This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Transcription

1 This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier s archiving and manuscript policies are encouraged to visit:

2 Animal Behaviour 84 (212) 1579e1587 Contents lists available at SciVerse ScienceDirect Animal Behaviour journal homepage: Making a trail: informed Argentine ants lead colony to the best food by U-turning coupled with enhanced pheromone laying Chris R. Reid a,b, *, Tanya Latty a,b, Madeleine Beekman a,b a Behaviour and Genetics of Social Insects Lab, School of Biological Sciences A12, University of Sydney, Camperdown, Australia b Centre for Mathematical Biology, University of Sydney, Camperdown, Australia article info Article history: Received 26 July 212 acceptance 4 September 212 acceptance 28 September 212 Available online 29 October 212 MS. number: R Keywords: collective behaviour decision making foraging Linepithema humile pheromone trail-laying ant The mechanisms by which self-organized systems like an ant colony reach collective decisions are poorly understood. Models explaining how trail-laying ants select the best food source and adapt to changes in their foraging conditions are often based on mechanisms that are either unrealistic or overly simplified. We studied the individual-level behaviour of the mass-recruiting Argentine ant, Linepithema humile, to determine the mechanisms that underlie the colony s ability to choose collectively the better of two food sources of different quality. Ants returning from a high-quality food source were more likely to U-turn back to it than ants returning from a lower quality food source. Those U-turning ants were also more likely to lay pheromone, and deposited pheromone at a higher rate than nonturning ants. We show that U-turning is pivotal to the rapid establishment of a strong pheromone trail to high-quality food sources, and that the trail is constructed from food to nest, not from nest to food. The pheromone-boosting U-turns were performed only during the initial period when the trail was being established. Once a trail was established, the trail was maintained by nonturning ants. U-turners appear vital in allowing a mass-recruiting colony to allocate workers rapidly to the food source of highest quality. Ó 212 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. Self-organized biological systems, such as flocks of starlings (Ballerini et al. 28), shoals of fish (Herbert-Read et al. 211; Katz et al. 211), swarms of honeybees (Camazine et al. 1999; Seeley 21; Diwold et al. 211) plagues of locusts (Bazazi et al. 28) and even slime moulds (Garfinkel 1987), make decisions in the absence of central control, and without any individual in the group possessing global information (Couzin 29). Animals living in large groups often engage in complex behaviour governed by the emergent collective intelligence of groups of simple agents (Bonabeau et al. 1999, page xi). For group-living animals to perform complex tasks such as collective decision making, individuals must develop behavioural mechanisms that allow those in possession of new information to elicit appropriate changes in the behaviour of the group (Edelstein-Keshet 1994; Bonabeau 1997). Perhaps the most commonly studied collective decision-making systems are the social insects, in particular the ants. Many species of ants use pheromones to coordinate foraging (Robertson et al. 198; Hölldobler & Wilson 199). Pheromones are mixtures of volatile compounds deposited on the substrate by individuals in response to certain stimuli. Pheromone trails are usually attractive to * Correspondence: C. R. Reid, School of Biological Sciences, University of Sydney, Camperdown, NSW 26, Australia. address: christopher.reid@sydney.edu.au (C. R. Reid). conspecifics, and are used to recruit workers to forage when the food source found is of high quality and too large to be exploited by a single individual (Hölldobler & Wilson 199). The volatile nature of pheromones means the trail evaporates and thus requires regular reinforcement (Goss et al. 1989; Beckers et al. 199). In mass-recruiting ants, workers not only follow trails, but also reinforce them as they travel by laying more of the attractive pheromone for other ants to follow and repeat the process (Hölldobler & Wilson 199; Bonabeau et al. 1998). This cycle of positive feedback allows the rapid build-up of foragers at profitable food sources. Positive feedback is often cited as the key factor in a colony s ability to focus on a single food source (Pasteels et al. 1987; Beckers et al. 199; Bonabeau et al. 1998). Albeit effective for problem solving in environments in which food sources do not change in time or space, positive feedback may, in theory, prevent an insect colony from responding appropriately to dynamic changes in its environment. Both earlier mathematical models (Goss et al. 1989; Nicolis & Deneubourg 1999; Camazine et al. 21) and experimental studies (Beckers et al. 1992; Sumpter & Beekman 23) suggest that once a trail is established, ants that rely solely on pheromone trails may be unable to reallocate their workforce if a better food source becomes available. This is because new trails cannot compete with the established trail, even if the new trail leads to a higher quality food source. Thus, the ants will be locked in to their original trail and be unable to adapt to subsequent /$38. Ó 212 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

