- Email: [email protected]
Wild, free-living hummingbirds can learn what happened, where and in which context
Animal Behaviour 89 (2014) 185e189
Contents lists available at ScienceDirect
Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav
Wild, free-living hummingbirds can learn what happened, where and in which context Sarah A. Jelbert a, b, T. Andrew Hurly c, Rachael E. S. Marshall a, Susan D. Healy a, d, * a
School of Psychology, University of St Andrews, St Andrews, U.K. School of Psychology, University of Auckland, Auckland, New Zealand c Department of Biological Sciences, University of Lethbridge, Lethbridge, Alberta, Canada d School of Biology, University of St Andrews, St Andrews, U.K. b
a r t i c l e i n f o Article history: Received 1 May 2013 Initial acceptance 30 May 2013 Final acceptance 11 December 2013 Available online 7 February 2014 MS. number: 13-00369R Keywords: hummingbird scene memory sequence timing what-where-which
Studies in the laboratory have shown that animals can combine multiple kinds of information to form integrated memories for rules and events. Less is known about how animals make use of these integrated memories in the wild. Here we tested whether wild, free-living, rufous hummingbirds, Selasphorus rufus, could learn to identify rewarded flowers in a naturalistic foraging situation, by remembering, over multiple exposures, what flower was rewarded, where and in which context. Birds were presented with boards on which four artificial flowers were mounted, one containing a food reward, the others containing water. Which flower (its colour and location) contained a reward was indicated in one condition by the presence of visually distinctive background boards and in a second condition by the sequential order in which the boards were presented. In both conditions, birds combined these pieces of information and learned to use the context to determine which of the four flowers was rewarded. Although they were not required to do so here, it is possible that these birds might be able to combine pieces of information to form integrated memories for single events. Ó 2014 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
It is well established that animals are capable of forming memories involving the integration of different types of information (Pearce, 1994; Pickens & Holland, 2004). Much of this research has addressed the neural underpinnings of these memories (see Rudy, 2009 for a review), typically, in the laboratory, training and testing animals with procedures including operant conditioning (Iordanova, Burnett, Aggleton, Good, & Honey, 2009; Iordanova, Good, & Honey, 2008; Penick & Solomon, 1991), object recognition paradigms (Mumby, Gaskin, Glenn, Schramek, & Lehmann, 2002; Norman & Eacott, 2005) and computer-based memory tasks (Browning, Easton, Buckley, & Gaffan, 2005; Eacott & Gaffan, 2005). Less is known, however, about how animals might make use of integrated memories in their daily lives (Clayton, Griffiths, Emery, & Dickinson, 2001; Clayton, Yu, & Dickinson, 2001). In this study, we investigated whether wild, free-living rufous hummingbirds, Selasphorus rufus, could form integrated memories for what was rewarded, where and in which context, in a naturalistic foraging situation. Hummingbird foraging is a plausible model for examining whether animals can combine different kinds of information into * Correspondence: S. D. Healy, School of Biology, Harold Mitchell Building, University of St Andrews, St Andrews KY14 7AU, U.K. E-mail address: [email protected] (S. D. Healy).
memories. Not only can these birds learn to avoid visiting recently emptied flowers (Healy & Hurly, 1995) and to return to flowers that contain a consistent reward (Gass & Sutherland, 1985; Healy & Hurly, 2003; Hurly, 1996), they can also remember both where and when flowers refill (Henderson, Hurly, Bateson, & Healy, 2006; Marshall, Hurly, & Healy, 2012) and associate this with variations in nectar quality (González-Gómez, Bozinovic, & Vásquez, 2011). Whether rufous hummingbirds, or any other animal foraging in the wild, can combine three pieces of information, including contextual cues, into memories is not yet known. The combination of elements into a memory is one of the prerequisites for episodic-like memory (Tulving, 1972), which involves the integration of what, where and when information (Clayton, Griffiths, et al., 2001) or what, where and which occasion information (Eacott & Easton, 2010; 2012) about a single event. If, in the current experiment, the birds could form a combined memory for these three features of episodic-like memories over repeated exposures, it would suggest that, in principle, they may also be able to form combined memories for single events. Potentially, this could provide a real-world paradigm for addressing the value of episodiclike memories. To assess whether hummingbirds can form combined memories, we based our experimental design on a scene-learning paradigm in which rhesus macaques, Macaca mulatta, learnt the
0003-3472/$38.00 Ó 2014 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.anbehav.2013.12.028
186
S. A. Jelbert et al. / Animal Behaviour 89 (2014) 185e189
location and visual features of rewarded and unrewarded symbols in a visual scene (Gaffan, 1994). In the first condition of our experiment hummingbirds were allowed to forage from sets of artificial flowers on two visually distinctive experimental boards, which were presented repeatedly, one at a time. Both boards comprised matching sets of four flowers, with two colours of flower placed in two locations on the board. A different flower, in a different location, was rewarded on each board. As the four flowers were the same on both of the boards, a bird could not identify which flower contained the reward by remembering only a flower’s colour (what) or its location (where). The reward could be found, however, if the bird formed an integrated memory of the flower’s colour, location and the context in which it was presented. In a second condition, we investigated whether hummingbirds could also form a combined memory involving a temporal context. As hummingbirds use timing cues in their daily lives (Henderson et al., 2006; Marshall, Hurly, Sturgeon, Shuker, & Healy, 2013) we hypothesized that they could integrate temporal sequential cues with other kinds of information. To test this we repeatedly presented birds with a single white board with four flowers, but we alternated the colour and position of the rewarded flower each time we presented the board. The birds had to attend to this alternating sequence in order to identify the rewarded flower. We