The regulation of ant colony foraging activity without spatial information

The Regulation of Ant Colony Foraging Activity without
Spatial Information
Balaji Prabhakar1, Katherine N. Dektar2, Deborah M. Gordon3*
1 Departments of Computer Science and Electrical Engineering, Stanford University, Stanford, California, United States of America, 2 Biomedical Computation, School of
Engineering, Stanford University, Stanford, California, United States of America, 3 Department of Biology, Stanford University, Stanford, California, United States of America

Abstract
Many dynamical networks, such as the ones that produce the collective behavior of social insects, operate without any
central control, instead arising from local interactions among individuals. A well-studied example is the formation of
recruitment trails in ant colonies, but many ant species do not use pheromone trails. We present a model of the regulation
of foraging by harvester ant (Pogonomyrmex barbatus) colonies. This species forages for scattered seeds that one ant can
retrieve on its own, so there is no need for spatial information such as pheromone trails that lead ants to specific locations.
Previous work shows that colony foraging activity, the rate at which ants go out to search individually for seeds, is regulated
in response to current food availability throughout the colony’s foraging area. Ants use the rate of brief antennal contacts
inside the nest between foragers returning with food and outgoing foragers available to leave the nest on the next foraging
trip. Here we present a feedback-based algorithm that captures the main features of data from field experiments in which
the rate of returning foragers was manipulated. The algorithm draws on our finding that the distribution of intervals
between successive ants returning to the nest is a Poisson process. We fitted the parameter that estimates the effect of each
returning forager on the rate at which outgoing foragers leave the nest. We found that correlations between observed rates
of returning foragers and simulated rates of outgoing foragers, using our model, were similar to those in the data. Our
simple stochastic model shows how the regulation of ant colony foraging can operate without spatial information,
describing a process at the level of individual ants that predicts the overall foraging activity of the colony.

Citation: Prabhakar B, Dektar KN, Gordon DM (2012) The Regulation of Ant Colony Foraging Activity without Spatial Information. PLoS Comput Biol 8(8):
e1002670. doi:10.1371/journal.pcbi.1002670
Editor: Iain D. Couzin, Princeton University, United States of America
Received January 23, 2012; Accepted July 15, 2012; Published August 23, 2012
Copyright: ß 2012 Prabhakar et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The work was funded by NSF grant IOS-0718631 to D.M.G, and by a grant from the CleanSlate program at Stanford University to B.P. and D.M.G. The
funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: dmgordon@stanford.edu

Introduction The most familiar example of feedback based on local
interaction in ants is recruitment to a food source using a
The fundamental question about the collective behavior of pheromone trail. In some ant species, an ant that finds food lays a
animals is how the actions of individuals add up to the dynamic chemical trail on its way back to the nest. Studies of the
behavior we observe. In many systems, including animal groups, algorithms used by ants in forming recruitment trails show that a
distributed networks are regulated using feedback based on local slight tendency on the part of other ants to move toward the trail
interactions. It is not yet clear how the analogies among diverse leads to the formation of trail systems [6,7,8] that can channel
complex systems reveal general underlying processes [1,2]. Here ants to the best food source [9] or trace the shortest path toward
we propose a simple stochastic model of collective behavior in the food [10]. Ants also use feedback from other forms of
ants. Our first goal is to account for the details of a particular interaction, such as brief antennal contact, in recruitment to food
system, as a step toward further insight on whether similar [11] and in other spatial decisions. The perception of the local
processes are at work in othersystems. A second goal of our work is density of digging ants generates branches in nest chambers [12].
to contribute to the study of collective behavior from the The rate of brief antennal contact as a cue to local density [13] is
perspective of evolutionary biology. If the outcome of collective used in spatial decisions, in combination with other information
behavior is ecologically important, then natural selection can act about location, such as the choice of new nest sites by acorn ants
on variation in that behavior. Modeling the parameters that [14,15].
produce collective behavior can provide the basis for detailed The regulation of activity by a simple stochastic process is
measures of variation among ant colonies. characteristic of many biological systems. Local interactions in
The best-studied algorithms for collective behavior in animals social insects, like those in other dynamical networks, can regulate
are those that regulate spatial patterns [3], based on local the flow or intensity of activity as well as its location or spatial
interactions that influence whether one animal stays close to pattern. For example, a social insect colony must adjust the
another [4]. Social insect colonies provide many fascinating allocation of individuals to various tasks, in response to changing
examples of collective behavior. There is no central control; no conditions [16]. Various models have been proposed to explain
insect directs the behavior of another. Like other social insects, the dynamics of the intensity of activity, or numbers of workers,
ants use local interactions to regulate colony behavior [5]. devoted to colony tasks [e.g. 17,18,19].

