BiologicalJournal ofthe Linnean Society (1995),55: 29-44. With 4 figures Convergence and historical effects in harvester ant assemblages of Australia, North America, and South America RODRIGO G. MEDEL Departamento de Ciencim Ecohgicas, Facultad de Ciencias, Universidad de Chile, Cmilla 653, Santiago, Chile Received 1 1 May 1994; acceptedforpublication 7 September 1994 In this paper I examine the extent to which contemporary ecological patterns in 42 harvester ant assemblages of three continents can be explained as a result of present-day environments or from differences in the history of each ant biota. The contribution of each factor to the overall variability of six community characters was evaluated by the ANOVA procedure. The method revealed absence of convergence in three-continent and pairwise-continent analyses in almost every community attribute that was measured. Significant convergence was detected only in the foraging score for the North America-South America comparison. This implies that the foraging mode used by ants for searching seeds is more similar within similar environments in the two continents than between different environments in the same continent. Significant historical effects were much more prevalent than convergence both in threecontinent and pairwise-continent comparisons. Abundance at baits, body size, and foraging distance were more variable within similar environments in different continents than between different environments in the same continent. The overall absence of convergence documented in this study suggests that constraints related to the evolutionary history of each species assemblage have inhibited convergent evolution in response to local selective pressures. ADDITIONAL KEY WORDS:-ANOVA - desert - convergent evolution - phylogeny. CONTENTS Introduction . . . . . . . . Material and methods . . . . . . Study sites and sampling methodology . Statistical procedure . . . . . Results . . . . . . . . . Convergence . . . . . . . Historical effects . . . . . . . . . . . . . . Discussion Acknowledgements . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 31 31 34 35 35 36 39 41 42 INTRODUCTION It is now widely accepted that species inhabiting similar habitats but different regions may evolve in similar ways, originating morphological and ecological 0024-4066/95/050029+16 $08.00/0 29 0 1995 The Linnean Society of London R.G. MEDEL 30 equivalents among independently evolved biotas. Most evidence giving support to the idea of convergence comes from tests on morphological and behavioural traits of individual species (e.g. Cody, 1974; Mares, 1976; Ricklefs & Travis, 1980), but analyses at the level of guilds or communities are more equivocal (Terborgh & Robinson, 1986). One of the most critical problems for testing convergence directly is that similarity and convergence are not equivalent concepts (Schluter, 1986). To demonstrate similarity among two independently evolved assemblages does not imply that they underwent convergence, because the same degree of similarity might have been attained from a wide range of initial ancestral states. Consequently, in the absence of knowledge about ancestral assemblages it may be difficult to conduct an unequivocal test for convergence. Instead, two indirect approaches have been used. The ‘outgroup’ approach contrasts the level of similarity observed among assemblages in similar habitats but in different regions with the level of similarity among the focal assemblage with a ‘control’ assemblage in a different habitat but in the same region. If the similarity among the hypothesized convergent species assemblages is higher than their similarity to the ‘control’ assemblage, it suggests that convergent evolution produced the actual similarity (e.g. Fuentes, 1976; Blonde1 et aL, 1984). The problem with this approach is that it is very sensitive to which assemblage is selected as ‘control’ (Orians & Paine, 1983; Wiens, 1991). A second approach involves quantifylng the relative contribution of habitat and history to the overall variation in the metric of interest by means of an analysis of variance (ANOVA) procedure (Schluter, 1986). This method has at least two important advantages over the ‘outgroup’ approach. First, by knowing the amount of variation required for convergence to be statistically significant, it is possible to assess degrees of convergence rather than all-ornone results. Because conclusions about convergence may differ depending upon which variable is analysed, the ANOVA procedure provides information about the relative sensitivity of each variable to exhibit convergence. Second, it permits one to quantify the relative contribution of habitat or some abiotic variable and historical factors to the overall variation in the character selected for analysis. These factors deserve special attention in view of recent assertions that studies of convergence may benefit from including the phylogeny of species and the history of different continents (Westoby, 1988; Mares, 1993). Despite the advantages of the ANOVA approach, its most severe shortcoming is the large amount of comparable field measurements of independently evolved biotas needed to carry out the analyses. Ideally, assemblages should be sampled in a variety of habitats or abiotic conditions using the same methodology. These requirements are rarely met and lack of replication within habitats and continents continues to be an important obstacle to estimating convergence by the ANOVA procedure. Harvester ants are among the most important granivorous taxa in deserts of the world (Brown et aL, 1979; Davidson et aZ., 1980). Previous global comparisons of the ant biotas of desert and Mediterranean ecosystems have failed to reveal similarities attributable to convergent evolution (e.g. Kusnezov, 1957; Hunt, 1973; Orians et aZ., 1977). Even more specific tests of