Naturwissenschaften DOI 10.1007/s00114-007-0329-9
ORIGINAL PAPER
Individual and collective choice: parallel prospecting and mining in ants Antony S. Aleksiev & Ben Longdon & Matthew J. Christmas & Ana B. Sendova-Franks & Nigel R. Franks
Received: 3 August 2007 / Revised: 8 November 2007 / Accepted: 15 November 2007 # Springer-Verlag 2007
Abstract Decision making is of crucial importance in the lives of both animals and humans. How decisions of group members scale up to group decisions is of great interest. Accordingly, we gave homeless ant colonies (n=67) in three experiments a choice between two nest sites (with small, big or mixed sand grains), each of which had to be excavated to be habitable. Among the colonies that chose only one of the new nest sites, all preferred the ones that could be excavated most easily and quickly. There are interesting parallels between the collective choice of mining sites and the ability of certain ants to select short cuts; both involve positive feedback. However, in this paper, we discuss a mechanism whereby collective co-ordination in the production of social infrastructure can occur in the absence of signalling. Keywords Ant . Nest excavation . Aggregation . Self-organisation . Temnothorax albipennis
Introduction Animals often benefit from belonging to groups (Wilson 1975; Trivers 1985). Such groups may be mobile, and a consensus is often achieved on where they go and where they A. S. Aleksiev : B. Longdon : M. J. Christmas : N. R. Franks (*) School of Biological Sciences, University of Bristol, Woodland Road, Bristol BS8 1UG, UK e-mail:
[email protected]
A. B. Sendova-Franks School of Mathematical Sciences, CEMS, University of the West of England, Bristol BS16 1QY, UK
stay (Franks et al. 2002; Conradt and Roper 2005; Couzin et al. 2005; Amé et al. 2006). How the decisions of individual group members scale up to group decisions is of great interest (Camazine et al. 2001; Dussutour et al. 2004; Conradt and Roper 2005; Amé et al. 2006; Couzin et al. 2005). Self-organisation theory (Camazine et al. 2001) may help to explain how individual, locally based, rules of thumb lead to collective patterns and processes. For example, individual behaviour self-organises a decision during short-cut selection in ants (Goss et al. 1989; Beckers et al. 1993; Camazine et al. 2001). Consider short-cut selection when an ant colony is separated from a food source by a single bridge with a short and a long branch. A colony is often able exclusively to choose the shorter branch if the worker ants lay recruiting trail pheromones whenever they are foraging (Goss et al. 1989; Beckers et al. 1993; see also Camazine et al. 2001). This occurs because an ant that happens to take the shorter path will get to the food and back more quickly than one that takes the longer path. Each ant reinforces its choices with its own pheromone trails. Hence, the shorter path becomes more and more attractive much more quickly than the longer path. No ant needs directly to compare the path lengths. Exclusive colony-level choice of the short path can occur in the absence of any globally knowledgeable individual. Rather, individuals simply use the rule of thumb—when a food source has been discovered, follow and reinforce the strongest pheromone trail (Camazine et al. 2001). In this study, we examine a fundamentally similar, but simpler, set of logically equivalent rules of thumb that might enable an emigrating ant colony to choose between nest sites that require excavation (Deneubourg and Franks 1995). Recent work on choice of aggregation sites (Amé et al. 2006) also provides a fascinating comparison with the work we will describe in this paper.
Naturwissenschaften
Temnothorax albipennis ants inhabit flat crevices in rocks that may need to be excavated to be habitable (Franks et al. 1992). Nest choice in T. albipennis has been investigated extensively (Franks et al. 2003a, b; for a recent review, see Visscher 2007) but not for nest sites that need to be mined. Previous work has looked at the choices these ants make over different sizes of sand grain when they occupy empty crevices and need to retrieve such material to build one or more protective walls (Aleksiev et al. 2007). The ants prefer large sand grains over smaller ones but generally construct mixed walls, composed of both sizes of grain, probably because these can be both stronger and more compact (Aleksiev et al. 2007). Nest sites needing excavation pose a fascinating new issue in terms of nest choice because the potential nests are not only initially uninhabitable but cannot even be surveyed. For example, T. albipennis are able individually to assess the floor area (Mallon and Franks 2000), the available headroom (Franks et al. 2003b, 2006a) and the width and abundance of entrances (Franks et al. 2006a) of empty nest sites. Individuals, encountering cavities that are so filled with building material that they cannot be entered let alone surveyed, can assess none of these attributes. So how, in such circumstances, are individual and collective decisions made?
