Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
OIKOS 110: 231 /240, 2005 Evolution of allometries in the worker caste of Dorylus army ants Caspar Schöning, Wanja Kinuthia and Nigel R. Franks Schöning, C., Kinuthia, W. and Franks, N. R. 2005. Evolution of allometries in the worker caste of Dorylus army ants. / Oikos 110: 231 /240. The worker caste of polymorphic ant species consists of individuals of strikingly different morphologies. Most studies of polymorphic species have focused on intracolonial allometries and their association with division of labour and task performance of workers of different physical subcastes. However, the factors driving the evolution of these allometries are poorly understood. Here, we analyse the importance of life-style (hypogaeic vs intermediate vs epigaeic) as a factor in the evolution of allometries of functionally important body traits in Dorylus army ants in the context of an evolutionary scenario according to which a hypogaeic life-style is the ancestral state in this group. To this end, we conducted a detailed comparative analysis of the allometries of nine characters for ten species belonging to all six currently recognized subgenera and showing different life-styles. Eight of the nine traits under consideration show a clear increase in relative size from hypogaeic to intermediate to epigaeic life-style. These results strongly suggest that shifts in ecological niche are linked with adaptations in these traits. The degree of overall differentiation among species is more pronounced in larger than in smaller workers. The pattern of division of labour in the epigaeic D. molestus indicates that two factors that may have caused this phenomenon are new food habits and an increased need for colony defence. C. Schöning, Inst. für Biologie, Neurobiologie, Freie Univ. Berlin, Königin-Luise-Strasse 28/30, DE-14195 Berlin, Germany ([email protected]). / Present address: Zoological Inst., Univ. of Copenhagen, Universitetsparken 15, DK-2100 Copenhagen, Denmark. / W. Kinuthia, Dept of Invertebrate Zoology, National Museums of Kenya, P.O. Box 40658, Nairobi, Kenya. / N. R. Franks, School of Biological Sciences, Univ. of Bristol, Woodland Road, Bristol, UK, BS8 1UG. Polymorphism in the worker caste of ants has been defined as non-isometric increase in size occurring over a sufficient range of adult size variation to produce individuals of distinctly different proportions (Wilson 1953, Oster and Wilson 1978). Workers of different physical subcastes show marked but flexible task specialization (Wilson 1980, 1984, Moffett 1987, Franks et al. 2001, Hughes et al. 2003), and this is thought to enhance the efficiency of the worker caste as a whole, because workers allocate tasks to themselves for which they are morphologically specially adapted (Oster and Wilson 1978, Bourke and Franks 1995). Physical subcastes are defined as groups of workers that underwent the same developmental programme (Wheeler 1991). The development of ant larvae is controlled by nutrition and inhibitory pheromones (Wheeler 1991), but in one species studied so far, different physical subcastes were partly associated with different patrilines within a colony (Hughes et al. 2003). The effects of environmental cues are mediated by hormones, which act at specific thresholds and cause a switch between alternative developmental pathways with associated patterns of gene expression (Evans and Wheeler 2001, Abouheif and Wray 2002). In this way, the critical size at which the larva begins metamorphosis and/or the growth parameters of tissues can be reprogrammed. Imaginal structures grow mostly only during metamorphosis once food intake has stopped, so that the Accepted 26 December 2004 Copyright # OIKOS 2005 ISSN 0030-1299 OIKOS 110:2 (2005) 231 allometries among different body parts reflect the outcome of competition between growing tissues for nutrients (Wheeler 1991, Nijhout and Wheeler 1996, Nijhout and Emlen 1998). Physical subcastes of polymorphic species are usually identified based on allometries and/or the size-frequency distribution of workers (Hollingsworth 1960, Raignier et al. 1974, Franks 1985, Cagniant 1991, Tschinkel 1998), but this approach has been criticised (Feener et al. 1988). As the size-frequency distribution of the standing worker population represents the sum of many separate (or continuous, if the queen(s) lay(s) eggs continuously) brood sets modified by size-dependent mortality rates (Calabi and Porter 1989, Wheeeler 1991), the use of size-frequency distributions as an indicators of developmental mechanisms is problematical. Moreover multimodality is hard to demonstrate statistically. It has been shown that complex allometries can be the product of reprogramming of growth parameters (Nijhout and Wheeler 1996), but it is nevertheless hard to detect physical subcastes from smoothly curvilinear graphs. Finally, as switches between alternative developmental pathways can occur in different tissues at different times or not at all in others, interpretations concerning subcastes will depend on the choice of morphometric characters. The allometries among body parts within ant colonies represent a form of static allometry and can often be well described by Huxley’s growth function y /b/xa (Huxley 1932). Artificial selection experiments and a great number of comparative studies have shown that allometries in holometabolous insects