Behavioral evolution in the major worker subcaste of twig-nesting Pheidole (Hymenoptera: Formicidae): does morphological specialization influence task plasticity?
Polymorphism frequently correlates with specialized labor in social insects, but extreme morphologies may compromise behavioral flexibility and thus limit caste evolution. The ant genus Pheidole has dimorphic worker subcastes in which major workers appear limited due to their morphology to performing defensive or trophic functions, thus providing an ideal model to investigate specialization and plasticity. We examined worker morphology, brood-care flexibility, and subcaste ratio in 17 species of tropical twig-nesting Pheidole by quantifying nursing by major workers in natural colonies and in subcolonies lacking minors, in which we also measured brood survival and growth. Across species, majors performed significantly less brood care than minors in intact colonies, but increased rates of brood care 20-fold in subcolonies lacking minors. Brood nursed by majors had lower survival than brood tended by minors, although rates of brood growth did not vary between subcastes. Significant interspecific variation in rates of brood care by major workers did not lead to significant differences in brood growth or survival. Additionally, we did not find a significant association between the degree of major worker morphometric specialization and rates of nursing, growth, or survival of brood among species. Therefore, major workers showed reduced efficacy of brood care, but the degree of morphological specialization among species did not directly compromise task plasticity. The compact nests and all-or-nothing consequences of predation or disturbance on colony fitness may have influenced the evolution of major worker brood-care competency in twig-nesting Pheidole.
KeywordsCaste evolution Division of labor Polyethism Response threshold Trade-off
Ecological specialization is diverse in insects. It ranges from host specialization in phytophagous and parasitic species (Jaenike 1990; Fry 1996; Timms and Read 1999; Novotny et al. 2002; Dyer et al. 2007) and other aspects of resource specialization (Wcislo and Cane 1996; Rana et al. 2002) to behavioral (Johnson 2003) and morphological specialization (Gronenberg et al. 1997; Gilbert and Jervis 1998). Specialization can provide fitness benefits: Resource specialists, for example, can be competitively superior and display faster decision making than generalists (Bernays and Wcislo 1994; Bernays 1998; Larsson 2005), although this may not always be the case (Dornhaus 2008). Specialization can also have negative consequences by reducing behavioral flexibility (West-Eberhard 2003; Rueffler et al. 2007). Trade-offs between specialization and flexibility are thought to play a role in divergent evolution and speciation (Futuyma and Moreno 1988; Caillaud and Via 2000; Nosil 2002). The existence of such trade-offs has rarely been empirically demonstrated, likely due to the difficulty of accurately quantifying specialization and flexibility. Social insects, particularly ant species exhibiting quantifiable morphological specialization in the worker caste, provide a useful model for examining the nature and consequences of this trade-off.
Polymorphism, the presence of morphological variation in the worker caste, has evolved in at least 44 of 329 extant ant genera (Hölldobler and Wilson 1990; Bolton et al. 2006). Workers of different size perform different tasks in genera such as Atta (Weber 1972; Wilson 1980), Messor (Davidson 1978), Formica (Brandão 1978; Kay and Rissing 2005), Solenopsis (Wilson 1978; Tschinkel 2006), Camponotus (Busher et al. 1985), Dorlyus (Gotwald 1978; Braendle et al. 2003), and Eciton (Schneirla 1971; Topoff 1971; Powell and Franks 2006). Caste theory hypothesizes that polymorphism increases colony ergonomic efficiency (Oster and Wilson 1978). Considering the large number of tasks in an ant colony, the proliferation of subcastes to optimize social coverage is predicted, but does not appear to occur in nature (Hölldobler and Wilson 1990). Instead, species with large complex colonies, such as leafcutter and army ants, have evolved a maximum of three distinct worker subcastes, while 87% of ant genera show little or no worker polymorphism (Hölldobler and Wilson 1990; Bolton et al. 2006). This suggests that the evolution of polymorphism may be limited by developmental constraints (Wheeler 1991; Fjerdingstad and Crozier 2006) or a trade-off between morphological specialization and behavioral flexibility (Oster and Wilson 1978).
The degree of behavioral flexibility of morphologically specialized workers and its fitness consequences in ants are poorly understood. Some studies have demonstrated that workers may assume the tasks of another subcaste under circumstances of colony need (Hölldobler and Wilson 1990; Schmid-Hempel 1992), but few have measured the competency of workers performing subcaste-atypical tasks. When 90% of optimal-size leaf cutting workers were removed from colonies of Atta cephalotes, larger- and smaller-size workers cut leaves without a significant reduction in foraging rate, although increased leaf cutting by the remaining 10% of the optimal size class may have compensated for the decreased performance of the compensatory size classes (Wilson 1983). Solenopsis invicta colonies altered to contain only media or major workers raised brood at lower rates than fully polymorphic colonies or colonies comprised of only small workers (Porter and Tschinkel 1985). Clearly, the role of trade-offs between specialization and flexibility in caste evolution requires further consideration. To address this question, the effectiveness of morphologically specialized subcastes at performing subcaste-atypical behaviors in a group of closely related species should be examined. We therefore studied the fitness consequences of morphological specialization and behavioral flexibility in 17 species of Pheidole, an ant genus exhibiting striking worker size variation and subcaste-based division of labor.