3 158 C. R. Reid et al. / Animal Behaviour 84 (212) 1579e1587 changes in their environment. In the aforementioned mathematical models, ants become locked in because each ant is assumed to behave in exactly the same way, environmental conditions are static, and the ant s foraging behaviour is governed by differences in trail pheromone concentration only. In contrast, more recent empirical studies have shown that mass-recruiting ants are often highly successful at adapting their foraging behaviour when conditions change, and are far from blind slaves to pheromone trails (Dussutour et al. 29a, b; Reid et al. 211). Thus, the behaviour of individual ants must be more sophisticated than previously assumed. In response to recent empirical results suggesting more flexible behaviour of trail-laying ants, two mathematical models have explored ways by which ants can adapt to changes in their foraging environment. Ramsch et al. (212) explored possible means by which ants construct new trails when existing trails leading to food sources become blocked. Their model suggests that when ants know in which direction they are travelling, they are quickly able to establish alternative trails connecting the food and the nest, even if this means that they temporarily move away from areas with the highest levels of pheromone. In effect the ants require a very basic compass system, in which they need discern only whether they are travelling towards or away from the nest, to adapt quickly to a dynamic environment. A second model by Jackson et al. (211) incorporates observations that some Pharaoh s ants, Monomorium pharaonis, consistently perform U-turns when returning from a food source (Hart & Jackson 26). By comparing in silico colonies that either did or did not contain a small proportion of U-turning ants (found to be 7% of foragers in M. pharaonis, Jackson et al. 24), Jackson et al. (211) showed that the presence of U-turners allowed colonies to exploit newly discovered food sources rapidly. They therefore concluded that U-turning ants could play an important role in species that rely on mass recruitment. We studied the behaviour of individual Argentine ants, Linepithema humile, during foraging. Our main aim was to elucidate what individual behaviour allows the colony to focus on food of high quality. We compared ants when foraging at high- or lowquality food sources, which were offered simultaneously in a Y-maze. We specifically examined the role of U-turning ants, those ants that on their return journey to the nest after feeding turn back towards the food source, in collective decision making. We compared, over the course of 1 h, the relative contribution to pheromone deposition of ants that did or did not perform U-turns after feeding and between those that U-turned after having fed on food of different quality. METHODS General Procedures We collected Argentine ants from the grounds of the University of Sydney, New South Wales, Australia, and housed them in 3 2 cm plastic boxes with Fluon-coated sides to prevent escape. Up to 2 cardboard-wrapped test-tubes, partially filled with water and cotton wool plugs, were placed in each box. Colonies were fed with a vitamin-enriched artificial diet (Dussutour & Simpson 28) twice per week, supplemented with mealworms. Subcolonies were constructed from the pool of collected ants for all experiments. Experiments were conducted in July 211. All statistical tests were carried out using JMP v8..1 (SAS Institute Inc., Cary, NC, U.S.A.) and IBM SPSS Statistics v19 (SPSS Inc., Chicago, IL, U.S.A.). Experimental Procedures We food deprived three broodless, queenless colonies of 1 workers for 3 days. After 3 days the colonies were given access to the apparatus shown in Fig. 1. Nests were connected by a 3 cm length of string to the underside of two glass slides forming a Y- maze. The ants had access to the feeders via 3 cm lengths of string connected to the far ends of the glass slides. The sides of the glass slides were covered with a piece of acetate sheet coated in Fluon leaving strips of glass 1.5 cm wide for the ants to walk on. As the ants traversed the inverted glass strips, they were filmed at a 3 angle and backlit, to ensure that each ant s reflection was also visible in the glass surface. Whenever the ventral surface of an ant s gaster touched its reflection in the glass, that ant was deemed to be laying pheromone. The rate of substrate contacts by the gaster (also known as trail marking ) is a well-established proxy for the deposition of pheromone (Van Vorhis Key & Baker 1986; Aron et al. 1989; Jackson & Chaline 27). By forcing ants to walk on the underside of the glass, we prevented erroneous inflation of pheromone-laying rates when the engorged abdomens of fed ants accidentally touch the glass. This is because when the ants are upside-down, their gasters naturally sit further away from the substrate, and must be deliberately raised against gravity to lay Glass slide Y-maze 9 Arena/food source Light Arena/food source 23 cm Camera 3 Nest 35 cm 35 cm Camera 3 Figure 1. Experimental set-up for filming the behaviour of individual ants in food choice experiments. Ants could travel from the nest to the underside of a glass slide via a string walkway. Here they were backlit and filmed at a 3 angle as they were forced to travel upside-down along one of two arms of a 9 Y-maze. Access to an arena (containing either a high-quality or low-quality food source) was gained by another string walkway at the end of each arm. The glass slide was situated 23 cm above nest level, and the nest and food sources were separated by a distance of 35 cm. A 5 cm length of the trail was contained within the camera s field of view.

