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.

The five recruitment systems of Oecophylla longinoda just described illustrate signal economy in a striking manner. Although the messages differ from one another strongly (see Table 7-3), they are built out of pheromones from just two organs, the rectal and sternal glands, together with a modest array of stereotyped movements and tactile stimuli. The specificity of each of the recruitment systems comes mostly from the combinations of the chemical and tactile elements. There is a primitive sort of syntax in this differentiation. For example, both recruitment to food and recruitment to new terrain are stimulated and guided by pheromones from the rectal gland. But food is further specified by head waving and mandible opening, while fresh terrain is specified (occasionally) by body jerking. There is no proof that the ants themselves perceive the difference, but it seems unlikely that such stereotyped movements would have been incorporated into the behavior if it was wholly lacking in function. The parallel nature of the recruitment systems is accompanied by a lack of any clear distinction on the part of the Oecophylla workers between colony defense and predation. When defenders vanquish invading ants, they remove them to the nest interior and convert them into food. The full sequence of responses does not appear to differ significantly from that following the encounter of ordinary prey insects within the territorial boundaries, although the response to conspecific invaders is considerably more massive.

A comparable degree of parsimony is widespread among behavioral categories of the most diverse kinds. Workers of Formica subintegra, a north American slavemaking ant, spray acetates from the Dufour's gland when they invade the nests of other Formica species. These substances serve to attract the Formica subintegra workers, but act to alarm the workers being assaulted, which succumb to panic and disperse (Regnier and Wilson, 1971). Workers of the fire ant Solenopsis invicta disperse venom as an aerosol by forming a droplet on the tips of their extruded stings and vibrating (“flagging”) the abdomen vertically. They perform this unusual act under two radically different circumstances, using different quantities of venom. When repelling other species of ants from the foraging arena, each worker dispenses up to 500 ng of the substance. Inside the nest, brood attendants dispense about 1 ng over the surface of the brood, possibly as an antibiotic. Defensive flagging of the first kind is distinguished by either a vertical orientation of the entire body with the head pointing downward--the “head stand”--or a horizontal orientation with the sting pointed at the intruder. Brood flagging is accomplished with a mostly horizontal orientation (Obin and Vander Meer, 1985). In yet another category of behavior, some species of Leptothorax recruit nestmates initially by “tandem calling,” in which the worker slants its abdomen upward and discharges poison gland secretions from the extruded sting. Nestmates are attracted, and as soon as one of them touches the calling ant, tandem running begins. Then the recruiter leads the nestmate to the target area (Möglich et al., 1974; Figure 7-36). An apparently identical calling behavior is employed by the reproductive workers of the slavemaker Harpagoxenus, a phylogenetic derivative of Leptothorax, in order to attract males (Buschinger, 1968; Buschinger and Alloway, 1979; phylogeny discussed by Buschinger, 1981). Finally, as we will show subsequently in describing trophallaxis, ponerine ants employ closely similar antennal signals in social greeting, recruitment invitation, and food solicitation (see page 000).

Modulatory communication
Communication in complex social systems is seldom characterized by a direct, all-or-none response. Signals do not always merely “release” behavioral responses of a particular kind, but instead often appear to adjust the behavior of nestmates towards one another in a manner appropriate to the surrounding environment. In such instances the communication may have relatively low informational content when measured simply by the number of bits transferred. Instead, it causes each colony member to tune its behavior in a flexible manner to its immediate circumstances. According to this interpretation, advanced by Markl and Hölldobler (1978) for social insects and subsequently elaborated in later studies by Markl (1983, 1985) and Hölldobler (1984c), outwardly inefficient communication systems serve different but no less important purposes than more direct, deterministic systems. They influence the behavior of receivers, not by forcing them into narrowly defined behavioral channels but by slightly shifting the probability of the performance of other behavioral acts. Put another way, the signals of such “modulatory communication” do not release specific fixed-action patterns in the familiar manner envisaged by ethologists, but instead alter the probability of reactions to other stimuli by influencing the motivational state of the receiver. Modulatory communication can be expected to be more frequent in the most complex animal societies, where a large number of members perform many different tasks at the same time. In this arrangement, a flexible program is required if the work force is to distribute its energy investment among the different tasks in an effective manner.

The best analyzed case of modulatory communication in ants is the case of stridulation in the enhancement of short-range recruitment by workers of Aphaenogaster albisetosa and Aphaenogaster cockerelli (Hölldobler et al., 1978; Markl and Hölldobler, 1978). These gracefully slender ants of the American desert are adept at retrieving large prey objects, such as dead insects, within short periods of time. After discovering an object too large to be carried or dragged by a single ant, a scout worker releases poison gland secretion into the air. Nestmates as far away as two meters are attracted and move toward the source. When a sufficient number of foragers has assembled around the prey, they gang-carry it swiftly to the nest (Figure 7-37). Time is of the essence, because the Aphaenogaster must remove food from the scene before formidable but less agile ant competitors, including fire ants and Iridomyrmex pruinosum, arrive in large numbers on the scene. Aphaenogaster workers, in addition to releasing the poison gland pheromone, also regularly stridulate by rubbing the sharp posterior edge of the postpetiolar tergite (scraper) against a file of horizontally arrayed ridges on the anterior end of the first gastric tergite. The chirping sounds generated last up to 200 msec each with frequencies between 0.1 and 10 kHz. Once the foragers encounter the vibration, they remain in the vicinity for up to twice as long as when no stridulation occurs. Ants perceiving the signals also start to encircle the prey sooner, and they are likely to release the attractive poison gland pheromone earlier. Overall, both the recruitment of workers and the retrieval of the food object are advanced by 1 to 2 minutes as a consequence of stridulation. The vibration thus serves as a communication amplifier in this particular circumstance, conferring a considerable advantage to the ants, which must race to acquire food in the highly competitive desert environment.

A similar enhancement of recruitment by stridulation has been reported in the European harvesting ant Messor rufitarsis, which uses the sound in conjunction with odor trails from its Dufour's gland (Hahn and Maschwitz, 1985). Other studies suggest the phenomenon may be widespread in Messor (Schilliger and Baroni Urbani, 1985; Buser et al., 1987). In the tropical Asiatic predatory ant Leptogenys chinensis stridulation is combined with pygidial gland secretions during colony emigration (Maschwitz and Schönegge, 1983). A second category of modulatory communication is drumming in the European carpenter ants Camponotus herculeanus and Camponotus ligniperda (Fuchs, 1976a,b). Workers strike the surface of the wooden chambers and galleries in which they live with their mandibles and gasters, producing vibrations that can be perceived by nestmates for 20 cm or more. Much of the behavior is classifiable as direct alarm communication, which we will describe shortly. But Fuchs showed that the drumming alters other behavioral responses as well, in a manner that becomes clear only with the aid of statistical analysis of transition probabilities between the different responses. To put the matter in a phrase, the behavior of some categories is “tightened up.” Transition probabilities are raised, and hence uncertainty reduced, when the ant is in particular initial states when the signal is received. The particular initial states include antennal waving (probably a condition of monitoring the environment), grooming, and running. No effect was observed by Fuchs when the Camponotus workers were either inactive or feeding. The meaning of the modulatory changes remains to be elucidated.

Modulatory communication appears to be a primitive phenomenon in ants and other social insects. Other communicative motor patterns in ants, including the abbreviated fast runs and body jerking employed during recruitment, might have evolved from early forms that merely modulated the behavior of nestmates in other categories. Some of them were then ritualized into specialized signals with direct effects of their own, usually working in combination with other signals--such as trail or alarm pheromones.

In the case of chemical signaling, modulatory communication resembles the primer effects of pheromones. One of the best documented examples of primer effects in ants and other social insects is caste inhibition, in which the detection of substances secreted by one caste (such as a queen or soldier) induces a larva or nymph to mature into a different caste. Primer pheromones also inhibit wing shedding in queens and ovarian development in both queens and workers. In both of these categories, the pheromones have a profound influence on behavior and communication in the receptor individuals. However, the time scale is far greater than that of ordinary modulatory communication. It spans a large part of the life cycle rather than just seconds or minutes.