Pheidole provides an ideal model for studying the relationship between worker morphology and behavior; the genus is hyperdiverse, containing more than 600 species in the New World and nearly 900 species worldwide (Wilson 2003; Bolton et al. 2006). The majority of Pheidole species exhibit complete dimorphism: A colony contains two morphologically distinct worker subcastes (larger major workers and smaller minor workers) with no intermediate size workers (Wilson 1971). Minor workers typically care for brood, maintain the nest, and forage. Majors, in contrast, appear to be specialized for colony defense, food storage, or seed milling (Wilson 2003), although their behavioral repertoire varies among species (Wilson 1984; Patel 1990; Sempo and Detrain 2004). Pheidole majors are not isometrically scaled up minors; instead, major worker morphology is at least partially dissociated from that of minors, particularly in terms of head, petiole, and alitrunk shape (Pie and Traniello 2007). Wilson (2003) hypothesized that the evolution of the major worker subcaste was a key innovation, which along with small body size and rapid colony reproduction lead to the striking diversification of Pheidole. He theorized that concentrating the morphological requirements for defense and food processing in major workers could have allowed the morphology of minors to be simple and energetically inexpensive, increasing competitive ability and thus the ecological success of Pheidole (Wilson 2003).
Specialized morphologies selected to enhance the defensive or trophic functions of Pheidole majors could impair their ability to perform other tasks. For example, fine manipulation of delicate eggs and larvae may be difficult to achieve with mandibles designed for attacking enemies or crushing seeds. It is reasonable to assume that majors that are more morphologically distinct from minors may be less able to perform minor typically tasks, although this has not been demonstrated. Pheidole major workers avoid brood in the presence of minors (Wilson 1985) and are in general rarely involved in brood care. Majors that are larger in size tend to have smaller behavioral repertoires that include fewer nursing acts (Wilson 1984). Nevertheless, majors in some Pheidole species have the ability to perform brood care when subcaste ratios are experimentally modified. Pheidole pubiventris majors greatly increased their rate of brood care and other tasks when minors were removed entirely (Wilson 1985) and majors of Pheidole morrisi performed more brood-care acts in colonies altered to contain more than 75% majors (Brown and Traniello 1998). Pheidole guilelmimuelleri and Pheidole megacephala majors could also be induced to increase brood care in colonies containing high proportions of majors (Wilson 1984). These results suggest that majors can be behaviorally flexible when required, but do not demonstrate that they are able to perform brood care as efficaciously as minors. There is limited evidence of the fitness correlates of flexibility of Pheidole majors: Pheidole bicarinata majors raised brood equally well as minors (Wheeler and Nijhout 1984), whereas majors in Pheidole pallidula and P. megacephala were unable to raise eggs to adults (Chang 1985; Sempo and Detrain 2004).
Sociometric variation may also be associated with morphological specialization and task plasticity in Pheidole. Colony demography appears to be more flexible than worker morphology and may be influenced by factors such as colony age (Oster and Wilson 1978), reproductive condition (Kaspari and Byrne 1995), and the predatory and/or competitive environment (Calabi and Traniello 1989; Passera et al. 1996; McGlynn and Owen 2002; Yang et al. 2004). Ecological conditions may therefore obscure adaptive interspecific variation in subcaste ratios. Nevertheless, interspecific differences in subcaste ratio have been found to correlate with morphology and behavior in Pheidole. For example, Pheidole species with energetically costly alates have higher proportions of majors (Kaspari and Byrne 1995), and in ten species of geographically disparate Pheidole, major worker repertoires significantly increased with the proportion of majors in a colony (Wilson 1984). Morphological specialization may therefore be associated with colony demography, either as a consequence of the reduced flexibility of specialized major workers or, as hypothesized by Wilson (2003), because the development of a behaviorally limited major worker subcaste reduces the proportion of such specialists needed to maintain colony efficiency. Additionally, larger colony size is generally correlated with smaller worker body size (Karsai and Wenzel 1998). Major workers that are more morphologically specialized than minor workers could require higher per capita resource investment, potentially reducing colony size.
If major workers are indeed a key innovation in Pheidole, then determining the effect of morphological specialization on their task plasticity and competency is crucial to understanding social evolution and diversification in the genus. To this end, we investigated nursing behavior in a diverse community of twig-nesting species of Pheidole in Amazonian Ecuador. We chose this community of tropical Pheidole because it is species rich, and morphological and behavioral differentiation may be found among coexisting ant species (Davidson 1977; Hansen 1978; Mehlhop and Scott 1983; Wetterer 1995; Kaspari 1996b; Johnson 2001). Coexisting species of Pheidole may be more likely to vary along a continuum of morphological and behavioral specialization than a random geographic sampling of species, as interspecific competition is though to limit similarity among sympatric species (MacArthur and Levins 1967; Tilman 1994). We hypothesized that major workers would contribute little to brood care in intact colonies, but vary interspecifically in degree of behavioral flexibility, measured as their ability to provide competent brood care in colonies with altered subcaste demographies. If majors indeed show brood-care flexibility, as suggested by previous work in Pheidole (Wheeler and Nijhout 1984; Wilson 1984; Brown and Traniello 1998; Sempo and Detrain 2004), we can then test the hypothesis that a morphological specialization/behavioral flexibility trade-off exists among the Pheidole species in our Amazonian community. We test four predictions of this hypothesis: (1) the specialized major worker subcaste will be less efficacious than the minor caste at brood care overall, (2) behavioral flexibility in major workers will be negatively correlated with increased species-level morphometric specialization, (3) the survival and growth of brood nursed by majors, as indicators of colony-level fitness, will be negatively correlated with increased species-level morphometric specialization, and (4) the proportion of major workers and colony size will be negatively correlated with the degree of species-level morphological specialization. Lack of support for these predictions will indicate a lack of evidence for a trade-off between morphology and behavior among our sympatric species.
