A comparative analysis of precision rescue behaviour in sand-dwelling

ants
Karen L. Hollis
a,
*
, Elise Nowbahari
b,1
a
Interdisciplinary Program in Neuroscience & Behavior, Mount Holyoke College, MA, U.S.A.
b
Laboratoire d'thologie Exprimentale et Compare, Universit Paris 13, Sorbonne Paris Cit, France
a r t i c l e i n f o
Article history:
Received 4 June 2012
Initial acceptance 23 July 2012
Final acceptance 28 November 2012
Available online 12 January 2013
MS. number: A12-00426R
Keywords:
altruistic behaviour
ant
Aphaenogaster senilis
Cataglyphis oricola
empathy
eusocial insect
Lasius grandis
Messor barbarus
Messor marocanus
rescue behaviour
A sand-dwelling Mediterranean ant, Cataglyphis cursor, recently was discovered to engage in two new
forms of rescue behaviour, behavioural patterns that require would-be rescuers to recognize what,
exactly, holds nestmates in place. That is, when sand digging and limb pulling, two well-known forms of
rescue in ants, did not result in release of victims ensnared with nylon thread and partially buried
beneath the sand, rescuers next began to transport sand away from the snare and to direct their
behaviour to the snare in particular, biting and tugging at the snare itself. To determine whether these
new forms of precisely directed rescue behaviour, as well as their exclusive delivery to nestmates, as in
C. cursor, were characteristic of other ants occupying similar ecological niches, we conducted experi-
ments with ve sand-dwelling Mediterranean ant species: Cataglyphis oricola, Lasius grandis, Aphae-
nogaster senilis, Messor barbarus and Messor marocanus. Our experiments revealed the full range of rescue
behaviour, including snare biting and sand transport, in two species, C. oricola and L. grandis. Both
species directed rescue exclusively towards nestmates, treating other individuals, even conspecics, with
aggression, thus highlighting the ants discriminative capacities. Differences in the performance of rescue
behaviour between these and the remaining species mirror differences in their ecology, including the
threat of predatory antlions. Finally, because this precisely directed and exclusively delivered rescue
behaviour in ants resembles behaviour that has been labelled empathy in rats, our results demonstrate
that what can appear to be complex, cognitively motivated behaviour might come about through much
simpler mechanisms.
2012 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Recently, Nowbahari et al. (2009) demonstrated that Cataglyphis
cursor, a common sand-dwelling Mediterranean ant, is capable of
performing forms of rescue behaviour that had not been previously
reported, despite a literature that dates as far back as Belts (1874)
monograph on ant behaviour. That is, when C. cursor ants were
presented with experimentally ensnared nestmates held in place
with nylon thread, ants not only engaged in sand digging and limb
pulling, both familiar forms of rescue behaviour in ants (e.g. Laeur
1940; Wilson 1958; Markl 1965; Blum & Warter 1966; Spangler
1968; Hangartner 1969), but also somehow were able to detect
what, exactly, held victims in place: when digging and pulling did
not result in the victims release, ant rescuers next transported
excavated sand away from the victims body, exposed the nylon
thread, and then immediately bit and tugged at the snare directly.
Carefully aimed, snare biting never was misplaced, even though the
snare was in direct contact with the victims body.
Nowbahari et al. (2009) argued that sand transporting and,
especially, snare biting might require slightly more complicated
mechanisms than simple sand digging and limb pulling. That is,
digging and pulling require only that ants are alerted by a nest-
mates alarm signal, follow the sensory gradient to the source, and
then commence digging and pulling once they make contact with
the victim, a series of behaviour patterns easily explained by simple
releasing mechanisms. However, it is difcult to see how these
same simple mechanisms could guide rescuers to the precise
location of whatever object is holding the victim in place while
ignoring other non-ant objects nearby, enable them to expose it
further by transporting sand away fromthis object, and then nally
target their bites to the object itself, transporting additional sand as
necessary. In addition, because C. cursor directed this rescue
behaviour only towards nestmates, never even approaching ants
belonging to nearby colonies of the same species, the call-for-help
signal, in this species at least, necessarily contains a component
unique to each colony. Such precisely directed and exclusively
* Correspondence: K. L. Hollis, Interdisciplinary Program in Neuroscience &
Behavior, Mount Holyoke College, 50 College Street, South Hadley, MA 01075-1462,
U.S.A.
E-mail address: khollis@mtholyoke.edu (K. L. Hollis).
1
E-mail address: Elise.Nowbahari@leec.univ-paris13.fr (E. Nowbahari).
Contents lists available at SciVerse ScienceDirect
Animal Behaviour
j ournal homepage: www. el sevi er. com/ l ocat e/ anbehav
0003-3472/$38.00 2012 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.anbehav.2012.12.005
Animal Behaviour 85 (2013) 537e544
targeted rescue behaviour would be expected to evolve only when
members of a particular species faced risks of possible entrapment,
and the benets of releasing conspecics exceeded what appear to
be particularly high costs of engaging in rescue behaviour.
Although the risks of entrapment would not be limited to sand-
dwelling ants like C. cursor, sand-dwelling species are likely
candidates for several reasons. The experiments with C. cursor were
prompted by eld observations in which forager ants often
attempted to rescue nestmates that had been partially buried under
