When trade-offs interact: balance of terror enforces dominance discovery trade-off in a local ant assemblage


Correspondence and present address: The University of Texas at Austin Section of Integrative Biology Brackenridge Field Laboratory 2907 Lake Austin Blvd Austin, TX 78703, USA. Tel.: (512)471 2825. Fax: (512)475 6286. E-mail: elebrun@mail.utexas.edu


  • 1Trade-offs underpin local species coexistence. Trade-offs between interference and exploitative competitive ability provie a mechanism for explaining species coexistence within guilds that exploit overlapping resources.
  • 2Omnivorous, leaf litter ants exploit a shared food base and occur in species-rich assemblages. In these assemblages, species that excel at usurping food items from other species are poor at finding food items first. In assemblages where some members are attacked by phorid fly parasitoids, host species face an additional trade-off between defending themselves against parasitic attack and maximizing their competitive abilities. Host species thus face two trade-offs that interact via the trait-mediated indirect interaction generated by phorid defence behaviour.
  • 3In this study we test for the existence of these trade-offs and evaluate the predictions of a model for how they interact in an assemblage of woodland ants in which two behaviourally dominant members are attacked by phorid fly parasitoids as they attempt to harvest food resources.
  • 4The major findings are that unparasitized species in the assemblage follow a dominance–discovery trade-off curve. When not subject to attack by phorid flies, host species violate that trade-off by finding resources too quickly for their level of behavioural dominance. In contrast, when attacked by their phorid parasitoids, the host species dominance drops such that they fall into the assemblage trade-off.
  • 5These results match the predictions of the balance of terror model, which derives the optimal host response to parasitism, indicating that the host species balance the competing fitness costs of reduced competitive dominance and loss of workers to parasitism. This result supports the view that understanding the structure of ecological communities requires incorporating the indirect effects created by trait plasticity.


Interspecific trade-offs are generally thought to be a requirement for species coexistence in communities at small spatial scales (Tilman 2000; Kneitel & Chase 2004). Trade-offs may occur between utilization of different resources or different resource densities (Tilman 1982; Grover 1997), between resource use and abiotic tolerance (Bestelmeyer 2000; Chase & Leibold 2003), between competitive ability and natural enemy invulnerability (Chase et al. 2002), and between colonization and competitive ability (Tilman 1994). In this study we examine the effects of two potentially general trade-offs on the coexistence of omnivorous ant species in an Arizona woodland community. The first trade-off occurs between the ability of an ant species to dominate food resources and its ability to discover new food resource, while the second trade-off occurs between the ability of an ant species to dominate resources and its ability to avoid the direct and indirect fitness effects of species-specific parasitoids. Both of these trade-offs are necessary to explain the pattern of species coexistence in the focal community and these trade-offs interact in a way that is consistent with Adler's (1999) balance of terror model.

The resource dominance–discovery trade-off arises from the way in which ants forage. Ant species are central-place foragers that partition the workers in their foraging force into two mutually exclusive behavioural roles: scouts and recruits. Scouts search the environment for new food sources, while recruits wait in the nest for notification of the discovery of a rich resource patch, harvest that patch, and defend it against other ant species. Thus ant colonies face a constraint in the allocation of foragers to tasks and this constraint determines the most profitable types of food items for a colony to collect. Species that invest heavily in recruits are slow to discover new resources and thus depend on large persistent food patches, even if it means fighting to defend or gain access to them; whereas species that invest heavily in scouts discover resources rapidly and depend on small ephemeral resource patches, yielding to other species at large food patches (Johnson, Hubbell & Feener 1987).

A resource dominance–discovery trade-off has been repeatedly observed in pairs or small subsets of interacting species (Lynch, Balinsky & Vail 1980; Perfecto 1994; Morrison 1996), but few studies have examined this trade-off among larger sets of interacting species (Fellers 1987; Holway 1999). This trade-off can allow for coexistence among a large number of competing species (F.R. Adler, E.G. LebBrun & D.H. Feemer, unpublished data), and is mathematically analogous to models of the competition–colonization trade-off among plant species and the virulence–transmission trade-off among parasite strains in parasite–host systems (Tilman 1994; Adler & Mosquera 2002). In these communities, a competitive subordinate can coexist successfully with a more dominant species by finding and consuming resources quickly.