Materials and methods
Study site and colony collection
Pheidole species studied and sample sizes for ethograms of intact, queenright colonies, and for the major worker behavioral flexibility experiment
Monocaste subcolonies (setup and at 24 h)
Monocaste subcolonies (at 1 week)
Monocaste subcolonies (at 2 weeks)
P. sp. nr. embolopyx
Twig nests were carefully opened above a Fluon®-lined container to extract all colony members. Adult ants and immatures were placed in artificial circular Petri dish nests (6 cm diameter × 1.5 cm height) constructed with moist dental stone bottoms to maintain humidity and placed inside larger Petri dishes (10 × 2 cm). Colonies were kept at ambient temperature (25–30°C) and humidity (>80%) in a screened laboratory at TBS. All colonies were fed ad lib protein, fat, and carbohydrate foods (sugar water, peanut butter, cookies, seeds, baby food, tuna, and termites). Dental stone was moistened twice daily. Voucher specimens, collected for morphological measurements and species confirmation, have been deposited in the Museum of Comparative Zoology at Harvard University.
Censusing colonies and quantifying subcaste behavior
Intact queenright colonies were censused within 48 h of collection to record demographics (colony size, number of immatures, major/minor, and brood/worker ratios). Because larval instars could not be reliably distinguished for all species, the developmental stages of brood were estimated categorically: small (eggs and microlarvae smaller than the head of a minor of the same species), medium (larvae smaller than the length of the head and thorax of a minor), large (larvae larger than the head and thorax of a minor), and pupae. These size groups approximate the three-to-four larval instars found in Pheidole (Hölldobler and Wilson 1990; Cassill et al. 2005). Behavioral data were collected from each colony 24–76 h after establishment in an artificial nest. A circular grid (diameter = 100 mm) of 1 × 1 cm quadrats was placed over each nest and the number of quadrats containing brood was counted. Nursing behavior was the main focus of the study; therefore, quadrats without brood were not observed. Brood piles were never sufficiently large to extend to the edge of the 100-mm circle, so only full quadrats contained brood. Ants occasionally moved brood piles when colonies were disturbed by movement or noise; when this occurred, data were discarded and the colony was allowed to acclimate for minimally 1 h before resuming observations. Each colony was observed for 60 min, divided equally among quadrats with brood. Each quadrat was observed continuously for the resulting time period. We used this method to quantify differences in rates of overall activity and brood-care acts between majors and minors under normal subcaste ratios, rather than estimate the total behavioral repertoire of either subcaste (Sempo and Detrain 2004).
The number of major workers, minor workers, and brood in each quadrat was recorded at the start of an observation period. All behaviors were recorded by subcaste. Quadrats were small enough so that the activity of all ants within could be continuously monitored. We did not record stationary or walking ants; all other behaviors observed were considered tasks and were recorded to assess activity rate and brood-care rate (see ESM Table S1 for a list of brood-care and nonbrood-care tasks). Food was present in the foraging chamber during data recording so that brood-feeding tasks could be observed. Activity rate was measured as acts/ant/hour to standardize for differences in colony size. Brood-care rate was measured as brood-care acts/ant/hour. Because of intercolonial variation in brood/worker ratios, we also investigated measuring brood care as acts/ant/brood/hour; however, this metric appeared bias by colony size (Spearman’s rho = 0.60, P < 0.0001). Brood-care acts/ant/hour did not appear biased by brood size (Spearman’s rho = −0.05, P = 0.672) and therefore was chosen. Subcaste behavior was recorded in 81 colonies using this method (Table 1; for average number of acts observed, see ESM Table S2).
Establishment of monocaste subcolonies
Demographic data used to determine subcaste ratio for 17 twig-nesting Pheidole species
Proportion of majors
P. sp. nr. embolopyx
Monitoring behavior, brood survival, and brood growth in monocaste subcolonies
Adult ants in each subcolony were observed for two 30-min periods, once at 24 h after subcolony establishment to immediately measure response to brood and once at 1 week to measure changes in nursing response. Brood care was again measured as acts/ant/hour. Brood piles in subcolonies were small, allowing continuous monitoring and recording of all brood-care tasks.
Brood were censused after 1 week and again after 2 weeks in monocaste subcolonies. Brood absent at a census were considered dead, because workers were observed consuming dead larva and pupae. We measured brood growth as the total proportion of brood that advanced in developmental stage and separately analyzed larval growth (molting from small to medium larva or from medium to large larva), pupation (metamorphosis from large larva to pupa), and adult eclosion (pupa to callow worker). Growth was inferred conservatively by comparing the original number and stage of brood placed in the subcolonies to the brood present at 1 and 2 weeks. In cases of ambiguity, we chose the simplest explanation for changes in brood ratios. For example, if a subcolony had five large larva and six pupa at week 1 and four large larva and six pupa at week 2 but no enclosed workers or deceased brood, we would score it as the death of one large larva rather than the death of a pupa and the pupation of a large larva. Subcolonies were checked daily for newly eclosed workers, which were removed and placed in the main colonies to maintain consistent subcaste ratios and age distributions. Records were made of all workers removed to ensure accurate data on brood development. If either of the two subcolonies had no surviving brood at 1 week, the colony was not followed to 2 weeks: 63 colonies remained at the end of 2 weeks; 16 colonies were discarded due to escapes or high brood mortality (Table 1).
