The Ants Chapter 7

CHAPTER 7. COMMUNICATION

Communication has been the most intensively studied subject in ant biology during the past twenty years, yielding a profusion of new results that have come to affect our understanding of social organization deeply. The demonstrated modes of communication are extremely diverse. There exist the expected tappings, stridulations, strokings, graspings, nudgings, antennations, tastings, and puffings and streakings of chemicals that evoke various responses from simple recognition to recruitment and alarm. To this list can be added other, often bizarre effects, such as the exchange of pheromones that inhibit caste development, the soliciting and exchange of trophic eggs and special secretions from the anal region, the acceleration or inhibition of work performance by the presence of other colony members, and programmed execution.

Researchers on communication in ants and other social insects have come to recognize the following twelve broad functional categories of responses:

1. Alarm

2. Simple attraction

3. Recruitment, as to a new food source or nest site

4. Grooming, including assistance at molting

5. Trophallaxis (the exchange of oral and anal liquid)

6. Exchange of solid food particles

7. Group effect:  either increasing a given activity (“facilitation”) or inhibiting it

8. Recognition, of both nestmates and members of particular castes

9. Caste determination, either by inhibition or by stimulation

10. Control of competing reproductives

11. Territorial and home range signals and nest markers

12. Sexual communication, including species recognition, sex recognition, synchronization of sexual activity, and assessment during sexual competition (see Chapter 3)

Chemical communication
If any single generalization applies to all of these categories, it is that chemical signals pervade them all. Thirty years ago, on the basis of early results then just emerging, Wilson (1958a) predicted the dominance of chemoreception in ant behavior as follows: “The complex social behavior of ants appears to be mediated in large part by chemoreceptors. If it can be assumed that ‘instinctive’ behavior of these insects is organized in a fashion similar to that demonstrated for the better known invertebrates, a useful hypothesis would seem to be that there exists a series of behavioral 'releasers,' in this case chemical substances voided by individual ants that evoke specific responses in other members of the same species. It is further useful for purposes of investigation to suppose that the releasers are produced at least in part as glandular secretions and tend to be accumulated and stored in glandular reservoirs.” With each improvement in bioassay design and organic microanalysis permitting the separation and bioassay of secretory substances, new evidence has been added to reinforce this early impression.

A generally accepted terminology has evolved to classify the functions of the chemical substances (Nordlund, 1981). A ''semiochemical is any substance used in communication, whether between species (as in symbioses) or between members of the same species (Law and Regnier, 1971). A ''pheromone is a semiochemical, usually a glandular secretion, used within a species; one individual releases the material as a signal and the other responds after tasting or smelling it (Karlson and Lüscher, 1959). An ''allomone is a comparable substance employed in communication across species, as for example a lure used by a predator in attracting its prey. It evokes a response that is adaptively favorable to the emitter but not to the receiver (Brown, 1968; Brown et al., 1970a). In contrast, the term ''kairomone was proposed by Brown et al. in 1970 to cover chemicals emitted by an organism that elicit a response adaptively favorable to the receiver but not to the emitter. Semiochemicals can be classified as olfactory or oral according to the site of their reception. Also, their various actions can be distinguished as either releaser effects (then we speak of “releaser pheromones”), comprising the classical stimulus-responses mediated wholly by the nervous system, or primer effects (induced by “primer pheromones”), in which endocrine and reproductive systems are altered physiologically. In the latter case the body is truly primed for new biological activity, responding afterward with an altered behavioral repertory when presented with appropriate stimuli (Wilson and Bossert, 1963).

The sum of current evidence, which will be described in the remainder of this chapter, indicates that pheromones play the central role in the organization of ant societies. In general, it appears that the typical ant colony operates with somewhere between 10 and 20 kinds of signals, and most of these are chemical in nature. This rule is illustrated very well by the fire ant Solenopsis invicta, perhaps the most thoroughly studied ant species in this respect. As summarized in Table 7-1, 13 signals of a communicative or quasi-communicative nature are employed. Of these, all but one or two are mediated through chemoreception.

Glandular sources. The typical ant worker is a walking battery of exocrine glands, developed to a degree well beyond that typifying nonsocial hymenopterans. More than ten of the organs have been implicated thus far in the production of semiochemicals. They vary greatly in form and distribution among the major groups of ants, as illustrated in Figures 7-1 to 7-20 to 7-20. A few, such as the sternal and rectal glands of Oecophylla, the independently evolved sternal gland of Onychomyrmex, the pygidial gland of Polyergus, and the cloacal gland of Camponotus, appear to be unique to particular genera and to have arisen de novo during the course of social evolution (Hölldobler and Wilson, 1978; Hölldobler, 1982b,d, 1984a). Others, such as the poison gland of the Formicinae and Pavan's gland (sternal gland) of the Aneuretinae and Dolichoderinae, are peculiar to higher groups and thus provide valuable clues for the reconstruction of ant phylogeny. At least one structure, the metapleural glands, characterizes the ants as a whole (Maschwitz, 1974; Hölldobler and Engel-Siegel, 1984). Still other glands are shared with aculeate bees and wasps, including those that are nonsocial. Their versatile employment as sources of pheromones illustrates the economy of evolution, or “Romer's rule” as it sometimes is called, whereby organs and new functions tend to arise by modifications of preexisting organs and functions rather than as true novelties. This evolutionary process has been repeated many times to create a confusing pattern of glandular form and function across the Formicidae.

The repeating pattern of communicative evolution can be partially deciphered by focusing on five of the key exocrine glands that occur widely through the ants and serve a variety of functions in different phylogenetic groups. These structures are the Dufour's gland, the poison gland, the pygidial gland, the sternal glands, and the mandibular glands. Our knowledge concerning them has been thoroughly summarized at successive intervals by Maschwitz (1964), Bergström and Löfqvist (1970, 1973), Blum and Hermann (1978a,b), Hölldobler and Engel (1978), Parry and Morgan (1979), Hölldobler (1982a, 1984c), Vander Meer (1983), Bradshaw and Howse (1984), Buschinger and Maschwitz (1984), Morgan (1984), Attygalle and Morgan (1985), and Schmidt (1986). The treatment of anatomy and biochemistry by Blum and Hermann is close to being exhaustive for the earlier period of this research.