compared the learning speed and errors the birds made in each of the two conditions to determine whether hummingbirds can form combined memories involving both visual and sequential contexts. METHODS Subjects The subjects in this experiment were eight wild, free-living, adult male rufous hummingbirds that had established their territories around commercial hummingbird feeders, containing 14% sucrose solution, set out in the spring. The experiment was conducted in a valley in the eastern Rocky Mountains, Alberta, Canada (49 210 N, 114 250 W). Rufous hummingbirds migrate to this valley to breed and males establish and defend feeding territories. The territory holder typically feeds from the commercial feeder every 10e15 min, chasing away intruding conspecifics. To enable individual identification, birds with established territories were caught, by putting a mesh cage around a feeder, and the white plumage on their chests was colour marked with nontoxic ink. The data were collected between 0800 and 2000 hours Mountain Standard Time, from May to July 2011. All of the work was carried out under permits from Environment Canada and Alberta Fish and Wildlife, with the ethical approval of the University of St Andrews and the University of Lethbridge, and was conducted in adherence with the ASAB Guidelines for the Treatment of Animals in Behavioural Research. Experimental Apparatus Prior to the experiment we trained birds to feed from the experimental apparatus, which was placed conspicuously within a bird’s territory. The apparatus consisted of a 30 35 cm piece of foam board, mounted onto a 1 m high stake, on which four artificial flowers were positioned. Two small holes were made at both the top and the bottom of each board (10 cm apart, and 3 cm from the edge of the board) into which artificial flowers were inserted (Fig. 1). The artificial flowers were formed from syringe caps inserted into the centre of coloured cardboard discs (6 cm diameter). The syringe cap created a well in the centre of the flower from which the bird could feed. The capacity of an artificial flower (600 ml) was substantially more than a hummingbird could drink in
Figure 1. One of the experimental subjects feeding from a board in the Visual Condition. Boards measured 30 35 cm. Photograph: T.A. Hurly.
a single visit. On each board only one of these four flowers contained desirable 25% sucrose solution and the remaining three flowers contained water, which the birds find distasteful. Experimental Procedure All birds took part in two conditions, a Visual-cue condition and a Sequential-cue condition, with the order counterbalanced across birds. Both conditions followed the same format, in which one rewarded flower on a board could be identified by attending simultaneously to the flower’s colour (what), the flower’s location (where) and the specific board presentation (which context). The same flowers were presented every time; however, a different flower was rewarded in each of the two contexts. We used two different colours of flower (e.g. green and pink) and presented one flower of each colour in each of two locations on the board (i.e. top or bottom, Fig. 2). Which position (left or right) that each flower colour occupied at the top and at the bottom was randomized by coin toss before each trial. Thus the experience for a typical bird would be that in the first context a pink flower at the top of the board was rewarded and in the second context a green flower at the bottom of the board was rewarded (Fig. 2). The boards were presented repeatedly until birds demonstrated that they had learnt the identity of both rewarded flowers, defined by reaching our criterion number of correct responses: choosing the rewarded flower six times in a row (three times on each board). If the criterion was not reached, an upper limit was set at 100 board presentations (50 presentations of each board), at which point the condition was terminated. Visual Condition In the Visual Condition, the which-context cue was a visually distinctive background scene, painted on to the experimental boards (Fig. 2). The order of the board presentations in the Visual Condition was pseudorandomized. In every four presentations, each board was seen twice but boards could be presented in any order within these sets (for example, 1-2-1-2 then 2-2-1-1). This ensured that the two boards were seen equally often throughout the experiment, but that the rewarded flower was not coupled with
S. A. Jelbert et al. / Animal Behaviour 89 (2014) 185e189
Visual Condition Background 1
Background 2
A
B
Sequential Condition Feeder
Presentation 1
Presentation 2 B
A
Figure 2. An example of the apparatus used in each condition. The letters A and B indicate the rewarded flowers. The exact position of the coloured flowers (left or right) in each location was varied according to a random sequence and was not tied to a specific context.
presentation order. Each bird was tested with different board pairs and different colours of flower, which were chosen to be conspicuous on both of the background designs. Sequential Condition In the Sequential Condition, which context was specified by the order in which the boards were presented. One of the coloured flowers in one location of the board was rewarded the first time a board was presented and the other colour and location were rewarded the second time a board was presented. A single white board was used for both presentations, eliminating the possibility that birds could use visual cues to identify the reward. To make the sequential cues salient, in the Sequential Condition a commercial hummingbird feeder was presented in between each pair of board presentations. The feeder was included to ‘set’ the sequence, enabling the two boards to be described as either the first presentation after the feeder or the second presentation after the feeder (the sequence was: FeederePresentation 1ePresentation 2; Fig. 2). The bird was required to visit the feeder before being given the next presentation of a board. In both conditions, the two different colours of flower were randomly presented on either the left or the right. To prevent birds from using the appearance of a specific flower as a cue to the reward, we used multiple versions of the flowers across the trials. In both conditions boards were removed from the stake immediately after each visit by the hummingbird and the next board was set up. The number of trials per day was dependent on how frequently the bird returned to feed from the apparatus and ranged from 17 to 72 trials per day (mean SE ¼ 37 0.47). The two conditions took from 3 to 6 days to complete per bird.
