Perception of collective path use affects path selection in ants

Animal Behaviour 99 (2015) 15e24

Contents lists available at ScienceDirect

Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav

Perception of collective path use affects path selection in ants Tomer J. Czaczkes a, *, Sandra Franz b, Volker Witte b, Jürgen Heinze a a b

€t Regensburg, Regensburg, Germany Biologie I, Universita €t München, Planegg-Martinsried, Germany Biologie II, Ludwig-Maximilians Universita

a r t i c l e i n f o Article history: Received 30 July 2014 Initial acceptance 17 September 2014 Final acceptance 6 October 2014 Published online MS. number: 14-00619 Keywords: ants collective decision making cuticular hydrocarbons home range markings negative feedback positive feedback social cues U-turns

Social animals rely heavily on social cues to make foraging decisions. In social insects such as ants, perceived use of paths by nestmates is an important cue which allows ants to adjust foraging behaviours. Ants that encounter other ants on a trail reduce trail pheromone deposition. This has been predicted to allow ants to preferentially select underused paths, and thus avoid overcrowding. Here we tested this hypothesis by providing ants with two identical paths to a feeder. On the treatment path we placed 10 ‘dummy ants’, i.e. glass beads coated in nestmate cuticular hydrocarbons, and on the control path 10 untreated beads. Contrary to expectations, ant colonies preferentially chose the treatment path. This preference was unrelated to pheromone deposition, as it arose before any pheromone was deposited. Ants performed more U-turns on the control path, and thus were more likely to switch paths if they entered the control path. Path preference disappeared when three of the untreated beads on the control path were replaced by dummy ants, demonstrating that it is the perceived absence of nestmates on a path, not the relative path use, that drives colony-level preference. By preferentially using paths containing nestmates, ants may benefit from increased information transfer and recruitment potential. The presence of nestmates on a path coupled with a lack of alarm pheromones may be a ‘reassurance’ that the path is safe and productive. Although ants have various mechanisms for coping with trail crowding, they in fact prefer paths that are already in use. © 2014 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

Animals use a variety of information sources when foraging. For social animals signals and additional social cues are particularly important, and can provide group members with useful information such as food quality, safety and patch productivity (Galef & Giraldeau, 2001). Often, several sources of social information will be available. The situation can be complicated if one source of social information affects the perception, sharing or impact of other sources of information. For example, mate choice is often informed not only by direct signalling but also by other, often inadvertent, public cues, such as the decisions of other conspecifics (Mery et al., 2009; Vakirtzis, 2011). Many animals, both vertebrate and invertebrate, take public information as well as private information into account when deciding how to respond to risky situations (Coolen, Dangles, & Casas, 2005; Phelps, Rand, & Ryan, 2007). Different sources of information, both public and private, may be used hierarchically or may interact additively or synergistically. Understanding how different sources of information affect group decisions poses a major challenge to biologists and ecologists (Schmidt, Dall, & Van Gils, 2010).

€t Regensburg, Uni* Correspondence: T. J. Czaczkes, Biologie I, Universita €tsstraße 31, D-93053, Regensburg, Germany. versita E-mail address: [email protected] (T. J. Czaczkes).

Social insects, such as mass-recruiting ants and honeybees, rely heavily on signalling for recruitment and the organization of € lldobler & foraging (Detrain, Deneubourg, & Pasteels, 1999; Ho Wilson, 1990). In a classic ant foraging bout, scouts that find rewarding food sources return to the nest depositing pheromone trails along the substrate. These trails act as a signal to other ants that food has been found, and provide an orientation cue that can be followed to the food. These ants in turn return while depositing more pheromone, which forms a feedback loop of ever-increasing recruitment strength, until counteracted by overcrowding, satiation or food depletion (Wilson, 1962). When there are two paths to a food source, amplification of pheromone concentration on the paths by positive feedback usually results in one of the paths being used almost exclusively (Dussutour, Fourcassie, Helbing, & Deneubourg, 2004; Goss, Aron, Deneubourg, & Pasteels, 1989; Sumpter & Beekman, 2003). If the trails are identical, small initial differences in pheromone concentration seed this choice. However, if, for example, one of the paths is longer, positive feedback will work more slowly on it, resulting in an ant colony ‘selecting’ the shorter path (Goss et al., 1989). Likewise, if one of the paths is narrower and can only support a limited number of ants, the wider path will be selected (Dussutour et al., 2004). Lasius niger ants could avoid choosing the narrower trail because a higher ant density, and thus more crowding, forces other ants onto the wider one.

http://dx.doi.org/10.1016/j.anbehav.2014.10.014 0003-3472/© 2014 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

16

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24

However, such a mechanism may not function if the bottleneck is distant from the choice point. However, ants, like most animals, use multiple sources of information. Ants do not follow trail pheromones blindly. For example, they can extract information from the geometry of their trail network or the scent of their surroundings (Dupuy, Sandoz, Giurfa, & Josens, 2006; Forster et al., 2014; Jackson, Holcombe, & Ratnieks, 2004). They can also access many types of social information, such as the scent of food returned by foragers (Czaczkes, Schlosser, Heinze, & Witte, 2014) or the presence of home range markings deposited by nestmates (Devigne & Detrain, 2006; Devigne, Renon, & Detrain, 2004). The presence and behaviour of other individuals is an important source of information for many animals (Coolen et al., 2005; Phelps et al., 2007), and ants also make use of this information: interactions with nestmates can affect which tasks workers perform (Gordon & Mehdiabadi, 1999; Greene & Gordon, 2007) or the preference of workers for different food sources (F. Roces, personal communication). Collective decisions during trail foraging are not mediated solely by initial trail choice. U-turns can play a large role in collective decision making. For example, when presented with a long and a short path to a single food source, L. niger ants performed more Uturns on the longer path, which deviated more strongly from the straight nestefood line (Beckers, Deneubourg, & Goss, 1992a). This may strongly contribute to the ability of these ants to select the shorter of two routes. Similarly, Argentine ants, Linepithema humile, and Pharaoh's ants, Monomorium pharaonis, also perform more Uturns as they deviate more from their original path, a behaviour also implicated in their ability to choose the shortest path to a food re cheau, Combe, Fourcassie , & Theraulaz, source (Garnier, Gue 2009; Helanter€ a, Walsh, & Ratnieks, 2011). In the Argentine ant, U-turns can also act to strengthen a pheromone trail, by allowing an individual ant to repeatedly mark the trail in one journey (Reid, Latty, & Beekman, 2012). As mentioned above, collisions on a highly crowded trail may force ants to turn around and select a different , 2006). Ants path (Dussutour, Nicolis, Deneubourg, & Fourcassie will perform U-turns when stepping off a pheromone-marked path, presumably in order to relocate the path (Czaczkes, Grüter, Jones, & Ratnieks, 2011). Gaining information from one source can affect how information in another modality is deployed. For example, encounters with nestmates on a trail affect pheromone deposition rates. In two recent studies, Czaczkes, Grüter, and Ratnieks (2013a, 2014) described a negative feedback effect in the foraging system of L. niger. Ants that encountered other ants on a trail, or that encountered dummy ants composed of black beads coated in nestmate cuticular hydrocarbons, reduced the amount of trail pheromone they deposited. Ants that encountered just 10 such dummy ants were 45% less likely to deposit trail pheromone. This effect was postulated to have several roles, including preventing unnecessary use of trail pheromone and maintaining foraging flexibility. Another possible role for this mechanism could be to allow ant colonies to avoid recruiting to crowded trails or trails with a bottleneck somewhere along their length. It was postulated that on more crowded trails ants will experience more head-on encounters, and thus reduce their pheromone deposition. This would then seed a difference in pheromone levels between more and less crowded paths, resulting in more ants choosing the less restricted path. On the other hand, this effect would also predict that on nonlimited path systems with identical paths, such as perhaps those used in previous dual-path experiments (Dussutour et al., 2004; Goss et al., 1989), a uniform use of both paths would be observed, as the negative feedback effect first favours one and then the other path in a homeostatic manner. Such a pattern, however, is not observed. Rather, a single path is usually chosen.