Synergism
We have shown how some signals can modulate the response of ants to others in unexpected ways. Many cases have also been discovered of multiple signals, sometimes transmitted through two or more sensory modalities, that operate in concert to evoke complex responses. A third, related phenomenon is true synergism, in which two signals have the same or a closely similar effect but a combination of the two causes a stronger reaction than comparable magnitudes of either one presented singly. An apparent example of this phenomenon has been discovered by Cammaerts et al. (1982) in the European species of the ant genus Myrmica. The mandibular glands of the workers contain mixtures of low-molecular-weight alcohols and ketones that serve as alarm pheromones, inducing nestmates to increase the linear speed and decrease the sinuosity (angular velocity) of their running. In Myrmica schencki a mixture of 3-octanol and 3-octanone evoked a stronger response when presented in natural concentrations than did either component alone. The difference is relatively slight, however, and its meaning not yet clear. Species of the arboreal dolichoderine genus Azteca produce mixtures of three cyclopentyl ketones in their pygidial glands. Each component causes some alarm response, but a mix of all three is far more effective than any one pheromone alone. In addition, some species add 2-heptanone, another powerful pheromone. Other Azteca use the heptanone exclusively, and a few rely entirely on the cyclopentyl ketones (Wheeler et al., 1975). In general, few efforts have been made to test for synergism in the communicative systems of ants. The effort might prove profitable, once the complete compositions of more pheromone complexes are known.

Caste and colony differences in communication
A large part of the ant social organization is based on differences among the various castes and subcastes in patterns of communication. The Brazilian myrmicine Pheidole embolopyx is typical in this regard. The minor and major workers, illustrated in Figure 7-38, display different responses during recruitment and defense of the nest and food sources. The caste-specific behaviors are nevertheless coordinated to create efficient reactions at the level of the colony. Only the minor workers lay odor trails, which originate from the poison gland and are used to direct the remainder of the colony to new nest sites and food finds. Both castes cooperate in defending food objects too large to transport to the colony, by forming clusters that encircle these bonanzas for periods of hours or even days. The minor workers protect the food from intruder ants by seizing their legs and pinning them down, while the majors attack their bodies directly. When the nest is disturbed, both castes communicate alarm by means of abdominal pheromones, which in the case of the major worker has been traced to the pygidial gland. The queen behaves in still another, radically different manner. She can be recruited with odor trails only when the colony is emigrating to new nest sites. While the colony is in a state of alarm, she relies for personal defense on her uniquely turtle-like body form, which includes a truncated abdomen and paired flange-like protrusions of the pronotum and first gastric segments. She crouches tightly into small cavities in a way that exposes a minimum of vulnerable body surface. The defense is enhanced by gelatinous sheaths secreted from enlarged cephalic glands onto the scapes and anterior portions of the head (Wilson and Hölldobler, 1985).

The differentiation in communicative behavior among ant castes is often underlain by strong biochemical differences. The soldier of Pheidole fallax produces skatole in a hypertrophied poison gland that fills about one-third of the entire abdominal cavity. This substance is used in defense and possibly also alarm communication. As in Pheidole embolopyx, the majors follow odor trails but do not lay them. The trails are voided instead by the minors from their poison glands. Conversely, these ants do not produce skatole (Law et al., 1965). In a third Pheidole species, Pheidole biconstricta, the minor workers produce large quantities of odorous secretion from greatly enlarged pygidial glands. This material, which appears peculiar to the caste, is used in both alarm and defense (Kugler, 1979). Queen and male ants are well known to synthesize an array of distinctive pheromones used during mating and, in the case of the queens, for the attraction of workers (see Chapters 3 and 6).

Acoustical communication
The use of vibrational signals is weakly developed in ants in comparison with communication by pheromones. It often occurs in conjunction with chemical signals. Most but probably not all acoustical signals are transmitted primarily through the soil, nest wall, or some other solid substratum rather than through the air. Two forms of sound production have been identified, body rapping against the substratum and stridulation, the latter employing files and scrapers clearly evolved for a communicative purpose. According to species, one or the other of four functions is served by sound signals in ants: alarm, recruitment, termination of mating by females, and as noted previously, the modulation of other communication and forms of behavior.

The production of sound signals by body rapping or “drumming” in social insects occurs most commonly in colonies that occupy wooden or carton nests, where substrate vibrations are transmitted with high efficiency in comparison with otherwise similar nest structures in soil. It is widespread among the arboreal species of the dolichoderine genera Dolichoderus and Hypoclinea and the formicine genera Camponotus and Polyrhachis (Markl, 1973). Workers of the carpenter ants Camponotus herculeanus and Camponotus ligniperda, the best-studied species to date (Markl and Fuchs, 1972; Fuchs, 1976a,b), can be launched into drumming by any moderate disturbance to their nests, including air currents (a sign that the nest has been breached), sound, touch, or chemical contaminants. The drumming ant strikes the substrate with its mandibles and gaster while rocking its entire body violently back and forth. Up to seven such strikes are delivered at 50 msec intervals. The signaling pattern is independent of the triggering stimulus. That is, the ants do not modify the drumming to identify the category of danger to the nest. The signals have a maximum energy content at 4 to 5 kHz and attenuate at the rate of 2 dB/cm. They carry through the thin wooden shells of the nest for distances of several decimeters or more. As we noted earlier, drumming is a modulatory signal, altering the transition probabilities of other kinds of behavior. The vibration is also an effective alarm signal. The response of workers in or close to the nest depends on their initial level of activity, as measured by their running speed. Those less active “freeze” into immobility, while those more active do the exact opposite--they increase their running speed, move toward the source of the vibrations, and attack any moving object in the vicinity. The acoustical signals are also modulatory in nature, in that they alter the transition probability from antennal waving, grooming, and running to other behavioral states. It is not yet known whether this latter alternation also serves alarm and defense or is merely “noise” in the social system.

Stridulation as opposed to drumming is a more sophisticated sound-producing behavior found in the ant subfamilies Ponerinae, Nothomyrmeciinae, Pseudomyrmecinae, and Myrmicinae. Entomologists define it generally as the rubbing of specialized body parts together to produce a “chirp.” The sound has been characterized by Broughton (1963) as “the shortest unitary rhythm-element that can be readily distinguished as such by the unaided human ear.”  In almost all ant species studied to date the chirps are specifically produced by raising and lowering the gaster (the most posterior discrete section of the abdomen) in such a way as to cause a dense row of fine ridges (the “file”) located on the middle portion of the first segment of the gaster, i.e., fourth abdominal segment, to rub against a scraper (“plectrum”) situated near the border of the preceding, third abdominal segment (Haskins and Enzmann, 1938; Forrest, 1963; Markl, 1968; Schilliger and Baroni Urbani, 1985). The basic details are illustrated in Figures 7-39 and 7-40. Two exceptions to the position of the stridulatory organ are known. The first is provided by the Australian ant Nothomyrmecia macrops (Taylor, 1978). The workers of this very primitive species bear a scraper on the third abdominal segment and a file on the fourth abdominal segment. This location is the same as in other ants, except that in Nothomyrmecia the scraper and file are on the ventral segment of the abdomen instead of the dorsal. The other exception occurs in the ponerine genus Rhytidoponera, which possesses a dorsal and ventral file on the fourth abdominal segment (Hubert Markl, personal communication). The arrangement is unique within the Hymenoptera as a whole, where the file is otherwise invariably on the fourth or fifth tergite (Brothers, 1975).