Measuring colony size, subcaste ratio, and worker morphology
Mean colony size and proportion of major workers were measured for each Pheidole species using colony censuses for 79 tested colonies, as well as 109 additional colony censuses (Table 2). These additional colonies were collected and censused at TBS, some during a separate study (Mertl et al. 2009), others during an earlier field season (January to May 2003). Only queenright colonies with at least one major were used to record colony size and subcaste ratio. Colonies lacking majors appeared to be incipient, with small numbers of nanitic minor workers, and these colonies (ten in all) were therefore excluded. Morphological measurements were recorded from one major and one minor in three colonies per species using IMAGE-J 1.38 (Rasband 2007) on digital images taken with a JVC digital camera attached to a Leica MZ16 stereomicroscope. High-resolution images were made using Auto-Montage (Syncroscopy, Division of Synoptics, LTD, Frederick, MD, USA). Only species for which at least three colonies were collected were included in morphological analyses. For identified species, we compared (when possible) our values to those of Wilson (2003) to validate correspondence between our measurements and type specimens. Four linear measurements were made on eight reliable landmarks: head width (HW), the maximum width of the head directly above the eyes; head length (HL), the distance from the top of the occipital lobes to the lower margin of the clypeus; petiole width (PW), the dorsal width of the petiole at its widest point; and the distance between the outer mandibular condyles (MC). The surface area of three mandibular characters was recorded: mandible area (MA), area of the dorsal surface of a single mandible, not including dentition, and the dorsal surface area of the first (ATA1) and second apical tooth (ATA2). These measurements were chosen to examine the morphology relevant to brood-care competence. The first three measurements (HW, HL, and PW) reflect difference in overall body size and head-to-gaster balance between majors and minors (Pie and Traniello 2007). Size and maintenance of body position may impact a worker’s ability to manipulate and carry brood. The other four measurements (MA, MC, ATA1, and ATA2) reflect details of mandibular morphology and dentition that may affect fine-scale manipulation required for turning, grooming, and feeding brood.
Statistical analyses evaluated subcaste trends across species as well as interspecific variation in nursing competence. For the former analysis, we pooled data from all colonies in all 17 species (n = 81 intact colonies, n = 79 monocaste subcolonies), and for the latter one, we examined only ten principal study species for which we had data on minimally three colonies (P. allarmata [n = 12], P. lemnisca [n = 11], P. pholeops [n = 9], P. horribilis [n = 7], P. amazonica [n = 7], P. scalaris [n = 5], P. embolopyx [n = 5], P. ALM023 [n = 4], P. gagates [n = 4], and P. metana [n = 3]; Table 1).
Activity rates and brood-care rates were significantly nonnormal for both subcastes in intact colonies and in monocaste subcolonies (Shapiro–Wilk test, P < 0.0001). Therefore, nonparametric Wilcoxon signed rank and Kruskal–Wallis tests were performed on untransformed data to compare rates of behavior. We used data from the 76 colonies for which both sets of data were available to make paired comparisons of brood-care behavior between intact colonies and monocaste subcolonies (Table 1).
Brood survival (total survival and brood-stage specific survival) and growth in monocaste subcolonies were also significantly skewed for both subcastes (Shapiro–Wilk test, P < 0.0001). The nonparametric Wilcoxon signed rank test was therefore used to compare subcaste performance in nursing. A two-tailed Spearman’s rho was used to compare brood survival and growth with brood-care rates, and the Kruskal–Wallis test was used to analyze interspecific differences.
Colony size and subcaste ratio were log transformed prior to analysis to improve normality and permit analysis of variance (ANOVA). Due to differences in sample size among the species used for interspecific comparison (four to 62 colonies; Table 2), we used the Brown–Forsythe test for equal variances to compare species differences prior to ANOVA (Brown and Forsythe 1974; Conover et al. 1981). Because variance in colony size was significantly different among species (F9, 166 = 2.69, P = 0.006), we used a Welch ANOVA to examine interspecific variation in colony size (Welch 1951). However, there was no significant interspecific difference in the variance in proportion of majors (F9, 166 = 0.76, P = 0.653) and no significant correlation between species sample size and the standard deviation of the proportion of majors (F1, 9 = 0.067, P = 0.803, adjusted R2 = −0.116). These two tests support our use of a standard ANOVA to examine interspecific differences in the proportion of majors.
To determine the effect of morphological specialization, we examined the degree of difference between majors and minors for each of our seven morphological measurements rather than the values themselves. We first subtracted the values of each metric for the minor from the values for the major and divided by the value for the minor for each colony measured. We used ANOVA to determine if significant differences in morphological specialization existed between species, using head width, a reliable indicator of minor and major worker size (Pie and Traniello 2007; D. Feener, unpublished data), as an estimator of morphological specialization. We then averaged the degree of difference values for the three colonies measured for each species to obtain the mean degree of specialization for each measurement for each species. These seven variables (mean of degree of difference between majors and minors for HW, HL, PW, MC, MA, ATA1, and ATA2 for each species) were entered into a principal components analysis (PCA). Components were extracted based on the relative percent variance criteria. Measurements were standardized by conversion to Z scores prior to PCA analysis of the unrotated correlation matrix (Gotelli and Ellison 2004).
Linear multiple regression analysis was used to examine effects of the morphological principal components from the PCA analysis on subcaste specific brood-care behavior (mean brood-care acts/ant/hour at 1 week), brood survival (mean percent at 2 weeks), brood growth (mean at 2 weeks) as well as the mean proportion of majors and mean colony size among species. Separate regressions were run for majors and minors. A linear model produced a good fit to the data, based on residual plots showing no correlation between residuals and predicted values and normally distributed residuals. Attempts to fit data to nonlinear models did not produce significant results.