convergence carried out in desert harvester ant assemblages have failed to reveal convergence between Australia and North America concluding that historical ’ CONVERGENCE IN HARVESTER ANT ASSEMBLAGES 31 differences among continents are more important in determining ant community structure (Morton & Davidson, 1988). The notion of historical effects has been used in the literature to denote either the influence of past local or regional contingencies on contemporaneous ecological patterns (e.g. Gould & Woodruff, 1990; Herrera, 1992), or as a synonym for ‘phylogenetic constraint’ to denote that the time-course of phylogeny sets limits on the evolution of individual species (e.g. Felsenstein, 1985; Brown & Maurer, 1987). Although the conceptual limits between these two categories of phenomena are difficult to trace unequivocally, in the context of this article I will consider historical effects in relation to the former meaning, that is, in relation to the regional or continental contingencies outside the realm of local ecological processes that potentially affect the structure of species assemblages. The aim of this paper is to test the hypothesis of convergence in harvester ant communities of Australian, North American and South American arid zones by examining the relative contribution of present-day environments and historical effects to community characters. I apply Schluter’s ANOVA method to test if the variability in the structure of harvester ant assemblages is greater between localities with different mean annual precipitation within a continent than that with similar precipitation on different continents, as expected from the hypothesis of convergence. Ideally, the hypothesis should be tested with reference to the past states of communities, but in the absence of such information, I will interpret the results as indirect evidence of community-level convergence. According to the ANOVA procedure, the absence of convergence need not imply that historical factors have been important. Instead, history will be important when the variability among localities that differ in precipitation within continents is significantly lower than the variability of localities with similar precipitation on different continents. MATERIAL AND METHODS Study sites and sampling methodology The data base included a total of 42 harvester ant communities inhabiting deserts of Australia, North America and South America (Fig. 1). Information for Australian and North American ant communities was gathered from Morton & Davidson (1988) and Davidson (1977), respectively. I obtained data for 16 South American ant communities, including 11 from the Monte desert (Argentina) and 5 from Chilean arid zones. The total number of communities considered for analyses by continent was 16 in Australia, 10 in North America and 16 in South America (Table 1). Field work in South America was carried out during the summer seasons of 1990 and 1991. It has long been recognized that in deserts, seed production and primary productivity are best estimated by the amount of precipitation (Beatley, 1969; Rosenzweig, 1968). Although seasonal and interannual variability in precipitation may be at least as important as the mean value to characterize habitats in arid zones (Crawford & GOSZ,1982; Ezcurra & Rodrigues, 1986), granivorous species richness and population density tend to be positively 1500 Figure 1. Map indicating the geographical regions and study sites in arid and semiarid zones of Australia, North America, and South America. south American communities are M y named in the Appendix. 1200 - AUSTRALIA ' 'm CONVERGENCE IN HARVESTER ANT ASSEMBLAGES 33 TABLE1. Number of harvester ant communities used in this work, by continent and precipitation intervals Continent Mean annual precipitation Australia North America South America 50-99 100-149 150-199 200-249 250-299 300-349 2 6 4 2 2 3 2 3 2 - 6 3 3 2 2 correlated with the mean annual rainfall in arid zones in the range of 0300 mm (Brown, 1975; Davidson, 1977; but see Abramsky & Rosenzweig, 1984). Several other variables such as vegetation structure and soil type may contribute importantly to characterize habitats in arid zones. However, the absence of truly comparable field measurements for the 42 communities in the three continents restricted the analysis to the local mean annual precipitation as a macroscopic habitat descriptor. South American study sites were selected after inspection of precipitation data files in the Ministerio de Obras Piiblicas for Chile and in the Servicio Meteorologico Nacional for Argentina. The range of precipitations included in the analyses was 119-338 mm for Australian localities, 76-276 for North American localities, and 43-332 for South American localities (see Appendix). Previous inspection of data revealed that intervals of 50 mm of mean annual precipitation were appropriate to evaluate variability between continents without losing replicates within continents. Consequently, each assemblage was assigned to a precipitation interval of 50 mm according to the mean annual precipitation recorded for the site. With this classification, all categories were represented in at least two continents, as required for analysis, except for the first precipitation interval (0-49) where only one assemblage was present. I decided to include such locality (Caldera, South America, 43 mm) in the next precipitation interval (50-99) (Table 1). In order to make the data from South America comparable with those of the other two continents, I used a sampling procedure similar to that used by Morton & Davidson (1988). At each study site, a grid of 80 petri dishes in a configuration of 8 x 10 was laid out. Dishes were buried at ground level and spaced regularly 5 m one from each other. Each petri dish was filled with cracked millet seeds and four strips of masking tape were set up to facilitate