4 C. R. Reid et al. / Animal Behaviour 84 (212) 1579e pheromone, making incidences of trail marking clearer (see Fig. 2 for images of fed and unfed ants on the inverted surface). Filming began when the first ant discovered either food source and ended after 1 h. String walkways were discarded after each use and slides were washed in ethanol (9%). We tested context-dependent differences in trail-laying behaviour on each arm of the Y-maze by having one arm lead to a feeder containing a high-quality (HQ) food source of 1 M sucrose solution, and the other arm leading to a feeder containing a low-quality (LQ) food source of.33 M sucrose solution. To ensure that our different feeders were indeed considered high quality and low quality by the ants, we first had to confirm that Argentine ants can discriminate between 1 M and.33 M sucrose sources. We food deprived 27 broodless, queenless colonies of 5 workers for 3 days. The colonies were then given access to a Y-maze, with one arm leading to a.33 M sucrose solution, and the other to a 1 M sucrose solution. We alternated the position of the 1 M food source between replicates. After 6 min, we counted the ants feeding at each arm, and calculated the proportion of foraging ants feeding at the 1 M food source. Colonies could easily discriminate between the two sources, focusing their foraging effort on the HQ food source significantly more often (observed probability: 77% of foragers on HQ; onetailed binomial test: P <.1). Data Collection and Analysis To estimate the strength of the trail ( traffic ), we counted ants crossing from one end of the field of view to the other within each 1 min period. We were interested in whether an ant s direction of travel influences its behaviour on a pheromone trail, so we compared ants travelling out from the nest (outbound ants) with ants travelling from the food source back to the nest (nestbound ants). Nestbound ants were those ants that entered the field of view from the direction of the food, and exited the field of view via the end closest to the nest. Outbound ants performed the opposite. Individual ants could also influence colony level decisions by performing U-turns (i.e. returning to where they came from instead of continuing their path). We therefore further divided all ants into two groups: nonturners, which entered the field of view from one end and left via the other end, and U-turners, which entered and left the field of view via the same end. We examined the likelihood that an individual ant would mark the trail by determining the proportion of ants that were observed to mark the trail with their gaster at least once during the time of observation. We calculated the rate at which ants mark the trail (mean marking rate) by taking the total number of marks observed within each group, divided by the total number of ants within that group. To generate a true estimate of the mean marking rate of all ants, we included ants that did not deposit pheromone. Video data were recorded using a Sony HDR-XR55 Handycam and a Sony HDR-XR1 Handycam, each at 25 frames/s. Owing to the small size of individual ants, observing individual trail-marking behaviour required a high degree of resolution, and hence the section of trail within the field of view was restricted to the full 5 cm of the inverted glass slide surface. All data reported here come from observations within this field of view. Observations took place while the posterior tip of the gaster of each ant was within the field of view. Control for U-turning Outside Field of View As we were only able to observe ants closely over a relatively short distance travelled, we needed to confirm that most ants indeed proceeded to either the nest (inbound) or the food (outbound) and did not make U-turns outside our field of view thus re-entering the filmed area. If ants classified as either outbound or inbound U-turned outside the area in which we filmed them, the same ant would be counted multiple times. We therefore repeated the experiment with food-deprived colonies of 1 workers (N ¼ 3). Our single alteration to the set-up in Fig. 1 was allowing the ants access to the upper side of the Y-maze, so that they could be filmed traversing the entire set-up from above. We examined the behaviour of nestbound ants on their way back from the food source in the first 2 min after food discovery, as these were found to be the most influential in previous experiments. We found that the vast majority of foragers leaving a food source returned to the nest without first revisiting the food source (LQ: 87.84%, N ¼ 288; HQ: 93.18%, N ¼ 132). As we repeated the set-up of the original food choice experiments, we could delineate the original field of view recorded in the experiments. We calculated the percentage of ants that, as they travelled from food source to nest, U-turned outside of this field of view, thus returning to the area in which they were filmed in our original set-up. This allowed us to determine how many ants would have been recorded multiple times. We counted % ants on the LQ arm and 3% of ants on the HQ arm that would have been observed more than once in our original set-up. We can therefore assume that ants classified as outbound or nestbound in our original set-up mostly continued their trajectory. RESULTS Colony-level Behaviour The mean time required for the colonies to choose one food source over the other in the Y-maze can be measured as the time to an increase in mean traffic on one arm