- Email: [email protected]
The social transmission of spatial information in homing pigeons
ANIMAL BEHAVIOUR, 1999, 57, 715–719 Article No. anbe.1998.1024, available online at http://www.idealibrary.com on
The social transmission of spatial information in homing pigeons THERESA BURT DE PERERA* & TIM GUILFORD†
*Animal Behaviour Research Group, Department of Zoology, University of Oxford †Instituto de Ecologia, Universidad de Nacional Autonoma de Mexico (Received 8 July 1998; initial acceptance 25 August 1998; final acceptance 21 October 1998; MS. number: 5937R)
We adapted a technique to explore the social transmission of spatial information in homing pigeons, Columba livia. Five demonstrator pigeons were first trained to find a food goal within an indoor arena. This arena consisted of nine lidded cups laid out within a 1212 grid on the floor. The task was to find the goal cup and flip the lid to obtain the food hidden within. Once the demonstrators had reached criterion the experiment proper began. During stage 1 of the experiment, 10 target birds, which had not previously been trained to find the goal, were introduced to the spatial task either in isolation or paired with a demonstrator. We measured how long they took to complete the task, the number of squares crossed on the grid, and the number of incorrect lids flipped. In stage 2, the target birds were introduced to the arena a second time, by themselves, and we compared the performance of the birds in the two treatments. The pigeons that had been introduced to the task with a demonstrator in stage 1 walked further and made more incorrect choices when searching for the food goal in stage 2 than the pigeons that were introduced to the task alone. This indicates that pigeons learn a spatial, food-finding task more effectively when performing the task alone than when accompanied by a knowledgeable conspecific. We discuss possible reasons for this in the light of previous experiments.
situation in which one may expect an individual to learn from others is when it is presented with a spatial task. However, to our knowledge, no experiments have aimed to test this ability in animals. An individual that can learn spatial information from others would be expected to be at an advantage as it may learn rapidly and make fewer errors than if performing a task by trial and error. We would expect such an ability to have wide ecological significance in an animal’s natural environment. This is particularly the case for animals such as homing pigeons, Columba livia, which are naturally gregarious and which are frequently faced with orientational problems. We investigated the ability of homing pigeons to learn a spatial task within a social learning paradigm, with the aim of discovering whether information about a spatial task can be transferred between a knowledgeable pigeon and an observer. Demonstrator pigeons were trained to find a food goal within an indoor arena. Target birds without this knowledge were then introduced to the same arena either in isolation, or paired with a demonstrator. The target birds were then tested on the same task to ascertain whether the presence of a demonstrator influences learning of the spatial task. The null hypothesis was that the presence of a knowledgeable conspecific during training would not influence the efficiency of a pigeon to complete a spatial, food-finding task.
During social learning experiments, observer animals have often been shown to learn tasks more effectively when viewing knowledgeable demonstrators than when learning by trial and error (Dawson & Foss 1965; Palameta & Lefebvre 1985; Nicol & Pope 1992; McQuoid & Galef 1993). However, paradoxically, other studies reveal situations in which animals learn less effectively when observing skilled demonstrators than when watching unskilled demonstrators or when there was no demonstrator present (Robertson et al. 1985; Giraldeau & Lefebvre 1986, 1987; Biederman & Vanayan 1988; Stammbach 1988; Beauchamp & Kacelnik 1991; Templeton 1998). Thus, there is a suggestion that in some cases relying on an experienced animal’s knowledge may prevent an individual from learning a task itself where in other circumstances a task may be learnt more effectively in the presence of a demonstrator. These findings have been most frequently described for discrimination tasks or for experiments in which an animal had to learn how to perform a novel action to receive a reward. A similar Correspondence and present address: T. Burt de Perera, Instituto de Ecologia, Universidade Nacional Autonoma de Mexico, Apartado postal 70-275, Cayoacan, 04510 Mexico D.F., Mexico. T. Guilford is at the Animal Behaviour Research Group, Department of Zoology, University of Oxford, Oxford OX1 3PS, U.K. 0003–3472/99/030715+05 $30.00/0
1999 The Association for the Study of Animal Behaviour
715
1999 The Association for the Study of Animal Behaviour
716 ANIMAL BEHAVIOUR, 57, 3
1
2
3
4
5
6
7
8
were gradually pushed across and the number of cups containing food was reduced to one. The position of this goal varied daily. Pretraining ended (after approximately 3 weeks) when all the experimental birds could flip off a lid from a fully covered cup.
9 10 11 12 A
7
1
5
B C D E
2
8
3
Goal positions
Experimental Training and Testing
F G H I
4
9
6
J K L Curtain
Figure 1. Representation of the arena. The numbers on the cups indicate the order in which the goals were presented. Note that these numbers indicating order correspond to those in Fig. 2.