PLOS Computational Biology | www.ploscompbiol.org 1 August 2012 | Volume 8 | Issue 8 | e1002670

Regulation of Ant Colony Foraging Activity

Author Summary model with new data, from field experiments, that show how the
rate of at which outgoing foragers leave the nest changes in
Social insect colonies operate without any central control. response to changes in the rate at which returning foragers go back
Their collective behavior arises from local interactions to the nest.
among individuals. Here we present a simple stochastic
model of the regulation of foraging by harvester ant Methods
(Pogonomyrmex barbatus) colonies, which forage for
scattered seeds that one ant can retrieve on its own, so Measures of harvester ant foraging activity
there is no need for pheromone trails to specific locations. Experiments manipulating forager return rate were performed
Previous work shows that colony foraging activity is in August 2009 and August–September 2010 at the site of a long-
regulated in response to current food availability, using term study since 1985 of a population of P. barbatus near Rodeo,
the rate of brief antennal contacts inside the nest between New Mexico, USA. In 2009 there were 33 trials in 9 colonies on 8
foragers returning with food and outgoing foragers. Our
days, and in 2010 there were 29 trials in 8 colonies on 5 days, of
feedback-based algorithm estimates the effect of each
returning forager on the rate at which foragers leave the which 4 were the same colonies as in 2009, for a total of 62 trials.
nest. The model shows how the regulation of ant colony All colonies were mature, more than 5 years old (ages determined
foraging can operate without spatial information, describ- by yearly census; methods in [21]).
ing a process at the level of individual ants that predicts Returning foragers were prevented from returning from the nest
the overall foraging activity of the colony. in minutes 4–7 of a 20-min observation; methods were the same as
in [28,29]. Rates of returning and outgoing foragers crossing an
imaginary line along the trail were measured from video film using
Here we present a simple stochastic model that explains the an image analysis system developed by Martin Stumpe (http://
process underlying the regulation of foraging activity in harvester www.antracks.org). This image analysis system made it possible to
ant (Pogonomyrmex barbatus) colonies. Foraging activity, the numbers measure foraging rates accurately on a shorter timescale than in
of ants currently foraging, changes from moment to moment previous work.
within a foraging period and from day to day. This species does Most colonies use more than one foraging direction on a given
not use pheromone trails to recruit to localized food sources. The day [21]. We filmed all trails and used the combined foraging rates
ants forage for seeds that are scattered by wind and flooding [20], for all trails. Foraging rates were calculated separately for the
not distributed in patches, and a single ant can retrieve a seed on periods before (0–240 sec), during (240–430 sec), and after (500–
its own. The model uses an algorithm based on local interactions 1100 sec) the removal of returning foragers, as in previous work
among individuals in the form of brief antennal contacts, without [28,29]. The small interval between 430 and 500 sec allows for the
any spatial information such as pheromone trails. time it takes ants passing the camera on the foraging trail to reach
Harvester ants searching for food in the desert undergo the nest. To correct for differences among trails in the distance