Spangler (1967) gave a characterization of stridulation in the harvester ant Pogonomyrmex occidentalis. As the gaster is pulled down, a relatively weak sound is produced that lasts about 100 msec and generates its principal energy at 1-4 kHz. The gaster is then jerked back up, producing a second and similar sound, also lasting about 100 msec but overall louder and containing in addition some higher frequencies (7-9 kHz). All of the instrumentally detected frequencies are within the range of human hearing, with the result that the stridulating Pogonomyrmex worker, when held close to the ear, seems to squeak faintly and almost continuously. In his independent studies of Atta cephalotes, Markl (1965-1973) found that the chirp produced differs in quality from that of Pogonomyrmex. It is much louder, as much as 75 dB at 0.5 cm from a major worker of Atta cephalotes, whereas the intensity is less than 2 dB in Pogonomyrmex occidentalis. It is also higher pitched. In a more recent study Masters et al. (1983), employing laser-Doppler vibrometry, determined that the gaster, on which the file is located, appears to be the principal sound radiating part of the ant. In their analysis of Atta sexdens, these investigators divided the energy of the radiated airborne sound into three frequency regions, the first two corresponding to low frequency (LF) and high frequency (HF) peaks of the vibration spectrum, and the third extending above about 20 or 30 kHz. They state: “In the lower two regions energy in the sound spectrum can be interpreted rather directly in terms of the measured vibration. Energy in the range 0.5 to 1.5 Hz can be attributed to the LF oscillation of the tooth impact rate. This low-frequency component has been found in other stridulatory insects and may be an important carrier of information. . . . In the range from about 5 to 20 or 30 kHz the energy comes mainly from HF oscillation. Generally the vibration signal peak at about 10 kHz appears at a slightly higher frequency in the airborne sound pressure spectrum. The shift to higher frequencies is attributable to the increasing radiation efficiency of the gaster as frequency rises. Radiation efficiency reaches a plateau value at a wavelength of 2πa, where a is the gaster radius, representing a frequency of about 40 kHz for a typical ant, and above this frequency radiation is about equally efficient at all frequencies, although of course the amplitude of vibration will be small at such high frequencies.”

It has long been known, thanks to the experiments of Fielde and Parker (1904) and Haskins and Enzmann (1938), that ants are nearly deaf to airborne vibrations but extremely sensitive to vibrations carried through the substratum--sensitive enough, in fact, to detect stridulation sounds even through well-packed soil. Markl found that workers of leafcutter ants are attracted to nestmates pinned to the ground as far away as 8 cm, when sound was the only signal that could be transmitted. Workers buried under the soil to a depth of 3 cm induced digging with their stridulation, while those buried more deeply, to a depth of 5 cm, only attracted workers to the vicinity.

Markl and his co-workers have discovered three functions of stridulation in various species and castes:

(1) In Atta at least, the stridulation serves as an underground alarm system, employed most frequently when a part of the colony is buried by a cave-in of the nest. The activity varies greatly among the castes. The large majors contribute very little, while nearly half of the medias most active in nest excavation also participate in rescue digging in response to stridulation (Markl, 1985). A similar function is implied for Pogonomyrmex badius, the workers of which readily stridulate when they are pinned by forceps, pressed beneath some object, or restricted to a small space in a container (Wilson, 1971). In Pogonomyrmex the signal works in conjunction with alarm pheromones of the mandibular gland. Higher concentrations of the latter release digging behavior (Wilson, 1958a).

(2) Stridulation is used during mating by young queens of Pogonomyrmex (Markl et al., 1977). These ants gather in nuptial swarms on the ground or vegetation, where substratal sound transmission is possible. The young queens stridulate vigorously when the spermatheca has been filled, and they struggle to escape from the swarms of males chasing them. This “female liberation signal” benefits both sexes by saving the males lost mating chances, while allowing the females to escape and commence the highly competitive business of colony founding--where time is of the essence. The signal is not used during courtship itself. In other words it does not aid the males in either finding or selecting their mates.

(3) In some species of Novomessor, Leptogenys, and Messor, stridulation enhances the effectiveness of pheromones during recruitment of nestmates to food finds, new nest sites, or both (Markl and Hölldobler, 1978; Maschwitz and Schönegge, 1983; Hahn and Maschwitz, 1985).

No evidence exists to rank the chirps of stridulation as anything more than simple unitary signals. In other words, ants do not “talk” by modulating sound through time. Although the sounds differ greatly from one species to the next, they do not appear to vary much within species or within the repertory of one worker ant through time (Forrest, 1963). Spangler (1967) showed that the two phases of the Pogonomyrmex occidentalis chirp, caused respectively by the up and down swings of the gaster, are interrupted at regular intervals, but the pulses thus created do not form any discernible temporal pattern. It is not even certain that such patterns could be preserved during transmission through the ground. An essential similar result was obtained in the analysis by Markl of sound production in Atta cephalotes. So far as we know, then, stridulation in ants produces a monotonous series of chirps with limited meaning.

In a survey of 1,354 ant species belonging to 205 ant genera, Markl (1973) demonstrated that the taxonomic distribution of the stridulatory organ is due to the interplay of phylogenetic inertia at the level of the subfamily and environmental adaptation at the level of the genus and species. On the one hand, the organ appears to be limited to four subfamilies, the Ponerinae, Nothomyrmeciinae, Pseudomyrmecinae, and Myrmicinae. The extinct, ancestral subfamily Sphecomyrminae bears no trace of the organ, so that stridulation may well be a derived trait within the ants as a whole. Within the Ponerinae, Nothomyrmeciinae, and Myrmicinae, the stridulatory organ tends to be present in genera and species that nest in the soil, and hence are both subject to cave-ins and able to perceive substratum vibrations. On the other hand, it tends to be absent in ants nesting in plants and rotten logs, leaf litter, and other soft ground materials that are poor sound transmitters. The Pseudomyrmecinae, which are almost all completely arboreal, are a puzzling exception to this rule. All 36 species examined by Markl possess a stridulatory organ. A similar paradox is presented by members of the myrmicine genus Crematogaster, all of which appear to possess the organ but many of which are primarily or exclusively arboreal. Procryptocerus, the most primitive genus of the exclusively arboricolous tribe Cephalotini, has a stridulatory organ, but the structure is lacking in the remainder of the cephalotines. It would thus appear, in accordance with Markl's principle, that the organ was lost secondarily as a consequence of arboreal life at an early period in the history of the Cephalotini.

Tactile communication
Several recent independent investigators, including especially Bonavita-Cougourdan, Hölldobler, Jaisson, Lenoir, and Torossian, have concluded that tactile signals serve several communicative functions; but also, despite their outward complexity, they convey only a limited amount of information. This relatively modest assessment of the role of touch is a considerable departure from that of Erich Wasmann (1899a), who interpreted the play of antennae on the bodies of sister ants as a complex “antennal language” playing a basic role in colonial organization. We now understand that a great deal of such antennation serves to receive information rather than to send it. Ants antennate the bodies of nestmates in order to smell them rather than to inform them.

Nevertheless, a role for touch by the antennae and forelegs has been well established in a few categories of communication (Lenoir and Jaisson, 1982). They have been firmly implicated in invitation behaviors that entrain most forms of recruitment. Typically one ant runs up to a nestmate and beats the body of the other ant very lightly and rapidly with her antennae, often raising one or both forelegs to touch the nestmate with these appendages in addition. The recruiter then turns and follows a recently laid odor trail, or lays a new one, or commences tandem running. In the case of tandem running, experiments have shown that the leader ant requires the touch of the follower ant before proceeding back toward the food find or new nest site (Hölldobler, 1971c, 1984b; Hölldobler et al., 1974; Maschwitz et al. 1974).

Because of the superficial complexity of tactile behavior, the observer is always at risk of over interpreting it. When one ant rushes up to another and partially mounts it, the function could be to bring chemical signals as close to the other ant as possible. And when it antennates the nestmate, the function could easily be chemoreception instead of signal transmission. Even so, the ritualized nature of many of the movements, including body jerking and intense antennal beating, suggests a true signal function even when direct experimental confirmation is still lacking.

The best documented mode of tactile communication is in stomodeal trophallaxis, the exchange of liquid food from the crop of one ant to the alimentary tract of another. As Hölldobler found (1967a,b; 1970a,b), the ability possessed by some myrmecophilous staphylinid beetles and other social parasites to induce ant workers to regurgitate to them, despite their outlandishly different anatomies, suggests that there must be some simple trick involved in the soliciting procedure. He was able experimentally to induce Myrmica and Formica workers to regurgitate by touching parts of their body with the tips of a human hair. The result was augmented by close observations of natural regurgitation, and the whole procedure can be summarized as follows. The most susceptible worker ant is one that has just finished a meal and is searching for nestmates with which to share its crop content. In order to gain its attention, a nestmate (or myrmecophile) has only to tap the donor's body lightly with antennae or forelegs. This causes the donor to turn and face the individual that gave the signal. If tapped lightly and repeatedly on the labium (lower mouth plate), the ant will regurgitate (see Figure 7-41). Some ants use their fore tarsi for this purpose, although the antennae may suffice (Figure 7-42). The myrmecophiles also employ their tarsi or antennae.

