Behavioral Ecology Vol. 13 No. 6: 827-831
© 2002 International Society for Behavioral Ecology
Inhibition of optimal behavior by social transmission in the guppy depends on shoaling
Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK
Address correspondence to J. Chappell. E-mail: jackie.chappell{at}zoology.oxford.ac.uk.
Received 10 August 2001; revised 1 March 2002; accepted 14 March 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Previous research has suggested social learning of foraging behavior can inhibit learning of the optimal behavior pattern. Based on their transmission chain design, we used small groups of guppies (Poecilia reticulata) to determine the degree to which the optimal behavior pattern was inhibited by socially learned information. A founder group was trained to take a long, energetically costly route to a food source. The members of this group were gradually replaced with naive conspecifics. Replicating the findings of the earlier researchers, it was clear that the behavior of the founders strongly influenced the behavior of the naive fish, probably through a process of local enhancement. When tested as a group, the naive fish chose the long route to the food source significantly more often than chance. Each naive fish was also tested in isolation. When tested alone, there was a significant tendency to choose the short route despite following the long route when tested as a group. These results suggest social learning does not inhibit learning of optimal behavior patterns but that a trade-off occurs when tested in the group condition. It is possible that the advantages for an individual fish of swimming with the shoal, and thus following the socially learned route, may have outweighed the potential energetic costs of taking this longer route.
Key words: allelomimesis, guppies, local enhancement, maladaption, Poecilia reticulata, social transmission, trade-offs.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Social transmission learning is described as the process whereby animal B learns "aspects of behavioral similarity" from conspecific A (Whiten and Ham, 1992
). It is widely accepted that social interactions can
influence the activities of animals resulting in efficient, adaptive behavior
(Galef, 1995,
1996). Information that can be
transmitted socially can refer to foraging and food collection, predation, and
mating (Bateson, 1988).
Intuitively, social learning is locally adaptive; it must result in the
transmission of "good" information relevant to the local
environment. If the local environment is relatively stable, the behavior
resulting from the information will be reinforced in the same way for both the
transmitter and receiver of the information, and thus only adaptive behavior
should be transmitted (Galef,
1995). However, theoretical arguments have been proposed against
Galef's theory. Feldman et al.
(1996), Laland et al.
(1996), and Laland
(1996) have suggested social
learning may not always be locally adaptive and may sometimes result in the
transmission of maladaptive information. Maladaptive information transmission
refers to the conveyance of behavioral patterns that reduce the fitness of the
receiver compared with the relative fitness of expressing an alternative
behavior pattern (Laland and Williams,
1998). Feldman et al.
(1996) suggested maladaptive
social learning is probable in variable environments, as the information
passed socially will lag behind the environmental changes to a certain degree,
thereby resulting in outmoded or inappropriate behavior patterns. Galef
(1995) maintained that no such
lag is likely because both the receiver and transmitter of information will
rapidly adjust their behavior to fit with the conditions to maintain the
behavioral reinforcement.
Laland and Williams (1998)
experimentally investigated whether social learning could result in
transmission of outmoded, maladaptive information. Their results supported
those of Feldman et al. (1996):
Social learning is not always adaptive in rapidly changing conditions where
the information may be outdated. Using the transmission chain design pioneered
by Curio et al. (1978), Laland
and Williams (1997) had
previously shown that the tendency to shoal in guppies (Poecilia
reticulata) facilitated social learning in a maze swimming task. Laland
and Williams (1998)
established founder populations of guppies and frequently manipulated their
environments so that they could be said to be rapidly changing. Founder fish
were trained to take a certain route to a food source, which initially was
short but was lengthened over time by moving the food farther away. The
founders were gradually replaced with naive fish that learned the route
preferences of their trained conspecifics. Even those naive fish that learned
to take the circuitous, energetically costly route to the food maintained the
route preference over the week-long testing period. Thus Laland and Williams
concluded maladaptive information could be passed through a population, as the
guppies maintained the preference of their founder population. Lone fish never
exposed to founders learned to take preferentially the shorter route much more
quickly than fish who learned from founders to take the longer, energetically
costly route.
