The Ants Chapter 19

AntWiki: The Ants --- Online

CHAPTER 19. WEAVER ANTS

Introduction

Among the thousands of social insects a few deserve to be called classic, because certain remarkable features in the behavior have prompted unusually careful and thorough studies. The honeybees, the bumblebees, the driver ants, the army ants, the leafcutter ants, the slave-maker ants, and the fungus-growing termites are all examples of classic social insects. The latest entry in this select group are the weaver ants of the genus Oecophylla. The ants are relatively large, with bodies ranging up to 8 mm in length, and exclusively arboreal. The workers create natural enclosures for their nests by first pulling leaves together (see Figure 19-1) and then binding them into place with thousands of strands of larval silk woven into sheets. In order for this unusual procedure to succeed, the larvae must cooperate by surrendering their silk on cue, instead of saving it for the construction of their own cocoons. The workers bring nearly mature larvae to the building sites and employ them as living shuttles, moving them back and forth as they expel threads of silk from their labial glands.

Perhaps the first description of the biology and remarkable nest construction of Oecophylla was made by Joseph Banks, who accompanied Captain Cook on the voyage of the H.M.S. Endeavour in 1768 to Australia. In his Journal, Banks (cited in Musgrave, 1932) describes his first encounters with the green tree ant in that part of New Holland now called New South Wales:

Of insects there were but few sorts, and among them only the ants were troublesome to us. Mosquitoes, indeed, were in some places tolerably plentiful, but it was our good fortune never to stay any time in such places. The ants, however, made ample amends for the want of the mosquitoes; two sorts in particular, one green as a leaf, and living upon trees, where it built a nest, in size between that of a man's head and his fist, by bending the leaves together, and glueing them with whitish paperish substances which held them firmly together. In doing this their management was most curious: they bend down four leaves broader than a man's hand, and place them in such a direction as they choose. This requires a much larger force than these animals seem capable of; many thousands indeed are employed in the joint work. I have seen as many as could stand by one another, holding down such a leaf, each drawing down with all his might, while others within were employed to fasten the glue. How they had bent it down I had not the opportunity of seeing, but it was held down by main strength, I easily proved by disturbing a part of them, on which the leaf bursting from the rest, returned to its natural situation, and I had an opportunity of trying with my finger the strength that these little animals must have used to get it down.

But industrious as they are, their courage, if possible excels their industry; if we accidentally shook the branches on which such nest was hung, thousands would immediately throw themselves down, many of which falling upon us made us sensible of their stings and revengeful disposition, especially if, as was often the case, they got possession of our necks and hair, their stings were by some esteemed not much less painful than those of a bee; the pain, however, lasted only a few seconds.

Oecophylla belongs to the subfamily Formicinae, the species which lacks a functional sting. The painful “sting” Banks describes is actually the bite of the sharp and powerful mandibles, perhaps intensified by irritating secretions from the mandibular glands. The whitish glue that firmly holds the leaves together was described 137 years later as the salivary gland secretions of the Oecophylla larvae (Doflein, 1905).

The construction of communal silk nests has clearly contributed to the success of the Oecophylla weaver ants. It permits colonies to attain immense populations, in spite of the large size of the workers, because the ants are freed from the spatial limitations imposed on species that must live in beetles' burrows, leaf axils (the area between the stems of leaves and the parent branch), and other preformed vegetative cavities. This advance, along with the complex recruitment system that permits each colony to dominate up to several trees at the same time, has helped the weaver ants to become among the most abundant and successful social insects of the Old World tropics (Hölldobler and Wilson, 1977a,c,d, 1978; Hölldobler, 1979, 1983). A single species, Oecophylla longinoda, occurs across most of the forested portions of tropical Africa, while a second, closely related species, Oecophylla smaragdina, ranges from India to Queensland, Australia, and the Solomon Islands (Lokkers, 1986). The genus is ancient even by venerable insect standards: two species are known from Baltic amber of Oligocene age, dating back to about 30 million years (Wheeler, 1914b). Oecophylla leakeyi, described from a fossil colony of Miocene age (approximately 15 million years old) found in Kenya, possessed a physical caste system very similar to that of the two living forms (Wilson and Taylor, 1964). In particular, the allometry by which the minor and major workers are differentiated is closely similar across the three species. Also, both the living and extinct species have a bimodal size-frequency distribution, with the major workers more common than the minor workers, and media workers present but scarce.

