The Ants Chapter 14

CHAPTER 14. SYMBIOSES BETWEEN ANTS AND PLANTS

Odoardo Beccari, in his pioneering monograph on myrmecophilous plants (1884), reported that the East Indian pitcher plant Nepenthes bicalcarata harbors ant colonies in the hollow stem of the same pitcher-shaped leaf by which it captures and digests other kinds of insects (Figure 14-1). The ants are free to roam over the carnivorous plant and adjacent terrain, gathering insects and other food items of their own. If this relationship is verified, the ants and the plant appear to be engaged in a trade-off of mutual benefit (Jolivet, 1986). The ants risk being eaten by the plant but they get a home; the plants surrender some tissue space and insect prey to the ants but they gain some protection from herbivores.

The Nepenthe] story is only one, admittedly very peculiar case among hundreds of ant-plant symbioses documented during the past 150 years of research. This topic has been the subject of rich and informative reviews during the past ten years, including systematic accounts of the plants (Huxley, 1980, 1982; Jolivet, 1986), an ecological analysis of the proven cases of ant-plant mutualism (Beattie, 1985; Benson, 1985; Huxley, 1986), and a brief summary of all aspects of the symbioses with a bibliography complete through 1981 (Buckley, 1982a-c).

The varieties of ant-plant symbioses
The angiosperms (flowering plants) and the ants have been closely associated throughout most of their respective histories. By the middle of the Cretaceous Period primitive sphecomyrmine ants were on the scene, while the angiosperms were diversifying and spreading around the world as the newly dominant form of terrestrial vegetation. An intricate coevolution of the two groups probably began during this time. Many of the plant species had come to depend on insects for pollination, and an even greater number of insect species subsisted on nectar and pollen obtained during the pollination process. A legion of other insects fed on the foliage and wood of the angiosperms. Plants responded in turn by evolving various combinations of thick cuticles, dense spines and hairs, and secondary defense substances such as alkaloids and terpenes.

Into this lively theater of coevolution the ants entered. As the Cretaceous drew to a close, the ants increased in diversity and abundance, seized new roles as pollinators and seed dispensers, and appropriated the plants as domiciles. An entomologist returning to early Eocene times, about 60 million years ago, would find familiar-looking ants swarming over familiar-looking vegetation.

Complex symbioses have been fashioned among the thousands of species of ants and plants. Often these relationships are parasitic, with one exploiting the other and giving nothing in return. In other cases they are commensalistic, with one partner making use of the other but, as in the case of ants occupying hollow stems, neither harming nor helping it. But of maximum scientific interest, some symbioses appear to be mutualistic; in other words, both partners benefit from the association. To put the matter as briefly as possible, ants use cavities supplied by the plants for nest sites, as well as nectar and nutritive corpuscles given them as food. They in turn protect their plant hosts from herbivores, distribute their seeds, and literally pot their roots with soil and nutrients. There is abundant evidence, which we will review shortly, that some pairwise combinations of ants and plants have coevolved so that each is specialized to use the other's services. This mutualistic linkage has produced some of the most elaborate adaptations known in nature.

Before taking up these several categories and the evidence they provide for coevolution, let us define the specialized terms that have grown out of the study of ant-plant mutualism. The definitions below represent the concensus that we believe can be drawn from current usages.

Ant garden. A cluster of epiphytic plants inhabited by ant colonies. To qualify as a true ant garden, the plants must benefit from the association. True ant gardens are known from both the Asian and New World tropics.

Ant plant. Also known as a myrmecophyte; a species of plant with domatia, or specialized structures for housing ant colonies.

Arils. Same as elaiosomes (q.v.).

Bead glands. Same as pearl bodies (q.v.).

Beccarian bodies. The pearl bodies (a special kind of food bodies) produced by the stipules or young leaves of the Old World tropical genus Macaranga, and consumed by resident ants.

Beltian bodies. The food bodies found on the tips of the pinnules and rachises of some New World species of Acacia, and consumed by the resident Pseudomyrmex.

Domatia. Also called myrmecodomatia; specialized structures, such as inflated stems, evolved by ant plants for the housing of ants.

Elaiosomes. Also called arils; specialized nutritive attachments on ant-dispersed seeds. Stimulated by the attractants, the ants transport the seeds to new locations, discard them after feeding on the elaiosomes, and hence aid in the dispersal of the seeds.

Extrafloral nectaries. Secretory organs, often no more than small patches of tissue, that produce sugary secretions, which possibly contain amino acids, attractive to ants and other insects. By definition, extrafloral nectaries are not involved in pollination, although they may occur on the flower outside the perianth.

Food bodies. Special nutritive corpuscles evolved by ant plants to feed ants; particular kinds of food bodies include Beltian bodies, Müllerian bodies, and pearl bodies.

Müllerian bodies. Food bodies produced by Cecropia trees on the trichilium (a pad at the base of the petiole) and consumed by the resident ants, which are usually Azteca but occasionally Camponotus balzani or Pachycondyla luteola.

Myrmecochory. The dispersal of seeds by ants stimulated by nutritive bodies (elaiosomes) or special seed attractants. Dispersal of seeds by granivorous ants without the aid of such specialized attractants is not included.

Myrmecodomatia. Domatia (q.v.).

Myrmecophily. The general condition of encouraging ants. In the system proposed by van der Pijl (1955) and Beattie (1985), myrmecophily is used to denote ant pollination. However, Jolivet (1986) and other authors have used the expression to refer to the condition of being an ant plant, in other words possessing domatia or the same as myrmecophytism. Since the presence of domatia is so often accompanied by extrafloral nectaries, food bodies, and other coadaptive traits, it seems more appropriate to use the expression “myrmecophilous” in the broadest sense, comprising both pollination and myrmecophytism and all the accouterments they use to attract and award ants.

Myrmecophytism. The condition of being an ant plant, in other words possessing ant shelters (“domatia”).

Myrmecotrophy. The transport of soil, litter, and other nutrient-bearing materials by ants that results in the feeding of the plant hosts.

Pearl bodies. A heterogeneous group of food bodies with a pearl-like luster and a high concentration of lipids. Sometimes called bead glands.

Protectionism and the Acacia case
In spite of the obvious intimacy of the associations between many of the tropical plant species and their ants, and the extraordinary anatomical features of the plants that seem to have no other function than to serve their guests, biologists for many years disagreed about the significance of the association. On one side stood the “protectionist school,” to use W. L. Brown's (1960a) expression. It was founded by Thomas Belt, whose observations on the ant acacias in A Naturalist in Nicaragua were the beginning of serious studies of myrmecophytism. Belt, and a majority of subsequent writers, including particularly Schimper, Wasmann, and the leading authority on the bull's-horn acacias, W. E. Safford, agreed that the ants provide the plants protection against their natural enemies. They also postulated that in the course of their evolution the acacias developed the hollow thorns, Beltian bodies, and foliar nectaries as devices to promote the welfare of the ants (Figures 14-2 and 14-3). In short, the protectionist authors believed the symbiosis to be mutualistic. The opposing “exploitationist school,” represented chiefly by Elisabeth Skwarra (1934) and William Morton Wheeler (1942), argued that only the ants benefited and that the various “myrmecophilous” structures of the acacias serve some other, still unknown function. This opposition of viewpoints, which are rather oversimplified as we have expressed them here, extended to discussions of other genera of ant plants as well.

Brown, in crystallizing the issue in 1960, developed new evidence favoring the protectionist hypothesis. He pointed out that Acacia is a very old and widespread genus containing no fewer than 700 species. Australia contains a majority of the species as well as the greatest phyletic diversity of any continent. It also contains one of the richest ant faunas of any comparable area in the world. Yet not a single potential myrmecophyte has been found among the acacias of Australia. Moreover, the Australian species have mostly lost their siniform stipules, in striking contrast to their congeners in other parts of the world. This geographic distribution of characteristics agrees with the known occurrence, in the recent geological past and at present, of large and effective faunas of browsing mammals. Brown inferred, in accordance with Belt's hypothesis, that the development of myrmecophytism and spininess in the African and New World Acacia species represents an adaptive response to the presence of such mammals. In other words, they must provide an effective deterrent to browsing. In Australia, where advanced browsing faunas have been unknown (at least in the recent geologic past), the species have either failed to develop myrmecodomatia and spines or else lost them secondarily.