187
considered to be the bird’s choice. The bird was allowed to continue probing until he found the rewarded flower although these subsequent choices were not used in the analysis. First choices were initially recorded in terms of the absolute What-Where-Which properties of that choice (i.e. the flower chosen was the pink flower at the top of the board in the first context). They were then coded in terms of what aspects of that choice could be considered correct and which aspects were errors. On each board one flower was correct (rewarded) and three flowers were incorrect (unrewarded); however, the three incorrect options represented three different types of error. As illustrated in Fig. 3, we classified each incorrect choice as one of the following error types: (1) the same colour as the rewarded flower for that context, but in the wrong location (correct What-Which, wrong Where), (2) the same location as the rewarded flower for that context, but the wrong colour (correct Where-Which, wrong What) or (3) a colour and location that were rewarded together, but not in this context (correct What-Where, wrong Which). By comparing the frequency of each error type we examined whether any aspect of the task was harder to learn than any other. We also compared how birds learnt the What-Where-Which features of the rewarded flower in the Visual and Sequential conditions. RESULTS In all cases where criterion was reached, the final six correct visits were excluded from the analyses. All statistical tests were performed using PASW 18.0 (SPSS Inc., Chicago, IL, U.S.A.). Where relevant, Bonferroni corrections were applied to account for multiple tests. Criterion The birds were highly successful in both conditions. All eight birds reached criterion in the Visual Condition, making an average of 36 visits to the experimental boards (mean SE ¼ 35.88 3.32; criterion visits excluded) before they achieved this. In the Sequential Condition, seven of eight birds reached criterion. The
Correct What-Where-Which
Correct Where-Which Incorrect What
A
Data Coding The data we recorded were the flowers chosen by the birds on each visit to an experimental board and the time of each visit. Birds were considered to have ‘visited’ a board when they probed one or more of the artificial flowers. The first flower probed on a board was
Correct What-Where Incorrect Which
Correct What-Which Incorrect Where
Figure 3. A schematic of the way in which we coded flower choices.
188
S. A. Jelbert et al. / Animal Behaviour 89 (2014) 185e189
birds that reached criterion made an average of 53 visits to the boards (mean SE ¼ 53.29 8.57). Seven of the eight birds reached criterion in the Visual Condition in fewer visits than in the Sequential Condition (paired t test: t7 ¼ 3.05, P ¼ 0.02) but the birds’ performances (percentage correct) in the two conditions were not significantly correlated (Pearson correlation: r6 ¼ 0.46, P ¼ 0.25). Choices and Errors In the Visual Condition, birds were no more likely to make one type of error more or less frequently than the other errors, suggesting that all aspects of the task were equally easy (related sample Friedman’s ANOVA: c22 ¼ 0:58, P ¼ 0.76; Fig. 4a). One bird did not make any Where errors in the Visual Condition, although all of the other birds made a range of errors of each type. In the Sequential Condition, however, the birds’ errors did vary (Friedman’s ANOVA: c22 ¼ 12:19, P ¼ 0.001; Fig. 4a), as the frequency of Which errors was greater than both What and Where errors (post hoc Wilcoxon signed-ranks tests: What: z ¼ 2.37, P ¼ 0.02; Where: z ¼ 2.52, P ¼ 0.01). This difference in the frequency of errors suggests that the birds in the Sequential Condition learnt the What-Where association with relative ease but took
60
(a)
Visual Condition Sequential Condition
50
40
30
Mean percentage of first choices
20
10
0 7
Correct
What
Where
(b)
Which First 10 visits Last 10 visits
6 5 4 3 2 1 0
Correct
What
Where
Which
Figure 4. (a) The mean þ SE percentage of first choices classified as correct, or as one of three types of error, in both the Visual and Sequential conditions (criterion visits excluded). N ¼ 8. (b) The mean þ SE number of first choices classified as correct, or as one of three types of error, in birds’ first 10 visits and last 10 visits to the boards in the Sequential Condition (criterion visits excluded). N ¼ 8.