With their huge colonies and ability to construct nests almost anywhere, the Oecophylla weaver ants have achieved a close control of their environment. The empire of the African species can be visualized by the following census made by Vanderplanck (1960) of only a part of a colony on Zanzibar: in 192 leaf nests there were 62,694 workers; 18,498 worker larvae and pupae; and 657 queen larvae and pupae. The total mature and immature population was estimated to fall between 115,000 and 164,000. Yet this was by no means a large colony. Populations of a half million more often occur, with nests extending through the crowns of up to three or more good-sized trees. Many nests contain scale insects that the weaver ants keep and protect for the sugary excrement they produce. Some of the peripheral nests are “barracks,” containing mostly aging workers that sally out to attack alien ants and other intruders as soon as they cross the territorial borders (Hölldobler, 1983). The workers use no fewer than five recruitment systems, comprising differing chemical and tactile signals, to organize territorial defense, foraging, and the exploration of new terrain. Some of the pheromones come from rectal and sternal glands of a type unique to Oecophylla (see Chapter 7).

So far as known, only one queen occurs in each colony, and she is extremely attractive to major workers (Figure 19-1). Stimuli from her head, evidently chemical in nature, induce the workers to regurgitate and present trophic eggs at frequent intervals. The queen suppresses the laying of viable eggs but not trophic eggs by the major workers; the effect is mediated by pheromones and persists in the corpses of the queens for as long as six months. When the live queen or her corpse is removed from the nest, some of the workers start to lay viable but unfertilized eggs. The head of the queen has exceptionally large pro- and postpharyngeal glands as well as fully developed maxillary glands and mandibular glands. Her thorax and abdomen are lined with clusters of glandular cells drained by long ducts that exit from the exoskeleton. Since the mother queen does virtually nothing but eat and lay eggs, it seems likely that this unusual armamentarium of exocrine tissue is the source of the queen's pheromones (Hölldobler and Wilson, 1983a).

The environmental control exercised by the weaver ants can be put to practical use. Records from southern China show that weaver ant nests have been gathered, sold and placed in selected citrus trees to combat insect pests for approximately 1,700 years. The same basic techniques were repeatedly noted in the classical Chinese literature between A.D. 304 and 1795, and the practice is continued today as an alternative to chemical control in the provinces of Guangdong and Fujian (Huang and Yang, 1987). The weaver ant used for this purpose is the Asian species Oecophylla smaragdina. This utilization of weaver ants is the oldest-known instance of the biological control of insects in the history of agriculture. Dennis Leston (personal communication) recommended employing the African species of weaver ant to control pests of tree crops such as cacao. Studies in Ghana have shown that the presence of weaver ants reduces the incidence of two of the most serious diseases of cacao, one caused by a virus and the other by a fungus. In both cases the pathogen is transmitted by mirid leaf bugs. The weaver ants evidently combat the diseases by attacking the bugs. The Oecophylla workers are also particularly effective in hunting insects that feed on the tissue and sap of trees.

Communal nest weaving

Our recent studies, building on those of other authors, have revealed an unexpectedly precise and stereotyped relation between the adult workers and the larvae. The larvae contribute all their silk to meet the colony's needs instead of their own. They produce large quantities of the material from enlarged silk glands early in the final instar rather than at the end, thus differing from cocoon-spinning ant species, and they never attempt to construct cocoons of their own (Wilson and Hölldobler, 1980; Hölldobler and Wilson, 1983b). The workers have taken over almost all the spinning movements from the larvae, turning them into passive dispensers of silk.

It would seem that close attention to the exceptional properties of Oecophylla nest-weaving could shed new light on how cooperation and altruism operate in ant colonies, and especially on how larvae can function as an auxiliary caste. In addition, a second, equally interesting question is presented by the Oecophylla case: How could such extreme behavior have evolved in the first place? As is the case with the insect wing, the vertebrate eye, and other biological prodigies, it is hard to conceive how something so complicated and efficient in performance might be built from preexisting structures and processes. Fortunately, other phyletic lines of ants have evolved communal nest-weaving independently and to variably lesser degrees than Oecophylla, raising the prospect of reconstructing the intermediate steps leading to the extreme behavior of weaver ants. These lines are all within the Formicinae, the subfamily to which Oecophylla belongs. They include all the members of the small Neotropical genus Dendromyrmex; the two Neotropical species Camponotus (Myrmobrachys) senex and C. (M.) formiciformis, which are aberrant members of a large cosmopolitan genus; Camponotus (Karavaievia) gombaki, C. (K.) texens, and probably other members of the tropical Asian subgenus Karavaievia; and various members of the large and diverse Old World tropical genus Polyrhachis.