METHODS The subjects were 15 1–2-year-old homing pigeons, both males and females, reared at the Oxford University Field Station, Wytham, U.K. These birds were kept at 90% of their free-feeding normal body weight, monitored by daily weighing, to increase their motivation to find food. To regulate their food during the times when they were not in an experiment, we housed the pigeons in a loft separate to the others in the colony. The loft measured 55 m and was outdoors. The birds were fed once a day on corn and mixed seed and received water ad libitum. We performed the study within a laboratory. A 1212 grid of squares was drawn on the lino floor (55 m) and nine ceramic feeding cups 8 cm in diameter with cardboard lids were positioned in this grid (Fig. 1). A curtain with a small viewing hole was hung at the entrance to the laboratory from one side to the other from 1 m above the ground to the floor. Pigeons entered the arena by being placed under this curtain. We kept the pigeons that were to be used in the experiment or training stage in holding boxes behind the screen until their release into the arena. The laboratory was lit overhead by fluorescent lights and the floor was cleaned between experimental trials.
Pretraining We initially trained the 15 pigeons individually to flip off the lids to receive a food reward of corn and mixed seed. This task was thought to be fairly natural as pigeons routinely search for hidden seed by using their beaks (Goodwin 1983). Initially all the cups contained food and a gap was left between the lid and cup enabling the pigeon to see the food. Switching the room light off before release and on again after a 5-s interval stopped the pigeons from flying on introduction to the test area. This technique was also used in the experimental tests. After the bird had fed for 1 min, the room lights were switched off and the bird was recaptured with the aid of a torch. We repeated this procedure three times a day. The lids
We randomly assigned the pigeons to two groups. Five birds were classified as ‘experienced’. These were trained to find food in one goal location per trial, and after training ended a criterion of five consecutive correct choices had been reached by each experienced pigeon. A correct choice was defined by a pigeon flipping the lid of the goal cup before any other. We introduced the birds into the arena four times per day and 4 days were needed for every bird to reach this criterion. The other 10 pigeons were grouped as ‘inexperienced’ and received no extra training. Note that all the birds were pretrained to flip lids; experience or inexperience relates to whether the birds were permitted to learn the position of the food goal. The experiment was divided into nine trials. The goal was in a different position in each trial; thus over the entire experiment each cup on the grid was defined as a goal once. Each trial was split into two stages.
Stage 1: training We initially allocated the target birds at random to one of two groups: (1) five pairs each comprising one experienced and one inexperienced bird; (2) five inexperienced, unpaired individuals. The target birds were those that were inexperienced with respect to the position of the goal. In this testing stage either the pair of experienced and target inexperienced birds were released into the arena together or the target inexperienced bird was placed individually into the arena. In the first case the experienced bird generally ran directly to the food goal, flipped off the lid and began eating. The inexperienced bird followed and also ate some of the food that had been produced by the demonstrator. In the other case the inexperienced bird had to find the goal by trial and error by searching a number of cups and flipping off their lids itself. After finding the goal, each bird was given 30 s to feed before recapture during which time the food was never depleted. After a 2.5-min retention period during which the birds were held in carrying boxes, stage 2 began.
Stage 2: testing For the testing stage we placed the food in the same goal as for training and replaced the lids. The target birds were again released and we recorded (1) the trajectories walked (calculated by counting the number of lines on the grid that the pigeon crossed before discovering the food; (2) the number of incorrect choices of lids flipped off; and (3) the time to find the goal. We compared birds that were unpaired and those that were paired in stage 1. We then trained the experienced pigeons to a new goal position for the next trial and the inexperienced birds
BURT DE PERERA & GUILFORD: SPATIAL LEARNING 717
were subjected to the other treatment. Each inexperienced bird was allocated to the paired and unpaired treatments approximately an equal number of times over the experiment, four times in one, five in the other.