In this series of experiments we explored how ants use information derived from encounter rates on a trail to make collective decisions. We also attempted to reconcile this discrepancy between theory and observation mentioned above. We experimentally manipulated perceived path use on path selection by ant colonies. By using a dual-path design with dummy ants on one path and unmarked beads on the other, we could test whether ant colonies preferentially used underused paths as predicted by the results of Czaczkes et al. (2013a), while keeping all other aspects of the paths identical. We used ant-mimicking beads to simulate a path in use, the precursor to a crowded trail, as opposed to simulating highly crowded paths. METHODS Study Species We studied six L. niger colonies collected on the LudwigMaximilian University Planegg campus, in Munich, Germany. Colonies were housed in plastic foraging boxes (40
30 cm and 20 cm high). The bottom of each box was covered with a layer of plaster of Paris. Each foraging box contained a circular plaster nestbox (14 cm diameter, 2 cm high). The colonies were queenless with 1000e2000 workers and small amounts of brood. Queenless colonies forage, make pheromone trails and care for brood, and are frequently used in foraging experiments (Czaczkes et al., 2013a; Devigne & Detrain, 2002; Dussutour et al., 2004; Evison, Petchey, Beckerman, & Ratnieks, 2008; Portha, Deneubourg, & Detrain, 2002). While being from a queenless colony may slightly affect the behaviour of the ants, this is unlikely to affect the result of this experiment, especially as brood was present in the nest (Herbers & Choiniere, 1996; Portha, Deneubourg, & Detrain, 2004). Colonies were fed three times per week with Bhatkar diet, a mixture of egg, agar, honey and vitamins (Bhatkar & Whitcomb, 1970). Colonies were deprived of food for 4 days prior to a trial in order to achieve uniform and high motivation for foraging. Water was provided ad libitum. Dummy Ant Preparation Dummy ants made by coating glass beads in cuticular hydrocarbons have been used in many studies of ant behaviour, to simulate both nestmates and non-nestmates (Akino, Yamamura, Wakamura, & Yamaoka, 2004; Czaczkes et al., 2013a; Michael J. Greene & Gordon, 2003; Ozaki et al., 2005). Thirty workers were removed from the colony to be tested, and chilled at 20  C for about 10 min. They were then placed in a 2 ml glass extraction vial (Sigma Aldrich) and covered in pentane to a level at least 1 mm above the ants. The vial was agitated for 5 min at 30  C so as to dissolve the cuticular hydrocarbons from the ants' cuticle. The ants were then removed from the pentaneeCHC solution, and 12 black glass beads (diameter 2.5 mm, height 1 mm; KnorrPrandell GmbH, Lichtenfels, Germany) were placed in the solution. The solution and beads were then agitated at 30  C until all the pentane had evaporated, thus coating the beads in CHCs. By then extracting the CHCs from the beads and comparing them to those of nestmates, we confirmed that the beads were indeed coated with nestmate CHCs, and that these CHCs were similar to those extracted directly from ants (see Appendix for details). Ethical Note The animals in this study are not subject to ethical standard laws in the country in which these experiments were performed. Moreover, there is little research into the humane treatment and

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24

euthanasia of insects (Gunkel & Lewbart, 2007). None the less, distress to the animals was minimized as much as possible: although colonies were deprived of food for 4 days prior to testing, this is well within what such colonies can tolerate before starvation begins. Ants from which CHCs were extracted were chilled at 20  C for 10 min before extraction, to minimize their distress. Marked ants to be discarded were frozen at 20  C.

17

and stepped onto one of the perpendicular paths of the maze was also noted. Any effects observable before this time cannot be due to trail pheromone effects, as unfed L. niger do not deposit pheromone trails (Beckers, Deneubourg, & Goss, 1992b; Mailleux, Buffin, Detrain, & Deneubourg, 2011). The number of U-turns performed until the first ant left the feeder and stepped onto one of the perpendicular paths of the maze was also noted. Lastly, we followed the first return trip of the marked ants from the feeder to the nest, and the horizontal path initially entered, the path from which the ant eventually left the maze and the number of U-turns performed on each path were noted. Five of the colonies were tested four times, twice with the control path on the left and twice with the control path on the right. One of the colonies was only tested three times. To ensure that the dummy ants were not attractive or repellent at a distance, a further experiment was performed. The experimental set-up was identical to that of experiment 1. However, all ants leaving the nest were collected as they stepped onto the perpendicular portion of a path. Thus, we effectively conducted a Ymaze experiment. Data collection began with the first ant to enter the maze, and stopped after 10 min or after so many ants had entered the maze that accurate data collection ceased to be possible. Four colonies were tested, two with the dummy beads on the right and two with the dummy beads on the left. In total 339 ants were tested. Ants showed no preference for the treatment path (Z ¼ 0.01, P ¼ 0.992), demonstrating that the ‘dummy ants’ were not attractive over a distance.

General Experimental Set-up Food-deprived colonies were given access to a maze with a large 1 M sucrose solution feeder at the opposite end. The maze consisted of two parallel paths, 10 mm wide and 200 mm long, joined at each end by a 90 Y-bifurcation with 40 mm long arms (see Fig. 1), thus forming a stretched hexagon. The maze and all other parts of the apparatus to which the ants had access were covered in disposable paper overlays. Equal numbers of black glass beads, either uncoated (control beads) or coated with nestmate CHCs (dummy ants), were placed on both paths. The first 10 ants to reach the feeder were marked with a dot of acrylic paint on the abdomen, so as to allow the behaviour of individual ants to be followed on their first return trip to the nest. The ants were then allowed to forage undisturbed for 1 h, during which the entire maze was filmed in high definition from above using a Sony Handycam CX190. From these videos we obtained information such as the number of ants on each path at various times, the time at which the first ants began returning to the nest and the number of U-turns. At the end of the experiment all marked ants were removed from the colony, the remaining ants returned to the nestbox, the maze cleaned with ethanol and the paper overlays replaced. Both control beads and dummy ants were discarded after every trial.

Experiment 2: Relative Number of Dummy Ants Present This experiment was conducted to examine whether it is the absolute lack of dummy ants driving the patterns found in experiment 1, or whether the patterns found are quantitatively related to the relative number of dummy ants on each path. The method was identical to that used in experiment 1, except that on the control path three of the untreated glass beads, at positions 1, 5 and 10 from the nest, were replaced with dummy ants. All six colonies were tested three times. Whether the treatment path was on the right or the left side was varied systematically.