longer to learn how the rewarded flower altered according to the context. An analysis of the choices on the first 10 and last 10 sequential board presentations (across the whole experiment) supports this view: in the first 10 presentations there was no difference between the average number of What, Where and Which errors made by the birds (Friedman’s ANOVA: c22 ¼ 1:41, P ¼ 0.56), whereas in the final 10 visits birds made no Where errors, a few What errors but a large number of Which errors (Friedman’s ANOVA: c22 ¼ 14:21, P < 0.001; Fig. 4b). Which errors could have been made for one of two reasons: either the bird made choices to both rewarded flowers without attending to the context or the bird repeatedly chose one of the flowers. Four of the eight birds in the Sequential Condition preferred one of the rewarded flowers (chi-square tests: c2 ¼ 6.23e 34.84, P < 0.05), choosing one flower at least twice as often as the other. Two birds selected one flower more than 1.5 times as often as the alternative flower, but their preference was not statistically significant (c2 ¼ 1.58e3.57, P ¼ 0.28e0.08), and the remaining two birds did not show any evidence of a preference (c2 ¼ 0.07e0.16, P ¼ 1e0.79). This demonstrates that the majority of birds made Which errors in the Sequential Condition by tending to select one of the rewarded options consistently, rather than by choosing between them both, until eventually learning to alternate their choices according to a sequence. DISCUSSION In this experiment, wild, free-living rufous hummingbirds were able to combine information about a rewarded flower based on What it looked like, Where it was and in Which context it was rewarded. They could do this when the context was specified with either a visual cue or a sequential cue. Importantly, they did this while also engaging in other behaviours such as displaying to females, perching, chasing away intruders, hawking insects or feeding from natural flowers. To our knowledge, this is the first time that this ability has been demonstrated in free-living animals. We were also able to compare the birds’ learning of two different context-specifying rules, one based on visual cues and the other based on order. Although birds were able to learn both, they found it harder to identify the rewarded flower when they had to learn the sequence in which it occurred than when the context was cued visually, and performances in these two conditions were characterized by different patterns of errors. In the Visual Condition birds made What errors, Where errors and Which errors with equal frequencies, which suggests that they found all three features equally easy to remember. However, in the Sequential Condition birds made relatively more Which errors than any other kind of error. This might suggest that it was easier for them to learn to use a visual rather than a sequential context in order to distinguish between events. This comparative difficulty of learning a sequencebased rule is somewhat surprising, considering that sequential visiting of flowers is the essence of trap lining, which is the way in which at least some species of hummingbirds are supposed to forage (Feinsinger & Colwell, 1978; Gill, 1988). It is possible that sequence-based rules are simply more cognitively demanding: an individual must not only learn the rule but also remember what has gone before, whereas in the visual cue scenario, the bird need only remember the rule. Although confirming this suggestion requires further experimental investigation, there is some evidence that temporal order may be harder to learn than visual context information because the two types of memory appear to be underpinned by different brain regions (Albasser, Amin, Lin, Iordanova, & Aggleton, 2012; Iordanova et al., 2009). Overall, in this study we found that hummingbirds can combine information about what was rewarded where and in which context
S. A. Jelbert et al. / Animal Behaviour 89 (2014) 185e189
in response to stimuli presented in their natural environment, after relatively few exposures. The ability to combine different types of information in memory is one of the prerequisites for forming episodic-like memories (Clayton, Griffiths, et al., 2001; Eacott & Easton, 2010), which involve remembering the What-WhereWhen, or What-Where-Which occasion, of single events. In our task, birds learned all three of the components that make up an episodic memory, albeit over multiple exposures. Although in this experiment it is likely that the hummingbirds used semantic, rulelearning memory to remember the identities of rewarded flowers (Tulving & Markowitsch, 1998), our findings raise the possibility that hummingbirds may be able to form combined memories for single events. If so, rufous hummingbirds could provide a model for investigating the role of episodic-like memories in natural foraging situations. Acknowledgments We thank the Universities of St Andrews (S.A.J. and S.D.H.) and Lethbridge (T.A.H.), the Natural Sciences and Engineering Research Council of Canada (T.A.H.) and NERC (R.E.S.M.) for funding. We also appreciate the comments made on the manuscript by two anonymous referees. References Albasser, M. M., Amin, E., Lin, T.-C. E., Iordanova, M. D., & Aggleton, J. P. (2012). Evidence that the rat hippocampus has contrasting roles in object recognition memory and object recency memory. Behavioral Neuroscience, 126, 659e669. Browning, P. G. F., Easton, A., Buckley, M. J., & Gaffan, D. (2005). The role of prefrontal cortex in object-in-place learning in monkeys. European Journal of Neuroscience, 22, 3281e3291. Clayton, N. S., Griffiths, D. P., Emery, N. J., & Dickinson, A. (2001). Elements of episodicelike memory in animals. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 356, 1483e1491. Clayton, N. S., Yu, K. S., & Dickinson, A. (2001). Scrub jays (Aphelocoma coerulescens) form integrated memories of the multiple features of caching episodes. Journal of Experimental Psychology: Animal Behavior Processes, 27, 17e29. Eacott, M. J., & Easton, A. (2010). Episodic memory in animals: remembering which occasion. Neuropsychologia, 48, 2273e2280. Eacott, M. J., & Easton, A. (2012). Remembering the past and thinking about the future: is it really about time? Learning and Motivation, 43, 200e208. Eacott, M. J., & Gaffan, E. A. (2005). The roles of perirhinal cortex, postrhinal cortex, and the fornix in memory for objects, contexts, and events in the rat. Quarterly Journal of Experimental Psychology Section B: Comparative and Physiological Psychology, 58, 202e217.