125
Mean number of squares crossed
(a)
100
RESULTS 75
50
25
0
1
2
3
4
5
6
7
8
9
2
3
4
5
6
7
8
9
500 (b)
Mean latency (s)
400
300
200
100
0
1
5
Mean number of incorrect choices
(c)
4
3
2
1
0
1
2
3
4 5 6 Goal position
7
8
9
For each of the three variables tested in stage 2, we calculated means for all birds in each of the nine trials (and hence the nine cup positions) when the target bird was either paired with a companion or unpaired in stage 1 (Fig. 2). We analysed the data using two-tailed Wilcoxon signed-ranks tests. To test whether the birds did learn the goal location in stage 1, we compared the number of lids flipped by the target birds, the number of lines on the grids crossed, and the latency to find the goal in stages 1 and 2. Note that it was possible to perform this test only when the birds were placed individually into the arena during stage 1. In the paired condition the demonstrator flipped the lids, so we could not compare lids flipped by the target birds in both stages. Wilcoxon signed-ranks tests revealed that in stage 2, fewer lids were flipped over in all nine trials (Z= 2.668, N=9, P=0.008) and fewer squares were crossed (Z= 2.201, N=6, P=0.028; note that we only began to record trajectories from trial 4), but latencies to find the goal did not differ significantly (Z= 2.073, N=9, P=0.382). These results indicate that the pigeons had learnt the position of the food goal in stage 1. Having discovered that learning did occur in stage 1, we were able to analyse the effects of demonstrator presence in stage 1 on the efficiency of the target birds in finding the food goal in stage 2. Wilcoxon signed-ranks tests revealed that there were no significant differences between the paired and unpaired treatments in stage 2 in the mean latency to find the goal over all nine trials (Z= 1.362, N=9, P=0.173; Fig. 2). This was also the case for the median latency (Z= 0.252, N=9, P=0.799). However, the mean number of squares crossed (Z= 2.666, N=6, P=0.008) and the mean number of incorrect choices (Z= 2.200, N=9, P=0.023) were significantly higher for paired birds (Fig. 2). Thus, the presence of an experienced demonstrator hindered a pigeon learning the position of a food goal. Note that there are missing data points in this experiment. In some cases birds flew into the air knocking off the lids so the data point was no longer valid (two cases). At other times the demonstrator bird did not appear motivated to find the food and the target bird discovered the food first by trial and error (four cases). Occasionally the target birds would also lack motivation to find the goal. The test was abandoned if a pigeon had not found the goal in 45 min (six cases). There was no apparent
Figure 2. (a) Mean+SE number of squares crossed in the arena (see Fig. 1), (b) mean+SE latency to find the goal, and (c) mean+SE number of incorrect choices for all 10 target birds in stage 2 over nine trials. In stage 1 of the experiment, target birds were released in the arena either with a bird who knew the location of the goal () or alone ( ). In stage 2 all birds were alone.
718 ANIMAL BEHAVIOUR, 57, 3
pattern in the occurrence of these missing data points, and as we used mean values per bird in the analysis the results were not affected. Compared with paired birds in stage 1, unpaired birds in stage 1 took longer to locate food (Wilcoxon signedranks test: Z= 2.497, N=10, P=0.012), covered more of the arena (Z=2.201, N=6, P=0.03), and visited more unrewarded cups (Z= 2.599, N=10, P=0.01), so one possibility is that indirect effects (such as the possibility that paired birds were in the arena for less time because their partner found the food quickly) are the cause of less effective learning in stage 2. However, when each unpaired bird’s mean performance in stage 1 was regressed against that in stage 2 we found no evidence that variation in stage 1 performance affected stage 2 performance with respect to time to locate food (R2 =0.037, N=10, P=0.595), number of squares crossed (R2 =0.079, N=10, P=0.220), or number of cups visited (R2 =0.065, N=10, P=0.476). As the latency to discover the goal in stage 2 was not significantly different between the two treatments we also checked that a bird that spent more time in the arena in stage 1 did not cross more squares in stage 2, but this was not the case (R2 =0.010, N=10, P=0.947). DISCUSSION On the basis of the trajectory data (the number of squares crossed) and the number of incorrect choices of lids flipped, we can conclude that pigeons learn the position of a food goal more effectively when performing the task alone than when accompanied by a knowledgeable companion. We supposed that there is an advantage to a pigeon in learning the position of a food goal in this experiment, particularly as the birds were kept below their free-feeding weight. It is initially surprising that pigeons that were inexperienced with respect to the position of the food goal did not learn from a demonstrator as effectively as they learnt the task individually. However, other studies have reported similar results. For example, in a recent study Templeton (1998) showed that observer starlings, Sturnus vulgaris, learned a discrimination task more effectively when they watched demonstrators that made incorrect responses than those that watched correct responses. This suggests that starlings do not learn such discrimination tasks from knowledgeable companions. Our results were consistent with others that report this seemingly paradoxical finding (Robertson et al. 1985; Giraldeau & Lefebvre 1986, 1987; Biederman & Vanayan 1988; Stammbach 1988; Beauchamp & Kacelnik 1991). Two general hypotheses may explain our results. First, the pigeon may learn nothing in stage 1 when placed with a knowledgeable bird (the experienced bird may inhibit learning). Second, the inexperienced pigeon may learn something different to the experimental task when placed with the experienced bird, which may be useful for obtaining food in the presence of the demonstrator but which does not help a bird locate the goal on its own. The observers would not, therefore, attend to features of the environment in the arena and relate these features to