Experiment 1: Dummy Ant Presence The aim of this experiment was to examine whether the presence of nestmates on a path, as simulated using dummy ants, affects collective path choice. On one path (designated treatment) 10 dummy ants were placed at 1 cm intervals (Fig. 1). On the other path 10 control beads were similarly placed. Lasius niger walk with their antennae spread about 4 mm apart and the beads were 2.5 mm wide. As a result, ants walking on the 10 mm wide walkway made antennal contact with most of the beads (mean 8.4 beads contacted per trip, SD 1.6). From the videos, the ants on each path were counted every minute. Data were collected blind, i.e. the person counting the ants on each path did not know which path contained the CHC-covered beads. To examine whether any behavioural influence of the treatment was due to differential trail pheromone deposition, the time until the first ant left the feeder

Experiment 3: Home Range Marking Presence Lasius niger ants passively deposit cuticular hydrocarbons as they walk over a substrate (Yamaoka & Akino, 1994). These are detected by ants visiting an area, and act as home range markings (HRMs), affecting ant behaviours including U-turning and pheromone deposition (Czaczkes, Grüter, Jones, & Ratnieks, 2012; Devigne & Detrain, 2002; Devigne et al., 2004). Although the

200 mm Feeder platform 40 mm To the nest

Black glass bead or dummy

10 mm

Position of dummy ant on control path in experiment 3

Figure 1. Experimental set-up: the stretched-hexagon maze, showing the position of beads placed upon it. In experiment 1 all the beads on one path were ‘dummy ants’ and all the beads on the other path were untreated control beads. In experiment 2 the red (grey in black-and-white images) beads at positions 1, 5 and 10 on the control path were dummy ants, and the other seven beads were untreated. The diagram, including the image of a L. niger worker at the entrance of the maze, is to scale.

18

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24

chemical nature of such HRMs has been reported to be different re, from that of normal cuticular hydrocarbons (Lenoir, Depicke s, & Detrain, 2009), it is conceivable that the Devers, Christide dummy ants in this experiment were being perceived not as fellow foragers, but as HRMs. To control for this, we tested the effect of HRMs on path choice. Paper overlays fitting the straight paths on the stretched-hexagon maze were left overnight in the foraging box of colonies. Ants explored these overlays, marking them with HRMs. This method has been previously used to acquire and test HRMs (Czaczkes et al., 2011; Devigne et al., 2004). The next day one path of the maze was covered with a marked paper overlay, and the other path covered with a fresh, unmarked overlay. Ants were allowed to forage via this maze for 30 min, and the ants on each path were counted every minute. Eight colonies were tested, and each colony was tested twice, with the HRMs being placed on alternating sides of the maze. Statistical Analyses Statistical analyses were carried out in R2.15.1 (R Core Team, 2012) using generalized linear mixed models (GLMMs; Bates, Sarkar, Bates, & Matrix, 2007). Following Forstmeier and Schielzeth (2011) we included in the tested models only factors and interactions for which we had a priori reasons for inclusion, which were typically path (treatment or control), time (1e60) and time period (first 10 min or last 50 min), and all the interactions between these factors. Time period was added as a factor since the number of ants on the paths increased for the first 10 min of the experiment, and then, owing to satiation of the colony, began to decrease slowly. Colony and trial were added as random effects, with trial effects being nested within colony effects, to control for repeated measures for each colony. We built models that predicted either the total number of ants on a path or the proportion of ants on a path, using the following model formulae: total ants ¼ time
path
time period þ random effects (experimentID/colony) and

Mean number of ants on each path

proportion of ants ¼ time
path
time period þ random effects (experimentID/colony)

To test whether the number of U-turns performed in the first few minutes of the experiment varied between experiments 1 and 2, the data from both were combined and the factor ‘experiment’ (0 versus 10 or 3 versus 10) was added. The number of U-turns performed was divided by the number of minutes until the first ant left the feeder to provide a U-turning rate per min. This results in the following model formula: U-turns/min ¼ path
experiment þ random effects (experimentID/colony) where ‘experiment’ is either 10 versus 0 or 10 versus 3, depending on how many dummy ants (0 or 3) were on the control path. When testing the attractiveness of the dummy beads at a distance, we simply tried to predict the number of ants choosing one direction according to where the beads were placed. Count data (the number of ants on each path) were modelled using a Poisson distribution. Proportion data (the proportion of ants on each path) were logit transformed following Shi, Sand Hu, and Xiao (2013), and then modelled using a Gaussian distribution. Binomial data (choice data at a bifurcation when testing for the attractiveness of dummy ants at a distance) were modelled using a binomial distribution with a logit link function. Interactions were explored using subsetting. For example, to explore an interaction between time and time period the data would be split into the two time periods (the first 10 min and the last 50 min), and then the effect of time tested in both these subsets. All P values are corrected using the BenjaminieHochberg method (Benjamini & Hochberg, 1995) to control for false discovery rates. RESULTS Experiment 1: Dummy Ant Presence Overall, there were many more ants on the treatment path than the control path (Fig. 2). The number of ants on both paths rose sharply in the first 10 min, and then dropped steadily over the following 50 min, as can be seen by a significant interaction between time period (first 10 min or last 50 min) and time (Z ¼ 19.56, P < 0.0001; Fig. 2). In the first 10 min the number of ants on both paths increased rapidly over time (Z ¼ 15.35, P < 0.0001), with the number of ants

Control path Treatment path

40

30

20

10

0 1

5

10

15

20

25

30

35

40

45

50

55

60

Time (min) Figure 2. Mean number of ants on each path every minute. Ten ‘dummy ants’ (black glass beads coated in nestmate cuticular hydrocarbons) were placed on the treatment path. Ten unmarked black glass beads were placed on the control path. Bars represent 95% confidence intervals for the mean.

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24

on the treatment path increasing much more rapidly than the number of ants on the control path (Z ¼ 5.57, P < 0.0001). This led to a nonsignificant trend for there to be more ants on the treatment path in the first 10 min (Z ¼ 1.76, P ¼ 0.078). In the last 50 min the number of ants on both paths dropped steadily over time (Z ¼ 35.96, P < 0.0001), and this rate was faster on the treatment path (Z ¼ 9.67, P < 0.0001), since the total number of ants at the start of this period was much higher (Fig. 2). None the less, over the last 50 min there were significantly more ants on the treatment path than the control path (Z ¼ 30.08, P < 0.0001). Mirroring the results above, the proportion of ants on the treatment side was higher than that on the control path (Fig. 3). In the first 10 min the proportion of ants on each path was higher on the treatment side (Z ¼ 3.77, P < 0.001). In the final 50 min the proportion of ants on each path was also higher on the treatment side (Z ¼ 14.027, P < 0.0001), and this pattern weakened slightly over time (Z ¼ 3.59, P < 0.001). Surprisingly, the preference of ants for the treatment path emerged even before the first ant had returned from the food source. Even when only considering the minutes before the first ant had left the feeder and stepped onto a straight section of one of the paths, there were more ants in total (Z ¼ 4.046, P < 0.0001) on the treatment side, which translates into a larger proportion of ants on the treatment side (Z ¼ 3.67, P < 0.001). To explore some possible explanations for the collective choice of the control path we tested whether the ‘dummy ants’ were attractive to ants over a distance, and whether the lack of ‘dummy ants’ resulted in ants performing more U-turns. U-turn rates before the first ant returns We found that U-turning rates were higher on the control path than the treatment path (Z ¼ 5.041, P < 0.0001; Fig. 4). This pattern was weaker but also present in experiment 2 (see below; Z ¼ 3.20, P ¼ 0.001). The interaction between path (control or treatment) and experiment (10 versus 0 and 10 versus 3) was nonsignificant (Z ¼ 1.44, P ¼ 0.15). Path choice and U-turn rates on first return Although more marked ants entered the treatment path than the control path first (84 and 66 ants, respectively), this difference was not significant (Z ¼ 1.42, P ¼ 0.156). We found that the proportion of ants for which the path first entered was also the path by which the ant left the maze was higher in ants that first chose the treatment path (Z ¼ 2.93, P ¼ 0.003). In other words, ants were