and a simultaneous decrease on the other arm. We plotted the mean traffic on each arm over each 1 min period (Fig. 3), and performed linear regressions between the means at each time point. The slopes of the regression lines were positive on both the HQ and LQ arms between 1 and 2 min (23. and , respectively). However, between 2 and 3 min the slope of the HQ regression line was positive (52.667), while the slope of the LQ regression was negative ( 7.333). We therefore split our results into two periods: the first 2 min, which we called the initial period, and the remaining 4 min, here called the final period of the experiment. We analysed the results from these two time periods separately. Traffic to both feeders increased rapidly in the initial period. After this time, traffic continued to increase to the HQ food source, while traffic to the LQ food source slightly decreased (Fig. 3). We compared the number of ants foraging at the HQ and LQ arms using chi-square tests. We used a heterogeneity chi-square test to determine whether the change in traffic followed the same pattern on both arms. The number of ants travelling to the HQ feeder was higher at every time point (P <.1; see Appendix Table A1) while the change in traffic over time was the same on both arms (P >.5; see Appendix Table A1). Individual-level Behaviour We were interested in determining whether some individuals in the group were more likely to lay pheromone than others, depending on whether the trail led to a low-quality or high-quality food source, their direction of travel, and also whether the propensity to lay pheromone changed over time. The proportion of ants marking (Fig. 4) is a measure of the percentage of foragers within each group that were observed to mark the trail with their gaster, and hence contribute to the pheromone trail. For the LQ arm, regardless of travel direction or time period, the relationship was always identical; significantly fewer U-turners marked the

5 1582 C. R. Reid et al. / Animal Behaviour 84 (212) 1579e1587 Figure 2. Example images of experimental data: (a) and (b) are ants moving in an outbound direction, (c) and (d) are ants moving in a nestbound direction, and (b) and (d) are examples of marking behaviour. Note the touching of the ant s gaster with its reflection in the glass in (b) and (d), compared to the distance between the gaster and its reflection in (a) and (c). Also note the distension of the gaster in (c) and (d) as a result of feeding. substrate than nonturners (Fig. 4, Appendix Table A2). On the HQ arm, however, the opposite relationship was observed for the nestbound ants in the initial period; U-turners were significantly more likely to mark the trail than nonturners (Fig. 4, Appendix Mean traffic (ants/ 1 min) LQ HQ Time (min) Figure 3. Traffic levels over the hour of experimentation. Data are presented as means for each 1 min period at the colony replicate level; error bars are SEM. The vertical dashed line delineates the observed trend difference between the period during which the trail was formed ( initial period ; 2 min) and the final period during which a trail was clearly established to the HQ food source (2 6 min). See text for further details. 5 6 Table A2). This proportion of nestbound U-turners marking the trail on the HQ arm dropped to the levels of the other groups in the final period (Fig. 4, Appendix Table A2). Having established the likelihood of marking for each group, we then examined how much influence each group of individuals had on the strength of the pheromone trail. The mean marks per ant (Fig. 5) indicates the mean marking rate, and hence relative contribution of pheromone to the trail, of foragers within each group. During the initial period, the nestbound ants returning from the HQ food source were the only group in which U-turners marked at a significantly higher rate than non-turners (Fig. 5, Appendix Table A3). The marking rate of nestbound U-turners on the HQ arm was three times higher than nestbound U-turners on the LQ arm. Later on in the experiment, during the final period, this boost in marking rate decreased, such that regardless of travel direction or food quality, U-turners marked at a significantly lower rate than nonturners (Fig. 5, Appendix Table A3). As U-turners were found to be major contributors to the pheromone trail, we examined whether there were more U-turners on the HQ arm than the LQ arm, and whether U-turning frequency changed over time. In the initial period, there were significantly more nestbound ants U-turning back to the HQ food source than the LQ food source: 32.2% (57/177) versus 16.3% (15/92), respectively (Fig. 5; two-sample binomial test: c 2 ¼ 1.269, P ¼.5). During the final period, this difference in the proportion of U- turners in nestbound ants was less pronounced: 13.97% (96/687) on the HQ arm, 8.76% (17/194) on the LQ arm (Fig. 5; two-sample binomial test: c 2 ¼ 1.881, P ¼.55). Absolute Contribution to the Pheromone Trail(s) The relative contribution of each class of ant to the pheromone trail can be observed in the data for marking rate (Fig. 5), but what was the absolute contribution by each class of forager to the total pheromone on the trail? We found a strong difference in the ants

