Colony size, nestmate density and social history shape behavioural variation in Formica fusca colonies

In ants, individuals live in tightly integrated units (colonies) and work collectively for its success. In such groups, stable intraspecific variation in behaviour within or across contexts (personality) can occur at two levels: individuals and colonies. This paper examines how colony size and nestmate density influence the collective exploratory behaviour of Formica fusca (Hymenoptera: Formicidae), in the laboratory. The housing conditions of the colonies were manipulated to vary the size of colonies and their densities under a fully factorial design. The results demonstrate the presence of colony behavioural repeatability in this species, and contrary to our expectations, colonies were more explorative on average when they were kept at lower nestmate densities. We also found that experimental colonies created from larger source colonies were more explorative, which conveys that a thorough understanding of the contemporary behaviour of a colony may require knowing its social history and how it was formed. Our results also convey that the colony size and nestmate density can have significant effects on the exploratory behaviour of ant colonies.

Such colony phenotypes can be shaped by numerous properties, such as the distribution of the individual level phenotypes that comprise the group (Pruitt, Grinsted, & Settepani, 2013).
In the Formicidae group, different species can live in colonies with many different sizes, from a few to 300 million workers (Kaspari & Vargo, 1995). During the development of an ant colony, the number of nestmates usually increases steadily and the rate of colony growth can be influenced by a variety of intrinsic (e.g., number of egg-laying queens) and extrinsic factors (e.g., predation) (Bourke, 1999). In general, larger colonies can produce more complex or impressive collective feats and larger colonies can often organize more efficient foraging networks as well (Merkle & Middendorf, 2004).
For instance, in Messor sanctus Emery, 1921 collective digging is positively correlated with colony size and the tunnel networks of larger colonies are conspicuously more complex (i.e., contained more edges and vertices) (Buhl, Gautrais, Deneubourg, & Theraulaz, 2004).
Similar observations have been made in social bees, wasps and social spiders (Avilés & Tufino, 1998;Donaldson-Matasci, DeGrandi-Hoffman, & Dornhaus, 2013). Larger colonies of honeybees have a more complex dance language, which renders them more efficient at recruiting foragers to novel food patches (Donaldson-Matasci et al., 2013). Empirical evidence suggests that the division of labour and task specialization within social groups increases with group size (Amador-Vargas, Gronenberg, Wcislo, & Mueller, 2015;Anderson & McShea, 2001;Bourke, 1999), though this is not always the case (Dornhaus, Holley, & Franks, 2009). In some species, small colonies invested more in the consensus decision process than the large ones (Cronin & Stumpe, 2014). In aggregate, one can observe that colony size can shape a variety of collective attributes in social insect colonies. It, therefore, stands to reach those changes in colony size that could shape the expression of colony phenotypes that are commonly evaluated in animal personality literature as collective exploration.
Fluctuations in colony size can influence the density of nestmates (number of individuals per nest size) which can be another key parameter affecting within-nest dynamics (Perna & Theraulaz, 2017).
However, these aspects are rarely examined in social insects under laboratory circumstances (but see in Cao & Dornhaus, 2008;Gordon, Paul, & Thorpe, 1993;Modlmeier et al., 2019;Pie, Rosengaus, & Traniello, 2004). Changes in colony size and social density, for instance, could alter the way information is transferred between workers, which in turn could shape colony activities (Beshers & Fewell, 2001;O'Donnell & Bulova, 2007) and group behaviour (Pacala, Gordon, & Godfray, 1996). O'Donnell and Bulova's (2007) simulation model indicates that density-mediated changes in social interactions shape the division of labour and the organization of work across colonies. Taken together, the available evidence suggests that social density within nests could be another important factor in determining the phenotypes that colonies express during staged behaviour assays.
A common measure of behavioural repeatability is the rate at which individuals move through a novel space or environment, normally referred to as exploratory behaviour or exploration (Arvidsson, Adriaensen, Dongen, Stobbeleere, & Matthysen, 2017;Dingemanse, Both, Drent, Oers, & Noordwijk, 2002). In this study, we investigated (a) the repeatability of the collective exploratory behaviour in laboratory colonies of Formica fusca Linnaeus, 1758 (Hymenoptera: Formicidae) and (b) how colony size and nestmate density influence exploratory behaviour, using a fully factorial experimental design. We examined colony exploration, in particular, because the exploration of the exterior environment is how colonies obtain information about potential food sources, competitors and predators (Clobert, Danchin, Dhondt, & Nichols, 2001;Devigne & Detrain, 2002). For example, Gordon (1996) investigated the exploratory behaviour of colonies of different sizes (Linepithema humile, [Mayr, 1868]) and found that the largest number of exploring workers were present in the largest colonies. Similar positive correlation was found between colony size and the number of foragers in the studies with Pogonomyrmex salinus Cole 1983 (Porter & Jorgensen, 1981), Cataglyphis cursor (Fonscolombe, 1846) (Retana & Cerda, 1991) and Temnothorax rugatulus (Emery, 1895) (Charbonneau & Dornhaus, 2015). We therefore predicted that the level of exploratory behaviour would be more intense in large and dense F. fusca colonies.