It remained for Janzen (1966, 1967, 1969), in a brilliant field experimental study in Mexico, to prove directly that the ants do indeed provide vital protection to the bull's-horn acacias. While conducting a pilot survey Janzen noted, as Bequaert (1922) had found earlier for Barteria fistulosa in the Congo, that Acacia cornigera shrubs and trees devoid of ants suffer greater damage from attacks by phytophagous insects than do their neighbors harboring ant colonies. They also tend to be overgrown by competing plant species. When Janzen removed the ants (Pseudomyrmex ferrugineus) with any one of three treatments, namely spraying with parathion, clipping the thorns, or extirpating entire occupied branches, he found that the acacias became decidedly more vulnerable to attack by their insect herbivores. Coreid bugs and membracids sucked on the shoot tips and new leaves; scarabs, chrysomelid beetles, and assorted caterpillars browsed on the leaves; and buprestid beetle larvae girdled the shoots. Moreover, other plants grew in more closely and shaded the stunted shoots. In nearby control trees, still occupied by Pseudomyrmex colonies, Janzen observed that the ants attacked the invading insects, in the great majority of cases successfully killing them or driving them off. Alien plants that sprouted within a radius of 40 cm of the occupied acacia trunks were chewed and mauled by the ants until they died. Other plants whose leaves or branches touched the canopy of the acacia were also attacked. Up to one-fourth of the entire ant population were active on the surfaces of the control plants, day and night, constantly patrolling and cleaning them. For the full year during which the experiment was continued, the biomass and growth of the unoccupied acacias steadily fell below that of the occupied ones. In the end it seemed unlikely that they could survive much longer, let alone bear seeds. Thus, Belt's view, that the Pseudomyrmex “are really kept by the acacia as a standing army,” was substantially confirmed.

Although experimental evidence is still lacking, it seems probable that the ants are also effective against browsing mammals. The Pseudomyrmex workers are extremely aggressive toward intruders of all sizes. They become alert at the mere smell of a cow or a man, and, when their tree is brushed or shaken, they swarm out and attack at once. Their stings are very painful, causing a lasting burning and throbbing effect. To brush against an occupied acacia, and thus to acquire a group of vicious, stinging ants on an arm or a leg, is a sensation very much like walking into a large nettle plant.

According to Janzen, Pseudomyrmex ferrugineus is an obligate plant ant that occupies at least five species of Acacia (chiapensis, collinsii, cornigera, hindsii, and sphaerocephala). Its life cycle conforms to the basic claustral pattern of ants generally. After the nuptial flight, which can occur in warm weather in any month of the year, the queen alights, sheds her wings, and searches for a nest site. For a Pseudomyrmex ferrugineus there can be only one such place: an unoccupied acacia thorn. If the thorn has not already been opened by a previous occupant, the queen gnaws a circular hole near the tip of the spine and enters. Then she lays 15-20 eggs and rears her first brood while remaining secluded in the thorn cavity. Although the exact duration of brood development is not known, it is evidently relatively short for an ant species, and the worker population increases at a rapid rate. Within seven months there are about 150 workers, and, three months later, twice this number. The worker population increases to about 1,100 in two years and to over 4,000 in three years. The largest colony collected by Janzen contained 12,269 workers and a single queen. In old colonies the queen is physogastric, heavily attended by workers, and accompanied by masses of hundreds of eggs and young larvae. The production of males and virgin queens begins during the second year and proceeds continuously thereafter. Workers belonging to the youngest colonies leave the protection of the thorn home only long enough to gather nectar and Beltian bodies and, at rare intervals, to take possession of nearby thorns. When their numbers reach 50 to 100, they begin patrolling the open plant surface in the vicinity of the nest thorns. When the population size reaches 200 to 400, the workers become more aggressive and start attacking and destroying other, smaller colonies in nearby thorns. They also become increasingly effective in warding off phytophagous insects that attempt to land in the vicinity. Finally, the dominant colony takes possession of the entire tree, wiping out all competitors in the process. A few colonies are also able to extend their territories to other acacias nearby.

The Pseudomyrmex ferrugineus colonies appear to subsist primarily on the Beltian bodies and foliar nectar obtained from the host trees. The larvae are fed in part on unaltered fragments of Beltian bodies in the following peculiar manner. The nurse worker first pushes the fragment deep within the trophothylax, the special food pouch located on the lower surface of the thorax just behind the head (and found only in pseudomyrmecine larvae; see Petralia and Vinson, 1979a). The larva then starts to rotate its head in and out of the trophothylax, chewing and swallowing the contents. Simultaneously it ejects a droplet of clear fluid, possibly containing a digestive enzyme, into the trophothylax. If the Beltian fragment protrudes from the opening of the pouch, a worker may remove it, cut it up, and redistribute it. From time to time workers also force open the trophothylax and regurgitate droplets of fluid into it. Whether this material consists of elementary crop fluid or some more specialized form of nutritive secretion is unknown. Occasionally the Pseudomyrmex workers succeed in capturing insect prey on the nest tree. It is possible that these too are fed to larvae, but they can form a source of protein only secondary in magnitude to the Beltian bodies.

The balance of mutualism
It is reasonable to conclude from existing evidence that the anatomy and physiology of the ant guests and plant hosts have been modified at least occasionally to enhance the mutualism in a manner consistent with the theory of evolution by natural selection. And as Beattie (1985) has said, the majority of ant-plant mutualisms evolve in response to selection, especially stress selection, on the plants rather than the ants.

The participating ant species, which vary in time and space, respond to plant rewards on a facultative basis, and traits evolved in specific response to plant rewards are infrequent. Thus, when the definition of coevolution requires reciprocal selection involving heritable traits, coevolution is remarkably difficult to demonstrate in ant-plant mutualisms and appears to be the exception rather than the rule. In fact, the evidence suggests that it is directional selection that drives the vast majority. This does not diminish the importance of their effects. The speed and energy with which ants harvest rewards such as elaiosomes suggest that they are crucial to the economy of ant colonies. This may be especially true when the colony is stressed, either by the external environment or by major internal demographic events such as reproduction. For the plants, we have repeatedly seen that ant services are potent forces that increase plant fitness. In many cases the failure of ant services leads to a variety of demographic failures, including poor seed set and poor seedling recruitment. Populations may contract, crash, or even become extinct as a result. Ant services, either on a continuous basis or as a density response, are crucial to a wide variety of plant species worldwide.

Essentially the same conclusion was reached by Longino (1986) for the ant gardens of Central and South America and Huxley (1980) for epiphytic ant plants generally. Similarly, Davidson et al. (1988) present evidence that ants nesting exclusively on ant plants have restricted diets, refusing baits accepted by many other kinds of ants. In a symmetrical manner, Asian species that have invaded introduced Cecropia species (hence, are not adapted to them) ignore the Müllerian bodies on which coevolved ant species feed in the American tropics.

Ants protect plants
Throughout their ranges, ants forage on vegetation in large numbers, searching for arthropod prey. The insects they collect include hemipterans, beetles, sawfly larvae, caterpillars, and other herbivores. Wood ants, in particular Formica aquilonia, Formica polyctena, and Formica rufa, are so effective as predators that they have been utilized in Europe for centuries to control forest pests. Entomologists have developed an entire technology since the late 1940s to culture and transplant the ants, and it has often proved cost-effective (Gösswald, 1951a, 1985; Grimalski, 1960; Khalifman, 1961; Adlung, 1966; Gösswald and Horstmann, 1966; Smirnov, 1966; Bradley, 1972; Finnegan, 1975, 1977; Skinner and Whittaker, 1981). An ordinary colony of Formica polyctena was observed to collect about six million prey items from one-third of a hectare per year (Horstmann, 1972, 1974), while one colony of Formica rufa monitored by Strokov (1956) gathered 21,700 sawfly larvae and moth caterpillars in a single day. Similarly dramatic data have been published for polyctena, pratensis, and rufa by Sörensen and Schmidt (1987). Formica polyctena was so proficient at protecting mountain birches against an outbreak of the geometrid moth Oporinia autumnata that green islands of intact trees 40 meters in diameter were left around each nest in the midst of a grey, mostly defoliated forest (Laine and Niemelä, 1980).

There is a potential debit side in the ledger of ant protection. When seed dispersers, such as mammals and birds, are driven away from trees by the resident ants, the fruit and seeds might fall to the ground undispersed. Just this effect was documented by Thomas (1988) in the case of West African fig trees by weaver ant colonies (Oecophylla longinoda).

Plants shelter ants: myrmecophily
The strongest evidence for ant-plant mutualism comes from the existence of domatia, or plant structures that serve no evident purpose other than to shelter ant colonies. Domatia increase the density of ants on the plant itself. It is easy to see how the phenomenon could arise in evolution. Ants are always quick to take advantage of whatever hollows and crevices plants have to offer. In most cases the shelters are not true domatia. They are adventitious, and the ants therefore live as parasites or commensals on the plants. These incidental nest sites can be divided into convenient categories as follows:

• Preformed cavities in live branches and stems, excavated by woodboring beetles and other insects and later occupied by ant colonies belonging to such diverse genera as Daceton, Podomyrma, Crematogaster, Azteca, Lasius, and Camponotus.

