Insectes
Sociaux. Paris
1990,
Volume 37,
n"
3, pp. 258-267
©
Masson, Paris, 1990
COLLECTIVE
DECISION
MAKING
THROUGH
FOOD
RECRUITMENT
R.
BECKERS, J.L. DENEUBOURG, S. GOSS and J.M. PASTEELS
Unit
of
Behavioural
Ecology, CP 231, Université Libre de Bruxelles
Campus
de la Plaine, Bld. du Triomphe,
J050
Bruxelles, Belgium
Reçu
le 5 juillet 1989 Accepté le 8 novembre 1989
SUMMARY
A séries
of experiments shows how the ant
Lasius
niger uses its
trail
recruitment
System
to select between two food sources. Simuitaneously presented with two
IM
sucrose
solutions it
concentrâtes
on one of them. When offered a
IM
solution together
with
a
O.IM
solution it
sélects
the richer source,
uniess
the trail to the
O.IM
source had
become welldeveloped
before the
IM
source was introduced. In the same situation,
however,
the group/mass recruting ant Tetramorium
caespitum
uses its more
individual
transmission
of information to switch to the
IM
source. A mathematical
model
describes
thèse
processes and its dynamics reflect the
expérimental
results.
RESUME
La
prise de décision collective à travers le recrutement alimentaire
Nous
présentons une série d'expériences qui montrent comment la fourmi Lasius
niger
peut utiliser son système de recrutement par piste afin de sélectionner une des deux
sources
de nourriture. Si on leur offre simultanément deux solutions
IM
de saccharose,
la
société concentre son activité sur l'une des deux. Si l'on offre deux solutions, une de
IM
et l'autre de
O.IM,
elle sélectionne la plus riche, à moins que la piste qui mène à la
source O.IM
soit déjà bien développée au moment où l'on introduit
la
source
IM.
Face
à
la même situation, la fourmi Tetramorium caespitum, qui recrute par groupe/masse,
utilise
son mode de transmission d'information plus individuel pour changer son exploi
tation
vers la source
IM.
Un modèle mathématique décrit ces processus, dont la dynamique
correspond
bien aux observations expérimentales.
COLLECTIVE
DECISION
MAKING
259
INTRODUCTION
Sociality
provides an opportunity for the exchange of information
concerning
profitable foraging locations, and numerous food recruitment
behaviors,
the means by which a number of foragers are directed to a food
source,
have been described for very
différent
species (Social Insects :
WiLSON,
1971 ; Ants : HOLLDOBLER, 1978
;
PASSERA, 1984 ;
SUDD
and FRANKS,
1987
; Termites : LEUTHOLD, 1975 ;
RICKLI
and LEUTHOLD, 1986 ; Honeybees :
LiNDAUER,
1961 ; VON FRISCH, 1967 ; SEELEY, 1985 ; Gregarious Caterpillars :
FITZGERALD
and PETERSON, 1983, 1988). In the case of birds, recruitment can
either
be active or
else
simply a spying behaviour in which one animal sees
another
feeding and moves to join in (WARD and ZAHAVI, 1973 ;
EMLEN
and
DEMOMG,
1975 ; BROWN, 1986 ;
GOTMARK
et al, 1986 ; GREENE, 1987). Recruitment,
however,
is not
only
a system adapted to the exploitation of a patchy envi-
_
ronment or of prey needing
coopérative"
exploitation, but can
also
be the
touchstone
of a collective decision-making system developed by animal so-
cieties,
which we can term the natural
sélection
of food discoveries. The
interplay
between recruitments to
différent
food sources
générâtes
complex
social décisions well
beyond the capacity of an individual.
Small différences
in
the degree of cooperativity of the recruitment mechanism, as defined by
simple
individual
rules,
can
lead
to totally
différent
social
décisions
and
patterns
of food exploitation.
Thèse
ideas are
illustrated
for social insects,
the
most socially integrated of animais, with a
séries
of experiments on
ants
recruiting to sugar sources and a simple
model.