189
Feinsinger, P., & Colwell, R. K. (1978). Community organization among neotropical nectar-feeding birds. American Zoologist, 18, 779e795. Gaffan, D. (1994). Scene-specific memory for objects: a model of episodic memory impairment in monkeys with fornix transection. Journal of Cognitive Neuroscience, 6, 305e320. Gass, C. L., & Sutherland, G. D. (1985). Specialization by territorial hummingbirds on experimentally enriched patches of flowers: energetic profitability and learning. Canadian Journal of Zoology, 63, 2125e2133. Gill, F. B. (1988). Trapline foraging by hermit hummingbirds: competition for an undefended, renewable resource. Ecology, 69, 1933e1942. González-Gómez, P. L., Bozinovic, F., & Vásquez, R. A. (2011). Elements of episodiclike memory in free-living hummingbirds, energetic consequences. Animal Behaviour, 81, 1257e1262. Healy, S. D., & Hurly, T. A. (1995). Spatial memory in rufous hummingbirds (Selasphorus rufus): a field-test. Animal Learning and Behavior, 23, 63e68. Healy, S. D., & Hurly, T. A. (2003). Cognitive ecology: foraging in hummingbirds as a model system. Advances in the Study of Behavior, 32, 325e359. Henderson, J., Hurly, T. A., Bateson, M., & Healy, S. D. (2006). Timing in free-living rufous hummingbirds, Selasphorus rufus. Current Biology, 16, 512e515. Hurly, T. A. (1996). Spatial memory in rufous hummingbirds: memory for rewarded and non-rewarded sites. Animal Behaviour, 51, 177e183. Iordanova, M. D., Burnett, D. J., Aggleton, J. P., Good, M., & Honey, R. C. (2009). The role of the hippocampus in mnemonic integration and retrieval: complementary evidence from lesion and inactivation studies. European Journal of Neuroscience, 30, 2177e2189. Iordanova, M. D., Good, M. A., & Honey, R. C. (2008). Configural learning without reinforcement: integrated memories for correlates of what, where, and when. The Quarterly Journal of Experimental Psychology, 61, 1785e1792. Marshall, R. E. S., Hurly, T. A., & Healy, S. D. (2012). Do a flower’s features help hummingbirds to learn its contents and refill rate? Animal Behaviour, 83, 1163e 1169. Marshall, R. E. S., Hurly, T. A., Sturgeon, J., Shuker, D. M., & Healy, S. D. (2013). What, where and when: deconstructing memory. Proceedings of the Royal Society B, 280. Mumby, D. G., Gaskin, S., Glenn, M. J., Schramek, T. E., & Lehmann, H. (2002). Hippocampal damage and exploratory preferences in rats: memory for objects, places, and contexts. Learning and Memory, 9, 49e57. Norman, G., & Eacott, M. J. (2005). Dissociable effects of lesions to the perirhinal cortex and the postrhinal cortex on memory for context and objects in rats. Behavioral Neuroscience, 119, 557e566. Pearce, J. M. (1994). Similarity and discrimination: a selective review and a connectionist model. Psychological Review, 101, 587e607. Penick, S., & Solomon, P. R. (1991). Hippocampus, context, and conditioning. Behavioral Neuroscience, 105, 611e617. Pickens, C. L., & Holland, P. C. (2004). Conditioning and cognition. Neuroscience and Biobehavioral Reviews, 28, 651e661. Rudy, J. W. (2009). Context representations, context functions, and the parahippocampal-hippocampal system. Learning and Memory, 16, 573e585. Tulving, E. (1972). Episodic and semantic memory. In E. Tulving, & W. Donaldson (Eds.), Organisation of memory (pp. 381e403). New York: Academic Press. Tulving, E., & Markowitsch, H. J. (1998). Episodic and declarative memory: role of the hippocampus. Hippocampus, 8, 198e204.