19

more likely to switch paths if they had entered the control path (29% of ants switched path if first entering the control path, compared to 10% of ants switching paths if first entering the treatment path). Experiment 2: Relative Number of Dummy Ants Present Again, the number of ants rose sharply in the first 10 min of the experiment and then dropped steadily in the last 50 min (Z ¼ 20.44, P < 0.0001; Fig. 5). In the first 10 min there were slightly more ants on the control path (Z ¼ 3.15, P ¼ 0.002). The rate at which the number of ants increased over time did not differ significantly between the two paths (Z ¼ 0.12, P ¼ 0.90). In the last 50 min there were more ants on the treatment path (Z ¼ 5.24, P < 0.0001), and the number of ants on the paths dropped slightly faster on the treatment path (Z ¼ 2.46, P ¼ 0.014). When considering the proportion of ants on each path, we found no significant effects of time period (Z¼ 0.51, P ¼ 0.73), time (Z ¼ 1.38, P ¼ 0.36), path type (Z ¼ 0.33, P ¼ 0.78) or any of their interactions (Z < 1.47, P > 0.36; Fig. 6). We found that U-turning rates were higher on the control path than the treatment path, but this effect was not as strong as that found in experiment 1 (Z ¼ 3.20, P ¼ 0.001). The interaction between path (control or treatment) and experiment (10 versus 0 and 10 versus 3) was not significant (Z ¼ 1.44, P ¼ 0.15). Experiment 3: Home Range Marking Presence The presence of HRMs had little or no effect on collective path choice. In terms of absolute number of ants on each path, we found no significant effect of path type alone (control or HRM marked; Z ¼ 0.025, P ¼ 0.98). There was, however, a significant interaction between path type and time period (first 10 min or last 20 min of the experiment; Z ¼ 2.37, P ¼ 0.029): although path type had no effect in the first 10 min (Z ¼ 0.025, P ¼ 0.98), slightly fewer ants were found on the HRM path in the last 20 min (Z ¼ 3.437, P < 0.001; Fig. 7). We found no significant effect of path type on the proportion of ants on either path (Z ¼ 1.66, P ¼ 0.193; Fig. 8). DISCUSSION The results from experiment 1 are clear and unexpected: ant colonies preferentially used paths on which workers sensed some

Mean proportion of ants on each path

0.9

Control path Treatment path

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 1

5

10

15

20

25

35 30 Time (min)

40

45

50

55

60

Figure 3. Mean proportion of ants on each path every minute. Ten ‘dummy ants’ (black glass beads coated in nestmate cuticular hydrocarbons) were placed on the treatment path. Ten unmarked black glass beads were placed on the control path. Bars represent 95% confidence intervals for the mean.

20

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24

U-turns/min before any ants left the feeder

NS 12

10 ** 8

***

6

4

2 Control Treatment 10 ‘dummy ants’ vs 10 unmarked beads

Control

Treatment

10 ‘dummy ants’ vs 7 unmarked beads & 3 ‘dummy ants’

Figure 4. U-turning rates of ants walking from the nest to the feeder in the period of time before the first ant left the feeder. As no ants had left the feeder, the higher Uturning rate on the control paths cannot be explained by different levels of trail pheromone. Symbols are means; bars are 95% confidence intervals. ***P < 0.001; **P < 0.01; NS: P > 0.1.

level of path use, and avoided using completely unused paths. This is the opposite of what was predicted from the results of Czaczkes et al. (2013a). The results of experiment 2 demonstrate that the preference for used paths was dependent on a qualitative difference in the number of perceived (dummy) ants. Under a quantitative difference, i.e. 10 versus 3 dummy ants, colonies' preference for the treatment path was weak or nonexistent. In experiment 1 colony-level path choice occurred very rapidly, within 3 min, and well before any reduction of pheromone deposition due to perceived path use could begin. This mirrors the path choice speed

reported by Dussutour et al. (2004). Since colony-level path selection was detectable before the first forager left the feeder, trail pheromones can be ruled out as the cause of path choice. Lasius niger workers do not deposit pheromone until they have sampled some food, either at a feeder or (when starved) via trophallaxis (Beckers et al., 1992b; Mailleux, 2006). Thus, the negative feedback effect described in Czaczkes et al. (2013a), in which head-on encounters reduce pheromone deposition and thereby cause ants to choose the less-used branch, does not come into effect until a choice is already made. The results can also not be explained by volatile attractiveness of the dummy ants, as we tested for such an effect, and found none. Visual attraction to one of the paths can also be ruled out, as in all experiments paths were visually identical. An alternative mechanism for path selection, for which we found strong support, is U-turning behaviour. More U-turns were performed on the path without dummy ants in the first few minutes of both experiments 1 and 2, before any ant had begun returning from the feeder. Furthermore, in experiment 1 ants on their first nestward trip performed more U-turns on the control path, and were more likely to switch paths if they initially entered the control path. It is perhaps not surprising that U-turns play a key role in path selection due to perceived path use, as they have been shown to play a key role in several collective decision mechanisms in ants (Beckers et al., 1992a; Dussutour et al., 2006; Reid et al., 2012). Although one information source may begin the decisionmaking process, other information sources may cause a decision to be maintained. In the current study, the initialization of colonylevel path choice seems to have been via the U-turning mechanism. However, two other mechanisms are likely to have played a role in the amplification of this path choice. First, as more ants began using the treatment path more trail pheromone would have been deposited on this path. This would then begin a positive feedback cycle, with more ants choosing the treatment path because it had higher trail pheromone levels, and so laying more pheromone on this path (Sumpter & Beekman, 2003). Another mechanism that is likely to amplify the initial path choice pattern is route memory. Although we could not follow individual ants over multiple foraging trips, it is likely that ants that have taken a particular path and were rewarded are likely to take the same path in the future (Aron, Pasteels, & Deneubourg, 1989; Grüter, Czaczkes, & Ratnieks, 2011; Harrison, Fewell, Stiller, & Breed, 1989). Increased U-turning resulting in path switching for ants that initially choose the control

Mean number of ants on each path

40 Control path Treatment path 30

20

10

0 1

5

10

15

20

25

30

35

40

45

50

55

60

Time (min) Figure 5. Mean number of ants on each path every minute. Ten ‘dummy ants’ (black glass beads coated in nestmate cuticular hydrocarbons) were placed on the treatment path. Seven unmarked black glass beads and three ‘dummy ants’ were placed on the control path. Bars represent 95% confidence intervals for the mean.

