Polygyny

The Ants

Queen Numbers and Domination

Introduction

The number of queens profoundly alters several of the key features of colonial organization, including the kinship of the nestmates, the rate of colony growth, and the number and distribution of nests. The past ten years have witnessed an explosive growth of knowledge about this complex subject, with investigators addressing the following questions:

• Why does the number of queens vary among species? The colonies of some species always have a single egg-laying queen. Those of others have up to thousands or, as in the case of the “supercolonial” Formica yessensis of Japan, millions of queens. In still other species the number ranges among colonies from one to many. In addition, the number often shifts through different stages of the colony life cycle.

• Which colony members control the number of queens? Is the number a consequence of dominance and elimination among queens, or do the workers regulate the number?

• How is the number controlled? Regulation of the queen population can be achieved by physical elimination, reproductive castration (perhaps mediated by pheromones), or emigration of supernumeraries.

• What are the consequences of varying queen numbers on the growth and genetic structure of colony populations?

It will be useful to begin with the generally accepted terminology of queen numbers in relation to the life cycle. Monogyny refers simply to the possession by a colony of a single egg-laying queen, or reproductive female--or "gyne," as it is often called in the myrmecological literature. Polygyny is the possession of multiple queens. Oligogyny is a special case of polygyny, in which two to several queens coexist in the same nest but remain well apart from one another (Hölldobler, 1962; Buschinger, 1974a). As a rule, oligogyny in ants is characterized by tolerance of workers toward supernumerary queens combined with intolerance among the queens, so that the queens space out in the same nest (Hölldobler and Carlin, 1985). The founding of a colony by a single queen is called haplometrosis; when multiple queens found a colony the condition is referred to as pleometrosis (Wasmann, 1910b; Wheeler, 1933b). The term metrosis can be used to refer generally to this biological variable. Monogyny can be primary, meaning that a single queen is also the foundress; or it can be secondary, in which multiple queens start a colony pleometrotically but only one of them survives. In a symmetric fashion, polygyny can be primary (multiple queens persist from a pleometrotic association) or secondary (the colony is started by a single queen and others are added later by adoption or fusion with other colonies).

True polygyny, in which two or more queens contribute to egg-laying, has been conclusively demonstrated in many genera in the Ponerinae, Myrmeciinae, Myrmicinae, Dolichoderinae, and Formicinae, both by direct observation of egg-laying and indirectly, through the electrophoretic identification of allozymes in the workers and second generation of reproductives (see for example Pamilo, 1982b, and Berkelheimer, 1984). Without such proof the mere presence of multiple queens in the same nest does not necessarily mean that the colony is polygynous. Queens that are winged are invariably virgin and not active egg-layers. These "alate" individuals are young and native to the nests in which they are found. They await the mating flight and, in most cases, the initiation of new colonies on their own. Even if supernumerary queens are wingless (called "dealate" if they have shed wings they earlier possessed), they still might not be egg-layers. In some species such individuals are inseminated but their oviposition is suppressed by the presence of a major egg-laying queen in the same nest. This form of reproductive latency, also called "functional monogyny," has been demonstrated in Formicoxenus and Leptothorax by Buschinger and his co-workers (1967, 1968c, 1979a; Buschinger and Winter, 1976; Buschinger et al., 1980b); and in Solenopsis invicta by Tschinkel and Howard (1978). Functional monogyny has also been reported in Myrmecina graminicola by Baroni Urbani (1968a, 1970). However, Buschinger (1970b) cautioned that these particular observations might have been due to laboratory manipulations, pointing out that functional monogyny has not been observed in colonies collected fresh from the field. In other species, for example Leptothorax acervorum, the dealate supernumeraries are unfertilized. Buschinger (1967) calls this condition “pseudopolygyny” to distinguish it from true, functional polygyny and latent polygyny. He notes that in Leptothorax, the infertile queens often act like workers by sharing ordinary labor in the nest. On the other hand, Ehrhardt (1970) found that dealated virgin queens of Formica polyctena often contribute to the male brood. Similar observations have been made in polygynous colonies of Solenopsis invicta (D. J. C. Fletcher, personal communication).

The origins of polygyny

Polygyny can arise by one or the other of three means: pleometrosis, with the multiple founding queens remaining together after the first workers appear; the adoption of extra inseminated queens after their nuptial flights; and the fusion of colonies. Different ant species have played upon these devices in various ways to produce a remarkable diversity of statistical patterns of queen numbers. In many species, the colonies employ them to create mixed strategies of colony founding and structure.

The most versatile species studied to date in this regard is the Australian meat ant Iridomyrmex purpureus (Hölldobler and Carlin, 1985). Most new colonies are founded by single queens, following the nuptial flight in the spring month of October. Sixty-five of 72 newly founded nests excavated near Canberra contained a single queen, six contained two queens, and one contained three queens. As the nearby inseminated queens scurry over the ground and then start to dig a burrow in the soil, they are attacked by many enemies. Greaves and Hughes (1974) reported losses of 80 percent or higher from ground-feeding birds alone. Many other queens are killed by hostile workers of their own species when they stray too near the established colonies. But, surprisingly, other queens succeed in digging their nest chambers in the immediate vicinity of mature Iridomyrmex purpureus. Not only are most of these females tolerated by the resident workers, they are often attended and protected. The workers even help them dig the chambers (see Figure 6-1). Such actions are likely to protect the foundresses from hostile ants, birds, and other predators. Meat ant workers are extremely aggressive, biting enemies and spraying them with poisonous secretions from their pygidial glands. Hölldobler and Carlin believe that the queens protected in this manner are the ones fortunate enough to settle near their natal nests, so that the foundresses are also likely to be absorbed into the mother colonies to become supernumerary queens. Thus variation in queen numbers appears to be due at least in part to the vicissitudes of foundress association and tolerance by neighboring mature colonies.

Variation in the number of queens, usually less complexly organized than in Iridomyrmex purpureus, is widespread but not universal in the ants. Among 46 nests of Aphaenogaster rudis excavated by Headley (1949), for example, no queens were found in 6 of the nests, a single queen was found in each of 38 nests, and two queens each were found in 2 nests. Talbot (1951) obtained similar proportions in the 71 additional nests of Aphaenogaster rudis. In 20 nests of Prenolepis imparis excavated in Missouri, Talbot (1943a) found no queens in 3 of the nests, a single queen in each of 15 nests, and two queens in each of 2 nests. In Florida, Tschinkel (1988c) found most colonies of P. imparis to be polygynous, with a range of 1-6 and a mean of 4 queens per nest. In Lasius flavus, colonies with more than one queen occur but are rare (Waloff, 1957). The same is true of harvester ants in the genus Pogonomyrmex; in one sample of 70 Pogonomyrmex rugosus incipient nests studied by Pollock and Rissing (1985), only four contained more than one queen. Mature colonies, however, contain only one queen (Hölldobler, unpublished observations). The level of polygyny can vary geographically within the same species. Some local populations of the fire ant Solenopsis invicta are primarily monogynous, others primarily polygynous (Tschinkel and Howard, 1978; Fletcher et al., 1980). Unfortunately, most species reported to be polygynous have not been studied carefully enough to establish that the queens are all fully fertile and inseminated.

Although precise data are not available for enough species to assess the Formicidae as a whole, it is our impression from a good deal of field experience that the mature colonies of the majority of species are strictly monogynous. It is reasonably supposed that properties in colony organization tend to bias species toward monogyny in the course of evolution, and that the tendency is reversed only when special ecological constraints are imposed on the species (Hölldobler and Wilson, 1977b; Oster and Wilson, 1978). Two such properties can be expected as a consequence of evolution by natural selection. First, queens of all kinds of social insects should prefer to retain personal reproductive rights and surrender none to their sisters or daughters, because they are more closely related to their daughters and sons than to their nieces, nephews, and grandoffspring. Furthermore, queens living in multiples are likely to contribute fewer offspring than if they were the sole egg-layers, an effect in fact documented experimentally in Leptothorax curvispinosus by Wilson (1974b) and in Plagiolepis pygmaea by Mercier et al. (1985a,b). Second, workers should prefer to have only one queen serving as the colony progenitrix. In species with colonies of small to moderate size, the rate of colony growth, and hence the amount of colony genetic fitness, is limited primarily by the number of workers and not by the number of queens, one queen usually being able to supply as many eggs as a worker force can rear. This effect has been demonstrated, for example, in the myrmicine genera Tetramorium (Brian et al., 1967), Myrmica (Elmes, 1973), and Leptothorax (Wilson, 1974b). Since extra queens would then be an unnecessary energetic burden on the colony, an especially significant factor during the colony's early growth, it should be of advantage to the workers to eliminate them. In short, one can reasonably expect the dominant queen and the workers to conspire to eliminate supernumerary queens during the early phase of colony growth and thus to attain a state of monogyny.