Statistical tests were performed using JMP version 5.0.1 (SAS Institute Inc., Cary, NC, USA, 1989–2002) and SPSS version 11.0.4 (SPSS Inc., Chicago, IL, USA, 1989–2005). We applied the method of Benjamini and Hochberg (1995) to account for multiple statistical comparisons. All P values greater than P = 0.009 were not considered significant based on the highest P value to satisfy the constraint P(21) = 0.009 ≤ (21/74) × 0.05 = 0.014.
Minor and major worker brood-care behavior in intact and monocaste colonies
Behavioral differences between major and minor workers
Minors reduced their frequency of brood care in monocaste subcolonies, a significant decrease in nursing rates below that in intact colonies at both 24 h (Z = −3.4, P = 0.001) and 1 week (Z = −2.6, P = 0.009; Fig. 2b). Minors showed no change in brood care between 24 h (4.4 ± 3.4 acts/ant/h) and 1 week (5.6 ± 6.7, Z = −1.0, P = 0.337). Twenty-four hours after subcolony establishment, majors performed significantly less brood care than minors (Z = −3.6, P < 0.0001). Rates of brood care were also significantly lower in majors at 1 week in monocaste groups (Z = −3.2, P = 0.001). Rates of brood-care behavior in monocaste subcolonies were not correlated with those recorded in intact colonies for majors (24 h: Spearman’s rho = 0.24, P = 0.039, 1 week: rho = 0.28, P = 0.023) or minors (24 h: rho = 0.22, P = 0.055, 1 week: rho = 0.18, P = 0.143).
Rates of larval feeding (oral trophallaxis and placing solid food) were significantly higher in minors at 24 h (majors 0.41 ± 0.87 acts/ant/h, minors 0.60 ± 0.64, Wilcoxon signed rank test, Z = −2.9, P = 0.004), as well as at 1 week (majors 0.65 ± 1.07, minors 0.74 ± 0.88, Z = −3.0, P = 0.003). At 24 h, brood were piled in all but seven major worker subcolonies and five minor worker subcolonies, with no significant difference between subcastes (χ2 = 0.36, P = 0.549). Brood were piled in all subcolonies at 1 week.
Interspecific differences in subcaste brood care
There was no significant interspecific variation in activity rate for majors (Kruskal–Wallis test, χ92 = 6.9, P = 0.65) or minors (χ92 = 9.5, P = 0.40) among the ten principal study species. Minor worker brood-care rates were not significantly different across species (χ92 = 11.7, P = 0.23). Majors, however, showed interspecific variation in rates of brood care in intact colonies (χ92 = 32.2, P = 0.0002). P. amazonica majors had the highest rate of brood care (0.49 ± 0.57 acts/ant/h), followed by P. metana (0.44 ± 0.19) and P. gagates (0.24 ± 0.21). In contrast, P. horribilis and P. ALM023 majors were never observed performing brood care (Fig. 1b).
Brood care (acts/ant/hour), survival, and growth in monocaste subcolonies
Brood-care activity (24 h)
Brood-care activity (1 week)
Brood survival (2 weeks)
Total brood growth (2 weeks)
P. sp. nr. embolopyxa
Competency of minor and major worker brood care
Brood survival and growth in monocaste subcolonies
Interspecific differences in subcaste brood-care competency
Brood survival was the highest in major worker subcolonies of P. allarmata (42 ± 25%), P. pholeops (42 ± 40%), and P. gagates (42 ± 21%). Majors of P. ALM023 produced the lowest survival of brood (15 ± 4%; Table 3). However, there was no significant interspecific variation in brood survival among the ten principal study species for either majors or minors at 1 (Kruskal–Wallis, χ92 = 7.7, P = 0.566 for majors; χ92 = 11.9, P = 0.222 for minors) or 2 weeks (χ92 = 3.5, P = 0.940 for majors; χ92 = 10.3, P = 0.326 for minors).
The highest growth rates for brood tended by majors were recorded in P. gagates (21 ± 4%) and P. embolopyx (21 ± 22%). The lowest was in P. scalaris (6 ± 6%; Table 3). There was no significant interspecific variation in overall brood growth for either majors or minors at 1 (χ92 = 4.6, P = 0.865 for majors; χ92 = 4.0, P = 0.909 for minors) or 2 weeks (χ92 = 3.0, P = 0.965 for majors; χ92 = 8.8, P = 0.451 for minors). Statistical power was low for interspecific comparisons of both brood survival and brood growth, ranging between 0.08 and 0.39.
Morphology, colony demography, and behavioral flexibility
Interspecific variation in morphometric specialization
The degree of difference in head width between majors and minors varied significantly among the ten principal study species (ANOVA, F9, 29 = 11.4, P < 0.001). The largest majors (P. horribilis, mean HW = 1.9 mm ± 0.06) had a head width nearly four times that of the smallest majors (P. scalaris, mean HW = 0.5 mm ± 0.05). Minor workers showed less size variation; in the largest species (P. horribilis, mean HW = 0.75 mm ± 0.02), the size of minors was roughly twice that of the smallest species (P. lemnisca, mean HW = 0.29 mm + 0.03). Three principal components had eigenvalues >1 and together explained 85.1% of the interspecific variation in morphological distance between majors and minors. PC1 explained 42.0% of the variation and represented differences in MA and HW. PC2 represented differences in HL and PW and explained 24.9% of the variance. PC3 represented differences in MC, ATA1, and ATA2 and explained 18.3% of the variance (ESM Table S4).