It is possible however, that the naive fish chose the energetically costly
route after swimming with founders because the antipredation benefits of
remaining with the shoal out-weighed the costs incurred in taking the longer
route. Laland and Williams
(1997) showed that lone fish
continued choosing the route of founders when the two routes were equivalent
in length, but they did not test lone fish in a situation in which they could
choose between two routes that were not equivalent. In this study, we
replicated the transmission chain design of Laland and Williams
(1997,
1998), but we tested fish both
in a group and alone after they have acquired a suboptimal route from the
founders.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Subjects and apparatus
Subjects were 35 adult female guppies. Twelve fish were trained as demonstrators in the founder group, eight were used in the experimental group, and eight were used in the control group. The remaining seven fish were retained as spares and held in the stock tank. They were not used in the experiment. All the fish were purchased from The Goldfish Bowl, Oxford, UK, and were of comparable size and age. Domestic-strain female guppies were used so that results would be comparable with those of Laland and Williams (1998
) and for the reasons
cited in that article. After purchase, the 35 fish were left in the stock tank
for 3 days to allow them to settle, during which time they were fed once daily
with a standard tropical-fish flaked food. We noted distinguishing markings on
the fish during this time to aid identification.
Four identical glass tanks (one stock tank and three experimental tanks)
measuring 45 x 30 x 30 cm were used with 23 cm depth of water at
24°C. A transparent PVC screen was placed in each experimental tank before
testing to divide it lengthways. The screens were 30 cm high, with two 6-cm
diam holes at either end to allow the fish to swim from one side of the tank
to the other (see Figure 1a,b).
The center of each hole was 11 cm from the top of the screen (top of hole 1 cm
below the water line) and 5 cm from the nearest end. A second, smaller piece
of PVC approximately 10-cm wide was used as a trap door to cover the holes as
required (see Figure 1a). To
mark each hole, a nontoxic, 2-cm wide band of colored electrical tape was
placed around the perimeter. One end was surrounded by blue tape and the other
red. This apparatus is similar to that used by Laland and Williams
(1998), except that the
screens used here are transparent rather than opaque (we address the possible
implications of this difference in the Discussion). Floating feeders were used
to feed the fish during experimental trials. These were small, transparent
plastic cones suspended from rings of 3 cm diam, as can be bought from aquatic
shops, secured onto the edge of the tank using suction pads. Freeze-dried
bloodworms (Chironomous spp.), (approximately one or two worms per
trial per fish) were placed in the cones. For all test trials the feeders were
positioned 11 cm from the nearest wall of the tank, just 3 cm from the edge of
the blue hole (see Figure 1b). The red route to the food source was therefore approximately three times as
long as the short route. This positioning was previously determined by Laland
and Williams (1998) to be the
minimum difference possible to allow the fish to distinguish between the short
and long routes.
|
Experimental procedure
The design of the experiment was mixed: independent measures between the
experimental and control groups, but repeated measures between the two
conditions of the experimental trials. Experimental fish observed the route
choices of a trained founder group before testing. The control fish had
comparable experience with the tanks and feeders but received no specific
training from the experimenter or other fish. All founders were removed from
the tanks before testing, so only naive fish were present during testing.
Experimental and control fish had no interaction with each other and were
tested as separate groups. The independent variable was the test condition;
each subject was tested both in isolation and in a shoal. The dependent
variable was the preferred route to the food source, either long or short. We
counterbalanced order effects of testing by testing half of the control and
experimental fish in isolation first, and half in the group condition first
(four fish on day 1 and four on day 2). The short route was marked by the blue
hole in all test trials, and the long route by the red hole. Laland and
Williams (1998) showed that
fish exhibit no color biases between red and green holes, and a pilot study
was performed before the current study to ensure the same was true for the red
and blue colors used in this study. Tests on the hole preferences of each of
five fish over 20 trials revealed that their choices did not deviate
significantly from a binomial distribution (all tests, p > .05).
This simplified the design of the experiment, allowing the long route to be
indicated by the red hole in all conditions with no counterbalancing
required.
Hypotheses to be tested
Because the previous experiments by Laland and Williams
(1997,
1998) showed that the shorter
route is energetically less costly, our null hypothesis was that fish tested
in both groups and in the shoal and isolated conditions should prefer the
short route. We had two alternative hypotheses: (1) if the observed behavior
is purely the outcome of suboptimal socially transmitted learning, fish should
take the long route in both conditions (shoal and isolated) in the
experimental group, but the short route in both conditions of the control
group; (2) if following the socially transmitted longer route is the outcome
of a trade-off between the benefits of shoaling and the energetic costs of
taking the longer route, fish in the experimental group should take the long
route when in a shoal and the short route when isolated, and fish in the
control group should follow the short route in both conditions.