Traits that allow an ant species to be competitively dominant may also make it vulnerable to attack by specialized parasitoids at both evolutionary and ecological scales (Feener 2000). Available host records suggest that a majority of ant parasitoids in the Phoridae are associated with ant species in groups that tend to be competitively dominant such as Camponotus, Pheidole and Solenopsis (Disney 1994). These species typically utilize pheromones to organize foraging, and have large colony sizes. Social pheromones used to harvest and defend food resources provide detectable and reliable olfactory cues that parasitoids can use to locate hosts (Feener, Jacobs & Schmidt 1996), and large colony sizes may make species more ‘conspicuous’ evolutionary targets for host switching, and host specialization. On an ecological scale, strong pheromone-based recruitment trails are more likely to lead to resource dominance compared with weak recruitment trails (Hölldobler & Wilson 1990); these trails are more likely to attract parasitoids (Folgarait & Gilbert 1999). The presence of parasitoids in turn triggers host defence mechanisms (e.g. hiding and defensive posturing), which reduce the ability of the dominant species to rapidly harvest resources and/or defend or usurp them from other species (Feener 1981; Feener & Brown 1992; Orr et al. 1995; Morrison 2000; LeBrun & Feener 2002; Orr, Dahlsten & Benson 2003).

Thus in omnivorous ant assemblages in which some members are attacked at food resources by specialist phorid parasitoids two interacting trade-offs may influence competition: a community wide trade-off between resource dominance and resource discovery ability, and a trade-off faced by species attacked by phorid flies between maximizing competitive ability and minimizing vulnerability to parasitism. Adler's (1999) balance of terror model explores this linkage between competition and parasitism.

The balance of terror model maximizes consumer fitness to derive the optimal competitive strategy of a guild of consumers competing for a common resource pool while subject to attack by specialist parasitoids. The critical structural assumption is a trade-off in which foraging strategies that enhance a consumer's competitive ability also increase its vulnerability to parasitism, exactly the situation faced by ant species attacked by recruitment trail-orienting phorids. The central result of the model is that, in the presence of parasites, the evolutionarily stable strategy for competing consumers is to converge on a common resource gathering potential. When consumers share two foraging strategies (such as recruitment and scouting) one of which confers increased vulnerability to parasitism (recruitment) this convergence takes the form of a shared trade-off between these two axes of competition (Adler 1999). In effect, consumer species maximize fitness by reducing their competitive abilities to match the assemblage in order to avoid excess mortality from parasitism.

In the ant assemblage that is the focus of this study, species regularly displace one another at food resources and the dominance hierarchy governing these transitions is linear and deterministic (LeBrun 2005). Two species of ants in this assemblage are attacked by phorid flies as they attempt to harvest food resources. In response, both species dramatically reduce the numbers of workers at food items and, as a result, phorid flies reduce their competitive dominance (LeBrun & Feener 2002). This study sets out to determine if a dominance–discovery trade-off exists in this assemblage. If so, does the host's competitive dominance in the presence of phorid flies converge on a level consistent with a community-wide dominance–discovery trade-off, as predicted by the balance of terror model?

Materials and methods

study system

Studies were conducted during August of 2000, June–August 2001, and July and August of 2002, in the Chiricahua Mountains, AZ. The study site (31°52′N 109°14′ W) is a semi-open woodland at 1800 m elevation (2002). This study focused on the nine species in the assemblage that were the most common ants in pitfall traps, representing 87% of all species recorded in traps. These nine species were also the most common species at cricket baits during the day and interactions between these species at baits constituted 82% of all competitive interactions observed (LeBrun 2005). We restricted our analyses to the nine most common species because this was the largest number of species that we could include and have sufficient numbers of behavioural interactions to reliably assess their dominance.