can evolve quite rapidly (Emlen and Nijhout 2000). As selection in social insects not only operates at the individual level but also at the colony level, the allometries found in ants and bumblebees differ from the allometries of solitary holometabolous insects in that they are an adaptive colony trait. Previous studies have shown that these vary between populations (Owen and Harder 1995, McArthur et al. 1996), between different families within colonies of a facultatively polygnous formicine species (Fraser et al. 2000), between young pleometrotic and haplometrotic colonies in facultatively polygnynous myrmicine species (Kenne et al. 2000), with colony size (Kenne et al. 2000, Nowbahari et al. 2000) and perhaps with season (Fig. 3 in Rissing 1987). While numerous studies have examined allometries within species, comparative analyses that may reveal which factors drive the evolution of these allometries and allow insights into the functional morphology of workers with different body proportions are generally sparse (but see Wehner 1983, 1987, Feener et al. 1988, Zollikofer 1994). The results of recent studies inspired by the size-grain hypothesis (Kaspari and Weiser 1999) indicate that the characteristics of different spatial niches in a habitat act 232 as selection pressures for the morphology of the ant species using these different niches (Espadaler and Gómez 2001, Parr et al. 2003, Farji-Brener et al. 2004). Farji-Brener et al. 2004 also suggested that colonies of polymorphic ant species could enhance their foraging success by sending out workers of different sizes to explore and forage in environments with different levels of rugosity. However, at least in some species such as polymorphic army ants, which by definition employ obligate collective foraging, workers of different sizes work together in the same microniche. If different spatial niches in a habitat do indeed favour specific ant worker morphologies, one would expect workers of all sizes in these species to show the same respective niche specific adaptations. The army ant genus Dorylus represents a well-defined monophyletic group (Bolton 1990, Brady 2003) with six currently recognized subgenera (Bolton 1995). The worker caste is highly polymorphic in all species studied so far (Emery 1901, Hollingsworth 1960, Raignier et al. 1974, Berghoff et al. 2002b). Species of the subgenera Dichtadia , Rhogmus, Typhlopone, Alaopone and Dorylus s.str. hunt and nest in the soil. Only under exceptional circumstances do workers of these species appear on the surface: when foraging workers have to cross hard-packed soil (C. Schöning pers. obs.) or encounter unusually rich food sources (Berghoff et al. 2002a), during interspecific battles (Leroux 1979, Gotwald 1995, p. 159) and when the males emerge from the nest (Wheeler 1922). The use of ecological niches by species of the subgenus Anomma is more complex. While the typical ‘‘driver ants’’ (e.g. D. wilverthi , D. molestus ) conspicuously forage on the surface and climb vegetation in search of prey, several species hunt in the leaf litter stratum (e.g. D. kohli , Wasmann 1904, Raignier and van Boven 1955; D. gerstaeckeri , Gotwald 1974). Gotwald (1978) proposed that the extant Dorylus species derive from an ancestor with a completely hypogaeic life-style and that a few species adopted an epigaeic foraging style secondarily. He based this inference on the observation that workers are eyeless and have a reduced number of maxillary and labial segments in all species regardless of life-style. The aim of the present study is to examine the importance of life-style (foraging stratum use and nesting habits) as a selection factor in the evolution of allometries in the Dorylus worker caste. We have chosen to analyse the allometries of such characters whose functional value can be demonstrated or at least be inferred with reasonable confidence. They should all be subject to diverse selection pressures in the different ecological niches and therefore likely reflect possible adaptations associated with ecological niche shifts. OIKOS 110:2 (2005) Division of labour In order to understand why morphological differentiation is more pronounced in larger workers (results) we examined additionally the pattern of division of labour in the epigaeicly hunting Dorylus molestus. Workers engaged in 14 different tasks (sample sizes 113 /1667 workers per task) were collected, using forceps or aspirator, from the same colony mentioned above over a period of 2 weeks to test for size-related division of labour and to check which tasks large workers specialize on. The tasks are listed and defined in the legend of Fig. 6. Material and methods Samples Specimens used for this study were collected during field work in Kenya and Uganda between March 2001 and September 2002 and stored in 70% ethanol. Additionally, samples from south east Asian representatives of the genus were provided by Stefanie Berghoff. All six currently recognized subgenera were thus covered. In four species, samples were collected from the migration trail or directly from the nest and therefore likely represent the entire size range of workers in that colony. However, for the remaining six species nests or migrations could not be found so that these samples include only foragers and thus most probably consist only of subsets of the worker caste size range. In all 10 species, workers were taken