The Ants Chapter 15

CHAPTER 15. THE SPECIALIZED PREDATORS

Physical adventure and instant scientific rewards await the entomologist who searches for new kinds of ants in unusual food niches or investigates the niches of previously little known species, which then, in one stroke, places these species within the larger context of ant biology. Here are several of our favorite examples of past successes:

• Jean Lévieux (1983b), digging below 20 cm in the soil of the Ivory Coast savanna, discovered an entire guild of blind, centipede-eating amblyoponine ants. They comprise five or six species of Amblyopone, including Amblyopone mutica, and Amblyopone pluto, as well as Apomyrma stygia, the first species of a new genus.

• Working in Guyana (known as British Guiana at the time) in 1920, Alfred E. Emerson found that the large ponerine ant Pachycondyla commutata is a specialized hunter of Syntermes, the largest termites of the New World. In his notes, published by W. M. Wheeler (1936c), he wrote: “Kartabo, B. G. Oct. 10, 1920. Between 5:30 and 6:00 p.m. I observed a large number of commutata (more than 100) raid a trail of Syntermes territus. The termites, of which soldiers and two kinds of workers were plentiful, formed an open trail on the ground. The ants, moving either in groups or singly, came down a small hill from their nest in the ground under low bushy vegetation about 150 ft. from the termites and stridulated violently while moving back and forth from their nest. They attacked both the soldier and worker termites by stinging them and then carried them off to the nest.”

• Deep in the Oropouche Cave of Trinidad, Wilson (1962d) rediscovered the one supposedly true cave-dwelling ant, “Spelaeomyrmex” urichi. He found workers gathering arthropod eggs. Soon afterward, he encountered the same species nesting in open rain forest. He demonstrated experimentally that Spelaeomyrmex urichi is really a member of the genus Erebomyrma and that it is in fact a skilled predator of arthropod eggs.

• When William L. Brown (1974b) explored the native ant fauna of Mauritius, he discovered a remarkable new species, entailing a major range extension of its genus. The prime locality to search for the last remnants of this endangered island fauna is Le Pouce, an elongate massif with a topside plateau covered with low, gnarled native forest. After one only partially successful trip to the plateau, Brown decided on a second effort:

"On the first of April, though I was scheduled to leave on an evening flight to Bombay, I tried Le Pouce again. A telephone call to Mr. J. Vinson, who had collected some of Donisthorpe's material, convinced me that the main path on the Le Pouce plateau should not be avoided.  I arrived there in the afternoon; the day was heavily overcast, threatening rain on the peaks, and it took me about an hour to walk up to the plateau.  Whereas the sunny Sunday in the scrub shade had yielded almost no ants foraging, I now found foragers on foliage and on the hard-packed earth of the trail every few meters of the way.  These were mostly Camponotus aurosus and Pristomyrmex spp. (= Dodous), native Mauritian species.  Before long, on the trunk of a small tree by the path, I found a sparse trail of bright red ants climbing the bark.  Closer examination revealed these to be predominantly the ectatommine I had collected on the previous Sunday, but interspersed with these were workers of Pristomyrmex bispinosus which, with their gasters partly curled under, looked remarkably like the ectatommines. It is hard to avoid the impression that some kind of mimicry involves these two species in this habitat. The ectatommines ascending the trunk nearly all carried in their mandibles whitish spherical objects that proved eventually to be arthropod eggs--probably spider eggs. I climbed the tree which was only about 5 meters high, and soon found the nest about 3 meters up. Where two of the gnarled branches crossed, a thick pad of lichens surrounded the place where they touched. Forcing the branches apart, I found a rotted-out pocket, evidently caused by their rubbing together in high winds. The cavity extended downward several centimeters into one of the branches, and it was full of the ectatommine ants with brood and many of the round white arthropod eggs; I estimate that I removed or saw at least 200 workers, and there may have been more. [Afterward, in an incident about which he did not write, Brown attempted to climb the small peak at the end of the plateau. Lightning struck 20 feet upslope; then heavy rain turned the ground into slippery mud. Brown fell and slid down the slope toward the brink of a high cliff. Hanging on to one small bush for a few minutes, he finally regained his purchase and worked his way back down to Port Louis. Another day in the life of the myrmecologist thus came to an end.--The authors]"

The bright red ant turned out to be an undescribed Proceratium (P. avium), notable for its aboveground foraging for eggs, in marked contrast to the other, wholly subterranean members of the genus. Also, the enlargement of its one-faceted eyes suggests that avium emerged as an epigeic forager late in evolution after the arrival of its ancestral stock on the remote island of Mauritius (see Figure 15-1).

The evolution of prey specialization
A large amount of such field work has provided many generalizations of importance in ant biology. The ants display among themselves almost every conceivable degree of specialization, from those that accept many kinds of insects and other arthropods to others limited to geophilomorph centipedes or some other, comparably restricted group. The roster of the most specialized species known to us is presented in Table 15-1. The phylogenetic distribution of this extreme class is curious. It is disproportionately concentrated in the Ponerinae, Leptanillinae, and the army ant subfamilies Dorylinae and Ecitoninae. The Myrmicinae rank a somewhat distant fourth. Remarkably, almost no specialized predators have been recorded in the large subfamilies Dolichoderinae and Formicinae. The single exception is Myrmoteras barbouri, which appears to be a specialist on collembolans. It uses its elongate trap jaws and sensitive trigger hairs to snare these elusive prey. A second species, Myrmoteras toro, accepts a wide variety of soft-bodied arthropods. It possesses trap jaws but lacks trigger hairs (Moffett, 1986c; see Figure 7-54).