In
a typical ant
trail
recruitment, a scout discovers a food source and
returns
to the nest, laying a chemical trail. At the nest, other foragers
detect
the trail (we neglect invitational
différences)
and follow it to the
source
(e.g. SUDD, 1960 ; WILSON, 1962 ; WILSON, 1971 ; Ants : HOLLDOBLER, 1978 :
PASSERA,
1984 ; SUDD and FRANKS, 1987). Those that arrive at the source
load
food
and return to the nest reinforcing the trail. Unsuccessful
trail-followers
explore
the foraging ground before either returning to the nest or finding
the
food source. As the trail is reinforced, more ants are recruited and
less
recruits lose
the trail, and the recruitment is thus autocatalytic
(CORNETZ,
1914
; PASTEELS et al, 1987). Termite trail recruitment is very similar
(LEUTHOLD,
1975 ; RICKLI and LEUTHOLD, 1986).
MATERIALS
AND METHODS
fn
each cxperimcnt, a hungry colony of
Lasius
niger vvas
présentée!
with two
constantly
rcplenished sucrose solutions, 60 cm apart and both 60 cm from the nest in an
80
X 80 cm arena. The number of ants around each source was counted every minute.
The
experiments
were
performed on the trail recruiting ant Lasius niger and the group/
trail
recruiting ant Tetramorium caespitum. In this latter species, the number of ants
recruited
in groups were counted. Also, as it has a smaller natural foraging arca, the
sources
were offered 30 cm apart, 30 cm from the nest.
260 R. BECKERS
et al.
RESULTS
When
a colony was offered two identical sources at the same time, after
a
short period of
equal
exploitation, a bifurcation is observed and one of the
sources
becomes much more exploited than the other {fig. 1 A). If a second
is
discovered after a recruitment is
well
under awav to a first source, the
second
always remains underexploited.
Globalîy
speaking, the colony
"décides"
to exploit one source fuUy and to keep the second in reserve, the
trail
to it being maintained by a
low
levai of exploitation.
C
A
:
B
:
C
:
1,0
0.8
0.6
' 1
'
1
0
20 40
60
TIME
(min)
i
\
\
\
1
M
-, 0.1
M
0.0
B
'i '
0.1M
AH.
1M
20
40
60
TIME
(min)
80
1
00
20
30 100
Fig.
1. —
Présents
experiments on the
trail
recruiting ant
Lasiiis
niger and
the group/trail
recruiting ant Tetra-
morium
caespitum. In each expe-
riment,
a hungry colony was pre-
sented
with two sugar sources, and
the
proportion of ants around each
source
is presented as a function
of
time.
two IM
sucrose sources introduced simultaneously, with L. niger. (7 experi-
ments)
;
IM
source introduced 50 mins after a 0.1 M source, with
L.
niger (3
experiments) ;
IM
source introduced 60 mins after a
O.IM
source, with T. caespitum. (3
experiments).
40
60
TIME
(min)
Fig.
1. — Présente des expériences sur la fourmi Lasius niger, qui recrute par piste, et sur
la
fourmi Tetramorium caespitum, qui recrute par groupe/piste. Dans chaque expé-
rience,
on présente deux solutions de saccharose à une société affamée, et on mesure
la
proportion de fourmis autour de chaque source en fonction du temps.
A
: deux solutions 1 M introduites simultanément, avec L. niger (7 expériences) ;
B
: solution
1M
introduite 50 min. après une solution 0.1 M, avec L. niger (3
expériences)
;
C
: solution 1 M introduite 60 min. après une solution 0.1 M, avec T.
caespinim
(3
expériences).
COLLECTIVE
DECISION
MAKING
261
What
happens if the colony is offered two sources of
différent
sugar
concentrations
? If the richer source is discovered before or at the same time
as
the poorer one it is alvvays the more exploited. However, if the richer
source
is discovered after the poorer one, then it is
only
weakly exploited
{fig.
I B). As before, the colony exploits the first source discovered, even
if
this means neglecting a richer source for a poorer one. When faced with
two
simultaneouslv discovered unequal sources it chooses the richer of the
two.
Identical
experiments with the trail recruiting ants Iridomyrmex
hiimilis
and
Pheidole pallidula,
ail
three having very
différent
life-styles, gave the
same
qualitative resuits.