from a single colony. Intraspecific variability is low compared to interspecific variability (C. Schöning unpubl.). For morphometric analysis about 100 uninjured workers each were chosen from the samples so as to represent the entire size range evenly. Callow workers were excluded. Details on origin and type of sample are summarized in Table 1. Representative specimens will be deposited in the Department of Invertebrate Zoology, National Museum of Kenya, Nairobi and the Natural History Museum, Berlin. Based on own observations or information given in the literature, species were categorized as exhibiting one of the following three life-styles: epigaeic (E) / hunting on the ground surface and in the vegetation; conspicuous swarm raids attracting ant birds; large, conspicuous earth nests. intermediate (I) / hunting in the leaf litter but not in the vegetation; swarm raids can be so conspicuous as to attract ant birds; nests have never been found (presumably as in hypogaeic species, below). hypogaeic (H) / usually hunting in the soil; normally not seen in the leaf litter or on the ground except when passing through areas with for hard packed soil; nests almost undetectable (Weissflog et al. 2000, Berghoff et al. 2002a, b, 2003 for information on D. laevigatus and D. cf. vishnui ). Morphometrics All measurements were taken using a MS 5 Leica stereomicroscope fitted with an ocular micrometer. The maximum possible magnification was used whilst keeping a structure within the range of the ocular micrometer (6.3 /64). We used methods recommended by Seifert (2002) to minimise measurement errors. Measurement error (ME) was assessed by repeated measurements and was quantified as % ME following Bailey and Byrnes (1990) for a set of 10 D. molestus individuals encompassing the entire size range. These were scored for all characters five times. A one-way analysis of variance was then performed to partition variance into within- and between-individual components. The definitions, abbreviations, percent measurement errors, and presumed functional significance of the characters under consideration are as follows: Maximum head width (HWmax, 0.27%) / maximum measurable head width. Wider heads can accommodate larger mandibular muscles and may therefore allow larger biting force (Paul 2001). Anterior head width (HWa, 0.62%) / distance between points where mandibles touch head capsule when fully expanded with forceps. A wider anterior head width allows the mandibles to be opened wider when biting predators or prey. Length of the funiculus of the antenna (FL, 1.74%) / length of the left funiculus, i.e. length of the distal jointed, sensory part, of the antenna. Funiculus length Table 1. Dorylus samples used for morphometric analyses. Abbreviations: H / hypogaeic, I / intermediate, E / epigaeic, N / workers from nest or migration trail, F / foraging workers. Subgenus Dichtadia Rhogmus Typhlopone Alaopone Dorylus Anomma OIKOS 110:2 (2005) Species Life-style Sample type Sample origin D. laevigatus D. sp. D. obscurior D. cf. vishnui D. affinis D. sp. D. kohli D. opacus D. wilverthi D. molestus H H H H H I I I E E N F F F N F F F N N Sitiawan, Malaysia (482?N 10085?E, 0 m asl.) Kibale NP, Uganda (08 34?N 308 22?E, 1500 m asl) Wamba, Kenya (O8N, 37 E, 1350 m asl) Kinabalu NP, Sabah, Malaysia (685?N 11683?E) Chogoria, Kenya (0813?49??S, 37837?54?? E, 1575 m asl) Kibale NP, Uganda (0834?N 30822?E, 1500 m asl) Mgahinga NP, Uganda (1828?S, 29840?E, 2200 m asl) Kibale NP, Uganda (08 34?N 30822?E, 1500 m asl) Kibale NP, Uganda (0834? N 30822?E, 1500 m asl) Chogoria, Kenya (0814?16??S, 37834?31??E, 1842 m asl) 233 should be important in trail following behaviour, especially because Dorylus workers are eyeless. Longer funiculi (and scapes) may enable individuals both to follow the pheromone trails at higher speeds and avoid collisions (Couzin and Franks 2003). Length of the scape of the antenna (SL, 0.02%) / Maximum straight line scape length excluding the basal condyle. Functional significance see FL. Mandible length (ML, 4.36%) / distance between apex of the left mandible to the point where it touches the head capsule when fully opened with forceps. Longer mandibles will allow more painful biting of predators and more efficient cutting of larger prey animals. Mandible gape width (MGW, 5.15%) / distance between mandible apices when both are fully opened using forceps. A wider mandible gape allows biting of larger objects (prey or predator). Front leg length (FLL, 1.5%) / maximum length of left hind leg from trochanter to tarsal tips in dorsal view with the leg fully extended. As prey items are carried slung beneath the body, front leg length should limit the maximum possible size of prey items. Furthermore longer legs may allow more efficient and faster locomotion. Hind leg length (HLL, 0.87%) / maximum length of left hind leg from trochanter to tarsal tips in dorsal view with the leg fully extended. Longer legs will allow more efficient and faster locomotion and more space for other assisting workers during cooperative prey item transport (Franks et al. 1999), but on the other hand increase the cross-sectional area needed by the individual and therefore impede entering crevices (Kaspari and Weiser 1999). The last point should be of particular significance for species hunting in the soil. Pronotum width (PW, 0.58%) / maximum pronotal width in dorsal view. This character