6 C. R. Reid et al. / Animal Behaviour 84 (212) 1579e Nonturners U-turners Time period 1 * *.8 Proportion of ants marking * * * * High quality Food source.6.4 Low quality.2 Outbound Nestbound Outbound Travel direction Nestbound Figure 4. Proportion of ants marking. We compared U-turners with nonturners in the initial period and the final period using two-sample binomial tests. Stars indicate significant differences between groups, P values can be found in Appendix Table A2. behaviour between the initial period and the final period. If we sum the number of trail markings recorded for each treatment group (essentially combining the mean marking rate data with the number of observations in each group, both reported in Fig. 5) we find that the absolute contribution of U-turning ants decreases in the final period, while the absolute contribution of nonturning ants increases (Fig. 6). This trend does not hold for the LQ food source because the trail to the HQ food source prevails, and instead we see a general decrease in the contribution of both nonturning and U-turning ants towards the LQ food source. DISCUSSION Many species of ant, including Argentine ants, use trail networks to organize their foraging (Deneubourg et al. 199; Hölldobler & Wilson 199). Our results provide new understanding about how colonies of trail-laying ants are able to allocate and redistribute their foraging effort within the trail network. We have confirmed the role of U-turns in the collective decision-making process of trail-laying ants. U-turners make up a small proportion of the total foraging population, but owing to the high rate at which they mark the trail with pheromone, combined with the increased time they spend on the trail, U-turning ants contribute disproportionately to the build-up of a pheromone trail. Compared with ants returning from the LQ feeder, ants that returned from the HQ food source were twice as likely to perform a U-turn. Ants U-turning towards the HQ food source had a 33% increased probability of marking the trail and marked the trail at a rate three times higher than ants U-turning towards the LQ food source. Moreover, this increased trail-laying behaviour was observed in the initial period, when the colonies were choosing to exploit one food source over the other. As a result, a pheromone trail to a food source of high quality builds up faster than a trail that leads to a source of lower quality. Importantly, the trail builds in strength from the food source rather than from the nest. This ensures that if a single ant discovers a new food source there is a high probability that a new trail will be established. Although not tested here, it is very likely that it is this U- turning behaviour that allows mass-recruiting ants to adapt quickly to dynamic changes in their foraging environment, by increasing the strength of incipient trails to compete with established trails in the network (Vittori et al. 24; Reid et al. 211). Future studies specifically designed to examine the mechanisms by which mass recruiters forage dynamically should incorporate the role of U- turns in their investigations. We are not the first to report the existence of U-turning in ants during collective decision making. So far, however, studies have mainly focused on corrective U-turns, where an ant has recently turned off the correct path, and reorients itself based on visual cues or pheromone gradients (Beckers et al. 1992; Dussutour et al. 26; Garnier et al. 29). Very few studies have examined the role of

7 1584 C. R. Reid et al. / Animal Behaviour 84 (212) 1579e1587 Nonturners U-turners Time period 6 * * * 4 High quality 2 Mean marks per ant * Food source 4 * * Low quality Outbound Nestbound Outbound Travel direction Nestbound Figure 5. Marking rates in mean marks per ant. Error bars are SEM. The data include nonmarking ants. Numbers at the base of each column denote the number of observations within each group. Stars indicate significant differences between the treatments adjacent to each other under the star. We analysed outbound ants separately to nestbound ants, and performed a Poisson regression with quality (HQ/LQ) and turning status (U-turn/nonturn) as the independent variables. To control for differences between colonies, we nested colony within quality. We also tested for an interaction between quality and turning status. We tested for overdispersion, and present the Firth-bias adjusted estimates in Appendix Table A3. noncorrective U-turns, which occur spontaneously in ants travelling in the correct direction. Hart & Jackson (26) began to examine the role of noncorrective U-turns in the Pharaoh s ant. They suggested a role for these seemingly inappropriate U-turns in trail maintenance, arguing that the greater amount of time spent on the trail by U-turners allows them to contribute more pheromone than nonturners. Jackson et al. (211) further explored the role of noncorrective U-turns in a mathematical model based on the observed behaviour of Pharaoh s ants and argued that the presence of U-turners that lay more pheromone than nonturners allows a colony to exploit new food discoveries more rapidly. Our study, performed on a different species of trail-laying ant, shows how differences over time in individual trail-laying behaviour leads to the collective choice of two simultaneously available food sources. We thus found strong empirical support for Jackson et al. s (211) model, suggesting that U-turning ants are an essential part of collective decision making in mass-recruiting ant species. Jackson et al. s (211) model suggests that U-turners are kept more informed of trail conditions by spending more time on the trail. By studying the effect of travel direction on