Observational visuospatial encoding of the cache locations of others by western scrub-jays (Aphelocoma californica)
- First Online:
- Cite this article as:
- Watanabe, S. & Clayton, N.S. J Ethol (2007) 25: 271. doi:10.1007/s10164-006-0023-y
- 146 Views
Western scrub-jays (Aphelocoma californica) hide food and rely on spatial memory to recover their caches at a later date. They also rely on observational spatial memory to steal caches made by other individuals. Successful pilfering may require an understanding of allocentric space because the observer will often be in a different position from the demonstrator when the caching event occurs. We compared cache recovery accuracy of pairs of observers that watched a demonstrator cache food. The pattern of recovery searches showed that observers were more accurate when they had observed the caching event from the same viewing direction as the demonstrator than when they had watched from the opposite direction. Search accuracy was not affected by whether or not the tray-specific local cues provided left–right landmark information (i.e. heterogeneous vs. homogeneous local cues), or whether or not the caching tray location was rotated. Taken together, these results suggest that observers have excellent spatial memory and that they have little difficulty with mental rotation.
KeywordsObservational learningMimicVisual perspective takingMental rotation
Impressive feats of spatial learning and memory have been observed for food-storing birds such as Clark’s nutcracker (Nucifraga columbiana) (Vander Wall 1982). Field and laboratory studies have both demonstrated that food-caching species encode the spatial location of their caches during a single, brief visit when the item is hidden and that they then rely, at least in part, on this memory when recovering their caches (the subject has been reviewed by Shettleworth 1995). Bednekoff and Balda (1997) found that Clark’s nutcrackers directed 82% of their “bill plunges” at cache sites during recovery trials. When these birds were presented with a 2×3 array of potential caches sites, only one of which contained a food cache, they made an average of less than one “error” (visit to a site that did not contain a food cache) before successful recovery. Western scrub-jays have a strong preference for spatial cues for cache recovery, although this strategy could be modified to use local pattern cues by training (Watanabe 2005). Some food-storing species accurately retain the spatial locations of several cache sites over long time periods (Hitchcock and Sherry 1990; Healy and Suhonen 1996; Brodin and Kunz 1997; Balda and Kamil 1992).
One advantage of food caching is that individuals can hide food when it is readily available, and recover the caches when supplies are less abundant. There is, however, a risk to this strategy because the caches are susceptible to pilfering by others. This pilfering or “kleptoparasitism” of food caches occurs with many avian species, between heterospecifics (e.g. Burnell and Tomback 1985; Gibb 1960) and between conspecifics (Brockmann and Barnard 1979; Heinrich and Pepper 1998; Ekman et al. 1995; Brodin and Ekman 1994). For many species pilfering of caches need not involve spatial memory, because the cache-raider might search randomly and steal those it encounters by chance, or it might steal the food immediately after storage. The second method of pilfering may incur a cost, however, because of the risk of aggression and this might therefore be a strategy only the most dominant birds use (Gibb 1960). So a more efficient method of pilfering might be to rely on observational learning of where the storer has cached and then wait until the storer has left the scene before raiding the caches. This latter strategy has been demonstrated for several food-caching corvids, including ravens (Heinrich and Pepper 1998; Heinrich 1999; Bugnyar and Kotrcshal 2001) and a variety of species of jay. In laboratory tests of cache recovery accuracy, Pinyon jays, Gymnorhinus cyanocephalus (Bednekoff and Balda 1996a), Mexican jays, Aphelocoma ultramarina (Bednekoff and Balda 1996b), and western scrub-jays (Clayton et al. 2001), perform extremely well when searching for caches they saw a conspecific make, suggesting that these birds also have excellent observational spatial memories. In the western scrub-jay study the observers and storers were housed individually in cages placed back to back, so the observers watched the caching event from the opposite perspective to that of the storer, and storer’s caching tray was placed in the equivalent place in the observer’s home cage during cache recovery. Despite this, the observers were almost as accurate as the storers’ in subsequent tests of cache recovery accuracy, showing the birds use allocentric spatial cues to remember cache sites, irrespective of body position at the time of encoding. Finding of food hidden by a demonstrator has been observed for mammals also. Mangabeys also had a similar behaviour (Coussi-Korbl 1994). Kangaroo rats also pilfer (Preston and Jacobs 2005). Pilfering is, therefore, observed in divergent species. Dominant pigs follow subordinates who have information about a place of hidden food and displace them when they reach the food (Held et al. 2000).