Once the food starts to flow, there is a heavy play of antennae between the donor and recipient, with that of the donor being more complex and variable. In both Myrmica rubra (Lenoir, 1982) and Camponotus vagus (Bonavita-Cougourdan, 1983; Bonavita-Cougourdan and Morel, 1984), the donor's paired funiculi are moved backward and forward, sometimes together and sometimes singly, to touch various parts of the dorsum of the recipient's head and, on occasion, the front ventral half of the recipient's head. The recipient directs its own funicular stroking to the anterior portion of the head, both the dorsal and ventral surfaces (Figure 7-43). The French investigators performed information analyses by measuring transition probabilities in the shifts from one antennal posture to another in both the donor and recipient ants. By this criterion very little transmission of information was recorded. One antennal posture did not lead to another within or across the repertories of the two ants in a regular sequence. In other words, signal A did not reliably cause response B, and so forth. In an important additional experiment, Bonavita-Cougourdan used radioactive gold to monitor the flow of liquid food between Camponotus pairs while recording the antennal postures of the donor. The onset of various positions brought no change in the flux. Overall, it seems unlikely that any of the postures convey particular messages to the receiver ants.

What then is the meaning of the elaborateness of antennal play? The paradox of complexity exceeding function is not unique to ants. Some other kinds of animal communication, for example the alternative courtship and territorial displays of many bird species, also appear to invest little individual meaning in the particular signals that are switched back and forth. We suggest that the chief role of variety in such signaling is dishabituation. Animals tend to habituate when they receive the same stimulus repeatedly; that is, their response threshold rises and the contact is thus more likely to be broken off. By constantly changing the parts of the head stroked, both the donor and the recipient are more likely to sustain the trophallactic exchange. It follows that when the donor has a large load it should use a more intense and variable repertory than otherwise. The same is true of the recipient if it is very hungry. Both correlations appear to be the case in nature. However, this fact alone does not prove dishabituation, because the correlations could also be simply a result of the general intensification of behavior with rising motivation.

Visual communication
The use of visual signals in ants is at best very minor and in fact not a single example has yet been solidly documented. This negative generalization is not simply the outcome of a physiological constraint. Many large-eyed ants have excellent form vision and are keen at detecting moving objects (Jander, 1957; Ayre, 1963b; Voss, 1967; Via, 1977; Wehner, 1981; Wehner et al., 1983). The South American formicine Gigantiops destructor (Figure 7-44) and African formicine Santschiella kohli have huge eyes that cover much of the sides of the head. Gigantiops workers are difficult to catch because they see approaching human observers from several meters away and flee by swiftly running and jumping away (the behavior of Santschiella, a very rare ant, has never been recorded). Workers of large-eyed ants generally do not respond to prey insects that are standing still, but run toward them as soon as they begin to move. Stäger (1931) noticed that when lone workers of Formica lugubris foraging in a field encounter an insect, they dash in erratic circles around it and attract other workers that happen to be in the vicinity. Stäger believed that the communication is mediated by vision alone and labeled the phenomenon “kinopsis.” Later Sturdza (1942) showed that the sight of a running Formica nigricans worker, apparently without reinforcement of other kinds of stimuli, is enough to set another worker running. Workers of Daceton armigerum, a large-eyed predatory ant living in the canopy of South American rain forests, rush to join other workers running around prey insects (Wilson, 1962a; see Plate 6). Unfortunately, neither Stäger nor Wilson eliminated the possibility of short-range chemical recruitment signals. It is now well known that workers of the Aphaenogaster cockerelli group of Aphaenogaster (Hölldobler et al., 1978), Oecophylla longinoda (Hölldobler and Wilson, 1978), Lasius neoniger (Traniello, 1983), and species of Formica (Horstmann et al., 1982; Horstmann and Bitter, 1984) are able to spread short-range recruitment pheromones over distances of at least several centimeters within a few seconds. This invisible chemical communication is accompanied by conspicuous running that could easily be misinterpreted as visual signaling.

Alarm communication
Alarm is the most difficult of behavioral responses merely to define, because investigators have used it as a portmanteau category into which all responses to danger are placed. In the broad sense worker ants are said to be in a state of alarm when they move away from a potentially dangerous stimulus, either calmly or in panic, or charge toward it aggressively, or simply mill about in a state of heightened alertness. Nevertheless, for each particular species in turn it is usually easy to devise a bioassay based on the precise reactions of the colonies, without having to take into account all the variations of behavior displayed by other species.

Most alarm signals are multicomponent. They typically consist of two or more pheromones, which often serve simultaneously to alert, attract, and evoke aggression. Acoustical signals are sometimes added to the pheromones, including especially stridulatory chirps that enhance attraction. The ants also occasionally touch nestmates with their antennae and forelegs during ritualized motor displays.

Alarm behavior is made additionally difficult to characterize because it so often blends into other major behavioral responses. Most ant species engage in some form of alarm-defense, in which the same chemicals are used to repel enemies and to alert nestmates. Examples of alarm-defense substances that have been identified in various ant species are citronellal, dendrolasin, dimethyl sulfate, and undecane. Some secretions are probably purely defensive, such as the formic acid employed by at least some formicine species. It is also possible that other compounds are purely communicative, especially those produced in minute quantities, but this specialized function remains to be documented in many cases of which we are aware. Probably the great majority of compounds used in alarm communication also serve in defense. Put in the sequence more likely to have occurred in evolution, many chemicals employed in defense have also been ritualized to some extent into alarm signals.

A second intergradient category that has been documented is alarm-recruitment. Alarm signals both alarm and attract in some species, while in others the alarm pheromones are combined with odor trails that lead nestmates toward or away from the source of the danger. Many species employ a single alarm-recruitment procedure to alert nestmates to both enemies and prey, and in fact the distinction between the two may be wholly blurred with reference to communication.

Whether joined with recruitment or not, alarm behavior can be conveniently classified into one or the other of two broad categories (Wilson and Regnier, 1971). In aggressive alarm some of the colony members (often the majors or “soldiers”) are drawn toward the threatening stimulus and seek to attack it. In panic alarm the colony as a whole flees from the stimulus or dashes around in erratic patterns. If disturbed strongly enough by the waves of alarm, they may even evacuate the nest.

A wide range of stimuli evokes alarm communication. Oddly, substratal vibrations and air currents disturb the entire colony and may even induce evacuation if severe enough, but they seldom trigger the release of alarm pheromones. Alarm communication is initiated far more predictably when an enemy penetrates the close environs of the nest. Certain stimuli work better than others in this context. In the phenomenon called “enemy specification,” dangerous species are more effective than other, less threatening ones in evoking a response. For example, minor workers of Pheidole dentata recruit majors to only a single fire ant forager approaching the nest (Wilson, 1976b). Those of Pheidole desertorum and Pheidole hyatti initiate panic alarm leading to nest evacuation when they become aware of the approach of army ants of the genus Neivamyrmex (Droual and Topoff, 1981). Such hair-trigger responses are not limited to living enemies. A single drop of water in the nest entrance of Pheidole cephalica is often enough to cause alarm-recruitment and the full retreat of the colony away from that part of the nest (Wilson, 1986c).

Alarm is easy to define but technically one of the most difficult forms of communication to study with precision. At lower stimulus intensities the responses may be subtle, entailing nothing more than increased alertness or an increase in running velocity combined with a reduced sinuosity in the direction taken (Cammaerts et al., 1982). It is further possible to get an alarm response from any volatile compound extracted from ants, if the concentrations used are high enough. A positive reaction cannot be used as evidence that the substance is an alarm pheromone if the experimental concentrations are higher than those that could be generated by the ants themselves. An essential part of the analysis is to create active spaces of the size and geometry resembling those expected under natural conditions. Investigators usually accomplish this end crudely by crushing single glands containing the pheromone.