was used as another size indicator (below). For D. (Rhogmus ) sp. only a reduced data set (PW, HWmax, SL, FL, HLL, DM) was recorded. All measurements are in millimetres. Measurement error is generally very low, but higher for MGW and ML which reflects the fact that these measurements are difficult to make. But the differences in MGW and ML between species are sufficiently large as to allow meaningful conclusions despite lower accuracy in these measurements. While any two linear measurements can be chosen to test for allometry and thus whether a given species complies with Wilson’s definition (above) of a polymorphic species or not, in a comparative approach one must have a general estimate of overall body size. DinizFilho et al. (1994) used principal component analysis in their study of polymorphism in Camponotus rufipes to identify which among ten morphological characters scaled most closely to isometry and could therefore be used as an indicator of overall size. However, when 234 allometric relationships are curvilinear rather than linear, principal component analysis cannot be employed in a straightforward manner (Diniz-Filho et al. 1994). We therefore used dry mass (DM, 0.02%) as an indicator of overall size, as this should also represent a good estimate of biomass investment and therefore allow conclusions as to how colonies invest biomass in workers of different proportions. Specimens were dried at 708C for 48 h and then weighed to the nearest 0.01 mg using a R 200 D balance (Sartorius GmbH, Göttingen, Germany). For the D. molestus workers engaged in various tasks PW only was measured. Statistical analysis As the aim of the present study was a comparison of the relative sizes of certain body traits between workers of different species regardless of the specific developmental programme that generates these workers, we did not attempted to identify physical subcastes. Instead, for each trait we first established the best fit model for raw linear data from all species combined over their common size range (0.20 mg /3.67 mg) as a function of DM1/3 by stepwise multiple regression. Higher order polynomials were included if all coefficients were significantly different from zero (p B/0.05, Zar 1996, Fraser et al. 2000). As the goal of conducting this multiple regression analysis was to develop a model that best predicts variability in the response, and there was no interest in studying the particular relationships between the response and explanatory variables of different orders, the problems due to multicollinearity can be ignored (Graham 2003). The relative residuals (absolute residual divided by predicted value) from this common regression model were then compared using the Kruskal /Wallis Test with life style as grouping factor and the Mann /Whitney U-test for pairwise comparisons between life-styles, as variances were heteroscedastic (Sokal and Rohlf 1995). Additionally, an identical analysis was run over the entire size range (0.05 mg /11.08 mg) displayed in our samples. Plots of raw data had initially indicated that the differences between species increased with DM both absolutely and relatively. To explore whether this is a real effect we used two tests. For each trait we identified the two species with the lowest and highest relative residual values over the common size range as established by the Mann /Whitney U-Test. We then calculated the ratios between the values predicted by the respective best-fit model over their common size range by determining the function f(x) /best-fit model of species with highest relative residual values/best-fit model of species with lowest relative residual values. This ratio indicates the maximum differentiation between the examined species OIKOS 110:2 (2005) for the trait under consideration as a function of DM1/3. In addition, for each trait the coefficient of variation (CV) was calculated for data from all species combined in 15 equally sized intervals over the common size range. It was then tested whether the CV increases with DM1/3. A higher CV implies stronger differentiation. In general, we did not use log-transformed data because (1) log-transformation data often did not show a normal distribution, (2) log-transformation yielded curvilinear relationships so that statistical analyses based on log-transformed data are as difficult to analyse as those based on raw (and cube-root transformed) data, and (3) a/b "/(log a)/(log b) so that the calculation of ratios would be confounded. To test for size differences between D. molestus workers engaged in different tasks again the Kruskal / Wallis Test with task as grouping factor and additionally the Mann /Whitney U-test for pair-wise comparisons were used (Sokal and Rohlf 1995). Significance levels were adjusted for multiple comparisons using the Dunn-Šidák method (Sokal and Rohlf 1995). All statistical tests were performed using STATISTICA ’99 Edition (StatSoft, Inc.). Results In all traits under consideration, life-style had a significant effect on the values of the relative residuals values (Kruskal /Wallis test, pB/0.001 in all cases). Except for pronotum width, workers of epigaeic species always had higher values than workers of intermediate species and these in turn showed higher values than workers of hypogaeic species (Fig. 1). Pronotum width values of hypogaeic workers were larger than in intermediate workers (Mann /Whitney