Larval helpers and age polyethism in ambrosia beetles

Abstract

Division of labor among the workers of insect societies is a conspicuous feature of their biology. Social tasks are commonly shared among age groups but not between larvae and adults with completely different morphologies, as in bees, wasps, ants, and beetles (i.e., Holometabola). A unique yet hardly studied holometabolous group of insects is the ambrosia beetles. Along with one tribe of ants and one subfamily of termites, wood-dwelling ambrosia beetles are the only insect lineage culturing fungi, a trait predicted to favor cooperation and division of labor. Their sociality has not been fully demonstrated, because behavioral observations have been missing. Here we present behavioral data and experiments from within nests of an ambrosia beetle, Xyleborinus saxesenii. Larval and adult offspring of a single foundress cooperate in brood care, gallery maintenance, and fungus gardening, showing a clear division of labor between larval and adult colony members. Larvae enlarge the gallery and participate in brood care and gallery hygiene. The cooperative effort of adult females in the colony and the timing of their dispersal depend on the number of sibling recipients (larvae and pupae), on the presence of the mother, and on the number of adult workers. This suggests that altruistic help is triggered by demands of brood dependent on care. Thus, ambrosia beetles are not only highly social but also show a special form of division of labor that is unique among holometabolous insects.

Division of labor enhances work efficiency and is fundamental to the biological evolution of social complexity (1). This conspicuous feature of social insects largely explains their ecological success (2). Task specialization in insects usually occurs between different age groups. Individuals either pass through consecutive molts during development (Hemimetabola; e.g., termites, aphids), with immatures resembling small adults and division of labor occurring typically between such larval nymphs that may show morphological specializations [e.g., first-instar soldier aphids (3)]; or individuals metamorphose by dramatically reorganizing their morphology during the pupal stage (Holometabola; e.g., bees, wasps, ants, beetles), which predestines larvae and adults to specialize in different tasks because of their morphological and physiological differentiation. Indeed, larvae of the eusocial Hymenoptera, for example, may produce nest-building silk [weaver ants (4)] and may supply adults with extra enzymes and nutrients [trophallaxis in several wasps and ants (2, 5, 6)]. However, all known cooperative actions of holometabolous larvae are apparently completely under adult control (2), and despite the potential for the evolution of highly specialized immature helper morphs, larvae of these taxa are largely immobile and helpless and in need of being moved, fed, and cared for by adults (7). In ambrosia beetles, however, in which cooperative breeding (8) and eusociality (9) also have been assumed, larvae can move and forage independently in the nest, which provides great potential for division of labor between larvae and adults.

Division of labor sets the stage for the origin of fungiculture in fungus-growing ants and termites (10). Ambrosia beetles originate from solitary or colonial ancestors, and their fungus agriculture may have coevolved with sociality (8, 10, 11). However, the role of division of labor is unknown. Ambrosia beetles live inside trees, which is a habitat extraordinarily favoring social evolution (12), apparently having fostered at least seven independent origins of fungiculture in beetles (13). Hence, they represent a unique model system to study the evolution of sociality in relation to fungiculture. Interestingly, ambrosia beetles vary in their mating system (inbreeding vs. outbreeding species) and ploidy level (haplodiploid vs. diploid species), which are factors that have been assumed to contribute to social evolution, although their respective roles in social evolution are controversial (1). The ambrosia beetle subtribe Xyleborini is characterized by regular inbreeding, haplodiploidy, and fungiculture (8, 14). High relatedness and haplodiploidy in combination with an extremely female-biased sex ratio are factors predisposing them to advanced sociality (1). Additionally, cooperation in fungiculture is likely because a single individual can hardly maintain a fungus garden on its own (10). Indeed, it was shown that adult female offspring delay dispersal from their natal gallery, which results in an overlap of eggs, larvae, pupae, and at least two generations of adult offspring within a colony (8). A helper effect of philopatric females has been indicated by the fact that the number of staying females that do not reproduce correlates positively with gallery productivity in Xyleborinus saxesenii Ratzeburg (8). Behavioral observations of ambrosia beetles within their galleries have been missing so far, however, because it is virtually impossible to observe them in nature inside the wood. The only report on eusociality in ambrosia beetles is not based on behavioral data, but reproductive roles have been inferred by destructive sampling of active nests of Austroplatypus incompertus (9). To facilitate observations of beetle behaviors inside galleries, we developed artificial observation tubes to contain entire colonies of reproducing beetles (15, 16). Here we use this breeding technique of X. saxesenii to ask (i) whether offspring produced in a gallery engage in alloparental brood care and fungus maintenance, (ii) whether different types of individuals specialize in divergent tasks, and (iii) whether decisions to help and to disperse relate to the number of potential beneficiaries and the number of potential workers present in the colony. Furthermore, we evaluate experimentally (iv) whether female dispersal depends on the presence of an egg-laying foundress, because her removal should affect the need for alloparental care. We compare our results with the patterns of sociality known from other major insect taxa.