Interspecific variation in colony demography
Interspecific differences in colony size (gray) and subcaste ratio (white) among ten principal study species
The relationship between morphology, colony demography, and major worker brood care
Our study of interspecific variation in nursing in Pheidole represents the first investigation of the relationship between subcaste morphological specialization and behavioral flexibility in a diverse clade of sympatric ant species. Our results demonstrate that major workers were less efficacious than minor workers at performing brood care, supporting the hypothesis of a trade-off between major worker specialization and brood-care flexibility in Pheidole. However, despite evidence of significant interspecific variation in behavior and morphology among major workers, we did not find a significant correlation between degree of morphological specialization and major worker brood care performance at the species level. Our work adds to a growing body of literature demonstrating reduced flexibility in specialized worker castes (Porter and Tschinkel 1985; Johnson 2005), suggesting its importance in tropical ecosystems with high levels of ecological specialization (Dyer et al. 2007). However, the extent of such trade-offs is still poorly understood in insect communities (Fry 1996). We interpret our results in light of the biology of tropical twig-nesting Pheidole and caste theory.
Upregulation of brood care by major workers
Major workers nursed significantly less than minor workers in intact colonies. Major workers of P. amazonica had the highest rate of brood care (0.49 acts/ant/hour), which was nevertheless substantially below the lowest rate recorded for minors (3.52 acts/ant/hour, P. pholeops). Major workers in monocaste subcolonies significantly increased their rates of nursing, but still remained significantly below those of minors even after 1 week of exposure to brood. Conversely, minor workers significantly decreased brood care in monocaste subcolonies, reducing rates of nursing by approximately 20% compared to those in intact colonies. We did not identify the cues or signals responsible for these changes in brood care. The arrangement of our parent and monocaste subcolonies presumably allowed workers to perceive pheromone signals from the colony and queen, but workers in subcolonies may have had reduced olfactory contact with the parent nest. Therefore, minors may have decreased their rates of brood care due to a reduced exposure to colony and/or queen pheromones. Nevertheless, identifying the cause of changes in nursing rates would not change our interpretation of the upregulation of brood care by major workers. Even if majors were in part responding to changes in pheromone concentrations, their significant increase in nursing from 24 h to 1 week suggests that they possess the ability to respond to colony need by upregulating brood care, perhaps in an effort to replenish the number of minors.
Major workers of all species in our study upregulated brood care in the absence of minor workers, consistent with other studies of Pheidole that have failed to identify any species whose major workers completely lacking the ability to nurse (Wheeler and Nijhout 1984; Wilson 1984, 1985; Brown and Traniello 1998; Sempo and Detrain 2004). This suggests that a modest level of nursing flexibility may not be energetically costly to maintain in Pheidole major workers. Brain tissue is considered costly to produce and maintain; mated queens of Messor and Pogonomyrmex, for example, show reduced volume of the optic lobes and medulla compared to virgin queens, presumably because these metabolically expensive tissues are less important after the mating flight (Julian and Gronenberg 2002). Maintenance of learning ability has also been shown to decrease the competitive ability of Drosophilia larva (Mery and Kawecki 2003; Mery and Kawecki 2004). However, the nursing ability of Pheidole majors may be related to routine social activities that require similar sensory capabilities. For example, major workers in all 17 Pheidole species were observed performing trophallaxis with minor workers in intact colonies; trophallaxis with brood may not require additional neural circuitry. Nursing behavior by majors may also be related to rescuing brood during nest disturbance or transporting brood during nest emigration. Differentiation between majors and minors occurs in the late larval instars (Wheeler and Nijhout 1984), so it is possible that maintenance of nursing behavior in major workers is a result of this shared developmental pathway. The elimination of brood-care flexibility in the major worker subcaste could be more costly than its retention.
Subcaste differences in nursing competence
Although major workers upregulated brood care in response to social needs, they were significantly less capable at tending brood than minor workers. Survival of mature (medium and large) larvae was particularly poor in major worker subcolonies; these larvae declined significantly faster than same-stage larvae tended by minors. Overall survival of brood was lower when immatures were in the care of major workers; this difference in the nursing competence of subcastes was significant at both 1 and 2 weeks. This suggests that major worker specialization results in reduced behavioral flexibility: Majors were less capable than minor workers at successfully raising larvae. Previous work on S. invicta demonstrating reduced brood production by major in comparison to minor workers (Porter and Tschinkel 1985) agrees with our finding. Further studies on other polymorphic ant genera are needed to confirm the generality of this apparent trend. Pheidole majors and minors were equally capable at raising pupae, perhaps because pupae do not require feeding. Rates of trophic tasks (oral trophallaxis and feeding solid food) were significantly lower for majors than minors, a result similar to preliminary data indicating that larvae raised by majors of P. morrisi gain less weight than larvae raised by minors (Brown and Traniello 1998). However, rates of larval growth, pupation, and eclosion did not vary significantly between majors and minors. Together, these results suggest that although majors are less able to maintain brood, those that survive develop more quickly. Nutritionally deficient larvae of holometabolic insects may pupate after they attain a lower critical weight (Safranek and Williams 1984; Davidowitz et al. 2003), and reduced colony nutrition in ants leads to smaller workers and sexuals (D. Cassill, unpublished data). If larvae nursed by Pheidole majors have similar nutritional deficits, they may pupate at a smaller size and eclose as smaller adults. Therefore, the lower quality of brood care provided by majors could nevertheless contribute to colony fitness.
