The Ants Chapter 13

CHAPTER 13. SYMBIOSES WITH OTHER ARTHROPODS

For more than a century it has been known that many species of insects and other arthropods live with ants and have developed a thriving symbiotic relationship with them. Most do so only occasionally, functioning as casual predators or temporary nest commensals. But a great many others are dependent on the ant society during part or all of their life cycles. These ant guests, commonly known as myrmecophiles, include a great variety of beetles, mites, collembolans, flies, and wasps, as well as less abundant representatives of a wide range of other insect groups (see Table 13-1). Myrmecophily is almost exclusively an invertebrate phenomenon. A respectable list of vertebrate species, including synbranch eels, frogs, lizards, snakes, birds, small and medium-sized mammals and even primates occasionally live with social insects or prey upon them. But very few are specialized for such an association (Myers, 1929, 1935; Hindwood, 1959; Scherba, 1965; Chew, 1979; Vogel and von Brockhusen-Holzer, 1984; Redford, 1987). The vast majority of obligate invertebrate symbionts are moreover arthropods, and it is among these organisms that the most striking adaptations for life with social insects have taken place. A number of these myrmecophiles make their homes in nests of the ants and enjoy all the social benefits of their hosts. Although the interlopers in some cases eat the brood, the ants treat the guests with astonishing tolerance: they not only admit the invading species to the nest, but often feed, groom and rear the guest larvae as if they were the ants' own young.

An ant colony possesses a complex system of communication (see Chapters 5 and 7) that enables it not only to carry out its collaborative activities in food gathering, brood care, and other social activities, but makes possible an instant recognition of nestmates and discrimination of foreigners. This identification and discrimination system functions like a social immune barrier: only colony members are allowed to enter the ant society, and alien individuals are harshly rejected. Nevertheless, by using various techniques, a considerable number of solitary arthropods have managed to penetrate ant nests. The fact that the ants treat many of these alien guests amicably suggests that the guests have broken the ants' communication and recognition code. In other words, they have attained the ability to “speak” the ants' language of mechanical and chemical cues.

History of myrmecophile studies
Before addressing this phenomenon in greater detail, it will be useful to review briefly the historical background of the study of myrmecophilous symbiosis. Erich Wasmann initiated the modern study of the subject. In 1894 and in subsequent publications he developed a classification that divides species into five behavioral categories, which reflect increasing levels of integration into the social system of their hosts:

1. Synechthrans. These arthropods are treated in a hostile manner by their hosts. They are predators for the most part, managing to stay alive by means of greater speed and agility or the use of defensive mechanisms such as repellent secretions and retraction beneath shell-like cuticular shields.

2. Synoeketes. These arthropods, which are also primarily scavengers and predators, are ignored by their hosts because they are either too swift or else very sluggish and apparently neutral in odor.

3. Symphiles. Also referred to occasionally as “true” guests, these symbionts are accepted to some extent by their hosts as though they were members of the colony.

4. Ectoparasites and endoparasites. These arthropods are conventional parasites. They live on the body surfaces of their hosts, or lick up their oily secretions, or bite through the exoskeleton and feed on their hemolymph, or penetrate the body itself.

5. Trophobionts. This category includes phytophagous homopterans, heteropterans, and lycaenid caterpillars that are not dependent on the social insects for food but instead supply their hosts with honeydew and nutritive glandular secretions. In exchange they receive protection from parasites and predators.

With the accumulation of more detailed information on the behavior of the symbionts in recent years, however, the Wasmannian classification has turned out to be considerably less than perfect. Many symbionts fit into more than one category. Symphiles, for instance, not only exist on the charity of their hosts, grooming and soliciting food from them, but also prey simultaneously on the hosts and their brood.

Several alternative schemes have been proposed to categorize the many life styles of myrmecophiles, by Delamare Deboutteville (1948), Paulian (1948), Akre and Rettenmeyer (1966), and Kistner (1979), respectively. In a more modern, ecological twist, Kistner distinguishes two major categories: (a) integrated species, “species which by their behavior and their hosts' behavior can be seen as incorporated into their hosts' social life”; and (b) non-integrated species, “species which are not integrated into the social life of their hosts but which are adapted to the nest as an ecological niche.”  Considering the complex diversity of myrmecophilous adaptations it is often difficult nevertheless to draw a clear distinction between “integrated” and “non-integrated” symbionts. Thus despite its considerable shortcomings, the original Wasmann nomenclature continues to be useful in labeling the majority of cases, and it is frequently employed as a kind of shorthand in the literature on social symbioses.

Diversity of myrmecophiles
The literature on myrmecophiles is enormous and growing each year, much of it consisting of incidental notes buried in taxonomic and ecological studies of selected genera and higher taxa. A very extensive review of solitary symbionts in insect societies was recently published by Kistner (1982). Much of the information on myrmecophiles is summarized in Table 13-1, which is based on the original table compiled by Wilson (1971), together with new information provided by Kistner's review and other publications.

Not surprisingly, the cumulative data reveal that certain taxa are much more preadapted for life as symbionts than others. The mites (Acarina), for example, are the foremost representatives among ectosymbiont species in terms of sheer numbers of individuals. Although there exist only a few quantitative faunistic investigations of symbionts in ant colonies, Rettenmeyer's study (1962a) of 150 colonies of army ants in Kansas and Panama is characteristic: he counted almost twice as many ectosymbiotic mites as all the other myrmecophiles taken together. Also, Kistner (1979) reports that he collected 3,288 mites from a single colony of the army ant Eciton burchellii. Mites find easy entry into nests either as scavengers that are too small and quiescent to be evicted by their hosts, or else as ectoparasites adapted for life on the body surface of the ants.

Myrmecophilous mites are rivaled in diversity by staphylinid beetles, a family of approximately 28,000 species. Staphylinids, like acarines, are predisposed to life in ant and termite nests by virtue of their preference for moist, hidden environments and the role they commonly assume as generalized scavengers and predators. The greatest variety of staphylinid symbionts is found in colonies of the “true” army ants, that is, the dorylines and ecitonines (Seevers, 1965; Akre and Rettenmeyer, 1966; Kistner and Jacobson, 1975; Kistner, 1979). But they also commonly occur in nests of nonlegionary ant species. Kistner (1982) lists 19 staphylinid genera recorded from nests of attine ants, 17 genera found with Solenopsis fire ants, and approximately 15 genera known to occur in nests of formicine ants. Diversity aside, probably the most abundant insect guests of social insects in general and army ants in particular are phorid flies. When all species are combined, as many as 4000 adults occur in a single ecitonine army ant bivouac (Rettenmeyer and Akre, 1968).

Another safe generalization is that by far the greatest diversity of species of myrmecophiles, measured either per host species or per host colony, is found with species that form exceptionally large mature colonies. The ultimate in this trend is found in the great faunas of symbionts that live with ecitonine and doryline army ants, the meat ants of the genus Iridomyrmex, and the leafcutters of the genus Atta, the nests of which normally contain from hundreds of thousands to millions of inhabitants. An exceptional variety of guests has also been recorded from large colonies of Hypoclinea in tropical Asia and the north temperate species of the Formica rufa group. By contrast, very few symbionts are known from nests of species with the smallest mature colony sizes, including the great majority of ponerine, dacetine and leptothoracine ants.

This last rule of population size lends itself readily to theoretical explanation. The insect colony and its immediate environment can be thought of as an ecological island, partitioned into many microhabitats that symbiotic organisms are continuously attempting to colonize (Wilson, 1971; Hölldobler, 1972). In general, species with the largest mature colony size possess the greatest diversity of ecological niches and also enjoy the longest average span of mature colony life. The more diverse the microhabitats presented by the host colony, the greater the potential diversity of symbiont species. Furthermore, a long colony life increases the probability that symbiotic propagules will penetrate a given colony at some time or other. It is also true that if colony size is large, the equilibrial population size of its symbionts will be proportionately large and their species extinction rate (measured as the number of symbiont populations going extinct per colony per unit of time) will be correspondingly low. In general: large colony size enhances three factors--long colony life, high microhabitat diversity, and low symbiont extinction rates--that reinforce each other to produce a higher diversity and abundance of symbiotic species.

The evidence reviewed by Wilson (1971) and Kistner (1979, 1982) shows that many species are able to maintain a viable population size only under the protection of large colonies. The great majority of very specialized species are uncommon, and many are rare and local in distribution.

The guests of army ants: an extremely rich community
The species of mites associated with army ants, according to Rettenmeyer, can be conveniently divided into two main groups: those that live in the refuse piles or “kitchen middens” of the ants, and those that live on the bodies of the ants or within their bivouacs. The latter group, the only true myrmecophiles in this case, have finely apportioned the environment through amazing feats of specialization. Several of the extreme adaptations are illustrated in Figure 13-1. Of special interest are the Circocyllibanidae, which are phoretic on adult workers but occasionally ride on larvae, adult males and queens (Elzinga and Rettenmeyer, 1975). They appear to live primarily on certain parts of the host bodies such as the mandibles, head, thorax and gaster of adult workers. Of equal interest are the Coxequesomidae which are specialized to live on the antennae or coxae of their Eciton hosts. The most extraordinary adaptation of all, however, is exhibited by the macrochelid, Macrocheles rettenmeyeri, which is a true ectoparasite. This mite feeds on blood taken from the terminal membranous lobe (arolium) of the hind tarsus of large media workers of its exclusive host species, Eciton dulcium. In the process it allows its entire body to be used by the host ant as a substitute for the terminal segment of the foot. Rettenmeyer (1962a) described its behavior as follows: “The mite inserts its chelicerae into the membrane, but it is not known if the palpi assist in holding to the ant. Presumably the mite feeds in this position and may be considered a true ectoparasite rather than a strictly phoretic form. . . . Army ants characteristically form clusters by hooking their tarsal claws over the legs or other parts of the bodies of other workers. Temporary nests or bivouacs of Eciton dulcium crassinode may have clusters several inches in diameter hanging in cavities in the ground. In small laboratory nests, when a worker hooked the leg with the mite onto the nest or another worker, the hind legs of the mite usually served in place of the ant's tarsal claws. The hind legs of the mite were never seen in a straight position but were always curved. In all cases the entire hind legs of the mite served as claws, not just the mite's claws. No difference was noted in the behavior of the ant whether its own claws or the hind legs of the mite were hooked onto some support. When no other part of the ant's leg was touching any object, adequate support was provided by the mite's legs for at least five minutes.” A second species, M. dibanos, was observed to parasitize Eciton vagans in the same manner.

Many of the mites found with Neotropical army ants are endowed with particular holdfast mechanisms, such as modified claws, empodia, specialized ridges and teeth on their enlarged dorsa, and highly modified posterior idiosomas that adapt the mites to specific regions of the hosts' bodies (Elzinga, 1978).

Mites, being much smaller than their hosts, function primarily as ectoparasites or scavengers within the army ant nests. Many of the insects, on the other hand, either scavenge or obtain part of their nutriment by licking the oily secretions of the bodies of the hosts, a specialized behavior referred to by Wheeler and a few other authors as “strigilation.” The nutritive value of strigilation is open to some question, at least in the case of the ecitophiles. Both Rettenmeyer and Akre have suggested that it also serves to transfer the host odor to the parasites. This hypothesis received support from the recent finding by Vander Meer and Wojcik (1982) that the scarabaeid beetle Myrmecaphodius excavaticollis acquires colony specific cuticular hydrocarbons from its Solenopsis hosts. Other insects feed on the booty brought in by the army ant host workers, or else prey on the hosts themselves, especially the defenseless larvae and pupae. As illustrated in Figure 13-1, the insect guests vary greatly in size and in the positions they take on the bodies of the host workers. They also vary in the locations they prefer within the bivouac as a whole and the ant columns during colony emigrations and raids.

A typically instructive case of positional specialization is provided by the aleocharine staphylinid Tetradonia marginalis. Akre and Rettenmeyer (1966) note Tetradonia marginalis “is the most common staphylinid found in refuse deposits of Eciton burchellii and in the emigrations. Sometimes more than 100 individuals can be found in a single large deposit of this host. Many of these myrmecophiles do not run along the emigration columns but apparently fly from one colony to another. Tetradonia marginalis was the species most frequently attacking uninjured, active workers as well as injured workers of both Eciton burchellii and Eciton hamatum along edges of columns and near bivouacs [Figure 13-2]. The staphylinids were not seen to attack workers running along in the middle of raid or emigration columns or those running quickly along the edges.”

During recent field work in La Selva, Costa Rica, we were able to confirm some of the observations of Akre and Rettenmeyer. We were especially intrigued by the beetles' spatial positions at the end of the emigration columns and the hunting techniques they employed while there. For an hour late in the evening of March 28, 1985, we watched as a large part of the emigration column of Eciton hamatum crossed a forest path. At the very end of the column, with no more than 100 stragglers on that portion of the trail, we spotted ten or more of Tetradonia marginalis running swiftly along the trail, abdomen curled up over the thorax. Most were attacking Eciton media workers, and one was attacking a major. All of the victims were slow and partly disabled, unable to walk swiftly. We speculated that they had fallen back due to injuries or sickness, and had been singled out by the beetles for attack. The staphylinids circled around, seizing the legs or tips of the abdomens, in the latter case seeming to bite the anal region of the ant. Some dragged their victims swiftly along the trail by a leg and even those appearing to bite the anal region started off in this manner within a few minutes. In one instance two beetles attacked an ant together, running along with it in a coordinated fashion. One staphylinid flew off when we tried to capture it.

Akre and Rettenmeyer (1966) made a distinction between “generalized” species of symbiotic staphylinids, having the typical appearance of most Staphylinidae and with no obvious modifications for life with ants, and “specialized” species, which include mimetic forms (to be described later in this chapter). The specialized staphylinids are distinguished by the possession of rigid abdomens that are held in a single position (Figure 13-3). Paradoxically, these insects are also less common than the nonmimetic forms in the host colonies. Akre and Rettenmeyer compiled an impressive list of consistent behavioral differences between the two kinds of species, which we have reproduced in Table 13-2. These data reveal a gradually increasing degree of behavioral integration with the host society, which may constitute the most obvious of the evolutionary pathways along which progress in symbiosis can be measured. The two kinds of species can be reasonably interpreted as progressive stages in an adaptation to the host ants and their way of life.

The ant colony as an ecosystem
In more general terms, the concept of an insect colony as an ecological island, or, to use a slightly more precise language, a partially isolated ecosystem, can be extended to gain a better understanding of certain aspects of the biology of the symbionts. The ant colony and its surroundings is richly structured into many diverse microhabitats, such as the foraging trunk routes, refuse areas, peripheral nest chambers and guard nests, storage chambers, brood chambers with separate areas for pupae, larvae and eggs, queen chambers, and the bodies of adult and immature inhabitants of the nest. These microhabitats are occupied by a diversity of symbionts, which show special adaptations to each of the niches in turn.

