The Ants Chapter 4

CHAPTER 4. ALTRUISM AND THE ORIGIN OF THE WORKER CASTE

By almost any conceivable standard the single most important feature of insect social behavior is the existence of the nonreproductive worker caste. The altruistic actions of this caste integrate the colony more tightly and make possible advanced forms of labor specialization. The baseline for the role of the worker is provided by the queen, which in most species still behaves in a primitive, totipotent manner resembling that of a solitary aculeate wasp. She alone traverses the whole life cycle of the species. Acting like a solitary insect, she leaves the mother colony, mates, and builds a nest. During this time her anatomy and physiology are essentially those of a solitary wasp, and her behavioral patterns are equally complicated. Only when the first brood of workers arrives does she become specialized, narrowing her repertory to an almost exclusively egg-laying role. In contrast, workers are specialized throughout their lives, with a large part of their repertory being devoted, start to finish, to the welfare of the queen and their siblings.

Altruism
Is it correct to call the nonreproductive workers “altruistic”? Some authors have begun to drop this admittedly value-laden word. They point out that the prescribing genes are selfish rather than altruistic, because if our conception of evolution by natural selection is true, it must follow by definition that genes persisting at the expense of others are selfish--even if they prescribe outwardly selfless behavior. Alternative expressions have been suggested for outwardly selfless behavior, such as social donorism (Williams and Williams, 1957), nepotism (Alexander, 1974) and reciprocation (Trivers, 1971). But if altruism is defined in the original lexical manner as self-denying behavior performed for the benefit of others, its application to ants and other social insects is justified, at least at the levels of the organism and colony. The key question, as we shall see, is how natural selection can produce selfish genes that prescribe altruism.

Altruistic behavior has been documented in multiple ways. To start, the great majority of ant workers make no effort to reproduce at all. Although the ovaries of young individuals are often active, the eggs they produce are more often than not trophic in nature, used to feed the larvae and queen, and unable to develop even if left unharmed. Older workers typically leave the nest to search for food outside the nest, where they find life very dangerous. Such foragers in the Idaho harvester Pogonomyrmex owyheei, which make up less than 10 percent of the worker population at any given time, undergo a weight loss of 40 percent and increased mandible wear. They are subject to intense predation and live an average of only 14 days after starting their forays (Porter and Jorgensen, 1981). In a study conducted by De Vita (1979), individual colonies of the California harvester Pogonomyrmex californicus were observed to suffer an average of 0.06 death per worker foraging hour due to fighting with neighboring colonies. The level of sacrifice while foraging approaches suicide in the formicine Cataglyphis bicolor, a scavenger of dead arthropods in the North African desert. At any given time about 15 percent of the workers are engaged in long, dangerous searches away from the colony, during which they are downed mainly by spiders and robber flies. They enjoy a life expectancy of only 6 days, but during that time each one retrieves food weighing 15 to 20 times its own body weight (Schmid-Hempel in Wehner et al., 1983; Schmid-Hempel and Schmid-Hempel, 1984). Porter and Jorgensen have referred to foragers in such extreme cases as constituting a "disposable" caste, since they are workers that exchange their lives for a high productivity on behalf of the colony.

The trade-off between individual sacrifice and colony welfare is even more clear-cut in the case of defense. Aging workers of the green tree ant of Australia (Oecophylla smaragdina), which are distinguished by reduced fat bodies and ovaries, emigrate to special “barrack nests” located at the territorial boundary of the colony. When Oecophylla workers from neighboring nests or other invaders cross the line, these guards are the first to rush to the attack (Hölldobler, 1983). (It can be said that a principal difference between human beings and ants is that whereas we send our young men to war, they send their old ladies.)

Altruistic behavior is sometimes accompanied by anatomical specialization. Workers of some Pogonomyrmex species possess reverse barbs on the sting, causing the venom apparatus and other portions of the viscera to come free when the ant moves away. This device appears to be used to defend the colony against vertebrates (Hermann, 1971). Sting autotomy or autothysis as it is sometimes called, also occurs in honeybees and in some genera of the social polistine and polybiine wasps, and thus constitutes a remarkable example of convergence in social behavior (Hermann and Blum, 1981). An even more bizarre suicidal defense is mounted by workers of a tropical Asian species belonging to the Camponotus saundersi group. The worker mandibular gland is hypertrophied, occupying not only a large part of the head capsule but extending all the way back to the tip of the abdomen. When the ants are sufficiently provoked, they contract their abdomens violently until their body wall bursts along the intersegmental membranes, discharging large quantities of the sticky mandibular gland secretion which traps the attackers (Maschwitz and Maschwitz, 1974; see Figures 7-25 and 7-26).

West-Eberhard (1979, 1981) has tempered the interpretation of worker altruism as the dominant mode of ant life by pointing out that competition among nestmates occurs more commonly than recognized in the earlier literature. This is particularly the case in those species in which workers are able to lay eggs and thus compete reproductively. In other words, workers are less than perfectly altruistic. Furthermore, natural selection at the level of the "selfish" individual might have played a role even in the evolution of division of labor, a social arrangement considered the exemplar of cooperation and harmony. West-Eberhard argues the case especially for the centrifugal pattern of temporal castes, in which older workers move away from the queen and brood and devote themselves more to outside work. The "selfish" worker, by staying close to the brood chambers while still young and while her personal reproductive value is highest, maximizes her potential to contribute personal offspring. As the ant ages and her fertility declines because of programmed senescence, her optimum strategy for contributing genes to the next generation is to enhance colony welfare through more dangerous occupations such as foraging and defense. This is an appealing hypothesis, and it should be kept in mind especially when considering ant species with more primitively organized, smaller colonies. But it loses most or all of its force in phylogenetically advanced species with extreme worker specialization for foraging and defense. In some species workers completely lack ovaries and hence are constitutionally exempt from individual selection. Examples include the members of Monomorium, Pheidole, Pheidologeton, and Solenopsis.

Kin selection
In order to explain the more clear-cut cases of worker altruism, it is necessary to turn to kin selection, which can be defined as the alteration of the frequencies of genes shared by relatives, through actions that favor or disfavor the survival and reproduction of the relatives. In other words, it is selection mediated by interactions among relatives. Kin selection is inferentially a powerful force in evolution. To take an extreme imaginary case, if an allele (that is, an alternate form of a particular gene) appears in a population that causes its bearer to act so as to triple the reproduction of one or more of the bearer's brothers and sisters, the allele will spread rapidly through the population. This will occur even if the allele-bearer completely sacrifices itself in the process, because many of its siblings also carry the altruistic gene. Thus undeniable altruistic behavior can become the norm in the population. Although the roots of the theory of kin selection go back to Darwin, it was developed most originally and forcefully by W. D. Hamilton in the 1960s (his now-famous seminal article was published in 1964) and then taken to new levels of sophistication by Bartz, Charlesworth, Charnov, Craig, Crozier, Feldman, Oster, Trivers, and others during the past twenty years. An especially important extension was made by Trivers and Hare (1976), who predicted a 3:1 ratio of energy investment in queens as opposed to males during the production of reproductives, at least in cases where workers control the sex allocation, a rare instance of a quantitative prediction in evolutionary biology. Progress has been such that kin selection in ants can be considered fundamental to general sociobiology. Kin selection is actually one of three hypotheses that have been erected at various times to explain the origin and evolution of eusociality. These hypotheses are: (1) Kin selection. By reducing personal survival and reproduction, workers nevertheless increase the survival and reproduction of genes they share with other members of the colony by common descent. Individuals suffer, but the colony flourishes and so do the genes (including altruistic genes) shared by common descent.

OR (2) Mutualism. In some fashion, individuals do better in personal survival and reproduction when they live in groups than when they live alone, even though they defer to other colony members and sacrifice on their behalf to some extent.

OR (3) Parental manipulation. One or both parents (actually, the mother in ants and other social hymenopterans) is able to neuter and control some of her offspring so as to produce a larger total number of offspring. The parents' personal fitness is raised even though that of some of the offspring is lowered. The case for kin selection can be most clearly made by contrasting it against the hypothesis of social mutualism first developed by Michener (1958) and subsequently elaborated by Lin and Michener (1972). The idea is simple to the point of seductiveness: cooperation will evolve if a group of individuals can reproduce better than a single individual under the same circumstances. Just this relation has been documented in bees and wasps, as follows: the larger the colony, the more nest cells, eggs, larvae, and pupae it produces. However, the relationship is not linear. As first noted by Michener (1964, 1969), the production of nest cells and immature forms per individual member falls off with an increase in group size. The relation, which for convenience can be called the reproductivity effect, appears to occur generally in the social insects (Wilson, 1971).