Two additional but doubtful cases have been reported outside the Formicinae. According to Baroni Urbani (1978b), silk is used in the earthen nests of some Cuban species of Leptothorax, a genus of the subfamily Myrmicinae. However, he was uncertain whether the material is obtained from larvae or from an extraneous source such as spider webs. Since no other myrmicine is known to produce silk under any circumstances, the latter alternative seems the more probable. Similarly, the use of silk to build nests was postulated for the Javan ant Technomyrmex bicolor textor, a member of the subfamily Dolichoderinae, in an early paper by Jacobson and Forel (1909). But again, the evidence is from casual field observations only, and the conclusion is rendered unlikely by the fact that no other dolichoderines are known to produce silk.

True silk weaving is easily confused with the construction of carton nests and shelters, which is a widespread practice in Ectatomma, Crematogaster, Pheidole, Solenopsis, Monacis, Azteca, Tapinoma, Lasius, and many other genera. These ants use bits of plant fibers and other detritus, often combined with soil, to construct galleries and chambers on the surface of vegetation. Although the structures are often quite elaborate, they do not contain silk.

In recent years we have studied the behavior of both living species of Oecophylla in much greater detail than earlier entomologists, and have extended our investigations to the other, lesser known nest-weaving genera, Camponotus, Dendromyrmex, and Polyrhachis. This research has made possible a preliminary characterization of the stages through which the separate evolving lines appear to have passed.

In piecing together the data, we utilized a now-standard concept in organismic and evolutionary biology, the phylogenetic grade (Hölldobler and Wilson, 1983b). The four genera of formicine ants we considered are sufficiently distinct from each other on anatomical evidence as to make it almost certain that the communal nest-weaving displayed was in each case independently evolved. Thus it is proper to speak of the varying degrees of cooperative behavior and larval involvement not as the actual steps that led to the behavior of Oecophylla but as grades, or successively more advanced combinations of traits, through which autonomous evolving lines are likely to pass. Other combinations are possible, even though not now found in living species, and they might be the ones that were actually traversed by extreme forms such as Oecophylla. However, by examining the behavior of as many species and phyletic lines as possible, biologists are sometimes able to expose consistent trends and patterns that lend convincing weight to particular evolutionary reconstructions. This technique is especially promising in the case of insects, with several million living species to sample. Within this vast array there are 8,800 known species of ants, most of which have never been studied, making patterns of ant behavior exceptionally susceptible to the kind of analysis pursued in the case of nest weaving.

The highest grade of cooperation

The studies conducted on Oecophylla prior to our own were reviewed by Wilson (1971) and Hemmingsen (1973). In essence, nest-weaving with larval silk was discovered in Oecophylla smaragdina independently by H. N. Ridley in India and W. Saville-Kent in Australia, and was subsequently described at greater length in a famous paper by Doflein (1905). Increasingly detailed accounts of the behavior of Oecophylla longinoda, essentially similar to that of Oecophylla smaragdina, were provided by Ledoux (1950), Chauvin (1952), Sudd (1963), and Hölldobler and Wilson (1977c).

The sequence of behaviors by which the nests are constructed can be summarized as follows. Individual workers explore promising sites within the colony's territory, pulling at the edges and tips of leaves. When a worker succeeds in turning a portion of a leaf back on itself, or in drawing one leaf edge toward another, other workers in the vicinity join the effort. They line up in a row and pull together, or, in cases where a gap longer than an ant's body remains to be closed, they form a living chain by seizing one another's petiole (or “waist”) and pulling as a single unit (Figure 19-2). Often rows of chains are aligned so as to exert a powerful combined force (Plates 22-24). The formation of such chains of ants to move objects requires intricate maneuvering and a high degree of coordination. So far as is known, it is unique to Oecophylla among the social insects.

When the leaves have been maneuvered into a tentlike configuration, workers form rows and hold the leaves together (Figure 19-3). Another group of workers carries larvae out from the interior of the existing nests and uses them as sources of silk to bind the leaves together (Plate 24). Previous studies (Wilson and Hölldobler, 1980) showed that the Oecophylla longinoda larvae recruited for this purpose are all in the final of at least three instars, and have heads in excess of 0.5 mm wide. However, their bodies (exclusive of the rigid head capsule) are smaller than those of the larvae at the very end of the final instar, which are almost ready to turn into prepupae and commence adult development. Thus the larvae used in nest-weaving are well along in development and possess large silk glands, but they have not yet reached full size and hence are more easily carried and manipulated by the workers.