Defined with reference to the variation in prey utilized from one species to the next (as opposed to variation in prey utilized by single species), the Ponerinae and Myrmicinae are the clear leaders. By and large, the prey utilized by the specialists are ground-dwelling arthropods that are relatively abundant but either swift moving or armed with formidable defenses. They include collembolans (springtails), centipedes, millipedes, oniscoid isopods, other ants, and termites. As we shall see, various of the specialized predators have developed morphological and behavioral devices to overcome these prey. Eggs of arthropods are an odd separate class that requires separate explanation.

It must be emphasized that most of the listings in Table 15-1 are based on inadequate data. In a few cases they are inferred from no more than one to several field records or laboratory experiments. A species is proven oligophagous by only one or the other of three methods. First, the species might be seen consistently to capture the preferred prey to the exclusion of almost all other potential prey with which it comes in contact. Such is the case for most of the ant and termite predators. The raids are usually conducted aboveground in such a conspicuous fashion that they have been witnessed repeatedly by entomologists, and specificity is not in doubt. Second, species in a few ant genera, including the ponerines Gnamptogenys, Leptogenys, and Myopias, store the inedible remains of their prey in kitchen middens contiguous with the nest chambers. The consistent presence of only one or a few kinds of organisms in the middens is prima facie evidence of oligophagy, especially if it can be combined with the discovery of at least a few fresh prey among the brood. The third method is the “cafeteria experiment,” which has been used to produce persuasive evidence for predatory specificity in Smithistruma and other dacetine ants (Wilson, 1953a), Belonopelta (Wilson, 1955b), and Prionopelta (Hölldobler and Wilson, 1986a). The technique is most useful for species that neither forage aboveground nor accumulate kitchen middens. Colonies of the species are first established in an artificial nest adjacent to a foraging arena. An aspirator is then used to collect arthropods, nematodes, and other small organisms in the terrain in the vicinity of the nest and these are placed live in the foraging arena, along with a scattering of litter and soil. A record is kept of the prey selected by the foraging ants. In the case of dacetines at least, the cafeteria experiment has produced results consistent with observations of undisturbed colonies and foragers in nature. Other, more sophisticated techniques of prey identification are possible, including classification of remains in the infrabuccal pockets of the predators and biochemical analysis of stomach contents. But with the exception of an early analysis of infrabuccal pellets of Pseudomyrmex by Wheeler and Bailey (1920), these approaches have not been attempted.

Oligophagy appears to be an evolutionarily derived condition, despite its concentration in the primitive subfamily Ponerinae. In fact, the majority of ponerine genera are polyphagous or, as in the case of Pachycondyla, comprise both polyphagous and oligophagous species. Some ponerines visit extrafloral nectaries, including Ectatomma tuberculatum (Weber, 1946b) and Paraponera clavata (Young and Hermann, 1980). At least one species, an African Odontomachus, attends aphids and coccids and even builds shelters over these “cattle” in the manner of myrmicine, dolichoderine, and formicine ants (Evans and Leston, 1971). And while it is true that most species of the primitive genus Amblyopone thus far studied, are specialists on centipedes, at least one, the common Australian species Amblyopone australis, is a generalized insectivore (Haskins and Haskins, 1951). Amblyopone workers do not accept sugary fluids, possibly a correlate of their largely subterranean habits that keeps them from contact with extrafloral nectaries. The very primitive Nothomyrmecia macrops, sole living member of the subfamily Nothomyrmeciinae, is a generalized insectivore, readily climbs bushes and trees, and accepts honey in the laboratory (Hölldobler and Taylor, 1983). Myrmecia, the sole living genus of the subfamily Myrmeciinae and nearly as primitive in overall anatomy as Nothomyrmecia, preys on a wide range of insects and spiders (Gray, 1971a,b, 1974a). Its workers, the fearsome “bulldog ants,” visit extrafloral nectaries to imbibe sugary secretions (see Chapter 14). They also fit an intuitive conception of primitive huntresses, employing stealth, excellent vision afforded by their large eyes, and the ability to spring through the air to capture their victims. Gray reports that foragers of Myrmecia varians “were observed to remain stationary on branches of Eucalyptus largiflorens for as long as 30 minutes with their body crouched close to the branch and mandibles wide apart, ready to spring at unsuspecting prey. One such was seen to capture the fast-moving bush-fly Musca vetustissima Walker, which had landed approximately three centimeters away. The worker sprang quickly, clasping the fly's head between its mandibles, and stung the prey to death. The worker then jumped off the branch onto the ground, a distance of nearly one meter, and returned directly to the nest 15 meters from the tree.” Limited specificity was observed in Myrmecia dispar, which accepts a wide range of insect prey in the winter but turns predominantly to other ants in the summer (Gray, 1971b). Finally, the relatively primitive myrmicine genera Hylomyrma and Myrmica are also generalized insect predators (Wilson, unpublished notes).