Contents lists available at ScienceDirect
Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav
Wild, free-living hummingbirds can learn what happened, where and in which context Sarah A. Jelbert a, b, T. Andrew Hurly c, Rachael E. S. Marshall a, Susan D. Healy a, d, * a
School of Psychology, University of St Andrews, St Andrews, U.K. School of Psychology, University of Auckland, Auckland, New Zealand c Department of Biological Sciences, University of Lethbridge, Lethbridge, Alberta, Canada d School of Biology, University of St Andrews, St Andrews, U.K. b
a r t i c l e i n f o Article history: Received 1 May 2013 Initial acceptance 30 May 2013 Final acceptance 11 December 2013 Available online 7 February 2014 MS. number: 13-00369R Keywords: hummingbird scene memory sequence timing what-where-which
Studies in the laboratory have shown that animals can combine multiple kinds of information to form integrated memories for rules and events. Less is known about how animals make use of these integrated memories in the wild. Here we tested whether wild, free-living, rufous hummingbirds, Selasphorus rufus, could learn to identify rewarded flowers in a naturalistic foraging situation, by remembering, over multiple exposures, what flower was rewarded, where and in which context. Birds were presented with boards on which four artificial flowers were mounted, one containing a food reward, the others containing water. Which flower (its colour and location) contained a reward was indicated in one condition by the presence of visually distinctive background boards and in a second condition by the sequential order in which the boards were presented. In both conditions, birds combined these pieces of information and learned to use the context to determine which of the four flowers was rewarded. Although they were not required to do so here, it is possible that these birds might be able to combine pieces of information to form integrated memories for single events. Ó 2014 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
It is well established that animals are capable of forming memories involving the integration of different types of information (Pearce, 1994; Pickens & Holland, 2004). Much of this research has addressed the neural underpinnings of these memories (see Rudy, 2009 for a review), typically, in the laboratory, training and testing animals with procedures including operant conditioning (Iordanova, Burnett, Aggleton, Good, & Honey, 2009; Iordanova, Good, & Honey, 2008; Penick & Solomon, 1991), object recognition paradigms (Mumby, Gaskin, Glenn, Schramek, & Lehmann, 2002; Norman & Eacott, 2005) and computer-based memory tasks (Browning, Easton, Buckley, & Gaffan, 2005; Eacott & Gaffan, 2005). Less is known, however, about how animals might make use of integrated memories in their daily lives (Clayton, Griffiths, Emery, & Dickinson, 2001; Clayton, Yu, & Dickinson, 2001). In this study, we investigated whether wild, free-living rufous hummingbirds, Selasphorus rufus, could form integrated memories for what was rewarded, where and in which context, in a naturalistic foraging situation. Hummingbird foraging is a plausible model for examining whether animals can combine different kinds of information into * Correspondence: S. D. Healy, School of Biology, Harold Mitchell Building, University of St Andrews, St Andrews KY14 7AU, U.K. E-mail address: [email protected] (S. D. Healy).
memories. Not only can these birds learn to avoid visiting recently emptied flowers (Healy & Hurly, 1995) and to return to flowers that contain a consistent reward (Gass & Sutherland, 1985; Healy & Hurly, 2003; Hurly, 1996), they can also remember both where and when flowers refill (Henderson, Hurly, Bateson, & Healy, 2006; Marshall, Hurly, & Healy, 2012) and associate this with variations in nectar quality (González-Gómez, Bozinovic, & Vásquez, 2011). Whether rufous hummingbirds, or any other animal foraging in the wild, can combine three pieces of information, including contextual cues, into memories is not yet known. The combination of elements into a memory is one of the prerequisites for episodic-like memory (Tulving, 1972), which involves the integration of what, where and when information (Clayton, Griffiths, et al., 2001) or what, where and which occasion information (Eacott & Easton, 2010; 2012) about a single event. If, in the current experiment, the birds could form a combined memory for these three features of episodic-like memories over repeated exposures, it would suggest that, in principle, they may also be able to form combined memories for single events. Potentially, this could provide a real-world paradigm for addressing the value of episodiclike memories. To assess whether hummingbirds can form combined memories, we based our experimental design on a scene-learning paradigm in which rhesus macaques, Macaca mulatta, learnt the
0003-3472/$38.00 Ó 2014 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.anbehav.2013.12.028
186
S. A. Jelbert et al. / Animal Behaviour 89 (2014) 185e189
location and visual features of rewarded and unrewarded symbols in a visual scene (Gaffan, 1994). In the first condition of our experiment hummingbirds were allowed to forage from sets of artificial flowers on two visually distinctive experimental boards, which were presented repeatedly, one at a time. Both boards comprised matching sets of four flowers, with two colours of flower placed in two locations on the board. A different flower, in a different location, was rewarded on each board. As the four flowers were the same on both of the boards, a bird could not identify which flower contained the reward by remembering only a flower’s colour (what) or its location (where). The reward could be found, however, if the bird formed an integrated memory of the flower’s colour, location and the context in which it was presented. In a second condition, we investigated whether hummingbirds could also form a combined memory involving a temporal context. As hummingbirds use timing cues in their daily lives (Henderson et al., 2006; Marshall, Hurly, Sturgeon, Shuker, & Healy, 2013) we hypothesized that they could integrate temporal sequential cues with other kinds of information. To test this we repeatedly presented birds with a single white board with four flowers, but we alternated the colour and position of the rewarded flower each time we presented the board. The birds had to attend to this alternating sequence in order to identify the rewarded flower. We compared the learning speed and errors the birds made in each of the two conditions to determine whether hummingbirds can form combined memories involving both visual and sequential contexts. METHODS Subjects The subjects in this experiment were eight wild, free-living, adult male rufous hummingbirds that had established their territories around commercial hummingbird feeders, containing 14% sucrose solution, set out in the spring. The experiment was conducted in a valley in the eastern Rocky Mountains, Alberta, Canada (49 210 N, 114 250 W). Rufous hummingbirds migrate to this valley to breed and males establish and defend feeding territories. The territory holder typically feeds from the commercial feeder every 10e15 min, chasing away intruding conspecifics. To enable individual identification, birds with established territories were caught, by putting a mesh cage around a feeder, and the white plumage on their chests was colour marked with nontoxic ink. The data were collected between 0800 and 2000 hours Mountain Standard Time, from May to July 2011. All of the work was carried out under permits from Environment Canada and Alberta Fish and Wildlife, with the ethical approval of the University of St Andrews and the University of Lethbridge, and was conducted in adherence with the ASAB Guidelines for the Treatment of Animals in Behavioural Research. Experimental Apparatus Prior to the experiment we trained birds to feed from the experimental apparatus, which was placed conspicuously within a bird’s territory. The apparatus consisted of a 30 35 cm piece of foam board, mounted onto a 1 m high stake, on which four artificial flowers were positioned. Two small holes were made at both the top and the bottom of each board (10 cm apart, and 3 cm from the edge of the board) into which artificial flowers were inserted (Fig. 1). The artificial flowers were formed from syringe caps inserted into the centre of coloured cardboard discs (6 cm diameter). The syringe cap created a well in the centre of the flower from which the bird could feed. The capacity of an artificial flower (600 ml) was substantially more than a hummingbird could drink in
Figure 1. One of the experimental subjects feeding from a board in the Visual Condition. Boards measured 30 35 cm. Photograph: T.A. Hurly.