In this experiment cache recovery accuracy has been compared for observers that watched the caching event from either the same or opposite perspective as the demonstrator. Thus, one observer saw where the demonstrator cached from the same perspective as the demonstrator whereas the other observer watched from the opposite perspective. When caching was complete, each bird was tested individually for its ability to relocate the caches the demonstrator had made. A variety of tests were conducted to establish whether the observers’ accuracy depended on whether they shared the same viewpoint as the demonstrator, to assess the effect of rotating the position of the tray through an angle of 90° or 180°, and to establish whether the task was easier if the local tray-specific cues surrounding the tray were heterogeneous (i.e. as opposed to homogeneous) so that the birds could use the relative position of these topographical and visuo-spatially distinct Lego Duplo structures to guide their recovery behaviour.
Materials and methods
Five male and four female adult western scrub-jays (Aphelocoma californica) were used in this study. Unlike the jays in the previous studies by Clayton and Dickinson (1998, 1999), these birds had not received extensive experience in tests of cache-recovery accuracy. One female was used as a demonstrator only, the rest of the birds were both demonstrators and observers and thus data were obtained from eight observers (three females and five males). Each bird was housed individually in a wire mesh cage (91×91×76 cm3). The maintenance diet consisted of a variety of foods including powdered IAMS dog food and peanuts, a mixed seed mix, shelled sunflower seeds, pine nuts, mealworms, boiled egg, and grapes. The food and water were provided ad libitum when the birds were not being tested. During caching and recovery trials the maintenance diet was removed at 17:30 on the previous day and returned to the bird’s home cage when the recovery trial was complete. Caching and recovery trials were conducted between 10:30 and 16:00, and during these trials the birds could consume wax worms, their most preferred food. The lights went off at 18:00 and came on again at 8:00 the following morning.
Each triad consisted of one demonstrator and two observers. The subjects were naive in the first three groups, but two of the demonstrators and one of the observers in these three triads subsequently became the observers and demonstrator in the fourth group to maximise the number of observers in the study.
A caching tray and a bowl filled with 40 wax worms, 10 peanuts and 10 husked sunflower seeds were placed in the central compartment of the experimental chamber and a demonstrator was introduced. Half of the surface of the tray (left or right) was covered by a transparent plastic strip and secured by bulldog clips, so that the demonstrator could not cache in the covered area. As the experimenter could vary the location of the plastic strip from trial to trial, this treatment prevented the demonstrator from repeatedly using the same cache sites throughout the experiment. The tray was placed on the left side or right side of the central compartment facing the partition, and the side in which the caching tray was placed was counterbalanced across triads. When the tray was placed on the left side, for example, an observer in the right compartment saw the tray from the same visual direction as the demonstrator, while an observer in the left chamber saw it from the opposite direction.
In two habituation sessions, the birds were allowed to eat food in the central compartment individually. They easily adapted to eat in the experimental chamber and did not show any sign of distress. During training and testing the two observers were introduced into their respective side compartments first, and then the demonstrator was placed in the central compartment and the caching trial began immediately. The demonstrator was allowed to eat and cache for 15 min, and was then removed from the chamber. The number of items cached and their location in the tray were recorded.