Under a wide range of conceivable conditions, independent colony foundation should be a second trait favored by natural selection. If entire colonies can be started by one or a few queens, mother colonies producing such females can deploy far more of them over greater distances than otherwise comparable colonies that reproduce by swarming. Each swarm drains off a part of the original worker force, and its dispersal range is limited by the difficulties inherent in mass orientation and mobility. Swarming is likely to be of advantage only if the survival rate of queens is overwhelmingly greater when they are accompanied by workers than when they proceed alone.

In addition, haplometrosis can be expected to be the preferred mode of independent colony foundation. Unless circumstances give a large advantage to founding in groups, each queen should attempt to start a colony well away from all possible rivals. The tendency just cited toward the restitution of monogyny by both the dominant queen and the first worker force means that a queen choosing to become a member of a group of n founding queens has a 1/n chance of surviving to be the nest queen.

Finally, colony foundation should be claustral whenever possible. The highest mortality of social insect workers occurs during foraging trips, and it is probable that the same is true of founding queens forced to leave their nests in order to search for food.

In sum, a first logical examination of the more abstract general properties of insect societies leads us to expect that natural selection will lead species to monogyny, haplometrosis, and claustral nest founding. Yet deviations from this expected pattern are many and exceedingly diverse. In the case of the fire ant Solenopsis invicta, for example, local strains of polygyne colonies have originated and flourished repeatedly in various parts of the United States since 1940 (Greenberg et al., 1985).

The adaptive significance of polygyny

An examination of the trends toward polygyny can, if examined on a broader theoretical context, shed light on the evolution of other aspects of social behavior. Let us begin with a comparison of two of the major groups of social Hymenoptera. Most ant species are monogynous and obligatorily haplometrotic. Within most major phyletic lines, polygyny and swarming appear to be evolutionarily derived conditions. Among the wasps, in contrast, only the temperate zone species of Vespa and Vespula are known to be obligatorily haplometrotic. Belonogaster, Mischocyttarus, and Polistes are sometimes haplo- and sometimes pleometrotic, while most or all of the Polybiini, containing 20 of the 26 known social wasp genera, are polygynous and reproduce by swarming (Evans and West-Eberhard, 1970; Spradbery, 1973). We suggest the following simple explanation for the difference. When an ant colony moves to a new nest site, for example following a disturbance by a predator, it must walk to the new site. The workers are wingless, and the queen must travel on the ground with them. Thus queens can afford to shed their wings following the nuptial flight and initiate claustral colony foundation. In all but a few species they histolyze their alary muscles to nurture the first worker brood within a completely closed cell. Since they never need to take flight again, the queens are able to take advantage of the surplus energy available in the alary muscles. Both wide dispersal and independent, claustral colony founding are within their reach. Since these are the optimum techniques under almost all conceivable conditions, most ant species have evolved to acquire them.

When wasp colonies are disrupted, they fly to a new nest site. Lengthy ground travel is not only unnecessary but would be disadvantageous for insects so fully adapted to life in the air. Because the queens must fly with them, they must "stay in shape" by not losing their flight muscles. The nest queens of Vespa lose the power of flight as they become older, presumably because of the weight of their ovaries, but this is the exception rather than the rule in social wasps. Thus wasp queens are less well equipped to be solitary foundresses. Since independently founding individuals find it disadvantageous to convert the alary muscles into energy for the brood, they must engage in the risky process of foraging for food. It appears to follow that wasp species should be more likely to rely on pleometrosis or even swarming, which is in fact the case.

Why, then, does polygyny occur at all in ants? It is evidently derived in evolution and has arisen repeatedly during the 100-million-year evolution of these insects, despite the fact that it has certain intrinsic disadvantages. When a phenomenon displays this kind of pattern, the biologist is justified in searching for unusual circumstances that have promoted its deviant form of evolution.

Endangered populations. Population extinction rates are likely to be highest in rare or locally distributed species. In ants and other social hymenopterans, in which the males are derived from unfertilized eggs, the effective size of the population of colonies (Ñ in the parlance of population genetics; see Li, 1955) is

[(4.5Nt)/Nc](MQ/1+2M)

where

Nt is the total number of adult individuals, of all castes, in the population of colonies;

Nc is the average number of adult individuals per colony;

Q is the average number of queens contributing offspring to individual colonies;

M is the average number of males that fertilized each queen (these individuals no longer need be living).

Ñ is the equivalent of the number of breeding individuals in an idealized nonsocial population with equal numbers of males and females, and it provides an exact measure for the estimate of inbreeding, gene loss through random drift, and probability of extinction. In ants as in nonsocial organisms the effective breeding size will be exactly equal to the term given if breeding is panmictic, that is, completely random among the reproductive individuals, and less than the term if it is not panmictic (Wilson, 1963, 1971).

Examination of this formula shows that the most efficient means of enlarging effective population size is by increasing the number of queens. Merely adding a single additional queen to a monogynous colony system, for example, has the effect of doubling the effective population size. And to double the population size will enormously increase the mean survival time of the populations under a wide range of environmental conditions and demographic constraints (MacArthur and Wilson, 1967). Consequently polygyny alone might "rescue" rare populations from quick extinction. Put another way, group selection could establish polygyny in systems of populations small enough to be subject to frequent extinction. If there are multiple rare species or rare populations belonging to the same species, those containing polygynous colonies would be more likely to survive than others composed exclusively of monogynous colonies.

The effective population size of some ant species is indeed very low. In the Erebomyrma urichi population occupying Trinidad's Oropouche Cave in 1961, it was estimated not to exceed 400 (Wilson, 1962d). The population of Lasius minutus at Hidden Lake, Michigan, nested in only 700 mounds during the 1950s. Since a single colony occupied an average of about 4.4 mounds (Kannowski, 1959a,b), the population could be estimated to contain approximately 160 colonies and an effective population size as low as 320 outside the reproductive season. The extreme rarity of many social parasites, coupled with patchy distributions, is well known to myrmecologists. Chamberlin (in Wheeler, 1910a) found Formicoxenus (= Symmyrmica) chamberlini only “in several parts of a ten-acre field" near Salt Lake City, Utah, in 1902, and only three colonies were located there. After an intensive search eighty years later Buschinger and Francoeur (1983) rediscovered a sparse population in the same area. The extreme parasite Teleutomyrmex schneideri has one population apparently limited to the east side of the isolated Saas-Fee Valley of Switzerland, between 1,800 and 2,300 meters elevation (Kutter, 1950a); a second local population has been discovered near Briançon in the French Alps (Collingwood, 1956).

A second means of enlarging the effective population size, which might be favored by group selection (or more precisely, interdemic selection) is the reduction in the average colony size, which increases the number of colonies for a fixed total number of individual ants. Still another is the promotion of outbreeding, by such mechanisms as the appearance of the male before the reproductive females in individual colonies (protandry), fully developed nuptial flights that carry the reproductive forms away from the natal nests, and the synchronization of nuptial flights to bring together males and females from different colonies at the time of mating.

Have in fact the rare species taken these steps? From

the evidence presented in Table 6-1, it can be seen that a significantly higher number of such species have multiple queens in comparison with related species. The inference that can be drawn is that selection is strong enough at the level of entire populations of colonies (as opposed to individual colonies), due to the danger of extinction through small population size, to force polygyny on the colonies. But the inference is weak since polygyny also occurs in many ant species that are abundant and widespread.

To summarize, one circumstance by which polygyny might have arisen in ants is by group selection acting on rare species, the category in which interdemic selection is generally most likely to be potent. However, an alternative explanation for polygyny in rare species is available: because of the small number of individuals, associated queens might be more closely related than is the case in larger, more typical populations. Hence, as originally suggested for Formica rufa by Williams and Williams (1957), altruistic cooperation among queens would be favored due to sister-group selection. At the present time there is no clear way to decide between these two competing hypotheses.

