paul Du Chaillu, criminals were exposed to Dorylus swarms as a cruel form of execution. While traveling in Ghana, Gotwald (1984-85) was told of an incident in which driver ants killed a baby left beneath a tree while its mother tilled the family garden.

Yet driver ants are not really the terror of the jungle as popularly conceived. Although the colony is an “animal” weighing in excess of 20 kilograms and possessing on the order of 20 million mouths, its raiders move over the ground at the rate of only a meter every three minutes. It is possible to watch the whole process at close range while seated comfortably in a camp chair (which of course must be periodically moved--carefully, and in the right direction!).

Although most Dorylus species have very broad dietary habits, at least a few are specialized predators. Two colonies of Dorylus gerstaeckeri studied by Gotwald (1974) fed only on earthworms. According to Leroux (1979), Dorylus of the “subgenus Typhlopone,” specifically either Dorylus badius or Dorylus fulvus, regularly raid colonies of Dorylus nigricans.

Dorylus and the other Dorylinae and Ecitoninae have triumphed as legionary ants over all their competitors. An idea of their diversity can be gained from the synopsis presented in Table 16-2. They not only outnumber other kinds of legionary ants in both species and colonies, but they tend to exclude them altogether. Cerapachyines, for example, are relatively scarce throughout the continental tropics wherever dorylines and ecitonines abound, but they are much more common in remote places not yet reached by these advanced army ants--for example, Madagascar, Fiji, New Caledonia, and most of Australia.

Army ants outside the Dorylinae and Ecitoninae
Because the dorylines and ecitonines are all fairly specialized in anatomy and social behavior, the best strategy for tracing the origins of migratory huntresses is the study of other ant taxa in early and intermediate stages of its evolution. Many of the species are less accessible by virtue of being rare or exotic, but important information has nevertheless accumulated steadily during the past thirty years.

For example, members of the little subfamily Leptanillinae have long been considered army ants. But they have remained an enigma, because they are very rare and local. Most of the 33 species are known only from one or a few workers or males (Baroni Urbani, 1977). Neither of the present authors has ever seen one alive, in spite of years of field work in all parts of the tropics. W. L. Brown, probably the most widely traveled and productive ant collector of all time, has found only one colony during more than forty years. The belief that leptanillines have legionary habits has come chiefly from their anatomy. The workers are tiny simulacra of dorylines or ecitonines in overall aspect, while the queens are typical dichthadiigynes remarkably convergent to those of the Dorylinae and Ecitoninae (see Figure 16-16). Like ecitonine army ant queens their gasters are richly endowed with intersegmental exocrine glands, believed to produce attractants for the workers (Hölldobler et al., in preparation). Recently a breakthrough was achieved by Masuko (1987), who succeeded in collecting eleven colonies of Leptanilla japonica in the broadleafed forest at Cape Manazuru, Japan. The colonies each contained about 100 workers and were strictly subterranean--a feature that perhaps explains the scarcity of leptanillines in collections. The ants are evidently specialized predators of geophilomorph centipedes, each of which is many times the size of a worker. The foragers follow trunk trails out from the nest, but it is not clear whether they move singly or in groups in the typical army-ant manner. The workers possess a large median sternal gland at the seventh abdominal sternite. Although no experimental proof has yet been attempted, it seems probable that the gland is the source of the trail pheromone (Hölldobler et al., in preparation). The brood cycle is tightly synchronized in a long annual cycle: eggs, young larvae, and pupae are present in the summer, while mature larvae appear during the winter and spring (the larvae pupate by summer).

In the laboratory, Leptanilla japonica colonies promptly emigrate whenever they are disturbed. The swiftness of their response suggests that changes in nest site often occur in nature, perhaps approaching the army ant pattern. The transport of larvae is highly evolved and stylized. The workers do not use their mandibles to hold the larvae in the manner of other ants. Instead, they use their lower mouthparts to grip a peculiar appendage projecting from the larval prothorax. Leptanilla larvae also possess a unique exudatory organ on each side of the third abdominal segment that provides hemolymph as a nutrient to the adults, especially the queens.

The Leptanilla japonica annual cycle has been interpreted by Masuko in the following way. While larvae are present, the workers intensively hunt centipedes (Figure 16-17a). The larvae consume these victims and grow quickly. During this period the queen's abdomen remains constricted. When the larvae mature, the colony passes into its oviposition phase. The queen feeds heavily on larval hemolymph from the abdominal exudatory organ and becomes physogastric (Figure 16-17b). Prey consumption ends, while the larvae pupate all at once and the queen lays the eggs of the next worker generation in a single batch.

The leptanilline ants are an especially promising group for the study of adaptive radiation of miniature group-predatory ants. Two new Asian genera have been recently identified by R. W. Taylor (personal communication). Protanilla wallacei has saber-shaped mandibles capable of being opened 180 degrees and a row of labral hairs that appear to serve as prey-seeking guides (Figure 16-18). Anomalomyrma kubotai has one of the most bizarre mandibles known in the ants: their blades are scoop-shaped and lined on the inner surface with thick, inward-directed pegs (Figure 16-19). The function of the latter armament is still unknown.