The body of the host ants. The bodies of ants are occupied internally by a diversity of endoparasites, including nematodes (Gösswald, 1938b), termatodes (Hohorst and Lämmler, 1962; Schneider and Hohorst, 1971) and cestodes (Muir, 1954; Plateaux, 1972; Buschinger, 1973b). This form of parasitism can affect the individual host's morphology (Gösswald, 1938; Kutter, 1958), metabolism (Kloft, 1949) and behavior (Hohorst and Graefe, 1961), but it usually has little impact on the social behavior of the host ants. We will therefore not discuss this subject in greater detail.

One endoparasite, however, deserves special mention, namely the tachinid fly Strongygaster globula (formerly called Tamiclea globula), whose larvae develop as endoparasites inside the gaster of colony founding queens of Lasius niger and Lasius alienus (Gösswald, 1950). The behavior of the infected queens is not noticeably affected, except that they cease to lay eggs. When the last instar larva of the parasite leaves the host's abdomen through the cloaca, it quickly pupates and is groomed and tended by the host ants as if it were a member of the ants' own brood (Figure 13-4). In contrast, the fly imago is not treated amicably by the ant and has to leave the incipient ant nest quickly. The ant queens infected with Strongygaster die after the parasites leave the nest.

The bodies of immature and adult army ants, as noted earlier, provide special microhabitats for a diversity of ectosymbionts. The same is true for nonlegionary ant species. Probably the best known examples are species of the antennophorid mite genus Antennophorus, which live with formicine ants of the genera Lasius and Acanthomyops (Janet, 1897b; Wasmann, 1902; Karawajew, 1906; Wheeler, 1910b). The mites ride on the bodies of the ants, shifting their positions when two or more are present on the same ant so as to produce a balanced load. When three mites are present, two ride on either side of the gaster, and the third on the lower surface of the head (Figure 13-5). When four are present two position themselves on either side of the abdomen and head, respectively. The workers first try to remove the mites from their bodies, but seem to give up after several unsuccessful attempts.

The Antennophorus live on food regurgitated by the ant hosts, either imbibing it as it passes between workers or soliciting it directly by stroking the mouthparts of workers with the long, antennal-like forelegs. In the latter case the mites evidently imitate the tactile signals used by the ants themselves. It has even been observed that mites attached to the gaster of one worker are able to solicit food from another. The mites prefer to mount newly eclosed workers. This habit confers a distinct advantage: young workers usually stay inside the nest, and they are attractive to older workers who frequently feed them by regurgitation, affording the mites plenty of opportunity to participate in the trophallaxis.

A remarkable analogy to Antennophorus in its ectosymbiotic adaptations has been discovered by K. Hölldobler (1928) in the phorid fly Metopina formicomendicula, which lives in nests of the thief ant Solenopsis fugax. The Metopina adult mounts thief ant workers and rapidly strokes the head and mouthparts of the ant with its forelegs. The worker usually responds by slightly raising her head, opening her mandibles, and regurgitating a droplet of food, which is then rapidly imbibed by the fly (Figure 13-6). The ants occasionally attack the phorids when they move through the nest, but they rarely succeed because the flies are too swift and elusive. When resting the Metopina frequently ride on the queen, where they are usually ignored by the workers. Karl Hölldobler (1951) categorized Metopina as a “synoekete of the evasive type.”

Trails and kitchen middens. Many guests of army ants, and especially the staphylinid beetles, regularly accompany raiding columns, where they prey on the captured booty of their hosts or on the host ants themselves. These myrmecophilous scavengers and predators, as exemplified by Tetradonia marginalis, often bring up the “rear guard” of the raiding columns, sprinting deftly along each twist and turn of the newly vacated trails. It has been known for a long time that guests of army ants can orient by the odor trails without further cues provided by the hosts. Since army ants frequently change the location of their bivouacs, it is important for the myrmecophiles to be able to follow the emigration trails. Akre and Rettenmeyer (1968) investigated this phenomenon systematically by exposing a large variety of guests of ecitonine army ants to natural odor trails laid over the floor of the laboratory arena. They found that nearly all of the species tested, including a haphazardly chosen sample of Staphylinidae, Histeridae, Limulodidae, Phoridae, Thysanura, and Diplopoda, were able to orient accurately by means of the trails alone. However, some myrmecophiles did better than others. For example, the staphylinids Ecitomorpha, Ecitophya, and Vatesus, which regularly accompany ant columns in close formation, followed experimental trails readily and accurately, whereas phorid flies, which under natural conditions exhibit a kind of meandering movement along emigration trails, deviated more frequently from the experimental trails. Other species, such as histerid beetles, which normally ride on host ants or booty in ant columns, did relatively poorly in the trail-following tests.

In some instances the guests were more sensitive to trail pheromones than the ants themselves. In choice tests the symbionts preferred trails laid by their own host species to those of other army ants, and sometimes they were repelled by the trails of the wrong species. However, a few instances were recorded in which the ecitophiles were less specific in their response. Interestingly, the behavior was partly correlated with the myrmecophiles' acceptance of several host species in nature. None of the symbionts showed an ability to distinguish the trails of its own host colony in competition with trails laid by other colonies belonging to the same species.

Trail following may also be a means by which parasites locate the host colony. Kistner (1979) observed a limuloid beetle Mimocetes as it flew into a raiding column of Dorylus, then zigzagged from one side to another for about 10 cm, and finally entered the thick swarm of driver ants. The cockroach Attaphila fungicola lives with the fungus growing ant Atta texana, the workers of which depend strongly on odor trails for orientation during foraging expeditions and colony emigrations. Moser (1964) demonstrated that Attaphila fungicola follows the trail pheromone of its host ants, which he extracted directly from the poison gland of the workers and laid arbitrarily along selected routes in the laboratory. This experiment substantiates an earlier observation by Bolivar (1905) of Attaphila schuppi following trails of Acromyrmex in the field in Brazil. Moser also noted that the cockroaches do not depend solely on the ability to disperse from nest to nest, because individuals have been observed riding on the backs of Atta texana queens during the nuptial flights of these ants. The myrmecophilous cricket Myrmecophila manni probably also disperses along the foraging trails of its host, the mound-building ant Formica obscuripes. Henderson and Akre (1986a,b) recently observed a total of 63 crickets, including both sexes and all nymphal instars, on foraging trails of Formica obscuripes more than 20 meters distant from the nests. During June and July these insects were seen running on the trails nearly every evening from dusk to almost midnight.

It is well known that caterpillars of the lycaenid Maculinea teleius change from phytophagous to myrmecophilous habits after casting the skin of the third larval instar. It was previously assumed that the myrmecophilous stage is carried by its host ants, Myrmica rubra, into the host nest, where the Maculinea larvae prey on the ant brood (Malicky, 1969). However, Schroth and Maschwitz (1984) recently found that fourth instar larvae of Maculinea teleius actively move into the host ants' nest, thus confirming an earlier observation by Chapman (1919). In laboratory experiments Schroth and Maschwitz demonstrated that the lycaenid larvae follow trail pheromones of their host ants when searching for the nests.

Trails and ant paths not only serve as orientation cues, they also inadvertently create a diversity of ecological niches within which the symbionts specialize. The Eciton army ants and their attending ecitophiles provide some of the most striking examples of modes of adaptive radiation. Another, wholly unexpected case involves the adult females of three nymphalid butterfly species, Mechanitis isthmia, M. doryssus, and Melinaea imitata, which regularly follow raiding swarms of the army ant Eciton burchellii. Ray and Andrews (1980) found that these butterflies feed on bird droppings in the swarm vicinity. Many species of birds, mostly in the antbird family Formicariidae, deposit droppings as they follow raiding swarms of Eciton burchellii to feed on insects flushed from the leaf litter by the ants (Willis and Oniki, 1978). Thus, army ant swarms provide an indirect yet reliable and relatively rich source of nutrients in the form of bird droppings. Although no experimental proof has yet been attempted, it is most likely that odors associated with the swarms may allow the butterflies to orient to the general vicinity of the droppings.

The trails and paths of nonlegionary ant species also harbor a diversity of myrmecophiles. One of the most fascinating examples is the nitidulid beetle Amphotis marginata, the “highwaymen” of the local ant world (Figure 13-7). These beetles occupy shelters along the foraging trails of the formicine ant Lasius fuliginosus during the day. At night they patrol the trails and successfully stop and obtain food from ants returning to the nest. Ants that are heavily laden with food are easily deceived by the beetles' simple solicitation behavior. The Amphotis adult induces an ant to regurgitate food droplets by using its short antennae to trap the ant's labia and rapidly drum on its head (Plate 16). Soon after the beetle begins to feed, however, the ant seems to realize it has been tricked and attacks the beetle. The beetle then is able to defend itself simply by retracting its appendages and flattening itself on the ground. With the aid of special setae on its legs it firmly attaches its lower body surface to the ground. The ant is then unable to lift the beetle or turn it over (Hölldobler, 1968, unpublished observations). This mechanism appears to give the beetle nearly complete protection. A somewhat similar parasitic pattern is exhibited by the calliphorid fly genus Bengalia. This paleotropical form has been reported to steal prey and larvae from columns of doryline driver ants and trails of a large number of nonlegionary ant genera (Bequaert, 1922; Maschwitz and Schönegge, 1980).

A diversity of insect predators settles near ant trails and preys on the ants. They employ an astonishing repertory of hunting techniques to capture their prey without being attacked or killed by the usually very aggressive ants. Some of the most specialized predators occur among the Hemiptera (Jordan, 1937). One remarkable hunting technique is employed by the reduviid bug Acanthaspis concinnula (Mühlenberg and Maschwitz, 1976). This species lives near the nests of the fire ant Solenopsis geminata. The bug captures Solenopsis workers, immobilizes them by injecting poison, and then sucks out the hemolymph. Then it places the shriveled corpses on its back. The accumulated shield of dead victims provides an excellent disguise and even attracts other ant workers, which approach to inspect their freshly killed nestmates. A similar “shielding behavior” has been described in some hunting spiders (Oliveira and Sazima, 1984). A wholly different method is employed by the reduviid bug Ptilocerus ochraceus (Figure 13-8) of Java, which feeds to a large extent on the dolichoderine ant Hypoclinea biturberculata, an extremely abundant ant in southeastern Asia (Jacobson, 1911). The predators take up a position in or near an ant path and offer an intoxicating attractant secreted from trichome glands on the underside of the bug's abdomen. On encountering a Hypoclinea worker the bug presents the trichomes by raising the front of the body, simultaneously lifting up its front legs, “. . . folding them in such a manner that the tarsi nearly meet below the head. The ant at once proceeds to lick the trichomes, pulling all the while at the tuft of hairs, as if milking the creature, and by this manipulation the body of the bug is continually moved up and down. At this stage of the proceedings the bug does not yet attack the ant; it only takes the head and thorax of its victim between its front legs, as if to make sure of it; very often the point of the bug's beak is put behind the ant's head, where it is joined to the body, without, however, doing any injury to the ant. It is surprising to see how the bug can restrain its murderous intention as if it was knowing that the right moment had not yet arrived. After the ant has indulged in licking the tuft of hair for some minutes the exudation commences to exercise its paralyzing effect. That this is only brought about by the substance which the ants extract from the trichomes, and not by some thrust from the bug, is proved by the fact, that a great number of ants, after having licked for some time the secretion from the trichomes, leave the bug to retire to some distance. But very soon they are overtaken by the paralysis, even if they have not been touched at all by the bug's proboscis. In this way a much larger number of ants is destroyed than actually serves as food to the bugs, and one must wonder at the great profligacy of the ants, which enables them to stand such a heavy draft on the population of one community. As soon as the ant shows signs of paralysis by curling itself up and drawing in its legs, the bug at once seizes it with its front legs, and very soon it is pierced and sucked dry” (Jacobson, 1911).

A similar waylaying technique is employed by larvae of the carabid beetles Sphallomorpha colymbetoides and Sphallomorpha nitiduloides, which prey on the Australian meat ants Iridomyrmex purpureus, also a dolichoderine species. Moore (1974) discovered that the beetle larvae dig burrows near the nests and paths of the meat ants and wait for passing Iridomyrmex workers. When an ant approaches, the larva lunges at her, grasping the ant's legs and holding her for one or two minutes. The struggling ant soon relaxes, apparently in a state of paralysis. Moore speculates that the larva might apply toxic secretions, possibly from its maxillary glands. This, however, has not yet been tested experimentally. Once the prey is motionless, the larva sucks out the hemolymph of the ant and then discards the shriveled corpse. Often several Sphallomorpha burrows are found grouped together. Moore believes that the aggregation improves the predators' hunting success, because captured meat ants discharge an alarm pheromone that attracts other ants from nearby.

A similar ecological niche near ant nests or ant paths is occupied by ant lions, which are members of the order Neuroptera. The larvae of these insects build funnel-shaped pit traps in loose soil, where they wait on the bottom for ants to slip and fall down to them (Wheeler, 1930). However, the ant lion larvae are not exclusively predators on ants. Of 222 prey of the species Myrmeleon immaculatus identified by Heinrich and Heinrich (1984), only 36 percent were ants.

Yet another, very different technique exploiting trails and kitchen middens is employed by the European staphylinid beetle Pella funesta (Hölldobler et al., 1981). In early spring the adult beetles deposit eggs near the kitchen middens of the formicine ant Lasius fuliginosus. The larvae develop in the refuse pile and pupate sometime during May to July. In July or August the adult beetles eclose. The young adult beetles then apparently emigrate, as indicated by a short phase of high diurnal locomotory and flight activity. After this period the beetles hunt the ants near the Lasius fuliginosus nest during the night, while remaining hidden in convenient shelters during the day. The Pella adults overwinter in dormancy inside the Lasius nest. At the end of the winter they enter a second diurnal activity phase during which mating takes place. After reproduction the beetles die, normally a few weeks before the new adult beetle generation ecloses in June.

The larvae of Pella funesta are as closely associated with the ants as the adults. They are found almost exclusively near or in the kitchen middens of the Lasius fuliginosus nest, where they subsist primarily on dead ants. While feeding, the larvae always attempt to stay out of sight, either by remaining entirely beneath the booty and devouring it from below, or by crawling inside the bodies of their victims. Occasionally two to four larvae feast on the same corpse. When the beetles become overly crowded, however, they attack and cannibalize each other (Figure 13-9).

