Limited dispersal and an unexpected aggression pattern in a native supercolonial ant

Abstract Understanding how social groups function requires studies on how individuals move across the landscape and interact with each other. Ant supercolonies are extreme cooperative units that may consist of thousands of interconnected nests, and their individuals cooperate over large spatial scales. However, the inner structure of suggested supercolonial (or unicolonial) societies has rarely been extensively studied using both genetic and behavioral analyses. We describe a dense supercolony‐like aggregation of more than 1,300 nests of the ant Formica (Coptoformica) pressilabris. We performed aggression assays and found that, while aggression levels were generally low, there was some aggression within the assumed supercolony. The occurrence of aggression increased with distance from the focal nest, in accordance with the genetically viscous population structure we observe by using 10 DNA microsatellite markers. However, the aggressive interactions do not follow any clear pattern that would allow specifying colony borders within the area. The genetic data indicate limited gene flow within and away from the supercolony. Our results show that a Formica supercolony is not necessarily a single unit but can be a more fluid mosaic of aggressive and amicable interactions instead, highlighting the need to study internest interactions in detail when describing supercolonies.

where limited dispersal leads to relatives aggregating together, leading to both increased kin competition and more possibilities for cooperation among relatives (Queller, 1992;Taylor, 1992). Understanding the relevant spatial scales of cooperation and competition requires knowledge on how individuals move across landscapes and interact with each other.
Ants are a good example of obligate eusociality, where a reproductive caste depends on help from a sterile worker caste (Crespi & Yanega, 1995). In several ant species, colony structures have grown beyond the ancestral simple family unit, from which eusociality originally evolved (Boomsma, 2009). In many ant taxa, dispersal is limited, and new nests are formed by budding from a parent nest, instead of long-range dispersal by flying sexual offspring (Cronin, Molet, Doums, Monnin, & Peeters, 2013). In some of these taxa, this leads to the formation of polydomous colonies where several nests are interconnected and work together as a single colony (Debout, Schatz, Elias, & McKey, 2007). Further, polydomy is often associated with polygyny (multiple reproducing queens per colony) and the colonies can grow extremely large and have extremely high numbers of queens. This leads to colony members being unrelated at the local scale (Boomsma, Huszár, & Pedersen, 2014;Helanterä, Strassmann, Carrillo, & Queller, 2009), which challenges traditional definitions of colonies as kin-selected cooperative units. Very large polydomous and polygynous colonies are often called supercolonies . In a supercolony, the cooperative colony spans larger spatial scales than a single individual can cross Pedersen, Krieger, Vogel, Giraud, & Keller, 2006), but individuals still recognize and treat each other as members of the same colony when brought together (Moffett, 2012). As an extreme example, a genetically homogenous single colony of the Argentine ant Linepithema humile has spread over the whole globe, but individuals living on different continents still behaved as one colony in behavioral experiments (Brandt, Wilgenburg, & Tsutsui, 2009;Holway, Suarez, & Case, 1998;van Wilgenburg, Torres, & Tsutsui, 2010).
Although the existence of supercolonies is something of an evolutionary paradox due to low relatedness within these cooperative units (Giraud, Pedersen, & Keller, 2002), they are ecologically very dominant (Human & Gordon, 1996). This has led to many supercolonial ant species becoming harmful invasive pests (GISD, 2019), the best-studied example being the above-mentioned Argentine ant (Giraud et al., 2002;Tsutsui & Case, 2001). It forms massive supercolonies, especially in its invasive ranges, where it has been able to spread without much competition (Tsutsui & Suarez, 2003;Wetterer, Wild, Suarez, Roura-Pscual, & Espadaler, 2009), and smaller supercolonies in its native ranges (Pedersen et al., 2006;Vogel, Pedersen, D'Ettorre, Lehmann, & Keller, 2009). Similarly, polygynous and polydomous colonies of the fire ant (Solenopsis invicta) reach very large sizes and densities, and are ecologically dominant in their invasive range (Ascunce et al., 2011;Ross, Vargo, & Keller, 1996). Supercoloniality has evolved independently in several other taxa across the ant phylogeny, some of which are not invasive . Studying the spatial scale and inner social organization of supercolonial societies in all of these taxa would give a fuller understanding of the evolution and maintenance of such high levels of cooperation (Robinson, 2014).
The ability of ant individuals to distinguish between group members and outsiders makes it possible to define the borders of supercolonies (Moffett, 2012). Individuals recognize their colonymates using olfactory cues such as cuticular hydrocarbons that can be both genetically determined and acquired from the environment and food (Ginzel & Blomquist, 2016;Howard, 1993;Vander Meer & Morel, 1998). As ants usually behave aggressively toward intruders, aggression assays are commonly used to study nest-and colonymate recognition, and provide a simple way to infer colony boundaries, and thus also spatial scales of potential cooperation (Roulston, Buczkowski, & Silverman, 2003). Even the largest supercolonies lack internest aggression within the colony, while they do behave aggressively toward other conspecific (super)colonies (Giraud et al., 2002;Holway et al., 1998;Thomas, Payne-Makrisâ, Suarez, Tsutsui, & Holway, 2007;Tsutsui, Suarez, Holway, & Case, 2000). However, not all supercolonies and supercolonial species reported in the literature have been rigorously tested for internest aggression or resource sharing (Hoffmann, 2014), but instead some of them have been described as supercolonial based only on genetic data and the spatial organization of nests .
Formica ants offer excellent possibilities for studying spatial scales of cooperation and the evolution of social organization, as this genus has large variation in social organization, from simple family units all the way to very large supercolonies (Ellis & Robinson, 2014;Helanterä et al., 2009;Rosengren & Pamilo, 1983;Rosengren, Sundström, & Fortelius, 1993). The largest reported example is a 45,000 nest supercolony of Formica yessensis (Higashi & Yamauchi, 1979), while supercolonies in other Formica species range from tens to a few thousands of nests (Markó, Czekes, Eros, Csata, & Szász-Len, 2012). In some cases, populations of the same species vary in their social structure, and, for example, in Formica exsecta, it seems that polygynous, polydomous, and supercolonial populations can arise from monogynous background populations (Seppä, Gyllenstrand, Corander, & Pamilo, 2004). Formica populations tend to be genetically viscous; that is, spatially close nests are more closely related than spatially distant ones, which is especially true in polygynous species and populations where young queens are often philopatric (Chapuisat, Goudet, & Keller, 1997;Rosengren et al., 1993;Sundström, Seppä, & Pamilo, 2005).
We investigate the nature and spatial scale of supercoloniality in the highly polydomous ant Formica pressilabris using behavioral assays and DNA microsatellite data. We test the hypothesis that behavioral colony borders correspond to the spatial and genetic structuring of a large nest aggregation, which is an underlying assumption of many previous studies. Based on this hypothesis, we expect nests at a densely populated F. pressilabris site to either belong to one supercolony without internest aggression, or possible internest aggression to occur between spatially or genetically distinct supercolonies competing with each other. We use genetic data to infer dispersal patterns within and outside of this supercolonial nest aggregation. As supercoloniality has previously been linked to limited dispersal, our hypothesis is that the dense nest aggregation is genetically somewhat separated from three other closely located study sites, and competition thus largely local. Additionally, we expect the philopatry of daughter queens to lead to genetic viscosity within our supercolonial site.

