Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
Ants work harder during consensus
decision-making in small groups
Adam L. Cronin1 and Martin C. Stumpe2
rsif.royalsocietypublishing.org
1
2
Research
Cite this article: Cronin AL, Stumpe MC. 2014
Ants work harder during consensus
decision-making in small groups. J. R. Soc.
Interface 11: 20140641.
http://dx.doi.org/10.1098/rsif.2014.0641
Received: 16 June 2014
Accepted: 24 June 2014
Subject Areas:
systems biology
Keywords:
ANTRACKS, group-size, house-hunting,
Formicidae, swarm intelligence
Author for correspondence:
Adam L. Cronin
e-mail: [email protected]
Electronic supplementary material is available
at http://dx.doi.org/10.1098/rsif.2014.0641 or
via http://rsif.royalsocietypublishing.org.
United Graduate School of Agricultural Sciences, Iwate University, 3-18-8 Ueda, Morioka 020-8550, Japan
AnTracks Computer Vision Systems, Mountain View, CA, USA
Individuals derive many benefits from being social, one of which is improved
accuracy of decision-making, the so-called ‘wisdom of the crowds’ effect. This
advantage arises because larger groups can pool information from more individuals. At present, limited empirical data indicate that larger groups outperform
smaller ones during consensus decision-making in human and non-human animals. Inaccurate decisions can lead to significant costs, and we might therefore
expect individuals in small groups to employ mechanisms to compensate for
the lack of numbers. Small groups may be able to maintain decision accuracy
if individuals are better informed than those in larger groups and/or by increasing the proportion of the group involved in collective decision-making relative
to larger groups. In this study, we use interactive computer vision software to
investigate individual contributions to consensus decision-making during
house-hunting in different sized groups of the ant Myrmecina nipponica. We
show that individuals in small colonies invest greater effort in the consensus
decision process than those in large colonies and should be better informed as
a result. This may act to ameliorate the limitations of group size, but could
leave smaller groups more susceptible to additional stresses.
1. Introduction
Animals benefit in a variety of ways from being social, but must overcome challenges associated with coordinating group members during collective actions
[1,2]. During tasks such as collective movements, groups must reach a consensus
over which action to pursue if group integrity is to be maintained [1,3–6]. These
decisions may be shared, with all group members contributing to the outcome, or
unshared, with a distinct leader always initiating the decision [7]. In a wide range
of animals including fish [8], some primates [5] and insects [9,10], decisions are
shared among a variable proportion of the group, and a group-level consensus
manifests as an emergent product of the combined actions of multiple individuals
[1,4]. Shared decisions allow information to be pooled, and this ‘wisdom of the
crowds’ effect has long been recognized as a means by which groups can make
more accurate decisions than individuals [3,11–16]. Group size varies markedly
between and within animal societies, however, raising the question of how
small groups perform relative to large groups, and how investment in decisionmaking varies with group size. While modelling studies predict larger groups
will outperform smaller ones during decision-making, there are few empirical
studies to support this, and fewer still that shed any light on how this benefit is
achieved at the individual level [17–19].
During consensus decision-making, individuals with information regarding a
possible course of action advertise their preference by ‘voting’ through a range of
means designed to recruit other group members. This voting behaviour is exemplified by the waggle-dance of honeybees [10], but includes chemical trails [20],
vocalizations [21], posturing [5] and a range of other means [22–24]. These
votes are mimetically reinforced by other group members or countered by votes
for alternative options, though eventually one option emerges as the one favoured
by the majority. The response to this feedback is often step-like, such that the propensity for an individual to perform an action increases markedly once a threshold
number of individuals, a quorum, is already performing that action [6]. Quorum
decisions are a widespread and effective means of collective decision-making
& 2014 The Author(s) Published by the Royal Society. All rights reserved.
Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
2. Material and methods
2.1. Colony collection and maintenance
Entire colonies of M. nipponica were collected from patches of moss
and the bases of ferns in broadleaf forest near Chitose City in
Hokkaido, northern Japan (428470 N 1418340 E, altitude approx.
100 m) in September 2012 and 2013. A total of 68 colonies were collected overall, with a mean colony size of 33 + 18 individuals.
Colonies were housed in 10 10 3 cm plastic boxes floored
with plaster that was kept moist. Each box contained a single nest
consisting of a microscope slide covered with a red filter mounted
on a 2 mm high circlet of foam with a small (3 mm) opening. Ants
were kept at room temperature (approx. 208C) and provided with
ad libitum water and sugar/water solution, and fed mealworm
pieces every few days. All ants were individually marked with
different coloured spots on the head, thorax and gaster using
Mitsubishi paint-marker pen paint applied with a fine brush. See
also [47,48] for basic housing and experimental methods.
2.2. Tracking of individuals during relocations
Colonies of M. nipponica can be induced to relocate to a new nest
site by removing their artificial home nest (which sits atop the
plaster). An identical nest was provided in a new box separated
by an empty navigation chamber (figure 1). Colonies rendered
homeless send out scouts to find a new nesting site. These scouts
‘vote’ on available sites by laying chemical trails to their favoured
site, which in turn attracts further scouts [47]. Once the number of
scouts at one site reaches a threshold, a ‘quorum’ is achieved [6].
This triggers a phase shift, and individuals rapidly undergo a
behavioural switch to conclude the relocation marked by the transport of brood to the new site. Brood transport does not begin until a
consensus decision has been achieved, and this is dependent on
achievement of a quorum of individuals at the new site [47,48].
The relocation process in M. nipponica can be divided into three
phases [48]. The ‘discovery’ phase is defined as the time from
removal of the home nest to the first ant entering the new nest.
The ‘assessment’ phase is defined as the time from first entry to
the new nest until the first quorum response. The quorum point
is defined as the moment at which the first ant to switch to
brood transport departs from the destination nest immediately
2
J. R. Soc. Interface 11: 20140641
remains unclear to what degree decision-making in social
insects relies on specialists. Furthermore, colony size can
influence the efficiency of a wide range of group tasks in
social insects ([45], but see [46]), and individual contributions
might also vary with group size.