An examination of Table 6-1 shows that decrease in colony size, another means of raising effective population size, in fact does not occur among the rare species. Furthermore, the data show no evidence of an increase in exogamy, the tendency of the sexual forms to mate with members of other colonies. Quite the contrary; the rare species have reduced the level of exogamy. Mating among members of the same colony, which has the effect of turning the colony into something resembling a self-fertilizing hermaphrodite, has long been recognized by myrmecologists as a trait of the rarest, parasitic species. The males are often apterous or subapterous, mating takes place in or near the nest, and the fecundated, winged queens then either disperse in search of new host colonies or else return to the old. Whether intracolonial mating characterizes other categories of rare species is an open question. Since the trait must cause a decrease in the effective breeding size of the population, just the reverse of what purely logical considerations concerning population size alone dictate, it is necessary to consider other possible advantages of such a design feature. At least one can be deduced: intracolonial mating certainly eliminates the loss of virgin reproductives normally occurring during dispersal and insures that the queens will be inseminated, however scarce the species. This advantage can easily outweigh disadvantages from inbreeding. Passing from random mating to perfect inbreeding reduces the effective breeding size by only half, a deficit that can be balanced merely by doubling the average number of nest queens. The exact extent of true brother-sister mating is unknown. Because of the additional trait of polygyny in these same species, the offspring of several matings probably breed with each other as a matter of course. In fact, Wesson (1939) did find that the queens and males of Protomognathus (= Harpagoxenus) americanus prefer to mate with unrelated individuals. It is even conceivable that parasitic species are less “adelphogamous” than previously assumed since it has not been established with certainty that true brothers and sisters within polygynous colonies really mate with each other at all. In polygynous colonies the opposite may be true.

Specialized nest sites. A much stronger correlation exists between polygyny and the manner in which colonies occupy nest sites. Truly polygynous species, most or all whose colonies contain multiple inseminated queens, fall into one or the other of three sets characterized by very different adaptation syndromes. (1) The first type is specialized on exceptionally short-lived nest sites. Such species are opportunistic in the sense employed by ecologists--they occupy local sites that are too small or unstable to support entire large colonies with life cycles and behavioral patterns dependent on monogyny. (2) The second type is specialized on nest sites that are scarce, evidently conferring advantages on queens willing to group together in lasting associations. (3) The third type is specialized on habitats--entire habitats as opposed merely to nest sites--that are long-lived, patchily distributed and extensive enough to support large populations. The three forms of specialization are not mutually exclusive; some ant species, for example Iridomyrmex humilis and Pheidole megacephala, possess features of the first and third types.

Consider first the class of opportunistic nesters. Colonies of Tapinoma melanocephalum, Tapinoma sessile, Paratrechina bourbonica, and Paratrechina longicornis often occupy tufts of dead grass, plant stems, temporary cavities beneath detritus in urban environments, and other local sites that sometimes remain habitable for only a few days or weeks. For example, colony fragments of Tapinoma sessile observed by Smallwood (1982) moved an average of every 12.9 days. Some species of Cardiocondyla excavate shallow tunnels and chambers in patches of soil in such places as the edges of palm trunks, sidewalks, and street gutters. Leptothorax curvispinosus typically nests in confined and relatively unstable preformed cavities in acorns, hickory nuts, galls, stems, and twigs. Lasius sakagamii favors unstable river banks, sandy and sparsely vegetated areas that are frequently flooded (Yamauchi et al., 1981). On the rocky islets of the Gulf of Finland, Formica truncorum occupies flimsy nest sites in which temperature varies in an unreliable manner (Rosengren et al., 1985). Colonies of Monomorium pharaonis and Tapinoma melanocephalum go so far as to invade houses to occupy cracks in walls, the lining of instrument cases, spaces in piles of discarded clothing, between leaves of books, and similarly unlikely microhabitats. Colonies of these species are characterized by extreme vagility--a readiness to move when only slightly disturbed and the ability swiftly to discover new sites and to organize emigrations. Their colonies are also typically broken into subunits that occupy different nest sites and exchange individuals back and forth along odor trails. Colonies of opportunistic nesters bud, disperse, and fuse again, as documented in Tapinoma erraticum by Dubuc and Meudec (1984) and various species of Leptothorax by Buschinger (1974a), Möglich (1978), Alloway et al. (1982), and Stuart (1985b). We suggest that it is the latter quality that gives polygyny a premium in opportunistic nesting and ties it closely to polydomy, the occupation of multiple nest sites. Because of the inevitable frequent fragmentation of the colonies, subunits probably lose contact with one another for long periods of time and occasionally forever. Having enough reproductive females to service most or all of the subunits means that the colony as a whole can exploit the rapidly fluctuating environment in which it lives. Other kinds of ants are not fragmented in this manner, and consequently a single queen suffices as the colony progenitrix.

The second class of polygynous nesters was recently defined by Herbers (1986a,b) and has only a single known example, the small North American myrmicine Leptothorax longispinosus. The species prefers hollow acorns and other preformed cavities on the ground, sites that are usually in short supply. Herbers found that the percentage of colonies that are secondarily polygynous increases with the local density of colonies, and she was able to exclude nest site fragility as a potent factor.

The third major class of fully polygynous ant species contains only a few species, but they include the great majority of examples that have been both well studied and do not qualify as opportunistic nesters. Thus, most fully polygynous species whose natural history is known to us are accounted for by the simple dichotomous classification proposed here. The habitats favored by species in the second category are first of all patchily distributed; they have distinctive qualities and are more or less isolated from each other. They are also extensive enough to support substantial populations of ants. Hence propagules of species that are specially adapted for such places encounter a potential bonanza when they succeed in colonizing one. The habitats are also relatively long-lived, giving a premium to the type of slow but thorough occupation made possible by polygyny and budding.

For example, the Allegheny mound-builder Formica exsectoides typically occupies persistent grassy or heath-like clearings. Such habitats are relatively scarce and patchily distributed, and many are fully occupied by dense unicolonial populations of Formica exsectoides. The “microgyna” form of Myrmica ruginodis (which is almost certainly a distinct species in its own right; see Pearson, 1981) shows a similar preference for scattered, very stable open habitats in England, some of which are known to have persisted at the same localities for as long as 200 years. The typical, or “macrogyna” form, which is haplometrotic and monogynous, favors less stable but more widespread habitats (Brian and Brian, 1955). Pseudomyrmex veneficus is specialized to occupy species of swollen-thorn Acacia that grow for long periods of time in areas of slow floristic succession. Because the acacias are able to expand into extensive thorn forests, the Pseudomyrmex veneficus colonies have the opportunity to build large unicolonial populations over a period of many years. Single populations may contain 20 million or more workers, rivaling Dorylus ants for the possession of the largest ant "colonies" in the world (Janzen, 1973b). Both are exceeded easily, however, by Formica yessensis, one supercolony of which in Hokkaido was estimated to contain 306 million workers and 1,080,000 queens (Higashi and Yamauchi, 1979; Higashi, 1983). "Tramp" species, those ants distributed widely by human commerce and living in close association with man, are typically polygynous. In a sense they can be said to have been preadapted for patchy but persistent and species-poor habitats created within man-made environments. Some of the best known species comprise unicolonial populations and spread largely or entirely by budding off of groups of workers accompanied on foot by inseminated queens. Examples include Monomorium pharaonis, Pheidole megacephala, Iridomyrmex humilis, and Wasmannia auropunctata.

The isolated-habitat specialists, as distinguished from the nest-site opportunists, are species that have difficulty locating their preferred habitats, but once having found a suitable place, are opposed by relatively less competition from colonies of their own or other ant species. Thus it is to the advantage of the founding colony to spread out as a continuous unicolonial population, occupying all of the habitable nest sites and foraging areas. An example involving the Argentine ant Iridomyrmex humilis is shown in Figure 6-2. In contrast, most other ant species have no trouble finding a suitable habitat, but such places are typically already saturated with other ant colonies. The best strategy of these ants is to send out large numbers of flying queens capable of founding new colonies independently. A tiny fraction of the propagules will locate some of the rare nest sites and foraging areas left unfilled; the colonies they produce will not find it profitable to try to spread outward through the surrounding occupied territories by budding in the unicolonial manner.

The full life cycle of unicolonial populations remains to be worked out in detail, although details of the structure and secular changes over a period of up to a few years have been studied in the huge colonies of Formica lugubris in Switzerland by Cherix (1980), Formica polyctena in the Netherlands by Mabelis (1986), and Formica aquilonia in the Soviet Union by Zakharov et al. (1983). A map of part of the Formica lugubris

colony is presented in Figure 6-3. Among the interesting variables to be studied as functions of the age and condition of the populations are the queen:worker ratios and the behavior of queens during nuptial flights. Scherba (1961) reported that most of the queens of Formica opaciventris, a North American mound-building, unicolonial species, mate within a few meters of their nest of origin and return at once, while a small minority, originating entirely from mounds near the edge of the population, fly out of sight in a direction away from the population. A very similar pattern occurs in at least one of its European equivalents, Formica polyctena, the life cycle of which was presented in Chapter 3 (see also Gösswald and Schmidt, 1960). We predict that in the early stages of population growth, the queen:worker ratio will start high (immediately following independent nest founding by one or more newly inseminated queens, usually by adoption of an alien host species), drop to a steady state as the population comes to saturate the habitat, then rise again as the habitat quality declines and a higher premium is placed on dispersal away from the habitat. It also seems likely that a fraction of the queens engaging in dispersal flights will increase when the habitat is saturated, and increase still more as the habitat declines.