In Oecophylla longinoda, all the workers we observed with spinning larvae were majors, the larger adults that possess heads between 1.3 and 1.8 mm in width. Hemmingsen (1973) reported that majors of Oecophylla smaragdina perform the weaving toward the exterior, while minor workers--those with heads 1-1.2 mm wide--weave on the inner surfaces of the leaf cavities. We observed only major workers performing the task in Oecophylla longinoda, but admittedly our studies of interior activity were limited. Hemmingsen also recorded that exterior weaving is rare during the daytime but increases sharply at night, at least in the case of Oecophylla smaragdina working outdoors in Thailand. We saw frequent exterior weaving by Oecophylla longinoda during the day in a well-lit laboratory, as well as by Oecophylla smaragdina outdoors in Queensland.

In order to work out the details of the spinning process, we followed the entire sequence through a frame-by-frame analysis of 16-mm motion pictures taken at 25 frames per second. The most distinctive feature of the larval behavior, other than the release of the silk itself, is the rigidity with which the larva holds its body. There is no sign of the elaborate bending and stretching of the body or of the upward thrusting and side-to-side movements of the head that characterizes cocoon-spinning in other formicine ant larvae, particularly in Formica (Wallis, 1960; Schmidt and Gürsch, 1971). Rather, the larva keeps its body stiff, forming a straight line when viewed from above but a slightly curved, S-shaped line when seen from the side, with its head pointing obliquely downward as shown in Figure 19-5. Occasionally the larva extends its head for a very short distance when it is brought near the leaf surface, giving the impression that it is orienting itself more precisely at the instant before it releases the silk. The worker holds the larva in its mandibles between one-fourth and one-third of the way down the larva's body from the head, so that the head projects well out in front of the worker's mandibles.

The antennae of the adult workers are of an unusual conformation that facilitates tactile orientation along the edges of leaves and other vegetational surfaces. The last four segments are shorter relative to the eight segments closest to the body than in other ants we have examined, including even communal silk-spinning formicines such as Camponotus senex and Polyrhachis acuata. They are also unusually flexible and can be actively moved in various directions in a fashion seen in many solitary wasps.

As the worker approaches the edge of a leaf with a larva in its mandibles, the tips of the antennae are brought down to converge on the surface in front of the ant. For 0.2 ± 0.1 sec (x ± SD, n = 26, involving a total of 4 workers), the antennae play along the surface, much in the manner of a blindfolded person feeling the edge of a table with his hands. Then the larva's head is touched to the surface and held in contact with the leaf for one second (0.9 ± 0.2 sec, n = 26). During this interval, the tips of the worker's antennae are vibrated around the larva's head, stroking the leaf surface and touching the larva's head about 10 times (9.2 ± 3.6, n = 26). At some point the larva releases a minute quantity of silk, which attaches to the leaf surface.

About 0.2 sec before the larva is lifted up again, the worker spreads and raises its antennae. Then it carries the larva directly to the edge of the other leaf, causing the silk to be drawn out as a thread. While moving between leaves, the worker holds its antennae well away from the head of the larva. When it reaches the other leaf, it repeats the entire procedure exactly, except that the larva's head is held to the surface for only half a second (0.4 ± 0.01 sec, n = 26); during this phase the worker's antennae touch the larva about 5 times (5.2 ± 2.4, n = 26). In other words, the workers alternate between a longer time spent at one leaf surface and a shorter time at the opposing surface.

To summarize, the weaving behavior of the Oecophylla worker is even more complicated, precise, and distinctive than realized by earlier investigators. The movements are rigidly stereotyped in form and sequence. The antennal tips are used for exact tactile orientation, a “topotaxis” somewhat similar to that employed by honeybee workers to assess the thickness of the waxen walls of the cells in the comb (Lindauer and Martin, 1969). The worker ant also appears to use its flexible antennal tips to communicate with the larva, presumably to induce it to release the silk at the right moment. Although we have no direct experimental proof of this effect, we can report an incidental observation consistent with it. One worker we filmed held the larva upside down, so that the front of the larva's head and its silk-gland openings could not touch the surface or be stroked by the antennal tips. The worker went through the entire sequence correctly, but the larva did not release any silk.

For its part, the larva has evolved distinctive traits and behaviors that serve communal weaving. It releases some signal, probably chemical, that identifies it as being in the correct phase of the final instar. When a worker picks it up, the larva assumes an unusual S-shaped posture. And when it is held against the surface of a leaf and touched by a worker's antennae, it releases silk, in a context and under circumstances quite out of the ordinary for most immature insects.