• Cavities in stems and branches that are naturally hollow or contain a pith soft enough to be easily excavated by ants. A legion of grasses, sedges, composites, and other herbaceous and shrubby plant forms provide this form of refuge and a great variety of ants occupy them. A typical field of sedges and weeds in Florida, for example, contains dense populations of one or more species of Pseudomyrmex, Colobopsis, Crematogaster, and Monomorium. The low trees and bushes of a Brazilian cerrado contain the same genera along with several species of Zacryptocerus.

• Natural or preformed cavities in bark. The bases of pine trees in the southern United States shelter an entire fauna of Hypoponera, Pheidole, Solenopsis, Crematogaster, Brachymyrmex, and other ant genera that nest adventitiously in the bark. Higher up can be found colonies of the small myrmicines Leptothorax bradleyi and Leptothorax wheeleri, which are specialized to this environment to some extent (Wilson, 1952). An ant restricted entirely to this microsite is Melissotarsus titubans, which occurs in the bark cavities of living trees in tropical Africa and Madagascar. Its body form and locomotory behavior are modified for existence in tight spaces. The workers walk with their middle legs held upright and touching the roofs of the galleries. If placed outside, in the open, they are unable to move around in a normal manner (Delage-Darchen, 1972a).

• Roots of epiphytes. The tangled root systems of orchids, gesneriads, and other tropical epiphytes are ideal nest sites for ants. One of the most profitable ways to collect ants anywhere in the world is to hold an epiphyte over a pan or ground cloth and strike the root system several times with a trowel. In Central and South America, this technique yields large numbers of colonies of Hypoponera, Gnamptogenys, Strumigenys, Nesomyrmex, and other genera, many belonging to rare or previously undescribed species.

• Ants frequently nest in galls formed by cynipid wasp larvae; the phenomenon has been observed in Europe (Torossian, 1972; Espadaler and Nieves, 1983) and North America (Wheeler, 1910a). Leptothorax obturator of Texas appears to be a specialist on this nest site (Longino and Wheeler, 1987).

• A few ant species, such as the large ponerines Ectatomma tuberculatum and Paraponera clavata of the New World tropics, construct earthen or carton nests vertically against the sunken portions of tree trunks. Thus the tree provides a partial wall of solid wood that is virtually invulnerable. Paraponera clavata prefers trees of the abundant legume Pentaclethra macroloba, but the relationship is not obligatory. It is possible that the ants benefit from both the well-formed buttresses of the Pentaclethra and the extrafloral nectaries in the foliage of this species (Bennett and Breed, 1985).

None of these diverse structures appears to be “designed” to accommodate ant colonies. All are ordinary anatomical features of the plants that the ants exploit, apparently in a unilateral manner. In contrast, the domatia listed comprehensively in Table 14-1 do appear uniquely to serve as ant nests. They are characterized by cavities that form independently of the ants (even in greenhouses, where no ants are present), adventitious roots and tubercles that absorb nutrients from waste material carried onto the cavities, and even holes or thin windows of tissue through which ants can more conveniently enter and leave. Some of the most distinctive forms are illustrated in Figures 14-4 through 14-8. Domatia are almost always occupied by ant colonies in nature. Furthermore, plant species with the most complex domatia are typically occupied by only one or a small number of species specialized to live with them. Finally, species with domatia usually also manufacture food bodies, which are unique structures with no known function other than the feeding of ants (see also Table 14-2). In short, strong circumstantial evidence indicates that domatia are structures specialized in evolution to promote symbioses with ants. Benson (1985) has suggested that the ant domatia of many ant species evolved from the sheltered feeding sites used by mealy bugs and other homopterous insects. Because of the honeydew produced by the homopterans, ants were attracted to these sites, which were enlarged and otherwise structurally modified into ant domatia.

Further evidence of coevolution is provided by the legendary ferocity of many of the guest ants. The vast majority of Pseudomyrmex species not occupying domatia are timid and flee even when their nest is broken apart. In sharp contrast, Pseudomyrmex triplarinus, an obligate resident of Triplaris americana, falls upon any intruder touching the nest tree without hesitation or mercy. To be stung by several of these ants within a few seconds is like encountering a nettle--you pull back at once. Or conversely, if you want to locate Triplaris quickly in an Amazonian forest, shake one sapling after another until one produces a swarm of the stinging ants. The Pseudomyrmex also attack and remove intruding insects, and they are significantly more efficient than the Crematogaster that also occupy Triplaris (Oliveira et al., 1987a). The species of Camponotus offer a similar dichotomy. Most retreat or offer limited resistance when the nest is disturbed by a human being. Possibly the shyest ant species in the world is the Amazonian Camponotus paradoxus, whose workers disperse and hide so quickly that it is difficult to catch any specimens at all. At the opposite extreme is Camponotus femoratus, an obligatory resident of epiphytic ant gardens in South America. Diane Davidson (personal communication) describes its behavior as follows:

When I approached to within 1-2 m of their nests, workers of this species typically began to run back and forth and frequently jumped or fell onto me. Workers of all size classes of this polymorphic species attempted to bite, but usually only the major castes were capable of breaking the skin with their mandibles and causing a stinging sensation by simultaneously biting and spraying formic acid into the wound. In addition, these workers often exhibited a second type of apparently aggressive behavior, which I will term “coughing” behavior. With mandibles held wide open and the prothoracic legs upraised, they brought their legs down abruptly in a jerking movement that resembled a cough.

Tetraponera tessmanni, the obligate pseudomyrmecine tenant of the verbenaceous creeper Vitex staudtii in West Africa, was described by Bequaert (1922) as “exceedingly vicious and alert. When its host plant is ever so slightly disturbed, the workers rush out of the hollow stalks in large numbers and actively explore the plant. Their sting is extremely painful and sometimes produces vesicles on the skin.” Even more redoubtable is the African pseudomyrmecine Tetraponera aethiops, the obligate tenant of the small flacourtiaceous tree Barteria fistulosa:  “As soon as any portion of their host plant is disturbed, they rush out in numbers and hastily explore the trunk, branches, and leaves. Some of the workers usually also run over the ground about the base of the tree and attack any nearby intruder, be it animal or man. All observers agree that the sting of the Tetraponera is exceedingly painful and is felt for several hours. Its effects can best be compared with those produced by female velvet ants.” The ant is feared by the natives of the Congo, who try to avoid the unpleasant task of cutting the small Barteria trees scattered through the forest. As a consequence individuals of Barteria fistulosa are often found standing by themselves in the center of clearings or near the sides of forest paths. The species is also abundant in secondary forest growth.

Not all myrmecophyte ants are this formidable. A few, such as the tiny Pheidole species that occupy Maieta and Piper, seem incapable of defeating most herbivores. Many authors have commented on the evident ineptness of these ants in the protection of their adopted plants. However, Letourneau (1983) points out that such ants can perform their service by removing the eggs and early developmental stages of the herbivores rather than face down the adults. She showed that Piper plants in Costa Rica occupied by Pheidole bicornis, their principal ant guest, suffered less damage than those deprived of the ants. The workers preferred to patrol new leaves, which are the most susceptible to insect damage. When Letourneau placed termite eggs on Piper bushes, the ants discovered more than 75 percent and dropped them off the plants within an hour. Risch et al. (1977) also found some evidence that the Pheidole chew through or push aside alien vines from their host plants, and they postulate that the ants also bring nutrients to the plant cavities as part of their nest material. No one has assayed the relative importance of these benefits to trees.

Competition for the ant plants is intense. Young plants are soon fully occupied by colony-founding queens of plant-ant species. It is common to find different internodes of very young Cecropia saplings occupied by one or more queens of Azteca. Sometimes the inhabitants belong to two species. As many as a dozen colonies are started in this manner in each tree. When the first workers emerge, they cut holes through the septa separating the internodes. Fighting and other forms of competition ensue, and all of the young colonies except one are either destroyed or perhaps even assimilated, so that in the end only one large colony and a single nest queen survive (Dan Perlman, personal communication). The same reductive sequence occurs in Pseudomyrmex ferrugineus, a resident of swollen-thorn acacias (Janzen, 1967), as well as Pseudomyrmex triplarinus, an obligate resident of Triplaris (Schremmer, 1984).