Results

Age- and Sex-Specific Behavior.

Gallery maintenance and brood care were allocated differently between different age classes within a nest (Fig. 1 and Table S1). The gallery was extended mainly by larvae (digging), which also reduces the spread of mold (SI Text and Fig. S1), whereas fungus care (cropping) and brood protection (blocking) were exclusively conducted by adults. Blocking was only performed by foundresses and mature females, and in 33 of 35 cases only one individual blocked at a time. In the 25 cases (in 19 of 93 galleries) when we did not observe a blocking female, the following behavioral scan revealed that larvae had crawled out of the gallery. At least 71 larvae ( = 3.7 larvae per gallery) died in this manner, which suggests that an important function of blocking is to prevent larvae from getting accidentally lost.

Fig. 1.

Division of labor and age polyethism between age and sex classes in X. saxesenii. Bars show the mean (±SE) proportions of time larvae, teneral females, mature females, and males performed different cooperative tasks. Statistically significant differences between the classes are denoted by different letters (P < 0.05; GEE, details in Table S1). Note scale differences between A–C and D–F.

Larvae and adults took different shares in hygienic behaviors. Only larvae compressed dispersed waste into compact balls (balling; Movie S1). Frass-balls, pieces of wood, or parts of dead siblings were moved within the gallery and pushed out of the entrance (shuffling; Movie S2) mainly by mature females but to some extent also by larvae and teneral females (Fig. 1 and Table S1). The ultimate waste disposer was always the mature female that was blocking the entrance. Cannibalism was directed toward adult beetles (1 case), pupae (3 cases), and larvae (37 cases) that were already dead (in most cases) or that did not respond to being groomed. In such cases the groomer (larva or adult) would use its mandibles to open the body of the groomed sibling within seconds. Allogrooming was shown by all stages and both sexes, and it seemed to be crucial for individual survival: in an experiment with 10 pupae that were singly kept either with one or six larvae, the pupae survived in five of five cases with six larvae, but they survived in only one of five cases with one larva (Fisher exact test: P = 0.048, n = 10; details in SI Text and Fig. S2). In the other four cases with one larva, the pupae were overgrown and killed by fungi. In addition, the body surface of single foundresses that had not yet successfully established a brood was overgrown by a fungal layer (mainly Paecilomyces variotii and Fusarium merismoides) within a few weeks, which caused the death of at least 7 of 29 solitary females. Whenever individuals of different stages encountered each other, they removed the visible fungal layer on each other's bodies by allogrooming. In males, allogrooming was frequently followed by a mating attempt; it may thus serve also to obtain information about female mating status.

Adult Female Behaviors Depending on Gallery Composition.

The proportion of time adult females exhibited cropping and allogrooming and the occurrence of blocking were higher during gallery stages when brood dependent on care was present in the colony (preadult and larval-adult gallery stages) than during other times (postlarval gallery stage; Table S2). Shuffling was shown equally often before, during, and after larval presence within a gallery, whereas adult female digging tended to increase per capita after pupation of the last larva. Cannibalism was not shown before the hatching of adult daughters, and it occurred more often when dependent offspring were still present. Dispersal of adult females seemed to be contingent on brood care demands: it significantly increased in the 10 d after a major proportion of larvae had completed their development relative to the 10 d before (Wilcoxon: z = −3, P < 0.001, n = 23; Fig. S3). Because brood numbers likely correlate with fungus productivity, dispersal could be triggered by alternative factors, however, like the quality of the wood/medium or fungus. Nevertheless, in summary there are hints that some behavioral tasks of adult females functionally relate to the demands of care-dependent brood.

In a second analysis we tested for the relationship between adult female behaviors and the numbers of adult females and brood (pupae and larvae) present in the galleries during the larval-adult gallery stage only (Table S2). Analyzed per capita, adult female digging, shuffling, cannibalism, and blocking were all independent of the numbers of adult females and younger nestmates. However, per capita allogrooming and fungus cropping activities significantly increased with increasing numbers of pupae and larvae, whereas with increasing adult female numbers allogrooming significantly decreased. Adult female dispersal correlated negatively with the number of dependent offspring [generalized estimation equation (GEE): P = 0.003; Fig. 2A and Table S2] and positively with adult female numbers (GEE: P < 0.001; Table S2).