How can the peculiar phylogenetic distribution of oligophagy be explained? The answer may lie in the interesting fact that all of the specialized predators have well developed stings, whereas a large percentage of the species that are generalized predators lack functional stings and rely exclusively on biting and the release of poisonous secretions. None of the many members of the subfamilies Dolichoderinae and Formicinae possesses a functional sting, and (with the exception of the very peculiar Myrmoteras barbouri) none is a specialized predator. It seems likely that stings are the best weapon to subdue formidable or elusive arthropods in a quick and clean manner. Put another way, it seems that the possession of a sting allows the species to be either a generalist or specialist, but the absence of a sting precludes the species from being a specialist.

The foraging strategies of the great majority of specialized predators fall into one or the other of two broad classes. One kind of species raids en masse. Groups of workers are directed along odor trails by scouts to the prey, which they then overwhelm in violent combat. All of the predators of ants use this method, including the army ants (subfamilies Dorylinae and Ecitoninae) and members of the genus Cerapachys. Other ants in the same category are predators of termites in the genera Pachycondyla, Leptogenys, and Megaponera; species of Leptogenys that capture isopods; and Leptanilla japonica and species of Onychomyrmex, which attack centipedes. The second principal group of specialists forage singly, most often relying on camouflage and stealth. They include most of the specialist myrmicines, such as Strumigenys, Smithistruma, other small members of the tribe Dacetini, and Eurhopalothrix in the tribe Basicerotini.

Another useful generalization is that all of the genera of specialized predators are either limited to the tropics or are at least disproportionately represented there. Except for an occasional Amblyopone, Proceratium, Smithistruma, or army ant in the genus Neivamyrmex, specialists are entirely absent from the northern half of the United States and Europe. Cerapachys and smaller dacetines are moderately common in portions of southern, warm-temperate Australia, but they and other specialists are far less well represented there in species and individuals than farther north in Queensland and Melanesia. This ecogeographic rule may in some way be related to another climatic correlation established experimentally by Jeanne (1979): rates of predation by ants, reflected by the harvesting of wasp larvae from bushes and trees, increase gradually southward from Michigan to the Brazilian Amazon. The numbers of ant species appearing at Jeanne's baits also increased, from 22 at the northern limit to 74 at the southern limit. The percentage of canopy-dwelling species, as opposed to those that forage onto vegetation from ground nests, also increase. It is well known that in the absence of such arboricolous forms, terrestrial ants forage upward into the vegetation in greater force. On Mont-Ventoux, in the south of France, Du Merle (1982) found that at least 23 of the 40 terrestrial ant species foraged on shrubs and trees, with about ten doing so intensively.

The evolution of the dacetine ants
The analysis by Brown and Wilson (1959) of the Dacetini, and others, supplemented by Wilson (1962a), Hölldobler (1981b), Dejean (1982, 1985a,b), and Masuko (1984), was one of the first attempts directed at any group of animals to correlate the evolution of social behavior with species-level adaptations in feeding. The dacetines are exceptionally well suited for such an analysis. The tribe is worldwide in distribution, and its 24 genera and 250 species vary enormously in size, morphology, and behavior.

Brown and Wilson showed that the main trend within the tribe Dacetini has been the shift from open foraging on the ground and low vegetation to an increasingly cryptic, subterranean existence. The change was associated in the early stages of dacetine evolution with a reduction in the variety of arthropods taken as prey. The most extreme modern forms specialize on entomobryomorph and sminthurid collembolans. There has been a concomitant reduction in body size, a possible shift from polymorphism to monomorphism, and an abandonment of recruitment trails to assemble masses of nestmates. Other anatomical changes include the development of fungus-like growths on the waist and other parts of the body, possibly containing substances that lure the collembolan prey, and a shortening of the mandibles.

The tendency to move from open, even partly arboreal foraging to cryptic, mostly terrestrial and subterranean foraging is not unusual in the ants. It is well-marked, for example, in the anatomically more primitive ponerine tribes Ponerini and Ectatommini. Haskins (1939) proposed the interesting theory that this trend in the ponerines resulted indirectly from competition with more recently ascendant, dominant aboveground foragers in the Dolichoderinae and Formicinae. A similar explanation is plausible in the case of the Dacetini. The early forms, most closely approached by Daceton and Orectognathus today, may have retreated when they came up against other myrmicines and the formicines during the great Tertiary radiation of the ants.

Whatever the prime mover in natural selection, the shift to a more cryptic existence incorporated important changes in hunting behavior of the dacetine ants. The relatively primitive long-mandibulate forms rely on the violent, trap-like action of their mandibles to secure prey (Figures 15-2 to 15-4). Their approach time to the prey is relatively short, and they often do not follow the mandibular snap with a stinging thrust. The short-mandibulate forms, on the other hand, have less shocking power in their mandibles. They rely more on stealth in approaching prey, holding on tenaciously after the mandibular strike and an immediate, consistent, and efficient use of the sting (Figures 15-5 and 15-6). The essential features of this evolutionary change is illustrated in the contrast between Strumigenys louisianae, a relatively primitive long-mandibulate species, and the phylogenetically more advanced, short-mandibulate Trichoscapa membranifera (Wilson, 1953a).