The Dufour's gland is usually a small gland, usually finger-shaped but sometimes bulbous or bifurcate in form, that opens at the base of the sting very near the egress of the poison gland (Figures 7-17 and 7-18). On morphological grounds it has been assumed that doryline, ponerine and myrmicine ants can discharge the contents of the Dufour's gland and poison gland independently (Whelden, 1960; Hermann and Blum, 1967a,b). In contrast, formicines were believed to release the contents of the two glands simultaneously, because no mechanisms were known that could close either one separately (Percy and Weatherston, 1974). Recently, however, Billen (1982b) discovered a closing apparatus of the Dufour's gland in several formicine ants (Formica sanguinea, Formica fusca and Lasius fuliginosus). Four sets of muscles play a part, two of which are directly attached to the slit-like glandular duct (Figure 7-19). Billen suggests that the opening of the Dufour's gland duct is achieved by active muscular contractions, while its closure is achieved by a passive return to the rest position of the thickened cuticular intima. A similar structure had been previously found by Beck (1972) in Formica sanguinea and Polyergus rufescens, but was not recognized as a closing mechanism. In a comparative ultrastructural study of the glandular epithelium of the Dufour's gland in ants, Billen (1986a) discovered remarkable differences in the cellular organization in eight ant subfamilies (Figure 7-20). Most of the Myrmicinae and Ponerinae possess a rather simple epithelium without special modifications. “In the African Dorylinae, the epithelium has a crenellated appearance and numerous basal invaginations, while the New World Ecitoninae have a very uniform epithelium with a basal layer of membrane foldings. Myrmeciinae, Pseudomyrmecinae and Dolichoderinae, each shows a different kind of apical microvilli, whereas Formicinae exhibit a characteristic subcuticular layer of mitochondria and a very thick basement membrane.” In the Myrmicinae the Dufour's gland produces only aliphatic hydrocarbons, but in dazzling variety among the various species, including such compounds as methylundecane, tridecane, hexadecane, hexadecene, and an array of farnesenes. In formicine ants the Dufour's gland is even more versatile. Aliphatic hydrocarbons are produced in abundance in most species, with n-undecane and n-tridecane typically present as major components and longer chain hydrocarbons as minor components. The alkanes are often accompanied by their corresponding alkenes, and in several species their dienes are also present. In addition, a great many oxygenated compounds occur in various combinations with the alkanes, especially in Lasius. They include alcohols, ketones, esters, acids, and lactones. The evolutionarily primitive function of the Dufour's gland and its basic set of alkanes is still uncertain. At least some of the compounds mediate alarm, recruitment, and sexual attraction among various species of ants. These communicative roles are clearly a derived condition within the Hymenoptera in general and the Formicidae in particular.

The poison gland apparatus typically consists of paired filamentous glands that converge into a single convoluted gland, which in turn empties into a thin-walled, sac-like reservoir or “poison sac.” The most evolved version is that found in the Formicinae. The convoluted gland is located on the dorsum of the poison sac, a condition unique within the Hymenoptera. The sac as a whole is also exceptionally large and it produces large quantities of formic acid by biosynthetic pathways that are now known (Hefetz and Blum, 1978). This simplest of all organic acids has for historical reasons been popularly regarded as characteristic of all ants, perhaps because it was one of the first natural products isolated in pure form, from the distillate of Formica workers in 1670. Nevertheless, it is evidently limited to the subfamily Formicinae.

The primary function of the poison gland in ants is the production of formic acid (in the Formicinae) or venom used in predation and defense. The primitive components, shared as a class with other aculeate hymenopterans, are proteinaceous. They are also neurotoxic, histolytic, or both in their effect--hence crippling to small invertebrate enemies and painful to human beings. This type of venom is the most common form in the anatomically more primitive ant subfamilies, namely the Ponerinae, Myrmeciinae, Pseudomyrmecinae, Dorylinae, and Ecitoninae. It is widespread among the tribes and genera of the Myrmicinae as well. Its effects are enhanced in bulldog ants of the genus Myrmecia by the addition of histamine and histamine-releasing factors. The “fire” in the venom of fire ants (Solenopsis), which indeed feels like a pinpoint burn, is caused by an unusual class of alkaloids, the piperidines, with 2,6-dialkylpiperidines composing the major components. In some species of myrmicines and formicines, a few constituents of the poison gland serve as recruitment or alarm substances. In Monomorium and Solenopsis at least, they are effective repellents against enemy ants and other arthropods.

In the Dolichoderinae the poison gland is typically reduced, and its function is replaced at least in part by the abundant toxic secretions of the pygidial gland. The homology of the pygidial gland has only recently been determined by anatomical studies. Its importance for ant biology has been enhanced by its newly discovered ubiquity and diversity in subfamilies additional to the Dolichoderinae. The history of research on the pygidial gland provides a cameo of the often haphazard way that knowledge of anatomy and behavior is acquired. In his magisterial study of the Myrmica rubra worker, Janet (1898) discovered the gland as a cluster of a few cells under the VIth abdominal tergite, with ducts leading to the intersegmental membrane between the VIth and VIIth tergites. After 80 years it was rediscovered as a well-developed organ in Novomessor by Hölldobler et al. (1976), and subsequently shown to be widespread among other genera of the Myrmicinae by Hölldobler and Engel (1978) and Kugler (1978b). Because the gland opens at the VIIth tergite (the pygidium), the name given it by Kugler, the pygidial gland, is now generally accepted (see Figure 7-11). Substances from the gland have been shown to function as alarm pheromones in three myrmicine genera. The large desert harvesters Aphaenogaster albisetosa and Aphaenogaster cockerelli (Figure 7-21) release strong-smelling components to evoke a form of “panic alarm,” which evidently serves to organize swift evacuations during the approach of army ants. Workers of Orectognathus versicolor, a highly predaceous Australian dacetine species, lay alarm recruitment trails to prey (Hölldobler, 1981b). Yet another evolutionary direction has been taken by the South American Pheidole biconstricta: minor workers produce large quantities of a secretion from their hypertrophied pygidial gland that are used in both chemical defense and aggressive alarm (Kugler, 1979). In Pheidole embolopyx, a Brazilian species, the major workers discharge alarm pheromones from the pygidial gland (Wilson and Hölldobler, 1985).