Intermediate steps

The existence of communal nest-weaving in Polyrhachis was discovered in the Asiatic species Polyrhachis (Myrmhopla) dives by Jacobson (Jacobson and Wasmann, 1905). However, few details of the behavior of these ants have been available until a recent study by Hölldobler (in Hölldobler and Wilson, 1983b).

A species of Polyrhachis (Cyrtomyrma), tentatively classified near doddi, was observed in the vicinity of Port Douglas, Queensland, where its colonies are relatively abundant. The ants construct nests among the leaves and twigs of a wide variety of bushes and trees (Figure 19-4). Most of the units are built between two opposing leaves, but often only one leaf serves as a base or else the unit is entirely constructed of silk and is well apart from the nearest leaves.

Polyrhachis ants have never been observed to make chains of their own bodies or to line up in rows in the manner routine for Oecophylla. Occasionally a single Polyrhachis worker pulls and slightly bends the tip or edge of a leaf, but ordinarily the leaves are left in their natural position and walls of silk and debris are built between them.

The weaving of Polyrhachis also differs markedly from that of Oecophylla. The spinning larvae are considerably larger and appear to be at or near the end of the terminal instar (see Figure 19-6). The workers hold them gently from above, somewhere along the forward half of their body, and allow the larvae to perform all of the spinning movements. In laying silk on the nest wall, the larvae use a version of the cocoon-spinning movements previously observed in the larvae of Formica and other formicine ants. Like these more “typical” species, which do not engage in communal nest-building, Polyrhachis larvae begin by protruding and retracting the head relative to the body segments while bending the forward part of the body downward. Approximately this much movement is also seen in Oecophylla larvae prior to their being touched to the surface of a leaf.

The Polyrhachis larvae are much more active, however, executing most of the spinning cycle in a sequence very similar to that displayed by cocoon-spinning formicines. Each larva begins with a period of bending and stretching, then returns to its original position through a series of arcs directed alternately to the left and right; in sum, its head traces a rough figure eight. Because the larvae are held by the workers, the movements of their bodies are restricted. They cannot complete the “looping-the-loop” and axial rotary movements described by Wallis (1960), by which larvae of other formicine ants move around inside the cocoon to complete its construction. In fact, the Polyrhachis larvae do not build cocoons. They pupate in the naked state, having contributed all their expelled silk to the communal nest. In this regard they fall closer to the advanced Oecophylla grade than to the primitive Dendromyrmex one, to be described shortly.

Polyrhachis ants are also intermediate between Oecophylla and Dendromyrmex in another important respect. The Polyrhachis workers do not move the larvae constantly like living shuttles as in Oecophylla, nor do they hold the larvae in one position for long periods of time or leave them to spin on their own as in Dendromyrmex. Rather, each spinning larva is held by a worker in one spot or moved slowly forward or to the side for a variable period of time (range 1-26 sec, mean 8 sec, SD 7.1 sec, n = 29). After each such brief episode the larva is lifted up and carried to another spot inside the nest, where it is permitted to repeat the stereotyped spinning movements. While the larva is engaged in spinning, the worker touches the substrate, the silk, and the front half of the larva's body with its antennae. However, these antennal movements are less stereotyped than in Oecophylla.

The product of this coordinated activity is an irregular, wide-meshed network of silk extending throughout the nest. The construction usually begins with the attachment of the silk to the edge of a leaf or stem. As the spinning proceeds, some workers bring up small particles of soil and bark, wood chips, or dried leaf material that the ants have gathered on the ground below. They attach the detritus to the silk, often pushing particles into place with the front of their heads, and then make the larvae spin additional silk around the particles to secure them more tightly to the wall of the nest. In this way a sturdy outside shell is built, consisting in the end of several layers of silk reinforced by solid particles sealed into the fabric. The ants also weave an inside layer of pure silk, which covers the inner face of the outer wall and the surfaces of the supporting leaves and twigs. Reminiscent of wallpaper, this sheath is thin, very finely meshed, and tightly applied so as to follow the contours of the supporting surface closely. When viewed from inside, the nest of the Polyrhachis ant resembles a large communal cocoon (Figure 19-4).

A very brief description of the weaving behavior of Polyrhachis (Myrmhopla) simplex by Ofer (1970) suggests that this Israeli species constructs nests in a manner similar to that observed in the Queensland species. The genus Polyrhachis is very diverse and widespread, ranging from Africa to tropical Asia and the Solomon Islands. Many of the species spin communal nests, apparently of differing degrees of complexity. Recently, Yamauchi et al. (1987) demonstrated that colonies of Polyrhachis dives on Okinawa construct multiple nests out of leaves on trees or grass on the ground and larval silk. They form truly multicolonial systems, with most nests further containing multiple queens. A steady traffic of workers, brood and queens flows between the nests. Further study of the behavior of Polyrhachis weaver ants should prove very rewarding.