The Strumigenys, as exemplified by Strumigenys louisianae, are bolder and more direct in their manner of stalking prey. This trait is permitted by their more efficient mandibles, which are extremely long and operate very much like a miniature spring trap. Prior to the strike the mandibles are locked into position at nearly 180 degrees. This is accomplished by special teeth at their bases that catch on the lateral lobes of the labrum. When the ants tense the retractor muscles alone the mandibles cannot move. But when the labral lobes are also dropped the mandibles snap shut. The worker Strumigenys approaching a collembolan moves slowly and cautiously. It spreads its mandibles to the maximum angle and exposes two long hairs that arise from the paired labral lobes. These hairs extend far forward of the ant's head and serve as tactile range finders for the mandibles. When they first touch the prey, its body is well within reach of the apical teeth. A sudden and impulsive snap of the mandibles literally impales it on the teeth, so that drops of hemolymph often well out of the punctures. If the collembolan is small relative to the Strumigenys, the ant lifts it into the air and then may sting it. All but the largest collembolans are quickly immobilized by this sequence of actions, and struggling is feeble and short-lived.

The Trichoscapa membranifera worker is much more circumspect than the Strumigenys when stalking prey. As soon as the ant becomes aware of the presence of a collembolan, it “freezes” in a lowered, crouching posture and holds this stance briefly. If the collembolan is to its back or side, the worker now turns very slowly to face it. Once it is aligned in this way, the ant begins a forward movement so extraordinarily slow that it can be detected only by persistent and careful observation. Several minutes may pass before the ant finally maneuvers over less than a millimeter's distance to come within a striking position, and it may remain in this position for as much as a minute or longer. Unlike the Strumigenys, the Trichoscapa open their mandibles only to about a 60 degree angle. Tactile hairs are present and eventually come to touch the prey. The mandibular strike is as sudden as that of the Strumigenys, but since it is usually directed at an appendage, it does not have the same stunning effect on the collembolan. These insects often struggle violently to escape, but the ants are very tenacious and retain a fast grip until they are able to sting their prey into immobility.

To summarize, Strumigenys louisianae relies on a comparatively swift approach to its prey followed by a fixed-action pattern that can be characterized as strike-lift-sting, with the last element occasionally being omitted if the prey is small. In contrast, Trichoscapa membranifera employs a more cautious approach followed by strike-hold-sting, with the last element inevitable. The Strumigenys pattern is apparently typical for long-mandibulate dacetines generally, while that of Trichoscapa is typical for the short-mandibulate groups.

The ecological significance of the difference between the two groups of dacetines is evidently the following: the Trichoscapa pattern, requiring less space for the operation of the mandibles, is generally associated with cryptic foraging. Masuko (1984) has discovered an extreme version of the short-mandibulate technique in Epitritus hexamerus (see Figure 15-5). These bizarre little ants are essentially ambush hunters. The mandibles are directed slightly upward from the plane of the head, and the dorsal apical tooth is especially long and sharp, enabling the ant to strike with particular effectiveness at objects above its head. The Epitritus forager hunts a great deal in small crevices within the soil. Because of the tightness of the passages, it usually encounters the prey in front. The ant immediately crouches and freezes, pulling the antennae completely back into the scrobes that line the sides of the head. The mandibles remain closed. Even though the prey (a collembolan or small centipede) may be very close by, the ant never moves toward it. Instead, it remains perfectly still for periods of 20 minutes or longer, waiting for the prey to step on its head. Then, with a sudden upward snap of its mandibles, it impales the victim on the long apical teeth.

Two recent studies suggest that there is a good deal more to the story of predation by short-mandibulate dacetines than we have just recounted. Masuko (1984) found that the workers of some species smear soil and other detritus on themselves in one or the other of two methods that appear to be unique in the ants generally. In one version, performed by species of Epitritus, Labidogenys, Pentastruma, Smithistruma, and Trichoscapa, the worker pauses on moist soil and scratches the surface repeatedly with its forelegs. It then rubs the forelegs against its head, alitrunk, and middle and hind legs. The gaster is next stroked with either the hind or fore legs, in the latter case with the abdomen flexed forward beneath the rest of the body. The greatest effort is directed toward the dorsal head surface. In the second version, witnessed by Masuko in a species of Smithistruma, the worker picks up minute fragments of organic material (in one case at least, probably arthropod feces). Masuko suggests that body smearing camouflages the odor of the foragers and allows them to close on the prey more efficiently.

Dejean (1985b) found that workers of Smithistruma emarginata and S. truncatidens are attractive to entomobryomorph collembolans, compared to workers of Pheidole and blank controls. The source of the attractant is unknown, but a good guess is the spongiform appendage of the waist (illustrated in Figure 15-4), which Dejean showed to be underlain by glandular cells. Another possibility is the head, especially the labrum.

A striking convergence to the Dacetini has occurred in at least two other, phylogenetically independent groups of myrmicine ants, the widespread tribe Basicerotini (Brown and Kempf, 1960; Wilson and Brown, 1984) and the Oriental genus Dacetinops (Taylor, 1985). These ants are similar to dacetines in body form, pilosity, and predatory behavior. And like the dacetines, they hunt small, soft-bodied arthropods.

Egg predators
At least four phyletic lines of ants specialize to some degree on arthropod eggs: the cosmopolitan ectatommine genera Proceratium and Discothyrea, the African ponerine Plectroctena lygaria, and the Neotropical myrmicine genus Erebomyrma. The least specialized of the oophagous genera is Erebomyrma. The most specialized are Proceratium and Discothyrea, which, as Brown (1957, 1979) first noted, appear to collect spider eggs exclusively. Workers of at least the species Proceratium silaceum use their peculiarly downward pointing gastral tips to tuck the slippery eggs forward toward the mandibles when transporting them back to the nest.