| Study species and sites
Formica (Coptoformica) pressilabris ( Figure 1) is a mound-building ant that lives on meadows and banks, builds nests of grass, and tends aphids for its main energy supply (Schultz & Seifert, 2007;Seifert, 2000) It founds new nests via temporary social parasitism with other Formica (Serviformica) species as its host, or via budding from a parent nest (Czechowski, 1975;Kutter, 1969). While monogynous colonies have been reported, secondary polygyny, where daughter queens stay in their natal nests, is common. A single nest can have hundreds of queens and grow up to over one meter in diameter, which is exceptionally large for such a small Formica species (Collingwood, 1979;Czechowski, 1975;Pamilo & Rosengren, 1984;Rosengren et al., 1993;Seifert, 2000). Colonies are also commonly polydomous with several interconnected nests and no aggression between nests (Collingwood, 1979;Czechowski, 1975;Seifert, 2000). Nest turnover is high in polydomous colonies; that is, new satellite nests are built regularly while old ones are abandoned (Bönsel, 2007).
We sampled a F. pressilabris population in Southern Finland in 2016. The sampled area consists of one 9-ha abandoned field (Särkkilen) with a large continuous nest aggregation (hereafter referred to as the supercolony site), and three closely located smaller sites with smaller nest aggregations (Storgård, Lillgård, Storsand; Figure 2). The three largest sites are former cropland, set aside for years or decades and now vegetated mainly by tall grass as well as some bushes and young trees. The smallest site is an edge area between cropland and forest. To assess whether the colonization potential of the area was fully used, we also extensively, although not exhaustively, searched for nests in other suitable habitats inside a 1 km radius from the supercolony site, but did not find additional colonies. Relatedness among worker nestmates estimated in a part of the supercolony site (r = .21 ± .02, determined with DNA microsatellites, Schultner, Gardner, Karhunen, & Helanterä, 2014) indicates polygyny and/or polydomy, that is, movement of individuals among nests. In our sampling, we counted nest mounds separated by more than 20 cm as separate nests. For our behavioral experiments and genetic analysis, we identified four spatially distinct parts of the supercolony site, with a clear gap between parts II and III ( Figure 2) and less clear gaps between the parts I and II, and III and IV, making the latter two divisions somewhat arbitrary ( Figure 2). We calculated nest densities first for the whole supercolony site and then separately for each of the four parts on it, as well as for the three smaller sites, using QGIS 3.4.1 (QGIS Development Team, 2018). For the whole supercolony site, we used the complete open area of the field for the calculation, whereas for its four parts and the three separate fields we used the areas of the polygons obtained by drawing straight lines between the outermost nests belonging to the respective areas. The age of the study population is not known, but the supercolony site has not been cultivated since the 1970s (landowner P. Forsbom, personal communication). Thus, the F. pressilabris population may have occupied the site for up to five decades.

| Behavioral assays
We performed behavioral assays in order to determine whether workers from the supercolony site behave differently toward their nestmates, conspecifics from other close and distant nests at the same site, conspecifics from other sites, and allospecific ants. These treatments are hereafter referred to as control, neighbor, distant, outside, and allospecific, respectively. We expected no aggression within the supercolony site, and that if any aggression were observed, it would occur across the nest-free gap in the middle of the site (Figure 2), as the gap may form a barrier between two separate supercolonies. We expected more aggression toward individuals from other sites than toward individuals from within the supercolony site. The closely related species F. exsecta should always be faced with aggression, as F. pressilabris has previously been shown to behave very aggressively against it (Czechowski, 1971).