Materials and methods We offered ant colonies binary choices between cavities that were either filled with large or small sand grains or a mixture of large and small sand grains. We collected 67 T. albipennis colonies in April and October 2005 from Dorset, England and allocated them randomly to one of three experiments. Colonies collected in October from our study site have a median size of 100 workers (Q1 =73, Q3 =136) and 153 brood (Q1 =84, Q3 = 202, Franks et al. 2006b). In April, colonies seem to be smaller probably because some exhibit seasonal polydomy (Partridge et al. 1997, including data on colony sizes). They are rarely, if ever, functionally polygynous (Partridge et al. 1997). We have no evidence that monogynous or queenless nests differ in their nest site preferences (personal observations), and therefore, in this study, we used a similar proportion of queen-right and queen-less nests as occurred in our collections. Experiment 1 used cavities filled with either only big sand grains or only small sand grains. Experiment 2 used cavities with only big grains or a mixture of big and small grains. Experiment 3 used cavities with either only small grains or a mixture of grain sizes. We used only dry sand throughout this study. Well-mixed sand was introduced into each cavity through a paper funnel.
Mixed grains consisted of a 1:5 ratio of big to small grains (with average diameters of 0.97 and 0.56 mm, respectively). This is an equal volume of big and small grains so they should have been equally apparent. The nest cavities were 70×30×1.1 mm and were formed from two microscope slides held apart by cardboard on three sides (Fig. 1). Black cardboard covers made them dark. The arena was a 220×220×22 mm Petri dish with Fluon\-coated sidewalls to prevent the ants from escaping. The new nest sites were 10 cm apart and equidistant from the old nest. The old nest was destroyed by removing its top microscope slide. Each replicate ran for 48 h, after which we noted which cavity the colony had chosen for its new nest site. Colonies were recorded as having “split” if there were one or more brood items in each of the new nests. They were recorded as having “failed to emigrate” if the colony still occupied the old destroyed nest site after 48 h. To measure the rate of excavation, we marked all the workers in another five colonies collected in June 2006 (see Partridge et al. 1997 for typical colony sizes) and gave them a choice between a cavity filled with large sand grains and a cavity filled with small sand grains. The dimensions of the nest cavities and the experimental arena were the same as in the other three experiments. The observation of individual mining from the two new nests took place when 3 to 4 h had elapsed from the removal of the top slide of the old nest. We recorded the time of deposition outside the nest for up to 20 consecutive events of sand grain removal by marked individuals for each of the two grain sizes for each of the five colonies. Deposition rates for excavated grains and their 95% confidence intervals were estimated from the slopes of linear regression models fitted to the above data sets for
Fig. 1 Two potential nest sites filled with grains. The left nest has big grains, the right small ones. The ants have begun to mine both nests but have chosen the nest with big grains. Black cardboard covers removed for photograph
Naturwissenschaften Table 1 Nest choice between cavities filled with only big grains or only small grains, between cavities filled with only big grains or a grain mixture and between cavities filled with only small grains or a grain mixture Choice of nest
Big
Mixed
Small
Splits
Failed to emigrate
P value
Big vs small Big vs mixed Mixed vs small
14 10 –
– 0 6
0 – 4
9 1 2
7 6 8
0.0001 0.0020 0.7539
P values are for two-tailed binomial tests for the respective choice.
each of the two grain sizes in each of the five individually marked colonies. All ten regression models fitted well, with variation explained by the model ranging between 76.2 and 99.6% and residual distributions not significantly different from normal except for two cases where the distributions were near-normal with P values of 0.023 and 0.027 (colony 1) from normality tests.