individual behaviour, we have shown a subtle but important difference; it is the ants that are most informed, nestbound from the food source, that are most likely to U-turn, and it is these informed individuals that guide the foraging effort of the colony. Even though we did not directly test the ants ability to adapt to dynamic changes during foraging, our study suggests two mechanisms that could allow the ants to do so. The first is a division of labour between U-turning and nonturning ants. As is usual for Argentine ants, foraging ants deposit pheromone when both outbound and nestbound (Van Vorhis Key & Baker 1986), and it is this continuous trail laying that contributes to the ants hypothesized tendency to make suboptimal foraging choices in mathematical models when a source of low quality is discovered before a high-quality source (Goss et al. 1989; Sumpter & Beekman 23). Becoming locked in to a suboptimal choice can be avoided if ants that have accurate information about a high-quality food source contribute more to the pheromone trail network than ants that have information about a low-quality food source. U-turning ants are such ants. Ants returning from the food source spend more time on the trail, are more likely to mark the trail with pheromone and mark the trail at

8 C. R. Reid et al. / Animal Behaviour 84 (212) 1579e Nonturners U-turners Outbound Travel direction Nestbound 6 5 Sum of marks/ 1 min High quality Food source Low quality 1 Time period Figure 6. The total sum of pheromone-laying marks. To control for the different lengths of each time period, we divided the total sum of marks by the number of 1 min blocks within each period (two for the initial period, four for the final period). a higher rate. Once a trail is established, (i.e. after the initial period), ants are less likely to perform U-turns and those that do reduce the rate at which they mark the trail. Thus U-turning ants appear to have a specialist role: to establish a trail rapidly when a food source is found. Once a trail is established, U-turning is no longer necessary. Future studies should examine the role of U-turning ants in redirecting the colony s foraging effort from a well-established trail leading to a low- or medium-quality food source, to an incipient trail leading to a newly discovered, high-quality food source. If U-turners perform the same function we have demonstrated here in influencing the collective choice to the best-quality food source, then the key role of U-turners in dynamic foraging will be confirmed. The second mechanism that allows a colony to (re-)allocate its workers rapidly to a more profitable food source is the construction of the trail from the food to the nest rather than the nest to the food, as hypothesized in the model by Jackson et al. (211). Some traillaying ants, such as Solenopsis saevissima, only lay pheromone when travelling back to the nest after inspecting a food source, and so construct a trail from the food to the nest (Wilson 1962). However, the fact that Argentine ant foragers deposit pheromone when travelling in both directions means that, all else being equal, trails will always be strongest at the nest end and weakest at the food end. This is because if a scout is successful in laying a trail back to the nest, both the trail itself and the pheromone laid by recruited workers will always ensure that more ants are recruited from the nest than are returning from the food source. Since the ants at the nest end also lay pheromone, the trial will be strongest at the nest end. Nest-favoured asymmetry in trail strength can be detrimental when relying on complex trail networks for navigation in a dynamically changing environment. If a new high-quality food source is discovered outside an existing trail network, incipient trails to the new food source must compete with the strong, established trails already in use within the network. If ants choose their direction of travel based on pheromone strength alone (as implemented in many models), foragers will rarely follow the new, weak trail and new food sources would remain unutilized. The constant retracing of steps and increased trail-marking behaviour exhibited by U-turners essentially means an individual can contribute as much pheromone as a small group of nonturning foragers within the same amount of time. Thus the small proportion of U-turning ants may well be able to overcome the constraints imposed by the pheromone s volatility on the initiation of new pheromone trails (Beekman et al. 21). Moreover, building the pheromone gradient from the food source back to the nest, or back to the intersection with existing trails in the network, ensures that the trail is strong when it reaches the nest or network. Increased trail strength will ensure more traffic on the new trail, which will lead to stronger reinforcement, expansion of the network and effective utilization of new resources.