collapsing sand and debris (Nowbahari et al. 2009). Yet another
potential opportunity for rescue occurs when ants fall into the traps
of pit-digging larval antlions (Neuroptera: Myrmeleontidae).
Antlions pit traps are found in some of the same habitats as sand-
dwelling ants and, although antlions prey upon many different
arthropods, ants are one of the most common prey items (Grifths
1980; Lucas & Brockmann 1981; Botz et al. 2003; Mencinger-Vracko
& Devetak 2008). Antlions pit traps, dug in loose sandy soils
(Devetak et al. 2005), are constructed in such a way to funnel prey
to the centre of the pit, where the predator is waiting; however,
prey sometimes are able to cling to the sides of pits for short
periods of time, despite attempts by antlions to dislodge their prey
by sand tossing (Hollis & Guillette 2011; Hollis et al. 2011). Thus,
rescue behaviour also may be an important antipredator defence.
Indeed, in several eld experiments with Formica workers reared as
slaves, Czechowski et al. (2002) demonstrated that enslaved ants
immediately rushed to the aid of heterospecic nestmates captured
by antlions and began to engage in sand-digging and limb-pulling
behaviour.
Despite its obvious benets, rescue behaviour entails costs that,
for some ant species, may be too high to pay. Rescue behaviour not
only is an energetically costly behaviour that prevents individuals
from devoting time to other important tasks, but also poses large
risks to rescuers (Nowbahari & Hollis 2010). For example, rescuers
themselves may become buried under sand as they attempt to aid
an entrapped nestmate. In addition, because some predators,
including antlions, are capable of seizing multiple prey items in
rapid succession, rescuers risk capture. Because of its costs, then,
we would predict that not all ant species are able to engage in
rescue behaviour. Thus, one goal of this study was to determine
how common the phenomenon of precisely directed rescue
behaviour is in ants, as well as to establish whether, like C. cursor,
foragers rescue only nestmates.
A second, related goal was to determine what might be some of
the ecological variables that promote rescue in ants. To address
both objectives, we conducted a comparative analysis of rescue
behaviour in ve species of sand-dwelling Mediterranean ants
representing two subfamilies: Lasius grandis and Cataglyphis o-
ricola (bicolour morph) from the subfamily Formicinae; and,
Aphaenogaster senilis, Messor barbarus and Messor marocanus from
the subfamily Myrmicinae. Because C. oricola not only inhabits the
same loose sandy soils as C. cursor, but also lives in close proximity
to pit-digging antlions, we expected that C. oricola would be likely
to show the same kinds of rescue behaviour as its close relative,
C. cursor. At the other end of the spectrum, Messor species inhabit
more compact soils than do Cataglyphis, and thus are probably less
likely to become trapped by collapsing sand. In addition, Messor
foragers form ant trails, long lines of individuals following marked
trails leading to food (Lpez et al. 1993; Hlldobler & Wilson 2009),
making accidental encounters with antlion pits, or any other
dangerous situations, far less likely.
Yet another goal of this study was to shed light on the
phenomenon of rescue behaviour more generally. Although anec-
dotes of rescue in nonhuman animals abound in the popular media,
the scientic literature contains only two experimental analyses of
rescue behaviour, one in ants (Nowbahari et al. 2009) and one in
rats (Bartal et al. 2011). Interestingly, the report of rescue behaviour
in rats describes behaviour analogous to what Nowbahari et al.
(2009) observed in ants. Although the authors themselves do not
use the word rescue to describe rats behaviour, preferring the
term empathy instead, our research with ants raises questions
about the necessary and sufcient conditions for interpreting
behaviour as empathic. That is, if ants respond to conspecics in
ways that appear nearly identical to those of a mammal, then future
experiments exploring empathy in animals will need to include
control groups that help scientists to distinguish between analo-
gous behaviour involving different mechanisms (Vasconcelos et al.
2012).
METHODS
Subjects and Materials
We observed the behaviour of ve sand-dwelling Mediterra-
nean ant species in the eld between April 2010 and June 2011 in
response to four kinds of test stimuli, namely a homocolonial
nestmate, a heterocolonial ant, or a heterospecic ant held in place
using a lter paper and thread snare, or an empty paper-and-thread
snare in a control test. Only colonies that showed high activity
levels and large numbers of foragers present on the surface near the
nest entrance were chosen for study. All data were collected during
ants active period, either in the morning, at least 1 h after foragers
had emerged from the nest but before ants retreated to the nest at
midday, or later in the afternoon, at least 1 h after foragers had re-
emerged but before ants retreated to the nest near the end of the
day. We observed one of the ant species, M. barbarus, in Argels-
sur-mer (Languedoc-Roussillon) with permission from Institut
National des Sciences de lUnivers; we observed all four of the
remaining ant species, L. grandis, C. oricola, A. senilis, and
M. marocanus, in the Doana Biological Reserve (Reserva Biolgica
de Doana, RBD) in Southwestern Spain. All eldwork in Doana
was formally approved by the ICTS-RBD program. Specimens of
each species were returned to the laboratory to ensure correct
identication (Bernard 1968; Tinaut 1991).
Filter paper snares were prepared in advance of all trials: using
a sewing needle, a small loop (0.5e0.75 cm, depending on the test