In this study, we shed light on these questions using
interactive computer vision software to analyse in detail the
consensus decision and colony relocation process at the individual level. We employ as our model system the small colony
Japanese ant Myrmecina nipponica. This species is ideally
suited to this purpose as entire colonies can be observed
under controlled laboratory conditions, and previous studies
have indicated that large and small colonies perform relocations
with similar efficiency [47], suggesting a capacity for adaptive
optimization of decision-making in colonies of different size.
Here, we explore three main questions: (i) how is individual
input to decision-making distributed within colonies, (ii) does
this differ in large and smaller colonies, and (iii) does the
decision process rely on specialists, that is, ants that consistently
take a majority role, or are roles flexible? We estimate individual
‘effort’ by quantifying ant movement and behaviour during the
house-hunting process, and contrast individual roles in colonies
of different natural and manipulated sizes.
rsif.royalsocietypublishing.org
[4,6] and can be tuned to environmental conditions to
emphasize decision accuracy or speed [25].
Consensus decisions are predicted to be more accurate in
larger groups because more individuals can collect more
information regarding the environment per unit time and
are better able to process this information [6,26,27]. However,
empirical data demonstrate that in some cases the proportion
of the group involved in decision-making may be quite low
(e.g. only around 5% of the colony act as scouts during
house-hunting in honeybees and Temnothorax ants [28,29]).
There may thus be considerable scope for adaptive variation
in the proportion of the group that contributes to the decision
process [30]. Furthermore, in addition to the question of how
many individuals contribute to the consensus process, there
is the question of who contributes. Information quality can
vary between individuals, and decision accuracy can be
maintained in smaller groups if the individuals that contribute to the decision are better informed [1,31]. In the extreme
case, a single expert can outperform a poorly informed collective [32]. In shared decisions, empirical data indicate
that the contribution of individuals can be shared equally
among all individuals or reside predominantly in a subset
of the group [5,33–36]. The strength of individual influence
may also vary with group size, with individuals having a
greater potential influence in smaller groups [37]. This
scope for variation in the proportion of individuals involved
in decisions, and the magnitude of the contribution that
individuals make to the decision process, may give smaller
groups some capacity to ameliorate the limitations of group
size. Indeed, Franks et al. [38] showed that small and large
colonies of Temnothorax ants were equally adept at solving
a nest selection problem, though in contrast, Ward et al. [18]
showed that larger groups made better and faster decisions
in fish. Understanding how individual contributions to consensus decision-making vary in conjunction with group size
should provide valuable insights into the mechanics underpinning collective processes, but this can only be elucidated
with detailed quantifications of individual contributions to
the decision process.
Social insects are excellent subjects for studies of collective
behaviour and the influence of group size on consensus
decision-making. Colony size varies over several orders of
magnitude, and the diversity of social systems represented is
unrivalled. Insects need to find a new home when their present
nest becomes unsuitable or during colony reproduction by fission [27,39], and studies of the consensus decision process
during house-hunting by bees and ants have been particularly
illuminating (reviewed in [4,9,10]). While leadership can be
highly variable in vertebrate groups, social insects are typically
regarded as being self-organized, such that group behaviour
emerges as the collective product of individual responses to
local information without the need for global control [3]. However, social insect colonies are often characterized by distinct
classes of individuals derived from different age groups, genetic lineages and behavioural or morphological castes, each
of which has the opportunity to contribute differentially to
the decision process. Indeed, it has long been recognized that
‘elites’ or ‘key individuals’ may contribute disproportionately
to collective processes [28,40–42]. On the other hand, for
effective exploitation of the ‘wisdom of the crowds’ effect in
decision-making, individual contributions should be independent, and increasing the diversity of individual contributors
can be more beneficial than adding expertise [43,44]. It thus
Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
a
2.4. Data analysis
a
N
D
prior to her maiden transport voyage. From the quorum point
until the end of the relocation is termed the ‘transport’ phase.
Relocations were deemed completed when all brood had
been transported.
Web cameras (Elecom Ucam-DLA200H) positioned above the
nest entrances were used with the motion-detection software iSpy
(www.ispyconnect.com) to track entry/exit events (figure 1). The
identity and timing of individual ants entering and exiting the
new nest(s) and transporting brood were manually scored from
videos. Ant movements in the navigation chamber were recorded
for the duration of the relocation using a high-resolution webcamera (Logitech C920). These videos were used to quantify
individual ant movements in detail using the software ANTRACKS
(www.antracks.org). This software accurately tracks individual
movements (x/y pixel coordinates for each video frame) as ‘trajectories’ representing one voyage (here from entry to exit of the
navigation chamber). Each trajectory was matched with data
from entry/exit events and assigned to an individually marked
ant, and then scored as whether or not it involved transport of a
brood item and whether or not it followed the pheromone trail.
Trajectories were considered to follow the trail if at least 80% of
the trajectory followed the most frequently used path in the navigation chamber. The most frequented path in each relocation
was inferred using the ‘density map’ function in ANTRACKS,
which in all cases clearly indicated the most frequently used
path in the navigation box between the source and destination
nests (see the electronic supplementary material).
2.3. Experimental approach
Relocations were performed with the goal of quantifying the
relative contribution of all individuals to the decision-making
process, exploring the influence of colony size on this distribution and investigating the adaptability of individual roles.
To this end, nine relocations were analysed in detail for six colonies. Relocations were divided into three classes: large colonies,
small colonies and split colonies. The former two categories
were natural colonies of varying size (34 ants for large colonies
and 15 – 18 ants for small colonies; table 1). Because natural
colonies of different sizes can differ in other size-correlated
attributes such as experience [49] we also performed relocations
with manipulated (split) colonies. Split colonies were the same
colonies as those used in large relocations, but were halved
in size (17 ants), such that small and split colonies were of
equivalent size. To standardize which individuals were removed
and control for possible stochastic effects of the removal of
key individuals, we deliberately selected the most active ants
(as adjudged by distance covered during full-size relocations)
to be removed.