| MATERIAL S AND ME THODS
Formica fusca is a facultative polygynous species (Czechowski, Radchenko, Czechowska, & Vepsäläinen, 2012;Seifert, 2018) being abundant in Central Europe, especially in areas with abundant dead wood. Its nests can be found in dead wood, soil or moss pads. This species relies on the swift discovery of ephemeral food patches and avoiding workers of dominant heterospecifics. Formica fusca is thus generally not aggressive, though particularly large colonies can defend some kinds of resources (Seifert, 2018).
Colonies of F. fusca were collected from Vámospércs, Hungary (47.524796 N, 21.886494 E) in May 2014, when the workers were active and brood was still not present in nests. These conditions made possible the collection of whole colonies and the accurate counting of workers and queens (workers: 85-635, queens: 1-10). We collected 14 colonies (hereafter: source colony) that were divided into 28 subcolonies (hereafter: experimental colonies). The subcolonies that were considered inappropriate for the observations were removed from the study (dead queen or workers), so finally 22 subcolonies were retained. Each new experimental colony contained only one fertilized queen (that laid eggs in the laboratory), to diminish the possible effects of queen number in determining colonies' exploratory behaviour. For housing the experimental colonies, we used four different sized artificial nests (made from concrete) with two colony size categories (25 or 135 workers). This produced two experimental densities of 0.25 or 0.5 worker/cm 2 in a 2 × 2 full factorial design (Table 1). Colonies were fed with Bhatkar diet (Bhatkar & Whitcomb, 1970) and were provided with water ad libitum. They were maintained at room temperature (22-25°C) with a natural light/dark cycle. Colonies were investigated in block design, where every block contained one nest from each of the four treatment combinations, respectively.

| Behavioural tests: Exploration of novel environment and objects
The exploratory behaviour of F. fusca colonies was investigated in circular arenas (diameter: 39.5 cm). Before the assays, we placed two pieces of lead into the arena as novel objects and two drops of sugar solution as a positive signal to instigate scouting. The novel objects were observable for the ants, but they could not remove them.
During the assay, the colony's nest and the arena were connected via a 10-cm-long plastic tube ( Figure 1). Thus, workers could freely enter the arena, explore it, and return to the nest. After the acclimatization period (20 min), which permitted the ants to investigate and use the tube connecting the nest with the arena, we opened the tube towards the arena and video recorded the movement of the exploratory workers for 40 min. The camera (Panasonic HC-V510) was focused on the exploratory arena, and the MWrap event recorder (Bán, Földvári, Babits, & Barta, 2017) was used to code the behaviours from the videos. At the time of the analysis, ants entering the arena were not distinguished individually, so every behaviour of workers from the same colony was analysed together.
The following variables were recorded: the latency of first worker entering into the arena (the "entrance," elapsed time between opening the connection and the first worker entering into the arena), the latency of the first return home to the nest ("lat. home," elapsed time from "entrance" to the first return to the colony) and the latency of the first visit to an object ("lat. sugar water" and "lat. object" separately, measured from "entrance"). The number of entries into the arena ("nr. entries") and the number of visits to objects ("nr. objects," TA B L E 1 Housing conditions (i.e., treatment) with the number of participating colonies in the behavioural tests (i.e., number of colonies tested, per nest types)

Nest type
Nest size (cm 2 ) Chamber size (cm 2 )  Table 1 for details). The sizes of the novel objects are proximate on the figure Arena Nest 10 cm 39.5 cm "nr. sugar water") were also counted. The number of homecoming workers ("nr. returnings") and the maximum worker number ("max.