The ant-termite arms race
Ants and termites are the “superpowers” of the insect world. In rain forest near Manaus, Brazil, they make up three quarters of the entire insect biomass (Fittkau and Klinge, 1973), a preponderance that probably holds in rough magnitude in many of the other major habitats of both tropical and temperate regions. Ants, being largely predators, are correspondingly the greatest enemies of termites. The two groups have undoubtedly been locked in a struggle for the greater part of the 100 million years of their coexistence, with ants comprising the active aggressors for the most part and termites the prey and resisters. In addition, both occasionally compete for scarce nest sites in rotting wood and leaf litter. They have engaged in a coevolutionary arms race that has produced the most elaborate weapons and battle strategies known in the animal world.

A large percentage of ant species, perhaps a majority--including Pheidole and Camponotus, the most speciose of all ant genera--prey on termites if given the opportunity. When a nest of termites is broken open by a large animal, windfall, or some other force, foraging ants nearby rush in to seize workers and nymphs before the termites can pull back and close the breaches in the nest. One of the most effective ways to locate the nests of some wide-ranging ant species, such as Gigantiops destructor and the species of Pachycondyla in tropical forests, is to scatter portions of a termite nest on the ground and track foraging workers as they carry termites home.

However, a few ant genera are specialized termite predators and need no assistance from investigators (see Table 15-1). The tropical zone of every continent has a guild of large ponerine ants that organize raids on termite nests. In particular, there are three species of Pachycondyla (= Termitopone) in Central and South America; Megaponera, Ophthalmopone, and Paltothyreus in tropical Africa; and Leptogenys of the processionalis group in tropical Asia. The raids of Megaponera foetens are one of the most impressive sights of tropical Africa, if one can scale one's image of size down a bit from rhinoceros and bongos, and they have attracted the attention of explorers and naturalists since the time of David Livingstone (1858). The expeditionary force of large black workers is led by a single scout. The raiders proceed along an odor trail in columns of two or more workers wide to the termite nest, whereupon, according to Arnold (1915), “the columns break up and pour into every hole and crack which leads into the invaded galleries. The method then adopted is as follows: Each ant brings to the surface one or more termites, and then re-enters the galleries to bring up more victims. This is continued until each has retrieved about half a dozen termites, which, in a maimed condition, are left struggling feebly at the surface. The whole army reassembles again outside, and each marauder picks up as many as it can conveniently carry, usually 3 or 4. The columns are then re-formed and march home.” While on the move the Megaponera frequently stridulate loudly enough to be audible to human observers as much as several meters distant. According to Arnold (cited by Wheeler, 1936c), the Megaponera colonies change their nest sites at frequent intervals, apparently in response to the need for fresh supplies of termites.

Paltothyreus tarsatus, as Arnold first noted, also raids termites. As in Megaponera, each Paltothyreus stacks as many termites in its mandibles as it can carry before starting home (Figure 15-7; Plate 18). Hölldobler (1984b) found that successful foragers, often carrying termites, recruit nestmates by laying chemical trails with secretions from intersegmental sternal glands located between the Vth and VIth, and VIth and VIIth abdominal sternites. According to Barry Bolton (cited in Longhurst et al., 1979b), Decamorium uelense is a small African myrmicine that appears to have arrived at the same specialization. The ants prey mostly on termites during the wet season, but during the dry season take a variety of soft-bodied insect larvae which are encountered underground. Decamorium uelense prefers Microtermes, a member of the fungus-growing subfamily Macrotermitinae. Microtermes species are litter-feeding termites that forage within their food sources of roots and plant debris. When scouts detect foraging Microtermes, they return to their nest and recruit a group of 10 to 30 nestmates. These ants attack and immobilize all termites they can capture by stinging them. Later a still larger group is recruited to retrieve the immobilized prey.

A wholly different strategy of termite hunting is employed by the Malaysian basicerotine ant Eurhopalothrix heliscata. The workers hunt solitarily through rotting wood housing the termites. They use their peculiar wedge-shaped heads, hard bodies, and short legs to press into tight spaces. They seize the appendages of the termites with their short, sharp-toothed mandibles, clasping these body parts of the prey even more tightly with the aid of heavily sclerotized labra that project like forks from between the bases of the mandibles (Wilson and Brown, 1984).

Beyond observations and a few experiments on the spectacular ponerine raiders, the myrmicine Decamorium, and the cryptic Eurhopalothrix huntresses, surprisingly little research has been directed at the other specialized termite predators. Many conjectures but few observations have been published on these ants, partly because their behavior is conducted in a secretive manner deep within the intact termite nests. One of the most common and least understood of the adaptive types are the thief ants, the small workers of which slip into the termite brood chambers to steal eggs and nymphs. All of the suspected species are myrmicines. They nest close to termite nests, sometimes in the walls of the termitaries themselves. They include Solenopsis (Diplorhoptrum) laeviceps, which is known to collect eggs of Nasutitermes under natural conditions (A. E. Emerson in Wheeler, 1936c); Erebomyrma urichi, which has been observed to steal Armitermes eggs in laboratory nests (Wilson, 1962d); the pantropical genus Carebara; the Neotropical genus Carebarella; the Asian genus Liomyrmex; and the African and Asian genus Paedalgus.