Once defined anatomically in the Myrmicinae, the pygidial gland was quickly located as well in the subfamilies Ponerinae, Myrmeciinae, Dorylinae, Pseudomyrmecinae, Aneuretinae, and Dolichoderinae (Hölldobler and Engel, 1978). Only the Formicinae lack the gland altogether, except in the slave raiding genus Polyergus, where it appears as an independent evolutionary development (Hölldobler, 1984a). A surprising find, however, was the recognition of the structure in the Dolichoderinae. Generations of researchers had diagnosed this subfamily in part by the possession of the supposedly unique “anal gland,” which produces strongly odorous secretions often referred to informally as the “Tapinoma odor”--after the dolichoderine genus Tapinoma. Now it is recognized that the anal gland is homologous to the pygidial gland of other ant groups. The finding has bearing on the phylogeny of several subfamilies. It has long been thought that the Aneuretinae are ancestral to the Dolichoderinae, on the basis of common features in external anatomy (Wheeler, 1914; Wilson et al., 1956). Recent studies by Traniello and Jayasuriya (1981a,b) of the pygidial gland, as well as the sternal gland, in the sole living species Aneuretus simoni, lend further support to this hypothesis. On the other hand W. L. Brown (quoted in litt. by Kugler, 1978) suggests that “the Aneuretinae might just be closer to the Myrmicinae than has been thought.” The anatomy and functions of the pygidial gland are at least consistent with this additional linkage. Furthermore, Blum and Hermann (1978b), noting similarities in the chemistry of the secretions of the mandibular glands in several myrmicine and dolichoderine species, concluded that “from an exocrinological standpoint, the Dolichoderinae have far more in common with the Myrmicinae than any other formicid subfamily.” Because Taylor (1978) considers the Nothomyrmeciinae ancestral to the Aneuretinae, it is noteworthy that the pygidial gland secretions of the very primitive Nothomyrmecia macrops elicit an aggressive alarm response in nestmates as well as a repellent effect on some other ant species occurring sympatrically with it (Hölldobler and Taylor, 1983). Thus in the Nothomyrmeciinae, Aneuretinae, Dolichoderinae, and Myrmicinae, the pygidial gland appears to produce alarm pheromones, defensive substances, or both.

The pygidial gland is both widespread and functionally diverse in the Ponerinae (Hölldobler and Engel, 1978; Jessen et al., 1979; Fanfani and Dazzini Valcurone, 1986). The secretions also play different roles from those of the nothomyrmeciine-myrmicine complex: in several species thus far studied, they elicit either recruitment or sexual attraction. In some species of Pachycondyla they are used in either tandem running or trail laying (Hölldobler and Traniello, 1980a,b; Traniello and Hölldobler, 1984; see Figures 7-22 and 7-23). In species of Leptogenys, Cerapachys, and Sphinctomyrmex, the pygidial gland substances are mixed with poison gland pheromones to produce odor trails (Maschwitz and Schönegge, 1977, 1983; Hölldobler, 1982b, and unpublished data). Finally, the results of preliminary experiments suggest that the pygidial gland is at least one of the sources of the trail pheromones in ecitonine army ants (Hölldobler and Engel, 1978).

A plethora of sternal glands, representing several independent evolutionary origins, has been discovered in ants. “Pavan's gland,” a well-developed, often paddle-shaped structure located beneath the VIIth sternite, is the source of the trail pheromone in the Aneuretinae and Dolichoderinae (see Figure 7-16). It consists of a medioventral sac between the VIth and VIIth abdominal sternites, which serves as the gland's reservoir, and a thick glandular epithelium on the anterior margin of the VIIth sternite (Traniello and Jayasuriya, 1981; Fanfani and Dazzini Valcurone, 1984; Billen, 1985b). It might well have originated in the primitive aneuretines, which in turn gave rise to the dolichoderines in late Cretaceous or early Eocene times (Wilson et al., 1956; Traniello and Jayasuriya, 1982). Many myrmicine species possess paired clusters of cells beneath the VIIth sternite, but their anatomy is so different as to suggest that they are not homologous with Pavan's gland (Hölldobler and Engel, 1978). Nothing is known at the present time concerning their function, although some circumstantial evidence reported by Cammaerts (1982) suggests that secretion obtained from the seventh abdominal sternite serves as an auxiliary trail pheromone in Myrmica.

The greatest variety of sternal glands has been encountered in the Ponerinae (Hölldobler and Engel, 1978; Jessen et al., 1979; Fanfani and Dazzini Valcurone, 1986). In the termite-hunting Paltothyreus tarsatus of Africa these structures occur beneath the intersegmental membranes that connect the terminal three abdominal sternites, and they produce pheromones for both the recruitment and orientation trails (see Figure 7-7). Workers of Onychomyrmex, an Australian genus unique among the amblyoponine Ponerinae for its legionary (army-ant) behavior, has a single large gland that opens between the Vth and VIth abdominal sternites (see Figure 7-8). Its secretions serve as a powerful trail and recruitment pheromone during predatory raids and colony emigrations (Figure 7-24). Other, nonlegionary amblyoponines investigated thus far (in the genera Amblyopone, Myopopone, Mystrium, and Prionopelta) lack the gland. Thus both the gland and the communication it serves appear to have evolved de novo in Onychomyrmex as part of the army ant syndrome.

Sternal glands found in some species of the subfamily Formicinae are also unique. One such structure, apparently limited to Oecophylla weaver ants, occurs beneath the VIIth sternite (Hölldobler and Wilson, 1977d, 1978). This gland consists of an array of single cells, which send short channels into cuticular cups on the outer surface of the sternite (see Figure 7-4). Its original function might have been to secrete lubrication for the seventh abdominal segment, which is frequently rotated when the ant raises the gaster to spray venom through the acidopore. The secretions also function as a short-range recruitment signal. A very different organ is the “cloacal gland” found in several Camponotus species consisting of a paired cluster of glandular cells located at the base of the VIIth abdominal sternite. Each cluster is associated with a major duct elaborated from an invagination of the cloacal chamber. The channels of the glandular cells of each cluster open in dense bundles into these two major ducts (see Figure 7-6). Experiments on Camponotus ephippium suggest that the secretions of the cloacal gland serve as recruitment pheromones (Hölldobler, 1982d).