species) of two-ply cotton thread was inserted near one end of
a strip (2 cmlong 1 cmwide) of clean lter paper; the ends of the
thread were tied loosely underneath the paper and left hanging
until needed in the eld. Prepared snares were kept in a clean
plastic box to avoid contamination and a clean snare was used for
each test.
Procedure
The testing procedures were similar to Nowbahari et al. (2009)
but adapted to eld conditions where noted below. For homo-
colonial (nestmate) tests, the ant victim was chosen from amongst
the foragers near the nest entrance, grasped with insect handling
tweezers, gently inserted under the thread loop on the lter paper
such that the thread encircled the victims pedicel (waist), and then
secured to the lter paper by pulling on the ends of the thread
underneath the paper. A single knot held the victim in place. Next,
the lter paper was trimmed to a length of approximately 1.0e
1.5 cm (depending on species size) and the ends of the thread
were clipped close to the knot such that no thread extended beyond
the edges of the paper. For heterocolonial tests, the victim was
chosen froma nearby colony of the same species, at least 30 mfrom
the test colony, and tied to the lter paper as described above.
Heterospecic tests were conducted in the same manner as het-
erocolonial tests, except that ant victims were chosen fromcolonies
K. L. Hollis, E. Nowbahari / Animal Behaviour 85 (2013) 537e544 538
representing different species whose nests were typically between
30 and 50 m from the test colony. In this way, the ant victimwould
represent a species with which the test species could have had
contact, at least potentially. For A. senilis, heterospecic tests were
conducted with both M. marocanus and L. grandis (a larger and
a smaller ant species, respectively). For C. oricola, heterospecic
tests were conducted with L. grandis, A. senilis and M. marocanus
(one smaller and two larger ant species, respectively). For L. grandis,
heterospecic tests were conducted with A. senilis (a larger
species). And, nally, both Messor species were tested with a single
species during heterospecic trials; M. marocanus was tested with
snared A. senilis ants, and M. barbarus was tested with snared
C. cursor ants. All control tests were conducted with a clean lter
paper containing a thread snare as described above; just prior to
the test, the paper and thread were trimmed and clipped exactly as
was done for trials with ant victims.
To conduct a test, the lter paper, either containing a victim or
left empty, was placed within 7e10 cm of the nest entrance and
covered with a thin layer of sand, such that the head and thorax of
each victim, or the empty loop, but not the lter paper, was visible.
Unlike the procedure described in Nowbahari et al. (2009), no arena
was used to conne the victim and potential rescuers; instead, the
5 min test began when at least one ant from the test colony either
contacted an ant victim, or walked within one body length of the
empty snare. Following the 5 min test, nestmate victims were
sequestered in a plastic box until any further trials on that day had
been conducted. Heterocolonial and heterospecic victims that did
not receive aggressive treatment, and thus were mostly uninjured,
were returned to their respective colonies following the test;
otherwise, they were destroyed. Used lter paper snares were
placed in a separate plastic box for experimentally contaminated
materials and disposed of at the end of each day.
Astopwatch was used to time each 5 mintrial andall behavioural
recording was done using an interval recording procedure, which is
especially useful when behaviour may be performed at low and
mediumrates (Martin & Bateson 1993). Each 5 min trial was divided
into 10 s intervals, and target behaviours were recorded for each
interval. For each target behaviour (see Table 1), the data were the
proportion of intervals in which that behaviour pattern was
observed. To ensure reliability, we collectedmultiple samples of each
test condition from at least three different colonies of each species.
Statistical Analysis
Each of the behaviour patterns was subjected to a univariate
ANOVAwith two between-subjects factors, namely species and test
type. In addition, because individual target behaviours cannot
capture overall levels of rescue or aggression, we computed
a combined rescue score and a combined aggression score. For
the combined rescue score, we computed the proportion of inter-
vals in which at least one of the four rescue behaviours occurred
(digging, pulling, snare biting or sand transport). For the combined
aggression score, we computed the proportion of intervals in which
at least one of the three aggressive behaviours occurred (aggressive
biting, dismembering or aggressive projecting). These combined
scores also were subject to ANOVAs. All ANOVAs were computed
using SPSS v.19 (SPSS Inc., Chicago, IL, U.S.A.). Signicant main
effects and interactions were explored with post hoc tests, namely
paired comparisons using Bonferroni adjustments for multiple
comparisons; all reported P values represent two-tailed tests.
Finally, we repeated our data analyses using nonparametric
statistics, that is, KruskaleWallis tests to evaluate main effects,
followed by ManneWhitney U tests to explore pairwise compari-
sons, and in all cases obtained exactly the same patterns of results
as are reported below with parametric analyses.
RESULTS
Rescue Behaviour
Combined rescue score
The ve ant species differed signicantly in their rescue
response to the test stimuli (Fig. 1). An ANOVA revealed signicant
main effects of species (F
4,285
44.12, P < 0.0001) and test type
(F
3,285
188.07, P < 0.0001), as well as a signicant interaction