Mean proportion of ants on each path

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24

21

0.9

Control path Treatment path

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 1

5

10

15

20

25

30 35 Time (min)

40

45

50

55

60

Figure 6. Mean proportion of ants on each path every minute. Ten ‘dummy ants’ (black glass beads coated in nestmate cuticular hydrocarbons) were placed on the treatment path. Seven unmarked black glass beads and three ‘dummy ants’ were placed on the control path. Bars represent 95% confidence intervals for the mean.

Mean number of ants on each path

35

Control path Treatment path

30 25 20 15 10 5 0 1

5

10

15 20 Time (min)

25

30

Figure 7. Mean number of ants on each path every minute. The treatment path was marked with colony home range markings. The control path was unmarked. No beads were placed on the paths. Bars represent 95% confidence intervals for the mean.

using already used paths, the chance of monopolizing discovered food is increased. There may also be less danger from predators or parasitoids on occupied paths (Feener & Moss, 1990). Another possible explanation for the preference for occupied paths is the potential for information transfer. When outgoing ants encounter returning foragers, important information may be transferred. For example, in the leafcutter ant Atta colombica, when outgoing workers encounter returning workers carrying a leaf they are themselves more likely to find and collect leaves (Dussutour, , 2007; Farji-Brener et al., Beshers, Deneubourg, & Fourcassie 2010). Moreover, ants are more likely to collect the same type of leaf they encounter on their outward journey (Howard, Henneman, Cronin, Fox, & Hormiga, 1996; Roces, 1994.). Even in ants such as L. niger, which retrieve their food internally, food residues smeared on returning ants may provide information for outgoing ants (Breton & Fourcassie, 2004). Indeed, social facilitation, whereby animals encountering others performing a certain action are then also more likely to perform that action, has often been described in social insects and other animals (Gordon, 1991; Gordon &

0.9 Mean proportion of ants on each path

path will result in more ants completing their journey to the feeder via the treatment path. They would then be more likely to take that path in the future. As path-switching probability becomes lower, the more successful visits an ant makes to the feeder (Czaczkes, 2012); thus memory coupled with U-turning would act as a ratchet, strengthening the choice of the treatment path. We have strong evidence for the proximate mechanism underlying preferential selection of used paths in ants, but what are the ultimate reasons for this selection? Why would ants prefer to use paths that are already being used? There are several possibilities. One is that group foraging only works with a minimum number of foragers (Beckers, Goss, Deneubourg, & Pasteels, 1989; Mailleux, Deneubourg, & Detrain, 2003), and that recruitment is more likely to be successfully triggered on an already used path. Moreover, recruitment need not take place from the nest; ants scouting outside the nest or on a trail can be recruited, which speeds resource exploitation speed and thus improves the ability of ants to monopolize a food source (Czaczkes & Ratnieks, 2012; Flanagan, Pinter-Wollman, Moses, & Gordon, 2013). Thus, by preferentially

Control path Treatment path

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 1

5

10

15 20 Time (min)

25

30

Figure 8. Mean proportion of ants on each path every minute. The treatment path was marked with colony home range markings. The control path was unmarked. No beads were placed on the paths. Bars represent 95% confidence intervals for the mean.

22

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24

Mehdiabadi, 1999; Klotz, 1986; Tolman, 1964). However, in our experiment the ants were encountering dummy ants which neither performed any behaviours nor could provide any information about potential food sources. Why then would ants none the less prefer the path with the dummy ants? One reason is that the preference for already occupied trails is beneficial, as there is the chance of useful information being gleaned from other foragers. Another possibility is that the simple presence of ants is an important social cue and information source. The presence of other ants, and lack of alarm pheromone, acts as valuable information about the safety of a path. As such, the presence of other ants might act as a ‘reassurance’ that the path is suitable and safe. A similar ‘reassurance’ effect was described by Greene and Gordon (2003; 2007), in whose study a high enough rate of returning scouts (or glass beads coated in scout CHCs) was required to trigger daily foraging in harvester ants, Pogonomyrmex barbatus. A ‘reassurance’ effect has also already been described for ant trail pheromones, which affected walking speed and sinuosity, and also increased route memorization speed (Czaczkes et al., 2011; Czaczkes, Grüter, & Ratnieks, 2013b). Home range markings, cuticular hydrocarbons passively laid down by walking ants, can also act as an important social cue. Such markings may also have a ‘reassurance’ effect on foraging ants, affecting pheromone deposition behaviour (Czaczkes et al., 2012; Devigne & Detrain, 2002; Devigne et al., 2004). The preference for foraging where others are already foraging is well documented in both vertebrates and invertebrates (Krebs, 1974; Lihoreau, Deneubourg, & Rivault, 2010; Ryer & Olla, 1995). For example, cockroaches, Blattella germanica, are attracted to odour cues of foraging conspecifics, and are more likely to remain at a food source when more conspecifics are present (Lihoreau et al., 2010; Lihoreau & Rivault, 2011). Such aggregations are thought to be of benefit to the individuals taking part for similar reasons as those discussed above, namely, information transfer (the location of resources worth exploiting due to quality or safety) and defence (increased alertness provided by many watchers, the ‘many eyes’ hypothesis, Lima, 1995). Moreover, the likelihood of an individual being killed by a predator is greatly reduced when part of a large aggregation (the ‘selfish herd’ hypothesis, Davies, Krebs, & West, 2012; Hamilton, 1971). This explanation does not, however, apply for eusocial insects since, for an individual worker, being killed or a sterile sister being killed would have an equal negative impact. The specific collective behaviour triggered en route to a food source that we describe here may well also take place in other path-using central-place foragers, such as termites. In this study we found little or no effect of HRMs on path selection. This is in spite of the fact that HRMs have been reported to have a role in catalysing recruitment by increasing pheromone deposition (Devigne et al., 2004), as well as in improving orientation, modulating recruitment behaviour and increasing walking speed (Detrain & Deneubourg, 2009). It seems that in this study, the ants responded to direct cues (contacting nestmates) but not indirect cues (HRMs). Direct cues may be more beneficial for tracking or responding to short-term changes (Czaczkes, Grüter, et al., 2014; Detrain & Deneubourg, 2009). HRMs are long-lived compared with trail pheromones, and thus risk providing outdated information. When deciding whether to continue following a path, ants may be more interested in current information (e.g. indicating that the branch leads to a food source that is currently productive) rather than long-term information (e.g. the branch led to a food source that was productive in the past). None the less, the apparent discrepancy in the HRM effect reported here and elsewhere (Devigne & Detrain, 2002, 2006; Devigne et al., 2004) is intriguing, and deserves further attention. The somewhat conflicting findings may simply be due to differences in experimental set-up. However,