A second intermediate grade is represented by Camponotus (Myrmobrachys) senex, which occurs in moist forested areas of South and Central America. With the closely related Camponotus (Myrmobrachys) formiciformis, it is one of no more than six representatives of the very large and cosmopolitan genus Camponotus known to incorporate larval silk in nest construction (although admittedly very little information is available about most species of this genus), and in this respect must be regarded as an evolutionarily advanced form (Figure 19-5). The most complete account of the biology of Camponotus senex to date is that of Schremmer (1972, 1979a,b).

Unlike the other weaver ants, Camponotus senex constructs its nest almost entirely of larval silk. The interior of the nest is a complex three-dimensional maze of many small chambers and connecting passageways. Leaves are often covered by the silken sheets, but they then die and shrivel, and thereafter serve as no more than internal supports. Like the Australian Polyrhachis, Camponotus senex workers add small fragments of dead wood and dried leaves to the sheets of silk along the outer surface. The detritus is especially thick on the roof, where it serves to protect the nest from direct sunlight and rain.

As Schremmer stressed, chains of worker ants and other cooperative maneuvers among workers of the kind that characterize Oecophylla do not occur in Camponotus senex. The larvae employed in spinning are relatively large and most likely are near the end of the final instar. Although they contribute substantial amounts of silk collectively, they still spin individual cocoons--in contrast to both Oecophylla and the Australian Polyrhachis. Workers carrying spinning larvae can be seen most readily on the lower surfaces of the nest, where walls are thin and nest-building unusually active. During Schremmer's observations they were limited to the interior surface of the wall and consequently could be viewed only through the nascent sheets of silk. Although numerous workers were deployed on the outer surface of the same area at the same time, and were more or less evenly distributed and walked slowly about, they did not carry larvae and had no visible effect on the workers inside. Their function remains a mystery. They could in fact be serving simply as guards.

Although Schremmer himself chose not to analyze the weaving behavior of Camponotus senex in any depth, we were able to make out some important details from a frame-by-frame analysis of his excellent film (Schremmer, 1972). In essence, Camponotus senex appears to be very similar to the Australian Polyrhachis in this aspect of their behavior. Workers carry the larvae about slowly, pausing to hold them at strategic spots for extended periods. They do not contribute much to the contact between the heads of the larvae and the surface of the nest. Instead, again as in Polyrhachis, the larvae perform strong stretching and bending movements, with some lateral turning as well. When held over a promising bit of substrate, larvae appear to bring the head down repeatedly while expelling silk. We saw one larva perform six “figure eight” movements in succession, each time touching its head to the same spot in what appeared to be typical weaving movements. The duration of the contact between its head and the substrate was measured in five of these cycles; the range was 0.4-1.5 sec and averaged 0.8 sec. During the spinning movements the workers play their antennae widely over the front part of the body of the larva and the adjacent substrate.

The nest-weaving of Camponotus senex, then, is basically the same as that of the Australian Polyrhachis. The only relevant difference between the two is that Camponotus senex larvae construct individual cocoons and Polyrhachis larvae do not.

A wholly new example of communal weaving in the intermediate evolution grade has been recently discovered by Maschwitz et al. (1985a) in the tropical Asian subgenus Karavaievia of Camponotus. The behavior has been observed in only two rare and previously undescribed Malaysian species, Camponotus (Karavaievia) gombaki and Camponotus (Karavaievia) texens, but judging from an earlier ambiguous note of Viehmeyer (1915) on Camponotus (Karavaievia) dolichoderoides, it is likely to be a general trait of this small but morphologically distinct group of species. The level of behavior in Karavaievia is approximately that of Camponotus senex. Both of the Malaysian species weave their nests with the aid of fully grown last-instar larvae, which they hold anterior to the midlength of the body. When stimulated by antennal strokings of the workers, the larvae swing their heads back and forth while expelling silk threads. Like the workers of Camponotus senex and species of Polyrhachis, those of Karavaievia do not conduct weaving movements of their own; they rely entirely on the initiative of the larvae held in their mandibles. As the weaving proceeds, a second group of workers gathers sand particles, bits of detritus, and plant fragments and inserts them into the fresh silk sheets. A third group bites into the loose silk, especially at the points of contact with the leaf surface, tightening and smoothing the sheet as a whole. A fourth group (in texens at least, which could be closely observed) transports honeydew-producing scale insects into the new pavilion. Thus a division of labor exists that is comparable to that displayed by the Oecophylla weaver ants during nest building.