Fig. 2.

(A) Adult female dispersal correlates negatively with the number of cared brood in the gallery (larvae and pupae; GEE: P = 0.003; statistical details in Table S2). (B) Effects of removal of the blocking foundress. The numbers of dispersing X. saxesenii adult females (medians and quartiles) are shown 4 d before and 4 d after the experimental treatments. Wilcoxon tests: *P < 0.05, **P < 0.01; Mann-Whitney U test: ***P < 0.001; sample sizes were 10 control galleries and 10 galleries from which the foundress was removed.

In 34 of 43 galleries with mature offspring, females delayed their dispersal from the natal nest after maturation. This philopatric period (i.e., the latency from the first appearance of a mature female within a gallery until the first dispersal event) correlated positively with the average number of dependent brood (eggs, larvae, pupae) per adult female present during this period (Spearman rank correlation: R_S = 0.379, P = 0.027, n = 34).

Effect of Foundress Removal.

Eggs are produced primarily by the foundress, and in 4 of 16 galleries dissected in the field eggs were produced also by at least one daughter (on average reproduced 23.9% of all females in these 4 galleries: range, 2–4 egg-layers on 4–22 females in total; details in ref. 17). If fewer eggs are produced, the need for alloparental care declines. Therefore, we predicted that the number of dispersing daughters will increase if the foundress is experimentally removed from the gallery. Our experimental interference raised the dispersal activity of daughters (relative to the same galleries before the manipulation; Fig. 2B) in the treatment (Wilcoxon: z = −2.637, P = 0.008, n = 10) and in the control groups (Wilcoxon: z = −1.82, P = 0.034, n = 10). Removal of the foundress, however, raised the dispersal of daughters much more strongly than the control situation (i.e., removal of the medium without the foundress; U test: z = −3.708, P < 0.001, n = 10 + 10).

Discussion

Here we report on division of labor and age polyethism in Coleoptera, which apparently relate to the fungiculture in ambrosia beetles. It confirms the predicted high degree of sociality of ambrosia beetles (9, 11). In X. saxesenii, all group members contribute to the divergent tasks of gallery maintenance and brood care, and there is correlative evidence that female offspring delay their dispersal depending on the number of potential beneficiaries present in their natal gallery. Tasks are typically shared differentially between larval and adult colony members, which resembles the division of labor reported from termites and other highly social insect taxa (Table 1). Among holometabolous insects, however, X. saxesenii is the only species to date known to exhibit an active behavioral task specialization between larvae and adults. Furthermore, adult daughters in this species specialize in different tasks than the colony foundress, and they seem to adjust their work effort flexibly to the varying size of the brood produced by the latter. Dispersal propensity of adult females (i) correlates positively with low numbers of dependent young and high numbers of adult workers, and (ii) is increased by experimental removal of the foundress. This suggests that philopatry may be related to indirect fitness benefits, because the group members sharing a gallery are all very highly related (for another xyleborine beetle see ref. 14). There is no morphological differentiation among adult females, and they are all fully capable of breeding and establishing their own gallery. However, dissections of all females in a number of field galleries showed that daughters cobreed in their natal gallery in only a quarter of colonies (17). In summary, the social and reproductive patterns of X. saxesenii conform to primitive eusociality, defined by overlapping generations, cooperative brood care, and some reproductive division of labor, despite totipotency in reproduction of all individuals.

Table 1.

Forms of division of labor in exemplary species of the most prominent social insect taxa