The pervasiveness of competition is indicated by the patchiness of distribution of the ants among plants of the same species. At the Manu National Park, Peru, Davidson et al. (1988b) found eight myrmecophyte species. Among 130 plants dissected, 127 contained ant colonies, and of these 126 belonged to one species only. In many instances the structure of the plant was found to bias which ant species can colonize it. Maieta guianensis, for example, has dense epidermal hairs (trichomes) covering the leaves and stems. Its usual guest ants, the tiny Pheidole minutula, are able to walk through the hairs without difficulty. A second and less common ant, a Crematogaster belonging to the victima group, apparently has to cut trails through the trichomes (Diane Davidson, personal communication). In addition, the narrow tunnel entrances leading into the pouch-like domatia are readily entered by the Pheidole queens during colony founding but not by the Crematogaster queens, which are forced to chew holes into the plant. Allomerus demararae enjoy a similar advantage in the occupancy of the myrmecophyte Cordia nodosa over species of Crematogaster and Azteca, which must cut trails through the trichomes just to move around (Davidson et al., 1988b). Put briefly, certain ant species occupy individual myrmecophytes predominantly and other species intrude occasionally, but only one species is found in each plant. The final result of this competitive pressure is that the myrmecophytes are saturated with ants.

Ant gardens
Perhaps the most complex mutualism between plants and ants is the ant garden, which is an aggregate of epiphytes assembled by ants. The ants bring the seeds of the epiphytes into their carton nests. As the plants grow, nourished by the carton and detritus brought by the ants, their roots become part of the framework of the nests. The ants also feed on the fruit pulp, the elaiosomes (food bodies) of the seeds, and the secretions of the extrafloral nectaries.

It is a curious fact that while epiphytic myrmecophytes are generally diverse and abundant throughout the tropics, they form ant gardens principally in Central and South America, with some examples having been recently found in tropical Asia by Diane Davidson (personal communication). First reported by Ule (1902), the gardens are typically round or ellipsoidal, and they range from 6 to more than 60 centimeters in greatest diameter; a typical example is shown in Figure 14-9. The ants construct irregular nest chambers divided by carton walls among their roots (Weber, 1943; Kleinfeldt, 1978, 1986). Specialized garden plants representing no fewer than 16 genera have been identified to the present writing (Buckley, 1982b; Davidson, 1988). At one locality in the Manu National Park alone, Davidson identified ten such specialists belonging to the following seven families: Araceae (Anthurium, 2 species; Philodendron), Bromeliaceae (Neoregelia, Streptocalyx), Cactaceae (Epiphyllum), Gesneriaceae (Codonanthe), Moraceae (Ficus), Piperaceae (Peperomia), and Solanaceae (Markea). Peperomia macrostachya was by far the dominant species, occurring in 76 percent of the gardens. It was followed by Anthurium gracile and Ficus paraensis, with 29 and 23 percent occurrence rates respectively. Seven other genera in six families were represented among the unspecialized, adventitious species in the gardens. In general, the dominant plants of gardens in South and Central America are five species of Codonanthe, followed in abundance by Aechmea mertensii and Anthurium gracile'' (Kleinfeldt, 1986).

The dominant ants in the ant gardens across Central and South America generally are members of the genera Crematogaster, Solenopsis, Azteca, Monacis, and Camponotus, with Anochetus and Odontomachus occurring much less commonly (Kleinfeldt, 1986). The overwhelmingly dominant species in the Peruvian Amazon are Crematogaster parabiotica (broadly defined) and Camponotus femoratus. These forms appear to be the most abundant ant species in the forest canopy generally (Wilson, 1987a; Davidson, 1988). Other, possibly adventitious ant-garden species belong to the genera Odontomachus, Anochetus, Monomorium, and Solenopsis (Mann, 1912a; Wheeler, 1921a; Macedo and Prance, 1978). Monacis debilis is commonly found in the same garden with one sibling species in the Crematogaster parabiotica complex and Camponotus femoratus with another. The species pairs nest in separate but contiguous chambers within the same garden, the condition originally called parabiosis by Forel (1898). The Monacis at least follow the Crematogaster out along their odor trails and aggressively displace them at food sites (Swain, 1980).

The tightness of the associations suggests that the plants and ants in the gardens have coevolved, but surprisingly little experimental evidence has been adduced to test this supposition. Kleinfeldt (1978) showed that Codonanthe crassifolia, one of the specialized garden plants, grows more quickly when in association with Crematogaster longispina than when alone. It is probable that the carton of ant nests provides a physical substrate laced with nutrients preadapting them for epiphytic growth. Longino (1986) observed that Crematogaster longispina occurs throughout the Atlantic rain forest of Costa Rica, often as a dominant element of the understory fauna. Its large, diffuse colonies manufacture loose, coarse-fibered carton to build large numbers of nests, shelters for scale insects, and galleries that connect all of these components. The bulk of the carton is placed under and around the roots and stems of aroids and gesneriads running up and down the tree trunks. Whenever new carton is laid down, there is usually a flush of newly sprouted epiphytes in the walls. As Longino points out, it is but a short further step in evolution to nurture an obligate garden species such as Codonanthe crassifolia.

How do individual symbiotic gardens get planted in the first place? Ule (1905, 1906) observed that ants retrieved the seeds of the epiphytes, and he proposed that they establish the gardens within the confines of their own nests to start the symbiosis. Wheeler (1921a) argued that the gardens might be autonomous instead, with the ants colonizing the plants later. The evidence has consistently favored Ule. Later authors, including Madison (1979), Kleinfeldt (1978), and Davidson (1988), have repeatedly observed that the symbiont ants are strongly attracted to the fruits and seeds of the garden epiphytes. After the workers consume the adhering fruit pulp and elaiosomes, they place the seeds near their own brood piles. Workers of Camponotus femoratus have been observed to build epiphytic gardens over the carton chambers of Crematogaster parabiotica in this manner. Davidson notes that the myrmecophyte seeds remain attractive even after they pass through the digestive tracts of frugivorous bats. She suggests that they contain pheromone-mimicking substances rather than just food material. Seidel (1988) has identified several of the compounds as 6-methyl-methylsalicylate, benzothiazole, and a few phenyl derivatives and monoterpenes. Many precedents exist in the behavior of myrmecophilous arthropods that gain entrance to their hosts' nests by pheromonal mimicry (see Chapter 13). It may also be significant that at least some ant species not associated with gardens are repelled by the seeds. Finally, Davidson (personal communication) tested Wheeler's competing hypothesis by setting out garden epiphytes in sites where they could be found by the symbiotic ants. None was colonized.

To summarize our existing knowledge of the ant gardens, the epiphytes restricted to the gardens appear to be truly adapted to this symbiosis. Their seeds are transported to favorable sites by the ants in response to what appear to be specialized attractive substances, and the subsequent growth of at least some of the species is enhanced by the presence of the ants. For their part, the ants are not so clearly adapted to benefit from the ant gardens. They feed on the extrafloral nectar, fruit pulp, and elaiosomes supplied by the plants. But this is not a restrictive diet, and in fact all of the garden ant species forage away from the gardens. The best evidence that some ants have coevolved with the plants is that the dominant garden species of South America, Monacis debilis, Camponotus femoratus, and the two Crematogaster in the parabiotica group, usually if not invariably nest in the gardens. They are behaviorally specialized to some extent for bringing the seeds of the epiphytes to their carton nests, in effect planting the gardens.

Plants feed the ants: food bodies
The expression “food body” or “food corpuscle” applies to any small epidermal structure that is collected and eaten by ants. The principal food bodies discovered so far are named and characterized in Table 14-2. They are extremely diverse in origin and form. For the most part they are best developed in myrmecophytes, in other words plants with domatia, a circumstance reinforcing the judgment that they are beneficial to the ants. However, relatively little is known about the actual use of the corpuscles. Ants have been seen collecting them from only a few of the plant species, and little effort has been made to document their consumption by either adults or larvae. Also, the biochemistry of attraction and nutrition of the food bodies remains largely unexplored. Nevertheless, Pachycondyla luteola, Camponotus balzani, and species of Azteca are wholly dependent on Cecropia and Pseudomyrmex on Acacia for their food, making it clear that the mutualism is obligate in at least one direction (Davidson et al., 1988). It is further true that the Cecropia and Acacia decline when deprived of their ants, so that the relation is truly mutualistic (Janzen, 1966, 1967, 1969; Schupp, 1986). The same is true of the myrmecophyte Triplaris americana (Davidson et al., 1988b).