The mandibular glands are a pair of thin-walled sacs filled typically with mixtures of alcohols, aldehydes, and ketones. Each of the two structures consists of a flattened glandular mass on the surface of a reservoir. The exit ducts are always connected to the mesal side of the mandibles and open near the anterior edge of the preoral cavity (Blum et al., 1968b). The glands vary relatively little through the Formicidae, although they are generally small in the Ponerinae and large in the Formicinae. In a few species they are hypertrophied in connection with special functions. For example, in a Malaysian species of the Camponotus saundersi group, they extend posteriorly all the way into the abdomen and are burst by muscle contractions during combat (Maschwitz and Maschwitz, 1974; Figures 7-25 and 7-26).

When the mandibles are carefully torn away from the head capsule of ant workers, the gland often (but not always!) pulls free in an intact condition, making its study much easier. Buren et al. (1970) pointed out that longitudinal mandibular grooves are widespread in ants and other aculeate Hymenoptera. This observation led to the oft-cited suggestion that these structures serve as channels for the outward flow of mandibular gland secretions. However, the grooves do not extend to the gland orifice and in any case are on the opposite side of the mandible from the glandular orifices, making a guiding function unlikely (Hermann et al., 1971).

The mandibular gland secretions of the ants as a whole are so chemically diverse as to preclude any generalization at this time. The substances manufactured by ponerines are especially diverse, including (according to species) organic sulfides, ketones, pyrazines, and a salicylate ester. The glands of myrmicines are a “veritable storehouse of ethyl ketones” and are further often accompanied by their corresponding carbinols, according to Blum and Hermann (1978a). Those of the Formicinae are dominated by terpenoid constituents. The functions of the mandibular gland secretions appear to be primarily if not exclusively defensive and alarm communication. In most species of ant the two roles are combined, but their relative importance varies greatly from one species to the next. Thus in a few species the glands are large, produce copious quantities of toxic secretions, and appear to have little behavioral impact on the ants. In other species the glands are small yet contain behaviorally very active components.

The metapleural glands (also called metasternal or metathoracic glands) are complex structures located at the posterolateral corners of the alitrunk (Figure 7-27). Each consists of a cluster of glandular cells, with each cell draining through a duct into a common membranous collecting sac. The collecting sac leads directly into the storage chamber or reservoir, which is a simple sclerotized cavity. Externally the metapleural glands are often marked by a pronounced vault or “bulla,” and a slit-shaped opening to the outside (Figure 7-27a). Brown (1968) suggested that the glands produce pheromones for recognition and identification of nestmates and alien species (Brown, 1968), and recently Jaffe and Puche (1984) claimed that in Solenopsis geminata metapleural gland secretions serve as territorial markers. This general explanation seems unlikely, because other investigations have found no evidence that secretions from the metapleural glands are involved in communication at all (Maschwitz et al., 1970; Maschwitz, 1974). Maschwitz and his collaborators did, however, demonstrate that in a number of ant species the metapleural gland secretions serve as powerful antiseptic substances that protect the body surface and nest against microorganisms. One active antibiotic component of Atta sexdens, for example, is phenylacetic acid, of which one ant carries an average of 1.4 µg at any given moment. In Crematogaster difformis the hypertrophied metapleural glands contain a mixture of phenols, including mellein (Attygalle et al., 1988b). The worker regularly releases small amounts of this mixture, which serve as an antiseptic. But when she is attacked by enemy ants, particularly at the highly vulnerable petiolar-postpetiolar region of the abdomen, she suddenly discharges large quantities of the metapleural gland secretions, which then function as a powerful repellent. Finally, in Crematogaster inflata, which also possess hypertrophied metapleural glands, the sticky secretions serve primarily as an alarm-repellent substance (Maschwitz, 1974; Figure 7-28).

It is generally assumed that the metapleural glands are a universal and phylogenetically old character of the Formicidae. Even the extinct species Sphecomyrma freyi of Cretaceous age appears to have possessed one (Wilson et al. 1967a,b). The organ is well-developed in the Ponerinae, Myrmeciinae, and in Nothomyrmecia macropsthe only living species of the primitive subfamily Nothomyrmeciinae. Only the species of a few genera, such as Oecophylla, Polyrhachis, and Dendromyrmex, as well as most Camponotus and certain socially parasitic ants have secondarily atrophied or completely lost the metapleural glands (Brown, 1968; Hölldobler and Engel-Siegel, 1984).

Design features of ant pheromones. It is not always the “purpose” of animal communication systems to maximize the information transmitted. In many cases, a simple yes-or-no signal is optional, for example, when nestmates are distinguished from aliens or workers broadcast a state of alarm. In others, such as the pinpointing of food discoveries by means of odor trails and waggle dances, the precision and hence the quantity of spatial information are at a premium. The optimal gain in transmission, in other words the number of group members contacted (Markl, 1985), also varies according to circumstance. Alarm signals are typically local, while caste-inhibitory signals are colony-wide.

Research on ant pheromones has revealed these and other design features of signals to be adaptations to the moment-by-moment needs of the colony. The theory of design is based on the concept of the active space, which is the zone within which the concentration of a pheromone (or any other behaviorally active chemical substance) is at or above threshold concentration (Bossert and Wilson, 1963; Wilson and Bossert, 1963). The active space is, in fact, the chemical signal itself. According to need, the space can be made large or small; it can reach its maximum radius quickly or slowly; and it can endure briefly or for a long period of time. These adjustments have been made in the course of evolution by altering the Q/K ratio, the ratio of the amount of pheromone emitted (Q) to the threshold concentration at which the receiving animal responds (K). Q is measured in number of molecules released in a burst, or in number of molecules emitted per unit of time, while K is measured in molecules per unit of volume. Where location of the signaling animal is relevant, the rate of information transfer can be increased either by lowering the emission rate (Q) or by raising the threshold concentration (K), or both. This adjustment achieves a shorter fade-out time and permits signals to be more sharply pinpointed in time and space by the receiver. A lower Q/K ratio characterizes both alarm and trail systems. The mathematical models based on diffusion and plume formation can be used to predict the form and duration of the active space or, conversely, either Q or K when the other parameter is known along with the elementary dimensions of the active space (Bossert and Wilson, 1963).

If part of the message is the location of the signaler, as it typically is in alarm, recruitment, and sexual communication, the information in each signal increases as the logarithm of the square of the distance over which the signal travels. In chemical systems it is the active space that must be expanded. An increase in active space can be achieved either by increasing Q or decreasing K. The latter is far more efficient, since K can be altered over many orders of magnitude by changes in the sensitivity of the chemoreceptors, while a comparable change in Q requires enormous increases or decreases in pheromone production as well as large changes in the capacity of the glandular reservoirs. The reduction of K has been especially prevalent in the evolution of trail systems and airborne sex pheromones, where threshold concentrations are sometimes on the order of only hundreds of molecules per cubic centimeter.