We collected workers from 16 F. pressilabris nests at the supercolony site, covering all four parts ( Figure 2). Additionally, we collected F. F I G U R E 1 The study species, Formica pressilabris pressilabris workers from four nests at one of the smaller study sites, and one nest at another location approximately 30 km SW of the study population (Tvärminne). We also sampled two F. exsecta nests from the main study area and one from Tvärminne. We collected 2-5 L of nest material and a minimum of 300 workers from each nest on June 29 and 30. In the laboratory, we immediately divided all 16 focal nests from the supercolony site into two boxes to control for the possibility that physical separation in the laboratory causes aggression in the assays. We reared the laboratory nests in room temperature for three to eight days, kept them moist, and fed them daily with a Bhatkar-Whitcomb diet (Bhatkar & Whitcomb, 1970).
We tested the reaction of workers from each of the 16 supercolony site nests against control, neighbor, distant, outside, and allospecific individuals ( Figure 3). We replicated each of the five treatments five times per nest with new arenas and new individuals, except for the allospecific tests. The latter we performed in the same arenas with the same individuals as the same-nest control treatments, as there never was any aggression in the control treatments.
Our preliminary experiments showed that F. pressilabris workers act passively in standard one-on-one aggression assays on neutral arenas, usually showing no interest toward each other. Therefore, we used experimental arenas (6.5-cm-diameter fluon-coated, newly purchased plastic cups) with 15 workers on their own nest material, simulating natural conditions with nestmates and familiar odors present. In an assay like this, the observed behavior is expected to correlate with the natural nest defense behavior, revealing whether the workers would allow visitors to enter their nest or not, as even submissive ant species defend their nests against intruders (Savolainen & Vepsäläinen, 1988;Vepsäläinen & Pisarski, 1982). If an assay this sensitive does not show aggression, this can be interpreted as very strong potential for cooperation: When ants are willing to share their nest, they probably cooperate in other ways, too.
After letting the 15 focal ants calm down (when they had stopped running around and did not show signs of alert, such as opening their mandibles), we introduced one worker from another laboratory nest box. We recorded the actions of the ants for one minute from the introduction, using a Canon EOS 550D DSLR camera with a Canon EF 100 mm f/2,8 macro lens. The distance between the ants and the lens was 48.5 cm. Ants did not react toward the camera. We performed the behavioral assays on July 3-7, 2016.
One of the authors (MI) watched and transcribed the videos in a randomized order at half speed and blindly regarding which nests and which treatments were represented in each video. The durations of antennation, trophallaxis, and biting events against F I G U R E 2 (a) The study area. The sampled subpopulations are indicated with letters. A: The supercolony field Särkkilen; B-D: The smaller fields, Storsand, Lillgård, and Storgård, respectively. (b) Locations of the nests on the supercolony field (circles). The nests used for genotyping are marked with filled circles, and the nests used for the behavioral experiments with red symbols. The red squares, diamonds, triangles, and circles represent parts I, II, III, and IV, respectively (see text). There are relatively large nest-free gaps between the parts (70 m between II and III; 31 m between III and IV), except for parts I and II, which are separated by a narrower (18 m the one introduced worker were recorded using the software JWatcher 1.0. We could not reliably score behavior frequencies or minor signs of aggression such as mandible opening due to the large number of individuals on the arena and the small size of the ants. The scoring was hierarchical in the sense that when even one of the fifteen focal individuals was aggressive, we did not score nonaggressive behavior at the same time, because aggression-related pheromones could affect the behavior of other individuals. In our analyses, we combined trophallaxis and antennation as nonaggressive inspecting behavior, because the two were sometimes hard to separate from each other and both have been shown to increase when individuals recognize the opponent as a non-nestmate in polydomous Formica paralugubris (Chapuisat et al., 2005;Holzer et al., 2006). Seven out of the 400 videos could not be analyzed due to damaged files.