Thus in a curious fashion the extreme polygynous species have evolved unicolonial populations as a means for trading dispersibility for potential colony immortality. Where competitive pressure is less, the species can afford to gamble on great longevity. The colonies can be fused into single units, and newly mated queens can more safely return to the parental nests. And since colonies are on the average longer lived, fewer empty nest sites will be available at any given time, making the dissemination of queens away from the parental nests less profitable. The results will be a dual pressure for the unicolonial species to suspend claustral nest founding altogether. The adoption of unicolonialism, with its increased degree of inbreeding and reliance on budding as the principal means of reproduction is an evolutionary step that parallels the adoption of apomixis and vegetative reproduction in nonsocial organisms. Both permit the rapid growth of population in habitats that are relatively free of competing ants but sparsely distributed.

An interesting phenomenon for which no explanation yet exists is the frequent coexistence of pairs of closely related species, one of which is monogynous and the other polygynous. Examples include the “macrogyna” and “microgyna” forms of Myrmica ruginodis, probably at least a few other European species of Myrmica (Pearson, 1981), the acacia ant Pseudomyrmex veneficus and an undescribed polygynous sibling form in Mexico (Janzen, 1973b), two apparent species placed under Crematogaster minutissima (Wilson, unpublished), Dorymyrmex insanus and Dorymyrmex flavopectus (Nickerson et al., 1975), Formica incerta and Formica nitidiventris (Talbot, 1948), and two apparently distinct species of Formica neorufibarbis in the White Mountains of New Hampshire (Wilson, 1971). A similar duality appears to exist in Formica “rufa,” which apparently consists of two sibling species, one strictly monogynous and the other polygynous (Kutter, 1977). It is also possible for at least the rough equivalents of monogynous and polygynous strains to occur as genetic polymorphs within the same species, as in the "A" and "B" colony types of Rhytidoponera confusa and Rhytidoponera chalybaea in Australia (Ward, 1983a,b).

Variable food supply. A third principal clue to the origin of polygyny, as well as mixed strategies among and within species with respect to the number of queens, has been provided by Briese (1983) in his study of an Australian species of Monomorium (= Chelaner) belonging to the rothsteini group. He followed two colonies over a two-year period in a semi-arid saltbush steppe in New South Wales. One, containing winged queens capable of flight, started new colonies by the conventional means of nuptial flights followed by claustral nest founding. The other, containing brachypterous (short-winged) queens unable to fly, started new colonies by fission. The brachypterous queens were either adopted back into the natal nest or escorted to new nest sites by their worker sisters. Both procedures led to polygyny with brachypterous queens. Flights by the normal-winged queens occurred after a favorable period of substantial rainfall that increased the food supply, principally seeds. Fission by brachypterous females followed a drought and reduced food supply (Figure 6-4). Briese considered the two kinds of colonies to belong to the same species, and their founding methods to constitute alternative responses of a mixed strategy:

When conditions are favourable, the production of fully alate queens, with sufficient body reserves to raise a first brood by themselves, would allow a colony to extend its genetic material over a much wider area with maximised chances of successful establishment. When stress is very severe queen production may cease altogether. However, under conditions of less severe food stress, the mode of colony foundation might alter, and brachypterous queens with less body reserves be produced. During colony foundation, these would be accompanied by groups of worker ants which act as food gatherers to support the initial brood. Such daughter colonies would remain in the same, relatively favourable area, still capable of supporting a colony, rather than facing the risk of failure through the alternative mode of random dispersal to areas of unknown favourability.

The attractiveness of Briese's hypothesis is that it can be tested, regardless of whether the two queen morphs represent different species or variants of the same species. Further studies of this and other species can judge whether there is indeed a propensity of ants to evolve fission in areas of wide and erratic fluctuations in climate, such as occur in the arid Australian interior. Furthermore, it can be determined whether species that are truly polyethic with response to the two strategies favor fission during hard times.

Gyny and species diversity

Having reviewed the possible ecological prime movers that lead to monogyny or polygyny, let us now briefly consider some of the broader implications of variation in the number of queens. Monogyny is closely associated with colony distinctness. Each colony occupies a separate nest site, and its workers either avoid or attack the members of other colonies they encounter during foraging. The loss of territorial boundaries in the case of highly polygynous, unicolonial species changes the rules drastically. Unicolonial species are notable for their high local abundance and the degree to which they appear to dominate the environment. Introduced populations of Pheidole megacephala, Wasmannia auropunctata, and Iridomyrmex humilis extirpate many other kinds of ants, although it is not known whether species occurring within the native ranges of the unicolonial ants have evolved competitive resistance to the point of being able to coexist. Habitats containing populations of Formica exsectoides have notably sparse ant faunas, and the related Formica exsecta and Formica opaciventris are known to exclude some other territorial ants, including other species of Formica, by aggression (Scherba, 1964; Pisarski, 1972). The same is true of Formica aquilonia and Formica polyctena, the “large-scale conquerors” of the Formica rufa group in Europe (Rosengren and Pamilo, 1983). It is not known to what extent the general occurrence of unicolonial species in species-poor habitats is a cause and to what extent it is an effect. The matter can be put as a question: have the unicolonial ants simply adapted to habitats that were species-poor in the first place, or does the formation of supercolonies provide them with a decisive competitive edge in habitats that would otherwise be species-rich? This problem seems eminently tractable to field analysis (see Chapter 11).

It is possible that gyny and the nature of territoriality also affect the patterns of species diversity and quite possibly the mechanism of speciation itself. Local faunas consisting of multicolonial ant populations, which comprise in turn chiefly monogynous and oligogynous colonies, seem to be more susceptible to increases in within-habitat species diversity, for three reasons. First, as previously noted, they appear to suppress other ant species less severely than do unicolonial populations, permitting the buildup of larger numbers of coexisting species. Second, because the colonies are much smaller in size, multicolonial species are able to specialize more on nest sites and food. For example, a colony of one species might do well with a single hollow stem, while a colony of another species is able to occupy the space beneath a nearby stone. Other species can be differentiated according to food, some preying exclusively on small arthropods, others mostly tending scale insects, and so forth. The relatively huge populations of unicolonial species cannot afford a narrowing of their niche. In order to survive they must remain broad generalists, which brings them into competition with a large number of specialists as well as other generalists.

The third reason why multicolonial species can be expected to exhibit higher within-habitat diversity is more subtle. Many studies have shown that colonies of ants are hostile to a degree inversely proportional to the degree of similarity to their competitors. That is, they are most aggressive to other colonies of the same species, somewhat less to other species in the genus, and least of all to forms that not only belong to other genera but differ strongly in size and behavior. This being the case, within-habitat diversity can be expected to be enhanced by the divergence of species odors between closely related species. Suppose that two newly formed, cognate species have arisen by genetic divergence during geographical isolation, and have recently come into contact along the boundaries of their respective geographical ranges. Suppose further that the species are sufficiently divergent in ecological requirements so that neither would exclude the other by means of nonaggressive preemption of resources. Yet the two forms are likely to remain in separate geographical ranges so long as their species odors are too similar, causing interspecific territorial exclusion to be as strong as intraspecific exclusion. The two species can penetrate one another's range only if one or both undergoes a divergence in the species-specific components of the colony odor. This would be a form of character displacement comparable to the divergence of identifying songs by which territorial bird species penetrate one another's ranges (Murray, 1971). The result is an increase in the within-habitat species diversity.

The penetration of territories belonging to alien monogynous species is made easier by the very fact that territorial colonies repel one another so effectively--to the extent that colonies are overdispersed in their statistical pattern of distribution. In many kinds of ants, such as members of the genus Pogonomyrmex, this pattern is geometrically very regular and is maintained by high levels of intercolonial aggression (Hölldobler, 1974, 1976a). If colonies of other species are different enough in species specific components of the colony odor to escape such aggressive response, and if they are also distinct enough in their foraging habitats not to be replaced through competitive exclusion, they might easily slip into nest sites located between those of the resident species.

In an earlier analysis (Hölldobler and Wilson, 1977b), we hypothesized that the degree of odor specificity associated with monogyny and intercolony territoriality lends itself to interspecific recognition, the reduction of interspecific territorial aggression by means of that recognition, and an increase in numbers of species that can coexist in the same habitat. Many polygynous ant species, and particularly those that are unicolonial, have surrendered some of this discriminatory power. As a result they remain aggressive toward a broader range of species, and it is not surprising to find very few unicolonial species occupying the same habitat, as well as an overall decrease in species diversity. Although it is theoretically possible for unicolonial ant species to build up between-habitat diversity by means of specialization on habitats as opposed to niches within habitats, very little appears to have occurred in nature--quite possibly due to the preemption of most kinds of habitats by monogynous, multicolonial ant species.