between species and test type (F
12,285
43.33, P < 0.0001). Explo-
ration of that interaction, using Bonferroni post hoc tests adjusted
for multiple comparisons, revealed that in the presence of an
ensnared nestmate, C. oricola and L. grandis engaged in very high
levels of rescue behaviour that were nearly identical to one another
(P 1.000), and which differed signicantly from the three
remaining species (P < 0.0001), all of which engaged in extremely
low levels of rescue (between 0.01 and 0.16 of the 5 min test).
Table 1
Operational denitions of target behaviour patterns in sand-dwelling ants
Behaviour Operational denition
Sand digging Ant positions itself within 2 cm of, and facing, the test stimulus, either the ensnared ant victim or the paper-and-snare control stimulus, and icks
sand backward, away from the test stimulus, using its anterior legs; the gaster (abdomen) is not exed, and the antennae are facing forward, in the
characteristic nonaggressive posture
Limb pulling Ant grabs limb of the ensnared ant victim with its mandibles and drags the limb backwards with frequent antennation; the gaster (abdomen) is not
exed, and the antennae are facing forward, in the characteristic nonaggressive posture
Snare biting Ant bites and tugs at the thread snare using its mandibles; the gaster (abdomen) is not exed, and the antennae are facing forward, in the
characteristic nonaggressive posture
Sand transport Ant picks up one or more sand particles, covering the lter paper or ensnared ant victim, with its mandibles and moves the sand at least one
body length from its original position; the gaster (abdomen) is not exed, and the antennae are facing forward, in the characteristic nonaggressive
posture
Paper tugging Ant bites and tugs at the lter paper using its mandibles, moving the paper away from the nest; the gaster (abdomen) is not exed, and the antennae
are facing forward, in the characteristic nonaggressive posture
Biting Ant bites the limbs, antennae or other body parts of the ensnared ant victim with its mandibles; the gaster (abdomen) is exed, curved underneath
the body, and the antennae are facing backward, in the characteristic aggressive posture
Dismembering Ant grabs limb or antenna of ensnared ant victim with its mandibles and drags the body part backwards; the gaster (abdomen) is exed, curved
underneath the body, and antennae are exed backward, in the characteristic aggressive posture.
Projecting Ant exes its gaster (abdomen), curving it underneath its body, and projects a liquid substance in the direction of the test stimulus; used in aggressive
encounters, this substance is formic acid in ants belonging to the subfamily Formicinae (C. oricola and L. grandis); in the remaining three species
(Aphaenogaster senilis, Messor barbarus and Messor marocanus), the substances are alkaloids from the poison gland and/or hydrocarbons from the
Dufour gland; the antennae are exed backward, in the characteristic aggressive posture
Retreating Ant orients towards test stimulus and immediately darts to a position inside the nest; subsequent retreats were scored whenever an ant appeared
at the nest entrance, or partially emerged, but immediately jerked back inside
K. L. Hollis, E. Nowbahari / Animal Behaviour 85 (2013) 537e544 539
Individual rescue behaviours
The same pattern of ndings was obtained when each of the
four individual rescue behaviours was subject to analysis, namely
signicant main effects of species and test type, as well as signi-
cant interactions between species and test type (see Table 2 for
results of these individual ANOVAs). Exploration of each signicant
interaction between species and test type, using Bonferroni post
hoc tests adjusted for multiple comparisons, revealed that in the
presence of an ensnared nestmate, both C. oricola and L. grandis
engaged in signicantly higher levels of digging and pulling (all
Ps 0.001), as well as the two newly discovered rescue behaviours,
namely snare biting (all Ps 0.001) and sand transport (P 0.044),
compared to extremely lowlevels of these same four behaviours in
the remaining three species (Fig. 2). However, as Fig. 2 also illus-
trates, C. oricola and L. grandis differed in the levels of these
individual behaviours, with C. oricola engaging in signicantly
more pulling (P 0.001) and snare biting (P 0.015) than
L. grandis, but L. grandis engaging in more digging (P 0.002) and
sand transport (P 0.001) than C. oricola.
Messor marocanus, which appears to have the highest levels of
rescue behaviour amongst the three essentially nonrescuer
species, did indeed differ signicantly from M. barbarus (P 0.015),
although not from A. senilis (P 0.081). However, M. marocanus
engaged in nearly as much paper-tugging behaviour as it did in
rescue behaviour (Fig. 2). Paper tugging was not included in the
combined rescue score because it appeared to be primarily a way to
remove the victim from near the nest. That is, once M. marocanus
revealed part of the lter paper by digging or pulling, the paper was
carried several centimetres away from the nest and the nestmate
was ignored for the remainder of the trial. Analysis of paper tugging
revealed a signicant species by test type interaction
(F
12,285
2.049, P 0.02), with M. marocanus engaging in signi-
cantly more paper tugging than any of the four remaining species
(all Ps 0.014).
Messor barbarus engaged in virtually no rescue behaviour
whatsoever. In one test, a single instance of snare biting occurred
and, in another, a single instance of sand transport. However, just
Homocolonial
Heterocolonial
Heterospecific
Control
1
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
L. grandis A. senilis M. marocanus M. barbarus C. floricola
0.9
M
e
a
n