desiccation, and the ants obtain water from metabolizing the fats between the camera and the nest, we found the average time, out
in the seeds they eat. Thus a colony must spend water to obtain of 5 observations, for an ant to travel back to the nest from the
water, as well as food. The intensity of foraging is regulated from point where the foraging trail was filmed. To adjust for the
moment to moment, and from day to day, to adjust foraging distance between the camera and the nest, we then subtracted this
activity to current food availability, while maintaining sufficient travel time from the counts for outgoing foragers and added it to
numbers of ants foraging to compete with neighbors for foraging counts for returning foragers.
area [21]. The average error in the accuracy of the image analysis software
A long-term study of the foraging ecology of this species has in counting foraging rate was 7.3%, estimated by comparing 66
shown how the moment-to-moment regulation of foraging is counts made by observers from 500 frames (about 17 sec) of 44
accomplished. Regulation depends on feedback from returning video films with counts made by the image analysis software. Most
foragers, who stimulate the outgoing foragers to leave on the next errors were due to an extraneous object or shadow in the film at
trip. Forager return rate corresponds to food availability, because the point at which ants crossed the imaginary line where ants were
foragers almost always continue to search until they find a seed, counted. There was no bias toward counting more or fewer ants
then immediately bring it back to the nest [22,23]. The more food than actually crossed the line.
is available, the less time foragers spend searching and the more
rapidly they return to the nest. Results
The crucial interactions between returning and outgong
foragers take place in a narrow entrance tunnel, 5–10 cm long, Model of the regulation of foraging activity
that leads to a deeper entrance chamber. Observations with a We first tested whether the return of foragers to the nest could
videoscope show that returning foragers drop their seeds in the be described as a Poisson process. To determine the distribution of
tunnel, and then other ants pick up the seeds and take them deeper intervals between the arrival of foragers at the nest, we used the
into the nest. Once the returning forager has dropped its seed, it data from the period before the removal of returning foragers (60–
becomes an outgoing forager, available to go out on its next trip. 240 sec) in 39 trials. We calculated the fit with an exponential
Experiments using artificial ant mimics coated with extracts of ant distribution of the empirical distribution of the interarrival times of
cuticular hydrocarbons [24,25], and experiments manipulating the the returning foragers, and measured the error using the total
rate of forager return [26,27,28,29] show that how quickly an variation distance [30] between the two distributions. We found
outgoing forager leaves on its next trip depends on its interactions that the return of foragers to the nest can be described as a Poisson
with returning foragers. Foraging activity is more closely regulated process: the distribution of intervals between returning foragers is
when foraging rates are high, above a baseline rate at which exponential (Fig. 1) and independent of the spacing between
foragers leave independently of the rate of forager return [29]. foragers in the previous interval. We found the mean (SE) error for
We developed a model that takes into account previous work on the 39 trials to be 0.056 (0.009) with a better fit at high foraging
the regulation of foraging. We compared simulations using the rates.