Another line of evidence is the evolutionary reduction of the myrmecophytic structures in localities lacking guest ants altogether. At least two species of Azteca were abundant on Hispaniola in early Miocene times, but the genus is entirely absent from the Greater Antilles today (Wilson, 1985f). Cecropia occurs through the Lesser Antilles to Puerto Rico, where Cecropia peltata grows entirely in the absence of resident ants (Janzen, 1973a). And in different populations from Trinidad northward through the Lesser Antilles to Guadaloupe, Cecropia peltata shows a progressive reduction of ant-related traits (Rickson, 1977).

In almost all of the principal myrmecophyte genera listed in Table 14-2, the food bodies grow spontaneously. They develop fully on plants reared in greenhouses in the absence of the guest ants. The exception is the genus Piper. In Piper cenocladum at least, the leaf-margin corpuscles are produced only when the plant is occupied by Pheidole bicornis. Food-body production declines precipitously when the ants are removed, and it commences again when the ants are restored (Risch and Rickson, 1981).

Plants feed ants: extrafloral nectaries
Extrafloral nectaries are sugar-producing organs that attract animals but do not promote pollination. They can occur according to species almost anywhere on the plant, including even the flower outside the perianth. When active they attract worker ants, which tend to defend them from other insects. Unlike domatia and food bodies, they are produced by an enormous diversity of plants, occurring in no fewer than 68 families (Elias, 1983). Their anatomical variety is correspondingly great, ranging from small groups of cells that can be located with the naked eye only by detecting the nectar droplets they secrete or the groups of ants that gather around them, to complex cavities filled with trichomes and opening to the outside by a slot or pore. A very large literature exists on the subject, and fortunately this has been well reviewed in recent years by Bentley (1977), Buckley (1982a,b), Elias (1983), Koptur (1984), Beattie (1985), Benson (1985), and Oliveira and Leitão-Filho (1987).

Worker ants treat extrafloral nectaries in the same way that they respond to aggregations of honeydew-producing insects and sugar baits (Figure 14-10). The more aggressive species defend the active nectaries, in some cases extending their territorial zone over the entire plant. In 1889 Wettstein performed a surprisingly modernistic experiment that showed the protective role of the attending ants in the European composites Jurinea mollis and Serratula lycopifolia. He excluded ants from the plants and recorded an increase in damage from beetles and hemipteran bugs. Many similar studies conducted in recent years have yielded the same result, for example that of Oliveira et al. (1987b) in the cerrado woodland of Brazil. The plant species tested have been variously temperate or tropical in origin; vines, shrubs, or trees; and occupants of either forests or grasslands. The ants tested belonged to several genera in the Myrmicinae and Formicinae. The kind of damage averted or reduced by the ants included destruction of flower parts by grasshoppers, seed predation by bruchid weevils, and withering of shoot tips by psyllid nymphs. In some cases the authors directly observed the ants killing or chasing away the herbivores.

There is also some evidence that plants time their secretions in a way that enhances the protective role of the nectaries. In Michigan the nectaries of the North American black cherry (Prunus serotina) are most active during the first three weeks after budbreak, at which time they attract large numbers of Formica obscuripes workers. Perhaps coincidentally, the same three-week period is the only time that eastern tent caterpillars (Malacosoma americanum), which are the major defoliators of the black cherry, are small enough to be captured and killed by the ants. It is a reasonable hypothesis that Prunus serotina has evolved to bring its extrafloral nectaries into play when they can indirectly inflict the greatest amount of damage on the caterpillars (Tilman, 1978).

Ants feed plants: myrmecotrophy
Ant nests are generally among the most favorable sites for plant growth. The ants turn and aerate the soil, add nutrients in the form of excrement and refuse, and hold the ambient temperature and humidity at moderate levels. Larger nests are often surrounded by more luxuriant and species-rich vegetation than similar but unoccupied sites nearby. The difference is especially conspicuous in deserts, grasslands, and the early successional stages of forests (Gentry and Stiritz, 1972; Beattie and Culver, 1977). And at least in the chalk grasslands of England (King, 1977) and the deserts of Arizona (Rissing, 1986), some plants are much more abundant around the nests and others much scarcer, creating a striking floristic heterogeneity. On the foreshores of salt lakes in the southwestern Baraba Steppe of the Soviet Union, ants construct large hummocks that play a key role in plant succession (Pavlova, 1977).

This general potency of ant nests for the stimulation of plant growth preadapts the myrmecophytes to draw nourishment from their guest ants. Just such a relationship has been documented in the epiphytic myrmecophytes of the genera Hydnophytum and Myrmecodia (see Figure 14-4). Janzen (1974) found that the workers of Iridomyrmex cordatus (= I. myrmecodiae) discard the remains of prey in the cavities lined with absorptive tissues, while sequestering their own brood in separate chambers lined with tough, nonabsorptive cells. The absorptive surfaces are dotted with small lenticular warts. Janzen suggested that each of the two zones serves a separate function, namely the feeding of the plant and the housing of the ant brood. Using radioactive tracers, Huxley (1978) and Rickson (1979) demonstrated that this differentiation is indeed the case. The pseudobulbs absorbed (32P) phosphate, (35S) sulphate, and (35S) methionine from waste material deposited by the Iridomyrmex, as well as various breakdown products of decomposing Drosophila larvae. Most of the activity was concentrated in the warted areas. In a word, the ants feed the plants.

Many of the most specialized are tropical epiphytes in open forests and savannas located on nutrient-poor soils. In such areas there are typically few other epiphytes, and when they occur they usually grow on top of the myrmecophyte. Janzen (1974), Huxley (1980), and Thompson (1981) have all suggested that myrmecotrophy allows the plants to penetrate harsh environments otherwise closed to epiphytes.

Further research will no doubt discover new botanical structures and physiological processes that serve the capture and absorption of nutrients supplied by ants. The flask- and bladder-like domatia of Dischidia, for example, are penetrated by networks of adventitious roots that almost certainly absorb nutrients. Crematogaster and other ants that live and accumulate organic detritus on these roots are in fact literally “potting” the Dischidia.

Myrmecotrophy is not limited to plants with domatia. It inevitably occurs whenever ants nest on and around epiphytes, carrying in soil, building carton, and discarding waste materials. An entire fauna of arboreal ant species favors epiphytes for nesting. Only a small fraction of the epiphyte species are myrmecophilous, in other words domatia-bearing, and fewer than one percent of the ant species are closely associated with domatia. Huxley (1980) has pointed out that the domatia-bearing epiphytes are mostly limited to tropical Asia. Their relative scarcity in the New World tropics may be a result of competition from the immensely successful tank bromeliads and the ant-garden epiphytes. In at least the case of the ant-garden epiphyte Codonanthe crassifolia, the study by Kleinfeldt (1978) revealed that the presence of ants promotes growth, very probably as a result of increased nutrient supply. It is a remarkable fact that orchids, the other dominant group of vascular epiphytes, are also dependent on symbiosis for their early nutrition, but in this case the symbiosis is a mycorrhizal association with fungi rather than the harboring of ants.

Ants disperse plants: myrmecochory
Harvesting ants do not manage to carry all the seeds they collect back to their nests, and they do not eat all of the seeds stored in their granaries. The result is that ants are a major, albeit fortuitous dispersal agent of plants. They are especially effective in deserts and grasslands, but many species, not necessarily specialized harvesters, play some role even in tropical forests.

A wholly different, outwardly “purposeful” category of dispersal is accomplished by plants through myrmecochory, the employment of attractive seed appendages and chemicals that induce the ants to transport the seeds without harming the embryo or endosperm. The phenomenon was first carefully analyzed, and the relevant term proposed, by Sernander (1906). The appendages, which are called arils or elaiosomes, are biochemically distinctive and often large in size (see Figure 14-10). They take various forms according to species, including girdles, sheaths, caps, and finger-shaped terminal extrusions. They have been derived from several kinds of tissues, including the raphe, pericarp, and receptacle (Sernander, 1906; Ridley, 1930; Berg, 1979). Fleshy in consistency, white in color, and containing lipids, protein, starch, sugars, and vitamins, they are carried by foragers back to the nests, where they are eaten by the adult workers and larvae (Beattie, 1985).