3. Results
All nine relocations were completed successfully, with a
total of between 119 and 469 trajectories recorded (mean ¼
260 + 98; table 1). As might be expected from the greater
number of individuals, large colonies tended to generate
more trajectories than split or small colonies (means:
large ¼ 347 + 112, split ¼ 189 + 77 and small ¼ 245 + 20;
table 1), though this difference was significant only between
large and split colonies (table 2).
3.1. Relocation mechanics and distribution of effort
3.1.1. Trajectory alignment, speed and distance
Trajectory speed increased both with time (t2137 ¼ 8.09, p ,
0.001) and individual experience (t2136 ¼ 5.60, p , 0.001),
3
J. R. Soc. Interface 11: 20140641
Figure 1. Experimental set-up for relocations. The apparatus consisted of
three nesting boxes (10 10 3 cm) linked via small holes. Ants relocated
from the source nest-box (S) to the destination nest (D) via the navigation
box (N). Ants were prevented from accessing other parts of the destination
nest-box other than the nest itself. Web cameras (a) recorded exit and
entrance events while a further high-resolution camera (b) recorded all movements in the navigation box (see also main text). The area available to ants
during the relocations is indicated in grey: the area in the destination box (D)
was enclosed within the artificial nest.
Data from ANTRACKS were converted into spatial objects using
the sp and trip packages in R v. 3.0.2 [50,51] after manual
scaling of distances. The mean speed and distance of individual
trajectories were obtained from these spatial objects. Individual contributions to the relocation process were assessed
using two metrics, (i) the distance covered by individuals and
(ii) the number of brood transports undertaken. The distance covered by each individual (hereafter termed effort) was recorded
only in the central navigation chamber, and an assumption of
our analysis is that activity beyond this area was directly proportional to activity within it. Observations suggest that this is
reasonable as the activity level of individual ants was consistent
across regions. Activity outside the navigation chamber was only
a small proportion of overall activity as movement in the source
nest-box was largely limited to between the site of the old nest
and the entrance to the navigation chamber, and the area of
the new nest was small. To account for variation in the duration of phases when comparing relocation events, distance was
scaled by relative duration of the relocation or phase (i.e. divided
by the proportionate difference between the focal and longest
relocation). This metric was termed ‘scaled effort’ and was our
primary measure of individual contributions to the decision
and relocation process.
Generalized linear mixed models (GLMM) were implemented using the lme4 package in R v. 3.0.2 [51] with colony
identity as a random factor. Analyses of quorum thresholds,
scout and transporters numbers were modelled using a Poisson
errors, whereas binary data such as trail following were treated
assuming a binomial error distribution. Other data including
speeds and durations were treated as Gaussian, and log transformed when plots of residuals indicated a violation of the
assumption of equal variance. Scaled effort data were continuous
and zero-inflated and thus we employed a compound Poisson
linear mixed model using the cplm package in R [52]. In analyses
of speed, individual identity was added as a nested random
factor within colony. In analyses of overall trends during relocations (§3.1), all potential explanatory factors were initially fitted
and removed in a stepwise fashion until the minimal adequate
model was achieved following Zuur et al. [53]. Factors included
transport status (carrying brood or not), experience (number of
trips by each individual), trail fidelity (following the trail or
not) and time (seconds since start of the relocation). In analyses
of group effects (§3.2), we used only group (large, small and
split) as a factor, with colony identity as a random factor.
These latter models were followed by Tukey post-hoc comparisons between groups using the glht function of the multcomp
package in R. Means are given as arithmetic mean + s.d. unless
otherwise stated.
rsif.royalsocietypublishing.org
S
b
Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
Table 1. Colony size, brood numbers and numbers of scouts and transporting ants in colonies and colony groups. The number of trajectories recorded in each
relocation is given for each phase and overall.
colony
size
scouts
transporters
brood
large
218
34
27
13
28
large
large
220
317
34
34
24
24
25
25
split
218b
17
14
split
split
220b
317b
17
17
small
small
209
223
small
224
discovery
trajectories
assessment
trajectories
transport
trajectories
total
trajectories
1
173
149
323
34
40
2
2
104
332
143
135
249
469
7
12
3
204
65
272
14
11
9
13
14
19
12
9
68
94
39
71
119
174
16
18
14
13
2
7
4
33
1
3
201
108
49
111
251
222
15
14
9
12
11
210
39
260
Table 2. Results of GLMM analyses with Tukey post-hoc comparison between colony groups. Significant results (at p , 0.05) are indicated in bold. Details of
statistical models used are given in the Methods section. Dataset indicates the unit of analysis, whether by colony, individual or trajectory.