| Statistical analysis of behavioral data
We analyzed the presence and absence of aggression explained by the different treatment classes with a binomial generalized linear mixed model (GLMM). In all of our models, we included both the nest of the focal workers and the nest of the introduced worker as random effects to account for the nonindependence of samples coming from the same nest. We excluded the "control" treatment as no aggression occurred in the within-nest controls. We used a beta GLMM to analyze the duration of aggression among the treatment classes where it occurred.
For the videos where no aggression occurred, we also used a beta GLMM to analyze the duration of nonaggressive inspecting behavior (antennation and trophallaxis combined). For this analysis, we excluded the treatment level "allospecific" due to low sample size (n = 3), and substituted five samples with a value of 0 (= no inspecting behavior) with a value of 1 (= a millisecond of inspecting behavior) to allow the use of the beta distribution, which cannot contain zeros. For the beta GLMM's, we measured the duration of aggression or inspecting behavior as the proportion of the total time that the introduced ant was in sight.
We further tested whether the geographical distance between two separate nests explains the presence or duration of aggression, or the duration of nonaggressive inspecting behaviors. We analyzed a subset of our behavioral data within the supercolony site (treatment levels "neighbor" and "distant"), using a binomial GLMM for the presence of aggression and beta GLMM for the duration of aggression and inspecting behavior. Geographical and genetic distances are collinear in our data (see below), and the effect of these two variables cannot be fully separated in our results. Therefore, we used only geographical distance as an explanatory variable in our analysis. As we did the genetic analysis only for a single individual per nest, using genetic distance would be more problematic in connection to the nest-level behavioral data. Additionally, conspecific aggression in ants correlates mostly with chemical distance (Martin, Vitikainen, Drijfhout, & Jackson, 2012) which may have both genetic and environmental components (Ginzel & Blomquist, 2016;Vander Meer & Morel, 1998).
Thus, as geographical distance contains information of both genetic and environmental factors, we deemed it more biologically relevant than genetic distance. We analyzed the behavioral data in R (R Core Team, 2013) with the package glmmTMB (Bolker et al., 2009).