the position of the food. Alternatively, learning may be overshadowed by a demonstrator. The overshadowing theory supposes that either the approach of the demonstrator to the goal or the demonstrator itself acted as a reliable indicator of food in stage 1. Thus, the demonstrator may have in some way overshadowed signals needed to find the food goal in the arena in stage 2. The pigeon is a social species which usually forages in a flock. Therefore, it may be an advantage for individuals to follow other more knowledgeable birds and not necessarily to learn the task themselves. This scrounging strategy (Giraldeau & Lefebvre 1986) may be a more efficient behaviour for some members of the flock than finding the food themselves. We call an example by which the overshadowing hypothesis may operate the passenger/ driver effect. This assumes that the observer does not attend to the position of the goal because it is simply following the demonstrator in the same way, it is said, that a passenger of a car will often be less able to describe a route taken than a driver. This is, however, only one hypothesis that may operate within the overshadowing theory. For example, another possibility is that the bird per se may have an effect. As this experiment was not designed to investigate why pigeons did not learn from companions, this question is open for further study. The mechanisms we have suggested are perhaps the most interesting from a social learning perspective, but other lower level effects are also possible and, although strictly still the results of the demonstrator’s presence in stage 1, would be useful to eliminate in future experiments. First, we have already suggested (and inferred unlikely) that unpaired birds may gain more experience than paired birds in stage 1 because they spend longer searching, cover more ground, or visit more unrewarded cups, and that this enhances performance in stage 2. Second, the presence of the conspecific itself at the baited bowl might have reduced the strength of the food bowl as an unconditioned stimulus in stage 1 either directly because of lowered food intake (even though the bowl was nondepleting), or indirectly through social stresses. Since the pigeons habitually feed at crowded food bowls, and were not seen to exclude each other in this experiment, this final mechanism, although possible, seems unlikely. The protocol in this experiment did not allow us to test learning by a target bird paired with an inexperienced bird in stage 1 as it would not have been possible to ensure that the experience of the target bird was consistent. The target bird might have discovered the food itself or the inexperienced partner could have found it. However, such a test would allow some of the effects described above to be eliminated, and should be considered in future experiments of this kind. In conclusion, our results suggest that pigeons learn a spatial, food-finding task more effectively when performing the task alone than when accompanied by a knowledgeable conspecific. Reasons for this include the possibility that learning is inhibited by a demonstrator, perhaps because the observers develop a scrounging strategy in the presence of a knowledgeable bird. Alternatively, the demonstrator (in some way) may have
BURT DE PERERA & GUILFORD: SPATIAL LEARNING 719
overshadowed signals needed to find the food goal in the arena in stage 2. To our knowledge this is the first attempt to investigate the ability of homing pigeons to learn a spatial task from companions. The results suggest that orientational information may not be learnt from others and instead pigeons learn more effectively when performing a task by trial and error. This finding has an important significance for homing pigeons which are often faced with orientational problems. Acknowledgments We thank Jackie Chappell and two anonymous referees for helpful comments on the manuscript and BBSRC for funding the project. References Beauchamp, G. & Kacelnik, A. 1991. Effects of the knowledge of partners on learning rates in zebra finches, Taeniopygia guttata. Animal Behaviour, 41, 247–253. Biederman, G. B. & Vanayan, M. 1988. Observational learning in pigeons: the function of quality of observed performance in simultaneous discrimination. Learning and Motivation, 19, 31–43. Dawson, B. V. & Foss, B. M. 1965. Observational learning in budgerigars. Animal Behaviour, 13, 470–474.
Giraldeau, L. A. & Lefebvre, L. 1986. Exchangeable producer and scrounger roles in a captive flock of feral pigeons: a case for the skill pool effect. Animal Behaviour, 34, 797–803. Giraldeau, L. A. & Lefebvre, L. 1987. Scrounging prevents cultural transmission of food-finding behaviour in pigeons. Animal Behaviour, 35, 387–394. Goodwin, D. 1983. Pigeons and Doves of the World. 3rd edn. Ithaca, New York: Cornell University Press. McQuoid, K. M. & Galef, G. G. 1993. Social stimuli influencing feeding behaviour of Burmese fowl: a video analysis. Animal Behaviour, 46, 13–22. Nicol, C. J. & Pope, S. J. 1992. Effects of social learning on the acquisition of discriminatory keypecking in hens. Bulletin of the Psychonometric Society, 30, 293–296. Palameta, B. & Lefebvre, L. 1985. The social transmission of a food-finding technique in pigeons. What is learned? Animal Behaviour, 33, 892–896. Robertson, H. A., Vanayan, M. & Biederman, G. B. 1985. Observational learning and the role of confinement in pigeons: suppression of learning as a function of observing the performance of a conspecific. Journal of General Psychology, 112, 375– 382. Stammbach, E. 1988. Group responses to specifically skilled individuals in a Macaca fascicularis group. Behaviour, 107, 241– 266. Templeton, J. 1998. Learning from others’ mistakes: a paradox revisited. Animal Behaviour, 55, 79–85.