9 1586 C. R. Reid et al. / Animal Behaviour 84 (212) 1579e1587 We have demonstrated that a small group of ants can play a pivotal role in decisions made at the colony level. This is not unknown in the study of collective behaviour, where many animal groups have been shown to contain a subclass of individuals that have a disproportionate influence over the group s decision making. During nest site selection in honeybees, Apis mellifera, the scouts responsible for the ultimate decision comprise only around 5% of the colony s population (Seeley & Buhrman 21), while in house-hunting Temnothorax ants, the decision is driven by about a third of the colony s workers (Pratt et al. 22). Similarly, a minority of informed individuals directs group decisions in primates deciding to move (Byrne 2), groups of homing pigeons, Columba livia, returning to their home roost (Biro et al. 26), foraging in shoals of fish (Reebs 2), and flying honeybee swarms (Beekman et al. 26; Schultz et al. 28; Latty et al. 29). Mass recruitment by trail-laying ants is often seen as the sum of many interactions by simple, uniform individuals. As a result, many mathematical models of ant foraging reduce the number of behavioural parameters to the bare minimum necessary to reconstruct empirical observations qualitatively. This simplification has contributed to the misconception that mass recruitment equals individual-unit simplicity. Such an approach is not usually applied to social insects that do not rely on trail pheromone recruitment (Mallon et al. 21; Seeley & Buhrman 21). Our study contributes to the growing realization that workers of mass-recruiting ant species are not only capable of, but also rely upon, complex individual behaviour for collective decision making. Acknowledgments We thank Pierre Humblot for assistance with early versions of the experiment, Sara Perry, Charlie Dafforn, Samantha Zaiter and Shuravi Paul (volunteers) for assistance with video analysis, and Benjamin Oldroyd, Mark Brown and the anonymous referees for comments on the manuscript. This work was supported by the Human Frontier Science Program (M.B.), The Australian Research Council (M.B. and T.L.), Natural Sciences and Engineering Council of Canada (T.L.), and the Centre for Mathematical Biology, University of Sydney (C.R.). References Aron, S., Pasteels, J. M. & Deneubourg, J. L Trail-laying behaviour during exploratory recruitment in the Argentine ant, Iridomyrmex humilis (Mayr). Biology of Behaviour, 14, 27e217. Ballerini, M., Cabibbo, N., Candelier, R., Cavagna, A., Cisbani, E., Giardina, I., Orlandi, A., Parisi, G., Procaccini, A., Viale, M., et al. 28. Empirical investigation of starling flocks: a benchmark study in collective animal behaviour. Animal Behaviour, 76, 21e215. Bazazi, S., Buhl, J., Hale, J. J., Anstey, M. L., Sword, G. A., Simpson, S. J. & Couzin, I. D. 28. Collective motion and cannibalism in locust migratory bands. Current Biology, 18, 735e739. Beckers, R., Deneubourg, J. L., Goss, S. & Pasteels, J. M Collective decisionmaking through food recruitment. Insectes Sociaux, 37, 258e267. Beckers, R., Deneubourg, J. L. & Goss, S Trails and U-turns in the selection of a path by the ant Lasius niger. Journal of Theoretical Biology, 159, 397e415. Beekman, M., Sumpter, D. J. T. & Ratnieks, F. L. W. 21. Phase transition between disordered and ordered foraging in Pharaoh s ants. Proceedings of the National Academy of Sciences, U.S.A., 98, 973e976. Beekman, M., Fathke, R. L. & Seeley, T. D. 26. How does an informed minority of scouts guide a honeybee swarm as it flies to its new home? Animal Behaviour, 71, 161e171. Biro, D., Sumpter, D. J. T., Meade, J. & Guilford, T. 26. From compromise to leadership in pigeon homing. Current Biology, 16, 2123e2128. Bonabeau, E Flexibility at the edge of chaos: a clear example from foraging in ants. Acta Biotheoretica, 45, 29e5. Bonabeau, E., Theraulaz, G. & Deneubourg, J.-L Group and mass recruitment in ant colonies: the influence of contact rates. Journal of Theoretical Biology, 195,157e166. Bonabeau, E., Dorigo, M. & Theraulaz, G Swarm Intelligence: From Natural to Artificial Systems. New York: Oxford University Press. Byrne, R. W. 2. How Monkeys Find their Way: Leadership, Coordination, and Cognitive Maps of African Baboons. Chicago: Chicago University Press. Camazine,S.,Visscher,P.K.,Finley,J.&Vetter,R.S House-hunting by honey bee swarms: collective decisionsand individual behaviors. Insectes Sociaux, 46, 348e36. Camazine, S., Deneubourg, J.-L., Franks, N. R., Sneyd, J., Theraulaz, G. & Bonabeau, E. 21. Self-organization in Biological Systems. Princeton, New Jersey: Princeton University Press. Couzin, I. D. 29. Collective cognition in animal groups. Trends in Cognitive Sciences, 13, 36e43. Deneubourg, J.-L., Aron, S., Goss, S. & Pasteels, J. M The self-organizing exploratory pattern of the Argentine ant. Journal of Insect Behavior, 3, 159e168. Diwold, K., Schaerf, T. M., Myerscough, M. R., Middendorf, M. & Beekman, M Deciding on the wing: in-flight decision making and search space sampling in the red dwarf honeybee Apis florea. Swarm Intelligence, 5, 121e141. Dussutour, A. & Simpson, S. J. 28. Description of a simple synthetic diet for studying nutritional responses in ants. Insectes Sociaux, 55, 329e333. Dussutour, A., Nicolis, S. C., Deneubourg, J. L. & Fourcassie, V. 26. Collective decisions in ants when foraging under crowded conditions. Behavioral Ecology and Sociobiology, 61, 17e3. Dussutour, A., Beekman, M., Nicolis, S. C. & Meyer, B. 29a. Noise improves collective decision-making by ants in dynamic environments. Proceedings of the Royal Society B, 276, 4353e4361. Dussutour, A., Nicolis, S. C., Shephard, G., Beekman, M. & Sumpter, D. J. T. 29b. The role of multiple pheromones in food recruitment by ants. Journal of Experimental Biology, 212, 2337e2348. Edelstein-Keshet, L Simple models for trail-following behavior; trunk trails versus individual foragers. Journal of Mathematical Biology, 32, 33e328. Garfinkel, A The Slime Mold Dictyostelium as a Model of