When ants encounter the Pella larvae they usually attack them. Almost invariably the larvae raise their abdominal tips towards the heads of the ants in a defensive maneuver. Pella larvae possess a complex tergal gland near the abdominal tip, which may secrete appeasement or diversion substances. At least the ants respond to the movement by stopping their attack and palpating the abdominal tip of the larva. In most cases this brief interruption is sufficient to allow the larva to escape. In general, Pella larvae are able to come into close contact with workers of Lasius fuliginosus without being attacked, and this is especially the case when the temperature is relatively low (14-17°C). Under these circumstances larvae have been observed licking the cuticle of live ants, including even the mandibles and mouthparts. Tracer experiments with radioactive labeled food did not, however, indicate that the larvae are able to solicit regurgitation from the ants. Their main food source seems to be dead ants and nest debris. They can easily be raised in the laboratory entirely apart from living ants, provided they are given a regular diet of freshly killed ants. The genus Pella was originally combined with Zyras and Myrmedon, but raised to full generic rank by Kistner (1972). Wasmann (1886, 1920, 1925, 1930) reported that all species he had studied, including cognata, funesta, humeralis, laticollis, lugens, and similis, live with Lasius fuliginosus. Of these, only humeralis uses alternate hosts, in this case species of the Formica rufa group. According to Wasmann all Pella species prey on ants, concentrating especially on disabled adults. In addition Wasmann observed that the beetles are active primarily during the night. In a more recent study Kolbe (1971) failed to find predatory behavior in P. humeralis and concluded that this species primarily feeds on dead ants. Similar observations were made with Pella japonicus, which lives with Lasius spathepus (Yasumatsu, 1937; Kistner, 1972). Kistner also observed that these beetles “ate small insects that are being transported by the ants.” However, he could not see the Pella eating live ants or fighting any of the ants on the trail. Hölldobler et al. (1981) confirmed that adult Pella funesta, Pella laticollis and P. humeralis live as scavengers, feeding on dead or disabled ants and debris discarded by the ants. However, they also observed that these beetles, in concert with Wasmann's claim, act as very effective predators on the ants. In the field and laboratory Pella beetles were seen hunting Lasius fuliginosus. The beetles chased after individual ants and pursued them for distances of up to six centimeters. When a beetle made a rear approach, it attempted to mount the ant and insert its head between the victim's head and thorax (Figure 13-7). When attacked the ant usually reacted by stopping abruptly and pressing the femurs rapidly and tightly against its body. Often this reaction threw the beetle off the back of the ant, allowing the ant to escape. Of 178 beetle onslaughts on Lasius fuliginosus workers observed during a period of 3 hours, only nine were successful. The beetles partially decapitated their victims, at least to the extent of severing their esophagus and nerve cord. Occasionally two or three beetles were seen chasing an ant. Once the ant was caught by a beetle the other beetles joined in subduing and killing it. Although individual Pella often tried to drag the prey away from the others, all the members of the pack usually ended up feeding on it simultaneously. No aggression among the beetles was observed on such occasions. When the beetles were starved for several days, they occasionally chased each other, even jumping on one another's backs, but they never engaged in cannibalism.

When foraging on the ants' garbage dumps or running along the ants' trails, Pella adults frequently encounter ants and are attacked by them. How do the beetles manage to escape their aggressive hosts? Usually they run around with their abdomens curved slightly upwards. When encountering an ant, the beetles flex the abdomen even more strongly. This is a typical and frequently described behavior of many staphylinid myrmecophiles and is commonly considered a defensive response. It has been suggested that during abdominal flexing the beetles discharge secretions from their tergal gland (Jordan, 1913; Kistner and Blum, 1971). The gland is located between the sixth and seventh abdominal tergites and is unique to the subfamily Aleocharinae, to which most myrmecophilous and termitophilous staphylinids belong (Jordan, 1913; Pasteels, 1968). The chemical composition of the tergal gland secretions of several species has been investigated and found to be extraordinarily diverse (Blum et al., 1971; Brand et al., 1973a; Kolbe and Proske, 1973; Hölldobler et al., 1981). Kistner and Blum (1971) suggested that Pella japonicus and possibly also Pella comes, both of which live with Lasius spathepus, produce citronellal in their tergal glands. This substance is also a major compound of the mandibular gland secretions produced by the host ants, for which it functions as an alarm pheromone. Pella japonicus tergal gland contents have not yet been chemically identified. Because irritated beetles seemed to smell like the ants' mandibular gland secretions, Kistner and Blum speculated that Pella japonicus produce citronellal in the tergal glands and thereby mimic the alarm pheromone of their host ants. They suggested that in this way the beetles can “cause the ants to reverse their direction, a reaction which allows the myrmecophiles to escape.” However, the investigations by Hölldobler et al. (1981) of the defensive strategy employed by the European Pella towards their host ants Lasius fuliginosus lead to a different interpretation. When irritated mechanically, Pella laticollis discharges a pungent brownish secretion from its tergal gland. A chemical analysis of the secretions did not reveal any resemblances to the mandibular gland secretions of the host ants. Furthermore, the beetles never employed the tergal gland secretions when they were attacking ants. They did so only when severely attacked themselves--and in particular when the ants seized their appendages. Ants contaminated with tergal gland secretions in these episodes usually displayed an aversive reaction, releasing the grip on the beetles and grooming and wiping their mouthparts and antennae on the substrate. But the beetles then had to escape quickly, because other ants nearby became alerted and rapidly converged on the scene, apparently in response to the alarm pheromones of their nestmates.

Although Hölldobler and his collaborators could not find any resemblance of the bulk of Pella tergal gland secretions to the mandibular gland secretions of Lasius fuliginosus, it is noteworthy that the Pella secretions did contain undecane, a hydrocarbon commonly found in the Dufour's glands of formicine ants and considered to be an alarm pheromone in Lasius fuliginosus (Dumpert, 1972). Even with this element, however, isolated tergal gland contents of P. laticollis elicited a repellent reaction rather than an alarm response in Lasius fuliginosus. Apparently the repellent effect of the quinones, the major components of the tergal gland secretions, is stronger than the attractant effect of undecane. When the ants' antennae were directly contaminated with the beetles' tergal gland secretions, they hung almost motionless, and the ants were disoriented for several minutes. Hence it is unlikely that the beetle uses its powerful repellent defensive system each time it makes a “routine” encounter with its host ants. Hölldobler and his co-workers concluded that in such situations the beetles rely on an appeasing or diverting defensive strategy, with the repellent defense being employed only as a last resort.

In early spring, when most of the Pella adults are close to the entrance of the ants' nest, they feign death when attacked by ants. The ants then either ignore the motionless beetles or carry them around and finally discard them on the refuse piles. Later in the season, when the activity of ants and beetles is much higher, the beetles employ a different appeasement technique. Each time they encounter ants they flex their abdomen forward and point the abdominal tip toward the head of their adversaries. Usually the ants respond by antennating the tip and licking it briefly (Figure 13-10). This ordinarily damps the ants' aggressions, allowing the beetle to escape. When on occasion the ants become persistent, the beetle extrudes a white, viscous droplet from the abdominal tip, which the ants eagerly lick up. Histological investigations revealed the abdominal tips of Pella to contain exocrine glandular structures, which Hölldobler has called the “appeasement gland” complex. It is not yet possible to assign one specific gland to the appeasement function.

Kitchen middens and peripheral nest chambers. We have described the genus Pella in some detail as representative of an early evolutionary grade in the myrmecophilous evolution of the aleocharine staphylinids. Species of Pella are specialized predators and scavengers, but they are less advanced in their myrmecophilic relationships than aleocharine species specialized on the kitchen middens, peripheral nest chambers, and brood chambers of the host ants' nests. A representative of the latter, more advanced evolutionary grade is Dinarda. The myrmecophilous habits of this genus have been known since Wasmann's first account published in 1894, and the genus has since been recorded with many species of Formica (Bernard, 1968). Kistner (1982) points out that the species-level taxonomy is still in poor condition, so that the best studied form, Dinarda dentata, might in fact comprise several species. The following account is based mostly on the form (or species) of Dinarda dentata that lives in nests of the red slavemaking ant Formica sanguinea of Europe (Figure 13-11).

The larvae of Dinarda are concentrated in the kitchen middens of the Formica hosts, where they feed on dead ants and debris. When they encounter worker ants they exhibit the same appeasingly defensive behavior described for Pella larvae (Figure 13-12). Not surprisingly, they also possess a similar complex glandular structure in the tip of their abdomens. Some adult Dinarda beetles are found in the kitchen middens, where they exist as scavengers. Most, however, patrol through the peripheral nest chambers, where they feed on arthropod prey brought in by the host workers. Wasmann (1894) reported that Dinarda also eats mites and occasionally even ant eggs and larvae.

Finally, studies of laboratory colonies have revealed that Dinarda taps into the liquid food flow of its hosts (Hölldobler, 1971a, 1973c). This parasitism occurs mostly in the peripheral chambers of the nest, where the bulk of regurgitation occurs between newly returning foragers and nest workers. The Dinarda beetles sometimes insinuate themselves between two workers exchanging food and literally snatch the droplet of food from the donor's mouth. They also use a simple begging behavior in order to obtain food directly from the returning foragers. The beetle approaches an ant and furtively touches its labium, causing the ant to regurgitate a small droplet of food (Figure 13-13). The ant, however, soon recognizes the beetle as an alien and commences to attack it. At the first sign of hostility the beetle raises its abdomen and offers the ant the appeasement secretions at the abdominal tip. The abdominal tip is quickly licked by the ant, and almost instantly the attack ceases. During this brief interlude the beetle makes its escape. Dinarda also possesses a well-developed tergal gland, but its repellent secretions are applied only as a last resort against ant attacks.

It is noteworthy that a myrmecophile belonging to a radically different taxonomic group obtains food in a similar manner to Dinarda. Janet (1896) discovered that the thysanuran Atelura formicaria, which is a very common symbiont of many European ant genera (Bernard, 1968), occasionally snatches food during food exchanges between worker ants. Like the Dinarda beetles, Atelura favors the peripheral chambers of the host nests.

Another example of a myrmecophile adapted especially to kitchen middens and peripheral nest chambers is the histerid beetle genus Hetaerius (Figure 13-14). As Wasmann (1886, 1905) first reported, the European species Hetaerius ferrugineus is most frequently found in nests of Formica fusca, where it feeds on dead and wounded ants. Occasionally it also consumes ant larvae and pupae. Most of the time the ants pay no attention to the Hetaerius. When they do attack an adult, the beetle exhibits death-feigning behavior by holding perfectly still with its legs closely appressed to its body. The ants frequently react to this nonviolent resistance by carrying the beetle around, licking it and finally releasing it. It has been suggested that Hetaerius has special trichome glands opening on the margins of the thorax (Escherich and Emery, 1897; Wasmann, 1903; Wheeler, 1908c). However, no histological investigations have been performed to confirm this assumption.

Wheeler (1908c) observed that adults of the North American Hetaerius brunneipennis solicit regurgitated food from the host ants. The beetle sometimes waves its forelegs toward passing ants, and by this action appears to attract their attention. A very similar behavior has been more recently recorded in H. ferrugineus. The soliciting beetle takes an upright position, stretching its forelegs widely apart and waving them slightly towards the approaching ant. The ant antennates and licks the beetle with her mandibles usually held relatively closed (Figure 13-15). Tracer experiments have revealed that the beetle also obtains small amounts of regurgitated food (Hölldobler, unpublished results). Thus, contrary to a suggestion originally made by Wheeler, Hetaerius brunneipennis does not represent a more advanced evolutionary state over H. ferrugineus (where food solicitation was previously unknown).

Brood chambers. The brood chambers constitute the optimal niche within an ant nest for a social food-flow parasite, because there the food of the highest quality is concentrated to be fed to the developing larvae, callow workers and queen(s). Moreover, the immature ant stages housed in the brood chambers provide the most valuable prey for specialized ant predators. The brood chambers are nevertheless difficult for parasites to penetrate, because they are fiercely defended by the ants. Very special adaptations are thus needed for myrmecophiles to exploit these sites as an ecological niche. This has been accomplished by some of the evolutionarily most advanced myrmecophilous staphylinids, of which the aleocharine beetles Lomechusa strumosa and several species of the genus Atemeles are the premier examples.

Lomechusa (Figure 13-16) lives with Formica sanguinea in Europe. Atemeles pubicollis (Figure 13-16), also a European species, is normally found with the mound-building wood ant Formica polyctena during the summer. But in the winter it inhabits the nests of ants of the myrmicine genus Myrmica. We know from the observations of Wasmann of more than 70 years ago that these beetles are both fed and reared by their host ants. The behavioral patterns of the larvae of these beetles are similar for Lomechusa and the various Atemeles species; in particular the larvae prey to a certain extent on their host ants' larvae. It is therefore astonishing that nurse ants not only tolerate these predators but also feed and protect them as readily as they do their own brood (Figure 13-17).

Hölldobler (1967a,b) demonstrated that the interspecific communication is both chemical and mechanical in nature. The beetle larvae show a characteristic begging behavior toward their host ants. As soon as they are touched by an ant, they rear upward and try to make contact with the ant's head. If they succeed, they tap the ant's labium with their own mouthparts. This apparently releases regurgitation of food by the ants. The ant larvae beg for food in much the same way but usually less intensively than the beetle parasites. With the aid of experiments using food mixed with radioactive sodium phosphate it has been possible to measure the social exchange of food in a colony. The results show that myrmecophilous beetle larvae present in the brood chamber obtain a proportionately greater share of the food than do the host ant larvae. The presence of ant larvae does not effect the food flow to the beetle larvae, whereas ant larvae always receive less food when they compete with beetle larvae. This disparity suggests that the releasing signals presented by the beetle larvae to the nurse ants may be more effective than those presented by the ant larvae themselves.

The Lomechusa and Atemeles larvae are also frequently and intensively groomed by the brood-keeping ants. Thus it appears probable that chemical signals are also involved in the interspecific relationship. The transfer of substances from the larvae to the ants was detected with the aid of radioactive tracers. These substances are probably secreted by glandular cells that occur beneath the integument of the dorsolateral area of each segment. The biological significance of the secretions was elucidated by the following experiments. Beetle larvae were first completely covered with shellac to prevent the release of the secretion. They were then placed outside the nest entrance, together with freshly killed but otherwise untreated control larvae. The ants quickly carried the control animals into the brood chamber. The shellac-covered larvae, on the other hand, were either ignored or carried to the garbage dump. It was found that for adoption to be successful, at least one segment of the larva must be free of shellac. Furthermore, after all the secretions have been extracted with solvents, the larvae are no longer attractive. When the extracted larvae are then contaminated with secretions from normal larvae, they regain their attractiveness. Even filter paper dummies soaked in the extract are carried into the brood chambers.