Nr. workers in the nests
workers") in the arena were strongly correlated with the "nr. entries" (Spearman correlation: rho = 0.978, p < .001, n = 82 and rho = 0.851, p < .001, n = 82, respectively), so we omitted them from subsequent analyses. We also recorded the encounters between the individuals (contact with each other: antennation or trophallaxis; "lat. interactions," "nr. interactions"). The behavioural tests took a month (from 13.06.2014 to 14.07.2014). The behavioural assay was repeated four times per colony, with approximately 8 days (mean = 8.48, median = 8, SD = 2.52) separating consecutive assays on the same colony. If the queen of a colony or the colony itself died, no further tests were performed with that colony. We only considered colonies that completed at least two trials (n = 22).

| Statistical analysis
All analyses were performed in the R interactive statistical environment (version 3.3.3, R Core Team 2017). We applied a principal component analysis (PCA) (base "stats" package and "prcomp" function) on the behavioural variables to test for possible correlations among all response variables. The PCA was carried out on the covariance matrix based on the average value of the behavioural variables. We only analysed those principal components (PCs) further that had eigenvalues greater than 1 (Kaiser-Guttman criterion; Jackson, 1993).
PCA scores (from first (PC1) and second principal components (PC2) were analysed using linear mixed-effect models (LMMs). The full models included experimental colony size, treatment density, original queen number and original colony size as fixed effects, and the source colony ID as a random factor. We investigated behavioural repeatability at the colonial level by calculating the repeatability (marked with R) of each variable using the rptR package (Stoffel, Nakagawa, & Schielzeth, 2017) with 1,000 bootstrap steps. p-Values were obtained by likelihood ratio tests, and confidence intervals (95%) were derived by 1,000 permutations. Repeatability values, which are significantly higher than zero, indicate a significant among-colony component to the observed intraspecific variation (Réale et al., 2007). "Nr. entries" and "lat. home" were log-transformed and analysed with Gaussian family, while "entrance," "lat. sugar water," "lat. object" and "lat. interaction" were binomized by the median and analysed with binomial family. "Nr. interactions," "nr. sugar water" and "nr.

| Correlations among the response variables
Principal component analysis is often used to test for the presence of behavioural syndromes (Dingemanse, Dochtermann, & Wright, 2010). In our case, PC1 explained 54.24%, while PC2 explained 21.88% of the total variance (76.12%). The first component was composed of five variables ("entrance," "lat. home," "lat. sugar water," "lat. object" and "nr. entries"), and it can be referred to as "activity in the arena." The higher score indicated that the colony showed a lower level of activity in the arena with higher latency times and fewer entries. The second component was composed of two variables ("nr. sugar water" and "nr. objects"), and it can be referred to as "exploration of the arena" (Table 2).

| Repeatability of the collective behaviour
We detected temporally consistent among-colony differences The repeatability of other behavioural traits was not significant (Table 3).

| Effects of the housing conditions
Principal component 1 scores were positively correlated with nestmate density (ß = 1.963, SE = 0.878, t = 2.236, p = .036) suggesting that colonies from nests with higher densities showed a higher level of activity in the arena during the examination. PC2 was not related to any of the housing condition parameters.
Colonies with high nest density treatment exhibited fewer en-

| Effects of queen number on behaviour
We could not detect any significant association between the number of queens within source colonies and any aspect of the exploratory behaviour of their descendent experimental colonies.