Let us now examine a particular case of a relatively uncomplicated aggressive alarm system. When a worker of the subterranean formicine ant Acanthomyops claviger is severely threatened, say attacked by a member of a rival colony or an insect predator, it reacts strongly by simultaneously discharging the contents of the reservoirs of its mandibular and Dufour's glands. After a brief delay, other workers resting a short distance away display the following response: the antennae are raised, extended, and swept in an exploratory fashion through the air; the mandibles are opened; and the ant begins to walk, then run, in the general direction of the disturbance (Regnier and Wilson, 1968). Workers sitting a few millimeters away begin to react within seconds, while those a few centimeters distant take a minute or longer. In other words, the signal appears to obey the laws of gas diffusion. Experiments have implicated some of the terpenes, hydrocarbons, and ketones as the alarm pheromones; these are shown in Figure 7-30. Undecane and the mandibular gland substances (all terpenes) evoke the alarm response at concentrations of 109-1012 molecules per cubic centimeter, reflecting a moderate amount of sensitivity as far as pheromones go. These same substances are individually present in amounts ranging from as low as 44 ng to as high as 4.3 µg per ant, and altogether they total about 8 µg. Released as a vapor during experiments, similar quantities of the synthetic pheromones produce the same responses. Apparently the Acanthomyops claviger workers rely entirely on these pheromones for alarm communication. Their system seems designed to bring workers to the aid of a distressed nestmate over distances of up to 10 cm. Unless the signal is then reinforced by additional emissions, it dies out within a few minutes. The Q/K ratios are on the order of 103-104. If the entire contents of the Dufour's gland, containing about 2.5 µg of undecane, are discharged as a puff from the poison gland, the diffusion model of Bossert and Wilson (1963) predicts that the pheromone signal will reach a maximum of about 20 cm in still air. If on the other hand only 0.1 percent was to be discharged, this active space can still reach a maximum of 2 cm. Hence the match with the observations of natural behavior is reasonably close.

The alerted Acanthomyops workers approach their target in a truculent manner. This aggressive defensive strategy is in keeping with the structure of their colonies, which are large in size and often densely concentrated in the narrow subterranean galleries. It would not pay the colonies to try to disperse when their nests are invaded, and, consequently, the workers have apparently evolved so as to meet danger head-on.

A different strategy, based on panic and escape, is employed in the chemical alarm-defense system of the related ant Lasius alienus (Regnier and Wilson, 1969). Colonies of this species are smaller and normally nest under rocks or in pieces of rotting wood on the ground; such nest sites give the ants ready egress when the colonies are seriously disturbed. Lasius alienus produce the same volatile substances as Acanthomyops claviger, with the exception of citronellal and citral. Their principal volatile component is undecane. When they smell the latter pheromone, the Lasius workers scatter and run frantically in an erratic pattern. They are more sensitive to undecane than the Acanthomyops workers, being activated by only 107-108 molecules per cubic centimeter. Thus in contrast to Acanthomyops claviger, Lasius alienus utilizes an “early warning” system and subsequent evacuation in coping with serious intrusion.

Alarm systems are nearly universal in ants, having been discovered in every species in which a test for the phenomenon was performed. This is true even of the species considered to be among the most primitive, Amblyopone pallipes (Traniello, 1982), Myrmecia gulosa (Robertson, 1971), and Nothomyrmecia macrops (Hölldobler and Taylor, 1983). By far the most common mode of alarm communication is chemical. Many species, such as the members of Acanthomyops and Lasius, employ alarm pheromones without the accompaniment of acoustical or tactile displays. In general, it appears that where acoustical and tactile signals exist they have been added onto chemical alarm and recruitment as modulators. However, this does not mean that such signals are specialized or derived with reference to the evolution of the ants as a whole. Stridulatory organs are present in Nothomyrmecia and many members of the primitive subfamily Ponerinae. Among members of the latter group, Amblyopone australis and Amblyopone pallipes utilize a tactile signal during a vibratory display in which the ants vigorously jerk the head and thorax up and down while coming into contact with nestmates (Hölldobler, 1977; Traniello, 1982). In [Stigmatomma pallipes|Amblyopone pallipes]] at least, mandibular gland pheromones also serve as weak attractants. They might serve in alarm, recruitment, or both--not enough experiments have been made to make a distinction.

A remarkable alarm behavior in the Australian bulldog ant, Myrmecia, was recently discovered by Hubert Markl (unpublished observations). On the basis of models of the two-dimensional random alarm process (Frehland et al., 1985), Markl and his collaborators give the following account: “Around an undisturbed nest one usually finds a small number of workers randomly distributed over a few square meters, sitting still or moving about slowly. If an intruder disturbs one of these 'sentinels,' it may trigger it into a frantic, erratic run for a number of seconds. If it thereby comes within sight of a second nestmate, this will lunge forward as if in attack. Upon direct contact, however, combat is avoided and now the second ant starts the same alarm-run while the first one may continue its own. Depending on density and distribution of ants over the guarded area and on number of guards initially stimulated, the alarm can spread two-dimensionally over an extended area. Soon, one or the other of the alarmed workers will return to the nest and there recruit additional forces to join the excited crowd. Intruders into the nest territory will thus be reliably detected and attacked by ants (who can sting ferociously) notwithstanding the stochasticity of the environment. After intruders have been driven out, the alarm subsides by not receiving more stimulating input and/or by having spread out over too large an area. Thus, with a small number of widely distributed guards, the colony can efficiently control a fairly large territory which cannot completely be overrun by any single individual, without engaging more work force than necessary at any given time.”

Most of the alarm pheromones identified to the present time are listed in Table 7-4. Their great structural diversity is a reflection of the antiquity and phylogenetic complexity of the ants themselves. Additional variety comes from the fact that most alarm pheromones also serve as defensive substances and have probably obtained their communicative function through ritualization many times independently. In some instances certain exocrine glands initially involved in alarm communication are secondarily hypertrophied and function as the major defensive devices of the ants (Buschinger and Maschwitz, 1984; Figure 7-45). Defensive chemicals, basic to the biology of social insects, are in turn tied into varying strategies that involve entire syndromes of anatomical structure and behavior. These syndromes are known to evolve across species in correlation with colony size, nest site, and other aspects of natural history. In short, the diversity of alarm pheromones is an expected consequence of phylogeny and the close linkage that exists in ants between defense and alarm. The one common feature the substances seem to share is molecular weight. The great majority of the compounds are in the C6 to C10 range. This is in accord with the prediction (Wilson and Bossert, 1963) that alarm pheromones should evolve in the lower molecular size range because of the need for substantial volatility and a lower Q/K ratio, permitting the rapid expansion and fade-out of the active space. Further, there is little need for privacy in communication, so that molecular complexity and size have not been at a premium during evolution.

Propaganda
Workers of the slavemaking ant Formica subintegra possess an enormously enlarged Dufour's gland loaded with approximately 700 micrograms of a mixture of decyl, dodecyl, and tetradecyl acetates (Regnier and Wilson, 1971). The substances are sprayed at defending colonies of other species of Formica during slave raids. They attract the Formica subintegra workers, but alarm and disperse the defenders. Hence they act at least in part as “propaganda substances.” The acetates are well designed for this purpose in accordance with the engineering rules for the evolution of pheromones. Having a higher molecular weight than most ordinary alarm substances, the Acanthomyops claviger (Regnier and Wilson, 1968), has not been tested explicitly in Formica.

In a remarkably parallel manner, workers of the myrmicine slavemaker Harpagoxenus sublaevis employ propaganda substances from their Dufour's glands to subdue workers of target Leptothorax acervorum colonies. One or more of the components causes the Leptothorax to fight their own nestmates, throwing the colony into greater confusion and rendering it more vulnerable to raids by the Harpagoxenus workers. An identical effect is produced by Dufour's gland secretions voided by the queens of Leptothorax kutteri, a workerless social parasite of Leptothorax acervorum (Allies et al., 1986), and queens of the slave raiding Polyergus breviceps (Topoff et al., 1988).