To summarize, the experiments show that the adoption of the Lomechusa and Atemeles larvae and their care within the ant colony depend on chemical signals. It may be that the beetle larvae imitate a pheromone which the ant larvae themselves use to release brood-tending behavior in the adult ants. This inference is supported by the fact that the beetle larvae and host ant larvae can be transplanted from the host species to closely related Formica species. But the ant species that do not accept the larvae of the ant hosts also reject the beetle larvae.

The question next arises as to how the ant colony manages to survive the intensive predation and food parasitism by the beetle larvae. The answer appears to be very simple. The beetle larvae are cannibalistic, and this factor alone appears to be effective in limiting the number of beetle larvae in the brood chambers at any time.

The larvae of both genera pupate in the summer, and they eclose as adult beetles at the beginning of autumn. The young Lomechusa adults leave the ant nest and after a short period of migration seek adoption in another nest of the same host ant species. Atemeles adults, on the other hand, emigrate from the Formica nest, where they have been raised, to nests of the ant genus Myrmica. They winter inside the Myrmica brood chambers and in the springtime return to a Formica nest to breed (Wasmann, 1910a; Hölldobler, 1970b). The fact that the adult beetle is tolerated and is fed in the nest of ants belonging to two different subfamilies suggests that it is able to communicate efficiently in two different “languages.”

The Atemeles face a major problem in finding their way from one host species to another. Formica polyctena nests normally occur in woodland, while those of Myrmica are found in grassland around the woods. Experiments have revealed that when Atemeles leave the Formica nest they show high locomotor and flight activity and orient toward light. This may well explain how they manage to reach the relatively open Myrmica habitat. Once they reach the grassland, the beetles must distinguish the Myrmica ants from among the other species present and locate their nests. Choice experiments in the laboratory have revealed that they identify the Myrmica nests innately by specific odors. Windborne species-specific odors are equally important in the springtime movement back to Formica nests.

Having found the hosts, the beetles must then secure their own adoption. The process comprises the four sequential steps depicted in Figure 13-18. First, the beetle taps the ant lightly with its antennae and raises the tip of the abdomen toward it. The latter structure contains the appeasement glands, the secretions of which are immediately licked up by the ant and appear to suppress aggressive behavior. The ant is then attracted by a second series of glands along the lateral margins of the abdomen. The beetle lowers its abdomen in order to permit the ant to approach this part of its body. The glandular openings are surrounded by bristles, which are grasped by the ant and used to carry the beetle into the brood chambers. By experimentally occluding the openings of the glands, Hölldobler found that their secretion is essential for successful adoption. For this reason he called them “adoption glands.” Thus, the adoption of the adult beetle, like that of the larva, depends on chemical communication, and it most probably entails an imitation of specific brood pheromones (Hölldobler, 1970b).

Before leaving the Formica nest the Atemeles beetle must obtain enough food to enable it to survive the trek to the Myrmica nest. This it obtains by vigorous solicitation from its hosts. The begging behavior is essentially the same toward both Formica and Myrmica (Figure 13-19; see also Figure 7-41). The beetle attracts the ant's attention by rapidly drumming on it with its antennae. Using its maxillae and forelegs it taps the mouthparts of the ant, thus inducing regurgitation. As noted previously, the ants themselves employ a similar mechanical stimulation of the mouthparts to obtain food from one another. It is thus clear that Atemeles is able to obtain food by imitating these very basic tactile food-begging signals.

What is the significance of the seasonal changes in hosts in Atemeles? There are good reasons for believing that Atemeles first evolved myrmecophilic relationships with Formica rather than with Myrmica. It seems likely that the ancestral Atemeles beetles hatched in Formica nests in the autumn and then dispersed, returning to other Formica nests only to overwinter. This simpler life cycle is followed by Lomechusa today (Wasmann, 1915b; Hölldobler, 1972). However, in the Formica nest the immature stages disappear during the winter, and consequently social food flow is reduced. In contrast, the Myrmica colony maintains brood throughout the winter. Thus, in Myrmica nests, larvae and nutrients from the social food flow are both available as high-grade food sources to the myrmecophiles. These circumstances, coupled with the fact that the beetles are sexually immature when they hatch, suggest why it is advantageous for the beetle to overwinter in Myrmica nests. While there the Atemeles complete gametogenesis, so that when spring comes the beetles are sexually mature. They then return to the Formica nest to mate and lay their eggs. At this time the Formica are just beginning to raise their own larvae and the social food flow is again maximized. The life cycle and behavior of Atemeles is thus synchronized with that of its host ants in such a manner as to take greatest advantage of the social life of each of the two species in turn.

The North American staphylinid myrmecophiles of Xenodusa seem to have a similar life history as Atemeles. The larvae are found in Formica nests and the adults overwinter in the nests of the carpenter ants of the genus Camponotus (Wheeler, 1911). It is undoubtedly significant that Camponotus, like Myrmica, maintains larvae throughout the winter. It may well be that the host-changing behavior of Xenodusa has the same significance as that inferred for Atemeles.

Wasmann considered Lomechusa strumosa to be the most evolutionarily advanced of the myrmecophilous staphylinids. We agree with this assessment, which is supported by a more recent analysis of the myrmecophilous behavior of Lomechusa (Hölldobler, 1971a, and unpublished results). Lomechusa also appears to have at least one migratory phase in spring, when, shortly after overwintering as an adult inside the Formica sanguinea nest, it travels to another nest of the same host species, where mating takes place (Figure 13-20). The adoption procedure into the new host ants' nest is even more complicated than that of Atemeles.

Like other aleocharine myrmecophiles, Lomechusa is equipped with well-developed tergal glands that produce defensive substances. The secretion consists of benzoquinone, methyl-benzoquinone, ethyl-benzoquinone and n-tridecane, with the latter compound accounting for more than 80 percent of the volatiles (Blum et al., 1971). This powerful repellent mixture, however, is normally employed only when the beetle is attacked by non-host ants. When approached by host ants Lomechusa behaves much more gently. It first presents the trichome structures on its legs, particularly the femora. The beetle then slowly circles around on the spot, with its legs widely extended so that the femoral trichomes are easily accessible. Simultaneously the beetle antennates the ants, bending its body sideways or backwards in order to reach the ants with its antennae. Next it slowly points its abdominal tip at the ants. Lomechusa possesses a battery of exocrine glands at the abdominal tip, and during the early adoption phase it frequently extrudes a whitish, viscous droplet which is eagerly licked up by the ants. The secretion is proteinaceous and contains no appreciable amounts of carbohydrates. The ants that feed on the material seem to grow calmer in the process. The Lomechusa now permit the ants access to trichomes along the lateral margins of the abdomen that project above the adoption glands. The ants lick and grasp the trichome bristles and finally carry or drag the beetle into the brood chambers of their nests. Inside the nest the Lomechusa continues to be tended by its host ants, despite the fact that it preys on the ants' brood (Figure 13-21). In addition the adult Lomechusa, like the larva, receives food directly from the nurse ants. Its soliciting behavior is markedly different from that of adult Atemeles. Lomechusa possess trichomes at the labrum, which are frequently licked by the ants. During this procedure the beetle uses its maxillae to stimulate the mouthparts of the ants, an action that seems to elicit the regurgitation of liquid food. In contrast to Atemeles, however, Lomechusa does not use its forelegs when begging for food. One gets the impression, as Wasmann first suggested, that the ants' feeding behavior toward adult Lomechusa resembles more the nursing behavior toward ant larvae than the trophallactic food exchange between adult ants. It is possible, but not yet proved, that Lomechusa is fed primarily with high-grade secretions from the postpharyngeal and labial glands of nurse workers--in other words, the same food given the larvae.

In short, Lomechusa strumosa presents its host ants with a complex variety of chemical signals, including imitations of brood-tending pheromones and other kinds of pseudopheromones. The species is thereby superbly well adapted to live with Formica sanguinea, especially in the brood chambers where it finds both food and a protected and regulated environment.

Most of the more casual observations of regurgitation between host and symbionts suggest that the exchange is exploitative, meaning that the liquid flows exclusively from the host to the symbiont. Experiments with radioactive tracers have proved this to be the case for the larvae and adults of Atemeles and Lomechusa (Hölldobler, 1967a,b, 1970b, and unpublished data). However, an exception has been reported in the case of Amorphocephalus coronatus. Adults of this brenthid beetle live with Camponotus in southern Europe (Figure 13-22). According to Le Masne and Torossian (1965) they receive food from some of the host workers and regurgitate part of it back to other host workers. This is the first reported example that could be construed as altruistic behavior on part of the symbiont beetles (Wilson, 1971). However, quantitative investigations are needed to see how much of the received food the beetle shares with other host ants. It could very well be that Amorphocephalus returns only very small portions of the food to the colony and that this “pseudo altruistic” behavior is one other mechanism of a symbiont to become fully integrated into the host colony.

Other adaptations of predators in the ants' brood chambers. Entomologists generally agree that a select few of the aleocharine staphylinids, including Atemeles and Lomechusa, have reached the evolutionary pinnacle of myrmecophilous adaptations. In mimicking the ants' important communication signals they are able to integrate themselves into the ant societies and exploit the hosts' brood chambers. There nevertheless exists a diversity of other arthropods that also preys on ant brood but uses a very different set of sophisticated techniques to invade the ants' nests. An especially instructive example is provided by the scarabaeid beetle genus Cremastocheilus (Figure 13-23).

Cremastocheilus is a North American genus comprising about 50 species (Krikken, 1976), various of which have been studied by Wheeler (1908b), Cazier and Mortenson (1965), Alpert and Ritcher (1975), and Kloft et al. (1979). Most recently Alpert (1981) has provided an insightful review of the ecology, behavior and evolution of the genus.

Various species of Cremastocheilus, the greatest number of which are found in the deserts of the southwestern United States and Mexico, use the formicine ants Formica, Myrmecocystus, Lasius and Camponotus, and the myrmicines Pogonomyrmex, Veromessor and Aphaenogaster as hosts. Ant nests in this arid area provide a high concentration of food resources and a refuge from predators and severe climatic change. Cremastocheilus larvae, like other scarabaeid larvae, feed primarily on decaying vegetable material. They develop within piles of vegetable matter inside ant nests, despite the apparent lack of any morphological adaptations for defense from their hosts. Alpert's field observations revealed few interactions between beetle larvae and ants. Experiments showed that the larvae are generally ignored by ant workers, while unrelated scarabaeid larvae of similar size used as controls were attacked immediately. Only highly excited ants were observed to attack Cremastocheilus larvae, which then defended themselves with powerful “mandibular strikes” and escape maneuvers. Alpert concluded that while Cremastocheilus larvae are not nutritionally dependent on ants (they can be raised in complete isolation), they gain protection from predators and desiccation simply by virtue of their residency in ant nests.

Cazier and Statham (1962) suggested that the adult beetles are first brought into the ant nests as booty by the foraging ants. The beetles, however, are so well protected by their heavily sclerotized cuticle that the ants are unable to kill them. Once inside the nest, the beetles themselves prey on the ant larvae (Cazier and Mortenson, 1965). This scenario has been basically confirmed by Alpert's (1981) field and laboratory observations. Although Cremastocheilus adults are capable of feeding on any species of ant larvae, only one or a few ant species are selected as hosts under natural conditions. This specificity appears to be a product of the host-searching process.

It was originally believed that secretions from glandular trichomes, located on prominent pronotal angles in all species of Cremastocheilus, attracted ants to beetles outside the nest and induced adoption (Cazier and Mortenson, 1965); but experiments conducted by Alpert did not confirm this assumption. Alpert released dead Cremastocheilus and other scarabaeid beetles with live C. stathamae near Myrmecocystus colonies. There was no significant preference by the ants for live C. stathamae, suggesting that special exocrine gland secretions are not involved in the adoption of the beetles by their ant hosts. Nor did glandular extracts elicit special attraction on the part of the ants. All the evidence supports the hypothesis that certain Cremastocheilus species are mistaken for prey items and thereby gain access to the nest. In fact, Cremastocheilus beetles commonly feign death when approached by ants, retracting their antennae into protective grooves and sticking out their heavily sclerotized legs. While some species gain entry into the host ants' nest by mimicking prey items, other species are able to march directly through the nest entrance or burrow through thatch piles into the interior. Often the beetles are attacked by ants inside the nest. Occasionally they are forcibly ejected, only to be carried in again by other ants. Attacks usually persist for several minutes, until the ants give up and the Cremastocheilus move to concealed corners in chambers or passageways. In the end the beetles move slowly toward the brood chambers, where they are completely ignored by the ants. Alpert reports “when the adult beetles (C. stathamae) approached ant brood, or while beetles were feeding, the host ants clearly licked the surface of the beetle. Although pronotal angles were specifically licked, most of the dorsal surface of the beetles was also involved.” Similar observations were made on other Cremastocheilus species. Alpert concludes that while adults of Cremastocheilus feed on ant brood, trichome secretions may distract workers, reduce aggression, and prevent workers from evacuating brood, while other glands serve to absorb colony odors.

Kloft et al. (1979) proposed that secretions from trichomes at the prosternal apophysis are sought by Camponotus castaneus and evoke food sharing behavior in the ants. They claimed that by transferring Cremastocheilus trichome substances to other workers, the entire colony could be appeased. In addition they suggested that the enlarged and projecting propygidial spiracles are glandular and “a possible source of secretions” attractive to the ants. By carefully sectioning Camponotus castaneus and members representing most of the other species groups, Alpert (1981) has shown that this conclusion is incorrect. The prosternal apophysis and propygidial spiracles are not glandular in the genus Cremastocheilus. Alpert nevertheless did find many new areas that bear exocrine glands. In a series of experiments with radioactive tracers, he further discounted the claim by Kloft et al. (1979) that trophallaxis occurs between host ants and Cremastocheilus. Furthermore, he was unable to find evidence of a “peace making” allomone transferred from trichomes to the colony.