Why has social existence evolved at all if belonging to a group diminishes personal reproduction? The answer is clearly the enhancement of group survival as opposed to individual survival. If an insect has a longer life as a member of a group than as a solitaire the advantage can more than compensate for its decrease in reproductive potency. And again, the evidence demonstrates that such enhancement exists. In the paper wasp Polistes canadensis (Pickering, 1980) and the honeypot ant Myrmecocystus mimicus (Bartz and Hölldobler, 1982), average lifetime reproduction per individual, taken as the summed products of survival probability and reproduction in each interval of time, increases with group size during colony founding. There are at least two reasons why increased survival should compensate for lowered reproduction among insects. First, groups are able to fend off competitors and other enemies more effectively, an advantage that has been well documented in the literature (Wilson, 1975b, and Chapter 6 of this book). Second, as noted, it is far safer to stay at home than to forage. A large percentage of colony members are able to remain in the nest, whereas all solitary hymenopterans must forage. Thus the mutualism hypothesis by itself does appear adequate to account for the origin of colonial existence. However, one of the three diagnostic traits of eusociality, the existence of a sterile worker caste, cannot be explained in such a manner. This is why kin selection has risen to importance in evolutionary theory. This concept was actually originated in a very general form, by Charles Darwin in The Origin of Species. Darwin, whose interest in social insects was strong in his later life (he consulted the early myrmecologist Frederick Smith at the British Museum about slave-making ants) had found in them the "one special difficulty, which at first appeared to me insuperable, and actually fatal to my whole theory." How, he asked, could the worker castes of insect societies have evolved if they are sterile and leave no offspring? This paradox proved truly fatal to Lamarck's theory of evolution by the inheritance of acquired characters, for Darwin was quick to point out that the Lamarckian hypothesis requires characters to be developed by use or disuse of the organs of individual organisms and then to be passed directly to the next generation, an impossibility when the organisms are sterile. To save his own theory, Darwin introduced the idea of natural selection operating at the level of the family rather than of the single organism. In retrospect, his logic seems impeccable. If some of the individuals of the family are sterile and yet important to the welfare of fertile relatives, as in the case of insect colonies, selection at the family level is inevitable. With the entire family serving as the unit of selection, it is the capacity to generate sterile but altruistic relatives that becomes subject to genetic evolution. To quote Darwin, "Thus, a well-flavoured vegetable is cooked, and the individual is destroyed; but the horticulturist sows seeds of the same stock, and confidently expects to get nearly the same variety; breeders of cattle wish the flesh and fat to be well marbled together; the animal has been slaughtered, but the breeder goes with confidence to the same family" (The Origin of Species, 1859: 237). Employing his familiar style of argumentation, Darwin noted that intermediate stages found in some living species of social insects connect at least some of the extreme sterile castes, making it possible to trace the route along which they evolved. As he wrote, "With these facts before me, I believe that natural selection, by acting on the fertile parents could form a species which regularly produce neuters, either all of a large size with one form of jaw, or all of small size with jaws having a widely different structure; or lastly, and this is the climax of our difficulty, one set of workers of one size and structure, and simultaneously another set of workers of a different size and structure" (The Origin of Species, 1859: 24). Darwin was speaking here about the soldiers and minor workers of ants.

Although J. B. S. Haldane, the noted British population geneticist, pointed out the implications of Darwin's insight, the modern genetic theory of kin selection and sterile castes was inaugurated by Hamilton. He recognized that there are two ways for alleles (alternative forms of a gene found on the same locus) to be passed to future generations. The first is by personal reproduction, in other words the production of sons and daughters. The measure of personal reproductive success, which monopolized the earlier theoreticians of population genetics, has come to be known as classical fitness. The second mode of gene descent is collateral, promoting the welfare of brothers, sisters, and other relatives besides offspring who possess the same alleles by reason of common descent. Hamilton recognized the importance of a measure he called inclusive fitness, which incorporates both the individual's personal reproduction (classical fitness) and its influence on the reproduction of collateral relatives. To avoid confusion, it is best to use the expression "kin selection" to refer to circumstances in which the reproduction of collateral relatives is involved. Yet the complete effects of kin selection on evolution cannot be evaluated without including the effects on personal reproduction. This notion, and the terminology expressing it, has been put in the most nearly standard form by Pamilo and Crozier (1982) and Pamilo (1984a,b). The ordinary (non-social) measure of classical fitness is: E(RS) W = 	_________________________ Average RS for Population where E(RS) is the average direct reproductive success of individuals possessing the genotype of interest. It measures the number of offspring the individual injects into the population, in comparison with the contribution from the remainder of the population. This is the most common measure of fitness encountered in the literature of population genetics. Inclusive fitness, on the other hand, incorporates two components: E(RS) + ∑[bjE(RS)] IF =	_________________________

Average IF for Population where the second term, ∑[bjE(RS)], is the effect on the reproduction of all of the collateral relatives. The quantity bj is the coefficient of relatedness, the probability that the relative j of the focal individual also possesses the allele of interest. In ordinary diploid systems, for example, bj is 1/2 for brothers and sisters; 1/4 for uncles, aunts, grandparents, and grandchildren; and 1/8 for first cousins. Outside this tight circle of close relatives, bj continues to fall swiftly, so that kin selection becomes a proportionately negligible force.

After Hamilton's original formulation, a great deal of confusion arose over the best definition and usefulness of measures of relatedness, but this appears now to be largely resolved. Michod and Hamilton (1980) demonstrated that the five principal different coefficients of relatedness invented to account variously for different degrees of penetrance, gene frequency, and inbreeding, are equivalent. Seger (1981) generalized the result to cases in which the average gene frequencies of the altruists differ from those of the recipients. The best intuitive way to think about the degree of relatedness and to approximate the coefficient of relatedness is to ask the following question: if the focal individual has an allele a, what is the chance that one of his relatives also possesses it? The pivotal idea can now be put as follows. If the allele a affects altruism in some manner, then self-sacrificing propensities have the potential to evolve. The allele is always "selfish"; it spreads through the population by promoting itself by means of the increased success of collateral relatives. The necessary minimal condition for this to happen is stated by "Hamilton's rule": C ____ <  b

B This says that the cost C (which is the loss in expected personal reproductive success due to the self-sacrificing behavior) divided by the benefit B (the increase in the relatives' expected reproductive success) must be less than b, the probability that the relatives have the same allele. Another way of expressing Hamilton's rule is to say that the benefit to relatives is discounted by their degree of relationship, so that the less the relatedness, the greater the benefit must be to counterbalance the cost. Consider, for example, a highly simplified network consisting solely of an individual ant and her sister. If the focal ant is altruistic she will perform some sacrifice for the benefit of her sister. She may share food, labor more in nest construction, or place herself between the sister and some enemy. The important consequence for the focal ant, from an evolutionary point of view, is her loss of personal genetic fitness, due to a reduced life span, fewer offspring, or both, leading in turn to less representation of the altruist's personal genes in the next generation. But three-fourths of the sister's genes are identical to those of the altruist by virtue of common descent. Suppose, to take the extreme case, that the altruist leaves no offspring. If her altruistic act enlarges the sister's personal representation in the next generation to a sufficient degree (in this case, by more than 50 percent), it will increase the three-quarters of the genes identical to those in the altruist, and the altruist will actually gain representation in the next generation. Some of the genes shared by such sisters will be the very ones that encode the tendency toward altruistic behavior. The inclusive fitness, in this case determined solely by the sister's contribution will be sufficient to cause the spread of the altruistic genes through the population. This general result can be changed somewhat by certain restrictive conditions, but in general it is surprisingly robust. One way to bend it a little is to regard the costs and benefits as multiplicative or "interactive" instead of additive (Uyenoyama and Feldman, 1981). In the original additive model IF = 1 - C + bB This quantity must be greater than one if the allele of interest is to spread, or 1 - C + bB > 1 which is rearranged to produce Hamilton's rule: C ____ < b

B In the interactive model proposed by Uyenoyama and Feldman the components are multiplied together: IF = (1 - C)(1 + bB) In this case the fitness is greater than one and the altruism allele will spread if C ______ < b B(1-C) The interactive model is more restrictive than the additive. In other words, some conditions exist under which the altruism allele would spread under the additive relation but not under the multiplicative model. However, the difference is not great, and it is minimal when the costs and benefits are both low. Furthermore, the additive model appears intuitively preferable to the interactive model. The cost to the altruist and a benefit to the relative are distinct events, unlike many interactive physiological processes within the same organism, and it is difficult to imagine how they could operate on inclusive fitness in a multiplicative way (Bartz, 1983). In short, Hamilton's rule is robust as a theoretical proposition.