the access of ants to seeds. Baits were offered to ants approximately 2 h before the first census. Baits were replenished as needed. To sample ant species that were active at different temperatures, censuses were conducted at soil temperatures of 25", 35", 45" and 50°C. A fifth census was made 2 h after sunset to include ants with nocturnal activity. Each census consisted of the visual inspection during 60 s of the species foraging at each bait. I considered an ant to be a seed-eater when it was observed foraging in the seed baits or transporting seeds to the nest. Ant species observed at seed baits were therefore included as part of the harvester ant assemblage, even 34 R. G . MEDEL though some species were not primarily seed-eating (e.g. Acromyrmex). Identification of species followed Kusnezov (1978) for Argentinean ants, and Snelling & Hunt (1975) for Chilean ants. Six community attributes were singled out for analyses. (1) Species richness, calculated simply as the number of species present in the assemblage. (2) Alpha diversity, calculated by the Shannon diversity index, H’ = - Cpilnpi, where pi is the relative abundance of each species. Abundance was measured as the number of baits occupied by species i as a proportion of the total number of baits occupied by all species in the grid. A bait was considered occupied by species i when it was present in the bait at any census period. Although most time individual colonies occupied only one bait, this measure gives weight to species with large colonies foraging in several baits simultaneously (e.g. Acromyrmex). (3) Number of occupied baits, calculated as the sum of the occupied baits by the total of species in the assemblage. (4) Community body size, calculated from the mean size of each species. Body size was estimated indirectly by measuring the head width of ants with a binocular microscope (precision 0.05 mm). Because size polymorphism is variable among communities (Davidson, 1978; Oster & Wilson, 1978), I measured the size of every species, regardless of whether they had been measured in other localities. An average number of 13 workers of each species was measured. Community body size was calculated by the index, BS = C$pi, where ti is the mean body size of species i, and pi represents its relative population abundance as previously defined. (5) Community foraging distance, measured as the linear distance from the nearest nest entrance to the baits foraged by workers. This community character was calculated by the index, D = Cdipi, where di is the average foraging distance of species i, and pi is the relative abundance of species i as before. (6) Community foraging score reflects the foraging mode used by any species while searching and collecting seeds. A score of 1 was assigned to species with individual foraging and a score of 0 to species with columnar foraging mode. A score of 0.5 was assigned to species having a mixed foraging. This assignment permits calculation of a foraging score as, S = Cfipi, where fi is the foraging score of the species i, and pi its relative abundance as previously defined. Statistical procedure Following Schluter (1986), the total variance of each community attribute was partitioned into the three components: KQtd= vp+V+V where is the total variance of the community character, V, is the fraction of total variance attributable to effects of precipitation, V, is the portion of total variance attributable to continental or historical effects, and V, is the random error component, measured as the mean square error in ANOVA. Because of the unbalanced sampling design, data were analysed within the framework of GLM for higher efficiency. I used an additive model with no interaction term among continent and precipitation intervals assuming that variance components were additive when ant assemblages were replicated within cells. This assumption proved to be reasonable after comparison of CONVERGENCE IN HARVESTER ANT ASSEMBLAGES 3.5 the MSe in a first ANOVA that included missing values and the MSe in a second ANOVA with missing values replaced with their predicted values from the first ANOVA (Schluter & Ricklefs, 1993). Variance components for each attribute were computed from the two-way ANOVA using community values as replicates (Sokal & Rohlf, 1981, see details in Schluter & Ricklefs, 1993). Since replicates in some cells included communities from the same continent, unknown effects of spatial autocorrelation among localities may influence the final results. In order to assess such possibility, I performed a second analysis that included the average of their community characters. I calculated convergence according to the index where C represents the fraction of the total variance in the community character that is among precipitation categories. C ranges from 0 to 1. When C = 0, it reflects total absence of convergence, when C = 1, it indicates complete convergence and 0 error within precipitation intervals. Similar to convergence, historical effects were calculated by where H reflects the proportion of the total variance that is among historical groups or continents. V, and Kod are defined as before. H may vary between 0 and 1 indicating from nil to complete importance of historical effects. The statistical significance of convergence and historical values was assessed by standard two-way ANOVA. I made three-continent and pairwisecontinent analysis for every community attribute. RESULTS Convergence The analysis of all three continents combined revealed non-significant C values for every community character both in replicated communities as in the average values (Table 2). Painvise continent analyses revealed that Australian and North American communities exhibited low and non-significant C values for every attribute in the replicate and the average analysis (Table 2). This finding gives support to previous