Two undescribed phorid fly species attack two of these common ant species as they attempt to harvest food resources. Apocephalus sp. 8 parasitizes Pheidole diversipilosa (Wheeler), a very abundant ant at this site. A. sp. 8 reliably appears at a large fraction of the food resources to which P. diversipilosa recruits (LeBrun & Feener 2002). Apocephalus sp. 25 parasitizes Pheidole bicarinata (Mayr). The abundance of A. sp. 25 is variable within and between years, ranging from essentially absent to abundant (E.G. LeBrun personal observation). Voucher specimens of all ant and phorid species discussed will be deposited in the Los Angeles County Museum of Natural History. Phorid species numbers follow the identification coding system of B.V. Brown (Los Angeles County Museum of Natural History collection).

measuring dominance and discovery ability

We measured dominance and discovery ability based upon interactions at large, immobile baits, a technique widely used to evaluate competition in ants (Fellers 1987; Holway 1999). Simulating natural food items, baits were adult, freeze-killed crickets Acheta domestica. Crickets were held in place using an insect pin.

We established 21 grids of 12 bait stations each. Bait stations were separated by 10 m. Grids were separated by no less than 20 m. At each bait station, we placed a 7 × 6 cm laminated index card with a fixed cricket at its centre. Bait stations were observed every 5 min for 3 h. We recorded the species and number of ants on the bait card and the species and number of any phorid parasitoids present. Baiting took place between 07.30 and 10.30 or between 14.30 and 18.00 local time. During these sampling periods ground temperatures ranged from 22·4 ± 0·6 to 30·1 ± 0·6 °C (mean ± SE) and all of the abundant, diurnal species in the assemblage were highly active (E.G. LeBrun, pers. obs.).

Discovery ability was assessed at the same bait stations as dominance. In order to avoid influencing discovery, measurements of discovery always preceded dominance trials. At each station, the first species to physically contact the bait was recorded as discovering the bait. Data for stations where multiple species were present on the bait card during the first observation of a worker touching the bait were discarded.

Behavioural dominance observations were carried out immediately after the discovery data were collected. Behavioural dominance was assessed using the outcome of interference interactions between the nine focal species. In this assemblage, a species’ behavioural dominance measured at resources of a particular size accurately reflects that species’ ecological dominance, the fraction of available resources in that size category a species captures (LeBrun 2005). Thus behavioural dominance rank provides an ecologically relevant measure of competitive status. Two categories of interference interactions were scored: (1) expulsion/escape when one species forcibly expelled another from a bait, and (2) retention/withdrawal when one species engaged in a failed attempt to take a bait from another. An attempt to usurp a resource occurred if a second, nonrecruiting species was present on the bait card for two or more consecutive observations (> 5 min) or if three or more workers of a second, mass recruiting species were simultaneously present on the bait card. Three or more workers appearing within one observation interval (5 min) provided a reliable indication that the ants were recruiting workers to the food item and thus an interference interaction had begun. Species that expelled another from a bait or retained a bait that a second species attempted to take were considered the winners of that interaction. To calculate dominance measures for species parasitized by phorids, observations were divided into those where phorids were present within one observation interval of the interaction (phorid present) and where phorids were not seen within one observation interval (phorids absent). For species not parasitized by phorid flies, only interactions with the two host species change between unparasitized and parasitized dominance measures. At the conclusion of the dominance trial all remaining bait was removed.

Forty-eight hours after all baiting used to assess both dominance and discovery was complete, a pitfall trap was placed at each bait location and left open for 48 h. Pitfall traps were 6 cm diameter plastic specimen cups partially filled with a solution of water, ethanol and soap. Pitfall traps provide a measure of foraging activity in ants that compares well with more labour-intensive litter sorting protocols (Andersen 1991). Any influence that baiting may have had on pitfall trap capture or any potential bias introduced by differences in ant species ‘trapability’ was eliminated by using only presence/absence information from the pitfall traps and not numbers of workers. This presence/absence information provided a measure of the relative access that each species had to the bait stations in the grids.

data analysis

We utilized a species’ departure from a null model to quantify its discovery ability. The null model consisted of the total number of baits discovered by the nine most common species divided by the total number of times each of these species was recorded in a trap. This proportion was multiplied by the number of traps each species was present to generate the expected number of baits that species would discover if all species were equally good discoverers (see Fig. 1). The residuals, the degree a species departed from this expected value, provided a measure of the species relative discovery abilities independent of their abundance in the environment. Discovery residuals provided results consistent with the proportion of baits first discovered but were not as sensitive to sample size.