U-test, Z/3.92, pB/ 0.0001) but do not differ between the latter and epigaeic workers (Mann /Whitney U-test, Z/ /0.16, p/0.87). These results also hold when the entire size range, instead of the common size range, is considered. As an illustration of the general pattern of increase in the traits, Fig. 2 shows bivariate plots of HLL and HWmax vs DM1/3. In all traits, the ratios between the species with the highest residual values and the one with the lowest values increased with DM1/3, but never in a simple linear way (Fig. 3). CV increased significantly with DM1/3 for all traits except SL (r2 /0.219 and p/0.078 for SL, r2 between 0.367 and 0.633, pB/0.02 for all others). D. molestus workers engaged in different tasks differed significantly in size (Kruskal /Wallis test, pB/0.0001, Fig. 4). The groups with the highest median values are workers transporting food item singly, leading workers in food transporting groups, workers in defensive posture along foraging trails and workers defending OIKOS 110:2 (2005) the nest. These are all significantly larger than all other groups. Discussion The evolution of allometric scaling relationships in the worker caste of ants has so far received little attention. Here we show that in Dorylus army ants differences in life-style are directly correlated with differences in the allometric scaling relationships with an unambiguous increase from hypogaeic to intermediate to epigaeic species in maximum head width, anterior head width, flagellum length, scape length, mandible length, mandible gape width, front leg length, and hind leg length. These results provide further support for the conclusions of other studies (Kaspari and Weiser 1999, Farji-Brener et al. 2004) that different spatial niches in a habitat select for specific worker ant morphologies. As traits of identifiable functional value were selected for our morphometric analysis, we can try to pinpoint the selection pressures that may have favoured the morphological changes. For workers hunting in the leaf-litter and on the surface longer legs will convey advantages with respect to locomotion and food transport (i.e. they should be able to cover long foraging distances faster and/or more efficiently and carry more food mass per unit body mass, Franks 1985, 1986), and at the same time they will be much less constrained by the possible costs associated with longer (hind) legs than workers hunting in the soil. The leaf-litter stratum may also hold many interstices yet these will be on average larger than interstices in the soil simply because soil is more densely packed. In the larger workers (dry mass /1 mg) the ratio between front and hind leg length increases from hypogaeic to intermediate to epigaeic species (Fig. 5). So the increase in hind leg length from hypogaeic to epigaeic is stronger than the increase in front leg length. This allows the conclusion that food and/or brood transport is indeed also of great importance for hypogaeic species but that longer hind legs are simply prohibitive for them when hunting in the soil. Longer legs and faster running speeds may in turn necessitate longer antennae to enable workers to follow the trail pheromone and/or avoid collisions with other workers running on the trail (Couzin and Franks 2003). Intermediate and epigaeic life-styles require more efficient defence as colonies will be more apparent to predators, more prone to disturbances during their raids and finally more vulnerable to thieves, in particular opportunistic birds which do not only catch animals escaping from the advancing swarm raid but also steal those that have been pinned down by the ants (C. Schöning pers. obs.). Longer mandibles, larger anterior head widths and as a result larger mandible gape widths should enable these species to inflict more 235 236 OIKOS 110:2 (2005) Fig. 1. Column plots of relative residuals for all traits from common regression models over common size range for workers with different life-styles (H / hypogaeic, I / intermediate, E / epigaeic). The whiskers encompass the lower 25% and upper 75% percentile range; the square indicates the median. The results of the Kruskal /Wallis-tests are given below. Fig. 2. Hind leg length (HLL) and maximum head width (HWmax) vs dry mass1/3 (DM1/3) for D. wilverthi (/), D. molestus (), D. opacus ("), D. kohli ('), D. sp. (j), D. (Rhogmus) sp. (2), D. affinis (^), D. vishnui (k), D. laevigatu s (m), D. obscurio r (I). painful bites on larger animals and thus pose a formidable deterrence. On the other hand, the features associated with locomotion and defence will also allow colonies to pin down and fragment larger prey animals. In summary, the inferred advantages associated with larger maximum head width, anterior head width, Fig. 3. Ratios between the predicted values of the species with the highest relative residuals and the predicted values of the species with the lowest relative residuals over their common size range. HLL (k), FLL ("), MGW (2), ML (j), FL (I), SL (^), HWa ('), HWmax ("), PW (/). OIKOS 110:2 (2005) funiculus length, scape length, mandible length, mandible gape width, front and hind leg lengths are certainly complementary, but several of them can and should be tested. Large D. molestus workers (pronotum width / 0.9 mm, this corresponds to a dry mass of about 1.6 mg) are disproportionately often engaged in dismembering food, food transport, and colony defence (Braendle et al. 2003), and a similar pattern can be found in D. wilverthi (Franks et al. 2001). As morphological differentiation among species is also more pronounced in workers of this size-class, this indicates that the requirement for efficiency in these tasks might have selected for divergent ecological niche-specific morphologies. According to the evolutionary scenario proposed by Gotwald (1978) the differences in life-style between extant Dorylus species are the result of a unidirectional shift from hypogaeic to epigaeic niches, but this scenario did not take into account more subtle differences in ecological niche use by Anomma species. Convergent evolution of similar characters is the strongest type of evidence for adaptation (Harvey and Pagel 1991). A species-level phylogeny of the genus Dorylus is lacking, and even the monophyly of Dorylus s.str. and Anomma are uncertain (Barr et al. 1985). Therefore, a more comprehensive analysis of morphological character evolution will only be possible once the phylogeny of the genus has been established. Species in the army ant subfamilies Ecitoninae and Aenictinae also show prominent differences in their ecological niches (Rosciewski and Maschwitz 1994, Gotwald 1995) and therefore offer additional opportunities to examine the link between scaling relationships of the morphological characters we examined and life-style. The mechanisms allowing coexistence of different army ant species in a given habitat are poorly understood. Resource partitioning according to food item size has been shown to be an important mechanism in harvester ant assemblages (Davidson 1977) but is unlikely to be a relevant phenomenon here because hypogaeic, intermediate and epigaeic Dorylus spp. all display extensive polymorphism. Rather, we suggest that species are using different strata of the same habitat for foraging and thereby reduce competition with each other. This, however, does not explain how different species displaying the same life-style can be sympatric. Detailed analyses of the food spectra of sympatric species hunting in the same stratum may reveal whether niches are partitioned according to prey type as has been suggested for neotropical army ants (Rettenmeyer et al. 1983). Foraging stratum use is not completely static. During their swarm raids in the leaf litter, workers of D. opacus and D. (Anomma ) sp. for example sometimes also come up on the leaf litter surface (but do not climb vegetation). On two occasions we observed antagonistic interactions between the epigaeic D. wilverthi and these 237 Fig. 4. Division of labour in D. molestus. Size (measured as pronotum width) of workers engaged in the following tasks (n given in brackets along with a more precise definition where necessary): (A) dropping soil particles on a slimy prey animal (144); (B) running on foraging trail (687); (C) removing soil particles from foraging trench during wet weather conditions (249); (D) participating in swarm raid (777, workers were collected from the front 20cm of the swarm); (E) running on migration trail towards new nest (1667); (F) defending foraging trail after disturbance (308, workers biting into the observer’s finger after blowing onto the trail with the finger placed 5 cm away from trail); (G) carrying soil particles out of nest during dry weather conditions (299)); (H) transporting food items in groups (294, all workers of the group combined); (I) carrying soil particles out of nest after heavy rain during previous night (273); (J) transporting food item alone (136); (K) transporting food item in group (113, front runners only); (L) standing alongside foraging trail in defensive posture (357); (M) cutting up earthworms (104); (N) defending nest after disturbance (142, workers biting into the observer’s hand after beating onto nest).The box encompasses the lower 25% and upper 75% percentile range; the internal line is the median; whiskers indicate the minimum and maximum values. The groups engaged in the various tasks are arranged according to increasing median values. two sympatric intermediate species. In both cases, workers of the intermediate species immediately withdrew to their usual hunting stratum into the leaf litter and into tunnels in the ground when the D. wilverthi swarm raids were approaching. This behaviour is different from the one displayed by colonies of the epigaeic species D. molestus during intraspecific encounters Fig. 5. Front leg length/hind leg length ratios vs dry mass for workers of hypogaeic, intermediate and epigaeic species. 238 which simply continue the swarm raid on the surface in another direction (C. Schöning pers. obs.). Thus species appear to be specialized for a particular foraging stratum and avoid competition with other specialized species when they leave it. In conclusion, our results show that morphological diversification in Dorylus army ants was most likely influenced by niche shifts to new foraging strata. Longer legs, antennae, and mandibles as well as wider heads are probably adaptations for efficient foraging and colony defence in intermediate and epigaeic microniches, respectively. The use of different foraging strata allows the complementary niche exploitation and thus coexistence of several army ant species in the same habitat. Acknowledgements / We would like to thank Randolf Menzel for making this study possible, Washington Njagi for steadfast help