Pleometrosis

The alliance of two or more queens during colony founding is a widespread but still far from universal habit in ants. Within particular species it is more or less an optional procedure. For example, ten percent of founding queens have been recorded as being in pleometrotic groups in Iridomyrmex purpureus (Hölldobler and Carlin, 1985), 80 percent in Lasius flavus (Waloff, 1957), 82 percent in Acromyrmex versicolor (Rissing et al., 1986), 89 percent in Messor pergandei (Rissing and Pollock, 1986), and 97 percent in Myrmecocystus mimicus (Bartz and Hölldobler, 1982).

These data suggest only coarse differences between species. In fact, the incidence of pleometrosis is usually flexible within particular species, rising under environmental conditions that promote crowding of the newly dealated queens. In the honeypot ant Myrmecocystus mimicus, queens are strongly attracted to each other after they have mated, and the pleometrosis is enhanced further by a tendency of the queens to start nests in close proximity (Bartz and Hölldobler, 1982). Even after the nest digging has begun, the foundresses aggregate still further, until the great majority of nests contain more than one queen, as illustrated in Figure 6-5. A closely similar pattern has been described in the fire ant Solenopsis invicta by Tschinkel and Howard (1983). After the fire ant queens have mated and cast their wings, they tend to settle on higher ground, away from rainwashed areas and puddles. Within the favored spots they also aggregate in a strongly non-random manner. A 64-fold increase in queen density across various sample areas near Tallahassee, Florida, was found to result in a 2.2-fold increase in the average number of queens per nest, with most burrows still containing one to three queens but a few containing ten or more. Variation in queen density is quite potent in affecting this entire aspect of social behavior. In the Tschinkel-Howard study it accounted for 70 percent of the variation in the mean number of queens per nest and 86 percent of the overall aggregation of queens in local areas. An even stronger initial clumping as an effect of microhabitat selection has been recorded in the leafcutter ant Acromyrmex versicolor by Rissing et al. (1986). Newly mated queens strongly prefer to settle under the shade of the scattered trees of the Arizona desert, where close proximity leads to a preponderance of joint occupancy in the newly excavated nests. The shaded and hence cooler environment permits a much higher survival of the incipient colonies.

While pleometrosis is very common among ants, it seldom leads smoothly to polygyny. In most cases studied to date, multiple queens are reduced to a single egg-laying queen, at least within local areas of the nest, shortly after the first brood of workers ecloses. Either the workers eliminate the supernumerary queens (Nothomyrmecia macrops, Myrmecocystus mimicus, some strains of Solenopsis invicta), or the queens fight among themselves and are reduced further by worker aggression (Messor pergandei), or the queens begin to fight and disperse to different parts of the nest, creating a condition of oligogyny (Iridomyrmex purpureus, Camponotus herculeanus, Camponotus ligniperda, Lasius flavus). Even when extra inseminated queens are tolerated in close association with the primary egg layer, their ovarian development is usually suppressed (species of Leptothorax and Formicoxenus, at least one strain of Solenopsis invicta). Primary polygyny, in which associations of founding queens survive to become multiple egg layers at close quarters in mature colonies, appears to be rare. In fact we know of two cases: the leafcutter [[Atta texana (Mintzer and Vinson, 1985a; Mintzer, 1987) and Pheidole morrisii (Stefan Cover, personal communication). In all other species whose colony ontogeny has been thoroughly studied, the polygyny is secondary, that is, originates when additional inseminated queens are adopted by an already existing worker force.

The association of queens during founding does not appear to be due to kin selection of the kind observed in foundress females of Polistes wasps. The latter insects preferentially associate with their former nestmates, which are likely to be full sisters (Ross and Gamboa, 1981; Post and Jeanne, 1982; Strassmann, 1983). The evidence so far points to lack of such discrimination in ants. In particular, foundress queens of Myrmecocystus mimicus and Messor pergandei associate freely with females that could not have come from the same nest. For example, Rissing and Pollock (1986) combined newly inseminated queens of Messor pergandei collected one to six kilometers apart. In all of nine replicates, the queens combined with no outward aggression to start a single cooperative nest. Another clue has been provided by the discovery that multiple queens of older Solenopsis invicta colonies are no more closely related to each other than to queens collected outside the nest (Ross and Fletcher, 1985a).

The key advantage that does accrue to multiple founding queens is the fact that in comparison with solitary founding queens they produce larger initial broods and worker force in less time and with less individual weight loss. This effect has been documented across a wide variety of ant genera, including Lasius (Waloff, 1957), Solenopsis (Wilson, 1966; Markin et al., 1972; Tschinkel and Howard, 1983), Camponotus (Stumper, 1962; Mintzer, 1979a); Tapinoma (Hanna, 1975); Messor (Taki, 1976); and Myrmecocystus (Bartz and Hölldobler, 1982). In addition, colonies starting with larger initial worker forces are more successful in brood raids or territorial fights directed against other incipient colonies. This advantage has been documented in Myrmecocystus mimicus by Bartz and Hölldobler (1982) and Messor pergandei by Rissing and Pollock (1986). No contrary cases are known.

Experiments by Bartz and Hölldobler (1982) on Myrmecocystus mimicus have allowed a more precise assessment of the advantage of pleometrosis as well as a test of optimization. As shown in Figure 6-5 the most frequent group size of founding queens in newly dug burrows is two to four. Under laboratory conditions, the maximum production of workers per founding queen--as opposed to workers produced by the entire founding group--is attained when the number of associated queens is about 3, corresponding very closely to the modal group size in nature (Figure 6-6). It also turns out that the mortality rate among the founding queens is lowest when the group size is 3 to 4, a second selection factor seemingly favoring this intermediate level. The explicit reasons why this group size does better have not been worked out. They might include convex curves of efficiency in brood care, along with steadily rising curves of mutual egg cannibalism (some of which was observed) and maximal resistance to disease.

Apart from optimal group size, Bartz and Hölldobler discovered an additional behavior favoring group formation over solitary founding in general. After the first Myrmecocystus workers appear, they start to raid nearby incipient colonies by transporting brood to their own nests. In laboratory enclosures that simulated the natural dispersal of incipient nests, all of the brood eventually ended up in a single nest. Workers frequently abandoned their own mothers in favor of the winning colonies. In 16 of 23 such experimental arenas, colonies starting with 5 queens prevailed in the end, while in 7 cases the winning group consisted of 3 queens. In no case did colonies founded by 1 or 2 queens prevail. Furthermore, in 19 cases the winning colonies were the ones with the largest initial worker force. This martial behavior foreshadows an even more remarkable competition that develops among mature Myrmecocystus mimicus colonies. These colonies conduct ritualized tournaments as part of the defense of their foraging territories. Opposing colonies summon their worker forces to the tournament area, where hundreds of ants perform highly stereotyped display fights. When one colony is considerably stronger than the other, in other words able to summon a larger worker force, the tournaments end quickly and the weaker colony is sacked. During these final incursions, the queen is killed or driven off and the larvae, pupae, callow and honeypot workers are transported to the raiders' nest (Hölldobler, 1976c, 1981a).

A closely similar pattern of prevalence of groups over solitaires was independently discovered in Solenopsis invicta by Tschinkel and Howard (1983). The mean number of founding queens per nest varied according to site from 1.1 to 3.4. In laboratory tests, reproductive performance was compared across groups of founding queens containing 1, 5, 10, and 15 individuals respectively. Groups with 5 queens produced the most pupae and workers, while productivity per queen was greatest at a lower number, somewhere between 1 and 5. The advantages of group life do not stop at the founding stage, however. When the colonies were maintained over a period of 100 days, with the worker populations reaching 200 to 1000 individuals, the most rapid growth occurred in the polygynous colonies. This differential persisted even after the workers had executed all but one of the founding queens. And as in Myrmecocystus mimicus, workers of Solenopsis invicta tend to move into whatever neighboring nest has the largest number of workers, leaving their mothers to die of starvation. Tschinkel and Howard suggest four advantages to pleometrosis in the fire ants, with accompanying documentation from laboratory and field studies:

(1) Earlier maturation and reproduction lead to a shorter generation time and higher population growth rate (Tschinkel and Howard, 1983).

(2) Fire ants are territorial and war with conspecific colonies along their mutual territorial boundaries, with the advantage presumably going to the larger colonies (Wilson et al., 1971).

(3) The coalition of queens in an incipient colony benefits those with the larger worker force, as just described (Tschinkel and Howard, 1983).