Figure 1.

Plot of the discovery residuals used as a measure of each species discovery ability. The line represents the null expectation of the number of baits a species is expected to discover for a given abundance. Species that lie above the line are better at discovering resources than expected and those below are worse. Asterisks indicate species attacked by phorid flies. Cs =Camponotus sansabeanus (Buckley), Di = Dorymyrmex insanus (Buckley), Fg = Formica gnava Buckley, Tn = Temnothorax neomexicanus Wheeler, Me = Monomorium emersoni Gregg, My = Myrmica cf. magniceps, Pb = Pheidole bicarinata, Pd =Pheidole diversipilosa, and Pp = Pheidole perpilosa Wilson.

A species’ behavioural dominance was calculated by entering the parasitized and unparasitized interaction data into Colley matrices (http://www.colleyrankings.com/matrate.pdf). The Colley matrix is a method used for rating college football teams based solely on win–loss information that incorporates the strength of the schedule of opponents each team encountered. This method does not employ any football-specific assumptions and is thus easily applicable to any system of interactions between multiple entities that can be characterized in a win–loss manner. It provides an improved measure of dominance to the proportion of interactions a species won because it adjusts the value of each win and loss by the dominance of the competitors a species interacted with. Therefore, a species dominance rating is independent of the identity of the competitors it faced. Because we observed that all species in this assemblage interact with most other species, this improved measure of dominance did not differ substantially from the unadjusted proportion of interactions won.

All dominance data are proportional and were compared with the nonparametric Spearman rank correlation. Given the expectation of a negative relationship between dominance and discovery ability, we employed one-tailed P-values to evaluate the significance of any dominance–discovery trade-off. In the analysis of the relationship between dominance and discovery ability, the discovery ability of a species is maintained as a constant across all analyses. Dominance values vary across analyses as a species dominance value depends upon which species it interacts with and the context of these interactions. The P-values for the three tests of the dominance–discovery relationship were adjusted for multiple comparisons using the sequential Bonferroni method (Sokal & Rohlf 2000). Statistical analyses performed using systat 10·2 (systat®, 2002).


dominance ability

The Colley dominance ratings for the nine focal species and the raw proportions of interactions each species won in the presence and absence of phorid flies are presented in Table 1. The Colley dominance ratings correlated strongly with the proportion of interactions won (rs = 0·98, n = 9, P < 0·0001). Dominance ratings calculated from observations made in 2000 and 2001 were consistent across years both for interactions influenced by the presence of phorid parasitoids (rs = 0·92, n = 9, P < 0·0005) and for interactions that occurred in their absence (rs = 0·98, n = 9, P < 0·0001).

Table 1.  Characterization of the behavioural dominance and resource discovery abilities of the nine most common species in the assemblage. Host species in bold
SpeciesResidual discovery*Proportion first discovered (95% CI)Colley dominance phorids absentColley dominance phorids presentColley dominance non-host speciesProportion won phorids absent (95% CI)Proportion won phorids present (95% CI)Proportion won non-host species (95% CI)
  • *

    Residuals from Fig. 1.

  • Fraction of sites that a species was present in the trap and was the first species to discover the bait at that site.

  • Fraction of competitive interactions that a species was observed in with the other eight species in the assemblage that it won.

Camponotus sansabeanus−19·70·060·580·690·650·580·780·75
(0·01–0·14)   (0·22–0·86)(0·31–0·98)(0·97–0·26)
Myrmica cf. magniceps−7·70·200·400·430·460·410·500·53
(0·11–0·31)   (0·19–0·63)(0·25–0·72)(0·77–0·25)
Pheidole perpilosa−7·20·120·810·880·860·920·910·91
(0·03–0·27)   (0·52–1·00)(1·00–0·49)(1·00–0·49)
Pheidole bicarinata2·10·330·710·62NA0·790·67NA
(0·17–0·51)   (0·49–0·94)(0·14–0·96) 
Monomorium emersoni3·30·320·510·680·640·500·790·76
(0·21–0·44)   (0·33–0·66)(0·61–0·91)(0·90–0·52)
Dorymyrmex insanus3·30·410·300·350·390·190·330·38
(0·19–0·63)   (0·02–0·48)(0·06–0·67)(0·77–0·04)
Temnothorax neomexicanus3·50·350·110·130·130·000·000·00
(0·20–0·51)   (0·00–0·22)(0·00–0·29)(0·00–0·31)
Formica gnava10·90·360·300·430·370·220·420·29
(0·27–0·45)   (0·10–0·37)(0·26–0·58)(0·46–0·11)
Pheidole diversipilosa11·60·360·790·30NA0·900·28NA
(0·27–0·45)   (0·75–0·97)(0·12–0·47) 