The duration of the signal can be shortened by an enzymatic deactivation of the molecules. When Johnston et al. (1965) traced the metabolism of radioactive queen substance, (E)-9-keto-2-decenoic acid, fed to worker honeybees, they found that within 72 hours more than 95 percent of the pheromone had been converted into inactive substances consisting principally of 9-ketodecanoic acid, 9-hydroxydecanoic acid, and (E)-9-hydroxy-2-decenoic acid. No comparable investigations have been conducted on the pheromones of ants, but they are likely to occur in systems requiring both a long reach and a rapid fade-out.

Communication can be enriched by variation in the response according to the concentration of the pheromone. In workers of the Florida harvester Pogonomyrmex badius, the principal alarm pheromone is 4-methyl-3-heptanone, which is stored in quantities of 0.2 to 34.0 µg (average: about 16 µg) in the mandibular gland reservoir (Vick et al., 1969; Nancy Lind, personal communication). Workers near the nest respond to threshold concentrations averaging 10^(10) molecules per cubic centimeter by moving toward the odor source; when a zone of concentration one or more orders of magnitude greater than this amount is reached, the ants switch into an aggressive alarm frenzy (Wilson, 1958a). The active space of the alarm can therefore be envisioned as a concentric pair of hemispheres. As the ant enters the outer zone it is attracted inward toward the point source; when it next crosses into the central hemisphere, it is excited into a frenzy. A very similar pattern of response to the same pheromone occurs in the leafcutter ant Atta texana and is illustrated in Figure 7-29.

The size of the pheromone molecules transmitted through air can be expected to conform to certain broad physical rules (Wilson and Bossert, 1963). In general, they should possess a carbon number between 5 and 20 and a molecular weight between 80 and 300. The a priori arguments that led to this inference are essentially as follows. Below the lower limit, only a relatively small number of molecules can be readily manufactured and stored by glandular tissue. Above it, molecular diversity increases very rapidly. In at least some insects, and for some homologous series of compounds, olfactory efficiency also increases steeply. As the upper limit is approached, molecular diversity becomes astronomical, so that further increase in molecular size confers no further advantage in this regard. The same consideration holds for intrinsic increases in stimulative efficiency, insofar as they are known to exist. On the debit side, large molecules are energetically more expensive to make and to transport across membranes, and they tend to be far less volatile. However, differences in the diffusion coefficient due to reasonable variation in molecular weight do not cause much change in the properties of the active space, contrary to what one might intuitively expect. The large number of ant pheromones identified to date conform to this rule of molecular size variation. Wilson and Bossert (1963) further predicted that alarm substances, which have no requirements for specificity and can be "read" by other species without harm to the sender, should have lower molecular weights than trail substances and other kinds of pheromones in which privacy is at a premium. The reason is that the smaller the molecule, the less likely it is to be unique. For example, there are vastly fewer variations possible on a 6-carbon alcohol than on a 12-carbon alcohol, or a 6-carbon alkane than on a 6-carbon nitrogen heterocycle. So far, this prediction has been vindicated in the Myrmicinae but not in the Formicinae. In the latter group, the alarm and trail substances overlap very broadly in their molecular weight, and they show no additional design features that conspicuously enhance or diminish their molecular specificity. The subject of molecular design in these substances remains a puzzle.

Because of the large numbers of species of ants and other social insects, and natural constraints on biosynthesis limiting molecular diversity far below the theoretical maximum, a considerable amount of convergence has occurred in pheromone chemistry. Examples of identical pheromones across species are given in Table 7-2. Since the insects listed are phylogenetically so remote from one another, every one of the pairings can be regarded as due to convergent evolution rather than to homology.

Some biochemical matches are nevertheless probably due to homology, with particular compounds having persisted over long periods of time through conservative biosynthesis and function. Possibly the most stable of all glands in this respect is the Dufour's gland, which often contains mixtures of terpenoid and straight-chain hydrocarbons that vary little from one genus to the next. Z,E-a-farnesene, for example, is the principal recruitment pheromone laid down in trails of the fire ant Solenopsis invicta, while two of its homofarnesene homologs serve as synergists (Vander Meer, 1986a,b). Myrmica lobicornis and Myrmica scabrinodis also produce Z,E-a-farnesene and homofarnesenes in their Dufour's glands, but these substances do not function as trail pheromones and their role remains unknown (Attygalle et al., 1983). Instead, the trail pheromone of Myrmica, produced in the poison gland, is 3-ethyl-2,5-dimethylpyrazine; it is also a poison gland product and a trail pheromone of Tetramorium caespitum and two species of Atta (see Table 7-5). Finally, the Dufour's gland contents of the large, primitive dacetine Daceton armigerum are followed by Solenopsis invicta (Wilson, 1962a), while the poison gland contents are followed by species of Acromyrmex and Atta (Blum and Portocarrero, 1966), which suggests that theDaceton armigerum Dufour's gland contains the farnesene and its poison gland the pyrrole, one of the trail pheromones identified in attine ants (Tumlinson et al., 1971, 1972). The comparative biochemistry of ant exocrine glands and their primitive and derived functions are fascinating but still relatively unexplored subjects.

Efficiency of semiochemicals. Possibly the chief advantage of semiochemicals over signals in other sensory modalities is the extreme economy of their manufacture and transmission. The sensory apparatus has evolved in some cases to respond to particular substances at a virtually quantal level, with only a few molecules striking the receptive membranes in each antennal sensillum every few seconds. The process is abetted by the existence of isomerism, in which relatively minor differences in the configuration of the same molecule generate new physical or chemical properties that are discernible by the ants. The most extreme form is optical isomerism, the existence of pairs of chemical compounds (enantiomorphs) whose molecules are nonsuperimposable mirror images. One configuration is capable of rotating plane-polarized light to the right, constituting the dextro or (+) form, and the other to the left, constituting the levo or (-) form. In leafcutter ants of the genus Atta, workers are 100-200X more sensitive to the natural, (+) enantiomer of 4-methyl-3-heptanone, an alarm pheromone, than they are to its (-) enantiomer (Riley et al., 1974a). Pogonomyrmex harvester ants are similarly more sensitive to the (+) enantiomer (Benthuysen and Blum, 1974).