Parental manipulation
Although kin selection successfully reaches beyond social mutualism to explain the origin of sterile workers, it is not the only conceivable explanation. Several writers have noted that a worker caste can arise if the mother is forceful enough to dominate and manipulate her own offspring. In other words, if the mother can rear enough additional daughters (b = 1/2) by enslaving some of them, it is to her advantage to sterilize them rather than let them depart to create a larger crop of granddaughters (b = 1/4). Such an arrangement, with sisters or other co-generational females enslaving each other, was proposed for example to explain the origin of eusocial bees by Michener (1974) and Michener and Brothers (1974). These authors noticed that queens of the primitively eusocial bee Lasioglossum zephyrum control other adult females by two simple behaviors. Other adult females are systematically nudged, an act that appears to be aggressive in nature and may have the effect of inhibiting ovarian development. The individuals most frequently nudged are the ones with the largest ovaries, and hence most able to compete with the queens. Nudging is followed by backing, in which the nudger retreats down the nest galleries, apparently attempting to draw the other bee after it. The effect is to maneuver the follower closer to the brood cells, where it can assist in the construction and provisioning of the cells used by the queens. It is not difficult to imagine, with Michener and Brothers, that sterile castes can evolve if certain allelic combinations arise that are very powerful in controlling nestmates. Alexander (1974) independently suggested that the exploitation of offspring by their parents has been a general force in the social evolution of insects. Any conflict between parents and their children, he argued, is likely to be resolved in favor of the parents, who are bigger, stronger, and more forceful in any episode of conflict. To summarize, we have two strong competing hypotheses for the origin of the sterile castes and, following that, the more complex forms of social organization. Which one is correct? Were the parents (meaning, in the case of the social Hymenoptera, the mothers) molded by natural selection to enslave their offspring? Or were the offspring shaped by natural selection to be willing helpers--or even further, to manipulate their mother in order to produce more brothers and sisters? Some potent logical arguments have been raised in favor of the parental manipulation hypothesis. One stresses the fact that some of the offspring become workers while others become queens. Such plasticity might suggest control on the part of the queen, because if there is no intrinsic advantage to a female to be a worker, all should choose to be queens. However, this argument is blunted by the evidence, examined in Chapters 3 and 8, that the ergonomic phase of colony growth postpones queen production only to allow a very large crop at a later date. The outcome serves both the classical fitness of the mother queen and the inclusive fitness of the workers.

A second argument, advanced by Charlesworth (1978), Charnov (1978), and Craig (1979) in connection with models for the spread of rare alleles for eusociality, identifies a circumstance that appears to make daughter enslavement easier. It goes as follows: from the mother queen's point of view, the offspring need only be half as good in raising their brothers and sisters as would be required if allowed to raise their own sons and daughters. The reason is that the queen is related to her own offspring (the brothers and sisters of the enslaved workers) by 1/2, but she is related to her grandsons and granddaughters (the would-be offspring of the workers) by only 1/4. As a consequence she can afford to degrade her offspring in the process of making them into workers, for example by starvation or hormone-mediated sterilization, even if it reduces their competence as helpers. Hence the matrix of relatedness favors the origin of a worker-caste by the evolution of manipulative behavior on the part of the queen--all other things being equal. The difficulty with the argument, and the parental manipulation theory generally, is that all things are not equal. If offspring have lower inclusive fitness as a consequence of being workers, any allele that prescribes resistance to the queen's machinations would be favored. Hence parental manipulation can be invaded by genes that prescribe its reversal as a social trend. On the other hand, as Craig has pointed out, mothers can also manipulate their offspring so that it is to the offspring's best interest to stay around and help. Parental manipulation of this kind devolves to kin selection, and brings us to the possibility that offspring "consent" to become workers.

Offspring consent
The opposing view to crude parental manipulation is of course offspring consent. Worker castes arise and are maintained because under certain circumstances inclusive fitness is enhanced by surrendering reproduction. Hamilton's original formulation of this idea permitted a surprisingly detailed prediction of some hitherto unexplained features of behavior in ants and other social Hymenoptera. Its explanatory power gave the model a great deal of initial appeal and, more generally, launched kin selection as an important idea in general sociobiology. The key step was the connection between the haplodiploid method of sex determination, which is universal in the Hymenoptera, and the remarkable prevalence of eusociality in this order. Haplodiploidy is the mode of sex determination in which males are derived from unfertilized (haploid) eggs and females from fertilized (diploid) eggs. The ultimate basis, however, is not the mere presence or absence of chromosomes but rather one or more sex-determining genes they bear. Haplodiploidy has a number of odd effects (Andersson, 1984). The parthenogenetic origin of hymenopteran males means that all alleles will be expressed in a homozygous condition (or, more precisely, hemizygous condition). As a result lethal and subvital alleles will be exposed each generation, and total genetic variability in the population will tend to be reduced. Hymenopteran species have between one-tenth and one-half the heterozygosity per individual as occurs in non-social insects. Also, the more advanced social hymenopteran species have less heterozygosity than do the solitary ones, possibly due to a higher degree of inbreeding caused by reduced gene flow among the population of colonies (Graur, 1985). However, the negative effects of haplodiploidy are true only for genes that are expressed in the male. Those limited in expression to female characters are theoretically expected to behave as though they existed in wholly diploid populations, enjoying the same potential variability and obeying the same equilibrium laws (Kerr, 1967). Another curious effect of haplodiploidy is that characters that are both under polygenic control and not sex limited should be more variable among males in sibling groups than among females. In fact, under the simplest possible conditions (panmixia and an absence of dominance and epistasis), the theory of polygenic inheritance predicts a genetic variance in males four times that of their biparental sisters. Since most characteristics are under polygenic control, it should therefore be a rule that males are more variable than virgin queens collected from the same colony. Kathleen Eickwort (1969) found this proposition to be true for ten external morphological characters which she measured in the paper wasp Polistes exclamans.

But the strangest of all consequences of haplodiploid sex determination is the asymmetries it creates in the relatedness among close relatives (Figure 4-1). The coefficient of relationship between sisters is 3/4, whereas between mother and daughter it is 1/2, the same as in diploid organisms. Sisters are exceptionally close because they share all of the genes they receive from their father (since their father is homozygous and produces genetically uniform spermatozoans), and they share on the average of one-half of the genes they receive from their mother. Each sister receives one-half of her genes from her father and one-half from her mother, so that the probability that a gene possessed by the focal females will be shared by a sister is From	From	Shared

father	mother	overall (1 x 1/2) + (1/2 x 1/2)   =    3/4 Hamilton reasoned that when the mother lives as long as the eclosion of her female offspring, her daughters can increase their inclusive fitness more by caring for their younger sisters than by caring to an equal degree for their own offspring. During the following twenty years that followed Hamilton's publication, the basic model received two major adjustments that were to strengthen its precision and hence its falsifiability. The first, by Trivers and Hare (1976), noted that whereas the advantage accruing to the care of sisters is inevitable, it is counterbalanced by an equal disadvantage that results from the rearing of brothers. This is due to the fact that brothers and sisters are related by only 1/4. The male comes from an unfertilized egg, whereas his sister comes form a fertilized egg. The coefficient of relatedness between the male and the haploid egg that produces his sister is 1/2. This is diluted when a sperm is added to the female-destined egg by an outside male, such that the relation between the female and her brother drops to 1/4. More precisely, the probability that an allele chosen in a focal female will also be present in the brother by immediate common descent is 1/4. In ordinary diploid organisms with a 1:1 sex ratio the degree of relatedness between a focal female and all her sisters and brothers is Related to	Related to	Related sister	brother	overall (1/2 x 1/2) +  (1/2 x 1/2)    =    1/2 This turns out to be identical to the haplodiploid system when that system includes a 1:1 sex ratio, in spite of the asymmetries in relationship: Related to	Related to	Related sister	brother	overall (3/4 x 1/2) +  (1/4 x 1/2)    =    1/2 Trivers and Hare pointed out that the advantage to rearing sisters as opposed to daughters could nevertheless be restored if the ratio of investment (approximated by dry weight) is altered to 3:1 in favor of sister production over brother production. The 3:1 ratio should be at equilibrium because the expected reproductive success of the males will then be three times that of the queen on a per-gram basis, balancing the one-third initial investment. This important result can of course be tested, one of the few cases in which evolutionary theory actually predicts a specific quantity rather than merely a trend or inequality. The substantial amount of research it has stimulated appears to favor the kin selection hypothesis, a conclusion we will discuss in more detail shortly.