conclusions about the absence of convergence in the structure of harvester ant communities in these two continents (Morton & Davidson, 1988). A similar situation was observed in the comparison between Australia and South America (Table 2). Similarly, ant communities in North America and South America, had low C values in most attributes except in foraging score, which had a significant 0.327 C value in the analysis including replicated communities and a significant 0.843 in the average community analysis (Table 2). Overall, these results revealed absence of convergence in the simultaneous three-continent comparison and in the three pairwise comparisons both in the replicate as in the average analysis. R. G . MEDEL 36 TABLE 2. Values of convergence (C) and historical effects (H) of harvester ant communities of Australia, North America, and South America. Indexes range from 0 to 1 indicating nil and complete convergence, and historical effect, respectively. R = analysis with replicated community characters within continents and precipitation interval. A = analysis with averaged community characters within continent and precipitation interval ~ ~~ ~ Australia North America South America Community variable Analysis Species richness Alpha diversity Number of occupied baits Body size Foraging distance Foraging score R A R A R A R A R A R A C 0.048 0.079 0.015 0.142 0.016 0.088 0.035 0.047 0.019 0.182 0.039 0.394 H 0.077t 0.400* 0.033 0.030 0.140* 0.203t 0.491*** 0.651*** 0.232** 0.379** 0.034 0.220' Australia North America C 0.064 0.115 0.088 0.195 0.167t 0.149 0.012 0.073 0.139 0.369 0.044 0.4707 H 0.132t 0.429* 0.051 0.022 0.044 0.039 0.392** 0.716** 0.032 0.256* 0.032 0.154t Australia South America C 0.051 0.577t 0.012 0.645t 0.053 0.124 0.077 0.280 0.033 0.065 0.069 0.119 H 0.002 0.089t 0.043 0.023 0.164* 0.430. 0.038 0.050 0.039 0.000 0.003 0.280* North America South America C 0.089 0.097 0.093 0.100 0.060 0.060 0.002 0.009 0.059 0.177 0.327* 0.843* H 0.122t 0.230t 0.012 0.016 0.009 0.109 0.728*** 0.844*** 0.497*** 0.485** 0.047 0.012 t P < 0.1, * P <0.05, **P<0.01, ***P<0.001 Historical effects I7i?-ee-continent analyses The overall analyses of the three continents revealed a marginally significant difference in species richness. This trend was amplified in the average analysis where assemblages revealed a significant historical effect (Table 2). In contrast, low and non-significant historical values in alpha diversity were detected (Table 2). A significant historical effect was found in the number of occupied baits with South American communities attaining the highest values in almost every precipitation interval (Table 2, Fig. 2). Because the sampling procedure gives weight to large colonies however, it is possible that these differences tend to reflect larger colonies rather than high population abundance. Community body size showed significant differences between the three continents in the two analyses which can be attributed to the high values exhibited by North American ant species (Table 3). This is graphically depicted in Figure 3 where North American communities conspicuously fall above the average line. Regarding foraging distance, significant historical effects were detected which can be attributed to the high distance covered by North American species while searching for and collecting seeds (Table 3, Fig. 4). Finally, ant assemblages did not show significant H values in community foraging score which suggests that ants utilize roughly similar foraging modes to search for seeds on a per-locality base. However, the average analysis revealed significant differences in the community values which suggests that on the average Australian species tend to forage individually rather than in a columnar form (Table 3). CONVERGENCE IN HARVESTER ANT ASSEMBLAGES 400 v) c .0 n U .- P) 5 200 37 0 t8 0 0 El 0 0 0 0 'c 0 0 z 0 I00 200 300 Mean annual precipitation (mm) Figure 2. Abundance (as reflected by the number of occupied baits) of Australian (A), North American (n),and South American ( 0 )harvester ant communities. The straight line connects the mean values by precipitation interval. TABLE3. Mean values of each community attribute in Australia, North America, and South America. Figures indicate meanfSE Continent Community variable Australia Species richness Alpha diversity Number of occupied baits Community body size Community foraging distance Community foraging score >r 'D 0 1.2 8.50f 1.04 1.54f 1.96 106.06f25.28 0.85 f0.12 2.95* 1.04 0.61 +0.18 A 0 North America South America 5.20* 1.40 1.30f0.32 115.70& 58.88 1.25 f0.12 4.92f 1.48 0.53f0.12 7.62f 1.76 1.58f0.28 204.06 f54.34 0.83 f0.06 2.19f0.60 0.50f0.08 El 0 0 n .->r C n t 0.0 A 0 - 8 0 E 0 A ha 0 0 0 6 A 0 A 0.4 I I00 I I 200 I I 300 Mean annual precipitation (mm) Figure 3. Body size values of Australian, North American, and South American harvester ant communities. The straight line connects the mean values by precipitation interval. Symbols as in Fig. 2. 38 R.G. MEDEL Mean annual precipitation(mm) Figure 4. Foraging distance values of Australian, North American, and South American harvester ant communities. The straight line connects the mean values by precipitation interval. Symbols as in Fig. 2. Pairwise-continent analyses The comparison between Australia and North America generally revealed low and non-significant H values except for some values in species richness, body size, and foraging distance (Table 2). There was a weak historical effect for species richness ( P = 0.067), with Australian communities having more species coexisting locally than North American communities (Table 3). As in the simultaneous three-continent analysis, this trend was amplified in the average analysis. Body sizes had high H-components of variance in the replicate and average analysis, with North American ants being significantly larger than their