A principal consequence of such acute sensitivity is the minute amounts of the pheromones needed at any given time. The extreme cases recorded thus far occur in the trail substances. The amounts of methyl 4-methylpyrrole-2-carboxylate found in each worker of the leafcutters Acromyrmex and Atta range according to species from 0.3 ng to 3.3 ng (Evershed and Morgan, 1983). Workers of Myrmica rubra contain 5.8 ± 1.7 ng of the trail substance 3-ethyl-2,5-dimethylpyrazine (Evershed et al., 1982). Even such trace amounts, while wholly undetectable to human beings without the aid of elaborate instrumentation, are sufficient to convey complete messages among ants. Tumlinson et al. (1971), the discoverers of methyl 4-methylpyrrole-2-carboxylate as the trail substance of Atta texana, estimated that one milligram of this substance (roughly the quantity in a single colony), if laid out with maximum efficiency, would be enough to lead a column of ants three times around the world.

The chief disadvantage of such chemical systems is the slowness of fade-out. When using pheromones alone, ants are not able to transmit a rapid sequence of signals in the manner of vocalizations or quickly changing visual signals. In order to replace signals, they must wait until the active space of the pheromones expands to maximum diameter, then shrinks back to the point of emission or is blown away by air currents. In many cases this property has been turned to the advantage of the insects. A long-standing active space is needed, for example, in the employment of colony odors and caste-identification substances, alarm pheromones, and trail substances. It is also possible to create sequential and compound messages either by a graded reaction to different concentrations of the same substance, as illustrated in the case of the Atta texana alarm system (see Figure 7-29), or by blends of signals. Let us now consider the latter, very interesting elaboration in some detail.

Pheromone blends. The following rule has emerged from surveys of the natural product chemistry of ants: individual exocrine glands usually produce mixtures of substances, which are moreover often complex in both constitution and function. A typical example is provided by the subterranean “citronella ant” Acanthomyops claviger of the eastern United States, depicted in Figure 7-30. The highly modified poison gland, typical of the Formicinae, appears to produce only formic acid, used in defense. But the multiple terpenoid aldehydes and alcohols of the hypertrophied mandibular glands serve in both defense and alarm. Among the homologous alkanes and ketones of the Dufour's gland, undecane is an alarm pheromone while the remaining components serve mostly or entirely in defense.

The identification of components has out-paced an understanding of their function, and differences among closely related species of ants have compounded the mystery. For example, the sibling species Tetramorium caespitum and Tetramorium impurum of Europe can be identified morphologically only by differences in the male genitalia, but their Dufour's gland secretions are quite distinct. The gland in Tetramorium caespitum is of moderate size and contains about 70 ng of C13 to C17 linear hydrocarbons together with a mixture of pentadecenes as major components; while that of Tetramorium impurum is smaller, containing 40 ng of the same mixture but with a prevalence of n-pentadecane and a sesquiterpenoid compound. Tetramorium semilaeve, a species anatomically more distinct from caespitum and impurum, has a still smaller gland (30 ng capacity) with a simpler mixture of hydrocarbons and the pentadecane present in much higher proportion (Billen et al., 1986). The functional significance of the mixtures and the differences discovered among these Tetramorium species remains unknown.

Only in a few ant species has the significance of chemical blending been clarified. The functions fall into one or the other of two categories. Either an increase in specificity allows one species to distinguish its own pheromones from those of others, or the production of multiple simultaneous signals allows the transmission of more complex messages.

An example of the first role, the promotion of privacy during communication, is provided by the trail substances of leafcutter ants. All of the species of Atta and Acromyrmex analyzed so far either produce methyl 4-methylpyrrole-2-carboxylate in their poison glands, or react to this substance, or both. Nevertheless, Acromyrmex octospinosus actively avoids trails of Atta cephalotes an effect that appears to be due to components that occur in blends with the pyrrole (Blum et al., 1964; Blum, 1982). This hypothesis has been verified in the case of Atta sexdens, which possesses the pyrrole (in addition to some minor components) but utilizes yet another substance, 3-ethyl-2,5-dimethylpyrazine, as its major trail pheromone (Cross et al., 1979).

In a closely parallel manner, the fire ant Solenopsis invicta relies primarily on Z,E-a-farnesene as its recruitment trail pheromone, with two homofarnesenes acting as synergists. All three substances are emitted from the Dufour's gland. It also produces two isomeric tricyclic homosesquiterpenes in small quantities. The closely related Solenopsis richteri, in contrast, relies entirely on the tricyclic homosesquiterpenes producing neither alpha-farnesenes nor homofarnesenes. Yet each species responds weakly to the trails of the other. The reason is that Solenopsis invicta is sensitive to some extent to its own tricyclic homosesquiterpenes, and it produces enough (over 50 femtograms per gland) to activate Solenopsis richteri (Vander Meer, 1986a,b).

A more precise delineation of the effects of pheromone blends has been made in Tetramorium caespitum by Attygalle and Morgan (1983). This myrmicine lays trails comprising two pyrazines, designated as VIII and X. As shown in Figure 7-31, workers respond maximally to a blend with a ratio of 3:7 of VIII to X. However, pheromone blending of this kind has not always arisen in evolution, even in the case of pyrazines. In eight species of Myrmica, the trail consists only of component VIII, which has been identified as 3-ethyl-2,5-dimethylpyrazine.

To summarize research on specificity, it is clear that related species, for example those similar enough to be placed within the same genus, have diversified repeatedly during evolution by creating variable mixtures of pheromones in the same exocrine gland. Some manufacture and rely wholly on single components, while others generate combinations that can be easily shifted during evolution to create optimal mixes peculiar to individual species. One important result is the enhanced privacy of communication within species, a feature that is of clear adaptive significance at least in the case of trail systems.