Myrmecochory is a very common, in fact almost worldwide phenomenon. Three plant groups have been identified to date in which it is especially common: early-flowering herbs in the understory of north temperate mesic forest (Culver and Beattie, 1978; Pudlo et al., 1980; Beattie and Culver, 1981; Handel et al., 1981); perennials in Australian and southern African dry heath and sclerophyll forest (Berg, 1975; Westoby et al., 1982); and an eclectic assemblage of tropical plants (Horvitz and Beattie, 1980). Of course myrmecochores occur elsewhere; they have been well documented, for example, in the North American desert (Solbrig and Cantino, 1975; O'Dowd and Hay, 1980). Also, many parts of the world, for example tropical Africa, have not been carefully explored for myrmecophory. However, in spite of such bias in sampling, there seems to be little doubt that myrmecochory is disproportionately rich in the three habitats cited. In a New York beech-maple woodland, Handel and his co-workers (1981) found that 13 of 45 herbaceous plant species present were myrmecochores. These plants possessed about half of the stems and 40 percent of the aboveground herbaceous biomass. The proportion of myrmecochorous species in some Swedish habitats is 40 percent (Sernander, 1906). Even more impressive is the occurrence of the phenomenon in sclerophyll vegetation growing on sterile soils. About 1,500 Australian species are known or thought to be myrmecochores (Berg, 1975) and 1,300 South African (Bond and Slingsby, 1983; Milewski and Bond, 1985), as opposed to only 300 in the entire rest of the world. Botanists who have analyzed myrmecochory have concluded that the shortage of nutrients, particularly phosphorus and potassium, is the key to its irregular geographic distribution (Westoby et al., 1982; Milewski and Bond, 1985). Quite possibly plants rely on the relatively small, inexpensive myrmecochores in preference to the larger, fleshy fruits favored by birds and mammals whenever nutrients are in chronic short supply.

The myrmecochorous plant species are phylogenetically diverse. In Australia alone they are scattered through no fewer than 87 genera in 23 families. That fact, plus the extraordinary variety in the embryological provenance of the myrmecochores themselves, indicates that myrmecophory has originated many times and undergone a great deal of convergence in evolution. A few of the independent phylogenetic pathways have in fact been worked out by Berg (1972, 1975, 1979).

The ants attracted to the myrmecochores are a similarly diverse lot, representing some of the dominant genera peculiar to each local fauna in turn: Aphaenogaster, Leptothorax, Myrmica, Tapinoma, Formica, and Lasius in north temperate forests (Culver and Beattie, 1978, 1980); Odontomachus, Pachycondyla, Pheidole, Azteca, and Paratrechina in at least one Neotropical forest locality (Lu and Mesler, 1981); and Rhytidoponera, Pheidole, Monomorium, and Iridomyrmex in Australia (Buckley, 1982b). At least two ant species, Pogonomyrmex californicus and Veromessor pergandei, are bivalent; they harvest some seeds for total consumption and disperse others after collecting them for their elaiosomes (O'Dowd and Hay, 1980).

The existence of an elaborate structure with no apparent function other than to attract ants implies that a substantial selective advantage accrues to the myrmecochores. Botanists have given a good deal of thought to what this advantage might be, and they have come up with five possibilities.

1. Avoidance of interspecific competition. If the ants carry the seeds of a particular plant species to sites where other, competing species grow less well, myrmecochory will be favored by natural selection. Handel (1978) has adduced some evidence for this effect in forest-dwelling Carex.

2. Avoidance of fire. As Berg (1975) has noted, myrmecochorous plant species are common in fire climax communities. It is possible that seeds carried into ant nests are protected during the frequent burnovers of their habitat. On the other hand, some species require the high temperatures generated by fires in order to germinate. Majer (1982) has shown that in southwestern Australia, ants transport many of the seeds to an intermediate depth where the temperature is high enough to germinate the seeds but not high enough to kill them. A similar finding was made by Bond and Slingsby (1983) in the heathland of the Cape Province.

3. Avoidance of parental competition. Young plants do less well near members of their own species, so that there is an advantage to dispersing well away from the parent. In the case of Ajuga at least, ants carry elaiosome-bearing seeds beyond the boundaries of parent clones, eliminating parental competition (Luond and Luond, 1981).

4. Avoidance of seed predators. Bond and Breytenbach (1985) found that seeds not dispersed by ants in the South African heathland are subject to much higher predation by small mammals. When seeds are concentrated they are evidently more attractive and easily discovered. Smith et al. (1986) demonstrated that ants rescue a large percentage of the seeds of the North American myrmecochore Jeffersonia diphylla from rodents.

5. Microsites with superior nutrients. Nests of ants, especially those belonging to large colonies, usually have higher levels of the nutrients important for plant growth (Gentry and Stiritz, 1972; Haines, 1978; King, 1977; Petal, 1978). It would seem to be advantageous for ant-dispersed seeds to find their way into the ant nests, which in fact is usually the case.

Beattie (1985) has evaluated these five selection pressures on the basis of the still sparse evidence. Although the factors are likely to vary in relative importance according to species and geographical location, some seem to be more generally potent than others. The interspecific competition model, for example, is compromised by its limitation to related species, some of which are myrmecochores and some not. Fire avoidance has so far been documented only by indirect evidence and is obviously limited to fire-dependent vegetation. Avoidance of parental competition is also a reasonable explanation, but it is easily confounded by any competing hypothesis that postulates an advantage to dispersal, such as the increased probability of predator avoidance. In the absence of optimality models that predict the best distances for particular circumstances, the parental-competition hypothesis is too good; it can be applied indiscriminately to almost all situations.

Nutrient enrichment and predator avoidance appear to be both the most plausible and the most readily testable of the suggested selection pressures. Unfortunately, few quantitative studies have been conducted at this writing to assay the effect of nutrient enrichment, and the results so far are mixed. Culver and Beattie (1980) obtained a positive result with two ant-dispersed species of violets in the chalk downland of southern England. Seeds planted in ant nests produced plants that were more numerous and larger in size than those planted in alternative, randomly selected microsites. Three years later almost all of the survivors were in the ant nests. In similar manner, Davidson and Morton (1981) found that ant-dispersed plants of the family Chenopodeaceae in arid scrubland of Australia do better on ant mounds than elsewhere. The effect varies according to the properties of the soil. In habitats underlain by red, crusty, alluvial loam soils, the chenopods are limited almost entirely to ant mounds, whereas in better aerated and drained sandy soils they form almost continuous stands. It thus appears that the Australian ants alter not only the nutrient levels but also the physical properties of the soils they inhabit. The relative contributions of the two factors will be difficult to weigh in most studies. This difficulty is underscored by the finding of Rice and Westoby (1986) that myrmecochorous species in nutrient-poor Australian sclerophyll vegetation grow in soils with no more nitrogen and phosphorous than soils around non-myrmecochorous species. Similarly, Horvitz and Schemske (1986) found that seedlings of the myrmecochore Calathea ovandensis of Mexico grow no better in pots of soil from ants' nests than in pots of soil from nearby random sites, despite the fact that the former are richer in nutrients.

In general, to conclude, it appears that nutrient enrichment is an important selection force in some habitats, especially where the soils are relatively sterile, but not in others--for reasons yet to be clarified. As Rice and Westoby say generally of the adaptiveness of myrmecochory,

Many explanations could account for the particular importance of myrmecochory in Australia and South Africa, or for the importance of myrmecochory in sclerophyll as compared with mesophyll vegetation. But few explanations can account for myrmecochory being more important in sclerophyll than in mesophyll vegetation in Australia and South Africa, but vice versa in North America and Europe. Because sclerophyll shrublands in Australia and South Africa are delimited by low-nutrient soils, while in North America and Europe they are climatically controlled, the nutrient-enriched microsite explanation could have accounted for this distribution. However, this explanation evidently does not hold for a representative cross section of Australian sclerophyll species. This throws open again the problem of finding an adaptive explanation that could account for the geographical distribution of myrmecochory.

Ants pollinate flowers (sparingly)
Given the abundance and antiquity of ants, it is puzzling to find that they play a relatively minor role as pollinators. Ants are important for a few flowering plants that display the “ant-pollination syndrome,” comprising the following traits: the plants grow in hot and dry habitats where ants are most abundant and active; the nectaries are morphologically accessible to the flightless ant workers; the plants are short and prostrate; pollen volume per flower is small in order not to stimulate self-grooming in the ant through excessive loading of pollen; and seeds are few per flower, thus requiring less abundant pollen transfers (Hickman, 1974). Examples include plant species occurring on the hot dry slope of the western Cascades of Oregon (Hickman, 1974), granite outcrops of the southeastern United States (Wyatt, 1981), and the Colorado alpine tundra (Petersen, 1977). Ant pollination of other kinds of species occurs, for example ground orchids with raised stems (Armstrong, 1979; Brantjes, 1981), but it is evidently much less common. The Australian orchid Leporella fimbriata is pollinated by pseudocopulation: winged males of the bulldog ant genus Myrmecia mistake the flowers for virgin queens and attempt to mate with them, picking up pollenia in the process (Peakall et al., 1987).