The second adjustment deserving special mention is due to Bartz (1982), who deduced that in a haplodiploid system it is theoretically possible to evolve either female workers or male workers but not both. As shown in Figure 4-2, male workers will be favored only if half or less of the reproductive investment made by the colony results in females, while at the same time the queen produces a large fraction of the males. Furthermore, the less the queen contributes to male production, the higher the proportion of females in the reproductive brood must be in order to compensate. "Bartz's rule" is in close enough accord with the facts to support the kin selection hypothesis. Pamilo (1984a) has confirmed the main result in a separate analysis, but he notes that it depends upon a population-wide sex ratio of 0.5. When the population of colonies deviates from this value, the result is less certain.

Testing the kin-selection theory
Few ideas in biology have been probed more aggressively and from so many directions as the theory of kin selection in the eusocial insects. The testing has been made still more rigorous by the presence of a strong alternative explanation of the evolution of eusocial behavior, namely the hypothesis of parental manipulation. The key critiques and reviews include those of Alexander and Sherman (1977), Andersson (1984), Bartz (1982, 1983), Craig (1979, 1980), Crozier (1977, 1979, 1982), Michod (1982), Page (1986), Pamilo (1982a, 1984a,b), Starr (1979), West-Eberhard (1982), and Wilson (1975b). How well has kin selection held up? In the sections to follow we will examine the most important features of ant biology that bear on the origin of the nonreproductive caste and attempt to weigh the empirical evidence accumulated to the time of writing. We will come out on the side of kin selection, but cautiously--because some new twist in theory or important empirical finding might yet overturn it. Which insects have become eusocial? The kin selection theory predicts that haplodiploid insects should show a higher incidence of eusociality than completely diploid insects. This appears to be the case in truth. Haplodiploidy is a characteristic of all of the Hymenoptera but is shared by only a few other arthropod groups, including certain mites, thrips, and whiteflies, iceryines and possibly other scale insects, and the beetle genera Micromalthus, Xylosandrus, and, perhaps, Xyleborus. Outside the Arthropoda, some nematodes and most rotifers are also haplodiploid. At the same time true eusociality is very nearly confined to the Hymenoptera. It has originated at least twice in the wasps, more precisely at least once each in the stenogastrine and vespine-polybiine wasps and probably a third time in the sphecid Microstigmus, eight or more times in the bees, and at least once or perhaps twice in the ants. In summary, it is evident that eusociality has arisen at least eleven times independently in the Hymenoptera. Quite probably this lower estimate will increase with the growth in knowledge of hymenopteran biology, especially that of tropical bees. Yet throughout the entire remainder of the Arthropoda, true eusociality is known to have originated in only one other living group, the termites or order Isoptera. Among the rest of the higher animal phyla, it is known only in the remarkable African mole rat Heterocephalus glaber (Jarvis, 1981). Aphids have evolved nonreproductive soldier castes no fewer than four times independently. They have also achieved an overlap of generations--but not cooperative brood care (Aoki, 1977, 1982, 1987). This dominance of the social condition by the Hymenoptera cannot be a coincidence. Of the 751,000 living insect species described to 1985 (Arnett, 1985), only about 103,000 or 14 percent belong to the Hymenoptera. Something close to this partition has persisted throughout at least the Cenozoic, further diminishing the possibility that the bias can be explained as a mere historic accident.

Several leading hymenopterists, namely Michener and Brothers (1974), West-Eberhard (1975), and Evans (1977a), have questioned the primacy of kin selection even while acknowledging its importance. They have urged the equivalent importance of at least two other preadaptations that exist in the Hymenoptera: the universal existence of mandibulate mouthparts and the frequent building of nests, especially in the aculeate wasps (as opposed to parasitoid wasps) and bees. Their case with reference to nest building is quite strong. Starr (1985) has added a third preadaptation: the aculeate hymenopteran sting, which is an effective defense against vertebrate predators. Only between 50,000 and 60,000 aculeate hymenopteran species have been described, yet they include all of the eusocial insects with the exception of the termites. The sting was needed in the earlier stages of eusocial evolution, Starr argues, because groups of individuals are generally more vulnerable to large predators that preferentially seek aggregations of prey and have the size and strength to overcome them. This seems to us to be a much less persuasive argument, but it cannot easily be rejected or confirmed. What is needed is a kind of accounting system in the origin of eusociality, perhaps in the form of a multiple regression equation that incorporates each of the factors including kin selection in order to provide a more precise measure of the inclusive fitness of the would-be queens and workers. When the averaged relative fitnesses rise to a certain "eusociality threshold" (Wilson, 1976e), the species is likely to evolve all three of the basic traits of a higher social (eusocial) insect, that is, cooperative brood care, overlap of at least two generations, and division of the group into reproductive and sterile castes. It would appear that two preconditions carried the eleven hymenopteran phyletic lines across the eusociality threshold. The first is the enhancement of kin selection by haplodiploidy. Mandibulate mouthparts and nest-building are not enough by themselves. Other kinds of arthropods, including many beetles, spiders, and orthopterans, build nests, manipulate objects skillfully with their mandibles and legs, and care for their young, occasionally in elaborate fashion, but none has attained the eusociality threshold. It is equally true that haplodiploid enhancement is insufficient on its own, because it occurs in other, nonsocial arthropod species, including all of the symphytan and most of the parasitoid hymenopterans as well as a majority of the bees and aculeate wasps. (Diploid males have been reported in the parasitoid genera Bracon, Nasonia, and Neodiprion, as well as Apis, Bombus, Melipona, and Trigona; see Page and Metcalf, 1982.) Do males ever serve as castes? In a manner consistent with Bartz's argument (see Figure 4-2), the worker caste of ants is universally female. Alexander (1974) and a few other writers have argued in contrary fashion that hymenopterous males, being stingless and otherwise highly specialized for reproduction throughout the order, simply are not able to evolve a worker-like anatomy and behavioral repertory. Hence phylogenetic inertia, rather than kin selection, can account for their lack of involvement. This reasonable-sounding explanation is considerably vitiated by the evidence of substantial evolutionary lability among male ants. In at least two species males have assumed a partial worker role, under special environmental circumstances that indicate that when selection pressures are strong enough to countervail the conventional components of inclusive fitness, evolution away from strict male idleness does occur. Males of the carpenter ant Camponotus herculeanus are exceptionally long-lived, because they are produced in the fall, overwinter in the nest, and if kept sufficiently cool through the following spring and summer, live on through a second annual cycle. Unlike most other ant males studied to date, they store food in their crops and regurgitate some of it back to workers and other males, thus participating in a key homeostatic role for the entire colony (Hölldobler, 1964, 1966). Male larvae of the weaver ant Oecophylla longinoda have well-developed silk glands. Like the worker-destined female larvae, they contribute the silk to nest construction rather than to construction of their own cocoons. Because the behavior is very specialized and clearly derived in evolution with reference to the ants as a whole, participation in nest building represents an important shift on the part of the males toward a new social role (Wilson and Hölldobler, 1980). Male anatomy is at least as malleable as behavior.