When the demonstrator had been removed from the room, each of the observers was allowed into the central compartment and given the opportunity to search in the caching tray and recover the demonstrator’s caches. To encourage the observers to enter the central compartment and search in the tray, some worms were placed on the floor of the chamber. During training trials, observers and demonstrators were allowed to recover the food from the cache sites in which the demonstrator had cached. On test trials, the demonstrator and the two observers within a triad were allowed to search for caches but the demonstrator’s caches were removed before the recovery sessions because some of the worms were alive and appeared on the kibbles. Fresh corn kibble was placed in the caching trays to eliminate possible traces of caching. During each recovery session one of the transparent partitions was exchanged for an opaque white partition, so the observer not being currently tested could not see inside the central compartment. The order in which the two observers were tested was counterbalanced between triads and across trials. The partition on the other side of the central compartment was opened and an observer was allowed into the central compartment to search in the caching tray. The recovery session ended after 5 min or when the observer found all the demonstrator’s caches (training trials), or searched in all the sites in which the demonstrator had cached (test trials), whichever occurred first. The observer was returned to its home cage and the tray was restocked with fresh nuts and worms in the original locations in which the demonstrator had cached them. The other observer was then allowed into the central compartment to search the tray under the same conditions. When its recovery session was complete the second observer was returned to its home cage and the caching tray was restocked with nuts and worms as before. Finally, the demonstrator was returned to the central compartment and allowed to recover its caches. This occurred about 15 min after the original caching episode. Thus the demonstrator recovered its caches in private, while both of the observers were back in their respective home cages.
Training and test trials
The fact that each pair of observers saw the same demonstrator from different perspectives enabled us to counterbalance the order in which birds watched from the same and different perspective as the demonstrator in training and test trials. Thus half the birds had the same perspective as the demonstrator first and then observed from the opposite perspective whereas for the others the order was reversed. We also varied whether or not the tray contained a heterogeneous or asymmetric arrangement of Lego Duplo blocks or whether the tray-specific Lego Duplo cues surrounding the caching tray were arranged homogeneously or symmetrically. Blocks of the same colour and height were used in the homogeneous or symmetrical arrangement. There was no local landmark within a tray. A symmetrical arrangement makes the task harder, because the bird cannot rely on local spatial cues about differences in the topographic arrangement of Lego Duplo blocks on the left and right sides of the tray. Finally, in some test trials the location of the caching tray was rotated through 90° or 180° as shown in Fig. 1 to test the accuracy of observers’ use of the topographical arrangement of the tray-specific cues in the absence of distal position cues about the absolute location of the caches. Note that the birds did not receive any training (i.e. rewarded) trials in which the caching tray was rotated to a new position.
Asymmetry training and rotation tests
During the initial training block (asymmetry training) which consisted of four caching and recovery sessions, the location of the tray was not altered between the caching and recovery periods, and all the caching trays contained an heterogeneous arrangement of Lego Duplo. The birds were then subjected to a test trial (AS-180 test) that differed from training trials in that no food was present at recovery, enabling us to assess the search patterns of all birds in the absence of any cues emanating directly from the food. AS-180 test also differed from the training in one other respect—the caching tray was rotated through an angle of 180° and placed at the opposite side of the central compartment (Fig. 1).
After the test trial, the birds received a second three sessions of the training in which they cached and recovered from a new set of caching trays, again with an heterogeneous arrangement of Lego Duplo blocks, but this time the observers were swapped around so that those that had watched the demonstrator from the same perspective now observed from the opposite perspective and vice versa. The birds were then subjected to the AS-180 test again.
Immediately after this test the birds received the asymmetry training again, in which each observer had the opportunity to observe the tray from the same and opposite perspective to that of the demonstrator, then received AS-90 test, in which the tray was rotated through an angle of 90° and placed at the far end of the wall in the central chamber during the recovery period. Thus, the location of the tray was rotated 90° anti-clockwise for the observer in the left compartment and 90° clockwise for the observer in the right compartment of each triad. Then, the positions of the observers were swapped and received the asymmetry training and AS-90 test again.