Australian counterparts (Table 3). This results gives support to previous conclusions of Morton & Davidson (1988) that Australian ants are smaller and more tightly distributed along the size gradient than North American ants. Foraging distance exhibited significant differences on the average but not on a per-locality base (Table 2). The comparison between Australia and South America revealed nine nonsignificant H values and only three significant historical values (Table 2). South American ants had almost two-fold greater abundance at baits than Australian ants (Table 3). Because these two continents did not show significant differences in species richness, results are consistent with previous findings that Australian ants tend to be locally more rare than ants of other continents (Morton & Davidson, 1988; Morton, 1993). Assemblages differed in the foraging score only in the average analysis (Table 2). Comparing North America with South America, significant historical effects were detected in body size and foraging distance (Table 2). However, South American communities had a marginally significant higher species richness than North American ones which is consistent in the replicate and average analysis (Table 2). North American ants exhibited significantly higher values of body size and foraging distance that South American ants (Table 3). This effect is sufficiently intense to be reflected on a per-locality base and on the CONVERGENCE IN HARVESTER ANT ASSEMBLAGES 39 average between continents as judged by the highly significant H values in the two analyses. DISCUSSION Community characters were more variable within precipitation intervalsbetween continents than between precipitation intervals-within the same continent, suggesting that the mean annual precipitations that characterize environments have a small importance in determining harvester ant community structure relative to the history of continents. But why has convergence not been detected in this particular study? The reasons may lie in the methodology utilized in testing convergence as well as historical biotic and abiotic differences among continents. Although the large number of communities within precipitation intervals and continents used in this study suggests that the observed variation is real, implicit in the ANOVA procedure is that any factor leading to an increase in the pooled variance of the community character will tend to underestimate convergence and historical effects. One such factor might be the geographical complexity within continents (Medel, unpublished manuscript). Both North American and South American arid and semiarid zones are disrupted by mountain ranges that create substantial local and regional variation in climate and vegetation. For instance, the data base of South American communities consists of two faunas separated by the Andes Mountain Range since approximately the Miocene when the Andes experienced its major uplift. The Andes Mountain Range has caused striking differences in the harvester ant fauna of the two regions, such as the higher seed removal rate and the four-fold greater species richness in the Monte desert in comparison to Chilean arid zones (Medel & Vasquez, 1994). Differences among the two deserts may be related to the high diversification of harvester ant genera such as Pogonomyrmex in Argentinian arid zones (Kuznezov, 1963), and to the colonization of other granivorous genera such as Pheidole and Ehmopheidole from tropical and subtropical areas to the Monte desert (Kusnezov, 1951a, 1951b). The geographical complexity occasioned by the Andes Mountain Range within South America may have increased the pooled variance in the ANOVA, rendering low C values in continent comparisons. Another potential factor for the absence of convergence is that localities of different continents, although similar in mean annual precipitation, may not be strictly comparable in other abiotic variables (Niemi, 1985; Wiens, 1991). Even though habitats appear similar regarding one abiotic factor, they differ importantly in other variables. It is quite possible that characterization of habitats based on mean annual precipitation has obscured more meaningful abiotic variables for arid and semiarid zones. For instance, Australian ants inhabit far more habitats than North American ants within a similar range of precipitation (Morton, personal communication), which suggests that there may be no necessary connection between precipitation intervals and habitats for harvester ants. In addition, rather than long-term averages of annual precipitation, the coefficient of variation within and between years may be more meaningful when rainfall is temporally and spatially unpredictable in 40 R. G . MEDEL arid regions (Crawford & GOSZ,1982; Ezcurra & Rodrigues, 1986; Armesto et al., 1993). Edaphic variables have been previously suggested as the most important abiotic factor underlying differences in the ant fauna between continents (Morton & Davidson, 1988). Australian soils tend to be less coarse and fertile than the soils of North American arid zones (Morton & James, 1988; Westoby, 1988). South American arid zones have been described as having infertile soils, ranging from clayish to rocky depending on the locality (Beek & Bramao, 1969). Chilean lowland arid zones are characterized by an assortment of broad valleys and mixed landscapes which have mainly infertile sandy soils devoid of vegetation (Killingbeck, 1993). At higher altitudes, soils are predominantly derived from volcanic ash, which also retain little water (Arroyo et al., 1988). The Argentinian Monte desert has landscape units of deep valleys surrounded by low mountains called ‘bolsones’. Each landscape unit consists of a variety of soils such as badlands, muddy spots, dunes, and small salt marshes, that create substantial variation across short distances (Morello, 1958; Mares et al., 1985). Although differences in edaphic conditions may be important in generating the observed differences among the three continents, in the absence of quantitative measures it is not currently possible to evaluate the effect of these abiotic variables by Schluter’s (1986) procedure. Contrary to the convergence effects, historical effects were statistically significant in several pairwise and three-continent comparisons. Phylogeny and degree of specialization of ants in seed consumption must have been important historical factors as suggested by differences in the taxonomic membership at each continent. In general, most harvester ant genera are concentrated in the subfamily Myrmicinae (Holldobler & Wilson, 1990). This situation is well represented in North American deserts where all species sampled belong to this subfamily. North American harvester ants consist of the genera Messor, Novomessor, Pheidole, Pogonomyrmex, and Solenopsis. A somewhat more diversified harvester ant fauna was observed in Australian arid zones where 66% of the species reported in Morton & Davidson (1988) belong to the Myrmicinae, and 34% to the Formicinae. Australian genera consist of Chelaner, Melophorus, Meranoplus, Monomorium, Pheidole, and Tetramorium, all classified as specialist granivores by Andersen (1991). South America had the taxonomically most diversified harvester ant fauna, with 52% belonging to the Myrmicinae, 24% to the Dolichoderinae, 19% to the Formicinae, and 5% to the Pseudomyrmecinae. The four genera that evolved a granivorous form in arid South America are Elasmopheidole, Pheidole, Pogonomymzex, and Solenopsis (Kusnezov, 1956). However, in addition to these genera, in South America I detected seed removal by Amomyrmex, Araucomyrmex, Brachymyrmex, Dorymyrmex, Forelius, Prenolepis, and Pseudomymzex (see Appendix). Although the biology of South American desert ants is poorly known, it is quite possible that most species inhabiting arid zones tend to be omnivorous, feeding on seeds opportunistically. Differences in the taxonomic diversity of ants involved in seed removal and the variable degree of specialization of ants may thus represent an important historical constraint precluding convergent evolution. Differences in the accompanying fauna may also be an important historical factor precluding convergence among continents. For instance, experimental CONVERGENCE IN HARVESTER ANT ASSEMBLAGES 41 studies among harvester ants, rodents and birds in North American deserts provide convincing evidence of intense resource limitation (Brown & Davidson, 1977; Davidson, 1985; Brown et aL, 1986; Thompson et aL, 1991). On the contrary, it has been suggested that seed limitation is not an important factor in determining the ant community organization in arid Australia (Briese, 1982; Westoby et aL, 1982) presumably due to the presently depauperate fauna of granivorous rodents (Morton, 1979; Morton & Baynes, 1985), and to the low importance of seed-eating birds (Morton, 1985). A relatively similar biotic setting is observed in South America. Absence of granivorous rodents and birds comparable to the specialized seed-eating species of North America characterizes South American deserts (Orians et aL, 1977; Capurro & Bucher, 1982; Brown & Ojeda, 1987). However, in the past, the Monte desert in South America did support bipedal and presumably seed-eating marsupials (Argyrolagidae), which became extinct in Pleistocene times (Simpson, 1970; Mares, 1976; Mares et aL, 1985). The involvement of different granivore taxa in the three continents suggests that ant community evolution might be subject to dissimilar selective pressures and proceed in different directions rather than toward similar states as predicted by the convergence hypothesis. Consequently, contemporary patterns may be the result of different species associations that evolved in the past under different circumstances and are not related to present-day environments. In this study I have examined the roles of convergence and history in determining the structure of harvester ant communities inhabiting deserts of Australia, North America and South America by using Schluter’s ANOVA. The overall absence of convergence documented in this study suggests that constraints related to the evolutionary history of each biota have inhibited convergent evolution in response to local selective pressures in harvester ant assemblages. Results give support to the assertion that the hypothesis of convergence should be tested mainly at the level of individual species rather than at the community or ecosystem level (Peet, 1978; Orians & Paine, 1983; Terborgh & Robinson, 1986). However, this does not imply that studies directed to test convergence are meaningless. On the contrary, the hypothesis continues to be of value for comparative studies because it permits separation of present-day environmental from historical factors. Although many regularities among communities can in principle be attributed to similar environments, perhaps attempts to derive generalizations such as convergence among independently evolved biotas run the risk of ignoring the importance of unique evolutionary events (Wiens, 1991). Future studies of convergence that quantify the contribution of historical factors such as phylogeny to the overall variance in ANOVA, would greatly increase our ability to understand the role of evolutionary history in the structure of present-day communities. ACKNOWLEDGEMENTS This paper has benefited greatly from comments made by Francisco Bozinovic, Mauricio Canals, Diane Davidson, Fabitin Jaksic, Pablo Marquet, Stephen Morton and Dolph Schluter. D. Davidson and S. Morton are also acknowledged for providing information of their work in Australian and North American deserts. Field work in South America was aided by Sergio 42 R. G. MEDEL Herrera and Rodrigo