Recruitment
Recruitment is defined as communication that brings nestmates to some point in space where work is required (Wilson, 1971). The ants have evolved an astonishing array of devices to assemble workers for joint efforts in food retrieval, nest construction, colony defense, and emigration to new nest sites. Various species employ idiosyncratic combinations of touch, stridulation, and chemical cues to achieve these ends. The category of recruitment is thus a very loose one, and in particular cases it often cannot be clearly distinguished from alarm and simple assembly. The Australian dacetine ant Orectognathus versicolor is typical in the way it mixes the functions. Workers employ trails laid from their poison gland to organize colony emigrations, during which their pheromones serve as a stimulative recruitment signal as well as a longer lasting orientation cue. On the other hand, trails laid from the Orectognathus pygidial gland appear to function as a short lasting alarm-recruitment signal, channeling workers to areas of disturbance near the nest. There is also some evidence that the pygidial gland pheromone is discharged by successful foragers as they enter the nest, increasing the rate at which nestmates exit and follow the poison gland trail (Hölldobler, 1981b).

The study of recruitment has recently grown into a substantial field of inquiry unto itself. During the 1950s and 1960s investigators emphasized the straightforward identification of the glandular source of the pheromones, while paying some attention to the details of trail-laying behavior (see reviews in Wilson, 1971; Blum, 1974b; Hölldobler, 1977; Parry and Morgan, 1979). Now they have begun to focus on an additional topic: the analysis of the organizational levels and ecological significance of recruitment. The possession of one kind of recruitment as opposed to another clearly constitutes an adaptation by individual species to particular long-term conditions in the environment. Researchers generally agree that the recruitment strategy makes little sense except with reference to the ecology of the species. Conversely, the ecology of most species cannot be fully understood without a detailed knowledge of their recruitment procedures.

The most prevalent form of recruitment behavior is chemical trail communication. Carthy (1950, 1951a,b) was one of the first to conduct an experimental study of trail laying in ants. He obtained strong circumstantial evidence that Lasius fuliginosus produces the trail pheromone in its hindgut, a result later confirmed by Hangartner and Bernstein (1964). Wilson (1959d, 1962b), working with the red imported fire ant Solenopsis invicta]] (= Solenopsis saevissima in part), provided the first bioassay methods to test trail-laying behavior even in the absence of a recruiting ant. He laid artificial trails made from different glandular extracts away from the nest entrance and from worker aggregations around food finds. By comparing the trail-following response of worker ants, he was able to identify the Dufour's gland as the source of the trail pheromone in fire ants. This technique has been subsequently used by many investigators and led to the discovery of a number of trail pheromone glands in different taxonomic groups (Table 7-5 and Figure 7-46). Most recently,there has also been rapid progress in the chemical identification of trail pheromones (Attygalle and Morgan, 1985; Table 7-6).

Wilson's (1962b) analysis also revealed for the first time the existence of chemical mass communication in ants, one of the most complex forms of social behavior occurring in the social insects. The details of the communication can be summarized as follows. When fire ant workers leave the nest in search of food, they may follow odor trails for a short while, but they eventually separate from each other and begin to explore singly. When alone they orient visually at least in part, a fact that was established by the following simple experiment. Single workers were first permitted to explore a foraging table in the laboratory for distances up to a meter from the nest. The only source of illumination was a lamp beamed from one side of the table. The workers were next allowed to discover a drop of sugar water. When they had fed and started home, laying an odor trail behind them, the light was switched off and a second light located on the opposite side of the table was simultaneously switched on. After the direction of the light source had thus been changed by 180°, the ants almost invariably performed a complete about-face. By alternately switching the light source from one side to another, individual workers could be marched back and forth like puppets and finally brought home at will.

When a fire ant discovers a food source, it heads home at a slower, more deliberate pace. Its entire body is held closer to the ground. At frequent intervals the sting is extruded, and its tip is drawn lightly over the ground surface, much as a pen is used to ink a thin line (Figure 7-47). The key recruitment pheromone in Solenopsis invicta is an alpha-farnesene, with synergistic effects being added by two homofarnesenes, possibly n-heptadecane, and a still unidentified component from the Dufour's gland. A second alpha-farnesene and an allofarnesene may also play minor roles as recruiters (Vander Meer, 1983, 1986a; see Figure 7-32). Each worker contains less than a nanogram at any given moment. Artificial trails made from a single Dufour's gland induce following by dozens of individuals over a meter or more. When the concentrated pheromone is allowed to diffuse from a glass rod held in the air near the nest, workers mass beneath it, and they can be led along by the vapor alone if the rod is moved slowly enough. When Wilson presented large quantities of the substance at the entrances of laboratory nests, he was able to draw out most of the inhabitants, including both workers carrying larvae and pupae and, on a single occasion, the mother queen.

While laying a trail, the fire ant worker sometimes loops back in the direction of the food find, but only for short distances, before turning toward the nest again. If another worker is contacted, the homeward bound worker turns to face it. It may do no more than rush against the encountered worker before moving on again, but sometimes the reaction is stronger: the recruiter climbs on top of the worker and, in some instances, shakes her body lightly but vigorously in a vertical plane. The movement may represent a modulatory signal that enhances the effect of the pheromone, but it does not appear to impart any essential information about the food find, because contacted workers do not exhibit trail-following behavior obviously different from those not contacted. Moreover, the pheromone is by itself sufficient to induce immediate and full trail-following when laid down in artificial trails.

Most workers encountering a freshly laid trail respond at once by following it outward from the nest. They are able to detect the farnesene pheromones by smelling the vapor over distances as great as a centimeter. The workers do not follow a liquid odor trace on the ground. Instead, they move through the vapor created by diffusion of the substances through the air. There is an active space, which theoretical calculations show to be semiellipsoidal in shape, within which the pheromone is detected by the ants (Figure 7-48). As follower workers travel through this “vapor tunnel,” they sweep their antennae from side to side, evidently testing the air for odorant molecules. In fact, they are able not only to detect these molecules in the gaseous state but also to move up gradients of molecular concentration, a process of orientation referred to as osmotropotaxis (Martin, 1964). When fire ants are presented with trail substance evaporating in still air from the tip of a glass rod, they run directly to the glass rod. The sensory mechanism enabling fire ants to orient in this fashion is still unknown. However, in a set of ingenious experiments, summarized in Figure 7-49, Hangartner (1967) was able to demonstrate the basis of osmotropotaxis in another trail-following ant species, Lasius fuliginosus. The method of following odor trails disclosed by these experiments makes it very unlikely that directional signals can be built into the trails. In other words, the odor streaks may or may not be tapered or shaped in some other way so as to point the way home--as discovered for example in Myrmica ruginodis trails by Macgregor (1948)--but it would be difficult for the follower ant to “read” this information. Additional experiments performed on Lasius, Acanthomyops, and Solenopsis have indicated in fact that no such information is transmitted (Carthy, 1951b; Wilson, 1962b). These findings refute the early hypothesis of Bethe (1898) that trails are effectively polarized, as well as the conjecture of Pieron (1904) that ants find their way back and forth by a special kinesthetic sense. Recently Moffett (1987a) discovered that food-laden workers of Pheidologeton diversus orient along chemical trails but determine the home direction according to the directions taken by other laden nestmates.

Our present knowledge makes it desirable to reassess the value of Auguste Forel's mysterious theory of the “topochemical sense.” Although this notion is frequently mentioned in the older literature, especially in connection with odor trails, there is general confusion over what it really means. Forel seems to have been talking about the perception of form and the spatial relation of discrete objects by means of smell, but his pronouncements on the subject were discouragingly obscure. Here is the clearest we have encountered: “By topochemical I mean a sense of smell which informs the ant as to the topography of the places surrounding it by means of chemical emanations, which give an odor to objects” (Forel, 1928: I, 116). The sense was said to be located in the terminal segments of the antennae. The perception of the spatial relation of objects implies an integration of sensory input in the central nervous system considerably more complex even than the following of an odor trail by osmotropotaxis. This has been well documented in the case of visual input (Jander, 1957; Wehner, 1981) but not yet in the case of olfactory input.