p
r
o
p
o
r
t
i
o
n

o
f

i
n
t
e
r
v
a
l
s
Figure 1. Rescue behaviour in sand-dwelling ants in response to different test stimuli. Mean SE proportion of intervals in which at least one individual engaged in rescue
behaviour (i.e. combined rescue score) in response to live test stimuli (a homocolonial nestmate, a heterocolonial ant or a heterospecic ant held in place using a lter paper and
thread snare) or a control test stimulus (an empty paper-and-thread snare). Five species of sand-dwelling Mediterranean ants were tested separately: Cataglyphis oricola, Lasius
grandis, Aphaenogaster senilis, Messor marocanus and Messor barbarus.
Table 2
Statistical analyses of individual rescue and aggressive behaviour patterns in sand-
dwelling ants using univariate ANOVA
Behaviour F df P Effect sizes
Digging
Species 18.29 4, 285 <0.0001 0.204
Test type 78.60 3, 285 <0.0001 0.453
Speciestest type 19.88 12, 285 <0.0001 0.456
Pulling
Species 41.58 4, 285 <0.0001 0.368
Test type 152.03 3, 285 <0.0001 0.615
Speciestest type 44.29 12, 285 <0.0001 0.651
Snare biting
Species 13.55 4, 285 <0.0001 0.160
Test type 42.34 3, 285 <0.0001 0.308
Speciestest type 14.31 12, 285 <0.0001 0.376
Sand transport
Species 25.96 4, 285 <0.0001 0.267
Test type 40.14 3, 285 <0.0001 0.297
Speciestest type 22.76 12, 285 <0.0001 0.489
Paper tugging
Species 1.88 4, 285 0.113 0.026
Test type 1.37 3, 285 0.252 0.014
Speciestest type 2.05 12, 285 0.02 0.079
Biting
Species 17.24 4, 285 <0.0001 0.195
Test type 144.75 3, 285 <0.0001 0.604
Speciestest type 30.90 12, 285 <0.0001 0.565
Dismembering
Species 63.31 4, 285 <0.0001 0.471
Test type 135.82 3, 285 <0.0001 0.588
Speciestest type 45.90 12, 285 <0.0001 0.659
Projecting
Species 16.82 4, 285 <0.0001 0.191
Test type 20.19 3, 285 <0.0001 0.175
Speciestest type 18.52 12, 285 <0.0001 0.438
Retreating
Species 6.51 4, 285 <0.0001 0.084
Test type 5.48 3, 285 <0.001 0.055
Speciestest type 6.69 12, 285 <0.0001 0.220
For each behavioural pattern, the entries under species, test type and spe-
cies test type refer to the source of between-subjects effects. Effect sizes refer to
partial eta-squared.
K. L. Hollis, E. Nowbahari / Animal Behaviour 85 (2013) 537e544 540
prior to the snare biting, a portion of the paper had been exposed
and ants were engaged in moving the paper away from the nest.
Once moved, the paper and nestmate were completely ignored.
Thus, like the behaviour of M. marocanus, rescue in M. barbarus
appears primarily as a way to remove an object from near the nest
entrance.
Tests with either a heterocolonial or heterospecic ant elicited
extremely low levels of the four rescue behaviours, if any at all, in
all ve species, which were not signicantly different from one
another (all Ps 1.000). Exactly the same pattern of results was
obtained when combined rescue scores were analysed instead (all
Ps 1.000); moreover, no differences were found between the
different heterospecic victims in species tested with more than
one species of heterospecic ant. Finally, control tests elicited
virtually no rescue behaviour in any of the ve species. None the
less, the failure to respond with rescue behaviour in control tests
cannot be attributed to the failure to detect the presence of the
lter paper, as ants from all species repeatedly walked over the
lter paper and antennated the empty thread loop.
Aggressive Behaviour
Combined aggressive score
Cataglyphis oricola and L. grandis, the two species that deliv-
ered the most rescue behaviour to nestmates, not only refused such
aid to conspecics from a different colony, but, instead, responded
to them with especially high levels of aggression. Moreover,
M. barbarus, which barely responded to nestmates with any formof
behaviour, either rescue or paper tugging, vigorously attacked
heterocolonial conspecics (Fig. 3).
An ANOVAof the combined aggressive score revealed signicant
main effects of species (F
4,285
42.16, P < 0.0001) and test type
(F
3,285
279.99, P < 0.0001), as well as a signicant interaction
between species and test type (F
12,285
54.86, P < 0.0001).
Exploration of that interaction, using Bonferroni post hoc tests
adjusted for multiple comparisons, revealed that in the presence of
a heterocolonial ant, C. oricola, L. grandis and M. barbarus engaged
in levels of aggressive behaviour that were similar to one another
(Ps 0.91), but which differed from the two remaining species (all
Ps < 0.0001), both of which engaged in signicantly lower levels of
aggressive behaviour.
Individual aggressive behaviours
This same pattern of results is reected in the analysis of two
individual aggressive behaviours, biting and dismembering of heter-
ocolonial victims. Exploration of the signicant interactions between
species and test type for biting and dismembering (see Table 2)
revealed that C. oricola, L. grandis and M. barbarus engaged in
signicantly higher levels of heterocolonial biting and dismembering
than did M. marocanus and A. senilis (all Ps 0.001; see Fig. 4a). No
signicant differences between species appeared in the frequency of
projecting (all Ps 1.000), which was rare in heterocolonial tests.
Compared to heterocolonial tests, a very different pattern of
aggressive behaviour across species was elicited by heterospecic
victims (see Fig. 4b). Although L. grandis responded aggressively to
both heterocolonial and heterospecic victims (P 0.246), neither
M. barbarus nor C. oricola did so (all Ps 0.001). Indeed, C. oricola
did not respond at all to heterospecics, even the comparatively
much smaller L. grandis victims, except to retreat from them.
Retreating by C. oricola took exactly the same form on every
heterospecic trial: As soon as the victim was positioned near the
nest, all nearby foragers ed back inside; repeatedly thereafter, one
or more foragers would appear at the nest entrance and then
immediately jerk back and disappear. This pattern of appearing and
retreating continued until the victimwas removed at the end of the
trial, whereupon multiple foragers emerged and resumed activity.
Statistical analysis conrmed that C. oricola showed signicantly
less aggression in heterospecic tests than all other species
(Ps 0.001) except, perhaps, for M. barbarus, which barely reached
a statistically signicant difference (P 0.048).
Figures 3 and 4 also showthat heterocolonial and heterospecic
victims produced dramatically different patterns of behaviour in
A. senilis. Whereas A. senilis responded with weak aggression to
conspecics from another colony, ants belonging to a different
species elicited much higher levels of aggression (P 0.001), which
were not signicantly different from those of the highly aggressive
L. grandis (P 0.247). None the less, the individual aggressive
behaviour patterns differed slightly between these two aggressive
species, with L. grandis engaging in signicantly more dismem-
bering and defensive projecting than A. senilis (both Ps 0.001).
Finally, aggression never was observed in the presence of
a homocolonial nestmate, except on very rare occasions in two of
the ve species, A. senilis and M. marocanus, nor was it ever
Digging
Pulling
Snare biting
Sand transport
Paper tugging
1
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
L. grandis A. senilis M. marocanus M. barbarus
0.9
0.8
M
e
a
n