In a striking counterpoint, ants tend to be the “scoundrel in the pollination drama,” because they dominate the nectaries while contributing little or nothing to pollination (Faegri and van der Pijl, 1971). Their antibiotic secretions from the metapleural and poison glands, used to suppress bacterial and fungal growth in the nests, are likely to interfere with pollen germination and pollen-tube growth (Iwahami and Iwadare, 1978; Nakamura et al., 1982; Beattie et al., 1984, 1985, 1986). These substances, originally characterized by Schildknecht and Koob (1970, 1971) and Maschwitz et al. (1970), include -indolacetic acid, phenylacetic acid, and -hydroxydecanoic acid (“myrmicacin”). Some plant species are said to have evolved defenses against the ants in the form of sticky belts and repellent substances in the petals and nectar (Faegri and van der Pijl, 1971; Janzen, 1977). However, the extent and effectiveness of these devices remains to be evaluated. Janzen suggested that as a rule ants do not visit flowers in lowland tropical habitats, and the reason is that floral nectar contains unpalatable substances. This generalization was refuted by Haber et al. (1981), who observed ants feeding on the floral nectar of 27 species of plants in Costa Rica. The ants also feed readily on nectar separated from the flowers and presented to them directly in the field, even though in some cases the liquid contained alkaloids and phenolic compounds.

Beattie (1985) has argued that the pollen-suppressing activity of myrmicacin and other ant substances is the fundamental constraint that has limited the evolution of ant pollination. Looked at from the ants' point of view, these insects have bought protection against fungi and bacteria in their underground nests at the cost of the nectar that would otherwise have been given them by the flowering plants. From the plants' point of view, the only species accepting ants as pollinators are those whose habitats leave them little other choice. Beattie has discounted the possibility that the peculiarities of ant foraging prevent them from serving as efficient pollinators, countering with the statement that “ants have sophisticated sensory and orientation systems and systematically visit plants of all sizes, from low herbs to tall trees, to harvest resources as diverse as insect prey, honeydew, extrafloral nectar, and seeds.” This is true insofar as it is stated, but Beattie overlooks another property of the foraging strategy of many species of ants that might serve as a fundamental constraint. Foragers of many if not most species display Ortstreue, returning daily to the same plant and even to the same branch or flower cluster (see Chapter 10). Some, including the omnivores most likely to serve as pollinators, even set up shelters over extrafloral nectaries and honeydew-producing homopteran colonies. Thus plants depending on ants for pollination are likely to find themselves being self-fertilized.

Ants prune and weed
Some specialized plant-dwelling ants protect their myrmecophyte hosts not only from herbivores but also from other plants that crowd in too closely. Pseudomyrmex ferrugineus workers attack and destroy any foreign plant that sprouts within 40 cm of the trunk of the Acacia in which they live, and they cut back vines and foliage of neighboring trees that touch the Acacia crown (Janzen, 1967). This pruning action has the effect of promoting the growth and survival of the host plant, but it also removes bridges over which alien ants can attack the resident colony. Davidson et al. (1988a) found that in Amazonian Peru, Crematogaster workers regularly encroach on Triplaris americana trees and attack the resident Pseudomyrmex dendroicus colonies by interfering with their foraging and even stealing their brood. The dendroicus go so far as to sacrifice valuable leaves of their own host plant when they serve as bridges for major invasions by enemy ants. In a suggestive parallel response, workers of Allomerus demararae prune vines on their Cordia nodosa host plant when they are pressured by the encroachment of alien ants.

Davidson and her co-workers tested the bridge-demolition hypothesis further by constructing many artificial (and indestructible) bridges of wire between Triplaris and Cordia trees and neighboring vegetation. The result on Triplaris was a significant increase in invasion by alien ants, almost all of which were members of the genus Crematogaster. The Cordia plants were protected by their trichomes.

Pruning is not practiced universally by myrmecophyte-dwelling ants. As Davidson and her co-workers discovered (see Table 14-3), the practice is almost entirely limited to species that sting opponents, as opposed to those that have non-penetrating stings and rely on poisonous chemical sprays and droplets. It is possible that stings are generally less effective against other ants than are chemical defenses, so that ants using them are forced to rely more on pruning and bridge demolition to fend off invasions.

Trade-off and compromise
Our account to this point has depicted myrmecophytism as a benefit accruing primarily to plants. The case seems especially strong in the extreme myrmecophytes, which are literally planted, fed, and protected by the ants. Yet there has been an undeniable reluctance on the part of plants generally to enter into such obligatory alliances, because only a small fraction of the genera and species in the world have done so. The explanation of why myrmecophytism is a minority phenomenon may lie in the tendency of arboreal ants to maintain herds of scale insects, mealy bugs, and other honeydew-producing homopterous insects (see Chapter 13). So whereas the ants bring gifts to their hosts, they exact a price in energy that the plant must donate to the homopteran sap-feeders.

However, the depredations of the homopterans are a price the plants can readily pay if the benefits conferred by the ants are high enough. Janzen (1979) has proposed that the maintenance of the herds are best regarded as a more or less fixed cost for the plant, the equivalent of maintaining alkaloids and other secondary substances as a standing defense system against herbivores. Some plants pay energy into the manufacture of the secondary substances, while others, the myrmecophytes, pay the ants to act like secondary substances. In both cases energy is invested to obtain a positive net yield of energy in the end. Janzen expressed this concept of a symbiotic balance sheet as follows:

In at least two complex and well-developed ant-plant mutualisms, African Barteria trees and Neotropical Cecropia trees, the ants maintain a standing crop of scale insects or other homopterans inside the hollow stems. These animals feed on the plant and provide a major food source for the ants with their bodies or honeydew exudates. The ants are obligate occupiers of the trees and protect the trees from herbivores and vines. The homopterans are zoological devices used by the plants to maintain an ant colony, the ants being directly analogous to the chemical defenses maintained (and paid for) by more ordinary plants.

Janzen concluded with the intriguing conjecture, “I would not expect there to be selection for traits that reduce the 'damage' done by the Homoptera to the level that would debilitate the ant colony and its protection of the tree.” Here we see an application of the concept of coevolution in purest form. The myrmecophyte and the guest ants have evolved by common pressures in natural selection, such that they do not let the homopteran populations explode and kill the tree, and the plants do not defend against the homopterans so vigorously that they starve the ant colony. How might such controls be enacted during evolution? It is well known that ants of various kinds often eat some of the homopterans they attend (Carroll and Janzen, 1973; Hinton, 1977). Among the myrmecophyte-dwelling ants, this practice is known to occur in at least one species: Pseudomyrmex triplarinus regularly kills some of its symbiotic coccids and feeds them to its larvae (Schremmer, 1984). At the same time, homopterans often do not build dense populations under natural conditions, especially those prevailing in tropical forests, so that resistance to them might be relatively easy and inexpensive (Beattie, 1985). Under such conditions, if indeed they occur with consistency, the plant species can “relax” and leave all forms of homopteran control, including that of the myrmecophiles, to the ants themselves.

Ecologists have only begun to investigate the subtle relations that can evolve between plants, their protector ants, and their enemies. In the lowland tropical forest at Los Tuxtlas, Mexico, for example, the perennial marantaceous plant Calathea ovandensis attracts ants with extrafloral nectaries, and the ants attack herbivores invading the plant. But the ants also tolerate one herbivore, the larvae of the riodinid butterfly Eurybia elvina, which provide them with attractive secretions from eversible glands on the dorsal surface of the eighth abdominal segment. Hence the Calathea plants are caught somewhere in the balance between the beneficial control of ordinary herbivores by the ants and the harmful tolerance of the butterfly larvae by the same ants. How do these combinations work out? Horvitz and Schemske (1984) showed that when ants were absent and Eurybia butterfly larvae present, seed production was lowered by 66 percent, the greatest loss recorded. When ants were present and Eurybia absent, the loss was only 33 percent, the least recorded. The remaining combinations, ants and Eurybia both present and ants and Eurybia both absent, yielded intermediate degrees of production loss. Of equal significance, there was considerable variation in effectiveness among the eight ant species recorded, with Wasmannia auropunctata conferring the most protection and Pheidole gouldi the least.

It has thus been born out that some ant species are better symbionts from the plants' point of view than others. This circumstance opens the possibility for some species to intrude as parasites into well organized mutualistic systems. One such parasite of a mutualism is Pseudomyrmex nigropilosa of southern Mexico and Central America (Janzen, 1975). It occupies swollen-thorn acacias but, unlike the other ten known members of the genus that are obligate residents of these myrmecophytes, Pseudomyrmex nigropilosa provides no protection for the plants. As a result either the acacias are soon killed by herbivores or the nigropilosa are replaced by one of the competent species of Pseudomyrmex. As expected of such a parasite, Pseudomyrmex nigropilosa produces reproductives earlier in the life of the colony than is the case with the mutualistic species. In other words, it breeds and disperses before the host dies as a result of its incompetence.