they may also be due to differences in the motivational state of the foragers (Czaczkes, Schlosser, et al., 2014). For example, ants constrained to paths or branches may ignore HRMs, whereas ants in open spaces may respond to them. Many laboratory studies have examined ant foraging under overcrowded conditions, and have described adaptations for avoiding overcrowding or its negative effects (Burd & Aranwela, 2003; Dussutour et al., 2007; Dussutour, Beshers, Deneubourg, & , 2009; Dussutour et al., 2004, 2006; Grüter et al., Fourcassie 2012). However, in the current study our dummy ants simulated a path that was used, but not crowded; the dummy ants barely hindered the movement of foragers on the paths. Our results show that path use under uncrowded conditions provides important information and strongly influences collective foraging before overcrowding occurs, and that in fact ants preferentially make use of paths that are already in use. Social cues, such as the presence of nestmates on a trail, may have wide-reaching implications not only for path use and pheromone deposition behaviour (Czaczkes et al., 2013a), but also for other aspects of foraging such as navigation and route memory formation. Acknowledgments Many thanks to Professor Herwig Stibor for providing a suitable room in which to conduct our experiments, and to the anonymous referees for comments on the manuscript. T.J.C was funded by an Alexander von Humboldt postdoctoral research fellowship. References Akino, T., Yamamura, K., Wakamura, S., & Yamaoka, R. (2004). Direct behavioral evidence for hydrocarbons as nestmate recognition cues in Formica japonica (Hymenoptera: Formicidae). Applied Entomology and Zoology, 39(3), 381e387. Aron, S., Pasteels, J. M., & Deneubourg, J. L. (1989). Trail-laying behaviour during exploratory recruitment in the argentine ant, Iridomyrmex humilis (Mayr). Biology of Behaviour, 14(3), 207e217. Bates, D., Sarkar, D., Bates, M. D., & Matrix, L. T. (2007). The lme4 package. Linear mixed-effects models using S, 4. Retrieved from http://cran.r-project.org/web/ packages/lme4/index.html. Beckers, R., Deneubourg, J., & Goss, S. (1992a). Trails and U-turns in the selection of a path by the ant Lasius niger. Journal of Theoretical Biology, 159(4), 397e475. Beckers, R., Deneubourg, J., & Goss, S. (1992b). Trail laying behaviour during food recruitment in the ant Lasius niger (L.). Insectes Sociaux, 39, 59e71. Beckers, R., Goss, S., Deneubourg, J. L., & Pasteels, J. M. (1989). Colony Size, communication and ant foraging strategy. Psyche: A Journal of Entomology, 96(3e4), 239e256. Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple resting. Journal of the Royal Statistical Society. Series B (Methodological), 57(1), 289e300. Bhatkar, A., & Whitcomb, W. H. (1970). Artificial diet for rearing various species of ants. The Florida Entomologist, 53(4), 229e232. Breton, J., & Fourcassie, V. (2004). Information transfer during recruitment in the ant Lasius niger L. (Hymenoptera: Formicidae). Behavioral Ecology and Sociobiology, 55, 242e250. Burd, M., & Aranwela, N. (2003). Head-on encounter rates and walking speed of foragers in leaf-cutting ant traffic. Insectes Sociaux, 50, 3e8. Coolen, I., Dangles, O., & Casas, J. (2005). Social learning in noncolonial insects? Current Biology, 15(21), 1931e1935. Czaczkes, T. J. (2012). Organisation of ant foraging (Doctoral thesis). Sussex, U.K: University of Sussex. Czaczkes, T. J., Grüter, C., Jones, S. M., & Ratnieks, F. L. W. (2011). Synergy between social and private information increases foraging efficiency in ants. Biology Letters, 7(4), 521e524. Czaczkes, T. J., Grüter, C., Jones, S. M., & Ratnieks, F. L. W. (2012). Uncovering the complexity of ant foraging trails. Communicative & Integrative Biology, 5(1), 78e80. Czaczkes, T. J., Grüter, C., & Ratnieks, F. L. W. (2013a). Negative feedback in ants: crowding results in less trail pheromone deposition. Journal of the Royal Society Interface, 10(81). Czaczkes, T. J., Grüter, C., & Ratnieks, F. L. W. (2013b). Ant foraging on complex trails: route learning and the role of trail pheromones in Lasius niger. The Journal of Experimental Biology, 216, 188e197. Czaczkes, T. J., Grüter, C., & Ratnieks, F. L. W. (2014). Rapid up- and down regulation of pheromone signalling due to trail crowding in the ant Lasius niger. Behaviour, 151(5), 669e682.