Findings of equal significance have recently been made by Maschwitz and his collaborators in the ponerine genus Leptogenys (Maschwitz, personal communication). They discovered several Malayan species that exhibit the complete army ant syndrome, including swarm raiding, temporary bivouacs and regular migrations to new hunting grounds. The brood cycle of the Leptogenys colonies is not synchronized, however. There were always all brood stages present in the colony, and the single ergatoid queen did not exhibit physogastric cycles. One species close to Leptogenys mutabilis (in the well-known processionalis group) was studied in considerable detail. The colonies each consist of more than 30,000 workers. During the day they stay in bivouacs, which are located in the soil or in the dense leaf litter. The foraging raids, often involving more than 20,000 ants, begin at dusk. They are not initiated by recruiting scouts or led by leader ants. A major raid fans out in one direction progressing approximately 5 m per hour. The average distance the fan front moves away from the bivouac site is 22 m (maximum 56 m). During one night a colony covers an area of approximately 300 m2. The swarm raiders hunt a broad spectrum of prey, including a variety of insects, spiders, chilopods, diplopods, earthworms, and flatworms. The colonies move to new hunting grounds on the average of every 1.5 days (range to 10 days) and travel an average distance of 28 m (maximum 50 m). Within the colony a host of myrmecophiles is found, including collembolans, flies, staphylinids, and isopods. These largely unstudied guests accompany the ants during colony emigrations (Figure 16-20).

Of equal novelty and interest are recent findings by Moffett (1984, 1986f, 1987a-c, 1988a,b) on the abundant Asian myrmicine Pheidologeton diversus. Colonies conduct swarm raids remarkably similar in some respects to those of Eciton burchellii and the Dorylus driver ants. The colonies are huge, with a single oversized queen and hundreds of thousands of workers. The foragers travel on one or two stable trunk trails that extend from 5 to over 100 meters from the nest. The trails sometimes last for weeks, with great numbers of ants traveling back and forth along them day and night. In this respect the Pheidologeton resemble many other kinds of ants that follow trunk trails to persistent food sources, such as the Messor and Pogonomyrmex that exploit seedfalls and the Formica and Lasius that visit clusters of honeydew-excreting aphids. But where the workers of these more conventional species wander away from the trails in large numbers to forage on their own, the Pheidologeton use the trails as departure points for column raids and occasionally swarm raids. As in other army ants bouts of solitary hunting are restricted to the front of the raid, and even these brief, forward sallies can be properly interpreted as part of a coordinated group effort. Workers otherwise almost never stray more than 5 centimeters away from the moving columns of ants. In a word, they forage very much like doryline and ecitonine army ants.

A Pheidologeton raid begins when ants move away as a group from the trunk trail. They form up into a narrow column that grows outward, like the pseudopod of an ameba, at the rate of 10-20 centimeters a minute. The explorers pay little attention to trails laid during previous raids, often traveling widely over previously unvisited terrain. After the column reaches between half a meter and 3 meters in length, some of the ants in the terminus spread out, and progress slows down. In a minority of the cases, however, the terminus blossoms out still further into a large, fan-shaped raid (see Figure 16-21). Behind the seething frontal edge of a full-blown raid, most of the ants run back and forth in a tapered network of feeder columns. These columns in turn funnel back into a single basal column, which lengthens as the swarm progresses.

Moffett found that the large raiding swarms each contain tens of thousands of individuals and reach as far as six meters from the trunk trails. They closely resemble the formations of the swarm-raiding Dorylus and Eciton but travel outward only one-fifth as fast or less (1.5-2 meters per hour as opposed to 10-20 meters per hour). The Pheidologeton, like Dorylus and Eciton, are able to conquer exceptionally large and formidable prey, up to and including frogs, by overwhelming them with the sheer force of numbers. They also carry large objects rapidly back to the nest with well coordinated group transport (see Figure 10-6). Because of these capabilities and the enormous size variation that occurs within the worker force of each colony (a 500-fold dry-weight ratio of largest to smallest workers), the ants are able to collect an impressive array of food (Figure 16-22 and Plate 21). While gangs are attacking earthworms, cockroaches, and other larger animals, smaller individuals are ferreting out tiny collembolans and flies. In addition, the Pheidologeton gather seeds and fruits of many kinds. In fact, usually about half of the material they carry to the nest is of vegetable origin. This complex foraging strategy has worked very well. Pheidologeton diversus and other, similarly behaving members of the genus are among the most abundant and ecologically dominant ants over a large part of tropical Asia.