Results Cavities filled only with big grains were strongly preferred to those filled either with small grains (two-tailed binomial test, N=14, with 14 exclusive choices of cavities filled with big grains, P=0.0001) or cavities filled with a mixture of the two grain sizes (N=10, with ten exclusive choices of cavities filled only with big grains P=0.0020; Table 1). Colonies showed no preference between nest cavities filled only with small grains and those with a mixture of the two grain sizes (four choices of the former cavity and six choices of the latter cavity; P=0.7539; Table 1). Colonies that split or failed to emigrate were excluded from the analyses. Notably, the colonies always began to mine both cavities. The deposition rate for excavated grains was significantly affected by grain size (two-way analysis of variance: F1,4 = 16.45, P=0.015) but also by colony (F4,4 =9.48, P=0.026). However, in all colonies, small grains were excavated faster than large grains (Fig. 2). This was the case notwithstanding the variation in grain deposition rates across colonies. Such variation between colonies is common in T. albipennis in particular (personal observations) and very likely also occurs in social insects in general. However, colony level comparisons are still rare, and we are only at the beginning of gathering evidence for this fascinating issue. It warrants further investigation. Each sand grain was removed one at a time by the excavating ants. This may have been associated with the sand being dry so that the grains did not stick together. The mean deposition rate was 0.025 grains per second for excavated large grains and 0.054 grains per second for excavated small grains. In other words, small grains were removed from the nest cavity just over two times faster than large grains. As individual large grains occupy a three times larger area and a
five times larger volume on average than small grains, it will be quicker for the ants to clear either a volume or an area of such larger grains than such smaller ones.
Discussion T. albipennis colonies, which made a choice, exclusively chose nest cavities filled only with big sand grains over those filled with only small sand grains or a 1:5 mixture of big to small grains. However, there was no clear preference between cavities with a 1:5 mixture of big to small grains and those filled only with small grains. T. albipennis colonies excavated small grains two times faster than large grains. The benefit of choosing a cavity filled exclusively with big grains over one filled exclusively with small grains is the speed at which the cavity can be mined and inhabited. In a foraging study (Aleksiev et al. 2007) in which the ants had to retrieve big or small grains to build a wall in an empty nest, colonies always preferred big grains. Such big grains are more profitable because the volume and area they occupy are, respectively, approximately five and three times greater than the volume and area occupied by a similar number of small grains. Therefore, although small grains were mined two times faster, the net profitability, measured either in total volume or in total area excavated per unit
Fig. 2 Deposition rate for grains excavated from the cavities of the two new nests—one filled in with small grains and the other with large grains in the five individually marked colonies; filled circle large grains; open circle small grains
Naturwissenschaften
time, from expelling big grains should be greater. Thus, the ants should be quicker at clearing a given space of big grains than they would be for small grains. In the cavities with mixed grains, small grains are five times more abundant than big grains. It is not surprising that the ants prefer cavities that are quick to excavate, because they are full with big grains, to cavities that are slow to excavate, because they have mixed grains, which are mostly small. This also explains the lack of choice between cavities with mixed grains and small grains. The speed at which these can be excavated will be similar; both had huge numbers of small grains. Thus, if ants can dig big grains at a rate c grains per unit time, then their rate of excavating small grains only and mixed grains will be 0.4c (2c/5, that is two times faster than big grains but five times less volume of grain) and 0.5c (c×1/6+2c/5×5/6) grains per unit time, respectively. Potentially, such calculations could be used for comparisons to test for positive feedback through recruitment. However, we did not observe recruitment in any of our 72 experimental colonies. When the ants have to retrieve grains for building, they bias their choice to big grains but always produce a wall with a mixture of grains (Aleksiev et al. 2007). Hence, they may have preferred cavities packed with mixed grains. They did not, possibly because the mechanism underlying the choice is probably differential excavation speed. Such a mechanism would favour nests filled exclusively with big grains. In each and every replicate of all experiments, the ants began excavating both cavities. This strongly suggests that individuals are not able immediately to determine that one cavity has greater potential for swift mining than the other. When these ants house-hunt for empty cavities, individuals can make comparisons (Pratt et al. 2002). In this study, individuals seem simply to start mining. This would be sufficient, however, for more rapid positive feedback, associated with the speed of excavation, to result in aggregation and collective choice. Imagine a race initially between individual miners: one ant in a cavity filled with big grains, the other ant in a cavity filled with small grains. The two ants work equally hard. However, the ant among the big grains could clear a space in which more than one of its nest mates could join it and start digging, at the same time that the ant among the small grains could clear a space to accommodate just one of its nest mates. Of course, multiplicative effects, of one individual digging out large grains and rapidly clearing a space that would accommodate another and then the two ants more quickly excavating space for a third and so on, would make this differential even greater. The result would be a collective choice of cavities with big grains even if no communication was involved (Deneubourg and Franks 1995).