Fish in the stock tank
Throughout the training and testing period, all fish remaining in the stock
tank were fed flake food once daily and bloodworm from a floating feeder twice
a week to familiarize them with it. Once a week they were also removed from
the stock tank and placed into one of the empty experimental tanks in
isolation to allow them to become accustomed to being alone in the tanks, as
they would be tested in isolation later.
Training of founder fish
After the 3-day settling period, 12 fish were randomly selected from the
stock tank and placed into one of the experimental tanks. For the next 3 days,
this founder group was fed a small quantity of bloodworms once a day from the
floating feeder. On day 4 the training of the founder group to swim through
the red hole (long route) to the food source commenced. The dividing screen
was placed into the tank, with all the fish on one side of it and the feeder
placed on the other. Initially, the feeder was placed almost opposite the red
hole to encourage the fish to swim through the target hole to reach the food
source (see Figure 1a). If any
fish approached the blue hole, it was blocked off with the trap door to
prevent them passing through, which acted as a mildly aversive stimulus (see
Laland and Williams, 1998).
When the fish had swum to the feeder, it was removed and placed on the other
side of the tank, the same distance from the red hole. This equaled one
shuttle trial. When training began, the fish were only required to complete
five such shuttle trials before the screen was removed. As the fish became
more competent at the task, they were required to complete 7 and eventually 10
shuttle trials every day. Some fish shuttled through the hole much quicker
than others, thus having more time to feed. It was necessary to use fewer
trials at the onset of training to prevent any of these fish from becoming
bloated and ill from receiving too much food. At the end of the shuttle
trials, we removed the screens and feeders.
As the fish became more proficient with the task, we moved the feeder away from the red hole toward its target position: 3 cm from the edge of the blue hole. The training period lasted for 28 days in total, but the fish were actually only presented with the apparatus on 21 occasions: every third or fourth day was a rest day. This again was necessary to prevent the fish from becoming bloated and ill. During the course of the training, three of the founders died. By the end of the training period the remaining nine founders shuttled rapidly from one side of the tank to the other via the target hole to reach the food source.
Establishing the experimental group
After completion of the founder training, one founder was removed from the
experimental tank, placed back into the stock tank, and replaced with a naive
fish. Twenty-four hours later, the screen and feeder were placed in the tank
with the 8 founders and 1 naive fish, and the 10 shuttle trials commenced,
with the absence of the trap door. At the end of the 10 trials, a second
founder was removed and another naive fish added. This was repeated over the
next 7 days, removing the remaining 2 founders on the last day, leaving 8
naive fish in the tank who had received training only from the founder fish,
not the experimenter.
Testing the experimental and control groups
Twenty-four hours after removing the last founders, we began the testing
trials. Fish 1 and 2 were removed from the tank and each placed into one of
the two isolation tanks. We recorded the route choices of fish 3 and 4 in the
shoal tank and of fish 1 and 2 in the isolated tanks over the 10-trial
free-choice experiment. These four fish were then switched between tanks and
the process was repeated: fish 1 and 2 were tested together in the shoal tank,
and fish 3 and 4 were tested alone in the isolation tanks. On day 2 of
testing, we repeated this process using the fish numbered 5-8.
The control test consisted of randomly selecting eight new, previously untrained fish from the remainder in the stock tank, numbering them from 1 to 8, and testing in exactly the same way as the experimental fish. In all tests, we terminated the trial if the fish did not move to the other side of the tank within 10 min.
We used this design principally because we were testing a specific
hypothesis generated by the experiment carried out by Laland and Williams
(1998), and so we wanted to
replicate their methodology. However, it should be noted that there are
potential problems of lack of independence because the shoal is tested
together (ideally, one would use a different shoal to test each fish).
Nevertheless, given that we are interested in whether fish make different
route choices when in a shoal and alone, and not why they prefer the long
route when in a shoal, we think that the results are still valid.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Table 1 summarizes the route choices made by all control and experimental fish during the 10 trials of both the isolated and group conditions. The data are given as proportions because not all fish completed all 10 trials. The table also shows how many trials each fish completed in both the isolated and group conditions.
|
It is apparent that, although the control group preferred the short route in both conditions, fish in the shoal condition of the experimental group preferred the long route, so our null hypothesis can be rejected. Furthermore, although the experimental fish preferred the long route when tested in the group condition, they chose the short route more often than the long (with the exception of fish number 5) when tested in isolation. This supports our second alternative hypothesis, suggesting that following the longer route when in a shoal is the outcome of a trade-off between the benefits of shoaling and the energetic costs of taking the longer route. Only seven fish participated in the isolated control trials because control fish number 7 showed severe stress at being alone in the tank. The data from this fish were not included in the analysis.