small versus split
small versus large
split versus large
z
z
dataset
variable
z
p
colony
number of scouts
20.22
0.973
3.11
0.005
23.31
0.003
proportion of scouts
number of transporters
21.04
1.40
0.545
0.336
21.43
3.99
0.319
<0.001
0.55
23.45
0.845
0.002
1.13
0.480
1.41
0.320
20.62
0.802
total duration
discovery duration
21.85
20.58
0.140
0.832
21.47
22.44
0.284
0.039
21.13
2.28
0.474
0.058
assessment duration
transport duration
21.35
20.52
0.349
0.849
20.86
1.34
0.651
0.342
21.18
210.44
0.449
<0.001
total distance
21.34
0.364
0.81
0.690
23.90
<0.001
discovery distance
assess distance
0.47
21.01
0.883
0.565
20.56
0.09
0.847
0.996
2.00
21.72
0.105
0.191
transport distance
transport efficiency
20.87
0.97
0.658
0.587
3.04
20.60
0.007
0.817
23.91
2.90
<0.001
0.001
number of trajectories
quorum thresholds
21.83
1.78
0.146
0.172
2.15
26.77
0.071
<0.001
211.67
26.26
<0.001
<0.001
transports per ant
21.44
0.310
21.35
0.358
20.32
0.944
scaled effort (overall)
scaled effort (discovery)
20.54
0.43
0.851
0.905
23.07
3.84
0.006
<0.001
3.12
24.03
0.005
<0.001
scaled effort (assess)
scaled effort (transport)
0.50
20.20
0.869
0.977
21.20
22.62
0.444
0.022
2.72
4.84
0.017
<0.001
mean trajectory speed
20.46
0.879
22.67
0.017
11.05
<0.001
proportion of transporters
individual
trajectory
and there was a corresponding decrease in distance of trajectories with time (t2140 ¼ 27.57, p , 0.001) though not with
experience (t2137 ¼ 0.22, p ¼ 0.823). Trajectories associated
with the trail were on average faster (t2137 ¼ 3.36, p , 0.001)
and shorter (t2140 ¼ 9.20, p , 0.001). There was no influence
of brood transport on speed (t2135 ¼ 1.51, p ¼ 0.131) or
p
p
distance (t2138 ¼ 20.005, p ¼ 0.996) of individual trajectories.
Trail fidelity increased over time (z ¼ 19.07, p , 0.001)
though not with experience (z ¼ 21.30, p ¼ 0.192), and
while only 18% (n ¼ 1539) of trajectories followed the trail
prior to the quorum point, this was markedly higher (85%;
n ¼ 802) after the quorum point (range: 3–70% before
J. R. Soc. Interface 11: 20140641
colony
rsif.royalsocietypublishing.org
group
4
Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
(a)
duration
(1000s of seconds)
5
rsif.royalsocietypublishing.org
versus 64–97% after, by relocation event). Ants in one relocation closely followed the box edge and this lead to an
unusually high-level pre-quorum trail fidelity (70%), whereas
the next highest was 28% for the split version of the same
colony. Trajectories in which ants were carrying brood were
more likely to be associated with the trail (93% versus 82%
following the quorum point: z ¼ 4.82, p , 0.001). Of 38
trips made by naive ants (those undertaking their first trajectory) following the quorum point, all but one was aligned
with the trail.
10
5
0
100
3.1.2. Distribution of effort within colonies
distance (m)
80
223
224
209
220b
317b
317
220
218b
split
small
colony number and size group
Figure 2. Stacked bar graphs of (a) the duration of relocation phases and
(b) distance covered in all summed trajectories for each phase. White bars
indicate the discovery phase, grey bars indicate the assessment phase, and
black bars indicate the transport phase. See table 1 for colony demographic
information.
(a)
8
6
4
2
0
15
10
5
large
split
224
223
209
317b
220b
218b
317
220
0
218
The results of inter-group comparisons are presented in table 2.
The time taken to locate the new nest site was highly variable,
and large colonies in general found new nests more rapidly
than small or split colonies (figure 2). Overall, however, relocations times did not differ significantly, as large colonies,
having more brood, also spent longer transporting. Analyses
of distance covered during each phase support this, with significantly more distance (summed over all trajectories) covered
during transport in large colonies when compared with both
small and split colonies, and correspondingly higher distance
covered overall (figure 2). Larger colonies employed more
scouts and transporters, and used higher quorum thresholds
than both small and split colonies (table 1 and figure 3). The proportion of scouts, however, did not vary with colony size
(means: 84 + 10 small, 74 + 5 large and 77 + 10 split).
218
large
quorum threshold
3.2.1. Duration of relocations and task allocation
40
0
(b)
3.2. Group size effects
60
20
speed (mm s–1)
An average of 78% (range 65– 93%) of colony members
recorded at least one trajectory prior to the quorum point
(i.e. functioned as scouts), whereas 48% (range 13 –76%) of
the colony contributed to brood transport. Scaled effort for
the entire relocation was higher in scouts than non-scouts
(z ¼ 11.31, p , 0.001) and transporters than non-transporters
(z ¼ 4.50, p , 0.001). The high trail fidelity following the
quorum point suggests that a functional trail was present
when a consensus decision was reached. As it is not possible
to observe trail laying in this species, we estimated traillaying effort by quantifying the number of trajectories
aligned to the final trail prior to the quorum point. This
value varied from 6 to 75 (mean 24 + 23) and suggests that
very low numbers of passages (6, 7 and 8 passages in the
three lowest cases) are sufficient to generate a functional
trail. The high passage rate in some colonies prior to the
quorum point (43 and 75) suggests that trail establishment
is not the principal requirement of the assessment phase, in
agreement with a previous study which showed that colonies
provided with pre-formed trails retained a long assessment
phase [47]. The route followed by trails varied from being closely aligned to one side of the navigation box to a more or less
straight line through the centre (see the electronic supplementary material). As for scouting and transporting, there was no
evidence that trail laying was monopolized by one or few individuals, as pre-quorum trail oriented trajectories were shared
among individuals, with 1–11 trajectories per individual
(mean: 2.5 + 1.9) and at least four separate individuals following a trajectory aligned to the trail before the quorum point in
all relocations.
J. R. Soc. Interface 11: 20140641
(b)
small
colony number and size group
Figure 3. Box plots of (a) trajectory speed and (b) quorum thresholds of
switching ants. Boxes indicate upper and lower quartiles, the horizontal
line indicates the median while whiskers indicate boundaries of maximum/
minimum points within 1.5 the interquartile range. Outliers are indicated
by points.
Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
(a)
(a)
6
40
percentage of
individuals
4
0
30
20
10
6
(b)
4
percentage of
individuals
2
4
3
40
30
J. R. Soc. Interface 11: 20140641
20
10
2
1
(c)
218
220
317
218b
220b
317b
209
223
224
0
large
split
small
colony number and size group
Figure 4. Mean (+s.e.) scaled effort per individual for each relocation for
(a) discovery phase, (b) assessment phase and (c) transport phases.
percentage of
individuals
(c)
mean scaled effort per individual (1000s ± s.e.)