Except for sparse data on Solenopsis and Erebomyrma, the case for termite thievery, or “termitolesty,” is circumstantial (Forel, 1901; Wheeler, 1936c; Ettershank, 1966). To wit, the ants live close to the termites; the workers are blind and are rarely if ever seen away from the termitaria; they are small enough to slip unobtrusively in and out of the termite nest chambers; and they have no other known source of food. In the case of Carebara, the colonies produce large masses of queens each year, and it is difficult to see where the food and energy come from if not from the termite hosts. An illustration of the queen and worker castes is presented in Figure 15-8. The queens are immensely larger than the workers. The weight difference is over 4000 times (Wheeler, 1922: 170), the largest known in the ants and matched within the social insects as a whole only by the very same termites that are the hosts of the Carebara. Several observers have found the workers clinging to the tarsal hairs of queens during nuptial flights. Arnold (1916) said of Carebara vidua, “It is probable that the dense tuft of hairs on the tarsi of the female serve an important purpose--that of enabling some of the minute workers to attach themselves to the body of the female when the latter is about to leave the parental nest. Several specimens of the female have been taken by me with one or more workers biting on to the tarsal fimbriae. I am inclined to suspect that the young queen cannot start a new nest without the help of one or more of the workers from the whole nest, on account of the size of her mouthparts, which would probably be too large and clumsy to tend the tiny larvae of her first brood, and that it is therefore essential that she should have with her some workers which are able to feed the larvae by conveying to them the nourishment taken from the mouth of the queen.” Wheeler (1936c) concurred, observing that the problems of a founding queen without the little helpers would be comparable to those of a hippopotamus struggling to feed infants the size of mice. However, the matter cannot be settled by such logic, however neatly expressed. Anyone who has seen a giant leafcutter queen such as Atta texana delicately separating and cleaning her minute eggs during colony founding knows that size alone is not a fatal impediment. Still more impressive is the fact that the founding queen of an Asian Carebara kept by Lowe (1948) succeeded in rearing a brood of 16 workers without the aid of helpers.

The unsolved mysteries of Carebara are symptomatic of our general ignorance of the myrmicine ants associated with termites. We need detailed studies of the actual behavior of the so-called termitolestic forms as they make contact with their termite hosts. Some of the species, such as Liomyrmex aurianus (C. F. Baker in Wheeler, 1914a) have been reported to live in the same chambers as the termites with complete amity. It is even remotely possible that ant species exist that are true termitophiles, accepted by the termites as pseudo-members of their own colony. One example suggested by Wheeler (1936c) is Stigmacros termitoxena, a peculiar Australian formicine ant discovered in a nest of the termite Tumulitermes peracutus. Here is Wheeler's account of the single collection made: “On September 18, 1931, near Mullawa, West Australia, I came upon a colony of diminutive termites nesting under a flat stone in earthen galleries which they had built in a bunch of dry grass. On breaking open one of the galleries I saw several ants of the same size and color as the somewhat more numerous termites and moving about among them. After carefully collecting the occupants of the gallery and making allowance for escaping individuals, the ant-colony was found to comprise only 25 to 30 workers and a single ergatomorphic female (queen). I failed to find any additional ants in the termitary and saw no traces of their brood. The female and more than half the workers attracted my attention because their gasters were enormously distended. . . This distension (physogastry) was not due to liquid stored in their crops but to an unusual accumulation of fat. . . The ants belong to an undescribed species of Stigmacros, an exclusively Australian genus of which eleven species are described and of which Mr. John Clark and I have taken quite a number of unpublished forms. Although I have examined hundreds of Stigmacros from numerous localities in Eastern, Southern and Western Australia, I have seen no traces of physogastry except in the Mullawa specimens. Since the ants were living in what appeared to be friendly relations with their hosts, I suspect that they are fed by the termite workers and that the physogastry of the female and so many of the workers, like the physogastry of the termite workers and queens, is a result of this feeding.” Wheeler's argument is far from convincing, but the phenomenon he suggests is so extraordinary as to warrant a special effort to rediscover and study in detail the biology of S. termitoxenus.

The termites have not stood still in the face of the furious onslaught of the ants. They have responded by evolving an array of weapons and complex tactics of their own (Deligne et al., 1981; Quennedy, 1975, 1984; Grassé, 1986; see Figure 15-9). It is a general characteristic of termite caste systems that the worker caste is morphologically uniform but behaviorally very diverse when different species are compared, whereas the soldier caste is morphologically diverse and behaviorally uniform. The workers construct the nests that vary so drastically among the termite species, and they are also responsible for the greatest part of the phyletic diversity in food habits, foraging styles, and other behavioral properties of the species. The soldiers, on the other hand, are wholly specialized for the single function of defense.