We applied one sample t tests to the data to test whether the distribution of their visits to the feeder by the long and short routes were significantly different from that expected by chance. In all cases, the distribution of visits was significantly different from that expected by chance (experimental group, shoal condition; t = 5.245, df = 7, p = .001, experimental group, isolated condition; t = -2.809, df = 7, p = .026, control group, shoal condition; t = -5.828, df = 7, p = .001, control group, isolated condition; t = -4.268, df = 6, p = .005; all tests two tailed). To test whether there was an effect of group (experimental or control) and condition (shoal or isolated testing) on the proportion of visits to the feeder via the long route, we applied a general linear model (Minitab 13, Minitab, Inc., USA) to the data. The model included a nested random factor, fish (condition), to account for the fact that condition was a repeated measure within a group. We found a significant interaction between condition and group (F1,14 = 55.55, p < .001). This suggests that the difference in the response of the fish between conditions depended on which group they belonged to (Figure 2). Fish exposed to founders trained to take the long route will only exhibit a tendency to take that route when swimming in a shoal. When alone, the fish suppress this acquired tendency, choosing the short route more than the long. For control fish, which were never exposed to any founder groups, there were no socially acquired preferences, and the preference for the short route was consistent over both conditions.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Our results replicate Laland and Williams's (1998
) finding that
experimental fish preferentially chose the route of their founder population
(here the long route), when tested in shoaling groups. This supports their
suggestion that one possible function of shoaling may be to encourage social
learning and that a maladaptive behavior may be transmitted through a
population. The results of the present experiment also showed that the same
fish switch strategies, significantly preferring the short route when tested
individually. This was true for all experimental fish except one, who,
although maintaining a preference for the long route, did use it less when
tested alone: fish 5 chose the long route on 100% of the group trials, but
only 80% when tested alone. The change in strategies between group and
isolated tests is not consistent with Feldman et al.'s
(1996) theory that social
learning inhibits learning of the optimal behavior pattern in changing
environments, as the optimal behavior was displayed by experimental fish when
tested alone. This implies that social learning is not maladaptive for an
individual fish overall, as they apparently retain the flexibility to show a
different (and energetically optimal) behavior when alone. The environment of
the founder population was manipulated to be variable; the food source was
moved away from the hole founders were trained to swim through, thus making
their hole preference maladaptive. It is inappropriate, though, to describe
the behavior of individual experimental fish as maladaptive in the group
tests, as staying with the shoal may have antipredation advantages
(Seghers, 1974) or provide
information about novel food sources
(Lachlan et al., 1998). This
gain could outweigh the energetic costs of taking the long route. That is,
experimental fish may have been trading off energetic costs against gains in
safety, information, or some other benefit of aggregation.
A similar switch in the behavior of individuals between social and
nonsocial conditions has also been observed in starlings. Vásquez and
Kacelnik (2000) presented
starlings with two depleting patches, which replenished when the alternative
patch was visited. The treatments differed in the contents of an aviary
adjacent to one of the patches: In the social condition it contained
starlings, in a control-flock condition it contained zebra finches, and in the
control condition it was empty. According to the marginal value theorem
(Charnov, 1976), starlings
should alternate between the two patches to maximize their rate of gain in all
of the conditions. However, although the starlings showed rate maximization as
expected while in the control and control-flock conditions, they spent more
time in the patch adjacent to the flock aviary when in the social condition
and, as a consequence, reduced their gain rate by about 24%.
One proposed mechanism that might allow groups to maintain collective
activity is allelomimesis ("the probability of an individual adopting a
particular behaviour or state is an increasing function of the number of
individuals already exhibiting that behaviour or state,"
Deneubourg and Goss, 1989:
296). Deneubourg and Goss
(1989) cite a number of
instances where different patterns of behavior can be generated by changing
the environmental conditions, while the rules that individuals use remain the
same. There are also instances in which maladaptive (i.e., non-rate
maximizing) behavior can be induced and maintained in a group, both in the
field and in simulations modeling the effect.
Laland and Williams (1997,
1998) used opaque dividing
screens for training and testing, whereas here the screens were transparent.