The social transmission of spatial information in homing pigeons THERESA BURT DE PERERA* & TIM GUILFORD†
*Animal Behaviour Research Group, Department of Zoology, University of Oxford †Instituto de Ecologia, Universidad de Nacional Autonoma de Mexico (Received 8 July 1998; initial acceptance 25 August 1998; final acceptance 21 October 1998; MS. number: 5937R)
We adapted a technique to explore the social transmission of spatial information in homing pigeons, Columba livia. Five demonstrator pigeons were first trained to find a food goal within an indoor arena. This arena consisted of nine lidded cups laid out within a 1212 grid on the floor. The task was to find the goal cup and flip the lid to obtain the food hidden within. Once the demonstrators had reached criterion the experiment proper began. During stage 1 of the experiment, 10 target birds, which had not previously been trained to find the goal, were introduced to the spatial task either in isolation or paired with a demonstrator. We measured how long they took to complete the task, the number of squares crossed on the grid, and the number of incorrect lids flipped. In stage 2, the target birds were introduced to the arena a second time, by themselves, and we compared the performance of the birds in the two treatments. The pigeons that had been introduced to the task with a demonstrator in stage 1 walked further and made more incorrect choices when searching for the food goal in stage 2 than the pigeons that were introduced to the task alone. This indicates that pigeons learn a spatial, food-finding task more effectively when performing the task alone than when accompanied by a knowledgeable conspecific. We discuss possible reasons for this in the light of previous experiments.
situation in which one may expect an individual to learn from others is when it is presented with a spatial task. However, to our knowledge, no experiments have aimed to test this ability in animals. An individual that can learn spatial information from others would be expected to be at an advantage as it may learn rapidly and make fewer errors than if performing a task by trial and error. We would expect such an ability to have wide ecological significance in an animal’s natural environment. This is particularly the case for animals such as homing pigeons, Columba livia, which are naturally gregarious and which are frequently faced with orientational problems. We investigated the ability of homing pigeons to learn a spatial task within a social learning paradigm, with the aim of discovering whether information about a spatial task can be transferred between a knowledgeable pigeon and an observer. Demonstrator pigeons were trained to find a food goal within an indoor arena. Target birds without this knowledge were then introduced to the same arena either in isolation, or paired with a demonstrator. The target birds were then tested on the same task to ascertain whether the presence of a demonstrator influences learning of the spatial task. The null hypothesis was that the presence of a knowledgeable conspecific during training would not influence the efficiency of a pigeon to complete a spatial, food-finding task.
During social learning experiments, observer animals have often been shown to learn tasks more effectively when viewing knowledgeable demonstrators than when learning by trial and error (Dawson & Foss 1965; Palameta & Lefebvre 1985; Nicol & Pope 1992; McQuoid & Galef 1993). However, paradoxically, other studies reveal situations in which animals learn less effectively when observing skilled demonstrators than when watching unskilled demonstrators or when there was no demonstrator present (Robertson et al. 1985; Giraldeau & Lefebvre 1986, 1987; Biederman & Vanayan 1988; Stammbach 1988; Beauchamp & Kacelnik 1991; Templeton 1998). Thus, there is a suggestion that in some cases relying on an experienced animal’s knowledge may prevent an individual from learning a task itself where in other circumstances a task may be learnt more effectively in the presence of a demonstrator. These findings have been most frequently described for discrimination tasks or for experiments in which an animal had to learn how to perform a novel action to receive a reward. A similar Correspondence and present address: T. Burt de Perera, Instituto de Ecologia, Universidade Nacional Autonoma de Mexico, Apartado postal 70-275, Cayoacan, 04510 Mexico D.F., Mexico. T. Guilford is at the Animal Behaviour Research Group, Department of Zoology, University of Oxford, Oxford OX1 3PS, U.K. 0003–3472/99/030715+05 $30.00/0
1999 The Association for the Study of Animal Behaviour
715
1999 The Association for the Study of Animal Behaviour
716 ANIMAL BEHAVIOUR, 57, 3
1
2
3
4
5
6
7
8
were gradually pushed across and the number of cups containing food was reduced to one. The position of this goal varied daily. Pretraining ended (after approximately 3 weeks) when all the experimental birds could flip off a lid from a fully covered cup.
9 10 11 12 A
7
1
5
B C D E
2
8
3
Goal positions
Experimental Training and Testing
F G H I
4
9
6
J K L Curtain
Figure 1. Representation of the arena. The numbers on the cups indicate the order in which the goals were presented. Note that these numbers indicating order correspond to those in Fig. 2.