(4) A colony's ability to survive adverse physical conditions is enhanced by larger size. Markin et al. (1973) found that Solenopsis invicta colonies not attaining a certain size by the onset of cold weather fail to survive the winter, and the same could be true of survival in the prolonged dry season of the Brazilian homeland of this species. A similar effect has been discovered in Myrmecocystus mimicus. Young colonies require a large number of workers in order to produce repletes. In the arid conditions of the American Southwest, the advantage provided by repletes in surviving the harshest seasons may be large enough to favor cooperation of the founding queens and their workers (Bartz and Hölldobler, 1982).

Tschinkel and Howard have further addressed the puzzling circumstance that founding queens readily join oversized groups of 15 or more, in which there is no hope of reproduction. They point out that Solenopsis invicta is far less abundant in southern Brazil, where it evolved, and populations are mostly restricted to disturbed or seasonally flooded habitats. Consequently mating flights and post-flight queen density would be low, pleometrosis much less frequent, and average group size low, so (unlike the case for Myrmecocystus mimicus) there might be little opportunity for selection to operate against excessively large groups. Yet selection for joining an available group would be strong.

Raiding and coalition of incipient colonies might be a commoner phenomenon in ants than previously realized. A third case has recently been reported in the desert granivore Messor pergandei by Rissing and Pollock (1986). Once again, as in Myrmecocystus and Solenopsis, pleometrotic colonies produce earlier brood and are more successful at raiding than haplometrotic colonies.

Demographic consequences of gyny and of dominance orders

The effects of variation in the number of queens will be fully understood only when the consequences of the variation on colony growth and survival are subject to quantitative measurements. Only a few studies have been designed thus far to acquire such data. Mercier et al. (1985a,b) found that the little European formicine ant Plagiolepis pygmaea is truly polygynous, averaging 17 laying queens per nest. The productivity of individual queens is very uneven, varying directly with their weight, their attractiveness to the workers, and the number of workers in the colony. When experimentally combined in pairs, their oviposition rate was approximately half that when kept alone with a fixed number (200) of workers. The implication from the combined studies was that the number of workers, rather than the number of queens, is the prime determinant of the total oviposition rate and hence the potential for colony growth. The same result was obtained by Wilson (1974b) for colonies of the ant Leptothorax curvispinosus. In Solenopsis invicta, on the other hand, polygynous colonies have higher brood:worker ratios than monogynous colonies, possibly implying that under some circumstances the number of queens is important (W. R. Tschinkel, personal communication).

By employing enzyme genetic markers in the polygynous Solenopsis invicta, Ross (1988) was able to assess directly the contributions of individual queens to the maternity of workers and female reproductives. He discovered that some queens that contributed substantial numbers to the worker pool, often added few if any daughters to the pool of female sexuals. Other queens' offspring developed primarily into reproductive females. The mechanism of this remarkable differential reproduction among associating queens is not yet known. However, Ross's results strongly suggest “that significant variability in short- as well as long-term reproductive success may occur among the distantly related queens associating in polygyne Solenopsis invicta nests.” Thus the common earlier view that queens of polygynous colonies perform about equally will probably have to be generally revised.

In most instances one queen was clearly dominant and the other subordinate. The dominant attempted to hold the subordinate in place, while the subordinate either struggled to work free or else lay quietly in a ball-like "pupal" posture. Several times the two appeared to be more evenly matched, so that neither could get a commanding position on top of the other. In both cases the combatants jockeyed and rolled around sluggishly for up to 15 minutes or longer. In one case three queens rather than two fought together briefly; one then walked away and left the struggle to the remaining pair.

Subordinates that managed to break free usually walked briskly a short distance from the scene. One was seen to depart from the brood chambers to the outer nest chambers, only to return after a few minutes--and suffer another attack. There was one piece of indirect evidence that damage was being inflicted during the battles. The mortality of the queens was much higher than that of the workers, the reverse of the usual case in ant colonies. By the end of the 17-day period during which fighting occurred, the number of queens dropped from 25 to 5. During this entire time no episodes of aggression were witnessed among workers or between workers and queens.

An apparent dominance hierarchy of subtler texture was demonstrated among queens of the polygynous Myrmica rubra by Evesham (1984a) and among three founding queens of the American carpenter ant Camponotus ferrugineus by Fowler and Roberts (1983). The Camponotus queens showed consistent (but at most marginally significant) differences in lunging behavior, displacement from original position, and consumption of eggs laid by other queens. When secretions of the resting queens were collected on absorbant paper and presented to separate groups of queenless workers, the order of attractiveness corresponded to the dominance order of the three queens. A strikingly mammal-like ritual was observed by Hölldobler and Taylor (1983) in the very primitive Australian ant Nothomyrmecia macrops: one queen stood above a rival, occasionally stepping on top of her while pointing downward with her head. The subordinate remained still in a crouching posture. The contacts did not lead to fighting of the kind observed among Nothomyrmecia workers from different

colonies (see Figure 6-7), but it was followed by the expulsion of the subordinate queen from the nest by the workers. Similar ritual postures have been observed among

queens of Myrmecocystus (Figure 6-8) and workers of Leptothorax allardycei by Cole (1981) and wasps of the genera Polistes and Mischocyttarus by Pardi (1948), West (1967), and Jeanne (1972).

Still more subtle indicators of rivalry have been discovered in other ant species. Queens of Leptothorax curvispinosus do not attack or threaten one another, but do consume the eggs of their rivals to differing degrees, so that a hierarchy of sorts emerges (Wilson, 1974a). Those of the Neotropical arboreal Procryptocerus scabriusculus are distinguished solely by the higher rates at which they solicit food from workers and larvae and groom themselves (D. Wheeler, 1984).

Interference among competing queens and reproductive workers is yet another factor that must be entered into the equations. In queenless Leptothorax allardycei groups at least, a large amount of time is spent in the dominance exchanges by the most competitive workers, in fact more than is spent on brood care. Cole (1986) has estimated that the cost to worker reproduction is 15 percent of the total in the case of time spent on brood care and 13 percent of brood care when measured as brood pieces tended per unit time.

Aggression and dominance have proved far more common within ant colonies than was appreciated as recently as 1970, and the interactions are very important in determining the numbers of reproductive females. The methods of control range from periodic all-out attacks to mammal-like dominance posturing, differential egg-eating of rivals, and the use of primer pheromones to suppress ovarian development.

An extreme case of fighting was recorded in a laboratory colony of the Asian termite predator Eurhopalothrix heliscata by Wilson and Brown (1984). The mother queen had evidently been lost during transport of the colony from the field, and unfertilized dealated queens were contending for control. Battles ensued almost continuously for a week; they then became increasingly intermittent and finally ceased altogether. On several occasions two pairs were locked in combat at the same time, and at various times either fully dealate or partly (but never fully) winged individuals were engaged. At least five individuals were evidently involved on different days, and the total number during the 17-day period may have been far higher. However, not all interactions were aggressive. Most of the time queens simply moved on past when they encountered each other, and on two occasions a dealate queen was observed grooming another.

In a typical aggressive episode the two individuals were locked together like wrestlers, with one trying to maintain a hold on the body of the other with all her legs while her opponent struggled to escape. Sometimes the aggressor rolled her own body around the head and anterior portion of the opponent's alitrunk; on other occasions she centered her attack on the alitrunk and occasionally farther back, on the gaster. Typically the aggressor pressed her mandibles downward and against the body of the opponent, evidently to gain added purchase. Several times Wilson and Brown saw one queen biting an antenna or portion of the vertex of the opponent. In most episodes the aggressor also pressed the tip of her gaster against the body of her adversary in an additional attempt to gain better purchase. No evidence of stinging was observed at this time, although the sting could easily have been extruded periodically and into the body of the adversary without being visible.

A still more subtle but pervasive phenomenon is the inhibition by the egg-laying queen of ovarian development in nestmates through the release of inhibitory pheromones. Substances producing this effect have been demonstrated in Plagiolepis pygmaea (Passera, 1980b), Oecophylla longinoda (Hölldobler and Wilson, 1983a), and Novomessor cockerelli (B. Hölldobler, N. Carlin, and E. P. Scovell, unpublished). The inhibition of the reproductive egg-laying by workers due to the presence of the laying queen has been widely demonstrated in other ants, and the experimental evidence suggests that it is likely to be mediated at least to some extent by inhibitory pheromones. Examples include Odontomachus haematodus (Colombel, 1972), Leptothorax tuberum (Bier, 1954b), Leptothorax (= Myrafant) recedens (Dejean and Passera, 1974), various species of Formicoxenus (Buschinger, 1979), and Myrmica rubra (Brian and Rigby, 1978). A second kind of pheromone produced by mated queens of Solenopsis invicta acts directly on virgin queens to prevent them from shedding their wings, histolyzing the wing muscles, and undergoing rapid oogenesis, all of which would convert them into reproductive rivals (Fletcher and Blum, 1981, 1983b; Fletcher and Ross, 1985).