Symmetry training and rotation tests
The birds then received another training trial (symmetry training) but this time the observers watched a demonstrator cache in a new caching tray which contained homogeneous or symmetric arrangement of Lego Duplo blocks. They were then allowed the opportunity to recover the demonstrator’s caches when the tray was placed in its original position. Homogeneous arrangement makes the task harder because the bird cannot rely on local spatial cues about differences in the topographic arrangement of Lego Duplo blocks on the left and right sides of the tray. After the training, the birds received another test trial (SY-180 test) in which the observers again watched a demonstrator cache in heterogeneous caching tray, but during recovery the location of the tray was rotated through an angle of 180° and placed at the opposite side of the cage. The birds then received a further symmetry training which was the same as the first training trial except that the birds now cached in a new tray with a different homogeneous arrangement of Lego Duplo blocks. This training trial was followed by another SY-180, also using a homogeneous arrangement of local tray-specific cues, and with the location of the caching tray rotated through 180° so that it was placed on the side of the cage opposite to that used before.
Novel tray test
Finally, in the NAS-180 test, a novel tray was introduced for each test and the location of the tray was rotated 180° as before. The NAS-180 test was repeated with swapped position of the observers. The objective of the final pair of tests was to establish whether or not the observers could generalise about the effects of tray rotation in the absence of specific training trials with a particular caching tray.
To establish that the observers were using observational spatial memory in these tasks we compared their performance in these tasks with a control trial in which the demonstrator cached in one of the heterogeneous caching trays as normal but the observers could not see where the demonstrator had cached. To do this the translucent partitions were replaced by opaque white screens so that the observers could hear the sound of the demonstrator caching but they could not see where the bird cached. All other procedures were identical with those described for other tests.
We compared the observers’ recovery behaviour with that expected if the observers were searching at random. Three indices were used in our analyses. First, we recorded which side of the tray the observers searched first. In any given trial, one side of the caching tray was covered by a transparent plastic strip so that the demonstrators could only cache in either the left or right half of the tray. If the observers remembered which side of the tray the demonstrator had cached in they should preferentially visit the correct side of the tray first compared with a random expectation that only a mean of 50% of the birds would visit the correct side in any given trial.
To establish how accurate their recovery performance was, two additional measures were taken, one based on how efficient a bird was at recovering the first cache and the other based on searching strategy throughout the recovery trial. The advantage of using these two measures rather than simply the proportion of items recovered is that we can directly compare training trials with test trials in which no food is present at recovery. To assess the former measure, we recorded the number of sites visited to find the first cache (or, on test trials, the first site in which an item had been hidden during the previous caching phase) and compared this value with that expected if the birds were searching at random. Previous observation showed that some birds had position bias of caching but that the bias was not consistent across days. Thus, the random search is most probable null hypothesis. Random expectation was calculated as q/p, where p is the number of sites in which the demonstrator had cached divided by 16, the total number of available cache sites, and q=(1−p). We used G square test (df=7) for this analysis.
Finally, we compared the accuracy of the observers’ searching preferences by calculating the number of cache sites visited as a proportion of the cache sites. If the demonstrator cached at two sites and the observer visited these two sites, the accuracy is 1.00. This search accuracy was compared with that expected if the birds were searching randomly, where random expectation was calculated by dividing the number of cached sites by the total number of possible cache sites (16). The expectation is 2/16 for the example above.
In summary, the results showed that western scrub-jays readily observe the caching behaviour of conspecifics and can rely on observational learning to remember where the demonstrator had cached. In control blind trials in which the observers could not see where the demonstrators cached performance plummeted to chance levels. Interestingly neither use of local cues that provided left–right landmark information (i.e. the asymmetry trays as opposed to the symmetry trays), or rotation of the caching tray location through 90°, altered the searching accuracy of the observers. The only rotation that significantly impaired the observers’ performance in terms of proportion of the side choice (Fig. 5) was when the tray location was rotated 180° during the first test (ASY-180). Interestingly, however, there was no evidence of impairment even with 180° rotation of the tray location on the final pair of test trials (NAS-180) suggesting the birds could learn to solve tasks with the 180° rotation with practice. These results suggest that observers have excellent spatial memory abilities and have little difficulty with mental rotation.