(b)
8
rsif.royalsocietypublishing.org
12
40
30
20
10
0
3.2.2. Variation in individual contributions
Individuals in large colonies invested more effort in discovery than in small or split colonies (figure 4). However, this
trend was reversed during the assessment and transport
phases, as split colonies invested significantly more effort
per individual than large colonies overall and during assessment and transport. The same was true for small colonies
except that there was no difference in effort during the assessment phase. Furthermore, ants moved faster in both small
and split colonies relative to large colonies (figure 3). The
number of transports per individual did not differ between groups. However, transport efficiency (transport per
individual per unit time) was higher in split colonies
than large colonies. Finally, the proportion of the total distance covered by each individual was considerably more
skewed in small and split colonies relative to large colonies
(figure 5), indicating that the increase in effort exhibited in
smaller colonies was focused in a few individuals rather
than being evenly distributed.
3.3. Role flexibility
Removing the most active ants from split colonies had no
observable influence on the performance of colonies during
the relocation process, and no significant differences were
found in any of the analysed metrics between small and split
colonies. This suggests that colonies are readily able to adapt
to changes in group size and that roles are highly flexible.
4. Discussion
Relative to their large counterparts, split colonies of M. nipponica
employed a lower quorum threshold, increased movement
speed, had higher transport efficiency, and invested more
effort per individual overall and during the assessment and
transport phases. In addition and probably because of having
5
10
15
percentage of total distance
20
Figure 5. Bar graph of mean percentage of the total distance covered by
percentage of ants, for (a) large colonies, (b) split colonies and (c) small
colonies.
less brood, split colonies required less time to relocate brood
and covered less distance during brood transport. These patterns were also largely consistent with unmanipulated small
colonies, suggesting they are not an artefact of the manipulation
(see also [47]), nor related to other colony-size-correlated factors
such as experience [28,54].
Colony-size related differences in the discovery and transport phases can be attributed to the likelihood of finding the
new nest with more scouts and differences in the number of
brood, respectively. However, the bulk of the relocation was
spent in information collection and collation during the
assessment phase. During the assessment phase, ants
accumulate knowledge of potential nesting sites by exploring
the accessible area. Colonies may eventually decide to nest
anywhere within the nesting box, but show a preference for
protected sites such as corners, and in this study we provided
and overwhelmingly superior choice in the form of the artificial nest. As a consequence, all colonies made ‘successful’
choices, though sub-optimal decisions can occur when the
available options are less distinct [55]. To determine the optimal site and collectively decide to move to it, both small and
large colonies had to assess an equivalent area of the experimental arena and, although smaller colonies had fewer
scouts with which to do this, assessment time and distance
did not differ between groups. If we make the assumption
that information accrued is proportional to distance covered,
it follows that small colonies are likely to have had a similar
collective awareness of their environment to large colonies,
though in small colonies this information was vested in
fewer individuals. Because individuals in small colonies
invested more effort (covered more distance) than individuals
Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
7
J. R. Soc. Interface 11: 20140641
while still retaining some individuals for brood protection
and maintenance. Indeed the high proportion of scouts
found here (78 + 9% of ants recorded at least one trajectory,
and 67 + 11% entered the new nest prior to the quorum
point) supports earlier data indicating that M. nipponica has
a more equally shared decision process than in other insects
so far studied [48] and is in this regard perhaps more akin
to some social vertebrates [33,34]. The number of scouts
also appears to scale with colony size in Temnothorax ants
[28,38] and honeybees [57], but is a considerably lower proportion of the colony: Dornhaus et al. [28] report that only
5–7% of ants in T. albipennis reach the new nest before commencement of transports and a similarly a low percentage of
scouts (5%) is found during nest relocation in honeybees [29].
Having additional resources may allow smaller colonies of
these and other species to allocate proportionately more individuals to information collection without (or in conjunction
with) increasing individual effort.
Split colonies were able to perform relocations successfully and on par with naturally small colonies despite the
most active ants having been removed. Because the impact
of individual contributions is expected to be larger in smaller
colonies [58], we might expect that removing the most active
half of the colony could have a particularly detrimental effect
in M. nipponica. However, this manipulation did not appear
to interfere with the consensus decision-making or relocation
process, suggesting both a lack of specialists and highly flexible roles in this species. This is in line with predictions that
larger societies of social insects are characterized by greater
individual specialization and a loss of individual totipotency
[59 –61]. Myrmecina nipponica is a small-colony ant and thus
may exhibit greater flexibility in individual roles than large
colony species. By contrast, Möglich & Hölldobler [24]
showed that in two species of large-colony formicine ants, a
small subset of the colony acted as relocation specialists,
and relocation behaviour suffered if these ants were removed.
However, Sumana & Sona [42] demonstrated the presence
of a single specialist during colony relocations in the smallcolony ant Diacamma indicum, indicating that colony size is
not a reliable predictor of specialist roles.
We demonstrate that individuals in small groups of
M. nipponica work harder than those in large groups during
the information gathering and voting stage of consensus
decisions. This may act to ameliorate the limitation of fewer
individuals with respect to the ‘wisdom of the crowds’ effect
in decision-making because the individuals involved are
better informed. To achieve an equivalent task at a comparable
rate and standard, individuals in smaller groups should work
harder than those in larger groups. For example, to attain the
same level of predator awareness, individuals in smaller
groups need to invest more effort in collective vigilance [62].