	Behavioral task-specialization
Social insect species	Between immatures* and adult stages	Among adult females	Cooperation by males	Mode of nesting and nourishment	Reference
Hemimetabola
Cryptotermes cavifrons (Kalotermitidae)	+	−	−	Wood dweller, gut symbionts	45
Hodotermopsis japonica (Termopsidae)	+	−	+	Wood dweller, gut symbionts	46
Reticulitermes fukienensis (Rhinotermidae)	+	−	?	Subterranean, gut symbionts	47
Macrotermes subhyalinus (Termitidae)	+	+	+	Subterranean and mound building, fungus cultivation	48
Pemphigus spyrothecae (Aphidae)	+	−	−	Plant galls, sap sucking	49
Kladothrips intermedius (Thysanoptera)	−	+	+	Plant galls, sap sucking	50
Holometabola
Lassioglossum zephyrum (Halictinae)	−	+	−	Ground nesting, pollen and nectar	51
Exoneura bicolor (Allodapinae)	−	+	−	Stalk-nesting, pollen and nectar	51
Apis mellifera (Apidae)	−	+	−	Cavity nesting, pollen and nectar	52
Liostenogaster flavolineata (Stenogastrinae)	−	+	−	Free nesting, arthropods and nectar	53
Vespula germanica (Vespinae)	−	+	−	Free nesting, arthropods and nectar	7
Atta texana (Formicidae)	−	+	−	Subterranean, fungus cultivation	2
Oecophylla longinoda (Formicidae)	−†	+	−	Tree nesting, arthropods and honeydew	4
Austroplatypus incompertus (Platypotinae)	−	+	+	Wood boring, fungus cultivation	9
Xyleborinus saxesenii (Scolytinae)	+	+	+	Wood boring, fungus cultivation	This study

+, present; −, absent; ?, unknown.

*Nymphal and larval stages.

^†Larvae of weaver ants produce nest-building-silk, but the weaving action is controlled by the worker ants (2, 4).

Division of Labor Between Larvae and Adults.

Task sharing in X. saxesenii is unequal between the sexes and age classes for most of the social behaviors described (Fig. 1 and SI Text). Regarding gallery hygiene, for example, larvae form balls from dispersed frass, whereas the transport and removal of these balls is mainly performed by adult females. Gallery extension is almost exclusively accomplished by larvae. A social role of larvae has never been reported in ambrosia beetles, and gallery maintenance and fungiculture have been attributed solely to the foundress (11, 15, 18). Larvae might serve a cooperative function not only in Xyleborini but also in other beetles, as anecdotal observations and speculations suggest from the Scolytinae [e.g., Dendroctonus sp.(19); the Xyloterini (20)], Platypodinae [e.g., Trachyostus ghanaensis (21); Doliopygus conradti (22)], and Passalidae [Passalus cornatus (23)]. The ultimate cause for the larval specialization in digging may relate to their repeated molting: as the mandibles gradually wear during excavation (for wood-dwelling termites see ref. 24), adult females showing extensive digging behavior would suffer from substantial long-term costs. In contrast, larval mandibles regenerate at each molt.

Teneral and mature females take over fungus care. Blocking of the gallery entrance is done exclusively by mature females and almost only by the foundress (see ref. 14 for similar observations in Xylosandrus germanus). Direct brood care by allogrooming is performed by all age classes. Males groom females at high rates, which may primarily serve courtship because mating attempts are always initiated by allogrooming (cf. 20). Males do not take part in other social behaviors except for low levels of digging and cropping. Asymmetries in relatedness caused by haplodiploidy should favor females to become helpers (25, 26). Inbreeding can nevertheless reduce relatedness asymmetries and thus favor also males to help (27). Male soldiers and brood-caring males have been documented in haplodiploid and sib-mating thrips and Cardiocondyla ants, respectively (27). In Xyleborini, very few males are produced [approximately 5–12% of offspring (28)], and males do not seem to contribute significantly to gallery function, hygiene, and fungus growth. Instead, they seem to specialize in their reproductive role by continually attempting to locate and fertilize their sisters. Nevertheless, both male larvae and adults are fully capable of performing all of the behaviors exhibited by females, as demonstrated in galleries that contain only males [approximately 2% of X. saxesenii galleries are founded by unfertilized females that produce solely male broods (16, 28)].

The life-history trajectory of X. saxesenii is most similar to that of social aphids and some termite families, where individuals serve as helpers (e.g., workers and soldiers) in their natal colonies before maturation (Table 1). Either sex may help in these taxa, and their flexible developmental period allows individuals to adjust the length of their immature stage to maximize their inclusive fitness. They may either remain in an immature helper phase at the nest or develop into adults that can disperse or reproduce in their natal colony (3, 29–32). Likewise, the developmental period of second- and third-instar larvae of X. saxesenii can vary between 4 and 17 d (16). In addition, there is a striking similarity of these taxa in ecology and mating patterns. They all inhabit defensible nests, often within wood, which favors local mating and inbreeding—conditions claimed as strongly favoring social evolution (12). In many of these cases helping tasks do not seem to curtail a helper's future reproduction (i.e., because helping is risk free and does not reduce a helper's energy stores), which may weaken the tradeoff between helping and future reproduction (33, 34).