In most species the soldiers are “mandibulate” forms, possessing large, heavily sclerotized heads, powerful adductor muscles, and sharp, elongate mandibles that seem clearly designed for biting and cutting rather than for such nondefensive functions as digging or handling of the brood. In Termes and several other genera of the “higher termites” (family Termitidae), the mandibles are shaped like scimitars, with points as sharp as needles. The soldiers of the kalotermitid genus Cryptotermes have cylindrical heads that serve as living plugs for the galleries of the nests. The mandibles are short and not particularly suited for defense. The head capsule, on the other hand, is thick-walled, heavily pitted, and truncate in front, so that it forms a barrier across the narrow galleries of the nest that any invader finds very difficult to get around. This “phragmosis” of the Cryptotermes soldiers closely parallels in structure and function that developed by certain ant species in the genus Camponotus and tribe Cephalotini. Even more bizarre are the “snapping” soldiers of Capritermes, Neocapritermes, and Pericapritermes. The mandibles are twisted, asymmetrical, and so arranged that their flat inner surfaces press against each other as the adductor muscles contract. When the muscles pull strongly enough, the mandibles slip past each other with a convulsive snap, in the same way that we snap our fingers by pulling the middle finger past the thumb with just enough pressure to make it slide off with sudden force. If the mandibles strike an ant or other insect, which seems to be the primary function of the action, a stunning blow is delivered. Even human beings receive a painful flick. The adaptation is similar to that evolved independently by ants of the genus Mystrium (Moffett, 1986a).

Finally, the premier combat specialists are the termite soldiers that employ chemical defense. When Protermes soldiers attack, for example, they emit a drop of pure white saliva which spreads between the opened mandibles. When they bite, the liquid spreads over the opponent. In general, the salivary glands of the soldiers are better developed than those of their worker nestmates, and they sometimes become grotesquely large in proportion to the rest of the body. The salivary reservoirs of Pseudacanthotermes spiniger swell out posteriorly to fill nine-tenths of the abdomen. The soldiers of Globitermes sulphureus, like those of at least one ant species of the Camponotus saundersi group (Buschinger and Maschwitz, 1984), are quite literally walking bombs. Their reservoirs fill the anterior half of the abdomen. When attacking, they eject a large amount of yellow fluid through their mouths, which congeals in the air and often fatally entangles both the termites and their victims. The spray is evidently powered by contractions of the abdominal wall. Occasionally these contractions become so violent that the wall bursts, shedding defensive fluid in all directions.

A separate strategy is the development of a “gun” consisting of a conical organ or “nasus” resembling a large nose on the front of the soldier's head. At its tip is an opening from which secretions of the frontal gland are discharged. Soldiers of some nasutitermitine genera are able to eject the material over distances of many centimeters. Their aim is quite accurate in spite of the fact that they are wholly blind.

Research during the past ten years has disclosed termites to be veritable factories of defensive chemicals. The frontal gland secretions of termitid soldiers, for example, constitute the richest source of monoterpene hydrocarbons known in the insects. Many idiosyncratic compounds have been identified in the secretions of rhinotermitid soldiers. One of them, 1-nitro-1-pentadecene, is the only nitro compound known among insect deterrent substances. This peculiarity of termite biology opens opportunities for new kinds of research. As Deligne and his co-authors (1981) say, “The dazzling variety of novel natural products that characterizes termite defensive secretions provides ideal compounds for studying the toxicology of compounds that may be highly selective insecticides. Hopefully, detailed investigations on the modes of action of these termite defensive products will be forthcoming in the near future.”

Untapped opportunities also exist in detailed studies of the relations of ants to termites. For example, Traniello and Beshers (1985) recently discovered that soldiers of a Neotropical termite, Nasutitermes costalis, are strongly recruited in response to intruders of the same species, while termites belonging to other species of Nasutitermes are less effective, and other termite genera are ignored. It will be useful to learn whether interspecific enemy specification occurs in termites as it does in ants, that is, whether especially dangerous enemies (in this case particular genera or species of ants) also trigger correspondingly stronger responses. In one suggestive study, Traniello (1981) found that the soldier caste of Nasutitermes costalis plays a major role in the foraging-defense system of the species in deterring attacks by ants on Nasutitermes costalis foragers. Conversely, we would also like to know if particular defensive techniques and secretions are broad-spectrum, in other words effective against a wide range of arthropod enemies, or whether they are tailored to combat certain kinds of ants. Might the specificity of predator-prey interactions and of competition between ants and termites play a role in community organization, as it does in ants?

Prey paralysis and prey storage
Workers of the South Asian ponerine Harpegnathos saltator are large ants with unusually long mandibles and large, widely separated eyes. Working as solitary huntresses, they capture cockroaches and spiders for the most part but also flies, butterflies, bugs, grasshoppers, cicadas and beetles (Maschwitz, 1981). Harpegnathos workers regularly sting the prey, even when offered a piece of a freshly killed cockroach (Figure 15-10). Maschwitz et al. (1979) provided experimental evidence that Harpegnathos saltator and Leptogenys chinensis paralyze prey by stinging and thereby are able to store prey a limited time. In one case the effect was observed to last for two weeks, and in no instance did the stung prey ever recover from the paralysis.