Figure 1. Poisson distribution of intervals between the arrival of successive returning foragers at the nest. The figure shows
representative data from one trial in which returning foragers were removed. Intervals are shown in video frames; each frame is 1/30 sec. The y-axis is
the probability that the interval i between successive returning foragers exceeds t frames. The solid blue line corresponds to the real data and the
dotted red line corresponds to the exponential fit y = e20.0326t with a mean separation time of 1/0.0326 or 30.67 frames, equal to 1.02 seconds.
doi:10.1371/journal.pcbi.1002670.g001

We began with a simple linear model in which the rate of denote the number of returning food-bearing foragers in the nth
outgoing foragers x(t) depends on the the rate of returning foragers time slot, and let Dn denote the number of outgoing foragers
plus a constant base rate: leaving the nest. The rate at which ants leave the nest in the nth slot
is an, n = 1, 2, …. We assume that an$a.0 for n = 1, 2, …, where
(1) l(t) = lac f(t,t) a is a parameter. The number of departures at the end of the nth
(2) x(t) = Poisson (lb)+Poisson (l (t)) time slot, Dn, was set equal to a Poisson random variable of mean
an. Given the an, the Dn are statistically independent of each other
where lb is a baseline rate of outgoing foragers, independent of and of An. The dynamics of an are described by:
the rate at which other foragers return; lac sets the number of
outgoing foragers per returning forager; and f(t, t) is the number of (3) an = max (an212qDn21+cAn2d, a), a0 = 0
returning foragers between times t2t and t. (4) Dn,Poisson (an)
For this initial, linear model, the most important parameter in
predicting the rate of outgoing foragers is t. Because this model In developing the model, we sought to facilitate its future use
integrates over all returns in time t, fitting the model requires us to to examine variation among colonies within this species, for
choose a value of t that gives an equally good fit to the observed example in the baseline rate at which ants leave the nest even
data over a range of foraging rates. This is difficult because when no ants return [29], and to take into account three features
foraging rates vary greatly among colonies, days, and moment-to- of the observed behavior of the ants in response to experimental
moment changes in the conditions that foragers encounter manipulation of foraging rate [28,29]. First, there is a lag in the
[5,27,29]. We thus elaborated this model into a second one that recovery of the rate of outgoing foragers in response to recovery
avoids the integration of instantaneous arrivals over a time interval of the rate of returning foragers after a decline [28,29] (Fig. 2).
that is not uniform across different foraging conditions. Moreover, Second, we introduced q, the extent to which each departing
the model below captures the effect of a single returning forager forager empties the nest entrance of available foragers, because
and so explains the process at the level of individual ants, as for observations with a videoscope inside the nest show that
example in other non-linear models that describe pheromone trail outgoing ants are crowded in a small tunnel, so that once each
foraging by ants [7]. outgoing ant departs it takes some time for the next outgoing ant
The model operates in discrete time: returning foragers are to move to the top of the tunnel where it can meet returning
observed in successive and equal time slots. We denote the rate of foragers. Third, we include the decay parameter d because
outgoing foragers as ‘a’, which increases by an amount c.0 for experimental results show that the response to returning ants is
each returning food-bearing forager. Alpha decreases by an weaker after some time has elapsed since the last ant returned
amount q.0 for each forager that leaves the nest, because the [24].
departure of each outgoing forager decreases the number of
outgoing foragers in the queue at the nest entrance available to Comparison of model and data
meet returning foragers. Alpha decays by an amount d.0 during We compared the simulated output of the model with the
each time slot, to reflect the lack of response to very low data from field experiments on the response of outgoing foragers
interaction rate [22]. Finally, a has a lower bound, a., to reflect the to a range of rates of returning foragers (Fig. 2). Using as input
observation that outgoing foragers leave the nest at a fixed low rate the data on the rate of returning foragers, we generated the
even when no foragers return for a while [28,29]. simulated rate of outgoing foragers, adjusting one parameter
We assume that arrivals occur at the beginning of time slots and and evaluating the resulting match with the observed rate of
departures occur at the end of time slots. For n = 1, 2, …, let An outgoing foragers.



Figure 2. Comparison of observed and simulated foraging rates. The rate of returning foragers was experimentally decreased by removing
the returning foragers, leading to a decrease in the rate at which outgoing foragers left the nest. Each figure shows data from one trial. Returning
foragers were removed from 240–420 sec, during the interval indicated by the horizontal black line, and then allowed to return to the nest
undisturbed for the remainder of the trial. The red line shows the observed rate of returning foragers, the blue line shows the observed rate of
outgoing foragers, and the green line shows the simulated rate of outgoing foragers. a, high foraging rate (mean rate returning foragers 0.807 ants/
sec); b, low foraging rate (mean rate returning foragers 0.169 ants/sec).