In a wholly different dimension, precision of communication has been improved by evolving mixtures of pheromones with different effects. When laying odor trails, workers of the imported fire ant discharge a medley of substances from the Dufour's gland, some of which are illustrated in Figure 7-32. The principal component for recruitment is Z,E-a-farnesene (I), which nevertheless requires the two homofarnesenes (III, IV) and a still unidentified component in order to attain the full activity observed from complete Dufour's gland extracts. Oddly, these substances remain inactive unless the ants have been primed by yet another, still unidentified component of the gland. Once the ants have encountered this unknown pheromone, they respond fully to artificial trails made entirely from the alpha-farnesene and homofarnesenes (Vander Meer, 1983, 1986a). A similar but less fully investigated Dufour's gland system has been discovered in the household ant Monomorium pharaonis. The principal component is faranal, or (6E,10Z)-3,4,7,11-tetramethyl-6,10-tetradecadienal. Several nitrogen heterocycles serve as supplementary attractants. Moreover, indolizines and pyrrolidines from the poison gland are attractive to the workers and may play a special role of their own, although their presence in odor trails remains to be proven (Ritter et al., 1977a,b).

The partitioning of foraging areas among sympatric species of the harvester ant genus Pogonomyrmex illustrates the involvement of both anonymous and specific semiochemicals in inter- and intraspecific territoriality. The relatively short-lived recruitment signal from the poison gland is, so far as known, invariant among Pogonomyrmex species. In addition to these anonymous recruitment trails, persistent trunk routes are established by clearing vegetation and marking with Dufour's gland secretions, which contain species-specific mixtures of hydrocarbons (Regnier et al., 1973; Hölldobler, 1976, 1986). The trunk routes also contain colony-specific chemical markers which, together with species-specific cues from the Dufour's gland, serve to channel the foragers of neighboring nests in diverging directions, effectively partitioning limited food resources. These examples illustrate that not all constituents of chemical communication signals need have the same functional significance. In many cases, one or several components act as key stimuli, triggering a basic anonymous response, while additional components add specificity (Hölldobler and Carlin, 1987). Undecane, for example, is apparently the active alarm signal in most ant species of the subfamily Formicinae, and it is also usually the most abundant product in formicine Dufour's glands. However, other hydrocarbons are also present, and the total mixture is often species-specific (Morgan, 1984). In Oecophylla longinoda, further specificity is added by droplets originating from the rectal bladder, which are used in colony specific territorial marking (Hölldobler and Wilson, 1978). Because rectal marking alters the probability of winning territorial conflicts (ants are more aggressive on ground that they have previously marked, and less so on ground marked by another colony) it fulfills the criteria for a modulator of the alarm response (Markl, 1985). To date the rigorous investigation of modulatory communication signals, defined as those which do not themselves release behavioral responses but which influence reactions to other signals, has been limited to cases in which one signal modulates another of a different modality (Markl, 1983, 1985; Hölldobler, 1984). However, different elements of cues in a single modality can also interact in this fashion; thus the paradigm also applies to multicomponent semiochemicals in which the additional information of specificity may be seen as modulating the response to anonymous chemical releasers (Hölldobler and Carlin, 1987).

If specificity is considered as a form of modulation, and modulatory functions presuppose the existence of the behavior being modulated, a possible evolutionary route to signal specificity can be envisaged. Hölldobler and Carlin (1987) argued that the production of simple semiochemicals releasing elementary anonymous reactions is subject to the inevitable impression of all biosynthetic processes. The resulting degree of variation may well be perceptible to the receivers' sensory system, but will have no effect on the response to the signal. However, should an adaptive advantage happen to correlate with any of the available variants, selection will favor individuals that respond differentially on the basis of these specific characteristics, in other words, that modulate the original response. For example, other Dufour's gland hydrocarbons will be released along with undecane. If, say, genetically similar colony members tend to produce similar hydrocarbon patterns, the signal may come to be modulated by this added specificity, informing workers whether nestmates or aliens are sending the alarm signal. Once the presence or proportions of additional components significantly affect the response to the basic releaser in an adaptive manner, selection can be expected to improve their distinctiveness and stereotypy. In fact, the exploration of variation in communication among colonies of the same species should prove fruitful in the future. Cherix (1983) found that two adjacent colonies of Formica lugubris in the Swiss Jura mountains possessed both qualitative and quantitative differences in their pheromones, including the presence or absence and proportionality of undecane, tridecane, and nonadecanol. Such within-species variation might come about as a result of genetic differences, or the succession of stages in colony growth, or previously unsuspected factors in nest environment.

Of equal importance, multiple pheromones permit the spread of different messages across space, especially when the mixtures are released from a single point. This paradoxical effect is made possible by the fact that chemical substances produce different active spaces. For example, if pheromone A is produced in larger quantities than pheromone B or is behaviorally more active, it will generate a larger hemispherical space that encompasses the similarly shaped active space of pheromone B. As the receiver ant approaches the point source, it first receives signal A and then signal B, responding in a sequence of actions.

Two cases of this interesting phenomenon are illustrated in Figure 7-33. Workers of the African myrmicine Myrmicaria eumenoides each deposit a single droplet of poison gland secretion near potential prey items. Unlike the venom of most other myrmicine ants, which are proteinaceous toxins, the gland contents of Myrmicaria eumenoides are monoterpene hydrocarbons, including b-myrcene, b-pinene, and limonene. The b-pinene generates the larger active space, which causes nestmates to move toward the droplet. At closer range, the limonene induces circling behavior, which deploys the workers around the prey so that they approach from many directions during the attack itself (Bradshaw and Howse, 1984). In the African weaver ant Oecophylla longinoda the multiple components of the mandibular gland secretion trigger a stepwise escalation of responses as the ants approach an enemy. In the outermost space, hexanal alerts the workers. Then 1-hexanol attracts them, and finally 3-undecanone and 2-butyl-2-octenal induces them to attack and bite any alien object in the vicinity (Bradshaw et al., 1975, 1979).

In addition to such multicomponent pheromones, multisource systems are commonplace in the ants. In such systems various compounds are released from multiple glandular sources. The substances may serve the same essential functions, as in the case of the alarm pheromones of Acanthomyops claviger (see Figure 7-30), but often the roles are different. In the Florida harvester ant Pogonomyrmex badius, for example, the recruitment pheromones are voided from the poison glands, while the homing pheromones originate at least in part in the Dufour's gland (Hölldobler and Wilson, 1970). Workers of the primitive Australian ant Myrmecia gulosa induce territorial alarm behavior in toto by pheromones from three sources: an alerting substance from the rectal sac, an activating pheromone from the Dufour's gland, and an attack pheromone from the mandibular glands (Robertson, 1971).

In what may be the ultimate evolutionary development, communication can be part of multimodal systems, which transmit signals through more than one sensory modality. The species with the most elaborate organization discovered thus far is Oecophylla longinoda, in which four of five recruitment systems incorporate pheromones (from the anal and sternal glands) together with specialized tactile signals. This example will be examined in some detail in the section now to follow, on ritualization.