Even more remarkable is the storage of living prey by the Australian ponerines belonging to the Cerapachys turneri group. During a raid on a Pheidole nest the Cerapachys workers briefly sting each Pheidole larva and pupa before they pick it up and carry it to the home nest (Hölldobler, 1982b; Figure 15-11). Some prey larvae are stored inside the nest of Cerapachys for two months or longer, but they do not pupate or visibly increase in size. Under the microscope the prey larvae can be seen to move their mouthparts slightly, indicating that they are alive. In a series of additional experiments Hölldobler further confirmed that larvae captured by Cerapachys are preserved alive but in a state of metabolic stasis. Approximately 30 Pheidole larvae collected from a Pheidole colony were put without workers in a small test tube, which was kept moist by a wet cotton plug. A second similar test tube contained 30 Pheidole larvae taken from a Cerapachys nest. The larvae from the Pheidole colony were all dead after two weeks. In contrast, all of the larvae from the Cerapachys colony were alive after two weeks, many of them still moving their mouthparts. There is one other interesting aspect concerning the paralysis of prey larvae by Cerapachys. The Pheidole larvae are small and tender and the powerful Cerapachys sting can easily pierce the larva and thereby kill it. Thus the injections of the paralyzing secretions through the sting has to be very subtle in order to preserve without killing. A clue is presented by the differentiated pygidium with its denticulate margins, which is present in all workers and queens of cerapachyine ants (Brown, 1975; see Figure 15-12). Hölldobler's morphological and histological investigations indicate that these denticuliform and spinuliform setae on the pygidium of Cerapachys turneri and Sphinctomyrmex steinheili are probably sensory setae, in particular mechanoreceptors. It is most likely that during the stinging process the mechanoreceptors signal the gaster tip's touch to the prey larva, allowing a close control of the sting's penetration. Many of the nonsocial aculeate Hymenoptera that paralyze prey by stinging are equipped with mechanoreceptors on the tip of the sting sheath (Oeser, 1961; Rathmayer, 1978). Hölldobler did not detect similar structures on the tip of the sting sheath of the cerapachyines Cerapachys and Sphinctomyrmex. The stinging of prey larvae appears to be a very stereotyped behavior. When Hölldobler shook a Cerapachys colony containing Pheidole larvae out of the nest chamber, Cerapachys workers retrieving Pheidole larvae almost invariably went through the typical stinging motion. They did not do this, however, when they picked up their own larvae. Workers, queens and larvae of Cerapachys all feed on the Pheidole larvae. The food storage system appears to enable these specialized predators to stay inside their nests for long intervals. Experiments have shown that they do not conduct raids so long as they have a good supply of prey larvae available.

The paralytic storage of prey appears to be quite common in ponerine ants. Amblyopone pallipes and species of Onychomyrmex paralyze captured centipedes by stinging (Traniello, 1982; Hölldobler and Taylor, unpublished observations). The immobilized prey can more easily be retrieved into the nest, where it also can be stored for some time. In fact, Hölldobler and Taylor found immobilized centipedes in Onychomyrmex bivouac nests, which did not exhibit any signs of feeding by the ants. Observations made by Masuko (1987, and personal communication) suggest that Leptanilla japonica also paralyze large centipedes. If the prey is too large to be transported, the entire colony moves to the immobilized prey to feed on it (see Figure 16-16a). Paltothyreus tarsatus workers frequently sting the captured termites before they stack them between their mandibles (Hölldobler, 1984b). It is possible that in this case surplus termite prey is also preserved by paralysis. Finally, Nothomyrmecia huntresses regularly sting their prey objects, even when the prey is small and can be easily subdued with the ant's powerful mandibles (Hölldobler and Taylor, 1983). Here too, it is suggestive that the surplus prey can be preserved and stored inside the nest, which makes the colony less vulnerable to climatic irregularities affecting hunting success.

Masters of camouflage
Until recently, the ants of the Neotropical genus Basiceros were thought to be relatively rare. Despite the relatively large size of the worker, no more than ten colonies had been collected, and almost nothing was known of the biology of the genus. However, after we had developed a “search image” for Basiceros manni in Costa Rica, based in part on the white larvae and pupae that stood out against dark nest material, we began to locate colonies with ease. The reason for the previous obscurity of Basiceros is the superb ability of the workers at camouflage. To the human eye, and presumably that of many visually orienting predators such as birds and lizards, the ants are difficult to see as they walk over the ground, and virtually invisible when standing still (see Plage 19). The effect is achieved in part by the fact that Basiceros manni workers are among the most slowly moving ants we have encountered in our joint lifetime of experience, which extends to more than 180 of the approximately 300 ant genera found worldwide. When observed in an undisturbed state, the entire worker force often stands perfectly still for minutes at a time, even holding their antennae in place. And when workers in motion are disturbed by being uncovered or touched by a pair of forceps, they freeze into immobility for up to several minutes (Wilson and Hölldobler, 1986).

The effect is enhanced by the gradual accumulation of fine particles of soil on the bodies of the worker. By the time the workers are old enough to forage, they blend in remarkably well with the soil and rotting litter over which they walk. The collection of the particles is accomplished by two layers of hairs on the dorsal surfaces of the body and outer surfaces of the legs: longer “brush” hairs with splintered distal ends, and shorter, feather-like “holding” hairs. The brush hairs evidently scrape or otherwise capture the soil particles, while the holding hairs help to keep them in place next to the surface exoskeleton (see Figure 15-13). The same phenomenon occurs widely through the other genera of the tribe Basicerotini, including Eurhopalothrix, Octostruma, and Protalaridris, as well as Stegomyrmex, the sole member of the tribe Stegomyrmecini (Hölldobler and Wilson, 1986c).