| D ISCUSS I ON
The study herein aimed to explore the extent to which F. fusca colonies exhibit temporal among-colony repeatability in their collective exploratory tendencies and how these tendencies are affected by the size of the colonies and the densities of workers within the nests.
The principal component analysis identified two behavioural axes, one is related to activity, while the other to exploration. Further analyses suggest that activity is influenced by worker density in the nests. The separate analyses of the behavioural traits also support this effect of colony density.
We detected behavioural repeatability among colonies in the total number of entries and the latency of the first worker to return to the nest. Both of these features are likely to be tied to the successful foraging activity of colonies (according to Gordon, 1983 andByrne, 1994). Consistent with this prediction, between-colony differences in collective foraging activity have been linked with colony success in several species of ants (e.g. Kühbandner, Modlmeier, & Foitzik, 2014, Modlmeier & Foitzik, 2011, Gordon, 2013, Bengston, Shin, & Dornhaus, 2017, Carere, Audebrand, Rödel, & d'Ettorre, 2018. Carere et al. (2018) found similar associations between worker deployment and return rates in F. fusca. We failed to detect consistent among-colony differences in the latency of visitations to staged patches of sugar water or foreign objects. We predicted that worker-worker interactions outside of the nest would be an important determinant of colonies' exploratory tendencies and in the exchange of information about the foraging landscape (Gordon, 1996;Gordon & Mehdiabadi, 1999;Pacala et al., 1996). However, we failed to detect repeatability in any of our metrics of worker-worker interactions (latency to first interaction, total interactions). Moreover, neither of these interaction metrics were associated with colony size, nest density or original queen number. This was at odds with our basic prediction that colonies containing more workers or even denser colonies would be, by default, exhibiting a larger number of worker-worker interactions, as a matter of packing. Recent work on Camponotus pennsylvanicus (De Geer, 1773) revealed that colonies of ants can maintain homeostasis in their worker-worker interaction rates in spite of changes to nest density (Modlmeier et al., 2019), which potentially conveys that there is an ambient and optimal level of interaction rates necessary for colony functioning. The fact that worker-worker interactions were also insensitive to colony size and TA B L E 2 Eigenvector loadings, eigenvalues, and proportion and cumulative proportion of variance for the principal components retained (PCs 1-3) nest density implies that F. fusca can also hone in the level of interactions needed for running the colonies efficiently.
Alterations in group size can have a variety of impacts on the collective phenotypes. However, in colonies of Temnothorax albipennis (Curtis, 1854), for instance, collective decision-making was not influenced by experimental alterations in colony size (Dornhaus & Franks, 2006).
Similarly, findings from social spiders convey that colony size does not affect the exploratory behaviour in Stegodyphus dumicola Pocock, 1898 (Keiser & Pruitt, 2014). Yet, alterations in group size are obviously influential in other systems. The results of Ward, Herbert-Read, Sumpter, and Krause (2011), for example indicate that larger fish groups (eastern mosquitofish, Gambusia holbrooki Girard, 1859) are better and faster in decision-making. We found some support that alterations in nest density can impact aspects of colony exploration; this was true for the latency of the first worker to return to the nest and for the total number of the entries. Here, smaller colonies exhibited longer latencies (i.e., later returns) and they showed more entries. Overall the exploration of the F. fusca colonies was influenced by the density. Similar density-dependent behavioural differences have been found in colonies of the polydomous T. rugatulus, where more intense foraging and scouting activity was observed under high-density conditions (Cao, 2013). Similarly, increasing group size enhances nest construction rates in Lasius niger (Linnaeus, 1758) (Toffin, Kindekens, & Deneubourg, 2010), digging rates in some termites (Bardunias & Su, 2010)

and the speed of worker returns in
Myrmica punctiventris Roger, 1863 (Herbers & Choiniere, 1996). In aggregate, it seems that there are cases when group size can strongly influence group behaviour, and in other cases, group phenotypes appear entirely resilient to alteration in group size. For now, the reasons behind such case-by-case outcomes remain elusive.
At odds with the findings from our experimental colonies, we detected an influence of the size of the source colony on the exploratory behaviour of our experimental colonies. Similar effects have been detected before (e.g., T. albipennis) (Dornhaus & Franks, 2006

ACK N OWLED G EM ENT
We would like to thank Miklós Bán, Attila Fülöp and Enikő Gyuris for providing technical help and Ferenc Báthori for assistance in the field. We are grateful for the helpful comments of the editor and the referees, which greatly improved the manuscript. The publication was supported by the EFOP-3.6.1-16-2016-00022 project.
The project is co-financed by the European Union and the European Social Fund. AT was supported by two "Bolyai János" scholarships

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

E TH I C A L N OTE S
Ethics approval was not required for the study. We worked with an unprotected species collected from an unprotected site.