The model has four parameters: a, c, q and d. We examined the fit here; however, empirical studies show that d may be an important
between model and data for one parameter, c. We thus fixed a., q parameter because it may vary by colony [29], or in response to
and d and varied c. As with any birth-death process, the ratio of c to e variation in environmental conditions that could affect the rate of
determines the distribution of {an}. We set q to 0.05 to keep the decay of chemical cues such as the cuticular hydrocarbons that ants
range of values of an within the range of observed foraging rates assess by antennal contact [24]. Similarly, a, the baserate of
(0.15 to 1.2 ants per sec). We set d to 0 for the simulations reported foraging, was very small, equal to 0.01 ants per second [27].



Figure 3. Correlation of rates of outgoing and returning foragers as a function of level of foraging activity. Each point shows, for one
trial, the coefficient of correlation between the smoothed rates of returning and outgoing foragers. The x-axis shows the mean rate of returning
foragers over the entire trial. Blue diamonds show the coefficients of correlation between observed rates of returning foragers and simulated rates of
outgoing foragers. Red squares show the coefficients of correlation between observed rates of returning foragers and observed rates of outgoing
foragers. The upper line shows the least-squares fit to the increase with foraging activity of the correlation between observed returning and
simulated outgoing foraging rates points (blue); the lower line shows the least-squares fit to the increase with foraging activity of the correlation
between observed returning and observed outgoing foraging rates(red).

To choose c for the given values of a, q, and d, we used for each or 2 remaining trials for that colony in the same year. We deleted
of the 62 experimental trials the data on returning foragers and one of the remaining trials in cases when foraging rates were much
equations (3) and (4) to generate a simulated rate of outgoing lower than for other trials with the same colony.
foragers. We swept across values of c from 0.01 to 0.25, and found The results of this comparison indicate that our estimate of best
the relative root-mean-square error (RMSE) [31] between the c generates the observed outgoing forager rate within the same
simulated and observed rates of outgoing foragers in each time range of error as the error generated by randomness in forager
interval for each of 200 iterations. Each iteration produces a departures. The mean (SE) change in RMSE among simulated
different output trace because of the independent Poisson random runs was 13% (0.0495) at low foraging rates and 2.6% (0.01) at
variable generated at equation (4). We chose the c with the lowest high foraging rates. The mean (SE) change in RMSE obtained by
average RMSE, over the 200 iterations, between simulated and varying u among trials of the same colony was 15.5% (0.019).
observed rates of outgoing foragers. The RMSE for the best c for We examined whether the correlation between rates of
each trial ranged from 0.237 to 4.9, and the mean (SE) RMSE for returning and outgoing foragers was at least as high in the
the 62 values chosen was 0.602 (0.077). simulation as in the data. To do this we compared the correlation
To evaluate how well our estimate of c captured, for a given between observed rates of returning and simulated rates of
colony, the effect of each returning forager on the rate of outgoing outgoing foragers, using the best value of c, with those between
foragers, we compared the error among runs of the simulation, observed rates of returning and observed rates of outgoing
due to randomness in the departures at equation (4), with the error foragers. We calculated the correlation between the observed
produced by varying c. To do this we compared the RMSE rates of returning foragers and the simulated rate of outgoing
between observed and simulated rates of outgoing foragers among foragers, in all 62 trials, by applying a moving-average rect filter
simulated runs with the RMSE for different values of c among [32] with a radius of 25 time slots, and then finding the empirical
trials of the same colony. We first found for each trial the mean correlation coefficent between the smoothed traces [31]. We
range in RMSE, between observed and simulated rates of calculated in the same way the correlation between observed rates
outgoing foragers, among 200 iterations of the simulation using of returning and outgoing foragers. The simulated rates of
the same value of c. To estimate the difference among runs of the outgoing foragers led to correlation coefficients higher than those
simulation, we used each trace of returning foragers to produce 3 between observed rates of returning and outgoing foragers (t-test,
different traces of simulated outgoing foragers (A,B,C). We found t = 8.98, p,0.001; Fig. 3).
the RMSE for A vs B and A vs C, and repeated this 200 times per Because previous work showed that the rate of outgoing foragers
trial at foraging rates ranging, as did the observed rates, from low tracks the rate of forager return more closely when foraging rates
(0.1 returning ants/sec) to high (1.2 returning ants/sec). We then are high, we examined how the correlation of rates of returning
found the mean RMSE when varying values of the best c among and outgoing foragers varies with foraging rate. The magnitude of
trials for the same colony. To do this we chose arbitrarily, for each the correlation coefficient between observed returning and
of the 13 colonies, 2 of the 3 or 4 trials. For those 2 trials we found simulated outgoing foragers increased with the mean rate of
the average of the values of c, then found the RMSE as above [31] returning foragers (Spearmann’s rank correlation, n = 62,
for simulated values of rates of outgoing foragers for each of the 1 z = 24.04, p = 0.0001, blue points in Fig. 3). The magnitude of