| DNA microsatellite genotyping and population genetics
To estimate gene flow among and within the four sites, we genotyped a single worker from 285 different nests, including 233 nests from the supercolony site ( Figure 2) and all nests from the three smaller study sites. We extracted the DNA using NucleoSpin Tissue extraction kits (Macherey-Nagel) and genotyped the samples with 14 DNA microsatellite markers originally developed for other Formica species (Chapuisat, 1996;Gyllenstrand, Gertsch, & Pamilo, 2002;Hasegawa & Imai, 2004;Trontti, Tay, & Sundström, 2003) using the protocol designed by Hakala et al. (2018). We scored the DNA microsatellite alleles with the software GeneMapper 5 (Applied Biosystems).

F I G U R E 3
The design of the behavioral experiment. In each assay, we tested the reactions of 15 ants from a focal nest toward one ant from another laboratory nest box. We collected ants and nest material into laboratory nest boxes (black arrows) and put 15 workers from a focal nest into each of five experimental arenas (gray arrows). Then, we introduced one worker from another laboratory nest into the arenas (dashed arrows). For each focal nest, we replicated this procedure five times with new ants and arenas. Control = introduced ant from the same nest, Neighbor = introduced ant from the same part of the supercolony field, Distant = introduced ant from a different part from the supercolony field, Outside = introduced ant from another field, Allospecific = introduced ant of a different species, F. exsecta sites, and among the four parts within the supercolony field using AMOVA's in GenAlEx 6.502 (Peakall & Smouse, 2012). To test for genetic viscosity, we performed Mantel tests using the package ecodist 2.0.1 (Goslee & Urban, 2007) in R.
We analyzed the genetic structure of the population with a Bayesian approach using the software STRUCTURE 2.3.4 (Pritchard, Stephens, & Donnelly, 2000), which clusters individual genotypes by the probability of similarity. To determine the most likely number of clusters (K), the analysis was run with K ranging from 1 to 7, using the admixture model and correlated allele frequencies. For each K value, we ran the analysis ten times with a burn-in of 100,000 steps for a run length of 300,000 steps. We estimated the most likely number of clusters by applying the delta K method with plotting the mean and standard deviation of the mean likelihood L(K) for each run in STRUCTURE HARVESTER (Earl & vonHoldt, 2012;Evanno, Regnaut, & Goudet, 2005). As the mathematical model used by STRUCTURE is not ideal with unbalanced sampling and groups with low sample sizes or patterns of isolation by distance (Kalinowski, 2011;Puechmaille, 2016), we repeated the genetic mixture analysis with a similar software, BAPS 6.0 (Corander & Marttinen, 2006;Corander, Siren, & Arjas, 2008;Corander, Waldmann, & Sillanpää, 2003). BAPS was allowed to find the most probable number of clusters with repeated runs (10 times K1-K7). As BAPS was unable to find any stable clustering without a spatial prior, we repeated the analysis as a spatial analysis with geographical coordinates. Subsequently, we performed an admixture analysis with the results of the spatial analysis.

| Mapping of the study area
The supercolony site had more than 1,300 nests, and the nest densities ranged from 254 to 401 nests/ha in the different parts of the site. The three smaller sites had 7, 16, and 29 nests, and the nest densities were 426, 25, and 25 nests/ha, respectively ( Figure 2, Table A1). We could not directly observe worker movement between nests, because F. pressilabris workers walk mostly on the ground surface under the grass cover or on grass stems, forming no visible paths between the nests. All parts of the supercolony site had many dense aggregations with nests situated very closely together. Often these aggregations had one or two large main nests and a few smaller ones. The distances between nest aggregations were often short, and the field is overall almost uniformly occupied by the species.

| Behavioral assays
There was no aggression in the same-nest controls, while the al- However, there was no significant difference between the aggression faced by ants from distant and outside nests (full test statistics in Table A2). The behavioral patterns were not consistent among the five replicates; instead, there were aggressive and nonaggressive interactions among and within all parts of the supercolony site. There were plenty of nonaggressive replicates also between distant nest pairs ( Figure 5).
In the assays where aggression occurred, its duration did not significantly differ among the treatments, except in one of the pairwise comparisons, where the effect size remained small ( Figure 6).
However, the allospecific treatment always had long aggression