The experiments focussed on the ability of observers to recover the caches made by other birds (demonstrators), capitalising on the fact that the birds could use local tray-specific cues to find the food to test the effects of viewing direction on cache recovery accuracy. The results suggest that the subjects have highly accurate observational memory even after rotation of the tray.
Much of the work on mental rotation comes from studies on human subjects, and particularly how these abilities develop in children. One of the first studies to investigate mental rotation in humans was conducted by Shepard and Metzler (1971). The subjects were shown 3D drawings of different objects and then asked whether sketches of 3D-rotated shapes were the same as, or different from, a particular object. The decision time of the subjects increased linearly with the angular disparity (degree of rotation) of the shapes. It has been suggested that the reason for the linear correlation between decision time and degree of rotation of the shape is because subjects mentally rotate one visual pattern into congruence with another.
The literature on comparative cognition contains surprisingly few studies of mental rotation in animals even though it is relatively easy to test in the laboratory using a delayed-matching-to-sample (DMTS) procedure in which the choice samples show one stimulus rotated in different orientations. Work on mental rotation in the dolphin (Tursiops truncatus) by Herman et al. (1993) and in sea lions (Zalophus californianus) by Mauck and Dehnhart (1997) suggested a human-like linear relationship between reaction time and angular rotation. Other studies on baboons (Pan panio) did not, however, reveal a significant correlation between angular rotation and reaction time (Hopkins et al. 1993). Interestingly, for pigeons (Columbia livia) there was no increase in reaction time depending on degree of rotation (Hollard and Delius 1982). The pigeons could easily discriminate 3D rotated objects when these objects were familiar to them (Watanabe 1997, 1999). This result suggests that some species of bird, for example pigeons and scrub-jays, might be particularly good at mental rotation. These experiments tested response to rotated stimuli in the same 2D place; the current experiment rotated the objects to a different 3D place. The results demonstrated 3D mental rotation ability in semi-natural setting. Perhaps this is not surprising given the 3D life of birds and this might have provided a strong selective pressure for improved mental rotation ability.
Other human experimenters (Huttenlocher and Pressen 1973, 1979; Pressen 1982) have directly compared performance on these mental rotation tasks with visual perspective taking tasks such as Piaget and Inhelder’s (1956) three mountain task in which the subjects are asked to select a picture showing another observer’s view. Baron-Cohen (1988) has argued that visual perspective taking can be performed “using a strategy of mental rotation on primary representations” (p. 394), thus relying on the ability to mentally rotate a 3D representation of an object array so it can be imagined from different perspectives to the way one actually views the object array at a given point in time. Others (Whiten and Perner 1991) have argued that visual perspective taking requires an understanding that another person’s point of view may be different from their own.
The observers’ behaviour can be explained in terms of allocentric spatial discrimination and mental rotation. Their behaviour can be explained in terms of visual perspective taking also. Because the spatial discrimination in this task was to locate the place of caching of the demonstrator, the observer birds may show visual perspective taking skills of understanding that another bird’s view of the caching event is different from their own. For visual perspective taking skills, the observers need combination of two abilities, namely discrimination of the behaviour of the conspecific and mental rotation. The observers’ view of the caching tray differs between caching and recovery and efficient cache recovery depends on their ability to mentally rotate the position of the caching tray at the time of recovery. The ability is particularly important after rotation of the tray.
These experiments were carried out in the department of experimental psychology, Cambridge University. The research was supported by Exchange Program of Keio University and Downing College, and The 21 Century COE Program (D-1) in Japan.