discovery ability

Discovery ability was assessed using the amount species departed from the null model: Expected Discovery = 0·28 × number of traps present (Fig. 1). The constant derives from the total number of baits discovered by the nine most common species divided by the total number of times each of these species was recorded in a trap. When multiplied by a species presence in the traps, this yields the null expectation for discovery given presence. The degree a species’ departs from this line represents its discovery ability relative to its competitors, independent of its abundance. This measure of discovery was highly correlated with the proportion of traps where a species was present and where it was the first to discover the bait (Spearman rank correlation: rs = 0·87, n = 9, P < 0·003) (Table 1). Unlike dominance, discovery was a highly variable process, with the average coefficient of variation of discovery time for the nine species being 79·8%.

dominance discovery trade-off

Examining only interactions between the seven species in the assemblage not parasitized by phorids revealed a significant, negative relationship between rank dominance and rank discovery ability (rs = −0·79, n = 7, one-tailed, adjusted P < 0·04) (Fig. 2a). However, when the host species were included, but only the interactions that occurred in the absence of phorid parasitoids were analysed, there was no relationship between dominance and discovery ability (rs = −0·30, n = 9, one-tailed, adjusted P = 0·22) (Fig. 2b). In contrast, when only interactions involving the host species where phorid parasitoids were present were examined, a significant, negative relationship between rank discovery ability and rank dominance ability re-emerged (rs = −0·77, n = 9, one-tailed, adjusted P < 0·02) (Fig. 2c).

Figure 2.

Plots of relationships between residual discovery ability (x axis) and Colley dominance scores (y axis) across the assemblage. A significant negative relationship indicates the existence of a trade-off between dominance and discovery ability. Discovery abilities are the same for all plots. Dominance scores depend upon the identity of the other species included in the analysis and whether the interactions occurred in the presence of phorid parasitoids. Species abbreviations are the same as in Fig. 1. Asterisks indicate species attacked by phorid flies. Ellipses are the 95% confidence ellipse for the relationship between dominance and discovery ability. (a) Plot of the species in the assemblage that are not hosts to phorid flies (rs = −0·79, n = 7, one-tailed, adjusted P < 0·04). (b) Plot of all nine species in the assemblage. Dominance rankings for interactions involving host species consist of only interactions where phorid flies were not present (rs = −0·30, n = 9, one-tailed, adjusted P = 0·22). (c) Same as b except only interactions involving host species where phorid flies were present are included in the generation of dominance ratings (rs = −0·77, n = 9, one-tailed, adjusted P < 0·02).


trade-offs in ant communities

Among the seven nonhost species, an assemblage-level trade-off between dominance and discovery ability exists (Fig. 2a), like that documented in other ant assemblages (Fellers 1987; Holway 1999). However, the trade-off disappears when interactions between all nine species that occurred in the absence of phorids are examined (Fig. 2b). This occurs because, when not attacked, the host species, P. diversipilosa and P. bicarinata, are too dominant for their level of discovery ability for there to be a significant negative relationship across the assemblage. In contrast, when only phorid-influenced interactions with the hosts are included, the host species dominance level is driven down to the level consistent with the trade-off among the nonhost species (Fig. 2c). A trade-off between dominance and discovery ability exists in this assemblage and it is reinforced by the host species’ behavioural defences against phorid attack.

Two interspecific trade-offs influencing resource competition operate in this assemblage: an inherent trade-off between the ability to find and control food resources and a trade-off between foraging and defence mediated by an inducible, behavioural response to parasitism. The interaction of these two trade-offs influences competition in a way that supports the balance of terror model (Adler 1999). When phorids are present, they drive the competitive ability of their host species down to match that of the rest of the assemblage, indicating that the host species balance the costs of reduced competitive dominance and loss of workers to parasitism.