a single visit. On each board only one of these four flowers contained desirable 25% sucrose solution and the remaining three flowers contained water, which the birds find distasteful. Experimental Procedure All birds took part in two conditions, a Visual-cue condition and a Sequential-cue condition, with the order counterbalanced across birds. Both conditions followed the same format, in which one rewarded flower on a board could be identified by attending simultaneously to the flower’s colour (what), the flower’s location (where) and the specific board presentation (which context). The same flowers were presented every time; however, a different flower was rewarded in each of the two contexts. We used two different colours of flower (e.g. green and pink) and presented one flower of each colour in each of two locations on the board (i.e. top or bottom, Fig. 2). Which position (left or right) that each flower colour occupied at the top and at the bottom was randomized by coin toss before each trial. Thus the experience for a typical bird would be that in the first context a pink flower at the top of the board was rewarded and in the second context a green flower at the bottom of the board was rewarded (Fig. 2). The boards were presented repeatedly until birds demonstrated that they had learnt the identity of both rewarded flowers, defined by reaching our criterion number of correct responses: choosing the rewarded flower six times in a row (three times on each board). If the criterion was not reached, an upper limit was set at 100 board presentations (50 presentations of each board), at which point the condition was terminated. Visual Condition In the Visual Condition, the which-context cue was a visually distinctive background scene, painted on to the experimental boards (Fig. 2). The order of the board presentations in the Visual Condition was pseudorandomized. In every four presentations, each board was seen twice but boards could be presented in any order within these sets (for example, 1-2-1-2 then 2-2-1-1). This ensured that the two boards were seen equally often throughout the experiment, but that the rewarded flower was not coupled with
S. A. Jelbert et al. / Animal Behaviour 89 (2014) 185e189
Visual Condition Background 1
Background 2
A
B
Sequential Condition Feeder
Presentation 1
Presentation 2 B
A
Figure 2. An example of the apparatus used in each condition. The letters A and B indicate the rewarded flowers. The exact position of the coloured flowers (left or right) in each location was varied according to a random sequence and was not tied to a specific context.
presentation order. Each bird was tested with different board pairs and different colours of flower, which were chosen to be conspicuous on both of the background designs. Sequential Condition In the Sequential Condition, which context was specified by the order in which the boards were presented. One of the coloured flowers in one location of the board was rewarded the first time a board was presented and the other colour and location were rewarded the second time a board was presented. A single white board was used for both presentations, eliminating the possibility that birds could use visual cues to identify the reward. To make the sequential cues salient, in the Sequential Condition a commercial hummingbird feeder was presented in between each pair of board presentations. The feeder was included to ‘set’ the sequence, enabling the two boards to be described as either the first presentation after the feeder or the second presentation after the feeder (the sequence was: FeederePresentation 1ePresentation 2; Fig. 2). The bird was required to visit the feeder before being given the next presentation of a board. In both conditions, the two different colours of flower were randomly presented on either the left or the right. To prevent birds from using the appearance of a specific flower as a cue to the reward, we used multiple versions of the flowers across the trials. In both conditions boards were removed from the stake immediately after each visit by the hummingbird and the next board was set up. The number of trials per day was dependent on how frequently the bird returned to feed from the apparatus and ranged from 17 to 72 trials per day (mean SE ¼ 37 0.47). The two conditions took from 3 to 6 days to complete per bird.