p
r
o
p
o
r
t
i
o
n

o
f

i
n
t
e
r
v
a
l
s
C. floricola
Figure 2. Rescue behaviour in sand-dwelling ants in response to homocolonial nestmates. Mean SE proportion of intervals in which ants showed rescue behaviour (sand digging,
limb pulling, snare biting or sand transport) or nonrescue behaviour (paper tugging) in response to a homocolonial nestmate. Five species of sand-dwelling Mediterranean ants
were tested separately: Cataglyphis oricola, Lasius grandis, Aphaenogaster senilis, Messor marocanus and Messor barbarus.
K. L. Hollis, E. Nowbahari / Animal Behaviour 85 (2013) 537e544 541
observed in any of the control tests. Statistical analyses conrmed
that no differences existed between species on either homocolonial
or control tests (all Ps 1.000).
DISCUSSION
We examined ants in nature to ensure that rescue behaviour was
neither conned to the single species previously studied (Nowbahari
et al. 2009), nor reected a laboratory artefact, and we succeeded on
both counts. In the presence of trapped nestmates, ants of both
C. oricola and L. grandis immediately approached the victims and
began to performmultiple behaviours to release them, including the
newly discovered, more complex forms of rescue, snare biting and
sand transport. Also similar to the laboratory experiments with
C. cursor, and critical to our understanding of rescue behaviour more
generally, C. oricola and L. grandis rarely if ever attempted to rescue
non-nestmates, even conspecics from nearby colonies. Indeed, not
only were heterocolonial conspecics attacked aggressively, but also
the patterns of aggressive behaviour towards heterocolonial and
heterospecic victims reveal ants capacity to discriminate between
different types of victims, as was demonstrated in C. cursor
(Nowbahari et al. 2009). Both C. oricola and L. grandis, as well as
C. cursor, belong to the subfamily Formicinae.
In stark contrast to the behaviour of the two new rescuer
species (C. oricola and L. grandis), the three remaining species in
this study, namely M. marocanus, M. barbarus and A. senilis
(subfamily Myrmicinae), rarely engaged in rescue behaviour.
However, note that the same rescue behaviour that occurred so
much more frequently, and, one might say, frenetically, in
C. oricola and L. grandis, including snare biting and sand transport,
appear to be within the behavioural capacity of the three non-
rescuer species. That is, although these nonrescuer species
neglected nestmate victims nearly all of the time, they nevertheless
are capable of performing precision rescue behaviour, as evidenced
by those occasions in which they executed the same behaviour
patterns, and in exactly the same way, as the two rescuer species.
Although additional comparative work is required, our analysis
of sand-dwelling ants behaviour reects some aspects of these
species behavioural ecology. Lasius grandis nests were located in
easily disturbed soils, which could result in ants becoming buried
under collapsing sand and debris. Evidence for such risks comes
from the frequent appearance of wild ungulates in the area (e.g. Sus
scrofa), which occasionally trampled nests. In addition, this
microhabitat was shared by many predatory pit-digging antlions.
Although C. oricola could sometimes be found in more compact
soils than L. grandis, C. oricola also located its nests in loose sandy
soils shared by antlions. Interestingly, both C. oricola ants and pit-
digging antlions are drawn to a common owering shrub of these
sandy soils, Halimium halimifolium. Antlions dig pits under the
overhanging branches of this shrub, which provides shelter from
wind and rain (Grifths 1986; see Hollis et al. 2011, for a review)
and C. oricola ants feed heavily on its petals (Cerd et al. 1996).
Very different fromthe microhabitats of the two rescuer species,
the nests of M. marocanus and M. barbarus were found in much
more hardened soils, nowhere near the vicinity of antlions pit
traps. In addition, both M. marocanus and M. barbarus workers
foraged in very large groups, forming ant trails that often extended
10e15 m or more from the nest entrance. These trails are estab-
lished only when foragers return to the nest after locating a source
of food (Cerdan 1989; Hlldobler & Wilson 1990) and thus could
effectively prevent them from encountering antlion pits and other
sources of danger. In addition, the steady stream of nearby nest-
mates might offer a kind of self-rescue in which trapped foragers
could grab onto the legs or bodies of passing nestmates. Although
L. grandis also makes trails, solitary workers often were observed
foraging at some distance from the nest.
In contrast to the four species just discussed, the behaviour of
A. senilis is puzzling. Although A. senilis almost always was found in
loose, sandy soils within a few metres of L. grandis, A. senilis rarely
engaged in rescue behaviour. However, A. senilis does not share the
same behavioural prole as the two Messor species either, even
though all three belong to the subfamily Myrmicinae. That is,
1
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
A. senilis M. marocanus M. barbarus
0.8
C. floricola
Homocolonial
Heterocolonial
Heterospecific
Control
L. grandis
0.9
M
e
a
n