the correlation coefficient for the observed returning and outgoing foragers with the observed rate of outgoing foragers (Fig. 3). In
rates did not increase significantly with the mean rate of returning addition, the coefficient of correlation with the observed rate of
foragers (Spearmann’s rank correlation, n = 62, z = 21.34, returning foragers increased significantly with foraging rate for the
p = 0.18, red points in Fig. 3). simulated rate of outgoing foragers but does not increase
significantly for the observed rate of outgoing foragers (Fig. 3).
Discussion The higher correlation in the simulation than in the data occurs
because our model (equations 3 and 4) does not capture all of the
Our model provides a way to evaluate quantitatively the factors that produce the actual rate at which ants leave the nest.
response of individuals to local interactions. It explains how ant For example, the structure of each nest probably affects the flow of
colonies regulate foraging activity from moment to moment, in ants in and out of the nest entrance, which in turn may affect the
response to current food availability, without any central control rate of interaction between outgoing and returning ants. A
or any spatial information on the location of food. Changes in a question for future work is whether nonlinear effects of nest
colony’s foraging activity from moment to moment, and from day structure influence the relation between overall foraging rate and
to day, show a predictable response to changes in forager return the correlation of the rates of returning and outgoing foragers. Our
rate, despite considerable stochasticity (Fig. 2). Our results show model assumes the same relation between the rate of incoming and
that much of these changes in colony foraging activity can be rate of outgoing foragers for all nests, and thus does not take into
explained by the effect of each returning forager on the probability account the local influence of nest structure. Weather conditions
that outgoing foragers leave the nest to search for food. also influence foraging activity, leading to day-to-day fluctuations
The model is analyzable because the distribution of returning in the foraging activity of a given colony [28].
foragers is well-approximated by a Poisson process at high foraging
The process described here is analogous to those operating in
rates. Like cars on highways, returning foragers travel at different
many other distributed networks, from computer networks to
velocities and overtake each other; as [33] shows, this produces a
neural integrators, that regulate activity through the rate of
Poisson process.
interaction [34,35]. Further work is needed to determine the
The simulated data provided by our model capture many of the
details of the correspondence among these analogous systems; for
features of a rich body of empirical results from a long-term study
example, in this system, the Poisson distribution of returning
of the foraging behavior of harvester ant colonies. The model
foragers is crucial.
provides a simulated rate of outgoing foragers, in response to the
The model presented here contributes to the study of the
rate of returning foragers, that is reasonably similar to the
evolution of collective behavior in harvester ants, because it can be
observed rate (Fig. 2), producing a close correlation between the
used to guide empirical measurement of differences among
rates of returning and outgoing foragers.
colonies in the regulation of foraging [29], by examining whether
Another similarity between the model and observation is in the
colonies tend to show characteristic parameter values. Harvester
contrast between colony behavior when food availability is high, so
ant colonies differ in foraging behavior, and such differences
that foragers find food quickly and the rate of forager return is
persist from year to year as the colony grows older [21,29].
high, and colony behavior when food availability is low, so that
Heritable variation among colonies in ecological relations, such as
foragers find food slowly and the rate of forager return is low.
Previous work on harvester ant foraging showed that rates of the regulation of foraging, is the source of variation in fitness [36].
outgoing foragers are more closely adjusted to rates of returning Future work will examine differences among colonies in the
foragers when foraging rates are high [29]. The same was true of response to interactions of returning and outgoing foragers. Small
our model. For example, Fig. 2 shows the data for two differences in the ants’ response to local interactions may lead to
representative cases. When foraging rates are high (Fig. 2A), on ecologically important differences among colonies that shape the
a day when food availability is high and foragers find it quickly, the evolution of collective behavior.
rate of returning and outgoing foragers is more closely matched
than when foraging rates are low (Fig. 2B), so that foragers find Acknowledgments
food more slowly and return less frequently. The closer fit between We thank Martin Stumpe for developing the image analysis software. We
outgoing and returning foragers at high foraging rates occurs thank Sam Crow, Jordan Haarmsa, Mattias Lanas, Doo Young Lee,
because the range of values of an tends to be much smaller and LeAnn Howard, Tom Yue, and Uyhunh Ung for their assistance with field
closer to a at low foraging rates than at high foraging rates, and work, and Noa Pinter-Wollman and William Flesch for comments on the
this causes Dn to be close to zero and, hence, less correlated with manuscript.
A n.
However, the simulation generally produces a closer correlation Author Contributions
between the rates of returning and outgoing foragers than is Conceived and designed the experiments: DMG. Performed the
observed in the data. The correlation coefficients for the observed experiments: DMG KND. Analyzed the data: DMG KND BP.
rate of returning foragers and the simulated rate of outgoing Contributed reagents/materials/analysis tools: DMG BP. Wrote the
foragers are higher than those for the observed rate of returning paper: DMG BP.