The invasion of a sufficiently dominant but inefficient intruder might endanger the very existence of plants dependent on ant services. According to Bond and Slingsby (1984), something approaching this ecological catastrophe is underway in South Africa. The Argentine ant Iridomyrmex humilis has invaded a portion of the fynbos, a local form of scrubland. This unusual habitat possesses a very large number of endemic plants, hundreds of which are myrmecochores and thus dependent on ants for their dispersal and interment in the soil. Iridomyrmex humilis, a dominant species of South American origin, has replaced the native ants where it has invaded the fynbos of the Kogelberg State Forest. In field tests, Bond and Slingsby found that the Iridomyrmex are much slower than the native ants in removing seeds of the proteaceous Mimetes cucullatus, a representative myrmecochore of the region. The Iridomyrmex also move the seeds shorter distances and then leave them on the soil surface, where they are quickly eaten by invertebrate and small vertebrate granivores. In one trial, 35 percent of the Mimetes seeds disseminated from depots germinated in Iridomyrmex-free sites, but fewer than one percent germinated in a nearby infested site.

Parasites of ants: microorganisms and fungi
Relatively little is known about the pathobiology of ants. The species studied most intensely to date is the red imported fire ant Solenopsis invicta, for which entomologists have sought--in vain--a biological control agent. Solenopsis invicta is evidently typical of ants in that microorganisms and fungal parasites are relatively scarce and few in species. They include a virus, one possible species of bacterium, a unicellular fungus (occurring in more than 90 percent of the colonies and not a serious pathogen), two microsporidians, and two neogregarines (Jouvenaz, 1986).

A peculiar haplosporidian parasite was discovered in the European thief ant Solenopsis fugax by Karl Hölldobler (1929b, 1933), which he described as Myrmicinosporidium durum. Because the microorganism takes on the shape of a little bowl when placed in fixative, the disease is called “Näpfchenkrankheit” in the German literature (Figure 14-11). The parasite was subsequently found in the genus Leptothorax (Gösswald, 1932; Buschinger and Winter, 1983a) and in Pheidole pallidula (Espadaler, 1982b). Recently Crosland (1988) discovered lemon-shaped objects in the bulldog ant Myrmecia pilosula, which appear to be the spores of a protozoan gregarine parasite.

Several kinds of fungi parasitic on ants have been studied by entomologists and mycologists, in some cases for generations. The first is the genus Cordyceps (Ascomycetes: Clavicipitaceae), the largest of the entomophagous fungi and often brightly colored as well. The multicellular mycelium pervades the host thoroughly, killing it. It often produces asexual spores or conidia, in which case the specimen has been placed in “Isaria,” a genus of convenience classified as one of the imperfect fungi. Eventually the mycelium sprouts a boll- or club-shaped organ on a stalk that protrudes as much as 10 centimeters outside the body of the host. The swollen terminus of this structure contains numerous ascocarps, each generating spores within elongate cells or asci (Thaxter, 1888; Rogerson, 1971; Evans and Samson, 1982, 1984; Samson et al., 1982). It is a startling experience to encounter a large ant, dead yet standing rigidly at attention with a Cordyceps sporophore raised above it like a flag (see Figure 14-12). At least some species of the genus occur on more than one species of insect host.

The fungus Desmidiospora myrmecophila was discovered by Thaxter (1891; cited in Bequaert, 1922) “growing luxuriantly on a large black ant fastened to the underside of a rotting log in Connecticut. The hyphae, much branched and septate, covered the host in a white flocculent mass; they emerged especially from between the abdominal segments, enveloping the insect more or less completely and extending a short distance over the substratum.” W. M. Wheeler later identified the ant as Camponotus pennsylvanicus. This fungus was recently rediscovered in Arizona on a worker of Camponotus semitestaceus by Clark and Prusso (1986).

A third group of fungi infecting ants are the members of the ascomycete order Laboulbeniales. These organisms are by far the most specialized of the fungi living on insects. They are transmitted by direct contact from one generation of hosts to the next. They are remarkable in that they neither kill nor visibly harm the insect hosts. Unlike the Cordyceps, they are inconspicuous and, according to Bequaert (1922), resemble minute usually dark-colored or yellowish bristles or bushy hairs, projecting from the integument either singly or in pairs. They are usually scattered, but sometimes they are so densely crowded in certain areas as to form a furry coating. The main body of the fungus is external to the insect. It consists of a small number of cells arranged in vertical rows and is attached by a blackened “foot” to the host's cuticle (Figure 14-13). In some species haustoria penetrate the integument and enter the body cavity (Benjamin, 1971).

Aegeritella roussillonensis is a recently discovered new species belonging to the Hyphomycetales. Balazy et al. (1986), who described it, found the fungus on live Cataglyphis cursor. A large number of the workers of a colony and even the queen can be infested. Sometimes the mycelium covers the whole body of the ant, but it is most common on the ant's mouthparts. The behavioral activity of the ants is seriously affected by the fungal infection. Other species of Aegeritella have been discovered in Formica in Europe and Camponotus in Brazil (Espadaler and Wisniewski, 1987).

Wheeler (1910c) posed the main theoretical question concerning parasitic fungi in ants in the following way: “At first sight ants would seem to be particularly favorable hosts for such parasites since these insects are in the habit of huddling together in masses in warm subterranean galleries, where the fungi might be supposed to develop luxuriantly and transmit their spores from ant to ant with great facility.”  Yet fungal parasites are scarce and do not appear to play any significant role in the demography of ant colonies. Wheeler supposed that the fanatical grooming behavior of ants was their principal defense against fungi. Workers are forever cleaning themselves and their nestmates with their tongues and forelegs (Wilson, 1971; Farish, 1972; Figure 14-14). The detritus collected in this way, including fungal spores, is compacted into small masses in the infrabuccal chamber, which is located on the floor of the mouth cavity. The masses are coughed up periodically in the form of “infrabuccal pellets” and carried out of the nest as waste material (Wheeler and Bailey, 1920; Eisner, 1957; Eisner and Happ, 1962). This behavior is undoubtedly an important part of the explanation, but there is more. As described earlier, we know that ants secrete a medley of potent antibiotic secretions from their metapleural gland and other exocrine organs in the body, which they then spread by grooming movements.

Possible microorganismic symbionts in ants
The tribe Cephalotini comprising exclusively arboreal ants of the New World tropics possess a peculiar mushroom-shaped proventriculus or gizzard that has long been thought to strain liquid food from the esophagus and crop posteriorly to the midgut (Eisner, 1957). Myrmecologists have often speculated (at least in conversation) that the unique anatomy is somehow associated with the unusual feeding habits in the cephalotines. Workers of Zacryptocerus rohweri and Zacryptocerus texanus do indeed collect and feed on pollen, at least in part (Creighton and Nutting, 1965; Creighton, 1967). On the other hand, Zacryptocerus varians is a general scavenger that can flourish in the laboratory on freshly killed insects and honey (Wilson, 1976a). Recently, Caetano and Cruz-Landim (1985) discovered that in the Brazilian species Zacryptocerus clypeatus and Cephalotes atratus, the anterior portion of the small intestine is nearly filled with fibrillar material. Close examination revealed the mass to consist of bacteria and filamentous, non-septate fungi. Many of the same kind of organisms occur in the posterior region of the midgut, just anterior to the principal aggregate of microorganisms located in the small intestine. The role of the bacteria and fungi is unknown, but Caetano and Cruz-Landim make the entirely reasonable suggestion that they exist in some form of alimentary mutualism with their ant hosts. Perhaps they assist in digesting pollen extracts or some other nutritive materials beyond the ordinary metabolic capabilities of ants.

Intracellular symbiotic bacteria have also been discovered in the ovarioles and midgut epithelium of Formica fusca and the carpenter ants Camponotus herculeanus and Camponotus ligniperda (Hecht, 1924; Lilienstern, 1932; Kolb, 1959; Buchner, 1965). Smith (1944) suggested that the symbiotic microorganisms either contribute vitamins to the hosts or in some way enable Camponotus to digest wood. However, biochemical analysis of the gut enzymes of both Camponotus herculeanus and Camponotus ligniperda and feeding experiments have clearly demonstrated that the carpenter ants do not digest wood (Graf and Hölldobler, 1964). The symbiotic role of the bacteria in the ant's body, if any exists, remains unknown.