Like Polyrhachis and Camponotus senex and unlike Oecophylla, the Karavaievia do not form chains of their bodies or pull leaves together as part of nest building. The silken “pavilions” are constructed directly onto broad leaf surfaces or (in the case of gombaki) between leaves that are left otherwise unmodified. Both species build multiple one-chambered pavilions in the foliage of one or more trees. Camponotus (Karavaeievia) texens resembles Oecophylla in its territorial domination of the nest trees. It excludes other ants and destroys or appropriates any scale insects found outside the pavilions.

The simplest type of weaving

A recent study of the tree ants Dendromyrmex chartifex and Dendromyrmex fabricii (both of these species =Camponotus nidulans has revealed a form of communal silk-weaving that is the most elementary conceivable (Wilson, 1981). The seven species of Dendromyrmex are concentrated in Brazil, but at least two species (chartifex and fabricii) range into Central America. The small colonies of these ants build oblong carton nests on the leaves of a variety of tree species in the rain forest (Weber, 1944b).

The structure of the nests is reinforced with continuous sheets of larval silk (Figure 19-6). When the nest's walls are deliberately torn to test their strength, it can be seen that the silk helps hold the carton together securely. Unlike Oecophylla larvae, those of Dendromyrmex contribute silk only at the end of the final instar, when they are fully grown and ready to pupate. Moreover, only part of the silk is used to make the nest. Although a few larvae become naked pupae, most enclose their own bodies with cocoons of variable thickness. Workers holding spinning larvae remain still while the larvae perform the weaving movements; in Oecophylla, the larvae are still and the workers move. Often the larvae add silk to the nest when lying on the surface unattended by workers. Overall, their nest-building movements differ from those of cocoon-spinning only by a relatively small change in orientation. And, not surprisingly, this facultative communal spinning results in a smaller contribution to the structure of the nest than is the case in Oecophylla and other advanced weaver ants.

Anatomical changes

The behavior of communally spinning ant larvae is clearly cooperative and altruistic in nature. If general notions about the process of evolution are correct, we should expect to find some anatomical changes correlated with the behavioral modifications that produce this cooperation. Also, the degree of change in the two kinds of traits should be correlated to some extent. And finally, the alterations should be most marked in the labial glands, which produce the silk, and in the external spinning apparatus of the larva.

These predictions have generally been confirmed. Oecophylla, which has the most advanced cooperative behavior, also has the most modified external spinning apparatus. The labial glands of the spinning larvae of Oecophylla and Polyrhachis are in fact much larger in proportion to the size of the larva's body than is the case in other formicine ant species whose larvae spin only individual cocoons (Karawajew, 1929; Wilson and Hölldobler, 1980). On the other hand, Camponotus senex larvae do not have larger labial glands than those of other Camponotus larvae. Schremmer (1979a) tried to fit this surprising result into the expected pattern by suggesting that the Camponotus senex larvae produce silk for longer periods of time than other species that weave nests communally, and therefore do not need larger glands. This hypothesis has not yet been tested.

Until recently, little was known about the basic structure of the spinning apparatus of formicines. Using conventional histological sectioning of larvae in the ant genera Formica and Lasius, Schmidt and Gürsch (1970) concluded that the silk glands open to the outside by three tube-like projections, or nozzles. They were indeed able to pull three separate silken strands away from the heads of larvae with forceps. However, our studies, which combine histology and the use of the scanning electron microscope, have led us to draw a somewhat different picture. In general, we found that the labial gland opens to the outside through a small slit with one nozzle at each end, as shown in Figure 19-7. This is the structure found in the Australian Nothomyrmecia macrops, which is considered to be the living species closest to the ancestors of the Formicinae, as well as in a diversity of formicines themselves. Among the formicines examined, including those engaged in communal nest-weaving, only Oecophylla has a distinctly different external spinning apparatus. The labial-gland slit of these extremely advanced weaver ants is enlarged into a single nozzle, incorporating and largely obliterating the lateral nozzles. As a result, it appears that each Oecophylla larva is capable of expelling a broad thread of silk--the kind of thread needed to create the powerful webs binding an arboreal nest together.