p
r
o
p
o
r
t
i
o
n

o
f

i
n
t
e
r
v
a
l
s
Figure 3. Aggressive behaviour in sand-dwelling ants in response to different test stimuli. Mean SE proportion of intervals in which at least one individual responded aggressively
(i.e. combined aggressive score) to live test stimuli (a homocolonial nestmate, a heterocolonial ant or a heterospecic ant held in place using a lter paper and thread snare) or
a control test stimulus (empty paper-and-thread snare). Five species of sand-dwelling Mediterranean ants were tested separately: Cataglyphis oricola, Lasius grandis, Aphaenogaster
senilis, Messor marocanus and Messor barbarus.
K. L. Hollis, E. Nowbahari / Animal Behaviour 85 (2013) 537e544 542
whenever A. senilis engaged in any form of rescue behaviour,
rescuers did so for extended periods rather than sporadically, as in
Messor. Moreover, A. senilis rescuers never engaged in paper
tugging, the behaviour used by Messor ants to move the victim
away from the nest. Thus, A. senilis does not completely t either
the rescuer or the nonrescuer prole. A close analysis of variables
that might have affected our observations, including weather
conditions, time of day or month, and locations of nests, revealed
no consistent patterns of this species tendency to rescue or not.
Anal comment speaks tothegeneralityof rescue across multiple
taxa and the care that researchers must exert when attempting to
explain how such complex behaviour as rescue comes about. In an
experiment very similar in design to that used by Nowbahari et al.
(2009) with ants, Bartal et al. (2011) showed that rats behaved
similarly when they encountered a restrained cagemate, pulling on
the victims tail and biting the restraining tube. Moreover, when
giventhe opportunity to free their cagemate by tippingtherestrainer
door, rats eventually learned to do so. In our eld study, too, and
despite efforts to prevent escape, ants occasionally released victims
by biting repeatedly through the thread. The authors of the rat study
made the claim that rescuers were able to understand their cage-
mates affective state, and that the results provide evidence of
intentional empathy, a higher level of empathic behaviour (de Waal
2011). However, the behaviour of ants reminds us how exceedingly
complex behaviour can come about through relatively simple
mechanisms. As Vasconcelos et al. (2012) argued, future experiments
exploringintentional empathyinanimals will needtoincludecontrol
groups that help scientists to distinguish between analogous
behaviour involving different mechanisms.
Acknowledgments
For helpful comments on our manuscript, we thank Alain Lenoir
and an anonymous referee. For access to Reserva Cientca de
Doana and funding of this project through two separate grant
applications, we thank the ICTS-RBD Program. Several individuals
at Doana were particularly helpful: Xim Cerd, Ph.D., was an
invaluable resource on all matters pertaining to ants and the
1
(a)
(b)
0.6
0.5
0.4
0.3
0.2
0.1
0
L. grandis A. senilis M. marocanus M. barbarus
0.9
Biting
Dismembering
Projecting
Retreating
Biting
Dismembering
Projecting
Retreating
Heterocolonial test
M
e
a
n

p
r
o
p
o
r
t
i
o
n

o
f

i
n
t
e
r
v
a
l
s
0.8
M
e
a
n

p
r
o
p
o
r
t
i
o
n

o
f

i
n
t
e
r
v
a
l
s
0.7
1
0.6
0.5
0.4
0.3
0.2
0.1
0
0.9
0.8
0.7
Heterospecific test
C. floricola
L. grandis A. senilis M. marocanus M. barbarus C. floricola
Figure 4. Aggressive behaviour in sand-dwelling ants in response to heterocolonial and heterospecic test stimuli. Mean SE proportion of intervals in which ants showed
aggressive behaviour (biting, dismembering or projecting) or retreated in response to a live (a) heterocolonial ant or (b) heterospecic ant. Five species of sand-dwelling Medi-
terranean ants were tested separately: Cataglyphis oricola, Lasius grandis, Aphaenogaster senilis, Messor marocanus and Messor barbarus.
K. L. Hollis, E. Nowbahari / Animal Behaviour 85 (2013) 537e544 543
Doana reserve, as well as a generous and gracious host. Begoa
Arrizabalaga and Rosa Rodrguez provided enormous help, not only
throughout the application process, but also for the duration of our
stays at Doana. Additional support of this project was obtained
through a Faculty Grant from Mount Holyoke College (K.L.H.), as
well as from Laboratoire dthologie Exprimentale et Compare,
Universit Paris 13 (E.N. and K.L.H.). Finally, Janelle Gagnon
provided statistical support and Cheryl McGraw provided technical
assistance to meet gure specications.
References
Bartal, I. B.-A., Decety, J. & Mason, P. 2011. Empathy and pro-social behavior in rats.
Science, 334, 1427e1430.
Bernard, F. 1968. Les Fourmis (Hymenoptera, Formicidae) dEurope Accidntale et
Septentrionale. Faune de lEurope et du Bassin Mditerranen. No.3. Pars: Masson.
Botz, J. T., Loudon, C., Barger, J. B., Olafsen, J. S. & Steeples, D. W. 2003. Effects of
slope and particle size on ant locomotion: implications for choice of substrate
by antlions. Journal of the Kansas Entomological Society, 76, 426e435.
Belt, T. 1874. The Naturalist in Nicaragua. London: J. Murray.
Blum, M. S. & Warter, S. L. 1966. Chemical releasers of social behavior. VII The
isolation of 2-heptanone from Conomyrma pyramica (Hymenoptera: For-
micidae: Dolichoderinae) and its modus operandi as a releaser of alarm and
digging behavior. Annals of the Entomological Society of America, 59, 774e779.
Cerd, X., Retana, J., Carpintero, S. & Cros, S. 1996. An unusual ant diet: Cataglyphis
oricola feeding on petals. Insectes Sociaux, 43, 101e104.
Cerdan, P. 1989. Etude de la biologie, de lcologie et du comportement des fourmis
moissonneuses du genre Messor (Hymenoptera, Formicidae) en Crau. Ph.D.
thesis, University of Aix-Marseille I.
Czechowski, W., Godzi nska, E. J. & Kozowski, M. W. 2002. Rescue behavior
shown by workers of Formica sanguinea Latr., F. fusca L. and F. cinerea Mayr
(Hymenoptera: Formicidae) in response to their nestmates caught by an ant
lion larva. Annales Zoologici, 52, 423e431.
Devetak, D.,