Honeybees
use a similar but
différent
recruitment mechanism
(LINDAUER,
1961
; VON FRISCH, 1967 ; SEELEY, 1985). A forager that has found a food source
personally
transmits the information concerning its location to nestmates
by
the dance ritual (in ant trail recruitment
Ihis
information is contained in
the
trail).
Honeybee
colonies when confronted with the same
décisions
as
described
above choose differentiy.
Whatever
the order of discovery, they
always
select the richer source and exploit two identical sources equally
(VELTHUIS, 1977
; SEELEY and
LEVIEX,
1987).
Honeybee
recruitment is similar in its
logic
to group recruitment in
ants,
in which a
successfui
scout personally guides a
small
group of nest-
Fig.
2. —
Présents
the number of T. caespitum
workers
recruited in groups to two
sequentially introduced
sucrose sources (0.1 M followed by 1 M). The arrows mark
the
moment in time at which each of the two food sources were introduced.
Fig.
2. — Présente le nombre d'ouvrières de T. caespitum recrutées en groupe vers deux
sources
de saccharose introduites en différé (0.1 M suivie par 1 M). Les flèches
indiquent
le moment auquel chaque source est introduite.
262 R. BECKERS
et ai.
mates lo
the source (e.g.
WILSON,
1971 ; in tandem recruitment, a third me-
chanism
close to group recruitment, a scout guides, personally and with care,
only
one recruit - MOGLICH et al., 1974). One
vvould therefore
expect the same
set
of
décisions
of choice as found in
honeybess. It
is, however, impossible
to
verify this as
ail
the species \ve know of that practice group recruitment
also
use trail recruitment. At the beginning of a recruitment, the recruiters
lead
groups to the food, but also
lay
a trail between the food and the nest. As
this
trail becomes strcnger, the groups
beccmc Icss fréquent
and trail
recruitment Lakes o\-er.
Thèse group/trail
recruiting species combine the two sets of
décisions
(PASTEELS
et ai, 1987). Like
honeybees
they exploit the richer source,
whether
dîscovered
first or second, and like trail
recmiting
ants they exploit two
equal
sources asymmetrically {fig. 1 C) (see also
FOWLER,
1987 for the same
qualitative
results for choice experim.ents between two
différent
food types
in
T.
caespitiun).
We show in figure 2 how groups oc T.
caespittim
appear at
the
discovery of each of two
sequentially
introduced sucrcse sources, only to
disappear
rapidly as mass recruitment takes
ov^er.
MATHEMATICAL
MODEL AND MONTE-CARLO SIMULATIONS
What
is behind
thèse différent
sets of
décisions
? A mathematical mode!
(see
also PASTEELS et al., 1987) enables us to
link
the collective
décisions
to
the
individual foragers' behavior which defines the
différent
recruitment
Systems
used and thus the model's kinetics.
Consider
a colony of N foragers, of which E are scouts or
lost
recruits.
Xj
are at food source i. and
N-E-IXi
are waiting in the nest to be recruited.
The
autocatlytic nature of the recruitment process can be represented by
stating
that the number of recruits per unit time to source i is given by the
product
between a rate constant,
a.,,
the number of foragers in the nest
and
the number of foragers actively involved in exploiting that source. Of
thèse
recruits, a fraction representing the recruitment accuracy,
fj,
reach the
source
and the
complément, l-fj,
become lost ants. The
scouts/lost
ants can
return
home on average everv
1/p
time units or can find one of the two
sources
every
1/c
time units. The foragers stay on average
1/b
time units at
the
source before returning to the nest. Thus ;
dXi/dt
=
a^Xil (N-SX.-E)
-
bXi
+ cE (i
=
1,2) (1)
dE/dt = Sfa,X^(l-fi)) (N-ZX,-E)
- pE - 2cE (2)
The
model's time
évolution
and stationary states reproduce the
compé-
tition
between the recruitments to the two sources. The
différence
between
the
three types of recruitment described lies in the recruitment accuracy,
f|.
(Note
that the
less
than perfect accuracy of
thèse
recruitment Systems may