METHODS The subjects were 15 1–2-year-old homing pigeons, both males and females, reared at the Oxford University Field Station, Wytham, U.K. These birds were kept at 90% of their free-feeding normal body weight, monitored by daily weighing, to increase their motivation to find food. To regulate their food during the times when they were not in an experiment, we housed the pigeons in a loft separate to the others in the colony. The loft measured 55 m and was outdoors. The birds were fed once a day on corn and mixed seed and received water ad libitum. We performed the study within a laboratory. A 1212 grid of squares was drawn on the lino floor (55 m) and nine ceramic feeding cups 8 cm in diameter with cardboard lids were positioned in this grid (Fig. 1). A curtain with a small viewing hole was hung at the entrance to the laboratory from one side to the other from 1 m above the ground to the floor. Pigeons entered the arena by being placed under this curtain. We kept the pigeons that were to be used in the experiment or training stage in holding boxes behind the screen until their release into the arena. The laboratory was lit overhead by fluorescent lights and the floor was cleaned between experimental trials.
Pretraining We initially trained the 15 pigeons individually to flip off the lids to receive a food reward of corn and mixed seed. This task was thought to be fairly natural as pigeons routinely search for hidden seed by using their beaks (Goodwin 1983). Initially all the cups contained food and a gap was left between the lid and cup enabling the pigeon to see the food. Switching the room light off before release and on again after a 5-s interval stopped the pigeons from flying on introduction to the test area. This technique was also used in the experimental tests. After the bird had fed for 1 min, the room lights were switched off and the bird was recaptured with the aid of a torch. We repeated this procedure three times a day. The lids
We randomly assigned the pigeons to two groups. Five birds were classified as ‘experienced’. These were trained to find food in one goal location per trial, and after training ended a criterion of five consecutive correct choices had been reached by each experienced pigeon. A correct choice was defined by a pigeon flipping the lid of the goal cup before any other. We introduced the birds into the arena four times per day and 4 days were needed for every bird to reach this criterion. The other 10 pigeons were grouped as ‘inexperienced’ and received no extra training. Note that all the birds were pretrained to flip lids; experience or inexperience relates to whether the birds were permitted to learn the position of the food goal. The experiment was divided into nine trials. The goal was in a different position in each trial; thus over the entire experiment each cup on the grid was defined as a goal once. Each trial was split into two stages.
Stage 1: training We initially allocated the target birds at random to one of two groups: (1) five pairs each comprising one experienced and one inexperienced bird; (2) five inexperienced, unpaired individuals. The target birds were those that were inexperienced with respect to the position of the goal. In this testing stage either the pair of experienced and target inexperienced birds were released into the arena together or the target inexperienced bird was placed individually into the arena. In the first case the experienced bird generally ran directly to the food goal, flipped off the lid and began eating. The inexperienced bird followed and also ate some of the food that had been produced by the demonstrator. In the other case the inexperienced bird had to find the goal by trial and error by searching a number of cups and flipping off their lids itself. After finding the goal, each bird was given 30 s to feed before recapture during which time the food was never depleted. After a 2.5-min retention period during which the birds were held in carrying boxes, stage 2 began.
Stage 2: testing For the testing stage we placed the food in the same goal as for training and replaced the lids. The target birds were again released and we recorded (1) the trajectories walked (calculated by counting the number of lines on the grid that the pigeon crossed before discovering the food; (2) the number of incorrect choices of lids flipped off; and (3) the time to find the goal. We compared birds that were unpaired and those that were paired in stage 1. We then trained the experienced pigeons to a new goal position for the next trial and the inexperienced birds
BURT DE PERERA & GUILFORD: SPATIAL LEARNING 717
were subjected to the other treatment. Each inexperienced bird was allocated to the paired and unpaired treatments approximately an equal number of times over the experiment, four times in one, five in the other.