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24 Czaczkes, T. J., & Ratnieks, F. L. W. (2012). Pheromone trails in the Brazilian ant Pheidole oxyops: extreme properties and dual recruitment action. Behavioral Ecology and Sociobiology, 66(8), 1149e1156. Czaczkes, T. J., Schlosser, L., Heinze, J., & Witte, V. (2014). Ants use directionless odour cues to recall odour-associated locations. Behavioral Ecology and Sociobiology, 68(6), 981e988. Davies, N. B., Krebs, J. R., & West, S. A. (2012). An introduction to behavioural ecology. Hoboken, NJ: Wiley-Blackwell. Detrain, C., & Deneubourg, J. L. (2009). Social cues and adaptive foraging strategies in ants. In S. Jarua, & M. Hrncir (Eds.), Food exploitation by social insects (pp. 29e52). Boca Raton, FL: CRC Press. Detrain, C., Deneubourg, J. L., & Pasteels, J. M. (1999). Information processing in social insects. Basel, Switzerland: Birkh€ auser. Devigne, C., & Detrain, C. (2002). Collective exploration and area marking in the ant Lasius niger. Insectes Sociaux, 49, 357e362. Devigne, C., & Detrain, C. (2006). How does food distance influence foraging in the ant Lasius niger: the importance of home-range marking. Insectes Sociaux, 53(1), 46e55. Devigne, C., Renon, A., & Detrain, C. (2004). Out of sight but not out of mind: modulation of recruitment according to home range marking in ants. Animal Behaviour, 67(6), 1023e1029. Dupuy, F., Sandoz, J.-C., Giurfa, M., & Josens, R. (2006). Individual olfactory learning in Camponotus ants. Animal Behaviour, 72(5), 1081e1091. , V. (2007). Crowding inDussutour, A., Beshers, S., Deneubourg, J.-L., & Fourcassie creases foraging efficiency in the leaf-cutting ant Atta colombica. Insectes Sociaux, 54(2), 158e165. , V. (2009). Priority rules Dussutour, A., Beshers, S., Deneubourg, J. L., & Fourcassie govern the organization of traffic on foraging trails under crowding conditions in the leaf-cutting ant Atta colombica. Journal of Experimental Biology, 212(4), 499e505. Dussutour, A., Fourcassie, V., Helbing, D., & Deneubourg, J.-L. (2004). Optimal traffic organization in ants under crowded conditions. Nature, 428(6978), 70e73. , V. (2006). Collective Dussutour, A., Nicolis, S. C., Deneubourg, J.-L., & Fourcassie decisions in ants when foraging under crowded conditions. Behavioral Ecology and Sociobiology, 61(1), 17e30. Evison, S. E. F., Petchey, O. L., Beckerman, A. P., & Ratnieks, F. L. W. (2008). Combined use of pheromone trails and visual landmarks by the common garden ant Lasius niger. Behavioral Ecology and Sociobiology, 63, 261e267. Farji-Brener, A. G., Amador-Vargas, S., Chinchilla, F., Escobar, S., Cabrera, S., Herrera, M. I., et al. (2010). Information transfer in head-on encounters between leaf-cutting ant workers: food, trail condition or orientation cues? Animal Behaviour, 79(2), 343e349. Feener, D., & Moss, K. (1990). Defence against parasites by hitchhikers in leafcutting ants: a quantitative assessment. Behavioral Ecology and Sociobiology, 26(1), 17e29. Flanagan, T. P., Pinter-Wollman, N. M., Moses, M. E., & Gordon, D. M. (2013). Fast and flexible: Argentine ants recruit from nearby trails. PLoS One, 8(8), e70888. Forster, A., Czaczkes, T. J., Warner, E., Woodall, T., Martin, E., & Ratnieks, F. L. W. (2014). Effect of trail bifurcation asymmetry and pheromone presence or absence on trail choice by Lasius niger ants. Ethology, 120(8), 768e771. Forstmeier, W., & Schielzeth, H. (2011). Cryptic multiple hypotheses testing in linear models: overestimated effect sizes and the winner's curse. Behavioral Ecology and Sociobiology, 65(1), 47e55. Galef, B. G., & Giraldeau, L.-A. (2001). Social influences on foraging in vertebrates: causal mechanisms and adaptive functions. Animal Behaviour, 61(1), 3e15. re cheau, A., Combe, M., Fourcassie , V., & Theraulaz, G. (2009). Path Garnier, S., Gue selection and foraging efficiency in Argentine ant transport networks. Behavioral Ecology and Sociobiology, 63(8), 1167e1179. Gordon, D. M. (1991). Behavioral flexibility and the foraging ecology of seed-eating ants. The American Naturalist, 138(2), 379e411. Gordon, D. M., & Mehdiabadi, N. J. (1999). Encounter rate and task allocation in harvester ants. Behavioral Ecology and Sociobiology, 45, 370e377. Goss, S., Aron, S., Deneubourg, J. L., & Pasteels, J. M. (1989). Self-organized shortcuts in the Argentine ant. Naturwissenschaften, 76, 579e581. Greene, M. J., & Gordon, D. M. (2003). Social insects: cuticular hydrocarbons inform task decisions. Nature, 423, 32. Greene, M. J., & Gordon, D. M. (2007). Interaction rate informs harvester ant task decisions. Behavioral Ecology, 18(2), 451. Grüter, C., Czaczkes, T. J., & Ratnieks, F. L. W. (2011). Decision making in ant foragers (Lasius niger) facing conflicting private and social information. Behavioral Ecology and Sociobiology, 64, 141e148. Grüter, C., Schürch, R., Czaczkes, T. J., Taylor, K., Durance, T., Jones, S. M., et al. (2012). Negative feedback enables fast and flexible collective decision-making in ants. Public Library of Science One, 7(9), e44501. Gunkel, C., & Lewbart, G. A. (2007). 13. Invertebrates. In G. West, D. Heard, & N. Caulkett (Eds.), Zoo animal & wildlife immobilization and anesthesia (pp. 147e158). Oxford, U.K.: Blackwell. Hamilton, W. D. (1971). Geometry for the selfish herd. Journal of Theoretical Biology, 31(2), 295e311. Harrison, J. F., Fewell, J. H., Stiller, T. M., & Breed, M. D. (1989). Effects of experience on use of orientation cues in the giant tropical ant. Animal Behaviour, 37(5), 869e871. Helanter€ a, H., Walsh, C. J., & Ratnieks, F. L. W. (2011). Effect of foraging trail straightness on U-turning probability in forager Pharaoh's ants. Sociobiology, 57(3), 445e449.

23

Herbers, J., & Choiniere, E. (1996). Foraging behaviour and colony structure in ants. Animal Behaviour, 51, 141e153. €lldobler, B., & Wilson, E. O. (1990). The Ants. Berlin, Germany: Springer Verlag. Ho Howard, J. J., Henneman, L. M., Cronin, G., Fox, J. A., & Hormiga, G. (1996). Conditioning of scouts and recruits during foraging by a leaf-cutting ant, Atta colombica. Animal Behaviour, 52(2), 299e306. Jackson, D. E., Holcombe, M., & Ratnieks, F. L. W. (2004). Trail geometry gives polarity to ant foraging networks. Nature, 432(7019), 907e909. Klotz, J. H. (1986). Social facilitation among digging ants (Formica subsericea). Journal of the Kansas Entomological Society, 59(3), 537e541. Krebs, J. R. (1974). Colonial nesting and social feeding as strategies for exploiting food resources in the Great Blue Heron (Ardea herodias). Behaviour, 51(1e2), 99e134. re, S., Devers, S., Christide s, J.-P., & Detrain, C. (2009). HydroLenoir, A., Depicke carbons in the ant Lasius niger: from the cuticle to the nest and home range marking. Journal of Chemical Ecology, 35(8), 913e921. Lihoreau, M., Deneubourg, J.-L., & Rivault, C. (2010). Collective foraging decision in a gregarious insect. Behavioral Ecology and Sociobiology, 64(10), 1577e1587. Lihoreau, M., & Rivault, C. (2011). Local enhancement promotes cockroach feeding aggregations. PLoS One, 6(7), e22048. Lima, S. L. (1995). Back to the basics of anti-predatory vigilance: the group-size effect. Animal Behaviour, 49(1), 11e20. Mailleux, A.-C. (2006). Starvation drives a threshold triggering communication. Journal of Experimental Biology, 209, 4224e4229. Mailleux, A.-C., Buffin, A., Detrain, C., & Deneubourg, J.-L. (2011). Recruitment in starved nests: the role of direct and indirect interactions between scouts and nestmates in the ant Lasius niger. Insectes Sociaux, 58, 559e567. Mailleux, A.-C., Deneubourg, J.-L., & Detrain, C. (2003). How does colony growth influence communication in ants? Insectes Sociaux, 50(1), 24e31.  Blanchet, S., Parejo, D., Coolen, I., et al. (2009). Mery, F., Varela, S. A. M., Danchin, E., Public versus personal information for mate copying in an invertebrate. Current Biology, 19(9), 730e734. Ozaki, M., Wada-Katsumata, A., Fujikawa, K., Iwasaki, M., Yokohari, F., Satoji, Y., et al. (2005). Ant nestmate and non-nestmate discrimination by a chemosensory sensillum. Science, 309, 311e314. Phelps, S. M., Rand, A. S., & Ryan, M. J. (2007). The mixed-species chorus as public information: túngara frogs eavesdrop on a heterospecific. Behavioral Ecology, 18(1), 108e114. Portha, S., Deneubourg, J., & Detrain, C. (2002). Self-organized asymmetries in ant foraging: a functional response to food type and colony needs. Behavioral Ecology, 13(6), 776e781. Portha, S., Deneubourg, J.-L., & Detrain, C. (2004). How food type and brood influence foraging decisions of Lasius niger scouts. Animal Behaviour, 68(1), 115e122. R Core Team. (2012). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.Rproject.org. Reid, C. R., Latty, T., & Beekman, M. (2012). Making a trail: informed Argentine ants lead colony to the best food by U-turning coupled with enhanced pheromone laying. Animal Behaviour, 84(6), 1579e1587. Roces, F. (1994). Odour learning and decision-making during food collection in the leaf-cutting ant Acromyrmex lundi. Insectes Sociaux, 41(3), 235e239. Ryer, C. H., & Olla, B. L. (1995). Influences of food distribution on fish foraging behaviour. Animal Behaviour, 49(2), 411e418. Schmidt, K. A., Dall, S. R. X., & Van Gils, J. A. (2010). The ecology of information: an overview on the ecological significance of making informed decisions. Oikos, 119(2), 304e316. Shi, P. J., Sand Hu, H. S., & Xiao, H. J. (2013). Logistic regression is a better method of analysis than linear regression of arcsine square root transformed proportional diapause data of Pieris melete (Lepidoptera: Pieridae). Florida Entomologist, 96(3), 1183e1185. Sumpter, D. J. T., & Beekman, M. (2003). From nonlinearity to optimality: pheromone trail foraging by ants. Animal Behaviour, 66(2), 273e280. Tolman, C. W. (1964). Social facilitation of feeding behaviour in the domestic chick. Animal Behaviour, 12(2e3), 245e251. Vakirtzis, A. (2011). Mate choice copying and nonindependent mate choice: a critical review. Annales Zoologici Fennici, 48, 91e107. €re Kohlenwasserstoff-Profile bei Wenczel, F. (2012). Saisonale einflüsse auf kutikula Lasius Arten (Hymenoptera: Formicidae) [The influence of season on the cuticular hydrocarbon profile of Lasius species] (Bachelors thesis). München: Ludwig €t. Maximillian Universita Wilson, E. O. (1962). Chemical communication among workers of the fire ant Solenopsis saevissima (Fr. Smith) 1. The organization of mass-foraging. Animal Behaviour, 10(1e2), 134e147. Yamaoka, R., & Akino, T. (1994). Ecological importance of cuticular hydrocarbons secreted from the tarsus of ants. In A. Lenoir, G. Arnold, & M. Lepage (Eds.), Les  Paris-Nord. Insectes Sociaux (p. 222). Paris, France: Universite