Aggression can lead not only to dominance orders and despotisms but to a mutual separation of the queens into different parts of the nest, in other words a spatial oligogyny. Hölldobler and Carlin (1985), who followed the transition from pleometrosis to large oligogynous colonies of Iridomyrmex purpureus, found that this process could be divided into three phases correlated with the size of the worker population. Soon after the first worker in laboratory nests emerged, the previously amicable co-foundress queens commenced aggressive displays. In this "phase I," they faced each other head-on while engaging in stereotyped bouts of rapid mutual antennation (see Figure 6-9), and occasionally shifted to threats with open mandibles and even biting. The encounters resulted in clear-cut dominance, with the winning queen holding her ground while the loser backed or turned away. The dominant queen also spent more of her time on the egg pile and contributed more of her own eggs to it. This phase, during which the queens remained together in an uneasy hierarchy, lasted for about a year, until the worker population reached 200-400 workers. Toward the end of the period, the antennation bouts became more frequent, and the queens began to leave the founding nest tubes for intervals of up to 3 minutes. In phase II, which lasted approximately 9 months, the separations became gradually longer, lasting from several hours at a time to over 30 days. Meanwhile, the size of the single colony kept under observation increased to 2,000 workers and dispersed out over many more nest tubes. During phase III, when the colony had reached a size of 2,500 to 3,000 workers, one of the queens moved permanently to a separate part of the nest area. The structure of colonies in the wild was consistent with this pattern of social change. Eleven mature nests were excavated in Australia, containing 10,000 or more workers; two queens were found in each of two of the nests, one in each of five nests, and none in the remaining four nests. The queens were clearly separated in different galleries within both of the colonies that contained two queens. When these two pairs were placed in small artificial nests with 500 of their workers, they soon initiated ritualized antennation bouts and subsequently moved apart in the same manner as observed in the laboratory colony.

Aggression and dominance hierarchies, sometimes including outright battles, have been described among workers of a few species when the queens are removed. Examples include the ponerines Platythyrea cribrinodis (Hölldobler and Wilson, unpublished observations) and Rhytidoponera metallica (Ward, 1986), and the myrmicines Leptothorax duloticus (Wilson, 1975a) and Novomessor cockerelli (B. Hölldobler, N. Carlin, and E. P. Scovell, unpublished). The Novomessor workers most frequently attacked were the ones with the greatest amount of ovarian development.

Dominance hierarchies among workers have also been recorded in queenright colonies of the myrmicines Leptothorax allardycei (Cole, 1981, 1986), Harpagoxenus americanus (Franks and Scovell, 1983), and Harpagoxenus sublaevis (A. Bourke, personal communication). In both species the high-ranking workers also have the greatest ovarian development and receive more food than the other workers. Even in the presence of the queen the Leptothorax allardycei dominants produce 20 percent of the eggs, which are unfertilized and hence destined to produce males if they develop. In the aggressive reactions the workers antennate each other, and the dominant ant often climbs above her opponent while (in Leptothorax at least) pummeling her with her mandibles. The subordinate responds by freezing, crouching, and drawing her mandibles back to the side of her head. Bourke (1988) discovered that in queenless worker groups of Harpagoxenus sublaevis, worker dominance behavior inhibits egg-laying by subordinates. In queenright colonies, on the other hand, the queen appears to restrict both dominance behavior and oviposition by workers, probably by chemical means.

Who is in charge: queens or workers?

The picture that has emerged in contemporary studies is the existence of a moderate amount of struggle within ant colonies. Queens and workers appear to be in general conflict over the management of the ratio of investment in new queens and males. In some species, under appropriate conditions, queens battle queens for principal reproductive rights. Workers compete with their nestmates for the same privileges in the absence of the queen and, in a few cases, even when the queen is present. As we have seen from the review just completed, all of this conflict falls into patterns. What does the pattern tell us about social organization, in particular about whose reproductive interests and genetic fitnesses are being served? The question can be focused to some extent by returning to the question partially explored in Chapter 4: which caste controls the reproductive activities of the colony, the queen or the workers?

A remarkable general phenomenon of ant biology that might be interpreted as dominance by the queen over the workers is the queen-tending pheromone. Queens are generally attractive to workers, who often form entourages around them, licking their bodies, feeding them with trophic eggs and regurgitated liquids, and guarding them against intruders. The queens of monogynous colonies, or at least those colonies in which one queen prevails as the egg layer, are more attractive than those in functionally polygynous colonies. In both comparisons of species and of growing colonies belonging to the same species, the queens are most attractive in the largest colonies. In Solenopsis invicta, the best-studied ant in this regard, no fewer than five behaviors are evoked by the secretions of the physogastric queens when the queens are displaced by the experimenter outside the nest (Glancey et al., 1982; Glancey, 1986):

1. Intense initial attraction toward the queen.

2. Formation of a dense cluster of workers around the queen.

3. Transport of brood to the queen and deposit of the brood next to her.

4. Formation of odor trails to the nest, often with several branches coalescing into a very wide trail terminating at the nest.

5. Guidance of the queen along one of the trails; if the queen does not walk along the trail under her own power, the workers drag her into the nest.

At least some of the components of the pheromone inducing this remarkable series of responses originate in the poison gland sac of the queen. Those characterized so far are two pyranones and a dihydroactinidiolide, the structures of which are illustrated in Figure 6-10. The queen pheromone of a second myrmicine species Monomorium pharaonis, has been identified as neocembrene. Found only in mated queens, it exercises a strong attraction on the workers, but no other responses have been reported thus far (Edwards and Chambers, 1984). In the European wood ant Formica polyctena, methyl-3-isopropylpentanoate, which is produced in the cephalic glands of the queen, inhibits aggression in workers (Francke et al., 1980).

These immediate and overt effects are only part of the total regime of influence of inseminated queens on workers, at least among some of the phylogenetically more advanced subfamilies of ants. As Brian and Hibble (1963) first showed, the presence of laying queens in colonies of Myrmica rubra also reduces the likelihood that larvae will become queens and hastens their development into workers. Some of these results are achieved by an increased tendency of the workers to bite and scar the growing larvae when a queen is present. Further studies by Brian and his co-workers have revealed a complex repertory of variable responses on the part of workers to different kinds of larvae, which is sensitive to whether queens are present or absent. Their propensity to lay eggs of their own is also strongly affected (see Table 6-2). The overall result is a system of checks and balances that inhibits worker reproduction and production of new queens when laying queens are present and promotes these two activities when laying queens are absent. The system sharpens the division of labor between the queens and the workers.

In view of the considerable evidence of queen control by pheromones, it is surprising that little attention has been given to the exocrine glands of this caste in ants. Whelden (1963) and Hölldobler and Rettenmeyer (unpublished observations) found that queens of New World army ants belonging to the genus Eciton are much more richly endowed with such structures than are workers of the same species. A similar difference has been noted in weaver ants of the genus Oecophylla (Hölldobler and Wilson, 1983a), the amblyoponine “army ants” of the genus Onychomyrmex (Hölldobler and R. W. Taylor, unpublished), and the leptanilline army ants of the genus Leptanilla. It is to be expected that the exocrine development of the queens will be greatest in species with the largest colonies, because the amount of pheromonal material required to reach the entire worker force is correspondingly greater.

Yet another indication that queens can exert control comes from the extraordinary discovery by Masuko (1986) that queens of the primitive ant Amblyopone silvestrii are vampires: they depend exclusively on hemolymph (blood) which they obtain by biting the larvae. Amblyopone workers do not lay trophic eggs or exchange liquid food by regurgitation, the two ordinary means by which ant queens obtain nourishment. Nor do the queens feed directly on the geophilomorph centipedes and other arthropod prey captured by the workers. Instead they use their sharp mandibles to pierce the dorsal integument of the forward part of the abdomens of older larvae, then drink the hemolymph leaking from the puncture. The bodies of the donors are scarred, but they appear otherwise unharmed. Masuko has observed similar behavior in queens of three Japanese species of the aberrant ponerine genus Proceratium, which ordinarily prey on the eggs of arthropods. In another study, Masuko (1987) found that queens of Leptanilla japonica, a minute army-ant-like species feed from special exudatory organs of the larvae. These structures, which are unique to Leptanilla and other Leptanillinae, resemble spiracles and are located on either side of the third abdominal segment.