125
Mean number of squares crossed
(a)
100
RESULTS 75
50
25
0
1
2
3
4
5
6
7
8
9
2
3
4
5
6
7
8
9
500 (b)
Mean latency (s)
400
300
200
100
0
1
5
Mean number of incorrect choices
(c)
4
3
2
1
0
1
2
3
4 5 6 Goal position
7
8
9
For each of the three variables tested in stage 2, we calculated means for all birds in each of the nine trials (and hence the nine cup positions) when the target bird was either paired with a companion or unpaired in stage 1 (Fig. 2). We analysed the data using two-tailed Wilcoxon signed-ranks tests. To test whether the birds did learn the goal location in stage 1, we compared the number of lids flipped by the target birds, the number of lines on the grids crossed, and the latency to find the goal in stages 1 and 2. Note that it was possible to perform this test only when the birds were placed individually into the arena during stage 1. In the paired condition the demonstrator flipped the lids, so we could not compare lids flipped by the target birds in both stages. Wilcoxon signed-ranks tests revealed that in stage 2, fewer lids were flipped over in all nine trials (Z= 2.668, N=9, P=0.008) and fewer squares were crossed (Z= 2.201, N=6, P=0.028; note that we only began to record trajectories from trial 4), but latencies to find the goal did not differ significantly (Z= 2.073, N=9, P=0.382). These results indicate that the pigeons had learnt the position of the food goal in stage 1. Having discovered that learning did occur in stage 1, we were able to analyse the effects of demonstrator presence in stage 1 on the efficiency of the target birds in finding the food goal in stage 2. Wilcoxon signed-ranks tests revealed that there were no significant differences between the paired and unpaired treatments in stage 2 in the mean latency to find the goal over all nine trials (Z= 1.362, N=9, P=0.173; Fig. 2). This was also the case for the median latency (Z= 0.252, N=9, P=0.799). However, the mean number of squares crossed (Z= 2.666, N=6, P=0.008) and the mean number of incorrect choices (Z= 2.200, N=9, P=0.023) were significantly higher for paired birds (Fig. 2). Thus, the presence of an experienced demonstrator hindered a pigeon learning the position of a food goal. Note that there are missing data points in this experiment. In some cases birds flew into the air knocking off the lids so the data point was no longer valid (two cases). At other times the demonstrator bird did not appear motivated to find the food and the target bird discovered the food first by trial and error (four cases). Occasionally the target birds would also lack motivation to find the goal. The test was abandoned if a pigeon had not found the goal in 45 min (six cases). There was no apparent
Figure 2. (a) Mean+SE number of squares crossed in the arena (see Fig. 1), (b) mean+SE latency to find the goal, and (c) mean+SE number of incorrect choices for all 10 target birds in stage 2 over nine trials. In stage 1 of the experiment, target birds were released in the arena either with a bird who knew the location of the goal () or alone ( ). In stage 2 all birds were alone.
718 ANIMAL BEHAVIOUR, 57, 3
pattern in the occurrence of these missing data points, and as we used mean values per bird in the analysis the results were not affected. Compared with paired birds in stage 1, unpaired birds in stage 1 took longer to locate food (Wilcoxon signedranks test: Z= 2.497, N=10, P=0.012), covered more of the arena (Z=2.201, N=6, P=0.03), and visited more unrewarded cups (Z= 2.599, N=10, P=0.01), so one possibility is that indirect effects (such as the possibility that paired birds were in the arena for less time because their partner found the food quickly) are the cause of less effective learning in stage 2. However, when each unpaired bird’s mean performance in stage 1 was regressed against that in stage 2 we found no evidence that variation in stage 1 performance affected stage 2 performance with respect to time to locate food (R2 =0.037, N=10, P=0.595), number of squares crossed (R2 =0.079, N=10, P=0.220), or number of cups visited (R2 =0.065, N=10, P=0.476). As the latency to discover the goal in stage 2 was not significantly different between the two treatments we also checked that a bird that spent more time in the arena in stage 1 did not cross more squares in stage 2, but this was not the case (R2 =0.010, N=10, P=0.947). DISCUSSION On the basis of the trajectory data (the number of squares crossed) and the number of incorrect choices of lids flipped, we can conclude that pigeons learn the position of a food goal more effectively when performing the task alone than when accompanied by a knowledgeable companion. We supposed that there is an advantage to a pigeon in learning the position of a food goal in this experiment, particularly as the birds were kept below their free-feeding weight. It is initially surprising that pigeons that were inexperienced with respect to the position of the food goal did not learn from a demonstrator as effectively as they learnt the task individually. However, other studies have reported similar results. For example, in a recent study Templeton (1998) showed that observer starlings, Sturnus vulgaris, learned a discrimination task more effectively when they watched demonstrators that made incorrect responses than those that watched correct responses. This suggests that starlings do not learn such discrimination tasks from knowledgeable companions. Our results were consistent with others that report this seemingly paradoxical finding (Robertson et al. 1985; Giraldeau & Lefebvre 1986, 1987; Biederman & Vanayan 1988; Stammbach 1988; Beauchamp & Kacelnik 1991). Two general hypotheses may explain our results. First, the pigeon may learn nothing in stage 1 when placed with a knowledgeable bird (the experienced bird may inhibit learning). Second, the inexperienced pigeon may learn something different to the experimental task when placed with the experienced bird, which may be useful for obtaining food in the presence of the demonstrator but which does not help a bird locate the goal on its own. The observers would not, therefore, attend to features of the environment in the arena and relate these features to