APPENDIX. THE SIMILARITY BETWEEN ‘DUMMY ANTS’ AND REAL ANTS This experiment was conducted so as to calibrate the number of ants from which cuticular hydrocarbons (CHCs) must be extracted, in order to produce 12 realistic ‘dummy ants’. First, the amount and

T. J. Czaczkes et al. / Animal Behaviour 99 (2015) 15e24

nature of the CHCs on real L. niger workers was measured. Next, various sets of dummy ants were created, using varying numbers of workers. The amount and nature of the CHCs on these dummy ants was then measured. Lastly, the values from real ants and dummy ants were compared, controlling for the difference in surface area of the ants and dummy ants. Measuring CHCs on real L. niger foragers Two nestmate workers were chilled for about 5 min at 20  C, placed in an extraction vial with 200 ml pentane, and agitated for 5 min. The ants were then removed, the pentane/CHC solution transferred to an autosampler vial (300 ml, Chromacol, 03-FISV) and the pentane allowed to evaporate. CHCs, being of high molecular weight, remained in the vial. We added 20 ml hexane containing a stable isotope internal standard (C36D66, 4 mg per 100 ml) to the autosampler vial, and the vial was agitated for 2 min. Samples were then analysed in a GCMS (see below for details). Five pairs of ants from three colonies were tested. Creating dummy ants Dummy ants were created by coating black glass beads (2.5 mm diameter, 1 mm height, KnorrPrandell GmbH) in cuticular hydrocarbons from nestmate ants. All beads were cleaned with pentane and then heated to 400  C for 2 h prior to use. To ascertain how many ants would be needed to cover 12 glass beads (10 for the trial and two spares) in a reasonable amount of CHCs three different quantities were tested. An extraction vial containing 20, 30 or 40 frozen workers each was filled with enough pentane to sufficiently cover the ants. The vials were then gently agitated for 5 min to dissolve the CHCs. After removing the ants, 12 glass beads were added to each vial. The pentane was then completely evaporated while being agitated at 30  C. Once all the pentane had evaporated, the quantity of CHCs on the beads was measured using an identical procedure to the one used to quantify CHC quantities on worker ants described above. A total of 18 measurements, six measurements per colony for three colonies, were made. For the behavioural experiments beads were created in an identical manner to that described above, using 30 ants per 12 glass beads (see results below), or using 40 ants per 16 glass beads. GCMS parameters and data collection We used a 6890N GC System equipped with a Restek Rxi-5MS column (30 m length, 0.25 mm ID, 0.25 mm film thickness) combined with a 5975 Mass Selective Detector (Agilent Technologies Inc., Apple Valley, MN, U.S.A.) operated using ChemStation software, version E.02.00.493. Injection was splitless over 1.0 min at 300  C under a pressure pulse of 16 psi for 0.5 min followed by

automatic flow control at 1.0 ml/min with helium as carrier gas. The oven program started isothermal at 120  C for 1 min, then increased rapidly by 20  C/min until 260  C, followed by a gentle temperature ramp of 3  C/min until 315  C and finally stayed isothermal for 5 min. The transfer line was held constantly at 320  C. A range of 50e500 amu was scanned after initial solvent delay of 3.8 min. Compounds were identified by their mass spectra and by retention time. Only typical CHCs were included in the analysis while contaminations were excluded. Quantities were calculated by integrating all peak areas and expressing each peak area relative to the internal standard. Results The surface area of a L. niger worker is different from the glass beads used in this experiment. An average worker has a total body surface of 9.8 mm2 (Wenczel, 2012). The total surface area of a glass bead was 19.24 mm2. We then compared the relative abundance of CHCs found per unit area on the beads made from the CHCs of 20, 30 and 40 ants with the relative abundance of CHCs found on worker ants. The relative abundance of CHCs on worker ants was greater than the CHC abundance on any of the beads (see Fig. A1). The beads created from the CHCs of 20 ants had a lower relative abundance of CHCs than the beads created using 30 or 40 ants (KruskaleWallis test: H ¼ 12.02, P ¼ 0.002). However, unexpectedly, the beads made from the CHCs of 30 ants had higher CHC abundances than those created using 40 ants (H ¼ 5.77, P ¼ 0.016). We thus chose to use the CHCs from 30 ants per batch of 12 beads, as this produced the beads with the largest CHC abundance and only required an intermediate number of ants to be killed per trial.

Relative CHC abundance (106) per individual per mm2

24

200

150

100 d 50 a

b

c

20

30

40

0 Directly measured ants

Number of ants used to create 12 'dummy ants' Figure A1. Relative abundance of cuticular hydrocarbons (CHC) measured from beads created using 20, 30 or 40 ants or measured directly from ants. Different letters indicate significant differences (P < 0.05). Boxes are interquartile ranges, horizontal lines are medians, circles are means and vertical lines are ranges.