Self-organization in Social Systems. Berlin: Springer. Garnier, S., Guerecheau, A., Combe, M., Fourcassie, V. & Theraulaz, G. 29. Path selection and foraging efficiency in Argentine ant transport networks. Behavioral Ecology and Sociobiology, 63, 1167e1179. Goss, S., Aron, S., Deneubourg, J. L. & Pasteels, J. M Self-organized shortcuts in the Argentine ant. Naturwissenschaften, 76, 579e581. Hart, A. & Jackson, D. E. 26. U-turns on ant pheromone trails. Current Biology, 16, R42eR43. Herbert-Read, J. E., Perna, A., Mann, R. P., Schaerf, T. M., Sumpter, D. J. T. & Ward, A. J. W Inferring the rules of interaction of shoaling fish. Proceedings of the National Academy of Sciences, U.S.A., 18, 18726e Hölldobler, B. & Wilson, E. O The Ants. Berlin: Springer-Verlag. Jackson, D. E. & Chaline, N. 27. Modulation of pheromone trail strength with food quality in Pharaoh s ant, Monomorium pharaonis. Animal Behaviour, 74, 463e47. Jackson, D. E., Holcombe, M. & Ratnieks, F. L. W. 24. Trail geometry gives polarity to ant foraging networks. Nature, 432, 97e99. Jackson, D. E., Bicak, M. & Holcombe, M Decentralised communication, trail connectivity and emergent benefits of ant pheromone trail networks. Memetic Computing, 3, 25e32. Katz, Y., Tunstrom, K., Ioannou, C. C., Huepe, C. & Couzin, I. D Inferring the structure and dynamics of interactions in schooling fish. Proceedings of the National Academy of Sciences, U.S.A., 18, 1872e Latty, T., Duncan, M. & Beekman, M. 29. High bee traffic disrupts transfer of directional informationinflying honeybeeswarms. Animal Behaviour, 78,117e121. Mallon, E. B., Pratt, S. C. & Franks, N. R. 21. Individual and collective decisionmaking during nest site selection by the ant Leptothorax albipennis. Behavioral Ecology and Sociobiology, 5, 352e359. Nicolis, S. C. & Deneubourg, J.-L Emerging patterns and food recruitment in ants: an analytical study. Journal of Theoretical Biology, 198, 575e592. Pasteels, J. M., Deneubourg, J.-L. & Goss, S Self-organization Mechanisms in Ant Societies (I) Trail Recruitment to Newly Discovered Food Sources. Basel: Birkhauser. Pratt, S. C., Mallon, E. B., Sumpter, D. J. T. & Franks, N. R. 22. Quorum sensing, recruitment, and collective decision-making during colony emigration by the ant Leptothorax albipennis. Behavioral Ecology and Sociobiology, 52, 117e127. Ramsch, K., Reid, C. R., Beekman, M. & Middendorf, M A mathematical model of foraging in a dynamic environment by trail-laying Argentine ants. Journal of Theoretical Biology, 36, 32e45. Reebs, S. G. 2. Can a minority of informed leaders determine the foraging movements of a fish shoal? Animal Behaviour, 59, 43e49. Reid, C. R., Sumpter, D. J. T. & Beekman, M Optimisation in a natural system: Argentine ants solve the Towers of Hanoi. Journal of Experimental Biology, 214, 5e58. Robertson, P. L., Dudzinski, M. L. & Orton, C. J Exocrine gland involvement in trailing behavior in the Argentine ant (Formicidae, Dolichoderinae). Animal Behaviour, 28, 1255e1273. Schultz, K. M., Passino, K. M. & Seeley, T. D. 28. The mechanism of flight guidance in honeybee swarms: subtle guides or streaker bees? Journal of Experimental Biology, 211, 3287e3295. Seeley, T. D. 21. Honeybee Democracy. Princeton, New Jersey: Princeton University Press. Seeley, T. D. & Buhrman, S. C. 21. Nest-site selection in honey bees: how well do swarms implement the best-of-n decision rule? Behavioral Ecology and Sociobiology, 49, 416e427. Sumpter, D. J. T. & Beekman, M. 23. From nonlinearity to optimality: pheromone trail foraging by ants. Animal Behaviour, 66, 273e28.

10 C. R. Reid et al. / Animal Behaviour 84 (212) 1579e Van Vorhis Key, S. E. & Baker, T. C Observations on the trail deposition and recruitment behaviors of the Argentine ant, Iridomyrmex humilis (Hymenoptera: Formicidae). Annals of the Entomological Society of America, 79, 283e288. Vittori, K., Gautrais, J., Araujo, A. F. R., Fourcassie, V. & Theraulaz, G. 24. Modeling Ant Behavior under a Variable Environment. Berlin: Springer-Verlag. Wilson, E. O Chemical communication among workers of the fire ant. Solenopsis saevissima (Fr. Smith). I. The organization of mass-foraging. Animal Behaviour, 1, 134e147. Table A1 Chi-square tests of heterogeneity for traffic c 2 df P period 1 min <.1 2 min c 2 of total <.1 Total of c <.1 Heterogeneity period 3 min <.1 4 min <.1 5 min <.1 6 min <.1 c 2 of total <.1 Total of c <.1 Heterogeneity Heterogeneity c 2 were calculated by calculating the absolute value of the difference between c 2 of Total and the Total of c 2. P values were calculated from the c 2 distribution of the c 2 values and degrees of freedom. Significant values in bold. Table A2 Comparison of U-turners to nonturners via two-sample binomial analysis of proportion of ants marking Travel direction Quality Time period Test statistic P Test statistic P Outbound LQ c 2 1;115 ¼ 8:792.3 c 2 1;237 ¼ 2:61 <.1 HQ c 2 1;199 ¼ 34:474 <.1 c 2 1;765 ¼ 2:71.15 Nestbound LQ c 2 1;92 ¼ 17:864 <.1 c 2 1;24 ¼ 36:9 <.1 HQ c 2 1;177 ¼ 5: c 2 1;687 ¼ 15:727 <.1 Significant values in bold. Table A3 Poisson regression analysis of marking rate data Travel direction Source of variance Time period Test statistic P Test statistic P Outbound Quality c 2 1;314 ¼ : c 2 1;12 ¼ 3:42.65 Turn c 2 1;314 ¼ 1: c 2 1;12 ¼ 19:166 <.1 Quality*turn c 2 1;314 ¼ : c 2 1;12 ¼ : Nestbound Quality c 2 1;269 ¼ 4: c 2 1;927 ¼ 3: Turn c 2 1;269 ¼ 5: c 2 1;927 ¼ 35:669 <.1 Quality*turn c 2 1;269 ¼ 11:364 <.1 c 2 1;927 ¼ 2: Significant values in bold.