A more important finding from the work on Solenopsis invicta involves the way in which mass communication is achieved. By mass communication Wilson meant information that can be transmitted only from one group of individuals to another group of individuals. In the case of Solenopsis invicta, the number of workers leaving the nest is controlled by the amount of trail substance being emitted by foragers already in the field. Tests involving the use of enriched trail pheromone showed that the number of individuals drawn outside the nest is a linear function of the amount of the substance presented to the colony as a whole. Under natural conditions this quantitative relation results in the adjustment of the outflow of workers to the level needed at the food source. An equilibration is then achieved in the following manner. The initial buildup of workers at a newly discovered food source is exponential, and it decelerates toward a limit as workers become crowded on the food mass because workers unable to reach the mass turn back without laying trails and because trail deposits made by single workers evaporate within a few minutes. As a result, the number of workers at food masses tends to stabilize at a level which is a linear function of the area of the food mass. Sometimes, for example when the food find is of poor quality or far away or when the colony is already well fed, the workers do not cover it entirely, but equilibrate at a lower density. This mass communication of quality is achieved by means of an “electorate” response, in which individuals choose whether to lay trails after inspecting the food find. If they do lay trails, they adjust the quantity of pheromone according to circumstances (Hangartner, 1969b; see Figure 7-47). The more desirable the food find, the higher the percentage of positive responses, the greater the trail-laying effort by individuals, the more the trail pheromone presented to the colony, and, hence, the more the newcomer ants that emerge from the nest.

Thus, the trail pheromone, through the mass effect, provides a control that is more complex than might have been assumed from knowledge of the relatively elementary individual response alone. This complexity is increased still further by the fact that the pheromone assumes different meanings in at least two different contexts. When colonies move from one nest site to another, a common event in the life of fire ants as well as many other kinds of ants, the new site is chosen by scout workers which then lay odor trails back to the old nest. Other workers are drawn out by the pheromone. They investigate the new site and, if satisfied, add their own pheromone to the trail. In this fashion the number of workers traveling back and forth builds up exponentially. In time the brood is transferred, the queen walks over, and the emigration is completed. The pheromone also functions as an auxiliary signal in alarm communication. When a worker is seriously disturbed, it releases some of the trail substance simultaneously with alarm substance from the head, so that nearby workers are not only alarmed but also attracted to the threatened nestmate.

Wilson (1962b) was able to measure the information transmitted by the odor trail of Solenopsis invicta and compare it with that transmitted by the waggle dance of the honeybee, which had been analyzed previously by Haldane and Spurway (1954). The method is straightforward. Workers were allowed to lay odor trails away from a food find (a drop of sugar water), and the starting point of the trail was marked in order to measure its distance from the food find--and hence, the initial error committed by the trail-laying ant. Similarly, the paths taken by the outward-bound follower ants and the distances of their nearest approaches on the trail to the food find were each recorded. The spatial precision of the two actions was then correlated and condensed into transmitted information, which is measured in bits. The bit is the amount of information needed to make a choice between two equiprobable alternatives. If n alternatives are present, a choice provides the following quantity of information:

H = log2n

Thus, in a most elementary communication system, the sending of one of n equiprobable messages reduces log2n amount of uncertainty. By definition, it provides that much information.

Suppose we ourselves were communicating information about direction, and suppose our system allowed us to transmit any one of 16 directions with perfect accuracy. Then the message “go north by northwest,” one of 16 equiprobable messages, conveys log216 = 4 bits of information. The essential results for honeybees and fire ants are shown in Figure 7-50. The key result is that both systems transmit about the same amount of information concerning both direction and distance. Also, perhaps coincidentally, they convey a comparable degree of precision as human beings when we think about spatial locations and express them verbally. Thus we can just about place direction to “north by northwest” mentally, and to no compass direction of finer degree; this also happens to be the performance of the honeybee and fire ant.

Why don't the social insects do better? As Haldane and Spurway noted in the case of the honeybee, “natural selection is always acting so as to reduce the error of the mean direction, while acting less intensely, if at all, to reduce individual errors which lead to a scatter round this direction.” They cite the analogous problem in naval gunnery, “where a superior force pursuing ships with less fire power should fire salvoes with a considerable scatter, in the hope that at least one shell will hit a hostile ship and slow it down.”  The same analogy at first glance appears even more relevant in the case of the fire ant. One of the chief problems faced by this species in the course of foraging is the recruitment of sufficient numbers of workers in time to immobilize small prey animals detected while passing through the colony territory. In laboratory arenas workers often succeed in capturing insects only because they deviate from odor trails that had been rendered inaccurate by the continued movement of the prey. Recently Deneubourg et al. (1983) have also pointed out the importance of “noise” in chemical trail recruitment in ants for simultaneously directing and spreading the foraging efforts.

There is an alternative and simpler explanation of the quantity of information transmitted by ant recruitment trails. It is possible that the level of accuracy of the pheromone system has been arrived at as a compromise between the utmost effort of the ants' chemosensory apparatus to follow trails accurately and, simultaneously, the need to reduce the quantity and to increase the volatility of the trail pheromones in order to minimize overcompensation in the mass response. The amount of the trail substances is indeed microscopic, which seems logical if the mass response is to be finely governed. Which of the two evolutionary hypotheses is closer to the truth cannot be determined until deeper physiological analyses have been made.

The mass communication system exemplified by Solenopsis invicta clearly represents an advanced evolutionary grade. In order to identify the more primitive forms of recruitment from which this system might have evolved, it is necessary to discover and characterize less sophisticated modes of recruitment communication. Tandem running, used variously during emigration and recruitment to food, is generally considered to be one such mode. Only a single nestmate is recruited at a time, and the follower has to stay in direct antennal contact with the leader. The pair then proceeds to the target site while remaining tightly linked together. The phenomenon was described for the first time, albeit sketchily, by the Swedish entomologist Gottfrid Adlerz in 1896, who saw a Leptothorax slave worker leading a Harpagoxenus sublaevis slavemaker to a new nest site (cited by Stuart, 1987d; and the phenomenon was confirmed by Buschinger and Winter, 1977). It was encountered a second time in the tropical Asian ant Camponotus sericeus by Hingston (1929), and then described in detail for the first time in Cardiocondyla by Wilson (1959b), who also first applied the expression “tandem running.” The ensuing 30 years have witnessed the discovery of this form of communication across a wide range of additional genera in the Ponerinae, Myrmicinae, and Formicinae (Table 7-7). Also, the first experimental analyses have been conducted on the signals employed by the ants.

One of the best understood species is the little European myrmicine Leptothorax acervorum (Möglich et al., 1974; Möglich, 1978, 1979). When a successful scout returns to the colony, it first regurgitates food to several nestmates. Then it turns around and tilts its gaster upward into a slanting posture (see Figure 7-36). Simultaneously it exposes its sting and extrudes a droplet of liquid. Nestmates are attracted by this “tandem calling” behavior, as Möglich and his co-workers have labeled it. When the first ant arrives at the calling ant, it touches the caller on the hind legs or gaster with its antennae, and tandem running commences. The recruiting ant leads the nestmate to the newly discovered food source. During tandem running the leader ant lowers the gaster, the sting remaining extruded. But the sting is not dragged over the surface, as in the case of Solenopsis invicta and other species that lay chemical trails from their stings. Instead, the Leptothorax acervorum follower keeps close antennal contact with the leader, repeatedly touching its hind legs and gaster. Whenever this contact is interrupted, for example when the follower accidentally loses its leader or is removed experimentally, the leader immediately stops and resumes the calling posture. Once again it stands still and points its abdomen obliquely upwards. It may remain in this posture for several minutes, continuously discharging the calling pheromone. Under normal circumstances the lost follower quickly orients back to the leader ant and tandem running is resumed. Closely similar behavior has been observed in Leptothorax muscorum and Leptothorax nylanderi.

Experiments on this interesting recruitment behavior have revealed tandem running to be mediated by signals in two sensory modalities:

(1) If a Leptothorax acervorum tandem pair has been separated, the leader immediately stops and resumes the calling posture. However, when the ant is carefully touched with a hair at the hind legs or gaster with a frequency of at least two contacts per second, the leader continues running to the target area. This experiment shows that the presentation of the tactile signals normally provided by the follower ant is sufficient to release “tandem running” by a leader ant.