Ritualization
The vast majority of cases of the origin of communicative systems in animals is based on ritualization, the evolutionary process by which a phenotypic trait is altered to serve more efficiently as a signal. Commonly, the process begins when some movement, anatomical feature, or physiological process that is functional in another context acquires a secondary value as a signal. For example, members of a species might recognize the opening of mandibles or the release of an odor as a threat. Alternatively they can interpret the turning away of an opponent's body in the midst of conflict as an intention to flee. During ritualization such movements (or odors, or visual features) are altered in a way that makes their communicative function still more effective. They acquire support in the form of additional anatomic structures or biochemical changes that enhance the distinctiveness of the signal. The movements also tend to become stereotyped and exaggerated in form. Finally, the receiving apparatus is modified to detect such ritualized signals with less ambiguity. In the case of trail systems of ants, the chemoreceptors have been modified to detect minute traces of the appropriate pheromone, which often occur in nanogram or even femtogram amounts.

The classic example of ritualization in the behavior of social insects is the waggle dance of the honeybee. The dance, first “decoded” by Karl von Frisch in 1945, is easy to understand if one thinks of it as a ritual flight, a scaled-down version of the journey from the nest to the food source. The essential element in the maneuver is the straight run, the middle piece in the figure-eight pattern. (The remainder of the figure-eight consists of a doubling back to repeat the straight run.) The dancing bee has just returned from several back-and-forth journeys to the target. The straight run it performs on the vertical surface of the comb is a miniaturized version of the outward flight that it now invites its nestmates to undertake. The angle between the straight run and the vertical line of the comb surface (in other words the line pointing straight up) indicates the direction of the target relative to the position of the sun. The duration of the straight run indicates the distance to the source: the longer the straight run takes to complete, the farther away the target. The straight run is rendered more conspicuous by a rapid waggling motion of the body, a typical case of an enhancing embellishment during ritualization. The dancing bees also produce a distinctive sound.

Relatively few communicative systems in ants have been analyzed explicitly with reference to their evolutionary origin, but suggestive evidence of ritualization has been adduced. One clear use is the invitational movements of Camponotus workers recruiting nestmates to new nest sites. Workers of Camponotus sericeus jerk their bodies back and forth vigorously in front of other workers, then seize them by the mandibles and pull them forward a short distance (see Figures 7-56 and 7-57). Those of Camponotus socius have taken the next step. They employ body jerking alone, evidently having entirely deleted the rudiments of physical transport during the early stages of recruitment. Food offering is also highly ritualized in Camponotus socius. After filling her crop with liquid food and returning to the nest, the scout shakes her body from side to side with her mandibles wide open, allowing nestmates to scan the lower mouthparts and odor of the recently ingested food (Hölldobler, 1971c; Hölldobler et al., 1974; see Figure 7-53). The ponerine Bothroponera tesseronoda of Sri Lanka, representing a separate evolutionary development, uses mandible pulling as a signal to initiate tandem running both to new nest sites and food finds (Maschwitz et al., 1974).

Additional evidence of ritualization can be found in the multiple recruitment systems of the African weaver ant Oecophylla longinoda, which are the most complex form of communication thus far discovered in the ants. Workers of this species, which construct arboreal nests in part from larval silk, utilize no fewer than five recruitment systems to draw nestmates from the nests to the remainder of the nest tree and to the foraging areas beyond (Hölldobler and Wilson, 1978; see Plate 5). These include (1) recruitment to new food sources, under the stimulus of odor trails produced by the scout from its rectal gland, together with tactile stimuli presented while the scout engages in mandible gaping, antennation, and head-waving; (2) recruitment to new terrain, employing pheromones from the rectal gland and tactile stimulation by antennal play; (3) emigration to new sites under the guidance of rectal gland trails; (4) short-range recruitment to territorial intruders, during which the terminal abdominal sternite is maximally exposed and dragged for short distances over the ground to release an attractant from the sternal gland; and (5) long-range recruitment to intruders, mediated by odor trails from the rectal gland and by antennation and intense body jerking. These systems exist in addition to the elaborate pheromone-mediated alarm communication described by Bradshaw et al. (1975).

The organization of the five recruitment systems is summarized in Table 7-3, and some of the behavior isillustrated in Figures 7-34 and 7-35. The forward jerking movement used during recruitment to enemies closely resembles maneuvering during the actual attack maneuvers themselves, and we have therefore interpreted the signals to be a ritualized version “liberated” during evolution to serve as a signal when a nestmate is encountered rather than an enemy. When workers recruit nestmates to food, they use a wholly different set of movements. They wave their heads laterally while opening their mandibles. The effect is evidently to waft food odors from the lower mouthparts to the antennae of the potential recipient.

Ritualization is not limited to tactile signaling. Chemical alarm communication evidently evolved from chemical defense behavior. Like many solitary insects, ants and other social insects use chemical secretions to repel predators and other enemies. In social insects, however, defensive reactions are closely linked with alarm communication, and quite often a single substance serves both functions. A well-documented example is citronellal, a mandibular gland product of Acanthomyops claviger (see Figure 7-30).

Acanthomyops and other formicine ants use hindgut contents as trail pheromones (Hangartner, 1969c), a procedure that might have evolved as a gradual ritualization of the defecation process. The final development, exemplified by the extraordinary rectal gland of the Oecophylla weaver ants, is the origin of a wholly new structure to generate the trail substances. In fact, Oecophylla workers employ the hindgut in two ways that could have evolved from defecation. The second application is the use of fecal material directly in territorial marking. The ants deposit fecal droplets more or less uniformly over the surface of the vegetation around their nests, rather than in refuse piles or other special zones. The droplets contain substances that are specific to their colony, and they permit the ants to determine from moment to moment whether they are in the vicinity of their own nests or on foreign terrain (Hölldobler and Wilson, 1978).

Signal economy and “syntax”
For two reasons ants can be intuitively expected to practice economy in the evolution of their communication systems. By this is meant the employment of a small number of signals relatively simple in execution and derived from a limited number of ancestral structures and movements. First, the small brain and short life span of ant workers limit the amount of information these insects can process and store. Second, the tendency toward signal evolution through ritualization restricts the range of potential evolutionary pathways.