The males of Formicoxenus and some species of Hypoponera, Cardiocondyla, and Technomyrmex are wingless and “ergatomorphic,” convergent to the worker of the species in overall body form (see Figure 12-18). The modification appears to be associated with the loss of between-colony dispersal by the males and a tendency toward pairing with female nestmates (Le Masne, 1956a; Terron, 1972b; Hamilton, 1979). In some species of Cardiocondyla, such as Cardiocondyla papuana and Cardiocondyla wroughtonii, they have evolved saber-shaped mandibles (Kugler, 1983; see Figure 4-3). The ergatomorphic males of Cardiocondyla wroughtonii fight among themselves until only one remains in the colony (Stuart et al., 1987a,b; Yamauchi and Kinomura, 1987). Fighting ergatoid males have also been recorded in the small ponerine Hypoponera punctatissima (Hamilton, 1979). However, no evidence has ever been adduced of participation of ergatomorphic males belonging to the three genera in any worker-like social function. Finally, it is probably relevant that termites, which are not haplodiploid, have both male and female workers. The social status of males is complicated somewhat by the coexistence in a few species of two morphological forms, the small “micraner” and the large “macraner.” The males of Solenopsis invicta have a bimodal size-frequency distribution; the small individuals are haploid and the large males are diploid (Ross and Fletcher, 1985b). Size dimorphisms also exist in Formica naefi (Kutter, 1957), Formica exsecta (Pamilo and Rosengren, 1984), Formica sanguinea (Agosti and Hauschteck-Jungen, 1988), and species of the Rhytidoponera impressa complex (Ward, 1983a), but the genetic basis has only been slightly studied. In the case of Formica exsecta at least, the two types are not haploid and diploid respectively. However, the micraners have a somewhat lower percentage of haploid brain cells, while chromosome numbers higher than 2N occur only in micraners (Agosti and Hauschteck-Jungen, 1988). The behavioral and ecological significance of the size dimorphism in all of the species remains unknown. Do males suffer higher mortality? Smith and Shaw (1980) have pointed out that because of their homozygous state, hymenopterous males always express lethal and subvital alleles in their genome and hence inevitably suffer higher mortality than the females. Even if selection has completely "cleaned out" such alleles by passing them through the male haploid filter, they will still appear each generation in very low levels due to the occurrence of new mutations. Hence the biased sex ratio attributed to kin selection might be due in part to new lethal mutations, provided the differential mortality occurs during or after the stage at which parental care is invested. The effect can be expressed as the probability that a male dies either due to a new mutation in its own soma or because of a new mutation picked up from the mother. This probability is P = 1 - exp(-3m) where m is the mutation rate per genome (not colony) genome per generation. The effect on the female:male ratio is displayed in Figure 4-4. The value of m for parasitoid wasps in the genus Apanteles is 0.035, giving a differential mortality of 10 percent. If a similar value holds for ants, the mutation-induced distortion of the ratio of investment would only be a small fraction of that expected from kin selection. The effect is further attenuated by the fact that the equation applies to the ratio of individuals, rather than to the ratio of biomass. Since individual queens are usually much heavier than males at eclosion, and receive proportionately larger amounts of food prior to the nuptial flight, the mortality ratio translates to a narrower biomass ratio. The importance of the equation is that it forms a baseline or null hypothesis against which kin selection can be more reliably tested. It is rendered much less tractable, however, by the fact that m must be considered separately for each phylogenetic group, and no values are yet available for ants. How closely related are the colony members? Here we encounter a potential difficulty of considerable significance for the kin selection hypothesis. Four key factors can be distinguished that determine relatedness within a colony: the number of laying queens, the degrees of relatedness among the laying queens, the number of males with which the laying queens mated, and the intensity of egg-laying by the workers. In some species a fifth, at least episodic factor is queen succession, which occurs in species that reproduce by colony fission and queen supercedure. Prominent examples include ecitonine army ants. John Tobin (personal communication) argues: "When colony fission occurs one of the daughter colonies will be headed by the mother queen.  However, the other daughter colony (or colonies, if there are more than two daughter colonies after fission) will get a new queen that is a sister of the workers, rather than of the mother; the effect will be that the workers in this colony will be raising not their sisters (or half-sisters) but their nieces (or half-nieces).  As the younger generation of workers (the daughters of the new queen) replaces the older generation (the daughters of the original queen), average intracolony relatedness will decrease, and will reach a minimum when the proportion of the two generations, or subfamilies, of workers is 1:1.  As the older workers continued to die away, average relatedness will increase back towards the pre-fission levels."

In Table 4-1 we have summarized the information collected to date on the degrees of relatedness. These data are mostly estimates based on the electrophoretic separation and identification of allozymes (different forms of the same enzyme) encoded by multiple alleles and treated as representative samples from the larger genotype. One important finding is the distant relatedness among queens in polygynous colonies. In Myrmecia pilosula and Formica sanguinea, the species for which data are available, the queens are less closely related than full sisters; in other words their b values are significantly below the theoretically expected 0.75. They are nevertheless as close as half sisters or first cousins. The other measurements taken so far match expectations. In strains of Rhytidoponera chalybaea and Rhytidoponera confusa that possess only one singly-mated queen per colony, b among workers does not depart significantly from 0.75, as expected. In queenless colonies of the same species with fertile workers serving as reproductives, the worker-worker b values are highly variable from one colony to another but always well below 0.75, as expected from the diverse provenance of the worker population in each nest. The difficulty raised for the kin selection theory by the estimates in Table 4-1 is the following. In polygynous colonies the degree of relatedness between selected individual workers and the cogenerational females and males in the same nest is less than would be the case if the individual workers had produced their own sons and daughters. If kin selection is a powerful force, what prevents evolution from leading to a more competitive state in which the workers (which have ovaries) try to take over reproduction? One obvious explanation is that polygynous colonies are so successful as smoothly operating units as to raise the inclusive fitness of focal workers above that prevailing if workers were egg layers. Just such a dissolution has occurred in some phyletic lines. All species of Diacamma, Dinoponera, and Ophthalmopone thus far investigated, as well as some species or genotypes in Rhytidoponera and Pristomyrmex, have discarded the queen caste and reverted to reproduction by workers (Peeters and Crewe, 1986b). In some cases a large percentage of the workers lay viable eggs during early stages of their lives. From one point of view, expressed with reference to Pristomyrmex pungens by Itow et al. (1984), the worker caste has been lost. It can even be questioned whether the colonies are classifiable as eusocial.

Another escape from the difficulty posed by polygyny is provided if workers in polygynous colonies could be shown to distinguish close kin from more distant kin. This ability has been demonstrated in social bees and ants, a subject to be reviewed in Chapter 5. Is the discriminatory activity strong enough to divide a polygynous colony into cliques of closely related individuals? All that would be required is for individual workers to favor their siblings over half-sisters and more distantly related nestmates among the reproductive larvae. Whether this occurs in ants remains to be learned. The occasional occurrence of distant relatedness raises questions about the origin of eusociality in ants, because some species of the very primitive ant genera Amblyopone and Myrmecia are polygynous. If true polygyny were the primitive condition for the Formicidae, it is difficult to see how the simple model of kin selection can explain the origin of eusociality. We clearly need many more data on the reproductive structure and relatedness among the colony members in primitive ant species in order to help resolve these important questions.