F I G U R E 4
Presence and absence of aggression by treatment with model predictions for the four treatment classes that we included in the analysis (binomial GLMM). The control treatment, which had no aggression, was excluded from the model. Among those treatment classes that were included, all other pairwise differences were significant, except the difference between distant and outside. Control = introduced ant from the same nest, Neighbor = introduced ant from the same part of the supercolony field, Distant = introduced ant from a different part from the supercolony field, Outside = introduced ant from another field, Allospecific = introduced ant of a different species, F. exsecta

| Population genetics
Of the original 14 DNA microsatellite markers, we used 10 for further analysis, as four had too much missing data or significant heterozygote deficiency at all study sites (details in Tables S1-S6). The pairwise F ST values show that the four parts of the supercolony site differed genetically from each other and that this differentiation was on a level similar to the differentiation among the other study sites (Table 1). Mantel tests showed minor but significant genetic viscosity when analyzing all samples in the study area (R = 0.06, 95% CI = 0.04, 0.08; p = .038), and also when analyzing only the samples from the supercolony site (R = 0.1, 95% CI = 0.09, 0.12; p ≤ .001).
In the Bayesian clustering for the study area (Figure 9a,b), the optimal number of genetic clusters was two according to STRUCTURE HARVESTER (ΔK = 99.9, details in Table S6