This is beautifully analogous to other ants being able to select the shorter of two paths to the same goal although no individual ant has necessarily assessed the lengths of both paths (Beckers et al. 1993). In both cases, any benefit snowballs. For nest excavation, however, the system is potentially even simpler—no communication would be needed. The ants would only need to use a simple rule of thumb such as “try to enter a cavity and start mining”. A cavity that was easy and fast to mine would be easier and quicker for others to enter; more ants would join in the excavation and it should be cleared more and more rapidly (Deneubourg and Franks 1995). We cannot rule out the possibility that these ants or others might also use attractive volatile recruitment pheromones in such situations. However, such pheromones have never, to the best of our knowledge, been described in T. albipennis. Another possible source of positive feedback would occur if the ants are less likely to leave a larger aggregation than a smaller one (Amé et al. 2006). These possibilities warrant further experimental quantification and will be one focus of our continuing work in this area. Positive feedback resulting from successful mining attracting further miners would be sufficient to explain all of the choices we have observed. So far, we have observed no initial recruitment to these filled cavities (personal observations). In nest choices between empty cavities, these ants, after making their own survey, may begin to recruit nest mates initially by tandem running (Franks and Richardson 2006). Only when a quorum has been achieved do they start to recruit passive nest mates by carrying them in to the new nest site (Pratt et al. 2002). Not only did we see no initial recruitment in these experiments but we also noted that many colonies split (thus occupying both cavities) or failed to emigrate at all (Table 1). Total emigration failures are much rarer when colonies are given choices of empty cavities; occurring in only about 1 in 20 replicates (NR Franks, personal observations). We suspect that colony splits and emigration failures are associated here with (1) the need to excavate the cavities, (2) the inability of the ants to survey them and (3) no initial recruitment. Simply, an ant that encounters a filled cavity cannot explore to make an assessment of all of that nest’s qualities, and hence, once engaged in mining, the ant may have neither the time nor the motivation to recruit to it. An ant engaged in mining may be so fully occupied by this time-consuming task that it has no time to spare for any other activity. Furthermore, as it cannot survey a cavity that is full of sand grains, it may have little motivation to recruit nestmates to it because it cannot measure many of the attributes that have been shown to be important in nest choice in this species such as floor area (Mallon and Franks 2000; Franks et al. 2006b) and the number and size of potential entrances to the cavity (Franks et al. 2006a).
Naturwissenschaften
Relatively large numbers of colonies split or failed to emigrate in many of these experiments. Both phenomena might be explained by the inability of the ants to assess and hence recruit to cavities that are full of debris. In nature, when we are collecting these ants, we rarely find any crevices in rocks, including uninhabited ones, that are completely filled by debris. Given our results, we would expect, all else being equal, colonies to emigrate preferentially to cavities that contain less material and can be assessed and more rapidly occupied. Furthermore, we have consistently found in choice experiments that the more similar two nest are, the higher the proportion of colonies that split (Franks et al. 2003b, 2006a, 2007). Therefore, one reason for the large number of colonies that split here might be that the two nest cavities were identical in every important respect, except for the size of the grains they contained. Indeed, they offered the same amount of head room, potential inside floor area and entrance width (Franks et al. 2003b). Hence, once the ants could begin to detect certain of these factors, the alternative nests might have seemed very similar, and this might have caused colonies to split. Moreover, certain colonies might either have been forced to split by the small amount of space offered at each new nest site, as it was slowly excavated, or indeed they may have been inhibited from moving at all. The ability of ants to select short cuts has inspired the so-called ant colony optimisation algorithms that are used in human telecommunications (Dorigo and Stützle 2004). Here, the algorithm of successful mining attracting more mining could be used in distributed concurrent surveying by relatively simple robots. Populations of probing, mining and aggregating robots could be used quickly to survey and find the best areas for swift and effective excavation. Little or no communication, or global knowledge, would be needed for the robots to find better sites, and the best site would be marked by the largest aggregation of these distributed robotic agents.