Do males from the same colony compete? Hamilton (1967) pointed out that another selection force that might bias sex allocation ratios in favor of new queens is local mate competition. Alexander and Sherman (1977) agreed and argued that this might be a more harmonious explanation than kin selection, when combined with control on the part of the queen. If a queen's sons habitually inseminated their own sisters, the queen could maximize her own fitness by producing only enough sons to ensure the insemination of all her daughters. In fact, Hamilton showed that such extraordinary sex ratios occur elsewhere in the Animal Kingdom. The extreme cases include certain mites and parasitoid wasps, where total incest is accompanied by very high female:male ratios. However, the local mate competition hypothesis is not favored by the natural history of most ant species, whose reproductive forms disperse widely and often form large mixed swarms assembled from hundreds or thousands of colonies (Hölldobler and Bartz, 1985; Woyciechowski and Lomnicki, 1987). Among the several species studied so far that possess large colonies and limited dispersal, mating has been shown by allozyme analysis to be either random (in Formica sanguinea and Formica transkaucasica) or to depart from randomness (in Formica pressilabris). No information is yet available concerning the degree of mating within single nest units (Pamilo, 1982c). Of equal importance, Nonacs (1986a) has systematically tested kin selection in opposition to local mate competition by examining the various phenomena in which they predict different outcomes. Kin selection is throughout the more consistent. In particular, the numerical sex ratios (as opposed to the biomass sex ratios) are close enough to 1:1 to be inconsistent with local mate competition, while the biomass sex ratios (as opposed to numerical ratios) are close enough in monogynous species to 3:1 to be harmonious with kin selection. Also, the biomass ratios of parasitic species, in which inbreeding does occur, is close to the 1:1 value expected from kin selection and not consistent with local mate competition. Finally, the evidence is overwhelming that ant males are capable of mating only once or at most two or three times. Furthermore, in some species of ants the queen requires the output of multiple males to fill her spermatheca (see Chapter 3). In short, the conditions do not exist for male ants to inseminate numerous closely related monoandrous females. How reliable are biomass measurements? Reproductive forms have been collected, dried, and weighed with little reference to the timing in the adult phase of the life cycle. Ignoring the life cycle is risky, because queens (gynes) and males undergo major body alterations from the time they eclose to the moment of the nuptial flights. Even worse, the time courses of the two sexes are exactly opposite: queens are fattened by the workers and grow steadily heavier, while males fast and grow lighter. Hence it makes a considerable difference as to when the two sexes are measured. In the European mound-building formicine Lasius flavus, for example, individual queens triple their weight during the first six days after emerging from the pupa. At the same time their fat content rises from 20 percent to 60 percent, while their water content falls from 75 percent to 45 percent (Nielsen et al., 1985a,b). In addition, the respiratory rate doubles. Except for their falling weight, males of the same species show little change during the same period. In addition, the weight-specific respiratory rate (µl O2/mg dry weight/hour) of queen-destined larvae is higher than in worker-destined larvae (Peakin et al., 1985). Similar results have been obtained independently for Lasius niger by Boomsma and Isaaks (1985; see also van der Have et al., 1988).

Ideally, then, measures of energy investments should include both construction (that is, growth in biomass) and maintenance (respiration during the tenure in the nest) over the life spans of the two sexes. This has been achieved in only one instance of which we are aware. In their audit of Lasius niger, Boomsma and Isaaks estimated that it takes 689 joules to make a queen and 90 joules to make a male. They found that the population means of energy investment in queens is lowered by the adjustment but remain in good agreement with the theoretical kin selection optimum of 3:1. The closest fit was in colonies taken from environments considered most favorable for Lasius niger. We suggest that failing exacting metabolic studies of this kind, the best time to collect data is in the middle period of adult life, when the queens have been fattened and the males still possess most of their reserves and original body weight. Failing even that, the aim should be to collect large enough samples to "wash out" the differences in the time courses of energy investment in females and males. Most investigators of the subject have in fact attempted to follow this second procedure. Is there sperm competition? Multiple matings are widespread in the ants, occurring most commonly in species with large mature colonies (Cole, 1983b). If the sperm of the two or more males are used extensively to fertilize the eggs, the degree of relatedness among the workers and between the workers and newly produced sexual forms will be diminished accordingly. If on the other hand there is sperm precedence, so that the sperm of one male or another dominates in fertilization, at least for intervals as long as the average longevity of females, the effect will be negligible. The relatively sparse evidence accumulated to date for solitary Hymenoptera and honeybees and Polistes wasps (reviewed by Crozier and Brückner, 1981; Page, 1986) indicates that sperm mixing is a general phenomena, although some bias in utilization occurs. Bias occurs in ants of the Formica rufa group, although whether through true competition or precedence is not yet known (Pamilo, 1982c). Who lays the eggs? The provenance of the eggs, whether from one queen, multiple queens, workers, or some numerical combination of these alternative sources, makes a great difference in the organization of the ant colony and the degree of relatedness of its members. The available data on worker-produced offspring in species with queens or ergatogynes (intermediates between queens and workers) are presented in Table 4-2. This information needs to be treated cautiously. Queens are sometimes difficult to find even when present, because they often crouch in hidden recesses of the nest. Also, the immature stages are sometimes prolonged by diapause. As a result it is easy to misidentify queen-produced workers and males as worker-produced. The result has been some confusing controversies in the literature (for recent reviews see Choe, 1988, and Bourke, 1988). For example, in proposing a cycle évolutif in the African weaver ant Oecophylla longinoda, Ledoux (1949, 1950) reported that many new colonies are started by workers that leave the territory of the mother colony and start colonies on their own. Some of the workers then lay small eggs 0.6 mm in length, most of which develop into workers and a few into queens. Ledoux suggested that the parthenogenesis in this case is apomictic, resulting in diploid eggs and that the eggs are small because they are ejected before normal meiosis can begin. However, Way (1954a), working in the field with free colonies, and Crozier (1971) and Wilson and Hölldobler (1980), experimenting on laboratory colonies, could find no evidence of thelytoky in workers. Groups of workers separated from the queen produced only males.

The general picture so far is that reproduction by workers is common but far from universal, and in most instances it is limited to the production of males in compliance with the ordinary working of haplodiploid sex determination. Thelytoky is relatively rare and it is often an infrequent facultative process in the species that display it. When Haskins and Enzmann (1945) made a careful attempt to obtain offspring from virgin queens of Aphaenogaster rudis, only 18 of 100 such individuals reared brood (mostly male) to the pupal stage, and only two females were brought to maturity among them. Bier (1952) found that a similar difficulty in rearing workers from worker-laid eggs of Lasius niger is due to the fact that the great majority of larvae coming from such eggs are actually male-determined and die at any early age. Eggs that are female-determined are viable but constitute only a tiny fraction of the total number laid. In spite of the considerable uncertainties in most cases of reported thelytokous parthenogenesis in ant workers, there exist at least two studies where worker thelytoky appears to be clearly demonstrated. One is Cataglyphis cursor where Cagniant (1979, 1982) and Lenoir and Cagniant (1986) provided solid experimental evidence for worker thelytoky, and the other is Pristomyrmex pungens where worker thelytoky seems to be the predominant mode of reproduction (Itow et al., 1984; Tsuji and Itô, 1986). In both cases the experiments were carefully controlled: the workers were reared from pupae and kept isolated from males. In addition histological and macroscopic dissections were employed to investigate possible inseminations, but no signs of inseminations were revealed (Suzzoni and Cagniant, 1975; Tsuji, personal communication). Male production by workers occurs mostly in queenless colonies, or, more precisely, orphaned colonies. Even when it is accomplished in the presence of the queen, the rate of production is usually small and limited to special circumstances. For example, workers of Odontomachus haematodus and Hypoclinea quadripunctata produce males only when separated to some degree within the nest from the mother queen (Colombel, 1974; Torossian, 1978). Leptothorax recedens lay male-destined eggs only during the first several weeks after hibernation, when the queen's inhibitory power is weak or absent (Dejean and Passera, 1974). Usually the sole Myrmica rubra workers that produce males are the callows, and this activity is most likely in the late summer in England, when there is a flush of such young individuals in the nest (Smeeton, 1981). "True" workers of the European slavemaker Harpagoxenus sublaevis exist, and Buschinger and Winter (1978) found 11 fertile individuals among 230 dissected; of these 7 had been laying eggs in the presence of the queen. But the workers in this unusual species differ from the much more fertile ergatogynes only in their lack of a seminal receptacle and hence ability to store semen for the normal generation of female offspring. A stronger exception to the general rule appears to be Leptothorax allardycei, whose workers produce 20 percent of the eggs in the presence of the queen (Cole, 1986). The overall picture of worker reproduction is one of marginal productivity and nearly universal inhibition by the laying queens. This circumstance is difficult to interpret with reference to the two major competing hypotheses, but it appears to favor parental manipulation. There would seem prima facie to be strong selection for individual worker reproduction in the presence of the queen. By producing males, the egg-layer trades brothers for sons and hence a degree of relatedness (b) of 1/4 for one of 1/2. The worker's sisters should be compliant, at least in the presence of a single queen who mated once, because by exchanging brothers for nephews they trade a relatedness of 1/4 for one of 3/8.