However, at present we have a limited understanding of how
the behaviour of individuals is modified and compiled in
groups of different size. Our study shows that individuals in
smaller groups may employ different strategies during collective actions than those in larger groups. Our findings also
raise some additional questions: firstly, it remains unclear to
what degree increasing effort can improve decision-making
in small groups, and if the scope for increased effort to improve
decisions is realized. Secondly, if individuals in small groups
are already working harder, this may limit the flexibility they
have in responding to additional environmental stresses
during decision-making, and as a result we might expect
rsif.royalsocietypublishing.org
in large colonies, these individuals should have accrued more
information and thus could be considered better informed.
Furthermore, the same individuals that perform the greatest
scouting effort are likely to be the ones with most opportunity to contribute to trail formation (i.e. generate social
information), and thus voting weight will also scale with
effort. In this way, small colonies of M. nipponica may be able
to ameliorate to some degree the disadvantage of small
colony size with respect to the ‘wisdom of the crowds’ effect:
small colonies have fewer individuals available to contribute
to the information pool but, by virtue of their greater effort, it
is likely that these individuals are better informed and have
higher voting weight. The degree to which this may enhance
decision accuracy in small colonies remains unclear but any
compensation for the lack of numbers is likely to be partial,
as currently available data suggest that smaller colonies of
M. nipponica suffer from reduced decision accuracy relative to
larger colonies (AL Cronin 2014, unpublished data).
Evidence that workers vary effort with colony size in
other species is contradictory: Dornhaus et al. [49] show
that the most active individual performed proportionately
more work in smaller colonies of Temnothorax albipennis
over a range of tasks, whereas honeybee foragers in large
colonies work harder than those in small colonies [56].
However, in many cases, workload is likely to scale with
colony size; foraging effort, for example will depend on the
number of brood, which in turn is correlated with colony
size. In this study, all groups had to perform a similar task
(finding a new home) and, while work during the transport
phase varied as brood numbers differed, an equivalent
amount of work was required to assess possible nesting
sites and decide on one of these. If effort is evenly distributed
among group members, we would expect a higher proportion of effort per individual in smaller colonies.
However, we found that not only proportionate effort, but
also absolute effort, was higher in smaller colonies. This
suggests that individuals in large colonies may only work
‘hard enough’ to maintain an acceptable level of decision
speed and accuracy and may retain a capacity for additional
effort if required. Individuals in small colonies, on the other
hand, may be at or close to their maximum work limit.
This difference could limit the ability of small colonies to
cope with additional stresses to the decision and relocation
process as might occur under harsh conditions or in complex
environments, and small colonies may suffer a higher cost to
decision accuracy as a result.
While individual effort was higher in smaller colonies of
M. nipponica, it appears that other characteristics scale with
colony size. As in previous studies of this species, quorum
thresholds and the number of scouts and transporters were
higher in larger colonies [47,48]. This indicates that while
ants increase individual effort in small groups, there is no
increase in the proportion of individuals involved in the
decision-making process. The number of scouts is likely to
be linked to the quorum threshold because colonies cannot
increase quorums without increasing the scouting force or
they risk achieving no decision at all [3]. Quorum thresholds
are thus linked to the number of individuals involved in
information gathering, and larger quorums should be associated with better decisions [25]. Why smaller colonies do not
allocate more individuals to the decision process is unclear,
though the high proportion of scouts in many relocations
suggests that these colonies may be at a functional limit
Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
such stress to have a more marked effect on smaller colonies.
These questions are currently under investigation.
Data accessibility. The datasets supporting this article have been
Acknowledgements. We are grateful to Thibaud Monnin, Andrew King
Funding statement. This work was supported by funds from the
United Graduate School of Agriculture Iwate University, and
Grant-in-Aid for Scientific Research JSPS KAKENHI #25440187
to A.L.C.
References
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
Conradt L, List C. 2009 Group decisions in humans
and animals: a survey. Phil. Trans. R. Soc. B 364,
719–742. (doi:10.1098/rstb.2008.0276)
Camazine S, Deneubourg JL, Franks NR, Sneyd J,
Theraulaz G, Bonabeau E. 2001 Self-organization in
biological systems, p. 560. Princeton, NJ: Princeton
University Press.
Conradt L, Roper JT. 2005 Consensus decision
making in animals. Trends Ecol. Evol. 20, 449–456.
(doi:10.1016/j.tree.2005.05.008)
Sumpter DJ. 2010 Collective animal behavior, p. 302.
Princeton, NJ: Princeton University Press.
King AJ, Sueur C. 2011 Where next? Group
coordination and collective decision making by
primates. Int. J. Primatol. 32, 1245 –1267. (doi:10.
1007/s10764-011-9526-7)
Sumpter DJ, Pratt SC. 2009 Quorum responses
and consensus decision making. Phil. Trans. R.
Soc. B 364, 743 –753. (doi:10.1098/rstb.
2008.0204)
Pyritz LW, King AJ, Sueur C, Fichtel C. 2011
Reaching a concensus: terminology and concepts
used in coordination and decision-making research.
Int. J. Primatol. 32, 1268 –1278. (doi:10.1007/
s10764-011-9524-9)
Ward AJ, Herbert-Read JE, Jordan LA, James R,
Krause J, Ma Q, Rubenstein DI, Sumpter DJ, Morrell
LJ. 2013 Leaders, and recruitment mechanisms in
the collective movements of damselfish. Am. Nat.
181, 748–760. (doi:10.1086/670242)
Pratt SC. 2010 Nest site choice in social insects. In
Encyclopedia of animal behavior (eds MD Breed,
J Moore), pp. 534–540. Oxford, UK: Academic
Press.
Visscher PK. 2007 Group decision making in nestsite selection among social insects. Annu. Rev.
Entomol. 52, 255 –275. (doi:10.1146/annurev.ento.
51.110104.151025)
Conradt L, Roper JT. 2003 Group decision-making in
animals. Nature 421, 155– 158. (doi:10.1038/
nature01294)
Wolf M, Kurvers RH, Ward AJ, Krause S, Krause J.