Entomologists continue to be intrigued by the possible evolutionary pathways that led to the specialized predaceous behavior of Cremastocheilus. One hypothesis is that Cremastocheilus has changed from what was almost certainly a herbivorous diet to become a fully carnivorous scarabaeid (Wilson, 1971; Kistner, 1982). This route might have started with the adaptation of larvae for development in ant nests. In typical scarabaeid fashion Cremastocheilus larvae eat vegetable matter in the soil, and because of the concentration of vegetation in ant nests and protection from predators, ant colonies became the new oviposition site. Such behavior is known in cetonine scarabaeids, such as Euphoria inda, which develops in Formica obscuripes nests in North America (Windsor, 1964), and the European species Cetonia curprea, the larvae of which also develop in Formica nests (Donisthorpe, 1927). Alpert (1981), on the other hand, argues that the adults of Ephoria and Cetonia are not preadapted for survival in ant colonies and did not evolve in this direction. Cremastocheilus adults, in contrast, have many morphological adaptations for integration into ant nests and an examination of other genera of Cremastocheilini gives the best understanding of how the genus Cremastocheilus might have evolved. In Alpert's view, “The tribe Cremastocheilini reflects an evolutionary route from adult predation on soft bodied insects to specialized feeding on ant brood and subsequent development of larvae in ant colonies.” The Indian cremastocheiline Spilophorus maculatus, in fact, is a predator on membracids and lives independently of ant colonies (Ghorpade, 1975). Genuchinus ineptus develops in plants of the liliaeceous genus Dasylirion of Mexico and the southwestern United States, and any association with ants is entirely accidental. The adults are general predators, feeding on many different soft bodied insects in the laboratory. Dasylirion plants are inundated with fly larvae, the most probable source of food for Genuchinus ineptus in the field. Genuchinus ineptus, in this interpretation, is preadapted for myrmecophilous existence as a predator on ant brood. The predaceous mouthparts of Genuchinus ineptus adults are almost identical to those of Cremastocheilus. There are other morphological features common to Genuchinus and Cremastocheilus which are apparently adaptations for predation. These include the flattened body shape, hard, pockmarked integument, retractable antennae, and protective mentum. The larvae of Genuchinus ineptus develops in leaf compost of Dasylirion plants, and the pupae are sheltered by protective cases. The Genuchinus differ from Cremastocheilus by remaining inside their cases until spring.

According to Alpert's reconstruction, the major evolutionary step taken by the genus Cremastocheilus was to specialize on ant brood. The development of pronotal angles and trichome hairs is a consequence of this specialization. Other genera of New World Cremastocheilini, such as Centrocheilus (Krikken, 1976), appear to occupy an intermediate level leading to this stage. Other species of Genuchinus from South America occur commonly in banana leaves and bromeliads. Ant nests are also frequently found at these sites, and beetles may have moved from the leaf litter into the ant colonies to promote their larval development. A study of the life history of Paracyclidius bennetti, which is found with ants on bromeliads in Trinidad, might provide insight into this evolutionary stage. The genus Cremastocheilus adaptively radiated into ant colonies inhabiting the southwestern deserts of North America. Today members from all five subgenera are found in Arizona, New Mexico, and Colorado. All species groups have at least one member from the Southwest, except for the castaneus group which is strictly eastern in distribution.

A major question remaining is whether after speciation and adaptive radiation species of Cremastocheilus are now at different evolutionary stages along a pathway toward full integration. Alternatively, differences in behavior and morphology might simply reflect adaptive radiation to the behavioral ecology of different species of ants. Alpert argues and we agree that the latter hypothesis is more likely, since all species of Cremastocheilus have the same basic relationship with ants.

Most species of the beetle family Pselaphidae are carnivorous, with members of the subfamilies Pselaphinae and Clavigerinae being additionally myrmecophilous. According to Park (1964) all of the approximately 250 species of the clavigerines are especially adapted to live in ant nests. The best studied genera are Adranes (Park, 1932; Akre and Hill, 1973; Hill et al., 1976; Kistner, 1982) and Claviger (Donisthorpe, 1927; K. Hölldobler, 1948; Cammaerts, 1974, 1977). All observations to date indicate that these clavigerine beetles live among the brood of their host ants, either preying on it or exploiting it in some other manner. Kistner (1982) has provided a table listing all the behavioral acts thus far observed in Adranes and Claviger. From these data we learn that Claviger testaceus (Figure 13-24), a frequent myrmecophile of the formicine genus Lasius, preys on ant eggs, larvae and pupae while at the same time receiving food from larvae and adult ants by means of trophallaxis (Figure 13-25). The beetles occasionally eat dead ants or booty brought into the brood nest by their hosts. Finally, Claviger adults are also frequently groomed by the ants. The host workers lick secretions from the numerous trichomes and other exocrine glands, the anatomy of which has been carefully studied by Cammaerts (1974). The ants rarely treat Claviger aggressively. When transplanted experimentally to new nests the beetles are accepted by a diversity of ant species--even those genera with which they have never been found to occur naturally (Donisthorpe, 1927; K. Hölldobler, 1948). When attacks do occur, almost invariably during the initial phase of adoption, the Claviger usually show death-feigning behavior. The ants often respond by grasping the beetles on their slender thoraces and carrying them around for a short while before releasing them again, usually in the midst of the ant brood. Claviger rarely suffer any harm from this treatment.

Cammaerts (1977) observed that Lasius workers regurgitate liquid food to Claviger testaceus after they have licked them intensively. He noted a similarity of the movements with those occasionally observed when ants lick prey objects before they place them as food among the ant larvae. From this observation he proposed the interesting hypothesis that Claviger produce substances from the trichome glands that smell like insect corpses, which cause the ants to place the beetles as “pseudo-prey” among the larvae. It is noteworthy in this connection that Akre and Hill (1973) observed that ant larvae lick the trichomes of the clavigerine Adranes taylori, although in contrast to the species of Claviger the Adranes do not eat ant larvae. Instead they depend on food regurgitated to them by the larvae. In both cases, however, the mimicking of prey signals would be an efficient way to gain access to the brood chambers.

One other category of myrmecophilous adaptation by brood predators deserves special mention. The larvae of the syrphid fly genus Microdon, which develop in ant nests, have unusual shapes (Figure 13-26). They resemble slugs or limpets and were originally described as mollusks, later as coccids, before they were correctly identified as syrphid larvae (for recent reviews of the literature see Duffield, 1981, and Kistner, 1982). Also, the larvae of at least two other nest parasites are convergent in this slug or limpet-like form: the larvae of the predaceous Australian lycaenid butterfly Liphyra brassolis and the larvae of several species of the Australian parasitic moths in the family Cyclotornidae (Dodd, 1902; Hinton, 1951). The genus Microdon is cosmopolitan and contains about 100 species, most of which have been recorded from nests of such diverse ant genera as Camponotus, Formica, Lasius, Polyergus, Aphaenogaster, Crematogaster, Monomorium, Pheidole, Iridomyrmex, and Tapinoma. Microdon larvae either feed on detritus (Akre et al., 1973) or prey on the host ants’ larvae (Hocking, 1970; van Pelt and van Pelt, 1972; Duffield, 1981; Garnett et al., 1985). Duffield (1981) reports that the first instar larvae of M. fuscipennis, which develop in nests of Iridomyrmex pruinosus, have never been observed eating ant larvae, although they may obtain food from the ants by larval trophallaxis. Second and third instar larvae, on the other hand, consume small ant larvae but never pupae. Frequently the adult ants pull their larvae away from the Microdon. Successful Microdon larvae move up and over the ant larvae, piercing the larval skin, emptying the body contents, and discarding the empty shell. Duffield observed third instar larvae consuming 8-10 ant larvae each during various 30-minute periods of observation.

How are the Microdon larvae able to survive inside the ants' brood chambers? The fly larvae are usually bulky and have a heavily reticulated skin, which appears to confer mechanical protection from ant attacks. Recently Garnett et al. (1985) made an interesting discovery that suggests a more subtle form of defensive adaptation on the part of certain Microdon species. These investigators studied three Nearctic species, Microdon albicomatus and M. cothurnatus in nests of Formica species, and M. piperi in Camponotus nests. The larvae are obligate predators on the brood of their hosts. However, they appear to feed almost exclusively on cocoon occupants, that is, on larvae already encased by a cocoon, or prepupae, or pupae. Attacks by ant workers on the Microdon larvae were extremely rare, but obvious acts of aggression did occur when second and third instars of M. albicomatus and M. corthurnatus were introduced into the nest of an inappropriate host, Camponotus modoc. This result indicates the presence of a pseudopheromone that mimics the colony odor of the hosts. Garnett et al. propose that at least the first and second instar larvae mimic ant cocoons. They observed when an ant nest is exposed to sunlight Microdon larvae compress their bodies laterally into the rough shape of an ant cocoon. They are then quickly picked up by workers and transported into undisturbed parts of the nest along with real cocoons. Third instar larvae are evidently too large and bulky to exhibit curling behavior, and they are never transported. “Considering that first and second instars of the three species of Microdon studied are physically similar and yet workers appear species-specific in their aggression, we assume that the aggressive mimicry is chemical as well as physical. If the first instar of Microdon gains access to a defenseless pupa in cocoon, it will not only find nourishment and protection from attack, but an opportunity to acquire recognition chemicals characteristic of the host. Upon its exit from the cocoon the first instar has gained the necessary chemicals to be accepted and even transported as brood. The possible dilution of such chemicals by growth of the larvae is partially compensated by the second instar continuing to attack and enter cocoons.”

Recently another syrphid parasite has been discovered in nests of the Australian weaver ant Polyrhachis doddi (Hölldobler, unpublished). It is an undescribed species of the genus Trichopsomyia (Thompson, personal communication). Its larvae, which resemble slugs even more strikingly than those of Microdon (Figure 13-27), live inside the ant nest where they prey on the ant brood. The parasitic larvae are ignored by the ants, but the adult fly, when eclosing from the pupa, is attacked by the ant workers and quickly leaves from the nest.

Wasmannian mimicry
Among the battalions of parasitic staphylinids that march with army ants are many species that strikingly resemble their hosts (Figure 13-28). This formicoid habitus is found almost nowhere else among the Staphylinidae. It has originated many times over through modifications of the abdomen and thorax that create an ant-like “petiole.” In addition, there is a strong tendency to resemble the hosts in the overall slender body form, in color, and even in the sculpturation of the body surface (see Figure 13-3). Seevers (1965) showed that “petioles” have been created in no fewer than seven ways in various groups of the Staphylinidae, with several of the modifications having been chosen by two or more groups independently. Although the ant-like species are accompanied in nests by staphylinids of a more generalized body shape, their large numbers and the remarkable degree of evolutionary convergence they encompass leave little doubt that some kind of mimicry is involved. The question is, exactly who is being fooled? Wasmann (1889, 1903, 1925), in first documenting the phenomenon from museum specimens, was persuaded that the form of the body, including the false petiole, is tactile mimicry that deceives the host ants. He further believed that the coloration is visual mimicry that deceives birds and other predators which might otherwise be able to pick the beetles out of the columns of running ants. The most prominent vertebrate predators along the columns of army ants are formicariid ant birds that follow the raiding swarms of Eciton or Labidus. They feed on arthropods flushed from leaf litter by the ants but avoid the army ants. Kistner (1979) describes Ecitomorpha nevermanni as one of the outstanding examples of such Batesian mimicry among ecitophilous staphylinids, whose color varies to match the color variation of its host, Eciton burchellii. “In Costa Rica and Panama, where Eciton burchellii is usually reddish-brown in color, so is Ecitomorpha nevermanni. At Tikal, Guatemala, where Eciton burchellii is nearly black, so is Ecitomorpha nevermanni. In Ecuador, both were bicolored.” Since the army ants have extremely poor vision, it is most likely that this remarkable color adjustment is an evolutionary adaptation to vertebrate predation. Nevertheless, Kistner (1966a,b; 1979), reflecting on his own field observations of staphylinid guests of the African driver ants, agrees with Wasmann's interpretation that visual mimicry evolved as a deceiving mechanism to fool the host ants in order to be accepted into their colonies. Species of Dorylomimus, Dorylonannus, Dorylogaster, and Mimanomma run with the ants in their columns. When a host worker encounters one of the beetles, it touches it lightly with its antennae. Kistner states, “this action is identical with that of an ant when it meets another ant. Typically, the antenna rubs along the ant and lingers at the petiole, then both ants move on. I have interpreted this palpation as a signal which tells the blind ants that the passing insect is another ant. I have further interpreted this with regard to the Dorylomimini, that the constriction of the abdomen is such that the same signal is evolved when palpated by the ants and both ants and myrmecophiles go their separate ways. Thus, the morphological constriction permits the myrmecophile to function within the colony as though it were an ant.” Kistner and Zimmerman (1986) has made a similar observation on the staphylinid genus Pheigetoxenus, a guest of the myrmicine swarm raiders of the genus Pheidologeton. This special form of tactile deception was designated “Wasmannian mimicry” by Rettenmeyer (1970).

Does Wasmannian mimicry really exist? We know that chemical identification is paramount in ants generally. When the surface odor of a worker ant is disturbed only slightly, the ant is swiftly attacked by its own sisters even though its morphology remains unchanged. Furthermore, there exist many more myrmecophiles which are fully integrated into the hosts' social system but do not resemble the host workers' shape in the least. Hölldobler (1970b, 1971a) found that when he artificially altered the shape and color of Atemeles or Lomechusa, the relationship of the beetles to their hosts was not affected. Thus it appears that communicative behavior and chemical mimicry are the essential ingredients for acceptance of social parasites. Although the hypothesis of tactile mimicry must not be dismissed out of hand, an alternative hypothesis is equally promising (Wilson, 1971). It is noteworthy that all cases of morphological resemblances of staphylinids occur in myrmecophiles that live with legionary ants. These ants and their guests spend most of their time on the surface, either in raiding or emigration columns. It is possible that predators watching the ant columns for edible morsels might be fooled by both the color of the beetles and their shape as well. Thus, all of the mimicry apparent to taxonomists may be visual and directed at predators outside the host colony. In fact there are numerous cases of arthropods that are not socially associated with ants yet strikingly resemble them in morphology and locomotion. Both circumstantial and experimental evidence strongly suggest that this similarity protects the mimics from predators (Reiskind, 1977; Oliveira, 1985; Mciver, 1987). Only by employing carefully designed experiments will we be able to resolve this question.

Some evidence exists meanwhile to suggest that tactile mimicry can serve as an ancillary mechanism for social integration in the host colony. We have already mentioned the case of the Microdon larvae that resemble the shape of ant cocoons and are transported by the ants in the same manner as these brood objects. Another remarkable case is that of the phoretic mite Planodiscus which attaches itself to the tibia of its host, Eciton hamatum. Kistner (1979) used scanning electron micrographs to demonstrate that the cuticular sculpturing of the mite's body is nearly identical to that of the ant's leg. Furthermore, the arrangement and number of setae on the mite approximates the arrangement and number of setae on the leg. “Thus when the ant grooms its leg,” Kistner observed, “the tactile stimulation will be similar to that of the leg itself. Since the mite clings to the leg by grasping setae, small movements of the mite during grooming would translate into small movements of the ant's setae. These would approximate movements caused by grooming activities themselves.”