Vasquez. Financial support was provided by Fundacion Andes and grants FONDECYT 0821-90 and DTI (Universidad de Chile) B 3539-9312. REFERENCES Abramsky Z, Rosenzweig ML. 1984. Tilman’s predicted productivity-diversity relationship shown by desert rodents. Nature 309: 150-151. Andersen AN. 1991. Seed harvesting by ants in Australia. In: Huxley CR, Cutler DF, eds. Ant-plant interactions. Oxford: Oxford University Press, 493-503. Armesto J, Vidiella PE, Gutibrrez JR. 1993. Plant communities of the fog-free coastal desert of Chile: plant strategies in a fluctuating environment. Revista Chilena de Historia Natural 66: 271-282. Arroyo MTK, Squeo FA, Armesto JJ, Villagrh C. 1988. Effects of aridity on plant diversity in the northern Chilean Andes: results of a natural experiment. Annals of the Missouri Botanical Garden 75: 55-78. Beatley JC. 1969. Dependence of desert rodents on winter annuals and precipitation. Ecology 50: 721724. Beek KJ, Bramao DL. 1969. Nature and geography of South American soils. In: Fitkkau EJ, Illies J, Hinge H, Schwabe GH, Sioli H, eds. Biogeography and ecology in South America. The Hague: Dr W. Junk, 82-112. Blonde1 J, Vuilleumier F, Marcus LF, Terouanne E. 1984. Is there ecomorphological convergence among Mediterranean bird communities of Chile, California and France? Evolutionary Biology 18: 141-2 13. Briese DT. 1982. Relationship between the seed-harvesting ants and the plant community in a semiarid environment. In: Buckley RC, ed. Ant-plant interactions in Australia. The Hague: Dr W. Junk, 11-24. Brown JH. 1975. Geographical ecology of desert rodents. In: Cody ML, Diamond JM, eds. Ecology and evolution of communities. Cambridge: Belknap Press, 315-341. Brown JH, Davidson DW. 1977. Competition between seed-eating rodents and ants in desert ecosystems. Science 196: 880-882. Brown JH,Maurer BA. 1987. Evolution of species assemblages: effects of energetic constraints and species dynamics on the diversification of the North American avifauna. American Naturalist 130: 117. Brown JH,Ojeda RA. 1987. Granivory: patterns, processes, and consequences of seed consumption on two continents. Revista Chilena de Historia Natural 60: 337-349. Brown JH, Reichmann OJ, Davidson DW. 1979. Granivory in desert ecosystems. Annual Review of Ecology and Systematics 10: 201-227. Brown JH, Davidson DW, Munger JC, Inouye RS. 1986. Experimental community ecology: the desert granivore system. In: Diamond JM, Case TJ, eds. Community ecology. New York: Harper & ROW,41-61. Capurro HA, Bucher EH. 1982. Poblaciones de aves granivoras y disponibilidad de semillas en el bosque chaqueiio de Chamical. Ecosur 9: 117-131. Cody ML. 1974. Competition and the structure of bird communities. Princeton, New Jersey: Princeton University Press. Crawford CS, Gosz JR. 1982. Desert ecosystems: their resources in space and time. Environmental Conservation 9: 181-195. Davidson DW. 1977. Species diversity and community organization in desert seed-eating ants. Ecology 58: 711-724. Davidson DW. 1978. Size variability in the worker caste of a social insect (Veromessor pergandei Mayr) as a function of the competitive environment. American Naturalist 112: 523-532. Davidson DW. 1985. An experimental study of diffuse competition in harvester ants. American Naturalist 125: 500-506. Davidson DW, BrownJH, Inouye RS. 1980. Competition and the structure of granivore communities. Bioscience 30: 233-238. Ezcurra E, Rodrigues V. 1986. Rainfall patterns in the Gran Desierto, Sonora, Mexico. Journal of Arid Environments 10: 13-28. Felsenstein J. 1985. Phylogenies and the comparative method. American Naturalist 125: 1-15. Fuentes ER. 1976. Ecological convergence of lizard communities in Chile and California. Ecology 57: 3-1 7. Gould SJ, Woodruff DS. 1990. History as a cause of area effects: an illustration from Cerion on Great Inagua, Bahamas. Biological Journal of the Linnean Society 40: 67-98. Herrera CM. 1992. Historical effects and sorting processes as explanations for contemporary ecological patterns: character syndromes in Mediterranean woody plants. American Naturalist 140: 42 1-446. CONVERGENCE IN HARVESTER ANT ASSEMBLAGES 43 Holldobler B, Wilson EO. 1990. The ants. Cambridge, Massachusetts: Harvard University Press. Hunt JH.1973. Comparative ecology of ant communities in Mediterranean regions of California and Chile. University of California: D. Phil. thesis. Killingbeck KT. 1993. Nutrient resorption in desert shrubs. Revista Chilena de Historia Natural 66: 345355. Kusnezov N. 1951a. El genero Pogonomyrmex Mayr. Acta Zoolbgica Lilloana 11: 227-333. Kusnezov N. 1951b. El genero Pheidole en la Argentina. Acta Zoolbgica Lilloana 12: 5-88. Kusnezov N. 1956. A comparative study of ants in desert regions of central Asia and South America. American Naturalist 90: 349-360. Kusnezov N. 1957. Number of species of ants in faunae of different latitudes. Evolution 11: 298-299. Kusnezov N. 1963. Zoogeografia de las hormigas en Sudamerica. Acta Zoolhgica Lilloana 19: 3-186. Kusnezov N. 1978. Hormigas argentinas: clave para su identijcadon. Tucumh, Argentina: Fundaci6n Miguel Lillo. Mares MA. 1976. Convergent evolution of desert rodents: multivariate analysis and zoogeographic implications. Paleobiology 2: 39-63. Mares MA. 1993. Desert rodents, seed consumption, and convergence. Bioscience 43: 372-379. Mares MA, Morello J, Goldstein G. 1985. The Monte desert and other subtropical semi-arid biomes of Argentina, with comments on their relation to North American arid areas. In: Evenari M, NoyMeir I, Goodall DW, eds. Hot deserts and shrublands. Amsterdam: Elsevier, 203-237. Medel RG, Vbquez RA. 1994. Comparative analysis of harvester ant assemblages of Argentinian and Chilean arid zones. Journal of Arid Environments 26: 363-371. Morello J. 1958. La provincia fitogeografica del Monte. Opera Lilloana 2: 1-155. Morton SR 1979. Diversity of desert-dwelling