Role of Individual Selection and Kin Selection.

We defined a behavior as cooperative if social partners potentially benefit from its performance, independently of whether this behavior entails net costs to the actor (35). This definition includes (i) mutualistic behaviors that are regarded as selfish acts generating benefits to other individuals (common goods) as a by-product (36), and (ii) altruistic behaviors that bring about net costs to the actor, which are compensated through indirect fitness benefits via kin-selection; they will thus only evolve in groups of relatives (25). In the course of social evolution and task specialization shaped by kin selection, mutualistic behaviors may lose their original function and change into truly altruistic behaviors (36). High relatedness and spatial separation between groups are very favorable to the evolution of cooperative behaviors (12), as long as local competition between relatives does not oppose this force (37).

In Xyleborini, ideal conditions for the evolution of both mutualism and altruism seem to prevail: (i) they are well preadapted to brood care because they originate from a beetle lineage (Scolytinae) with parental care (38); (ii) they breed in isolated galleries that are founded by one reproductive that dominates offspring production [i.e., the foundress (17)]; (iii) they are haplodiploid and mate predominantly among full siblings [e.g., inbreeding coefficients of approximately 0.9 in Xylosandrus germanus Reiter (8)], which increases relatedness within natal colonies; and (iv) they disperse solitarily after maturation to reproduce elsewhere. In addition, (v) colony members benefit greatly from cooperation due to their dependence on fungiculture; and (vi) the wood used as a resource for shelter and substratum for fungi is virtually nondepreciable, which renders resource competition negligible. Larval digging, for example, which might be regarded as a by-product of selfish larval feeding, reduces competition because it generates a common good (i.e., space and substratum for fungiculture). Similarly, other group members benefit from gallery hygiene resulting as a by-product from the seemingly selfish adult feeding activities cropping and cannibalism. By contrast, balling, shuffling, allogrooming, blocking, and adult digging may rather be altruistic. These behaviors apparently benefit other group members at the expense of time and energy costs to the actors, without immediate benefits to the latter. Particularly dangerous are allogrooming and blocking, because they expose the performers to pathogens, parasites, and predators.

Adult X. saxesenii females showed a strong incentive to stay and cooperate in a productive natal nest. Partly this may be selfish, because up to one quarter of females may get the chance to reproduce (17), and females might build up reserves during their philopatric period for subsequent dispersal and nest foundation. An experimental study in the ambrosia beetle Xyleborus affinis suggests, however, that females do not build up reserves during their philopatric period but rather suffer direct fitness costs (39). Alternatively, indirect fitness benefits may be involved in helping to raise siblings (8). Three results suggest that adult female cooperation is triggered by the demands of brood dependent on care: (i) helping effort of adult females rose with a greater number of brood dependent on care, (ii) adult females dispersed at a higher rate with an increasing number of workers in the colony, and (iii) dispersal rate of females increased in response to experimental removal of the foundress, which indicates that delayed dispersal of females is not primarily motivated by the potential to breed in the natal gallery (see also ref. 8).

In conclusion, the high degree of sociality in ambrosia beetles seems to result from a combination of four major factors: (i) parental care as a preadaptation for the evolution of sociality in the ancestors of modern ambrosia beetles (38); (ii) very high relatedness within families due to haplodiploidy and inbreeding; (iii) a proliferating, monopolizable resource providing ample food for many individuals, which needs to be tended and protected; and (iv) high costs of dispersal (for another scolytine beetle see ref. 40) due to the difficulties of finding a suitable host tree, of nest foundation, and a successful start of fungiculture (11, 16), which render predispersal cooperation particularly worthwhile. X. saxesenii larvae are predisposed to assume certain tasks like balling of frass and gallery enlargement (digging) because of their body morphology and the frequent renewal of mandibles by molting. Thus, behavioral tasks are shared between larval and adult stages. This has not been shown for beetles, and the described division of labor between immature and adult stages seems to be unique among holometabolous social insects at large.

Materials and Methods

Study System.

X. saxesenii galleries are founded by individual females that transmit spores of the species-specific ambrosia fungus Ambrosiella sulfurea Batra in their gut from the natal to the new gallery (41, 42). This fungus forms a yellow layer of fruiting cells on the surface of gallery walls (Fig. S4A). After landing on a tree trunk, the foundress excavates a straight tunnel into the xylem with a small egg niche at its end. As soon as fungus beds emerge, she feeds on them and starts egg-laying (16). During offspring development the egg niche is enlarged to a single flat brood chamber of up to a few square centimeters in size and ≈1-mm height. There, most of the fungus garden grows, and individuals of all age classes live in close contact with each other (Fig. S4C). In this study we bred X. saxesenii in glass tubes filled with artificial medium that mainly contained sawdust (for details on this method see SI Text and ref. 16).