References
1. Camazine S, Deneubourg J-L, Franks N R, Sneyd J, Theraulaz G, et al. (2003) Self- 6. Detrain C, Deneubourg J-L (2008) Collective decision-making and foraging
Organization in biological systems. Princeton: Princeton University Press. 529 p. patterns in ants and honeybees. Adv In Insect Phys 35:123–173.
2. Gordon DM (2007) Control without hierarchy. Nature 4468:143. 7. Sumpter DJT, Pratt SC (2003) A modelling framework for understanding social
3. Sumpter, D.J T (2010) Collective animal behavior. Princeton: Princeton insect foraging. Behav Ecol Sociobiol 53:131–144.
University Press. 298 p. 8. Sumpter DJT, Beekman M (2003) From nonlinearity to optimality: pheromone
4. Torney C, Neufeld., Couzin ID (2009) Context-dependent interaction leads to trail foraging by ants. Anim Behav 66: 273–280.
emergent search behavior in social aggregates. Proc Natl Acad Sci USA 9. Beckers R JL, Deneubourg JL, Goss S, PasteelsJ M (1990)Collective decision-
52:22055–2206. making through food recruitment. Insect Soc 37:258–267.
5. Gordon DM (2010) Ant encounters: interaction networks and colony behavior. 10. Dorigo M, Di Caro G, Gambardella LM (1999) Ant algorithms for discrete
Primers in complex systems. Princeton: Princeton University Press. 184 p. optimization. Artif Life 5:137–172.