Another form of tactile mimicry has been described in the myrmecophilous cricket Myrmecophila acervorum (Figure 13-29) by K. Hölldobler (1947). Incidentally, this is the European symbiont with the longest history of investigation; the first description of its behavior was by Paolo Savis (1819), and it was examined in increasing detail by Wasmann (1901), Schimmer (1909), and K. Hölldobler (1947). In North America, Myrmecophila nebrascensis and M. manni were studied by Wheeler (1900) and most recently by Henderson and Akre (1986a-c). In many ways the findings on M. manni parallel those made on M. acervorum. Henderson and Akre (1986b), however, found that M. manni reproduces sexually, with males exhibiting dominance hierarchies during mating. In contrast, no males at all are known from Myrmecophila acervorum, and it has therefore been suggested that this species reproduces by parthenogenesis (Schimmer, 1909; K. Hölldobler, 1947).

Myrmecophila acervorum, the species with possible tactile mimicry, lives with many different host species. Karl Hölldobler discovered two morphs of different sizes. The larger “major” morph is found primarily in nests of ant species with larger workers, such as Formica, Camponotus, Myrmica; while the smaller “minor” morph occurs with species that have smaller workers, such as Tetramorium and Lasius. All developmental stages live in ant nests. The adults and nymphs are extraordinarily versatile in their feeding habits. They prey variously on ant eggs, lick adult ants and other myrmecophiles such as Hetaerius and Claviger, snatch food from food exchanging ants, and solicit the regurgitation of liquid food by ant workers.

When the crickets are newly introduced into an ant nest, or the colony to which they belong is disturbed, they are usually treated in a hostile manner by the worker ants. They are then able to escape death only through swift and nimble running. But the ant aggression usually subsides as soon as the crickets adjust their locomotory pattern to the movement patterns of the undisturbed host ants. Karl Hölldobler noted that when Myrmecophila were transferred from an ant species with a relatively slow movement pattern, such as Myrmica rubra, to a species that generally exhibits a more vivacious “temperament,” such as Formica fusca, the crickets alter their own locomotory behavior to resemble that of their new hosts. The transformation usually takes a few days, during which the crickets stay mostly in hiding and apparently acquire the host specific colony odor. Hölldobler proposed that in addition to this locomotory convergence, the Myrmecophila employ a tactile mimicry. Although the cricket does not look like an ant overall, portions of its body resemble parts of the ants' bodies. K. Hölldobler elaborated his tactile mimicry concept with a metaphor. Suppose, he said, that we lived in a completely dark room and oriented primarily by means of the tactile sense in our hands. Among hundreds of us dwells one creature that is very differently constructed but has appendages resembling human hands, and it also manages to mimic our body movements and to touch us with a humanoid caress. This creature would be perceived by us as a fellow human being until some crucial behavioral mistake unmasked it as an alien. Thus the ants do not tolerate the crickets because they perceive the presence of the intruders as pleasant or comfortable, as Wasmann suggested, or because the crickets are too swift and elusive, as Wheeler and Schimmer proposed. Instead, the host workers are fooled into classifying the crickets as fellow ants. As soon as they notice the deception they hunt the crickets, even after having tolerated them for weeks or months.

Kistner and Jacobson (1975) redefined Wasmannian mimicry to cover all mimicry of social releasers, including behavioral imitation of solicitation signals and the imitation of ant pheromones by myrmecophiles. Wasmann himself would probably not have accepted this usage, because it contradicts his theory--now abandoned--about amical selection and symphilic instincts (see K. Hölldobler, 1948). Nevertheless, we agree that the expanded concept of Wasmannian mimicry is reasonable and honors the name of a great myrmecologist, who was the most important pioneer in discovering the world of symbiosis in insect societies.

An especially well-studied species displaying Wasmannian mimicry, by the modernized definition, is Myrmecaphodius excavaticollis, a scarabaeid beetle that lives principally with Solenopsis fire ants and secondarily with dolichoderine ants of the genus Iridomyrmex. Vander Meer and Wojcik (1982) noted that when these little beetles first move into a host nest, they exhibit a passive defensive behavior that allows them to acquire cuticular hydrocarbons specific to their host species. These adsorbed substances then apparently enable the beetles to become integrated into the host colony. This conclusion is supported by an increasingly convincing body of evidence that cuticular hydrocarbons are part of the species and colony recognition system in social insects (see Chapter 5).

All developmental stages of Myrmecaphodius excavaticollis have been found within the nests of the host ants. The adults variously prey on ant larvae, feed on dead ants and booty, and induce workers to regurgitate liquid food to them. They conduct dispersal flights throughout the year. After alighting they seek access to a suitable host colony, which may not belong to the same species they originally came from (Wojcik et al., 1978).

Vander Meer and Wojcik (1981) report that the hydrocarbons, which constitute 65 to 75 percent of the cuticular lipids of Solenopsis invicta and Solenopsis richteri, differ conspicuously among the four host species of Myrmecaphodius excavaticollis. But the beetles are able to shed the hydrocarbons of one Solenopsis species and acquire the pattern of another. In addition to species-specific characteristics, the hydrocarbon mixtures probably also contain colony-specific patterns. This, in part, explains how Myrmecophodius is able to infest a variety of species and colonies. Vander Meer and Wojcik demonstrated this effect with the following experiment. Beetles from Solenopsis richteri colonies were isolated for two weeks, and then introduced into colonies of Solenopsis invicta. After five days, the beetles were removed and analyzed for cuticular hydrocarbons. It was found that Myrmecaphodius excavaticollis had acquired the cuticular hydrocarbons of its new hosts (Figure 13-30). Vander Meer and Wojcik observed that “The same phenomenon occurred when previously isolated beetles were introduced into [Solenopsis geminata]] and Solenopsis xyloni colonies. Although the switching of hydrocarbon patterns from one host to another weakens the likelihood that they are synthesized by the beetle, we also found that freshly killed isolated beetles had acquired Solenopsis invicta hydrocarbons within two days after exposure to the ant colony. These data eliminate biosynthesis as a possibility and support a passive mechanism of hydrocarbon acquisition. When initially introduced into a host colony, the Myrmecaphodius excavaticollis were immediately attacked. The response of the beetles was to play dead and wait for the attacks to cease, or they moved to an area less accessible to the ants. Within two hours after introduction into a host colony, the beetles' cuticle contained 15 percent of host hydrocarbons. The accumulation of hydrocarbon continued up to four days until the beetles' cuticle contained about 50 percent host hydrocarbons. Beetles surviving this long were generally no longer attacked.”

In a suggestively parallel manner, K. Hölldobler (1947) observed that about four days are required for the cricket Myrmecophila acervorum to become fully integrated into the host colony, at which time it is able to groom the host ant workers. Furthermore, Rettenmeyer (1961a) raised the interesting possibility that the guests of army ants pick up the colony odor in an active fashion while grooming their hosts. It is certainly true that grooming by symphilic staphylinids and histerids cannot serve entirely for the ingestion of edible cuticular material. Akre and Torgerson (1968) have described the elaborate grooming rituals of a staphylinid guest of Neivamyrmex as follows: “While Probeyeria and Ecitophya straddled their host across the longitudinal axis of the body, Diploeciton assume a position parallel to, but slightly to one side of and on top of the ant. To position itself, a beetle grasps with its mandibles the scape of an antenna of an ant close to its base. It then positions its body parallel to the body of the ant and uses the first and third legs on the lower side of the body to brace against the substrate. The three legs on the other side of the body then straddle the ant. The mesothoracic leg on the bottom curls under and around the thorax of the ant. This places the sternum of the beetle's thorax against the side of the thorax of the ant as though riding 'sidesaddle.' In this position the beetle rubs the ant with its legs. The mesothoracic lower leg rubs the bottom of the thorax and the upper legs rub on the dorsal area of the ant; the prothoracic leg usually rubs the head of the ant, the mesothoracic leg rubs on the thorax and gaster, while the metathoracic leg is used sparingly to rub the gaster of the ant. The rubbing strokes are rather slow and alternate between stroking the body of the ant and the staphylinid's own body. The front leg is rubbed on the head and thorax, both middle legs are rubbed on the elytra and the globular portion of the myrmecoid abdomen, while the metathoracic leg was rubbed only rarely on the abdomen” (Figure 13-31). The stroking movements of the staphylinid and histerids calm the ants and in some instances, according to Akre and Torgerson, seem to paralyze them partially. However, this is not another case of a fatal seduction of the kind worked by the bug Ptilocerus, for the ecitophiles do not attack their adult hosts. They feed only on the immature stages, which interestingly enough are not recipients of the grooming ritual.

We infer that this kind of integration into the ant societies represents an evolutionarily less advanced grade than that of Atemeles or Lomechusa. Although these symbionts may also acquire host colony-specific cuticular hydrocarbons, they can be easily transferred from one colony to the other so long as both belong to at least the same species group. This trait probably is based on the ability of both the larvae and the adults of Atemeles and Lomechusa to produce pseudopheromones that mimic the brood pheromones of their hosts. It has long been known that brood pheromones dominate or mask colony specific recognition cues in ants (Hölldobler, 1977; Carlin and Hölldobler, 1986), making it easy to transfer ant brood from one colony to another belonging to the same species. The mimicry of brood pheromones thus appears to be the most advanced mechanism of social-parasitic integration, one that opens up the brood chambers, the richest ecological niche within an ant colony.

Symbioses between ants and lycaenid butterflies
The larvae of many species of lycaenid butterflies are closely associated with ants (Hinton, 1951; Atstatt, 1981a; Henning, 1983a; Cottrell, 1984; Pierce, 1987). The majority appear to be mutualistic, although a number are parasitic. An example of a parasitic species is the Large Blue (Maculinea arion) of Europe. The larva feeds on wild thyme and is attended by ants until it reaches the fourth instar. Then it crawls down onto the ground and wanders about until it meets a worker of the ant Myrmica sabuleti (Thomas, 1980). When touched by the ant the larva deforms its body into a striking new shape: it retracts its head and swells its thoracic segments up while at the same time constricting its abdominal segments, giving the body a hunched, tapered look (Figure 13-32). Apparently the altered body form serves as a signal to the ant, which may or may not work in concert with substances that resemble larval pheromones of the ants.

Whatever the nature of the key lycaenid stimuli, the ant now picks the caterpillar up and transports it into the nest. Once ensconced in the brood chambers, the caterpillar turns carnivore, feeding heavily on the host ant larvae. When it reaches full maturity, it pupates and overwinters in the nest, finally to emerge as an imago the following June (Frohawk, 1915; Cottrell, 1984). A very similar myrmecophilous behavior occurs in species of the lycaenid genera Lepidochrysops and Aloeides, but only Lepidochrysops larvae appear to consume ant brood (Henning, 1983a,b).

A remarkable variation on the Maculinea pattern was recently discovered by Schroth and Maschwitz (1984). The freshly eclosed fourth instars of Maculinea teleius leave their host plants, Sanguisorba officinalis, and actively search on the ground for the trails of their host ants, Myrmica rubra. Once they encounter the chemical trails they follow them until the host ants' nests are reached. The Maculinea teleius larvae seem to enter the Myrmica nests on their own. Schroth and Maschwitz never observed the lycaenid caterpillars being carried by the ants. Only the fourth instar larvae respond to the ants' trail pheromone and then only during a brief “sensitive phase” after eclosion. Caterpillars that have already lived two days inside the Myrmica nest no longer follow the ants' chemical trails.

Other parasitic species occur among the Asiatic and Australian lycaenids. For example, the caterpillars of Liphyra brassolis prey on larvae of the green tree ant, Oecophylla smaragdina (Dodd, 1902). Its larvae, unlike other lycaenid parasites, do not appear to be welcome guests in the host ants' nests, but rather defend themselves by their remarkable morphological adaptation against ant aggression. As we mentioned earlier, the larvae look like mollusks, and they are protected by an extremely hard and thick cuticle (Johnson and Valentine, 1986). When the butterfly hatches, it is covered by a dehiscent cloak of white, grey and brown scales. As the Oecophylla attempt to seize the butterfly during its egress from the nest, they get only mandibles and antennae full of scales. According to Hinton (1951), a similar adaptation occurs in several other lycaenids, including species of Maculinea, as well as the South American pyralid Pachypodistes goeldi.

The majority of myrmecophilous lycaenid species apparently live in exclusively mutualistic interactions with ants, and it seems likely that the parasitic life pattern has evolved from ancestral mutualistic relationships. Many species of New World riodinid butterflies have similar ecological relationships with ants (Ross, 1966; Callaghan, 1977, 1979, 1981; Schremmer, 1978; Robbins and Aiello, 1982; Horvitz and Schemske, 1984), and for the purposes of our discussion, we have included the Riodininae as a subfamily of the Lycaenidae (Ehrlich, 1958; Kristensen, 1976; Vane Wright, 1978; but see Eliot, 1973). However, as more is learned about the morphology and ecology of the riodinids, it may be found that they are a distinct taxonomic group whose ant associated characteristics are convergently derived with those of the Lycaenidae.

Several morphological structures in lycaenid larvae are important in the ant-lycaenid associations (Figures 13-33 and 13-34). Scattered over the surface of the larvae and pupae are many epidermal glands, called lenticles or “pore cupolas,” the significance of which was first noted by Hinton (1951) and subsequently described in detail by Malicky (1969), Downey and Allyn (1979), Kitching (1983), Pierce (1983), Wright (1983), Franzl et al. (1984), and Kitching and Luke (1985). These perforated structures have been found in all lycaenid subfamilies investigated, including the Riodininae. They are derived from innervated glandular setae that are believed to exude either ant attractants or appeasement substances. Epidermal extracts of several species have been bioassayed and shown to contain substances that are highly attractive to ants (Henning, 1983b; Pierce, 1983).

The larvae of many species also possess the so-called Newcomer's gland (called the dorsal nectar organ or honey gland by some authors), which is located on the dorsum of the seventh abdominal segments. This organ occurs in most species of the Polyommatinae and Theclinae but is absent in other subfamilies, including the Lycaeninae, Miletinae, Liphyrinae and Poritiinae. The Newcomer's gland presumably developed from dorsal epidermal pores of a kind common in the lycaenids (Malicky, 1969; Kitching and Luke, 1985; Fielder, 1987). In some groups, possibly analogous honey-gland structures have evolved from epidermal pores, such as the “perforated chambers” of the Curetinae (DeVries et al., 1986), the “pseudo-Newcomer's gland” of the Miletinae (Kitching, 1987), and the nectar organs of the Riodininae (Ross, 1966; Callaghan, 1977; Schremmer, 1978; Cottrell, 1984).