On the other hand, both the kin selection and parental manipulation hypotheses predict some conflict among the workers for whatever reproductive rights they assume. In fact, dominance hierarchies have been reported among the workers of the little stem-dwelling myrmicine Leptothorax allardycei (Cole, 1981) and the slave-maker Harpagoxenus americanus (Franks and Scovell, 1983). But otherwise observations of conflict among workers have been limited to queenless colonies, in which not only the queen's contribution is ended but the entire reproductive future of the colony is put at risk. Who controls the investments? Because the predicted ratios of investment are quantitative, they provide the most rigorous comparison of kin selection and parental manipulation. Trivers and Hare (1976) distinguished three kinds of societies that should possess characteristic investment ratios if kin selection is operating: In monogynous colonies, the ratio of investment should be 3:1 in favor of the female reproductives, providing the workers have evolved so as to control the investment in a way that maximizes their inclusive fitness. Put another way, the workers should see to it that the male biomass constitutes 25 percent of the total invested in reproductive adults as a whole. The reason: workers are weighing degrees of relatedness of 3/4 for their sisters against 1/4 for their brothers. In polygynous colonies, the investment ratio can be expected to subside toward 1:1 (50 percent male investment), because the workers are no longer all sisters. Indeed, as revealed by recent allozyme studies (Table 4-1) the average degree of relatedness among the workers from individual colonies is usually well below 1/2 and often below 1/4. In mixed colonies of parasitic and host species, the ratio should also be 1:1, because the parasite queens are selected to control the ratio, whereas the captive and usually queenless workers of the host species cannot be selected to resist manipulation by the parasite. They are unable to evolve resistance because they leave behind no offspring or otherwise aid in the continuance of their own species. In a recent analysis of the rapidly accumulating studies on investment ratios, Nonacs (1986a) found the Trivers-Hare predictions based on kin selection to be "remarkably robust," whereas the parental manipulation hypothesis has fallen short in several key areas. The main results are depicted in Figure 4-5. The percentages of biomass investment in males, with the standard error, are 0.282 ± 0.06 for the monogynous species, 0.522 ± 0.09 for the polygynous species, and 0.515 ± 0.02 for the parasitic species, all quite close to the equilibrium figures predicted by the kin selection model. The numerical percentages are 0.579 ± 0.09, 0.687 ± 0.09, and 0.628 ± 0.03 for the same three groups of species. These are again respectably close to the predicted value of 50 percent. It will be recalled that the latter, i.e. numerical, sex ratio is expected to be close to unity in a fully outbreeding system, because males are generally able to inseminate only one queen.

It is a curious feature of the investment ratio strategies that they apply in a given species to the population of colonies as a whole rather than to individual colonies. As illustrated in Figure 4-6, colonies tend to specialize on one sex or the other in any given reproductive season. Nonacs (1986a,b) found that the proportion of males decreases with an increase in the biomass of new reproductives in the colony as a whole. However, the correlation is not primarily with the size of the colony. When Nonacs parcelled out the biomass of reproductives, there was almost no relationship between the number of workers in the colony and the percentage of investment in males. These trends led him to conclude that colonies invest mostly in males when resources are low (and only a small reproductive crop can be raised) and mostly in females when resources become more plentiful. This conclusion seems supported by the independent finding of Rosengren and Pamilo (1986) that colonies of Formica aquilonia and Formica rufa bias the sex ratio investment toward females when the environment is most favorable. The input of the resources thus serves as a proximate cue to the colony on where to place its surplus energy income. The effects summed over the population of colonies as whole yield the 3:1 or 1:1 ratios. If a large population of colonies is at equilibrium, then all sex ratios will be equally fit, as shown in a general proof by Taylor and Sauer (1980). Hence, even colonies that produce all males or all females cannot be assumed to be acting against the inclusive fitness of its workers. The intuitive basis for this argument is as follows. In an equilibrial population, with either 3:1 or 1:1 ratios overall, the colony that produces mostly males will find itself compensated during mating because some other colonies are producing excess females. What evolves over a period of time, in theory, is the sensitivity of individual colonies to resource levels. These thresholds are selected so that the colony acts "appropriately" with respect to the population as a whole. In short, the sex ratios chosen at different times by single colonies represent an evolutionary stable strategy of mixed responses of the kind suggested by Maynard Smith (1982). Two detailed studies of variation among colonies of the same species have shed further light on the evolution of investment ratios. Herbers (1984) analyzed colony units of Leptothorax longispinosus with variable numbers of queens. She found that the fewer the queens, the closer the investment ratio approached 3:1 in favor of females. Queenless groups had the highest ratio of all. The result can be interpreted as kin selection operating amidst a conflict between the queens and the workers. When no queens are present, the workers boost the ratio without interference. When many queens are present, they push the ratio back down to a level more in their favor. However, the 1:1 ratio tends to be more in the interest of the workers in this case, so conflict might well be attenuated at the extreme multiple-queen end of the scale. Nonacs (1986b) interprets the result as a response of the polygynous colonies to fewer resources caused by a smaller proportion of workers. Whichever of the two interpretations is correct, Herbers and Nonacs agree that kin selection is an important force in Leptothorax longispinosus and that some amount of conflict between the queens and workers does occur. Ward (1983b) has conducted an equally instructive study on two species of Australian ponerine ants belonging to the Rhytidoponera impressa group (Rhytidoponera chalybaea and Rhytidoponera confusa). Both of these species has two types of colonies. Type A colonies each have a single queen, and their overall mean proportional investment is 0.82 in favor of the females. Type B colonies are queenless and are serviced instead entirely by reproductive workers. They are therefore usually polygynous. Their mean proportionate investment in queens ranges from 0.35 to 0.72. The two types of colonies are intermingled. The higher the relative frequency and density of the type B (queenless) colonies in particular local populations, the higher the proportionate investment by the neighboring type A (queenright) colonies in new queens. In fact, their average investment has risen above the 0.75 level predicted by kin selection theory in some localities, presumably as a local adaptive response to the larger number of males being generated in nearby type B colonies. This interpretation is supported by the fact that one population of Rhytidoponera purpurea, which also belongs to the impressa group and consists entirely of type A colonies, had a proportionate investment of 0.74, in close accordance with the kin selection model. As Ward points out, there are two ways by which type A colonies could compensate their investment ratios in the manner observed. First, there could be some form of communicating by which they gauge the density of type B colonies. Such an assessment among colonies is not impossible. Hölldobler (1976c, 1981a) has shown that it occurs during the ritualized tournaments among colonies of the honeypot ant Myrmecocystus mimicus. Alternatively, the investment ratios of type A colonies might have evolved in local populations in response to stable environments and equilibrial densities of type B colonies. This explanation is made plausible by the fact that the rainforest populations of Rhytidoponera chalybaea and Rhytidoponera confusa are in habitats that have persisted for thousands of years.

Finally, Bourke et al. (1988) provided solid evidence supporting the prediction made by Trivers and Hare that mixed colonies of parasitic and host species invest at a ratio of 1:1. They found that the mean proportion of dry weight investment in queens belonging to populations of the monogynous slavemaker Harpagoxenus sublaevis is 0.54. Using allozyme analysis they also confirmed that each queen mates with only one male, that female nestmates are full sisters (coefficient of relatedness 0.73 ± 0.07), that inbreeding does not occur, and that queen and worker siblings are not genetically differentiated. Recently, Frank (1987) presented an alternative model, which attempts to explain the phenomenon that small colonies of some ant species tend to produce predominantly males, while large colonies produce mostly female alates. This model, which Frank calls "constant male hypothesis" suggests "that when there is any local mate competition, however slight, small colonies are favored to make mostly males, and large colonies are expected to increase their investment in females as their total brood increases." As of now, however, no convincing behavioral or genetical evidence exists indicating that even a weak trend toward local mate competition in mating swarms of ants. In an important study of Lasius niger, van der Have et al. (1988) excluded mate competition in favor of kin selection. In marginal populations, investment ratios tend to drop from 0.75 toward 0.50, in part due to higher production of males by workers. Similarly, Elmes (1987b) found that the largest and most reproductively successful colonies of Myrmica sulcinodis have close to 0.75 investment ratios, which drop toward male bias in marginal, polygynous populations. Finally, it is surely significant that the larger and more complexly organized the monogynous ant colonies, and hence the less preponderant the physical presence of the queen, the more care the workers devote to her. Put another way, as colonies grow larger the workers appear to treat the queen more as a valuable resource and less as a rival. Among the more arresting spectacles of the ant world are the dense retinues of certain species that guard the mother queen as she moves from one nest site to another. In extreme cases, such as species of the legionary ponerine Simopelta, the doryline and ecitonine army ants, the leptanilline army ant Leptanilla, Solenopsis fire ants, Pheidologeton marauder ants, and Oecophylla weaver ants, the entire body of the queen is covered by a seething shell of these guards (see Figure 4-7). In ecitonine army ants, the retinue extends as much as a meter in front of the queen and two meters behind her and contains up to five times the number of workers found in an ordinary two-meter segment of the column of emigrating ants. Epigaeic species, in other words those whose colonies move mostly above ground and in sight of predators, have the largest retinues (Rettenmeyer et al., 1978). In retinue-forming ant species investigated so far, the queen possesses unusually extensive exocrine glands that evidently produce pheromones attractive to the workers.