| Gene flow and dispersal
Our population genetic data indicate limited dispersal within the supercolony site. Even though within-nest relatedness is low, probably due to polygyny and mixing of individuals among adjacent nests (Schultner et al., 2014), individuals do not seem to mix effectively across the entire supercolony site, not even winged sexual individuals. Such genetic viscosity at a site less than 1 km across shows that these ants disperse mostly over very short distances. Local mating between nestmates or individuals from neighboring nests must be common, as the population would otherwise not remain even weakly genetically viscous. F. exsecta, which is closely related and ecologically similar to F. pressilabris, has similar levels of genetic viscosity in its supercolonies (Seppä et al., 2012). In Formica paralugubris, a supercolony was more viscous than the surrounding nonsupercolonial population, which, just as our results, suggests that supercoloniality is linked to locally reduced dispersal (Chapuisat et al., 1997).
Importantly, data on gene flow, such as in the above-mentioned cases, do not provide any information about failed dispersal attempts. Workers in existing nests have a key role in determining which sexual individuals can establish themselves as reproducers at the supercolony site. Aggressive behavior toward individuals from distant nests could make their establishment as reproducers hard, even if they tried.
On a slightly larger spatial scale, the supercolony site is not genetically distinct from surrounding smaller polydomous colonies.
Instead, pairwise F ST values among the small sites and among the different parts of the supercolony site are in the same range. Our data suggest that longer-range dispersal among different sites is frequent enough to keep population structuring low (Table 1), although dispersal seems to be limited within the supercolony site. Ongoing long-range dispersal ensures that the supercolony is not a closed population, and extends the scale of competition beyond single sites, which may give some selective advantage to the individuals from the supercolony on larger spatial scales (Kennedy, Uller, & Helanterä, 2014;Pedersen et al., 2006). This contrasts with previous findings in F. exsecta, where polydomous colonies were genetically more different from surrounding monodomous and monogynous colonies than these were from each other (Gyllenstrand, Seppä, & Pamilo, 2005;Seppä et al., 2004), suggesting that polydomous colonies could form closed populations without much dispersal outwards.
Unfortunately, with our data we cannot assess whether dispersal is sex-biased. Male-biased dispersal is common in polygynous ants (Hakala, Seppä, & Helanterä, 2019) and in Formica overall . In socially parasitic Formica species, such as our study species, queen dispersal is further complicated by the fact that queens cannot found their nests alone, but have to parasitize a nest of their host species or possibly another nest of their own species (Buschinger, 2009;Czechowski, 1975). patches around the supercolony in our study area, which speaks for limited colonization. Overall, the dispersal strategy of F. pressilabris seems twofold: Risky long-range dispersal combined with high levels of queen philopatry and potential to spread locally through shortrange dispersal by foot. Such a dual strategy seems to be the rule in polygynous and polydomous Formica  and exists in other ant taxa too, for example, in Crematogaster pygmaea (Hamidi et al., 2017).
In addition to the current study, positive correlations between aggression and spatial distance have been reported, for example, within polydomous sites of F. exsecta (Katzerke et al., 2006) and Myrmica rubra (Fürst, Durey, & Nash, 2012;Garnas, Drummond, & Groden, 2007), although later Chen et al. (2018) did not find such a correlation in M. rubra. Importantly, Chen et al. (2018) had, through a previous set of aggression assays, assigned colony borders prior to testing for this relationship and suggest that the correlations between geographical distance and aggression found by Garnas et al. (2007) and Fürst et al. (2012) may be attributed to mixing nest pairs belonging to the same and different colonies in one analysis. Our supercolonial site may indeed consist of multiple supercolonies instead of one, but the lack of clear patterns and the overall low levels of aggression revealed in our study ( Figure 5) lend little support to the existence of clear and persistent borders at our study site. Finally, aggression has been shown to increase with internest distance also in monodomous species (Beye, Neumann, Chapuisat, Pamilo, & Moritz, 1998) and among distinct colonies or sites (Holzer et al., 2006;Pirk, Neumann, Moritz, & Pamilo, 2001;Rosengren, Cherix, & Pamilo, 1986;Zinck, Hora, Châline, & Jaisson, 2008), making these kinds of behavioral patterns hard to interpret.
The different behavioral assay methods used in many of the studies discussed above make direct comparisons of the results difficult (Roulston et al., 2003). Our method, where 15 individuals on their own nest material met one introduced ant, is very sensitive to aggression as it simulates an alien ant suddenly appearing in a nest, and there are many ants that can react. Based on our pilot experiments, we consider it likely that F. pressilabris would have shown even less, if any, aggression in some more commonly used assay types, such as one-on-one tests on neutral arenas. Even in the absence of aggression, our results show that workers spend more time inspecting any non-nestmates than nestmates. Thus, workers may distinguish between nestmates and more distant individuals, as is also suggested by increased antennation and trophallaxis in F. paralugubris (Chapuisat et al., 2005;Holzer et al., 2006), and increased antennation in argentine ants (Björkman-Chiswell et al., 2008).
While a lack of aggression between nests is commonly interpreted as a sign of shared colony identity, it does not necessarily mean that two nests share resources (Buczkowski, 2012;Giraud et al., 2002;Heller, Ingram, & Gordon, 2008). In large nest aggregations, it is relevant to ask what the true spatial scale of cooperation and competition is (Pedersen, 2012). At our supercolony site, aggression did not always occur even when testing over the nest-free gap in the middle of the field, showing that workers are willing to let individuals from relatively far away enter their nests ( Figure 5). Czechowski (1975) found that workers can move at least 20-30 m between nests in supercolonies of F. pressilabris, but such movements were considerably rarer than movements among nearer nests. In F. exsecta, workers from polydomous colonies forage on trees less than ten meters from a central nest (Sorvari, 2009).
Thus, given the width of the gap (~70 m) at our main study site, worker movements over it should be rare, and we consider true cooperation, such as transfer of brood or resources, over this gap unlikely. If resources are shared over relatively limited distances also at our study site, the two field halves should be considered functionally separate colonies, even though they are not clearly distinct based on either genetic or behavioral data. We agree with Heller et al. (2008) and Lester and Gruber (2012)

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
All authors designed and conceived the study. MI collected the data.
SMH and MI analyzed the data, prepared figures and tables, and drafted the manuscript. All authors reviewed drafts and approved the final manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data used in this study are archived in the Dryad Digital Note: Fixed effect levels: Control = introduced ant from the same nest, Neighbor = introduced ant from the same part of the supercolony field, Distant = introduced ant from a different part from the supercolony field, Outside = introduced ant from another field, Allospecific = introduced ant of a different species, F. exsecta. The aggression models do not include the level "Control," for it had no aggression. The model for antennation does not include the level "Allospecific" due to it having aggression in all but three samples and thus too small a sample size for analyzing the nonaggressive behaviors. In all of the models, the nest IDs of both the host and the visitor were included as crossed random effects.