11. Mailleux AC, Buffin A, Detrain C, Deneubourg J-L (2010) Recruiter or recruit: 24. Greene MJ, Gordon DM(2003) Cuticular hydrocarbons inform task decisions.
who boosts the recruitment in starved nests in mass foraging ants? Anim Behav Nature 423:32.
79:31–35. 25. Greene MJ, Gordon DM (2007) Interaction rate informs harvester ant task
12. Toffin E, Di Paolo D., Campo A, Detrain C, Deneubourg J-L (2009) Shape decisions. Behav Ecol 18:451–455.
transition during nest digging in ants. Procs Natl Acad Sci U S A 106:18816– 26. Gordon DM (2002) The regulation of foraging activity in red harvester ant
18620. colonies. Am Nat 159:509–518.
13. Gordon DM, Paul REH, Thorpe K. (1993) What is the function of encounter 27. Schafer RJ, Holmes S, Gordon DM (2006) Forager activation and food
patterns in ant colonies? Anim Behav 45:1083–1100. availability in harvester ants. Anim Behav 71:815–822.
14. Pratt S(2005) Quorum sensing by encounter rates in the ant Temnothorax 28. Gordon DM, Holmes S, Nacu S (2008) The short-term regulation of foraging
albipennis. Behav Ecol 16:488–496. inharvester ants. Behav Ecol 19:217–222.
15. Pratt S, Sumpter D (2006) A tunable algorithm for collective decision-making. 29. Gordon DM, Guetz A., Greene MJ, Holmes, S. (2011) Colony variation in the
Proc Nat Acad Sci U S A 103:15906–15910. collective regulation of foraging by harvester ants. Behav Ecol 22: 429–435.
16. Gordon D M (1996) The organization of work in social insect colonies. Nature 30. Durrett R (1991) Probability: Theory and Examples. Cambridge Series in
380:121–124. Statistical and Probabilistic Mathematics. New York: Cambridge University
Press.
17. Pacala SW, Gordon DM, Godfray HCJ (1996) Effects of social group size on
31. Papoulis A (1965) Probability, Random Variables, and Stochastic Processes.
information transfer and task allocation. Evol Ecol 10:127–165.
New York: McGraw-Hill. 841 p.
18. Anderson C, Ratnieks F (1999) Task partitioning in insect colonies. 1. Effect of 32. Foley J, van Dam A, Feiner S, Hughes J (1995) Computer Graphics: Principles
colony size on queuing delay and colony ergonomic efficiency. Am Nat and Practice, 2nd Edition. Reading: C. Addison-Wesle. 1175 p.
154:521–535. 33. Breiman L (1963) The Poisson tendency in traffic distribution. Annals Math Stat
19. Karsai I, Schmikl T (2011) Regulation of task partitioning by a ‘‘common 34: 308–311.
stomach’’: a model of nest construction in social wasps. Behav Ecol 22:819–830. 34. Alizadeh M, Atikoglu B, Kabbani A, Lakshmikantha A, Pan R, et al. (2008)
20. Gordon DM (1993) The spatial scale of seed collection by harvester ants. Data center transport mechanisms: congestion control theory and IEEE
Oecologia 95: 479–487. standardization. In: Proceedings of the 46th Annual Allerton Conference on
21. Gordon D M, Kulig AW (1996) Founding, foraging and fighting: colony size and Communications, Control and Computing.
the spatial distribution of harvester ant nests. Ecol 77:2393–2409. 35. Goldman MS, Compte A, Wang X-J (2009) Neural integrator models. In:
22. Gordon DM (1991) Behavioral flexibility and the foraging ecology of seed-eating Encyclopedia of Neuroscience 6. Squire LR, editor. Oxford: Academic Press.
ants. Am Nat 138:379–411. pp. 165–178.
23. Beverly B, McLendon H, Nacu S, Holmes S, Gordon DM(2009) How site 36. Dingemanse NJ, Kazem AJN, Reale D, Wright J (2010) Behavioural reaction
fidelity leads to individual differences in the foraging activity of harvester ants. norms: animal personality meets individual plasticity. Trends Ecol Evol 25:81–
Behav Ecol 20:633–638. 90.


The regulation of ant colony foraging activity without spatial information

More Related Content

What's hot (20)

Viewers also liked (9)

Similar to The regulation of ant colony foraging activity without spatial information (20)

More from Yury Chemerkin (20)

Recently uploaded (20)

The regulation of ant colony foraging activity without spatial information