A fruitful direction to take in ant biology will be to examine investment priorities among different classes of colony members. In principle, we should expect workers to maximize their personal inclusive fitness by favoring particular castes and life stages over others. They might provide special care to their own sisters as opposed to half-sisters and more distantly related relatives. A few tantalizing pieces of evidence concerning such discrimination have surfaced already. Petersen-Braun (1982) found that workers of the house ant Monomorium pharaonis respond most aggressively to the queens that are less productive, with attacks turning fatal when the colony breeding cycle is threatened (Petersen-Braun, 1982). In this study, however, no specific discrimination based on relatedness could be detected. A similar phenomenon has been reported in the fire ant Solenopsis invicta by Fletcher and Blum (1983a,b). Lenoir (1981) reports that when workers of Lasius niger discover lost brood, they retrieve the different stages in the following order: large larvae and pupae, then small larvae, and finally eggs. This is the correct sequence predictable from the differing amounts of payoff (in healthy adults produced) expected from these stages. In particular, pupae have already received the most energy and will enter the adult generation with the least additional care, whereas eggs still require large amounts of energy investment and future time. Unfortunately, however, the tests were not calibrated in a manner that allowed choices by the ants to be made on a per gram basis.

Eusociality and chromosome numbers
Finally, Sherman (1979) pointed out colonies of social insects tend to be more harmonious when the members are more closely related; therefore mechanisms to reduce within-colony genetic variance should be favored by natural selection operating at the level of the colony--or more precisely, at the level of the queen. One way of reducing variance is to increase the number of chromosomes, so that large blocks of genes held together by linkage are broken up and their constituents distributed more evenly among colony members during meiosis and fertilization. Increased cross-over should produce the same result. The data assembled by Sherman appeared to confirm that species of eusocial insects, including ants and termites, do have higher chromosome numbers than the solitary species phylogenetically closest to them. In a recent summary Crozier (1987b) supports this result for ants and other social insects, even though one species, a bulldog ant in the Myrmecia pilosula complex, has a haploid number of only one (Myrmecia is exceedingly variable in this respect, with another species, Myrmecia brevinoda, possessing a haploid number of 42). As Crozier points out, there is no way at present to judge which came first, that is, whether high chromosome numbers predisposed certain phylogenetic lines toward eusociality, or the reverse. He is certainly correct in suggesting that "sociogenetics," the blending of genetic and sociobiological analysis at the chromosome and genic levels, is a still inchoate field of great promise. Ultimately it will change our view of the way in which eusociality originated and is sustained.

Overview
During the past twenty years, the analysis of kinship and investment strategies in ants and other social insects has matured into one of the more sophisticated enterprises of evolutionary biology. During this time opinion concerning the dominant selective force of colonial evolution has shifted twice. Throughout the 1960s and into the 1970s, enthusiasm prevailed for the pure kin selection hypothesis, which appeared to explain the origin of the sterile worker caste in a uniquely robust manner. Sentiment in this direction was reinforced when Trivers and Hare (1976) added the concepts of conflict among colony members and the optimum 3:1 queen-to-male ratio of investment. Then the fortunes of kin selection declined as Alexander and Sherman (1977) pressed the opposing parental manipulation model. It seemed that theoreticians might dispense with a complex calculus of investments if the mother could be shown to force some of her daughters into a slave-like worker status with a reduced inclusive fitness. Doubts about kin selection were reinforced by flaws discovered in the statistical analysis of the early investment data, along with a battery of new theoretical arguments that made parental dominance seemingly easier to attain in the course of evolution. Alexander and Sherman further argued that biasing toward female production could be explained as the result of local mate competition, wherein the queen produced fewer males because a smaller force is needed to inseminate the cogenerational crop of sisters and other close kin. Crozier (1977) pointed out that in order for kin selection to create a sterile worker caste, as opposed to merely maintaining it, the ancestral pre-formicids would have to be able to tell the sex of larvae before such species could create the favorable 3:1 ratio. In other words, the sensory cart must be put in front of the behavioral horse. Charlesworth, Charnov, and Craig independently noted that the mother need only make offspring half as efficient at rearing other brood in order to serve her own interests, so that worker formation is easier by means of parental manipulation than had been earlier imagined. Altogether, kin selection took a heavy beating during the late 1970s. Then, in the 1980s, the tide turned back. As Crozier himself cautioned (in 1982), "don't discard kin selection in favor of parental manipulation." More and more pieces had begun to fall into place to create a pattern favorable to kin selection within a setting of parent-offspring conflict. From Bartz's rule to the cumulative investment measurements analyzed by Nonacs, as well as other contributions we have reviewed in the present chapter, details of colony life histories and caste structure have been elucidated that seem difficult to interpret by any other existing model, including parental manipulation. Furthermore, it began to be appreciated that parental manipulation can take a form concordant with positive kin selection rather than opposed to it, if the mother changes the environment and rearing conditions so as to improve the inclusive fitness of her "enslaved" daughters.

This interesting saga, reminiscent of some of the best historical controversies in physics and molecular biology, is far from over. Too many surprising new theoretical and empirical discoveries have been made during the past ten years to suppose that all competing explanations, including parental manipulation, can be discarded for good. One of the more desirable targets for future research is the close comparison of primitive and advanced ant taxa. Anatomically primitive genera with relatively simple social organizations, especially Amblyopone and other members of the Amblyoponini, as well as Myrmecia and Nothomyrmecia, should be compared in detail with evolutionarily advanced genera such as Formica and Pheidole. It is likely that differences in the degree of conflict and investment patterns will be found to occur and that they can be used to discriminate more confidently among the competing hypotheses. The basis of this last prediction is as follows. Charnov (1978) first pointed out that the conditions for the beginning of eusociality must have been very different from the conditions of its maintenance in advanced eusocial species. Indeed, many of the difficulties envisioned by Alexander, Crozier, Pamilo, and other writers have centered around the origin of the worker caste. It is entirely possible that parental manipulation played a role in spreading the first rare genes that capacitated worker development. Alternatively, the first workers could have been females parasitizing their mothers by choosing settlement in the home environment over the risk of creating new nests. As Bartz (1982) pointed out, the stay-at-homes were not required to bias their reproductive siblings into a 3:1 investment ratio in order to evolve into workers themselves. It was only when the queen produced all of the males herself that a female-biased ratio was necessary to create degrees of relatedness favorable to the evolution of female workers. If daughters produced any male reproductives at all, then the requirement for a biased investment ratio was sidestepped. If a female raised a nephew rather than a brother, she traded a degree of relatedness of 1/4 for one of 3/8. Consequently her average relatedness to a reproductive brood composed of sisters, brothers, and nephews would have been greater than the relatedness to a brood composed of only sisters and brothers, and under many combinations, greater than her relatedness to a brood she could have expected to raise had she started her own nest. Once eusocial species reached a more advanced stage, with large colonies and specialized worker subcastes, the rules changed. It is likely that worker reproduction and conflict lost most of their profit in comparison with the maintenance and fine-tuning of colony organization. As far back as the late Cretaceous Period the ants, as we know them, may have reached a point of no return. The earliest stages of eusociality have not been found in any living species of ant and no solitary aculeates are known that might have originated from ants through a secondary loss of eusociality. The bees and wasps remain by far the best insect groups for studying the origin of the worker caste. They offer a graded series of living species that range from completely solitary to completely eusocial. But the extraordinary phylogenetic spread of the ants and the immense diversity of their social systems make them the most favorable group for the reconstruction of the middle and advanced grades of eusocial evolution.