the position of the food. Alternatively, learning may be overshadowed by a demonstrator. The overshadowing theory supposes that either the approach of the demonstrator to the goal or the demonstrator itself acted as a reliable indicator of food in stage 1. Thus, the demonstrator may have in some way overshadowed signals needed to find the food goal in the arena in stage 2. The pigeon is a social species which usually forages in a flock. Therefore, it may be an advantage for individuals to follow other more knowledgeable birds and not necessarily to learn the task themselves. This scrounging strategy (Giraldeau & Lefebvre 1986) may be a more efficient behaviour for some members of the flock than finding the food themselves. We call an example by which the overshadowing hypothesis may operate the passenger/ driver effect. This assumes that the observer does not attend to the position of the goal because it is simply following the demonstrator in the same way, it is said, that a passenger of a car will often be less able to describe a route taken than a driver. This is, however, only one hypothesis that may operate within the overshadowing theory. For example, another possibility is that the bird per se may have an effect. As this experiment was not designed to investigate why pigeons did not learn from companions, this question is open for further study. The mechanisms we have suggested are perhaps the most interesting from a social learning perspective, but other lower level effects are also possible and, although strictly still the results of the demonstrator’s presence in stage 1, would be useful to eliminate in future experiments. First, we have already suggested (and inferred unlikely) that unpaired birds may gain more experience than paired birds in stage 1 because they spend longer searching, cover more ground, or visit more unrewarded cups, and that this enhances performance in stage 2. Second, the presence of the conspecific itself at the baited bowl might have reduced the strength of the food bowl as an unconditioned stimulus in stage 1 either directly because of lowered food intake (even though the bowl was nondepleting), or indirectly through social stresses. Since the pigeons habitually feed at crowded food bowls, and were not seen to exclude each other in this experiment, this final mechanism, although possible, seems unlikely. The protocol in this experiment did not allow us to test learning by a target bird paired with an inexperienced bird in stage 1 as it would not have been possible to ensure that the experience of the target bird was consistent. The target bird might have discovered the food itself or the inexperienced partner could have found it. However, such a test would allow some of the effects described above to be eliminated, and should be considered in future experiments of this kind. In conclusion, our results suggest that pigeons learn a spatial, food-finding task more effectively when performing the task alone than when accompanied by a knowledgeable conspecific. Reasons for this include the possibility that learning is inhibited by a demonstrator, perhaps because the observers develop a scrounging strategy in the presence of a knowledgeable bird. Alternatively, the demonstrator (in some way) may have
BURT DE PERERA & GUILFORD: SPATIAL LEARNING 719
overshadowed signals needed to find the food goal in the arena in stage 2. To our knowledge this is the first attempt to investigate the ability of homing pigeons to learn a spatial task from companions. The results suggest that orientational information may not be learnt from others and instead pigeons learn more effectively when performing a task by trial and error. This finding has an important significance for homing pigeons which are often faced with orientational problems. Acknowledgments We thank Jackie Chappell and two anonymous referees for helpful comments on the manuscript and BBSRC for funding the project. References Beauchamp, G. & Kacelnik, A. 1991. Effects of the knowledge of partners on learning rates in zebra finches, Taeniopygia guttata. Animal Behaviour, 41, 247–253. Biederman, G. B. & Vanayan, M. 1988. Observational learning in pigeons: the function of quality of observed performance in simultaneous discrimination. Learning and Motivation, 19, 31–43. Dawson, B. V. & Foss, B. M. 1965. Observational learning in budgerigars. Animal Behaviour, 13, 470–474.
Giraldeau, L. A. & Lefebvre, L. 1986. Exchangeable producer and scrounger roles in a captive flock of feral pigeons: a case for the skill pool effect. Animal Behaviour, 34, 797–803. Giraldeau, L. A. & Lefebvre, L. 1987. Scrounging prevents cultural transmission of food-finding behaviour in pigeons. Animal Behaviour, 35, 387–394. Goodwin, D. 1983. Pigeons and Doves of the World. 3rd edn. Ithaca, New York: Cornell University Press. McQuoid, K. M. & Galef, G. G. 1993. Social stimuli influencing feeding behaviour of Burmese fowl: a video analysis. Animal Behaviour, 46, 13–22. Nicol, C. J. & Pope, S. J. 1992. Effects of social learning on the acquisition of discriminatory keypecking in hens. Bulletin of the Psychonometric Society, 30, 293–296. Palameta, B. & Lefebvre, L. 1985. The social transmission of a food-finding technique in pigeons. What is learned? Animal Behaviour, 33, 892–896. Robertson, H. A., Vanayan, M. & Biederman, G. B. 1985. Observational learning and the role of confinement in pigeons: suppression of learning as a function of observing the performance of a conspecific. Journal of General Psychology, 112, 375– 382. Stammbach, E. 1988. Group responses to specifically skilled individuals in a Macaca fascicularis group. Behaviour, 107, 241– 266. Templeton, J. 1998. Learning from others’ mistakes: a paradox revisited. Animal Behaviour, 55, 79–85.