Behavioral Recordings and Analyses.

In total, 93 of roughly 500 galleries were founded successfully in the laboratory (i.e., eggs were laid), and in 43 of these galleries individuals reached adulthood. These galleries were used for behavioral analyses. We distinguished 11 behaviors (Table S3), which comprised seven cooperative behaviors that apparently raise the fitness of colony members. Every second to third day we performed scan observations of all individuals visible within a gallery (details in SI Text).

Age- and sex-specific behavior.

We used GEEs (details below) with an exchangeable correlation structure of the response variable within a cluster (= gallery identity) to identify effects of the four classes of individuals (larvae, teneral females, mature females, males) on the proportion of time spent with a certain behavior, using binomial error distributions (Fig. 1 and Table S1).

Adult female behaviors depending on gallery composition.

To check for potential effects of gallery age on behavior, we discriminated between the following successive stages of gallery development: (i) preadult gallery stage: founder female and dependent offspring (larvae and pupae) with or without eggs present (n = 2 galleries); (ii) larval-adult gallery stage: all age classes (mature/teneral beetles, larvae, and pupae) with or without eggs present (n = 31 galleries); and (iii) postlarval gallery stage: only adults without eggs, larvae, or pupae present (n = 16 galleries). Variation in sample size of galleries was caused by the fact that not all age classes were visible in every gallery. We used a first series of GEEs to identify the effect of the particular stage of gallery development on the proportion of time adult females (= teneral and mature females) spent with a certain behavior. In a second GEE series performed only with data of the larval-adult gallery stage, we analyzed whether and how task performance of adult females related to the number of adult females present and to the number of pupae and larvae present in the gallery (Table S2).

Female dispersal.

For each gallery we measured the retention period between the first appearance of a mature daughter (fully sclerotized, ready to disperse) within the gallery and the first female dispersal event (i.e., the female had left through the gallery entrance and sat on top of the medium, where it tried to escape through the cap). Using a Spearman rank correlation analysis we tested whether the dispersal delay interval related to the mean number of offspring attended, divided by the mean number of adult females present at that time.

Influence of foundress on offspring dispersal.

In the entrance tunnel foundresses either block (sit still and fully close the tunnel) or move back and forth when shuffling material to the dumps. To determine the influence of the foundress’ presence on the behavior and dispersal propensity of mature females, we experimentally removed the first centimeter of the entrance tunnel when a female was present in there (n = 11 galleries), at a stage when eggs, larvae, and adult daughters were present together with the foundress. We determined the reproductive status of the removed female by dissection to check whether we had successfully removed the foundress. We excluded one gallery from the treatment group, where we found an immature female blocking instead of the foundress. In the control group (n = 10 galleries) we removed the first centimeter of medium when no female was present in this part of the entrance tunnel. Experimental galleries were randomly assigned to treatment and control groups. Dispersal of the daughters was measured in both groups for 4 d before and 4 d after the intervention by collecting the females on top of the medium that had left the gallery through the entrance and tried to disperse through the cap of the tube (Fig. 2B).

Statistics.

We used a series of GEEs [lmer in R (43)], which are an extension of generalized linear models with an exchangeable correlation structure of the response variable within a cluster (= gallery identity), to analyze effects of dependent variables on correlated binary response variables (proportional data were transformed to binary data) and to identify factors affecting the relative behavioral frequencies per class (larvae, adult females, and males) (44). First, we tested whether the larvae, teneral females, mature females, and males show different tendencies to express the cooperative behaviors (Fig. 1 and Table S1). Second, we compared these frequencies between foundresses and their mature daughters (SI Text and Table S4). In a third series of GEEs we determined the influence of a particular developmental stage of the gallery (preadult, larval-adult, and postlarval gallery stages) on the relative behavioral frequency per class (Table S2). Finally, we modeled whether larvae and adult female numbers affected the relative frequencies of cooperative behaviors in adult females (Table S2). For the removal experiment we compared behavioral frequencies and dispersal activity between the groups using Mann-Whitney U tests, and within groups using Wilcoxon matched-pairs signed-ranks tests (Fig. 2B). All statistical analyses were performed with SPSS version 15.0 and R version 2.8.1 (43).