2013 Accurate decisions in an uncertain world:
collective cognition increases true positives while
decreasing false positives. Proc. R. Soc. B 280,
20122777. (doi:10.1098/rspb.2012.2777)
Simons AM. 2004 Many wrongs: the advantage of
group navigation. Trends Ecol. Evol. 19, 453– 455.
(doi:10.1016/j.tree.2004.07.001)
Condorcet JANC. 1785 Essai sur l’application de
l’analyse à la probabilité des décisions rendues à la
pluralité des voix. Paris: Imprimerie Royale.
15. Galton F. 1907 Vox populi. Nature 75, 450 –451.
(doi:10.1038/075450a0)
16. Clément RJG, Krause S, von Engelhardt N, Faria JJ,
Krause J, Kurvers RHJM. 2013 Collective cognition in
humans: groups outperform their best members in
a sentence reconstruction task. PLoS ONE 8, e77943.
(doi:10.1371/journal.pone.0077943)
17. Herbert-Read JE, Perna A, Mann RP, Schaerf TM,
Sumpter DJ, Ward AJ. 2011 Inferring the rules of
interaction of shoaling fish. Proc. Natl Acad. Sci. USA
108, 18 726–18 731. (doi:10.1073/pnas.
1109355108)
18. Ward AJ, Herbert-Read JE, Sumpter DJ, Krause J.
2011 Fast and accurate decisions through collective
vigiance in fish shoals. Proc. Natl Acad. Sci. USA
108, 2312–2315. (doi:10.1073/pnas.1007102108)
19. Conradt L. 2011 When it pays to share decisions.
Nature 471, 40 –41. (doi:10.1038/471040a)
20. Detrain C, Deneubourg JL. 2009 Social cues and
adaptive foraging strategies in ants. In Food
exploitation by social insects: ecological, behavioral
and theoretical approaches (eds S Jarau, M Hrncir),
pp. 29– 51. Boca Raton, FL: CRC Press.
21. Bousquet CAH, Sumpter DJT, Manser MB. 2011
Moving calls: a vocal mechanism underlying
quorum decisions in cohesive groups. Proc. R. Soc. B
278, 1482–1488. (doi:10.1098/rspb.2010.1739)
22. Pratt SC. 2005 Behavioral mechanisms of collective
nest-site choice by the ant Temnothorax
curvispinosus. Insectes Soc. 52, 383–392. (doi:10.
1007/s00040-005-0823-z)
23. Black JM. 1988 Preflight signalling in swans: a
mechanism for group cohesion and flock formation.
Ethology 79, 143–157. (doi:10.1111/j.1439-0310.
1988.tb00707.x)
24. Möglich M, Hölldobler B. 1974 Social carrying
behavior and division of labor during nest moving
in ants. Psyche 81, 219–236. (doi:10.1155/
1974/25763)
25. Franks NR, Dechaume-Moncharmont FX, Hanmore E,
Reynolds JK. 2009 Speed versus accuracy in
decision-making ants: expediting politics and policy
implementation. Phil. Trans. R. Soc. B 364,
845 –852. (doi:10.1098/rstb.2008.0224)
26. Sasaki T, Pratt SC. 2012 Groups have a larger
cognitive capacity than individuals. Curr. Biol. 22,
R827 –R829. (doi:10.1016/j.cub.2012.07.058)
27. 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. (doi:10.1098/rstb.2002.1066)
28. Dornhaus A, Holley J-A, Pook VG, Worswick G,
Franks NR. 2008 Why do not all workers work?
Colony size and workload during emigrations in the
ant Temnothorax albipennis. Behav. Ecol. Sociobiol.
63, 43 – 51. (doi:10.1007/s00265-008-0634-0)
29. Seeley TD, Morse RA, Visscher KP. 1979 The natural
history of the flight of honey bee swarms. Psyche
86, 103 –113. (doi:10.1155/1979/80869)
30. Franks NR, Richardson TO, Stroeymeyt N, Kirby RW,
Amos WMD, Hogan PM, Marshall JAR, Schlegel T.
2013 Speed – cohesion trade-offs in collective
decision making in ants and the concept of
precision in animal behaviour. Anim. Behav. 85,
1233– 1244. (doi:10.1016/j.anbehav.2013.03.010)
31. King AJ, Cowlishaw G. 2007 When to use social
information: the advantage of large group size in
individual decision making. Biol. Lett. 3, 137 –139.
(doi:10.1098/rsbl.2007.0017)
32. Katsikopoulos KV, King AJ. 2010 Swarm intelligence
in animal groups: when can a collective outperform an expert? PLoS ONE 5, e15505. (doi:10.
1371/journal.pone.0015505)
33. Leca JB, Gunst N, Thierry B, Petit O. 2003
Distributed leadership in semifree-ranging white
capuchin monkeys. Anim. Behav. 66, 1043–1052.
(doi:10.1006/anbe.2003.2276)
34. Meunier H, Leca JB, Deneubourg JL, Petit O. 2006
Group movement decisions in capuchin monkeys:
the utility of an experimental study and a
mathematical model to explore the relationship
between individual and collective behaviours.
Behaviour 143, 1511– 1527. (doi:10.1163/
156853906779366982)
35. Sueur C, Deneubourg JL. 2011 Self-organization in
primates: understanding the rules underlying
collective movements. Int. J. Primatol. 32,
1413– 1432. (doi:10.1007/s10764-011-9520-0)
36. Burns ALJ, Herbert-Read JE, Morrell LJ, Ward AJ.
2012 Consistency of leadership in shoals of
mosquitofish (Gambusia holbrooki) in novel and
familiar envionrments. PLoS ONE 7, e36567. (doi:10.
1371/journal.pone.0036567)
37. Herbert-Read JE, Krause S, Morrell LJ, Schaerf TM,
Krause J, Ward AJW. 2013 The role of individuality
in group movement. Proc. R. Soc. B 280, 20122564.