Interspecific variation in major worker morphology and nursing competence
We found significant interspecific differences in the degree of brood care by majors in intact colonies and significant interspecific variation in morphometric specialization between subcastes, but no significant effect of major worker morphology on rates of brood care, growth, or survival in monocaste subcolonies. Although our pooled results suggest that the morphological specialization of major workers reduces their brood care competency, interspecific comparisons do not provide additional evidence of such a trade-off. This does not imply that a trade-off does not exist in Pheidole; the relatively low statistical power of our regression suggests that detecting it may require a larger sample size. However, it is possible that morphological specialization is not a significant correlate of brood-care competence. Although we assumed that morphological specialization would strongly reduce nursing ability, other traits may have a greater impact on brood-care behavior. For example, neuroanatomy and neurochemistry change with worker age and subcaste in Pheidole dentata and may be causally related to task performance (Seid and Traniello 2005; Seid et al. 2008; Muscedere et al. 2009). Successful brood care also depends on the ability of workers to detect and respond appropriately to larval hunger cues (Cassill and Tschinkel 1995). Majors could have reduced olfactory sensitivity and/or responsiveness to such cues, rendering them less capable nurses. Major workers could also lack the glandular anatomy required for maintaining larval nutrition and hygiene. An interspecific analysis of neurological and/or physiological traits of majors rather than external morphology could demonstrate causal relationships to behavioral flexibility.
We found no correlation between rates of brood care and survival for immatures tended by majors or minors. This could be due to the greater importance of detection of and timely response to larval cues and/or the presence of worker pheromone to larval survival than rates of brood-care activity per se. Our result also suggests that the requirements for nursing vary by brood stage, some needing less care to survive. However, our assessment of brood-care rate using half-hour observation periods may not have been adequate, especially if nursing behavior was sporadic rather than continuous. This does not invalidate our conclusions, as our results for brood survival and growth remain unbiased. Rather, it suggests that experimentally estimated rates of brood care may be only weakly correlated with brood health, emphasizing the need to measure fitness indicators such as brood survival and growth.
Variation in major worker task flexibility in litter-nesting Pheidole
There was a surprising lack of significant interspecific variation in nursing competence among major workers of our Pheidole species. This may be due to high variation among colonies, as shown by the low statistical power of our interspecific comparisons. However, it may also be due to the nesting habits of the study species. We selected sympatric twig-nesting species based on the assumption that ecological interactions were responsible for sociobiological differentiation in Pheidole communities, leading to stronger and more easily interpretable interspecific variation than would be found in a random geographic sample. Indeed, major workers in our Amazonian species showed significant morphological variation, and field observations indicated that majors varied behaviorally as well. For example, P. allarmata and P. horribilis majors were rarely observed to forage, whereas majors of P. amazonica, P. embolopyx, and P. gagates consistently foraged and defended food sources (Mertl et al., unpublished data). We also found significant interspecific sociometric variation, measured as colony size and subcaste ratio. This indicates that selection has diversified twig-nesting Pheidole species in the morphology and behavior of major workers and subcaste investment patterns. However, selection for nursing flexibility would require that colonies experience predation or disturbance in which the worker population of a colony is partially eliminated. This may not occur in twig-nesting species because twig nests are spatially compact. Potential colony dangers such as army ant predation, flooding, and other disturbance (Byrne 1994; Kaspari 1996a; Mertl et al. 2009) may have all-or-nothing consequences leading to either total colony mortality or intact survival. For this reason, twig-nesting Pheidole may not commonly experience situations in which the colony would benefit from brood-care flexibility in major workers, leading to a lack of selection-driven differentiation in this trait.
Seven species were rare at our study site and could not be included in our interspecific comparisons due to small sample sizes. However, in two of these species, the response of major workers to the need for brood care was striking. Pheidole cephalica and Pheidole fissiceps rarely nest in twigs; P. cephalica was typically found in large rotting logs, while P. fissiceps used a broad range of nest sites, including logs, mud, and cavities of plants (Mertl et al., unpublished data). Majors of P. cephalica produced the highest rates of brood survival (90%) and brood growth (50%) of all species, whereas majors of P. fissiceps scored among the lowest (10% survival, 10% growth; Table 3). These preliminary results suggest that selection along the specialization/nursing flexibility axis could be stronger among Pheidole with larger colony size and diffuse nest structure. Therefore, we cannot conclude that interspecific variation in the competence of major worker brood care does not exist in Pheidole, both because of the low statistical power of our interspecific comparisons and the homogenous nesting habit of the species we tested.
Sociometric variation and colony-level regulation of task performance
Our hypothesis predicted proportionally fewer majors in Pheidole species having specialized majors, but this prediction was not supported. Instead, there was a positive correlation between subcaste morphological divergence and proportion of majors in a colony, mainly due to the relationship between mandibular characteristics and subcaste ratio. This suggests that the proportion of majors in a colony is not reduced in species with more morphologically specialized majors, but may actually increase in major workers with mandibular specializations. It appears that relatively simple models suggesting reduced colony investment in specialized majors due to flexibility constraints or colony efficiency do not explain subcaste ratios in these species. Indeed, caste structure rarely follows the predictions of such models in empirical studies (Schmid-Hempel 1992). Additionally, the relationship between individual and colony-level specialization and flexibility is not always intuitive in Pheidole. P. dentata colonies show enemy-specific recruitment of major workers to fire ants (Solenopsis spp.; Wilson 1976), but colonies do not respond to extended fire ant pressure by increasing production of major workers (Johnston and Wilson 1985). Yet, P. dentata colonies “learn” to respond defensively to a novel competitor (Tetramorium caespitum) and recruit more majors after repeated assaults (Carlin and Johnston 1984). P. pallidula, in contrast, readily increases soldier production in response to cues associated with intraspecific competition (Passera et al. 1996). Defensive strategy in this species also involves both majors and minors (Detrain and Pasteels 1992). Subcaste ratios are influenced by a wide range of factors (Oster and Wilson 1978; Calabi and Traniello 1989; Kaspari and Byrne 1995; Passera et al. 1996; McGlynn and Owen 2002; Yang et al. 2004), leading to high intercolonial variability, which renders species-typical subcaste patterns difficult to ascertain. Therefore, colony sociometry and the regulation of subcaste investment can add plasticity to the social phenotype, albeit inconsistently, and may in part explain why generic trends in major worker individual flexibility proved difficult to identify in our study.