The uncertain climb toward cooperation

In order to summarize existing information on the evolution of communal spinning, the grades in Table 19-1 are defined according to the presence or absence of particular traits associated with communal nest-weaving. We believe that it is both realistic and useful to recognize three such stages. It is also realistic to suppose that the most advanced weaver ants, those of the genus Oecophylla, are derived from lines that passed through lower grades similar to, if not identical with, those exemplified by Dendromyrmex, Polyrhachis, and Camponotus.

On the other hand, we find it surprising that communal nest-weaving has arisen only four or so times during the one hundred million years of ant evolution. Even if new cases of this behavior are discovered in the future, the percentage of ant species that weave their nests communally will remain very small. It is equally puzzling that the most advanced grade was attained only once. The separate traits of Oecophylla nest-weaving provide seemingly clear advantages that should predispose arboricolus ants to evolve them. The remarkable cooperative maneuvers of the workers allow the colony to arrange the substrate in the best positions for the addition of the silk bonds and sheets. By taking over control of the spinning movements from the larvae, the workers enormously increase the speed and efficiency with which the silk can be applied to critical sites. For their part the larvae have benefited the colony by moving the time when they produce silk forward in the final instar, thus surrendering once and for all the ability to construct personal cocoons while simultaneously allowing workers to carry and maneuver them more effectively because of their smaller size.

The case of Dendromyrmex is especially helpful in envisioning the first steps of the evolution in behavior that culminated in the communal nest-weaving of Oecophylla. Although the contribution of the larvae to the structure of the nest is quite substantial, the only apparent change in their behavior is a relatively slight addition to their normal spinning cycle, so that the larva releases some silk onto the floor of the nest while weaving its individual cocoon. It is easy to imagine such a change occurring with the alteration of a single gene affecting the weaving program. Thus, starting the evolution of a population toward communal weaving does not require a giant or otherwise improbable step.

There is another line of evidence indicating the general advantage of communal nest-weaving and hence a relative ease of progression. We discovered that both male and female larvae contribute silk to the nest in the cases of Oecophylla (Wilson and Hölldobler, 1980) and Dendromyrmex (Wilson, 1981); male contribution has not yet been investigated in Polyrhachis and Camponotus. Because cooperation and altruism on the part of male ants is rare, males are always worthy of close examination. As Bartz (1982) showed, natural selection in the social Hymenoptera will favor the evolution of either male workers or female workers, but not both, and the restrictive conditions imposed by the haplodiploid mode of sex determination--used by all Hymenoptera--favor all-female worker castes. In fact, the sterile workers of hymenopterous societies are always female (Oster and Wilson, 1978). In boreal carpenter ants of the genus Camponotus, where the males do contribute some labor to the colony, it is in the form of food-sharing, an apparent adaptation to the lengthy developmental cycle of Camponotus. The males are kept in the colonies from late summer or fall to the following spring, and it benefits both the colony and the individual males to exchange liquid food (Hölldobler, 1966).

The contribution of silk by male weaver-ant larvae is a comparable case. When the queens of Oecophylla and Dendromyrmex die, some of the workers lay eggs, which produce males exclusively (Hölldobler and Wilson, 1983a). Such queenless colonies can last for many months, until the last of the workers has died. During this period it is clearly advantageous for male larvae to add silk to the nest, for their own survival as well as that of the colony as a whole.

In summary, then, weaver ants exemplify very well an important problem of evolutionary theory: why so many intermediate species possess what appear to be “imperfect” or at least mechanically less efficient adaptations. Two hypotheses can be posed to explain the phenomena that are fully consistent with the manifest operation of natural selection in such cases. The first is that some species remain in the lower grades because countervailing pressures of selection come to balance the pressures that favor the further evolution of the trait. In particular, the tendency for larvae to collaborate in the construction of nests could be halted or even reversed in evolution if surrendering the ability to make cocoons reduces the larva's chance of survival. In other words, the lower grade might represent the optimum compromise between different pressures.

The second, quite different hypothesis is that the communal weavers are continuing to evolve--and will eventually attain or even surpass the level of Oecophylla--but species become extinct at a sufficiently high rate that most such evolutionary trends are curtailed before they are consummated. Even a moderate frequency of extinction can result in a constant number of species dispersed across the various evolutionary grades.

At present we see no means of choosing between these two hypotheses or of originating still other, less conventional evolutionary explanations. The greatest importance of phenomena such as communal nest-weaving may lie in the prospects they offer for a deeper understanding of arrested evolution, the reasons why not all social creatures have attained what from our peculiar human viewpoint we have chosen to regard as the pinnacles of altruistic cooperation.


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

The Ants - Table of Contents