The origin of legionary behavior
Wilson (1958e) argued that the key to understanding the origin of legionary behavior lies in the adaptive significance of group raiding. Earlier writers had stated repeatedly that compact armies of ants are more efficient at flushing and capturing prey than are assemblages of foragers acting independently. This observation is certainly correct, but it is not the whole story. There is another, primary function of group raiding that becomes clear only when the prey preferences of the group-raiding ants are compared with those of predatory ants that forage in solitary fashion. Most nonlegionary ponerine species for which the food habits are known take living prey of approximately the same size as their worker caste or smaller. As a rule they must depend on proportionately small animals that can be captured and retrieved by lone foraging workers. Group-raiding ants, on the other hand, feed on large arthropods or the brood of other social insects, prey not normally accessible to ants foraging solitarily. Thus, the species of Leptanilla, Onychomyrmex, and the Leptogenys diminuta group specialize on large arthropods; those of Eciton and Anomma prey on a wide variety of arthropods that include social wasps and other ants; species of Simopelta raid other ants; and Megaponera foetans raids termites. The doughty little ecitonine Neivamyrmex harrisii evidently specializes on fire ants of the genus Solenopsis (Mirenda et al., 1980).

With this generalization in mind, it is relatively easy to reconstruct the steps in evolution leading to the full-blown legionary behavior of the Dorylinae and Ecitoninae. The following scheme is modified from Wilson (1958e):

1. Group-recruitment raiding is developed to allow specialized feeding on other social insects. In this case group raids are initiated by successful scouts, which lead a raiding party to the discovered prey. This form of raiding is represented today in some Leptogenys species (Maschwitz and Mühlenberg, 1975; Maschwitz and Schönegge, 1983), Pachycondyla (Termitopone) laevigata (Hölldobler and Traniello, 1980b), at least one Cerapachys species (Hölldobler, 1982b), and also Megaponera foetens (Longhurst et al., 1979a).

2. Group raids are initiated more autonomously, without the stimulus of recruiting leaders. They are usually more massive than those raiding parties led by leader ants. This more advanced form of group raiding evidently developed to allow predation on large arthropods and other social insects and to cover a larger hunting area. Group raiding without frequent migrations occurs in Pheidologeton and possibly some of the Ponerinae. From his studies on Pheidologeton, Moffett (1988b) has suggested that the use of trunk trails in conjunction with recruitment was a forerunner of autonomous group raids.

3. Migratory behavior is either developed concurrently with group-raiding behavior, or it is added shortly afterward. The reason for this new adaptation is that large arthropods and social insects are more widely dispersed than other types of prey, and the group-predatory colony must constantly shift its trophophoric field to tap new food sources. With the acquisition of both group-raiding and migratory behavior, the species is now truly “legionary,” that is, an army ant in the full functional sense at least roughly equivalent to the Dorylinae and Ecitoninae. Some of the group-raiding ponerines, including the Malayan Leptogenys just described, have evidently reached this adaptive level. Colony size in these species averages larger than in related, nonlegionary species, but it does not approach that attained by Eciton and Anomma.

4. As group raiding becomes more efficient, large colony size becomes possible. This stage has been attained by many of the Dorylinae and Ecitoninae, including the species of Aenictus and Neivamyrmex and at least some of the column-raiding Eciton (dulcius, hamatum, mexicanum, vagans).

5. The diet may be expanded secondarily to include other smaller and nonsocial arthropods and even small vertebrates and vegetable matter; concurrently, the colony size becomes extremely large. This is the stage reached by the driver ants of Africa and tropical Asia (Dorylus), the species of Labidus, and Eciton burchellii, most or all of which also utilize the technique of swarm raiding as opposed to column raiding. It is also partially approached by species of the myrmicine genus Pheidologeton, particularly Pheidologeton diversus.

Two new twists have been added to the scheme of evolutionary grades proposed by Wilson. The discovery of swarm raiding in Pheidologeton diversus was a complete surprise. Besides being the first such case in the Myrmicinae, it shows that species can evolve to swarm raiding of the Eciton burchellii type without migratory behavior. The Pheidologeton evidently maintain themselves in stable foraging domains and nest sites by an extraordinarily broad diet, made possible by a heavy reliance on seeds and fruits and the ability to capture animal prey of almost all kinds and sizes.

A second complication is the correlation proposed by Gotwald (1978, 1982) between the soil zone in which legionary ants live and the degree of their prey specialization. In general, species that hunt underground or beneath the surface of rotting stumps and logs (hypogaeic foragers) are more specialized than those that hunt on the surface of the ground and vegetation (epigaeic foragers). Most legionary ant species are hypogaeic and hence relatively specialized. The majority of Aenictus for which data are available are predators of the immature stages of other ants. Hypogaeic Dorylus feed mostly on termites and other ants, and at least one or two species are predators of earthworms; and the species of Neivamyrmex, many of which are subterranean, specialize on ants. The known epigaeic Aenictus and Dorylus are generalized feeders, as are Pheidologeton diversus. If the correlation suggested by Gotwald holds, it is superimposed on the evolutionary grades defined by colony size, group raiding, and nomadism to account for part of the variance in degrees of prey specialization among species. However, the data are too fragmentary to be certain. We have little or no information on the great majority of species of dorylines and ecitonines, especially the less accessible hypogaeic species. The relation between diet, microhabitat, and social organization of the legionary ants is still a relatively unexplored and very promising subject. Rettenmeyer et al. (1983) have noted that as many as twenty ecitonine species are found together in some New World tropical forests. How, we may ask, can so many “huns and tartars” coexist in the same place?