Spernjak, A. & Janzekovic, F. 2005. Substrate particle size affects pit
building decision and pit size in the antlion larvae Euroleon nostros (Neuro-
ptera: Myrmeleontidae). Physiological Entomology, 30, 158e163.
Grifths, D. 1980. The feeding biology of ant-lion larvae: prey capture, handling and
utilization. Journal of Animal Ecology, 49, 99e125.
Grifths, D. 1986. Pit construction by ant-lion larvae: a costebenet analysis.
Journal of Animal Ecology, 55, 39e57.
Hangartner, W. 1969. Carbon dioxide, a releaser for digging behavior in Solenopsis
geminate (Hymenoptera: Formicidae). Psyche, 76, 58e67.
Hlldobler, B. & Wilson, E. O. 1990. The Ants. Berlin: SpringereVerlag.
Hlldobler, B. & Wilson, E. O. 2009. The Superorganism: the Beauty, Elegance, and
Strangeness of Insect Societies. London: Norton.
Hollis, K. L. & Guillette, L. M. 2011. Associative learning in insects: evolutionary
models, mushroom bodies, and a neuroscientic conundrum. Comparative
Cognition and Behavior Reviews, 6, 24e45.
Hollis, K. L., Cogswell, H., Snyder, K., Guillette, L. M. & Nowbahari, E. 2011.
Specialized learning in antlions, pit-dwelling insect predators, shortens
vulnerable larval stage. PLoS One, 6, e17958, http://dx.doi.org/10.1371/
journal.pone.0017958.
Laeur, L. J. 1940. Helpfulness in ants. Journal of Comparative Psychology, 30,
23e29.
Lpez, F., Acosta, F. J. & Serrano, J. M. 1993. Responses of the trunk routes of
a harvester ant to plant density. Oecologia, 93, 109e113.
Lucas, J. R. & Brockmann, H. J. 1981. Predatory interactions between ants and
antlions (Hymenoptera: Formicidae and Neuroptera: Myrmeleontidae). Journal
of the Kansas Entomological Society, 54, 228e232.
Markl, H. 1965. Stridulation in leaf-cutting ants. Science, 149, 1392e1393.
Martin & Bateson, P. 1993. Measuring Behaviour: an Introductory Guide. 2nd edn.
Cambridge: Cambridge University Press.
Mencinger-Vracko, B. & Devetak, D. 2008. Orientation of the pit-building antlion
larva Euroleon (Neuroptera, Myrmeleontidae) to the direction of substrate
vibrations caused by prey. Zoology, 111, 2e8.
Nowbahari, E. & Hollis, K. L. 2010. Rescue behavior: distinguishing between
rescue, cooperation, and other forms of altruistic behavior. Communicative and
Integrative Biology, 3, 1e3.
Nowbahari, E., Scohier, A., Durand, J.-L. & Hollis, K. L. 2009. Ants, Cataglyphis
cursor, use precisely directed rescue behavior to free entrapped relatives. PLoS
One, 4, e6573, http://dx.doi.org/10.1371/journal.pone.0006573.
Spangler, H. G. 1968. Stimuli releasing digging behavior in the western harvester
ant (Hymenoptera: Formicidae). Journal of the Kansas Entomological Society, 41,
318e323.
Tinaut, A. 1991. Contribucin al conocimiento de los formcidos del Parque Nacional
de Doana (Hymenoptera, Formicidae). Boletn de la Asociacin espaola de
Entomologa, 15, 57e63.
Vasconcelos, M., Hollis, K., Nowbahari, E. & Kacelnik, A. 2012. Pro-sociality
without empathy. Biology Letters, 8, 910e912.
de Waal, F. B. M. 2011. Empathy in primates and other mammals. In: The Oxford
Handbook of Social Neuroscience (Ed. by J. Decety & J. T. Cacioppo), pp. 87e107.
New York: Oxford University Press.
Wilson, E. O. 1958. A chemical releaser of alarm and digging behavior in the ant
Pogonomyrmex badius (Latreille). Psyche, 65, 41e51.
K. L. Hollis, E. Nowbahari / Animal Behaviour 85 (2013) 537e544 544