From its first description by Newcomer (1912), the lycaenid gland was thought to secrete sweetish liquid which is imbibed by the ants. Maschwitz et al. (1975) provided the first analytical proof for this hypothesis. They found that the secretions of the Newcomer's gland of the European species Lisandra hispana consist of a 13.1-18.7 percent solution of fructose, sucrose, trehalose, and glucose, as well as minor quantities of protein and a single amino acid, methionine. The hemolymph of the lycaenid caterpillar has a total carbohydrate concentration of only about two percent. Thus the Newcomer's gland provides the ants with a highly enriched sugar solution.

The Newcomer's gland of most species of Curetinae, Polyommatinae, and Theclinae is flanked on either side by a pair of eversible tentacles, which are also called “lateral organs” or “tentacular organs.” Similar structures exist in species of the subfamily Riodininae. However, they are absent in the Lycaeninae. Their function is unclear. It has been variously supposed that they produce substances attractive to the ants, signal the presence of the Newcomer's gland, or serve in defense when the dorsal organ is depleted (Downey, 1962; Claasen and Dickson, 1977). Recently evidence has been adduced to indicate that tentacular gland secretions serve at least in part to evoke alarm behavior in the attending ants (Henning, 1983b; DeVries, 1984; DeVries et al., 1986; Fiedler and Maschwitz, 1988c).

Other intriguing anatomical structures await investigation. Setae have been discovered on the surface of lycaenid larvae and pupae that are variously shaped like clubs, mushrooms, trumpets, clover blossoms, and multibranched hydroids (Downey and Allyn, 1973, 1979; Wright, 1983; Kitching and Luke, 1985; Fiedler, 1987). However, no connection has yet been experimentally established between these structures and the lycaenid-ant associations. In addition, the pupae of a number of species in the Lycaenidae possess stridulatory organs that can cause vibrations of the substrate on which the pupae are attached (Downey, 1966). In some instances, an audible sound is produced. Late instar larvae of certain species also possess stridulatory organs (Pierce et al., 1987; DeVries, personal communication). It seems likely that these stridulations play some role in interspecific communication with ants, and, for those species of Lycaenidae whose larvae aggregate, in intraspecific communication between aggregating larvae.

Aside from the case of Maculinea and a few other well-studied genera, the exact nature of the symbiosis between lycaenids and ants remains problematic. Malicky (1969, 1970a,b) for example, believed that there is no mutualistic symbiosis at all between ant and lycaenids, and that the so-called myrmecophilous organs such as the pore cupolas and Newcomer's glands serve solely as defensive devices against ant predation. Malicky argued that one of the most important features determining whether a lycaenid associates with ants or not is the “biotope” in which the larva lives. Both Malicky and Atstatt (1981b) pointed out that the structural form of plants may influence the distribution of ants, and consequently, the distribution of lycaenids. Hence Atstatt suggested that woody perennials, or “apparent” plants are more likely to be frequented by ants, and are therefore more likely to serve as host plants for myrmecophilous lycaenids. Atstatt (1981a) suggested that the capability of lycaenids to appease ants and thereby avoid predation opened up habitats which are dominated by ants. He considered this “selection for enemy free space” one of the major evolutionary forces that has shaped the lycaenid-ant associations. This pathway may well have been followed by the Lycaeninae, Miletinae and some species of the Riodininae, whose larvae are equipped with the pore cupola organs but lack the Newcomer's gland or other nectar organs. However, this is a relatively minor group within the Lycaenidae, comprising relatively few species in comparison with the highly diverse Polyommatinae and Theclinae (Eliot, 1973). Members of the latter two subfamilies possess both Newcomer's gland and cupola organs. Most enjoy a very close symbiotic relationship with ants, which appears to have contributed in turn to their remarkable species multiplication (Pierce, 1984, 1985, 1987). Increasingly strong evidence now exists that many of the associations are mutualistic. Evidence to that end was first provided by Ross (1966), who pointed out that larvae of the riodinine Anatole, which are normally associated with Camponotus colonies, are preyed upon by other ant species if left unattended. A similar behavior was recently reported in the larvae of the Australian lycaenid Ogyris genoveva, which are attended by Camponotus consobrinus and preyed upon by Iridomyrmex purpureus (Samson and O'Brien, 1981). It remained for Naomi Pierce and her collaborators, however, to demonstrate conclusively the ants' protective effect on the survival of lycaenid larvae (Figure 13-35). When larvae of Glaucopsyche lygdamus were experimentally isolated from their attending ants in the field, they were far more likely to disappear from the host plants. More exactly, they were more prone to drop off the plants or succumb to parasitoids. Ant-tended larvae during the study was four to twelve times more likely to survive to pupation than an otherwise similar group of untended larvae (Pierce and Mead, 1981; Pierce and Easteal, 1986).

The females of some myrmecophilous lycaenids actively search for oviposition plants already frequented by host ants. This curious behavior was first indicated in the field studies by Atstatt (1981b) and laboratory experiments by Henning (1983b) and has been demonstrated conclusively in a series of field experiments by Pierce and Elgar (1985) on the Australian species Jalmenus evagoras. This lycaenid is known to feed on at least 16 species of Acacia while being tended by several species of the dolichoderine ant Iridomyrmex. Jalmenus females are far more likely to lay egg clusters on plants that contain their attendant ants than on plants kept free of ants. Ovipositing females respond to the presence or absence of ants before they alight on the potential food plant. However, once they have landed, they are equally likely to lay eggs whether or not they encounter ants. Pierce and Elgar describe the symbiosis as follows:

If newly colonized plants are near nests of the attendant ant Iridomyrmex sp. 25, as the plants in our experiments were, then newly hatched clusters of pioneer larvae may be able to attract a sufficient number of tending ants quickly enough to survive and thus establish a new food plant. This process may be facilitated in several ways. First, A. irrorata, like many other acacias, has extrafloral nectaries (three near the tip and one at the base of each leaf), and plants are patrolled regularly by nectar seeking ants, including workers of Iridomyrmex sp. 25. Larvae of Jalmenus evagoras that hatch out on these plants are thus likely to be discovered quickly by ants. Second, eggs of Jalmenus evagoras are laid in clusters that hatch synchronously, thereby creating an aggregation that is a potentially more attractive food source than only a single larva. Third, like many other social insects, workers of Iridomyrmex sp. 25 recruit nestmates to food resources in numbers commensurative with the quality of those resources (see Wilson, 1971). A single worker discovering a cache of Jalmenus evagoras larvae may be able to recruit enough nestmates sufficiently quickly to tend the larvae and successfully protect them against predators.

Nevertheless commencing life on a new food plant without the benefit of pre-existing conspecific juveniles that have attracted attendant ants involves considerable risk for young larvae of Jalmenus evagoras. Ant exclusion experiments performed in 1981 (Pierce, 1983) revealed that the first and second instar larvae were preyed upon far more rapidly when they occurred on plants by themselves than when they occurred on plants that contained final instar larvae and pupae that had attracted large numbers of ants. One way Jalmenus evagoras may circumvent this problem is by ovipositing adjacent to ant tended homopterans. Atstatt (1981b) suggested a similar function for ant tended homopterans on food plants utilized by Ogyris amaryllis, although McCubbin (1971) and Das (1959) described situations in which myrmecophilous homopterans appeared to exclude lycaenids from potential host plants.

In addition to acquiring an immediate attentive ant guard, larvae of Jalmenus evagoras that hatch out besides myrmecophilous membracids gain a further advantage in the form of food. On numerous occasions we observed larvae of Jalmenus evagoras feeding on the honeydew secretions of homopterans. The first and second instars sometimes even ride on the backs of adult membracids. This phenomenon of honeydew feeding has been described for several lycaenids (e.g., Hinton, 1951). Although other lycaenids actually prey on homopterans (e.g., Cottrell, 1984) we found no evidence of this in Jalmenus evagoras.

Investigators over the years have described or suggested the existence of ant-dependent oviposition behavior in no fewer than 47 species of Lycaenidae. As a rule, the more dependent a lycaenid species is on its attendant ants, the more likely it is to possess ant-dependent oviposition behavior. It would seem to follow that “the propensity of female lycaenids to oviposit in response to myrmecophilous homopterans, when they occur on novel plants that are not the same species as the butterfly's original food plant, could have important implications for the host plant range of species that use ants as well as plants as cues in oviposition” (Pierce and Elgar, 1985). Indeed, a comparison of 282 species of Lycaenidae reported in the literature confirms that species tended by ants feed on a wider range of plants than those not tended by ants.

A recent observation by Ulrich Maschwitz (personal communication) is consistent with this generalization. During studies in Malaysia, he saw females of the polyommatine lycaenid Anthene emolus depositing large clusters of eggs (50-100 eggs) directly upon or close to the silk pavilions constructed by the weaver ants Oecophylla smaragdina. Within a few hours the larvae hatched and moved into the pavilions, where they fed on the surface tissue of the leaves. The second and third instar larvae left the pavilions and were subsequently carried by the ants to “grazing sites,” usually young shoots or blossoms of the host plants that are rich in carbohydrates and amino acids. The caterpillars were continuously attended by Oecophylla workers, which chased off approaching parasitic flies and wasps and frequently milked the secretions from the Newcomer's gland of the caterpillars. Anthene emolus is a common lycaenid in Malaysia which feeds on a wide variety of plants, including Nephelium litchi (Sapindaceae) and Cassia fistula (Caesalpiniaceae). They are, however, always associated with Oecophylla smaragdina.

Pierce discovered another important correlation between ant attendance and diet. A comparative survey from the literature showed that ant tended lycaenids usually feed on plants that fix nitrogen and are rich in proteins, while untended lycaenids feed on other kinds of plants (Pierce, 1984, 1985). Pierce and her collaborators demonstrated that the larvae of Jalmenus evagoras and two of its congeners J. ictinus and J. pseudictinus secrete concentrated amino acids in addition to carbohydrates as rewards for attendant ants, and that these amino acids are produced in species-specific profiles (Pierce et al., 1989). The larvae and pupae of Jalmenus evagoras in particular produce at least 14 different free amino acids among which serine is dominant. Secretions from the dorsal organ contain serine in concentrations ranging from 19-50 mM, which is at least an order of magnitude greater than that of amino acids found in most extrafloral nectaries and is comparable to concentrations found in the salivary glands of many social insects (Maschwitz, 1966; Hunt et al., 1982). Even when larvae are raised on four different host plants, the amino acid profile from the dorsal organ remains the same, with serine the primary component. Radioactive labeling experiments and bioassays with synthetic mixtures confirmed that the amino acids secreted by both the larvae and the pupae are an extremely attractive food source for workers of the attendant ant species (Pierce et al., 1989). Nitrogen is a limiting component in the diet of most insects (Mattson, 1980). The need to secrete surplus protein to attract attendant ants could explain why lycaenids associated with ants feed predominantly on protein-rich plants such as legumes, and even on protein-rich parts of plants, including flowers, terminal foliage, and seed pods. Thus a major constraint to reliance upon ants as a mobile defense force may well be the nutritional quality of the host plant.

Recently Fiedler and Maschwitz (1988a) witnessed host ants (Tetramorium caespitum, Plagiolepis pygmaea) recruiting nestmates to third and fourth instar larvae of the lycaenid Polyommatus coridon. This recruitment effect could be significantly weakened by closing off the Newcomer's gland with glue. Fiedler and Maschwitz measured the quantities of secretions produced by the Newcomer's gland and provided energetic calculations which strongly indicate that the symbiotic relationship between Polyommatus larvae and their attending ants is mutualistic. In the case of Tetramorium caespitum, their reasoning was as follows. They noted that an average sized colony of Tetramorium caespitum contains 11,000 workers (Brian et al., 1965; Brian, 1979b), whose monthly basal metabolic rate at 15°C is approximately 47.8 kJ (Peakin and Jasens, 1978). The monthly metabolic rate of one worker is therefore 4 J. The foraging area of a Tetramorium caespitum colony covers approximately 40 square meters. In a dense growth of the host plants, Hippocrepis comosa, Fiedler and Maschwitz found about 20 Polyommatus larvae per square meter. This corresponds to a monthly production of approximately 70-140 mg sugar per square meter, or a chemical energy of 1.1-2.2 kJ. Assuming that about 25 percent of the colony's foraging area is populated with Polyommatus larvae, in accordance with field observations, the energy amount obtained from lycaenid larvae is estimated to be 11-22 kJ, which is approximately a quarter to a half of the total monthly basal metabolic rate of the colony. Thus 100 P. coridon larvae could deliver enough sugars to cover the monthly energy needs (at 15°C) of 1,400-2,800 workers of Tetramorium. Similar calculations were obtained for the ants Lasius alienus and Plagiolepis pygmaea.

Pierce et al. (1987) examined other costs and benefits of ant associations for the lycaenid Jalmenus evagoras. They demonstrated that the degree of mortality caused by predation and parasitism is so severe as to force an obligate dependency on the part of the lycaenids: larvae and pupae deprived of attendant ants cannot survive. In addition to providing protection, attendant ants shorten larval development substantially while extending the duration of the pupal phase only slightly, thereby reducing the time that larvae are exposed to the threat of predators and parasitoids. On the other hand, energetic costs of associating with ants result in a reduction of adult weight and size, traits important to mating success in males and fecundity in females of Jalmenus evagoras (Elgar and Pierce, 1987).

Pierce et al. (1987) have provided two lines of evidence suggesting that the host ants receive substantial rewards for their efforts. “First, pupae that are tended by ants for only 5 days lose 25% more weight than their untended counterparts, and develop significantly more slowly. This indicates that pupae are supplying rewards for ants by diverting metabolic resources from metamorphosis. Second, the mean dry weight of an individual worker of Iridomyrmex anceps is about 0.4 mg, and our estimate of biomass removal from a tree containing 62 juveniles was about 405 mg. If there is a 10% rate of biomass conversion from one trophic level to the next, then foraging on a single tree containing about 60 juveniles of Jalmenus evagoras can result in the equivalent production of almost 100 new workers of Iridomyrmex anceps in one day.”