(doi:10.1098/rspb.2012.2564)
38. Franks NR, Dornhaus A, Best CS, Jones EL. 2006
Decision making in small and large house-hunting
ant colonies: one size fits all. Anim. Behav. 72,
611–616. (doi:10.1016/j.anbehav.2005.11.019)
39. Cronin AL, Molet M, Doums C, Monnin T, Peeters C.
2013 Recurrent evolution of dependent colony
J. R. Soc. Interface 11: 20140641
1.
rsif.royalsocietypublishing.org
and one anonymous referee for comments on the manuscript
and Ayuka Nishimura and Honami Oohashi for help with scoring
behaviour from videos.
8
uploaded as part of the Supplementary Material.
Downloaded from http://rsif.royalsocietypublishing.org/ on July 28, 2017
41.
42.
44.
45.
46.
47.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
PLoS ONE 8, e64668. (doi:10.1371/journal.pone.
0064668)
Eckert CD, Winston ML, Ydenberg RC. 1994 The
relationship between population size, amount of
brood, and individual foraging behaviour in the
honey bee, Apis mellifera L. Oecologia 97,
248–255. (doi:10.1007/BF00323157)
Schaerf TM, Makinson JC, Myerscough MR, Beekman
M. 2013 Do small swarms have an advantage when
house hunting? The effect of swarm size on nestsite selection by Apis mellifera. J. R. Soc. Interface
10, 20130533. (doi:10.1098/rsif.2013.0533)
Dornhaus A, Powell S, Bengston S. 2012 Group size
and its effects on collective organization. Annu. Rev.
Entomol. 57, 123–141. (doi:10.1146/annurev-ento120710-100604)
Anderson C, McShea DW. 2001 Individual versus
social complexity, with particular reference to ant
colonies. Biol. Rev. 76, 211–237. (doi:10.1017/
S1464793101005656)
Crespi BJ, Yanega D. 1995 The definition of
eusociality. Behav. Ecol. 6, 109–115. (doi:10.1093/
beheco/6.1.109)
Gautrais J, Theraulaz G, Deneubourg J-L, Anderson C.
2002 Emergent polyethism as a consequence of
increased colony size in insect societies. J. Theor. Biol.
215, 363–373. (doi:10.1006/jtbi.2001.2506)
Michelena P, Deneubourg JL. 2011 How group size
affects vigilance dynamics and time allocation
patterns: the role of imitation and tempo. PLoS ONE
6, e18631. (doi:10.1371/journal.pone.0018631)
9
J. R. Soc. Interface 11: 20140641
43.
48.
Ecol. Sociobiol. 67, 1643–1651. (doi:10.1007/
s00265-013-1575-9)
Cronin AL. 2012 Consensus decision making in the
ant Myrmecina nipponica: house-hunters combine
pheromone trails with quorum responses. Anim.
Behav. 84, 1243–1251. (doi:10.1016/j.anbehav.
2012.08.036)
Dornhaus A, Holley JA, Franks NR. 2009 Larger
colonies do not have more specialized workers in
the ant Temnothorax albipennis. Behav. Ecol. 20,
922 –929. (doi:10.1093/beheco/arp070)
Bivand RS, Pebesma EJ, Gómez-Rubio V. 2008
Applied spatial data analysis with R, p. 374. Berlin,
Germany: Springer.
R Core Team. 2013 R: a language and environment
for statistical computing. Vienna, Austria: R
Foundation for Statistical Computing.
Zhang Y. 2013 Likelihood-based and Bayesian
methods for Tweedie compound Poisson linear
mixed models. Stat. Comput. 66, 743 –757. (doi:10.
1007/s11222-012-9343-7)
Zuur AF, Ieno EN, Walker NJ, Saveliev AA,
Smith GM. 2009 Mixed effects models and
extensions in ecology with R, p. 574. New York, NY:
Springer.
Dornhaus A, Franks NR. 2006 Colony size affects
collective decision-making in the ant Temnothorax
albipennis. Insectes Soc. 53, 420–427. (doi:10.
1007/s00040-006-0887-4)
Cronin AL. 2013 Conditional use of social and
private information guides house-hunting ants.
rsif.royalsocietypublishing.org
40.
foundation across eusocial insects. Annu. Rev.
Entomol. 58, 37 –55. (doi:10.1146/annrev-ento120811-153643)
Oster GF, Wilson EO. 1978 Caste and ecology in the social
insects. Princeton, NJ: Princeton University Press.
Robson SK, Taniello JFA. 1999 Key individuals and
the organisation of labour in ants. In Information
processing in social insects (eds CC Detrain,
JM Pasteels, JL Deneubourg), pp. 239–259. Basel,
Switzerland: Birkhauser.
Sumana A, Sona C. 2013 Key relocation leaders in
an Indian queenless ant. Behav. Process. 97,
84 –89. (doi:10.1016/j.beproc.2013.03.006)
King AJ, Cheng L, Starke SD, Myatt JP. 2012 Is the
true ‘wisdom of the crowd’ to copy successful
individuals? Biol. Lett. 8, 197– 200. (doi:10.1098/
rsbl.2011.0795)
Lorenz J, Rauhut H, Schweitzer F, Helbing D. 2011
How social influence can undermine the wisdom of
crowd effect. Proc. Natl Acad. Sci. USA 108,
9020–9025. (doi:10.1073/pnas.1008636108)
Anderson C, Ratnieks FLW. 1999 Task partitioning in
insect societies. I. Effect of colony size on queueing
delay and colony ergonomic efficiency. Am. Nat.
154, 521–535. (doi:10.1086/303255)
Dornhaus A. 2008 Specialization does not predict
individual efficiency in an ant. PLoS Biol. 6,
2368–2375. (doi:10.1371/journal.pbio.0060285)
Cronin AL. 2013 Synergy between pheromone trails
and quorum thresholds underlies consensus
decisions in the ant Myrmecina nipponica. Behav.