Acknowledgements studentship.
ASA was supported by an ORS postgraduate
References Aleksiev AS, Sendova-Franks AB, Franks NR (2007) The selection of building material for wall construction in ants. Anim Behav 73: 779–788 Amé J-M, Halloy J, Rivault C, Detrain C, Deneubourg J-L (2006) Collegial decision making based on social amplification leads to optimal group formation. Proc Natl Acad Sci USA 103:5835–5840
Beckers R, Deneubourg J-L, Goss S (1993) Modulation of trail laying in the ant Lasius niger (Hymenoptera: Fromicidae) and its role in the collective selection of a food source. J Insect Behav 6:751–759 Camazine S, Deneubourg J-L, Franks NR, Sneyd J, Theraulaz G, Bonabeau E (2001) Self-organization in biological systems. Princeton University Press, Princeton New Jersey Conradt L, Roper TI (2005) Consensus decision making in animals. Trends Ecol Evol 20:449–456 Couzin ID, Krause J, Franks NR, Levin SA (2005) Effective leadership and decision-making in animal groups on the move. Nature 433:513–516 Deneubourg J-L, Franks NR (1995) Collective control without explicit coding: the case of communal nest excavation. J Insect Behav 8: 417–431 Dorigo M, Stützle T (2004) Ant colony optimization. MIT, Boston, MA, USA Dussutour A, Fourcassié V, Helbing D, Deneubourg J-L (2004) Optimal traffic organization in ants under crowded conditions. Nature 428:70–73 Franks NR, Wilby A, Silverman BW, Tofts C (1992) Self-organizing nest construction in ants: sophisticated building by blind bulldozing. Anim Behav 44:357–375 Franks NR, Pratt SC, Mallon EB, Britton NF, Sumpter DJT (2002) Information flow, opinion-polling and collective intelligence in house-hunting social insects. Phil Trans R Soc Lond B 357: 1567–1583 Franks NR, Dornhaus A, Fitzsimmons JP, Stevens M (2003a) Speed versus accuracy in collective decision-making. Proc R Soc Lond B 270:2457–2463 Franks NR, Mallon EB, Bray HE, Hamilton MJ, Mischler TC (2003b) Strategies for choosing among alternatives with different attributes: exemplified by house-hunting ants. Anim Behav 65:215–223 Franks NR, Richardson T (2006) Teaching in tandem-running ants. Nature 439:153–153 Franks NR, Dornhaus A, Metherell BG, Nelson TR, Lanfear SAJ, Symes WS (2006a) Not everything that counts can be counted: ants use multiple metrics for a single nest trait. Proc R Soc Lond B 273:165–169 Franks NR, Dornhaus A, Best CS, Jones EL (2006b) Decision-making by small and large house-hunting ant colonies: one size fits all. Anim Behav 72:611–616 Franks NR, Hooper JW, Dornhaus A, Aukett PJ, Hayward AL, Berghoff SM (2007) Reconnaissance and latent learning in ants. Proc R Soc Lond B 274:1505–1509 Goss S, Aron S, Deneubourg J-L, Pasteels JM (1989) Self-organized shortcuts in the argentine ant. Naturwissenschaften 76:579–581 Mallon EB, Franks NR (2000) Ants estimate area using Buffon’s needle. Proc R Soc Lond B 267:765–770 Partridge LW, Partridge KA, Franks NR (1997) Field survey of a monogynous leptothoracine ant (Hymenoptera, Formicidae): evidence of seasonal polydomy? Insectes Soc 44:75–83 Pratt SC, Mallon EB, Sumpter DJT, Franks NR (2002) Quorum sensing, recruitment, and collective decision-making during colony emigration by the ant Leptothorax albipennis. Behav Ecol Sociobiol 52:117–127 Trivers R (1985) Social evolution. Benjamin/Cummings, Menlo Park, CA, USA Visscher PK (2007) Group decision making in nest-site selection among social insects. Annu Rev Entomol 52:255–275 Wilson EO (1975) Sociobiology: the new synthesis. Belknap, Cambridge, MA, USA