during field work, William Gotwald, Jr. for kindly identifying several ant specimens, Stefanie Berghoff for providing specimens, Daniel Kronauer for helpful comments on the manuscript and Scott Powell, Xavier Espadaler and Bernhard Seifert for fruitful discussions. We are grateful to the government institutions of Kenya (Kenya Wildlife Service, Ministry for Science, Education and Technology) and Uganda (Uganda Wildlife Authority, National Council for Science and Technology) for granting research permission. Financial support was provided through doctoral fellowships by the Federal State Berlin and the DAAD to C.S. References Abouheif, E. and Wray, G. A. 2002. Evolution of the gene network underlying wing polyphenism in ants. / Science 29: 249 /252. Bailey, R. C. and Byrnes, J. 1990. A new, old method for assessing measurement error in both univariate and multivariate morphometric studies. / Syst. Zool. 39: 124 /130. Barr, D., van Boven, J. and Gotwald, W. H. Jr. 1985. Phenetic studies of African army ant queens of the genus Dorylus. / Syst. Entomol. 10: 1 /10. Berghoff, S. M., Weissflog, A., Linsenmair, K. E. et al. 2002a. Foraging of a hypogaeic army ant: a long neglected majority. / Ins. Soc. 49: 133 /141. Berghoff, S. M., Weissflog, A., Linsenmair, K. E. et al. 2002b. Nesting habits and colony composition of the hypogaeic army ant Dorylus (Dichthadia ) laevigatus Fr. Smith. / Ins. Soc. 49: 380 /387. Berghoff, S. M., Gadau, J., Winter, T. et al. 2003. Sociobiology of hypogaeic army ants: characterization of two sympatric Dorylus species on Borneo and their colony conflicts. / Ins. Soc. 50: 139 /147. Bolton, B. 1990. Army ants reassessed: the phylogeny and classification of the doryline section. / J. Nat. Hist. 24: 1339 /1364. Bolton, B. 1995. A new general catalogue of the ants of the world. / Harvard Univ. Press. Brady, S. 2003. Evolution of the army ant syndrome: the origin and long-term evolutionary stasis of a complex of behavioral and reproductive adaptations. / Proc. Natl Acad. Sci 100: 6575 /6579. Braendle, C., Hockley, N., Brevig, T. et al. 2003. Size-correlated division of labour and spatial distribution of workers in the driver ant, Dorylus molestus. / Naturwissenschaften 90: 277 /281. Bourke, A. F. G and Franks, N. R. 1995. Social evolution in ants. Monographs in behavioral ecology. / Princeton Univ. Press. OIKOS 110:2 (2005) Cagniant, H. 1991. Etude de populations marocaines de la superespece Camponotus cruentatus Latreille. Mise en evidence de Camponotus obscuriventris sp. n. / Revista Española Entomologı́a 67: 35 /54. Calabi, P. and Porter, D. 1989. Worker longevity in the fire ant Solenopsis invicta . / J. Ins. Physiol. 35: 643 /649. Couzin, I. D. and Franks, N. R. 2003. Self-organized lane formation and optimised traffic flow in army ants. / Proc. R. Soc. Lond. B 270: 139 /146. Davidson, D. W. 1977. Species diversity and community organization in desert seed-eating ants. / Ecology 58: 711 /724. Diniz-Filho, J. A. F., von Zuben, C. J., Fowler, H. G. et al. 1994. Multivariate morphometrics and allometry in a polymorphic ant. / Ins. Soc. 41: 153 /163. Emery, C. 1901. Studi sul polimorfismo e la metamorfosi nel genere Dorylus. / Memorie della Reale Accademia delle Scienze dell’ Istituto di Bologna 9: 183 /201. Emlen, D. J. and Nijhout, H. F. 2000. The development and evolution of exaggerated morphologies in insects. / Annu. Rev. Entomol. 45: 661 /708. Espadaler, X. and Gómez, C. 2001. Formicine ants comply with the size-grain hypothesis. / Funct. Ecol. 15: 136 /139. Evans, J. D. and Wheeler, D. E. 2001. Gene expression and the evolution of insect polyphenisms. / BioEssays 23: 62 /68. Farji-Brener, A. G., Barrantes, G. and Ruggiero, A. 2004. Environmental rugosity, body size and access to food: a test of the size grain hypothesis in tropical litter ants. / Oikos 93: 165 /171. Feener, D. H. Jr., Lighton, J. R. B. and Bartholomew, G. A. 1988. Curvilinear allometry, energetics and foraging ecology: a comparison of leaf-cutting ants and army ants. / Funct. Ecol. 2: 509 /520. Franks, N. R. 1985. Reproduction, foraging efficiency and worker polymorphism in army ants. / In: Lindauer, M. and Hölldobler, B. (eds), Experimental behavioural ecology. Fischer Verlag, pp. 91 /107. Franks, N. R. 1986. Teams in social insects: group retrieval of prey by army ants. / Behav. Ecol. Sociobiol. 18: 425 /429. Franks, N. R., Sendova-Franks, A. B., Simmons, J. et al. 1999. Convergent evolution, super-efficient teams and tempo in Old and New World army ants. / Proc. R. Soc. Lond. B 266: 1697 /1701. Franks, N. R., Sendova-Franks, A. B. and Anderson, C. 2001. Division of labour within teams of New World and Old World army ants. / Anim. Behav. 62: 635 /642. Fraser, V. S., Kaufmann, B., Oldroyd, B. P. et al. 2000. Genetic influence on caste in the ant Camponotus consobrinus. / Behav. Ecol. Sociobiol. 