Subcaste evolution and hyperdiversity in Pheidole
The relationship of major worker morphology and task flexibility to colony sociometry and fitness in Pheidole is complex, and the patterns that emerge from the limited available literature are unclear. Our results suggest that specialization in major workers has reduced their behavioral flexibility, although it is not certain that morphometric specialization is the direct cause of reduced nursing competency. The lack of interspecific variation in major worker competency also suggests that nesting biology may have been an important factor in the evolution of division of labor in the genus. If major worker nursing plasticity does not have significant consequences for litter-nesting Pheidole due to their small spatially compact nests, the specialization of majors in these species may not be constrained by the need for competent nursing flexibility. The combination of dimorphism and nesting habit might have allowed Pheidole to increase competitive ability, niche diversity, and ecological dominance, although more work is needed test this hypothesis. Additional investigations into the causes and consequences of morphological evolution and plasticity in Pheidole major workers, including comparative analyses with other polymorphic litter-nesting genera, could begin to unravel the complex interaction between ecology, morphology, sociometry, and behavior that favored the diversification of Pheidole.
We thank the directors and staff of Tiputini Biodiversity Station for their assistance in obtaining permits and support in the field. We thank Clifton Meek, Megan Johnson, Brian Henry, Noah Reid, Scott Appleby, Elise Koncsek, Frank Azorsa Salazar, Wendy Mertl, and Winston McDonald for field and laboratory assistance. We are very grateful to Stefan Cover for confirming Pheidole identifications, to Dr. Gary Alpert for instruction on digital imaging, to Dr. Marcio Pie for suggestions on morphometric analyses, and to Dr. Marc Seid for assistance developing ethogram techniques. We thank Dr. Michael Kaspari, Dr. Michael Sorenson, and two anonymous reviewers for constructive comments on the manuscript. This work was funded by a National Science Foundation Graduate Research Fellowship awarded to ALM. JT was funded by NSF Grants IOB 0725013 and 0724591 awarded to J. Traniello and W. Gronenberg. Voucher samples were collected and transported under permit 017-IC-FA-PNY-MA issued to ALM by the Ecuadorian Ministry of the Environment. This work complied with all current laws of Ecuador and the USA. The authors declare that they have no conflict of interest.
- Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B 57:289–300Google Scholar
- Bolton B, Alpert G, Ward PS, Naskrecki P (2006) Bolton’s catalogue of ants of the world 1758–2005. Harvard University Press, Cambridge, MAGoogle Scholar
- Busher CE, Calabi P, Traniello JFA (1985) Polymorphism and division of labor in the Neotropical ant Camponotus sericeiventris Guerin (Hymenoptera: Formicidae). Ann Entomol Soc Am 78:221–228Google Scholar
- Chang VCS (1985) Colony revival, and notes on rearing and life history of the big-headed ant. Proc Hawaii Entomol Soc 25:53–58Google Scholar
- Gotelli NJ, Ellison AM (2004) A primer of ecological statistics. Sinauer Associates, Sunderland, MAGoogle Scholar
- Hölldobler B, Wilson EO (1990) The ants. Harvard University Press, Cambridge, MAGoogle Scholar
- Johnston AB, Wilson EO (1985) Correlates of variation in the major/minor ratio of the ant, Pheidole dentata (Hymenoptera: Formicidae). Ann Entomol Soc Am 78:8–11Google Scholar
- Mertl AL, Ryder Wilkie KT, Traniello JFA (2009) Impact of flooding on the species richness, density and composition of Amazonian litter-nesting ants. Biotropica. doi:10.1111/j.1744-7429.2009.00520.x
- Oster GF, Wilson EO (1978) Caste and ecology in the social insects. Princeton University Press, Princeton, NJGoogle Scholar
- Rasband WS (2007) ImageJ. v 1.38. U.S. National Institutes of Health, Bethesda, MDGoogle Scholar
- Schneirla TC (1971) Army ants: a study in social organization. Freeman, San Francisco, CAGoogle Scholar
- Tschinkel WR (2006) The fire ants. Harvard University Press, Cambridge, MAGoogle Scholar
- Weber NA (1972) Gardening ants: the attines. American Philosophical Society, Philadelphia, PAGoogle Scholar
- Welch BL (1951) On the comparison of several mean values: an alternative approach. Biometrika 38:330–336Google Scholar
- West-Eberhard MJ (2003) Developmental plasticity and evolution. Oxford University Press, New York, NYGoogle Scholar
- Wilson EO (1971) The insect societies. Harvard University Press, Cambridge, MAGoogle Scholar
- Wilson EO (1978) Division of labor in fire ants based on physical castes (Hymenoptera: Formicidae: Solenopsis). J Kans Entomol Soc 51:615–636Google Scholar
- Wilson EO (2003) Pheidole in the new world: a dominant, hyperdiverse ant genus. Harvard University Press, Cambridge, MAGoogle Scholar