Therefore, in this experiment, the fish had access to visual information about
the location of the feeder on the opposite side of the screen, relative to the
red and blue holes. This information was available to all control and
experimental fish on all trials. This feeder location information is
noteworthy for two reasons. First, this visual information can be used by fish
with no previous experience of the maze to aid learning of the optimal
behavior pattern. This is apparent from the significant tendency of control
fish to choose the shorter route rather than initially responding at chance
level to the holes. Second, presuming experimental fish could also see that
the feeder was located nearer to the blue hole, it is interesting that they
continued to use the red hole preferred by the founder population. Again, this
replicates an observation of Laland and Williams
(1998: 498), who noted that
"subjects with no prior conditioning oriented themselves toward the hole
where it would appear they could see the feeder, and yet still swam the long
route with the rest of the group." This emphasizes the importance to an
individual fish of remaining with the shoal. It seems likely that the
transparent screen allowed the fish to identify the best route, and when the
trade-off between shoaling and energy costs was removed by removing the shoal,
the fish used the energetically cheaper route. It also suggests that learning
of the optimal behavior was independent of the socially learned route
behavior.
This study does not support the conclusions of Feldman et al.
(1996) and Laland
(1996) because the behavioral
responses of isolated fish were not maladaptive. The potential trade-off
between energy costs and the benefits of shoaling behavior requires further
investigation. In natural environments, animals would experience a greater
selection pressure to behave optimally because of the increased risk of
predation and/or prey escaping if foraging was not efficient. Here a quicker
rate of change of socially learned behavior patterns might be expected due to
reinforcement (Galef, 1995).
So, even in a changing natural environment, it is unlikely maladaptive
socially learned behavior patterns will persist for long, and we have shown
that these socially learned behavior patterns do not inhibit the individual
learning of adaptive behaviors.
|
ACKNOWLEDGEMENTS |
|---|
We thank Julian Howe for his help in the lab and Rebecca Smith, Steven Siller, Suzie Pomfret, Nick Bayly, and Esteban Fernández-Juricic for their help on the original manuscript.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Bateson PPG, 1988. The active role of behavior in evolution. In: Evolutionary processes and metaphors (Ho MW, Fox S, eds). Chichester, UK: Wiley; 191-207.
Charnov EL, 1976. Optimal foraging, the marginal value theorem. Theor Popul Biol 9: 129-136.[ISI][Medline]
Curio E, Ernst U, Vieth W, 1978. The adaptive significance of avian mobbing II. Cultural transmission of enemy recognition in black-birds: effectiveness and some constraints. Z Tierpsychol 48: 184-202.
Deneubourg JL, Goss S, 1989. Collective patterns and decision-making. Ethol Ecol Evol 1: 295-311.
Feldman MW, Aoki K, Kumm J, 1996. Individual vs. social learning: evolutionary analysis in a fluctuating environment. Anthropol Sci 104: 209-223.
Galef BG Jr, 1995. Why behaviour patterns that animals learn socially are locally adaptive. Anim Behav 49: 1325-1334.
Galef BG Jr, 1996. Introduction. In: Social learning in animals: the roots of culture (Heyes CM, Galef BG Jr, eds). San Diego, California: Academic Press; 3-15.
Lachlan RF, Crooks L, Laland KN, 1998. Who follows whom? Shoaling preferences and social learning of foraging information in guppies. Anim Behav 56: 181-190.[ISI][Medline]
Laland KN, 1996. Is social learning always locally adaptive? Anim Behav 52: 637-640.
Laland KN, Richerson PJ, Boyd R, 1996. Animal social learning: towards a new theoretical approach. In: Social Learning in animals: the roots of culture (Heyes CM, Galef BG Jr, eds). San Diego, California: Academic Press; 129-154.
Laland KN, Williams K, 1997. Shoaling generates social learning of foraging information in guppies. Anim Behav 53: 1161-1169.[ISI][Medline]
Laland KN, Williams K, 1998. Social transmission of
maladaptive information in the guppy. Behav Ecol
9: 493-499.
Seghers BH, 1974. Schooling behavior in the guppy (Poecilia reticulata): an evolutionary response to predation. Evolution 28: 486-489.[ISI]
Vásquez RA, Kacelnik A, 2000. Foraging rate versus sociality in the starling Sturnus vulgaris. Proc R Soc Lond B 267: 157-164.[Medline]
Whiten A, Ham R, 1992. On the nature and evolution of imitation in the animal kingdom: reappraisal of a century of research. Adv Study Behav 21: 239-283.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  I. Coolen, A. J.W. Ward, P. J.B. Hart, and K. N. Laland Foraging nine-spined sticklebacks prefer to rely on public information over simpler social cues Behav. Ecol., September 1, 2005; 16(5): 865 - 870. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||