47: 188 /194. Gotwald, W. H. Jr. 1974. Foraging behavior of Anomma driver ants in Ghana cocoa farms. / Bulletin de I’Institut Fondamental d’Afrique Noire (Ser. A) 36: 705 /713. Gotwald, W. H. Jr. 1978. Trophic ecology and adaptation in tropical Old World ants of the subfamily Dorylinae. / Biotropica 10: 161 /169. Gotwald, W. H. Jr. 1995. Army ants / the biology of social predation. / Cornell Univ. Press. Graham, M. H. 2003. Confronting multicollinearity in ecological regression. / Ecology 84: 2809 /2815. Harvey, P. H. and Pagel, M. D. 1991. The comparative method in evolutionary biology. / Oxford Univ. Press. Hughes, W. H. O., Sumner, S., Van Borm, S. et al. 2003. Worker caste polymorphism has a genetic basis in Acromyrmex leafcutting ants. / Proc. Natl Acad. Sci. 100: 9394 /9397. Huxley, J. S. 1932. Problems of relative growth. / John Hopkins Univ. Press. Hollingsworth, M. J. 1960. Studies on the polymorphic workers of the army ant Dorylus (Anomma ) nigricans Illiger. / Ins. Soc. 7: 17 /37. Kaspari, M. and Weiser, M. D. 1999. The size-grain hypothesis and interspecific scaling. / Funct. Ecol. 13: 530 /538. OIKOS 110:2 (2005) Kenne, M., Dejean, A., Fénéron, R. et al. 2000. Changes in worker polymorphism in Myrmicaria opaciventris Emery. / Ins. Soc 47: 50 /55. Leroux, J. M. 1979. Sur quelques modalités de disparition des colonies d’Anomma nigricans Illiger dans la région de Lamto (Côte d’Ivoire). / Ins. Soc. 26: 93 /100. McArthur, A. J., Adams, M. and Shattuck, S. O. 1996. A morphological and molecular review of Camponotus terebrans (Lowne). / Austr. J. Zool. 45: 579 /598. Moffet, M. W. 1987. Division of labor and diet in the extremely polymorphic ant Pheidologeton diversus. / Natl Geogr. Res. 3: 282 /304. Nijhout, H. F. and Wheeler, D. E. 1996. Growth models of complex allometries in holometabolous insects. / Am. Nat. 148: 40 /56. Nijhout, N. F. and Emlen, D. J. 1998. Competition among body parts in the development and evolution of insect morphology. / Proc. Natl Acad. Sci. 95: 3685 /3689. Nowbahari, B., Feneron, R. and Malherbe, M. C. 2000. Polymorphism and polyethism in the formicine ant Cataglyphis niger. / Sociobiol. 36: 485 /496. Oster, G. and Wilson, E. O. 1978. Caste and ecology in the social insects. / Princeton Univ. Press. Owen, R. E. and Harder, L. D. 1995. Heritable allometric variation in bumble bees: opportunities for colony-level selection of foraging ability. / J. Evol. Biol. 8: 725 /738. Parr, Z. J. E., Parr, C. L. and Chown, S. L. 2003. The size-grain hypothesis: a phylogenetic and field test. / Ecol. Entomol. 28: 475 /481. Paul, J. 2001. Mandible movements in ants. / Comp. Biochem. Physiol. (A) 131: 7 /20. Raignier, A. and van Boven, J. 1955. Etude taxonomique, biologique et biométrique des Dorylus du sous-genre Anomma . / Ann. Mus. R. Congo Belge, n.s. 48 (Sci. Zool.) 2: 1 /359. Raignier, A., van Boven, J. and Ceusters, R. 1974. Der Polymorphismus der afrikanischen Wanderameisen unter biometrischen und biologischen Gesichtspunkten. / In: Schmidt, G.H. (ed.), Sozialpolymorphismus bei Insekten. Fischer Verlag, pp. 668 /693. Rettenmeyer, C. W., Chadab-Crepet, R., Naumann, M. G. et al. 1983. Comparative foraging by neotropical army ants. / In: Jaisson, P. (ed.), Social insects in the tropics. Université Paris-Nord, pp. 59 /73. Rissing, S. W. 1987. Annual cycles in worker size of the seedharvester ant Veromessor pergandei . / Behav. Ecol. Sociobiol. 20: 117 /124. Rościszewski, K. and Maschwitz, U. 1994. Prey specialisation of army ants of the genus Aenictus in Malaysia. / Andrias 13: 179 /187. Seifert, B. 2002. How to distinguish most similar insect species / improving the stereomicroscopic and mathematical evaluation of external characters by example of ants. / J. Appl. Entomol. 126: 445 /454. Sokal, R. R. and Rohlf, F. J. 1995. Biometry. 3rd ed. / W.H. Freeman. Tschinkel, W. R. 1998. Sociometry and sociogenesis of colonies of the harvester ant, Pogonomyrmex badius : I. Worker characteristics in relation to colony size and season. / Ins. Soc. 45: 385 /410. Wasmann, E. 1904. Zur Kenntniss der Gaeste der Treiberameisen u. ihrer Wirthe am obern Congo nach den Sammlungen und Beobachtungen von P. Herrn Kohl, C.S.S.C. bearbeitet. / Zool. Jahrb. Suppl. 7: 611 /682. Weissflog, A., Sternheim, E., Dorow, W. H. O. et al. 2000. How to study subterranean army ants: a novel method for locating and monitoring field populations in the South East Asian army ant Dorylus (Dichthadia ) laevigatus Smith, 1857 with observations on their biology. / Ins. Soc. 47: 317 / 324. Wehner, R. 1983. Taxonomie. / Funktionsmorphologie und Zoogeographie der saharischen Wüstenameise Cataglyphis 239 fortis (Forel 1902) stat. nov. Senckenbergiana biologica 64: 89 /132. Wehner, R. 1987. Spatial organization of foraging behaviour in individually searching desert ants, Cataglyphis (Sahara) and Ocymyrmex (Namib Desert). / In: Pasteels, J. M. and Deneubourg, J. L. (eds), From individual to collective behaviour in social insects. Experimentia Supp. 54, pp. 253 /270. Wheeler, D. E. 1991. The developmental basis of worker caste polymorphism in ants. / Am. Nat. 138: 1218 /1238. Wheeler, W. M. 1922. Ants of the American Museum Congo expedition. A contribution to the myrmecology of Africa. II. The ants collected by the American Museum Congo Expedition. / Bull. Am. Mus. Nat. Hist. 45: 39 / 269. Wilson, E. O. 1953. The origin and evolution of polymorphism in ants. / Q. Rev. Biol. 28: 136 /156. Wilson, E. O. 1980. Caste and divison of labor in leaf-cutter ants I. The overall pattern in Atta sexdens. / Behav. Ecol. Sociobiol. 7: 143 /156. Wilson, E. O. 1984. The relation between caste ratios and division of labour in the ant genus Pheidole. / Behav. Ecol. Sociobiol. 16: 89 /98. Zar, J. H. 1996. Biostatistical analysis. 3rd ed. / Prentice-Hall. Zollikofer, C. P. E. 1994. Stepping patterns in ants. II. Influence of body morphology. / J. Exp. Biol. 192: 107 /111. Subject Editor : Jan Lindström 240 OIKOS 110:2 (2005)