mammals: a comparison of Australia and North America. Journal of Mammalogy 60: 253-264. Morton SR 1985. Granivory in arid regions: comparison of Australia with North and South America. Ecology 6 6 1859-1866. Morton SR 1993. Determinants of diversity in animal communities of arid Australia. In: Ricklefs RE, Schluter D, eds. Species diversity in ecological communities: historical and geographical perspectives. Chicago, Illinois: University of Chicago Press, 159-169. Morton SR, Baynes A. 1985. Small mammal assemblages in arid Australia: a reappraisal. Australian Mammalogy 8: 159-169. Morton SR, Davidson DW. 1988. Comparative structure of harvester ant communities in arid Australia and North America. Ecological Monographs 58: 19-38. Morton SR, James CD. 1988. The diversity and abundance of lizards in arid Australia: a new hypothesis. American Naturalist 132: 237-256. Niemi GJ. 1985. Patterns of morphological evolution in bird genera of New World and Old World peatlands. Ecology 66: 1215-1228. Orians GH, Paine RT. 1983. Convergent evolution at the community level. In: Futuyma DJ, Slatkin M, eds. Coevolution. Sunderland, Massachusetts: Sinauer, 43 1-458. Orians GH, Cates RG, Mares MA, Moldenke A, Neff J, Rhoades DF, Rosenzweig MI, Simpson BB, Schultz JC, Tomoff CS. 1977. Resource utilization systems. In: Orians GH, Solbrig OT, eds. Convergent evolution in warm deserts. Stroudsburg, Pennsylvania: Dowden, Hutchinson & Ross, 165224. Oster GF, Wilson EO. 1978. Caste and ecology in the social insects. Princeton, New Jersey: Princeton University Press. Peet RK. 1978. Ecosystem convergence. American Naturalist 112: 441-444. Ricklefa RE, Travis J. 1980. A morphological approach to the study of avian community organization. Auk 97: 321-338. Rosenzweig ML. 1968. Net primary productivity of terrestrial communities: prediction from climatological data. American Naturalist 102: 67-74. Schluter D. 1986. Tests for similarity and convergence of finch communities. Ecology 67: 1073-1085. Schluter D, Ricklefs RE. 1993. Convergence and the regional component of species diversity. In: Ricklefs RE, Schluter D, eds. Species diversity in ecological communities: historical and geographical perspectives. Chicago, Illinois: University of Chicago Press, 230-240. Simpson GG. 1970. The Argyrolagidae, extinct South American marsupials. Bulletin of the Museum of Comparative Zoology 139: 1-86. Snelling RR, Hunt JH. 1975. The ants of Chile (Hymenoptera: Formicidae). Revista Chilena de Entomologia 9: 63-129. Sokal RR, Rohlf FJ. 1981. Biometry: the principles and practice of statistics in biological research. Second edition, San Francisco, California: Freeman. Terborgh J, Robinson S. 1986. Guilds and their utility in ecology. In: Kikkawa J, Anderson DJ, eds. Community ecology: pattern and process. Melbourne, Australia: Blackwell, 56-90. Thompson DB, Brown JH,Spencer WD. 1991. Indirect facilitation of granivorous birds by desert rodents: experimental evidence from foraging patterns. Ecology 72: 852-863. Westoby M. 1988. Comparing Australian ecosystems to those elsewhere: what is the significance of evolutionary history? Bioscience 38: 549-556. R. G. MEDEL 44 Westoby M, Cousins JM,Grice AC. 1982. Rate of decline of some soil seed populations during drought in western New South Wales. In: Buckley RC, ed. Ant-plant interactions in Australia. The Hague: Dr W. Junk, 7-10. Wiens JA. 1991. Ecological similarity of shrub-desert avifaunas of Australian and North America. ECO~OQ 72: 479-495. APPENDIX South American ant species involved in seed removal from baits. Figures indicate the abundance of each species by locality, measured as the number of occupied baits when the five censuses are considered simultaneously (see Materials and methods for details). Localities Species Acromyrmex lobicornis Acromyrmex mpersus Araucomyrmex tener Araucomyrmex goetschi Brachymyrmex giardii Brachymyrmex laevis Brachymyrmex longicornis Dotymyrmex enrifer Dotymyrmex exsanguis Ehopheidole subaberrans Forelius grandis Pheidole sp. Pogonomyrmex brevibarbis Pogonomyrmex cunicularius Pogonomyrmex inermis Prenolepis longicornis Pseudomyrmex denticollis Sohopsis gayi Solenopsis granivora Solenopsis saevissima Solenopsis sp. 1 2 0 0 0 19 0 0 0 0 0 66 0 0 0 0 17 0 0 25 0 0 0 0 0 0 0 0 54 0 0 0 3 22 0 0 0 0 0 0 38 0 3 13 0 46 0 3 0 23 0 0 0 1 4 0 69 12 7 15 0 0 79 0 0 3 42 0 4 0 0 6 2 0 1 0 0 5 6 13 0 0 4 5 5 0 0 0 0 25 51 8 4 0 0 42 35 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5 1 6 0 0 2 2 0 0 1 8 5 0 0 4 1 2 0 36 0 0 27 1 0 0 1 3 0 10 55 0 0 5 1 0 0 27 18 0 0 40 40 0 0 0 0 0 0 66 0 0 0 4 0 10 0 0 0 0 50 0 0 0 0 0 63 0 0 55 24 0 0 0 0 0 0 0 0 4 0 0 0 0 0 5 8 0 2 1 0 0 0 0 0 27 0 0 40 0 0 0 0 0 69 0 3 58 0 0 0 4 0 0 0 0 0 0 3 1 3 0 0 1 8 0 26 30 Localities 1. Caldera, Chile (27"10'S, 70"50'W), 43 mm. 2. El Balde, Argentina (30°40S, 68"30'W), 66 mm. 3. Villa Mazh, Argentina (28"40'S, 66"30W), 80 mm. 4. Castilla, Chile (27"50'S, 70"35'W), 84 mm. 5. Retamito, Argentina (32"04'S, 68O35'W), 87 mm. 6. Zonda, Argentina (31"40S, 68"45'W), 87 mm. 7. Cachiyuyo, Chile (29"04'S, 7O05O'W), 123 mm. 8. San Jose de Jachal, Argentina (30"10S, 68"40'W), 128 mm. 9. Huaco, Argentina (3O0O5'S, 68°25'W), 128 mm. 10. La Serena, Chile (29"52'S, 71"15'W), 162 mm. 11. Agua del Medio, Argentina (29"30'S, 68"25'W), 194 mm. 12. Villa Uni6n, Argentina (29"40'S, 68°10W), 194 mm. 13. Mascasin, Argentina (31"30'S, 67"10W), 270 mm. 14. Auc6, Chile (31"45'S, 7l0O0'W), 297 mm. 15. Andalgala, Argentina (27"45'S, 66'25'W), 307 mm. 16. Uspallata, Argentina (32"40'S, 69"25'W), 332 mm. 0 0 0 0 0 30 0 0 0 3 0 24 28 27 58 0 0 0 0 0 0 0 0 5 9 2 0 4 21 3 7 1 0 9 0 0 0 0 4 0 8 0 0 4 0 0 0 57 0 3 0 56 0 0 0 0 20 24 0 0 37 0 0 0 2 51 77 0 0 0 0 0 0 65 45 3 21 0 61 0 0 0 0 24 0 0 0 0 1 4 67 0 0 52 0 0 0 0 8 0 0 9 5 0 53 0 39 0 0 65 0 0 0 0 0 0 11 40 62 0 0 0 8 38 55 65 0 0 0 9 1 0
© Copyright 2026 Paperzz