(2) The calling pheromone originates from the poison gland. In the studies by Möglich and his co-workers, ants were strongly attracted to dummies that had been contaminated with poison gland secretions but not to those contaminated with Dufour's gland secretions. Further experiments revealed that the poison gland substance not only functions as a calling pheromone but also plays an important role during tandem running itself by binding the follower ant to the leader. Möglich et al. (1974) found that the leader could easily be replaced by a dummy contaminated with poison gland secretions. Gasters of freshly killed ants from which the sting and venom glands had been removed could not replace a leader ant. However, when these body parts were contaminated with secretions of the poison gland, they functioned effectively as leader dummies.

The discovery of tandem calling with pheromones in Leptothorax throws considerable light on the evolution of chemical recruitment techniques in myrmicine ants generally. It now seems very plausible that the highly sophisticated chemical mass recruitment performed by Solenopsis, Monomorium, Pheidole, Pheidologeton, and certain other myrmicine ants was derived from the chemical tandem calling behavior of the Leptothorax mode. With the exception of Crematogaster, which produces a trail pheromone in the tibial glands of the hind legs, all other myrmicine species generate the pheromone from one of the sting glands. As Morgan (1984) has pointed out, the trail pheromone is found mostly in the poison gland of species with a proteinaceous venom, and mostly in the Dufour's gland in species with strongly alkaloidal venom (as in Solenopsis and Monomorium). Some ant species have lost the ability to sting but still retain a proteinaceous poison secretion. It is likely that pheromonal calling, during which an alerting and attracting pheromone is discharged through the sting into the air, was one of the first steps that led to chemical trail laying and mass communication in the later evolution of myrmicine ants.

Chemical calling, which as we noted in the discussion of signal parsimony closely resembles tandem calling, is also employed by reproductive castes of the social parasites Doronomyrmex kutteri and Harpagoxenus sublaevis (Buschinger, 1968a, 1971a,b). Harpagoxenus uses tandem running during slave raids (Buschinger and Winter, 1977; Buschinger et al., 1980a; see Figure 12-13). The manufacture of sex pheromones in the sting glands is widespread among myrmicine ants, having been documented in Harpagoxenus, Leptothorax, Monomorium, Xenomyrmex, and Pogonomyrmex (for review see Hölldobler and Bartz, 1985). It is therefore reasonable to suppose that in at least some myrmicine phyletic lines, sex attractants and recruitment pheromones had a common evolutionary origin (Hölldobler, 1978). In fact, the same substances may have originally served as sex pheromones in one context and as recruitment signals in another. Nearly identical modes of recruitment communication have been discovered in other ant subfamilies. But since the anatomical origins of the pheromones are different, we have to assume that the patterns themselves, however outwardly similar, have evolved independently several times. The parallel is especially striking between the Ponerinae and Myrmicinae.

Even some ponerine species that are exclusively solitary foragers employ tandem running recruitment during nest relocations. In fact, it is possible that tandem running recruitment, like social carrying, first evolved as a device to lead stray nestmates back to the nest. Many ponerine ants construct relatively simple nests in soil that are friable and prone to cave-ins. It would be advantageous for workers to be able to lead nestmates to intact portions of the nests with little delay. Several species of the ponerine genus Pachycondyla employ tandem running when recruiting nestmates to new nest sites or food sources (Maschwitz et al., 1974; Traniello and Hölldobler, 1984). A recruiting worker of Pachycondyla obscuricornis “invites” a nestmate with a stereotyped display that leads to the formation of the tandem pair. The nestmate then follows the recruiter to the new nest by keeping in close contact with her hind legs. If the contact is broken, as for example when the follower accidentally loses the leader or is removed experimentally, the leader immediately stops. Only after it receives the tactile signal on its hind legs or gaster does it continue to travel towards the target area. During tandem running, secretions from the pygidial gland released by the leader ant provide a chemical bond between it and the follower (Hölldobler and Traniello, 1980a). In the workers of some other ponerine species, including at least a few Bothroponera, Hypoponera, and Diacamma, which also recruit by tandem running, the pygidial gland secretion is evidently not involved (Jessen and Maschwitz, 1986; Hölldobler, unpublished data).

Pachycondyla obscuricornis is a solitary forager that uses tandem running only during nest emigration. In contrast Pachycondyla laevigata is an obligate predator on termites, and conducts massive group raids when foraging. The raids are organized by a powerful trail pheromone discharged from the pygidial gland by scout ants (Hölldobler and Traniello, 1980b; see Figures 7-22 and 7-23). The shift in diet from evenly dispersed to strongly clumped food sources (termites) apparently led to the use of the pygidial gland secretions in mass communication and the transformation of the pheromones into stimulatory and orienting signals. As we showed earlier, queens of some myrmicine species display sexual calling behavior identical to the tandem calling behavior of the workers. Again, the parallel to the Ponerinae in this respect is striking: here the pygidial gland appears to have a functional repertory identical to that of sting glands in the Myrmicinae. Depending on the species and behavioral context, the secretions of the ponerine pygidial glands can function as tandem running pheromones, recruitment trail pheromones, or sex pheromones (Hölldobler and Haskins, 1977).

In short, comparative studies have revealed several strikingly convergent pathways in the evolution of mass communication in ants (see also Jaffe, 1984). On the other hand, recent morphological and behavioral findings indicate that communication by chemical trails in ants is considerably more diverse overall than previously assumed. As illustrated in Figure 7-46, no fewer than 10 different anatomical structures have been identified in various species of ants as sources of trail pheromones. In the Ponerinae alone four different trail pheromone glands have been identified. It is obvious that trail communication has evolved many times independently. Even in the same subfamily the mechanisms and anatomical structures for trail communication have diverged considerably. In the termitophagous species Megaponera foetens and in several group-raiding Leptogenys species, the trail pheromone originates from the poison gland (Fletcher, 1971; Maschwitz and Mühlenberg, 1975; Longhurst et al. 1979a). In some Leptogenys species at least, a second recruitment pheromone originates in the pygidial gland (Maschwitz and Schönegge, 1977; Hölldobler, unpublished data). In the myrmecophagous group-raiding species Cerapachys turneri, the trail pheromone originates from the poison gland, but appears to be reinforced by a chemical recruitment signal from the pygidial gland (Hölldobler, 1982b). In several species of the legionary and group raiding genus Onychomyrmex, nest emigrations and raids are organized by trails laid with secretions from a large sternal gland located between the Vth and VIth abdominal sternites (Hölldobler et al., 1982; see Figures 7-8 and 7-24). Finally, the African stink ant Paltothyreus tarsatus, which is a scavenger and termite predator, employs tandem running during nest relocation and trail communication during foraging. The trail pheromone, which functions mostly as an orientation cue, originates from intersegmental sternal glands located between abdominal sternites IV and V, V and VI, and VI and VII (Hölldobler, 1984b; see Figure 7-7). Trail communication appears to be much more common in the Ponerinae than previously believed. Recently Overal (1986) has provided circumstantial evidence that the Neotropical species Ectatomma quadridens employs chemical trails in nest emigrations and recruitment of foragers. Stephen Pratt (personal communication) discovered trail communication in Ectatomma ruidum and identified the Dufour's gland as the source of the trail pheromone. Similarly, Breed et al. (1987) demonstrated that the giant Neotropical ponerine Paraponera clavata exhibits graded recruitment responses, depending on the type, quantity, and quality of the food source. Their observations suggest that a trail pheromone is used in orientation by individual ants, independent of recruitment, very similar to that documented in Pachycondyla tesserinoda by Jessen and Maschwitz (1985, 1986). On the other hand Pachycondyla tesserinoda recruits by tandem running, while Paraponera clavata appears to have a recruitment system similar to that of Paltothyreus. In other words, the recruitment signal by which nestmates are stimulated to leave the nest and follow the orientation trail appears to be separate from the trail pheromone (Hölldobler, 1984b).