Yet another apparent mode of queen control is aversion to nest queens by workers of Leptothorax curvispinosus (Wilson, 1974a). The workers regularly feed the queens by regurgitation, but they respond more commonly to them by withdrawing for a distance of two millimeters or more after making antennal contact. Sometimes they back off or turn aside and walk away at a normal gait. But with equal frequency they run away rapidly. When a queen moves into and through a crowd of workers, they "explode," scattering away from the queen and clearing a path in front of her. The workers seem to avoid her head in particular and to pay little attention to the rest of her body. The result is that the laying queens are able to move more freely around the nest. In particular, they gain easy access to the larvae, whose salivary secretions form part of their diet.

To summarize this review on queen actions, a wide diversity of anatomical structures and behaviors has been discovered in ants that might be interpreted as mechanisms by which queens control the workers. But this functional explanation is by no means certain. In each case it could be argued that the workers are not really "controlled," in the sense that a conflict exists between them and the queens and is resolved in favor of the queen. Quite the contrary: the workers may in fact simply be monitoring the presence of the queens and respond in ways that increase their own inclusive fitness. It can even be said that Amblyopone and Proceratium larvae surrendering their blood to the queen are not acting in any way fundamentally different from Leptothorax larvae surrendering their personal salivary secretions. Overall, the question of queen control is moot with reference to conflict and the queen's effect on the rest of the colony.

We now turn to evidences of worker control. In Chapter 4 it was shown how a conflict does arise in monogynous colonies between the queen and her daughter workers concerning the optimal ratio of investment in new male and female reproductives, and how the workers settle this dispute to their own advantage. There are many other circumstances in which the workers are usually far removed from any possible control of the queen other than pheromonal signals, and hence they are able to make decisions by themselves on a moment-by-moment basis. One of the most clear-cut is the harvesting and distribution of food. Fire ant workers (Solenopsis invicta) are probably typical of this caste in ants in the way that they partition the different kinds of food. During experiments conducted by Howard and Tschinkel (1981), foragers and nurses passed sugar to other workers, amino acids to larvae and the queen, and soy bean oil to larvae and other workers in equal amounts. During the elimination of supernumerary queens of Solenopsis invicta and Monomorium pharaonis, the workers select the losing contenders and execute them (Petersen-Braun, 1982; Fletcher and Blum, 1983a).

A particularly interesting case of queen selection by workers seems to occur in army ants. Franks and Hölldobler (1987) argued that during colony fission, workers should, in theory at least, select the queen that would enable them to maximize their own inclusive fitness (Macevicz, 1979; Franks, 1985). This calculation is complicated by three factors: (1) workers that accompany sister queens, rather than their mother queen suffer the fate of raising nieces and nephews rather than brothers and sisters; (2) at some stage in the life of a colony, workers have to reject their maternal queen on the grounds of her senility; and (3) virgin queens may not be full sisters of the workers, because army ant queens probably mate more than once in their life time. If all the virgin queens and all the workers were to be full sisters, then the workers should unanimously select the potentially most fertile queens, preferring their mother to a sister. However, if the maternal queen mates more than once in her life time, as has been suggested for both Eciton burchellii (Rettenmeyer, 1963) and species of Dorylus (Raignier and Van Boven, 1955), then each patrilineal group of workers should prefer that their own full sister become one of the new queens. However, this alternative would be possible only if workers can discriminate between full and half sisters. That degree of kin recognition has been suggested in honeybees (Getz and Smith, 1983; Page and Erickson, 1984; Noonan, 1986; Visscher, 1986). It has also been claimed that colony division in honeybees is associated with the segregation of workers into sororities (Getz et al., 1982). Furthermore, worker bees are known to distinguish between individual queens on the basis of their odors (Boch and Morse, 1974, 1979). Breed (1981) has shown that the rate of acceptance of foreign queens is correlated with the degree of genetic relationship among the queens involved in the transfers. However, few studies have been conducted on whether ants can discriminate between individual nestmates, on the basis of the relatedness, so that very little can be concluded at the present time (see Carlin et al., 1987b). Furthermore, although army ant queens may mate with two or more males, they may do so only once each year (Rettenmeyer, 1963a). It is thus possible that new queens and the majority of the worker population are full sisters.

At present we do not know if kinship influences how workers choose new queens. What is clear, however, is that whether workers are able to recognize their full sisters or not, there will be very strong selection for workers to discriminate between queens on the basis of the potential fertility and survivorship.

Army ant queens have to be both exceptionally vigorous and productive. Eciton burchellii queens, for example, may live six years, during which time they produce some three million workers and walk between successive bivouacs a total distance of 60 kilometers (Franks, 1985). Because of the huge worker mortalities during foraging, army ant colonies grow relatively slowly, and three years on the average elapse between bouts of sexual reproduction. The workers must select highly fecund and long lived queens in order to realize any inclusive fitness at all.

Studies on colony foundation by multiple queens in other species of ants suggests that workers are skilled at choosing the most fertile and attractive queens, and that kinship is of minor importance in the rejection of supernumerary queens (Bartz and Hölldobler, 1982). Although almost nothing is known about the genetic constraints on worker choice of queens in army ants, a considerable amount is known about the proximate mechanisms by which workers discriminate between queens.

Schneirla (1956a) observed conflicts between workers who were associated with different virgin queens even when the latter were still maturing larvae. He noted that individual queen larvae are often separated by considerable distances within the bivouacs and each larva is surrounded by a cluster of "satellite workers." Adjacent worker groups can come into conflict, occasionally resulting in the deaths of some potential queens. Schneirla (1971) further observed that the first queens to emerge are more attractive and likely to be successful. The first young females will have had longer to produce their pheromones and to win the allegiance of the workers. That workers may come into conflict over the allegiance to queens does not necessarily mean that they are forming sororities. Such conflict may be a mechanism by which workers can compare the strength of their advocacy for certain queens and hence the queens' attractiveness. Therefore such competition might have resulted purely from colony level selection. The phenomenon of workers changing their queen-allegiances has been demonstrated by experiments in which the old queen was removed from the bivouac when the sexual larvae were mature. Under these circumstances the parental queen will be readmitted to her colony in an unequivocal manner only if she is met by the group of workers previously affiliated with her. Otherwise she may be segregated in a tight cluster of workers and eventually abandoned (Schneirla and Brown, 1952; Schneirla, 1956a,b). On the other hand, Eciton colonies will accept workers of alien colonies only if the transplanted workers have been isolated from their queen for a number of days (Schneirla, 1971). After a few days a colony deprived of its queen will fuse with a colony of the same species possessing a viable queen. Thus, it appears that queen pheromones unite and coordinate the huge army ant society. Furthermore, the chemical basis of the queen's attractiveness and signature was demonstrated by a simple experiment: workers were more attracted to paper discs upon which the queen had been previously sitting than to control discs (Watkins and Cole, 1966). These various observations provide strong circumstantial evidence that army ant workers can recognize and choose particular queens during colony fission, making their selection on the basis of relative queen attractiveness. But even in such cases, it is difficult to conclude that the workers are really exerting their will over that of queens. Even when they execute supernumeraries, they work in the interest of the winning queens and not just themselves.

Aggression and dominance summarized

The evidence concerning conflict within ant colonies can be summarized overall as follows. Under many but not all circumstances there is direct aggression among queens, resulting in either dominance hierarchies, oligarchies by a few queens spaced through the nest, or the eviction or execution of supernumeraries. Dominance is not always based on physical domination. It is sometimes mediated by pheromones or some other still undiscovered, circuitous signals. Often it takes the form of the inhibition of ovarian development in subordinates who are otherwise left unmolested. When the laying queen is removed in a few species, virgin queens present at the time then contend for dominance.

Similarly, there is often clear-cut dominance behavior among workers. In rare cases hierarchies leading to differential reproduction occur among members of this caste even in the presence of laying queens. More often, the phenomenon emerges when laying queens are absent.

Conflict between queens and workers is less certain. In monogynous species, an inherent conflict arises between the two castes in the ratio of investment in new queens and males, and the difference has been resolved evolutionarily in the favor of the workers. But in other cases, the interactions between queens and workers are too ambiguous to be clearly labeled as "queen control" or "worker control." Even such apparently explicit behaviors as ovarian suppression in workers and execution of supernumerary queens might well benefit the inclusive fitness of both castes simultaneously. It is a mistake to interpret individual forms of interaction in vertebrate terms, in which dominant individuals are able to raise their genetic fitness. "Control" and dominance in ant colonies, at least between the queen and worker caste, must be viewed within a far tighter, more complex social context where (as experience has taught us) first appearances are often very deceptive.

Hölldobler, B. and Wilson, E. O. 1990. The Ants. Cambridge, Mass. Harvard University Press. Text used with permission of the authors.

The Ants - Table of Contents