Finally, Pierce and her collaborators point out that there are several benefits for larvae and pupae of Jalmenus evagoras by aggregating. “If a threshold number of ants is necessary to protect the larvae and pupae, then aggregating is one mechanism by which Jalmenus evagoras could simultaneously increase its collective defense and decrease the amount of food that each individual would need to produce to attract that defense. For example, first instars can gain more ants by joining a group of any size than by remaining alone, and solitary second and third instars can have a higher number of attendant ants by joining the mean size group of about 4 larvae. Moreover, aggregation is not automatic, but occurs in response to ants: young instars that are not tended by ants are less likely to form groups than their tended counterparts. It is interesting and probably significant that most lycaenid species that lay eggs in clusters and whose larvae aggregate have complex and apparently obligate associations with ants (Kitching, 1981; Pierce and Elgar, 1985).” Larval aggregations in lycaenids are mainly known from Australian and South African species (Clark and Dickson, 1971; Common and Waterhouse, 1972). One North American species was recently discovered by Webster and Nielsen (1984) who found that the third and fourth instar larvae of Satyrium edwardsii aggregate during the day at the base of the host plant within conical shelters of detritus constructed by the ant Formica integra. The lycaenid larvae feed nocturnally on the leaves and are usually attended by the ants. Other Satyrium species are also tended by ants but do not form larval aggregations.

The main cost of aggregation for the lycaenids may be competition. Indeed, Pierce et al. found many larvae of Jalmenus evagoras that had starved on their host plants after consuming all the available foliage. Furthermore, larvae that occur in groups are likely to be more conspicuous to their predators and parasitoids. The costs and benefits of larval aggregation from the ants' perspective remain to be assessed experimentally, although Pierce et al. note that: “Ants benefit energetically from aggregation in Jalmenus evagoras:  larvae and pupae that occur in clusters are easier to collect secretions from and to defend. However, if larvae and pupae can regulate the amount of secretion they produce then individuals in groups may be able to provide less food for ants than they would on their own while still receiving the same degree of defense. Furthermore, aggregations may become so attractive to predators (such as other ant species) that the ants themselves are endangered.”

To summarize briefly, it has now been convincingly documented that in the case of at least several myrmecophilous lycaenids, ants protect the larvae and pupae against parasitoids and predators and in return they are rewarded with nutritious secretions provided by these guests.

The evolution of the myrmecophilous anatomical structures and behavior in the lycaenids has been discussed by Hinton (1951) and Malicky (1970a,b). Myrmecophily is judged a primitive trait among the living lycaenids, while its absence in some species is considered to be a secondary loss in evolution. The unusually thick cuticle of lycaenid larvae may be a first adaptation to myrmecophily. The Newcomer's gland is also primitive, having been lost in a considerable number of species as the result of decreased interaction with ants. Malicky proposed that the pore cupolas evolved after the Newcomer's gland. When pore cupolas became increasingly efficient as an appeasement device, the Newcomer's gland regressed. The overall evolutionary scheme by Hinton and Malicky has been endorsed at least in part by more recent authors, including Fukuda et al. (1978), Pierce (1983, 1987), and Henning (1983a,b), and challenged at least in part by Maschwitz and Fiedler (1988).

In addition to associating with ants, a number of lycaenids have changed from phytophagous to carnivorous habits, or parasitize ants in other ways (Figure 13-36). A phytophagy in the Lycaenidae has recently been reviewed by Cottrell (1984), who argues that the switch from phytophagy to carnivory has arisen at least eight separate times within the Lycaenidae. Many members of the Miletinae, for instance, prey on homopterans. Several authors, including Fukuda et al. (1978), Cottrell (1984), and Maschwitz and Fiedler (1988) have suggested that the proclivity of lycaenids to occupy areas inhabited by ants and their homopteran cultures may have led to the evolution of carnivorous life patterns.

Recently Maschwitz et al. (1985b,c) studied three species of the miletine genus Allotinus in Malaysia, and Kitching (1987) observed a fourth species in Sulawesi. Maschwitz and his collaborators found adult males and females of the butterfly imbibing honeydew from ant attended aphids, coccids, and membracids, and they observed females depositing eggs into the midst of loose aphid associations. The ants did not behave aggressively toward the honeydew-stealing butterflies, which suggests that the adult Allotinus enjoy some form of chemical appeasement protection. The larvae prey on the homopterans, and the older larvae of A. unicolor also imbibe honeydew. Miletine caterpillars in general are also endowed with chemical appeasement substances secreted from the pore cupolas, but they lack the nectar organs (Malicky, 1969; Kitching, 1987). In some species the larvae are visually camouflaged, resembling the substrate in which they live. This feature is more likely to be a protection against vertebrate predators than against ants.

Another mode of parasitic behavior in lycaenids has been described by Maschwitz et al. (1984). Many species of the paleotropic plant genus Macaranga (Euphorbiaceae) live in symbiosis with the myrmicine genus Crematogaster. The ants protect the plants from many herbivorous enemies, and in return the plants provide food-bodies and nesting space, and they make possible the culturing of honeydew-producing homopterans (see Chapter 14). Maschwitz et al. discovered that several species of the lycaenid genus Arhopala intrude parasitically into this symbiosis system. With the aid of their myrmecophilic organs (pore cupolas, nectar organ, and tentacle organ) the caterpillars overcome the aggressivity of the ants and feed on the Macaranga leaves without disturbance. Moreover, the caterpillars and their pupae are protected against predators and parasitoids by the ants. Maschwitz et al. describe the intricate balance of costs and benefits as follows: “The secretion of the Newcomer's gland fluid seems to play only a minor role when compared to the honeydew excretion of the scales. This means that these larvae are pests of the plant and by that they harm the scales as well as the protecting ants. There is no severe damage of the plants because the eggs are only oviposited in low numbers. This is a typical pattern of well adapted parasites, which do not destroy their hosts. Against vertebrate enemies which are searching for their prey the larvae are protected by camouflage. There is a strong species specific host plant selection by the butterflies.”

The evolution and maintenance of truly mutualistic symbioses between lycaenids and ants depends on the costs and benefits of this relationship for both partners (Pierce and Young, 1986; Pierce, 1987). Mutualism has been a conspicuously successful strategy for the butterflies. By far the majority of all lycaenid species has evolved in those subfamilies where the larvae are endowed with pore cupolas, nectar organs, tentacular organs, and other myrmecophilous adaptations.

The trophobionts
A great majority of the members of the three phylogenetically most advanced ant subfamilies, the Myrmicinae, Dolichoderinae, and Formicinae, attend homopterans to some extent. To employ one last term from Wasmann, the ants can be said to have entered into trophobiosis with the homopterans. As trophobionts, the homopterans resemble many of the lycaenid symbionts in a basic way: they obtain their own food and pass some of it on to their hosts. However, unlike the lycaenids that secrete substances from specialized exocrine glands, the honeydew provided for ants by homopterans is an excretion derived from a digestive process (see Plate 17).

When aphids feed on the phloem sap of plants, they pass a complex mixture of nutrients, including sugars, free amino acids, amides, proteins, minerals and vitamins, through their gut and back out through the anus. During this passage the phloem sap changes as some of its components are absorbed and others are converted or added by the aphid (for recent reviews see Ziegler and Penth, 1977; Kunkel and Kloft, 1977; Dixon, 1985; Maurizio, 1985). According to Maurizio, O.2-1.8 percent of the honeydew dry weight is nitrogen and 70-95 percent of the nitrogen are amino acids and amides. The mixture of nitrogen compounds in the honeydew is largely identical to that in the phloem sap. Measurements made on Tuberolachnus salignus by Mittler (1958) show that as much as one-half of the free amino acids are absorbed by the aphid's gut. In a few cases the honeydew contains amino acids which are not present in the phloem sap. Presumably these are metabolic products added by the aphids (Gray, 1952; Ehrhardt, 1962).

By far the largest percentage (90-95 percent) of the honeydew dry weight consists of carbohydrates. Sugars from the phloem sap are partly absorbed or converted, and the diverse mixtures of sugars contained in the honeydew are often species-specific. They usually comprise fructose, glucose, saccharose, trehalose and higher oligosaccharides. Trehalose, which is the blood sugar of insects, composes up to 35 percent of the total sugar amount in the honeydew. Typical honeydew sugars also include the trisaccharides fructo-maltose and melezitose, with the latter making up 40-50 percent of the total sugar. Other sugars detected in honeydew are maltose, raffinose, melibiose, turanose, galactose, mannose, rhamnose, and stachyose. In addition the honeydew contains other classes of substances, including organic acids, B-vitamins, and minerals.

When unattended by ants, many aphids dispose of the honeydew droplets by flicking them away with their hind legs or caudae, or by expelling them through contractions of the rectum or entire abdomen. The honeydew then falls upon the vegetation and ground below. Similar substances are excreted by several other groups of sap-feeding Homoptera, including scale insects (Coccidae), mealybugs (Pseudococcidae), jumping plant lice (Chermidae = Psyllidae), tree hoppers (Jassidae, Membracidae), leafhoppers (Cicadellidae), froghoppers or spittle insects (Cercopidae), and members of the “lantern-fly” family (Fulgoridae). Sometimes honeydew accumulates in large enough quantities to be usable by man. The manna “given” to the Israelites in the Old Testament account was almost certainly the excretion of the coccid Trabutina mannipara, which feeds on tamarisk. The Arabs still gather the material, which they call “man.” In Australia, chermid honeydew is collected as food by the aborigines. Referred to as “sugar-lerp,” up to three pounds can be harvested by one person in a single day. It is no surprise, therefore, to find that ants also gather honeydew of all kinds. Many, perhaps most, species collect it from the ground and vegetation where it falls. But many others have developed the capacity to solicit the honeydew directly from the homopterans themselves.

Most aphid species associated with ants insert their stylets into the phloem of the host plant (Kloft, 1953, 1959a, 1960a,b,c; Kunkel, 1967). Although they can suck up limited amounts of liquid, the aphids appear to depend chiefly on the turgor pressure of the phloem, which forces sap up their stylets (Mittler, 1957; Kunkel and Kloft, 1985). To process a large volume of phloem sap and discard the excess as honeydew evidently costs the aphid fewer calories than a more nearly total extraction from smaller quantities of sap. The amounts of honeydew produced by individuals are often prodigious. First instar nymphs of Mittler's Tuberolachnus extracted honeydew at the rate of seven droplets per hour, each droplet containing 0.065 l, and the total output per aphid was 133 percent of the aphid's weight every hour. Other species that have been analyzed are slightly more modest, their hourly output ranging from 1.9 to 13.3 percent of body weight per hour (Auclair, 1963).

Most myrmecophilous homopterans, especially aphids, have special structural and behavioral adaptations for life with ants (Way, 1963; Kunkel, 1973; Kunkel and Kloft, 1985). Aphids frequently associated with ants tend to have poorly developed cornicles, a reduced cauda, and at most a thin coating of wax filaments. However, a few ant-attended species have large cornicles and others are densely covered with wax. In the case of one such species, Prociphilus fraxini, the ants simply remove wax from the bodies of the aphids (Zwölfer, 1958; Kunkel, 1973). And where most pseudococcids are covered with wax, the ants often clean the symbionts bare. Kunkel (1973) notes that myrmecophilous aphids generally have more setae on the dorsal body surface and tibiae. Anal setae in particular are very numerous in myrmecophilous aphids. They form a basket (“trophobiotic organ”) that holds the honeydew droplet until it is imbibed by the ants (Zwölfer, 1958; Kunkel, 1973).

Experiments by El-Ziady and Kennedy (1956) and Johnson and Birks (1960) revealed that the presence of ants belonging to the genera Lasius and Paratrechina delays the production of alates in populations of species of Aphis and hence postpones their dispersal and, from that, increases their population density. In addition, alate aphids have long wings that render it more difficult for the ants to collect the honeydew droplets from the anus. Kleinjan and Mittler (1975) demonstrated that if extracts of the mandibular glands of Formica fusca are applied to aphid nymphs, the proportions of the population exhibiting winglessness increases. Since it is known that winglessness is caused by juvenile hormone, the ants' mandibular gland secretions may well contain material simulating juvenile hormone. Finally, Kunkel (1973) reported that the wings of alate aphids are sometimes bitten and crumpled by their ant guardians.

Numerous authors, beginning with Pierre Huber (1810), have reported that honeydew ejection of certain homopterans has been modified under the attendance of ants (see reviews by Way, 1963; Kunkel, 1973). As a rule, symbiotic homopterans ease out the droplets of honeydew when solicited by ants rather than ejecting them away from their bodies. Individuals of the black bean aphid, Aphis fabae, show the following typical specialized responses in the presence of ants: the abdomen is raised slightly, the hind legs are kept down instead of being lifted and waved as in unattended aphids, and the honeydew droplet is emitted slowly and held on the tip of the abdomen while it is being consumed by the ants. If a droplet is not accepted, the aphid will often withdraw it back into the abdomen. Kunkel (1973) interprets the repeated alternate extrusion and withdrawal of honeydew droplets as a signal to the ants that the aphid is ready to defecate. In many cases the required stimulus for honeydew emission is a simple, mechanical one. According to Mordwilko (1907) and Kunkel (1973), many (but not all) of the symbiotic aphid species can be induced to emit a droplet merely by brushing the abdomen with some delicate object in imitation of the caressing movements of the ant's antennae and forelegs. In other species the aphids appear to exchange mechanical signals with the ants prior to releasing a honeydew droplet. For example Lachnus roboris lifts the hind legs when an ant approaches (Figure 13-37). Kloft (1959, 1960a) has made the intriguing suggestion that the rear of the aphid's abdomen resembles the head of an ant worker offering food, with the cornicles representing the opened mandibles, the cauda symbolizing the ant's extruded labium, and the waving of the aphid's hind legs providing an imitation of the antennal movements of the ant. These stimuli induce the visiting worker to mistake the aphid for a donor ant in the special way that any animal makes a mistake when confronted with a small but vital set of releasers out of context. The solicitation that follows, according to Kloft, is just a slight perversion of the ordinary trophallaxis occurring between sister workers. Kloft hypothesizes that this interspecific communication behavior in aphids evolved from a nonspecific defensive behavior. Approximately 75 percent of all species of the Aphidina, when disturbed, kick with their hind legs (Kunkel, 1973). Trophobiotic associations evolved independently many times within the Aphidina and in some cases the leg kicking may have been ritualized to become an interspecific trophobiotic communication signal. This ingenious hypothesis still needs to be tested experimentally. In any case, interspecific communication is not a general prerequisite of trophobiotic interactions. Coccids and mealybugs are attended with equal fervor and precision, yet their appearance and behavior make them appear wholly different from their ant hosts (Eckloff, 1978; Figures 13-38 and 13-39). It is entirely possible that the subtle resemblances between certain aphids and the heads of ants are just coincidental.