Examples of animals trading-off risk and foraging opportunities abound, but typically these are qualitative responses such as risk modulating habitat preference or the choice to forage or not to forage (Lima & Dill 1990). However, the host species in this system perform a much more subtle balancing act, dropping their dominance level to match the assemblage trade-off, but no farther. A hypothesis explaining why the dominance level of the host species drops to match the assemblage trade-off arises from a consideration of the opportunity costs of foraging. The optimal response to parasitism by the host is the one that minimizes the summed fitness consequences of both the direct fitness effect of parasitism, loss of workers, and the indirect fitness effect, the loss of resources to competitors. Host species, such as P. diversipilosa, that lie far above the assemblage trade-off between dominance and discovery ability, experience great competitive success and, as a result, a relatively rich resource environment. Optimal foraging theory reveals that the costs of giving up foraging opportunities in relatively rich environments are lower than in poorer environments (Stephens & Krebs 1986). Thus, for species far above the curve, a greater reduction in dominance is required to offset the fitness cost of the loss of a worker to parasitism than is required of species closer to the curve. Nonacs & Dill (1990) provide a similar example of a subtle balancing of risk and opportunity costs of foraging. In this case, the ant species trades off predation risk with patch quality assuming more risk as the difference in sugar concentration between a poor- and high-quality patch increases. Perhaps in social insects where the individuals at risk are sterile workers, these graded responses to mortality risk are more likely because the direct and indirect fitness costs of risk are of a more similar magnitude than in reproductively competent organisms.

Given the nonhost species in the assemblage are constrained by a dominance–discovery trade-off, how do the hosts achieve such high dominance and discovery values in the absence of their parasitoids? Extremely large colony size could allow a species to swamp constraints imposed by task allocation, allowing it to maintain high dynamic forager density enabling rapid resource discovery and still swiftly recruit large numbers of workers to newly discovered resources (Johnson et al. 1987). However, the social organization, and colony sizes of the host species in this system appear unremarkable, making it unlikely that their ability to quickly find and retain food resources is the result of large colony sizes. Alternatively, in order to meet the dominance–discovery trade-off, the reduction in dominance required in order to defend against parasitism may force species to enhance their discovery ability in a costly manner. Host species could accomplish this by decreasing the size of their foraging ranges. This would increase the dynamic density of scouts within their range (scouts per m2 per min), speeding resource discovery, but reducing the total amount of food available to a colony of the host species. Consistent with this hypothesis is the observation that the average distance over which the two host species in this system recruit to resources is shorter than that of any of the other focal species (E.G. LeBrun unpub. data).

trait-mediated indirect interactions

The reduction of the host's competitive ability in the presence of the attacking phorid, constitutes a trait-mediated indirect interaction (TMII) (Abrams 1995; LeBrun & Feener 2002). Because the strength of the host's defence determines the magnitude of the reduction in their competitive dominance, the trade-off between competition and parasitoid vulnerability and the trade-off between dominance and discovery interact via the TMII induced by phorids. This study provides an example of a TMII influencing a trade-off thought to underpin coexistence. In general, TMIIs may have large effects on community organization if, as in this case, the trait that is subject to modification is central to a trade-off maintaining species coexistence. This result reinforces the view that a complete understanding of ecological communities must incorporate trait plasticity and the indirect interactions that result.


Thanks are due to E. Lyon and J. Piasecke for providing field assistance. F. Adler, B. Brown, P. Coley, D. Clayton, D. Holway, S. Menke and M. Thomas provided helpful comments on the manuscript. F. Adler suggested the Colley matrix correction. B. Brown and S. Cover helped identify phorid and ant species, respectively. Any mistakes are the authors. K. Krejs’ support was invaluable. The American Museum of Natural History's South-western Research Station and its staff greatly facilitated this research. Financial support for this study was provided by a NSF Graduate Fellowship, a NSF Dissertation Improvement Grant, and the University of Utah, Department of Biology to EGL and NSF grant DEB-0316524 to DHF.