187
considered to be the bird’s choice. The bird was allowed to continue probing until he found the rewarded flower although these subsequent choices were not used in the analysis. First choices were initially recorded in terms of the absolute What-Where-Which properties of that choice (i.e. the flower chosen was the pink flower at the top of the board in the first context). They were then coded in terms of what aspects of that choice could be considered correct and which aspects were errors. On each board one flower was correct (rewarded) and three flowers were incorrect (unrewarded); however, the three incorrect options represented three different types of error. As illustrated in Fig. 3, we classified each incorrect choice as one of the following error types: (1) the same colour as the rewarded flower for that context, but in the wrong location (correct What-Which, wrong Where), (2) the same location as the rewarded flower for that context, but the wrong colour (correct Where-Which, wrong What) or (3) a colour and location that were rewarded together, but not in this context (correct What-Where, wrong Which). By comparing the frequency of each error type we examined whether any aspect of the task was harder to learn than any other. We also compared how birds learnt the What-Where-Which features of the rewarded flower in the Visual and Sequential conditions. RESULTS In all cases where criterion was reached, the final six correct visits were excluded from the analyses. All statistical tests were performed using PASW 18.0 (SPSS Inc., Chicago, IL, U.S.A.). Where relevant, Bonferroni corrections were applied to account for multiple tests. Criterion The birds were highly successful in both conditions. All eight birds reached criterion in the Visual Condition, making an average of 36 visits to the experimental boards (mean SE ¼ 35.88 3.32; criterion visits excluded) before they achieved this. In the Sequential Condition, seven of eight birds reached criterion. The
Correct What-Where-Which
Correct Where-Which Incorrect What
A
Data Coding The data we recorded were the flowers chosen by the birds on each visit to an experimental board and the time of each visit. Birds were considered to have ‘visited’ a board when they probed one or more of the artificial flowers. The first flower probed on a board was
Correct What-Where Incorrect Which
Correct What-Which Incorrect Where
Figure 3. A schematic of the way in which we coded flower choices.
188
S. A. Jelbert et al. / Animal Behaviour 89 (2014) 185e189
birds that reached criterion made an average of 53 visits to the boards (mean SE ¼ 53.29 8.57). Seven of the eight birds reached criterion in the Visual Condition in fewer visits than in the Sequential Condition (paired t test: t7 ¼ 3.05, P ¼ 0.02) but the birds’ performances (percentage correct) in the two conditions were not significantly correlated (Pearson correlation: r6 ¼ 0.46, P ¼ 0.25). Choices and Errors In the Visual Condition, birds were no more likely to make one type of error more or less frequently than the other errors, suggesting that all aspects of the task were equally easy (related sample Friedman’s ANOVA: c22 ¼ 0:58, P ¼ 0.76; Fig. 4a). One bird did not make any Where errors in the Visual Condition, although all of the other birds made a range of errors of each type. In the Sequential Condition, however, the birds’ errors did vary (Friedman’s ANOVA: c22 ¼ 12:19, P ¼ 0.001; Fig. 4a), as the frequency of Which errors was greater than both What and Where errors (post hoc Wilcoxon signed-ranks tests: What: z ¼ 2.37, P ¼ 0.02; Where: z ¼ 2.52, P ¼ 0.01). This difference in the frequency of errors suggests that the birds in the Sequential Condition learnt the What-Where association with relative ease but took
60
(a)
Visual Condition Sequential Condition
50
40
30
Mean percentage of first choices
20
10
0 7
Correct
What
Where
(b)
Which First 10 visits Last 10 visits
6 5 4 3 2 1 0
Correct
What
Where
Which
Figure 4. (a) The mean þ SE percentage of first choices classified as correct, or as one of three types of error, in both the Visual and Sequential conditions (criterion visits excluded). N ¼ 8. (b) The mean þ SE number of first choices classified as correct, or as one of three types of error, in birds’ first 10 visits and last 10 visits to the boards in the Sequential Condition (criterion visits excluded). N ¼ 8.
longer to learn how the rewarded flower altered according to the context. An analysis of the choices on the first 10 and last 10 sequential board presentations (across the whole experiment) supports this view: in the first 10 presentations there was no difference between the average number of What, Where and Which errors made by the birds (Friedman’s ANOVA: c22 ¼ 1:41, P ¼ 0.56), whereas in the final 10 visits birds made no Where errors, a few What errors but a large number of Which errors (Friedman’s ANOVA: c22 ¼ 14:21, P < 0.001; Fig. 4b). Which errors could have been made for one of two reasons: either the bird made choices to both rewarded flowers without attending to the context or the bird repeatedly chose one of the flowers. Four of the eight birds in the Sequential Condition preferred one of the rewarded flowers (chi-square tests: c2 ¼ 6.23e 34.84, P < 0.05), choosing one flower at least twice as often as the other. Two birds selected one flower more than 1.5 times as often as the alternative flower, but their preference was not statistically significant (c2 ¼ 1.58e3.57, P ¼ 0.28e0.08), and the remaining two birds did not show any evidence of a preference (c2 ¼ 0.07e0.16, P ¼ 1e0.79). This demonstrates that the majority of birds made Which errors in the Sequential Condition by tending to select one of the rewarded options consistently, rather than by choosing between them both, until eventually learning to alternate their choices according to a sequence. DISCUSSION In this experiment, wild, free-living rufous hummingbirds were able to combine information about a rewarded flower based on What it looked like, Where it was and in Which context it was rewarded. They could do this when the context was specified with either a visual cue or a sequential cue. Importantly, they did this while also engaging in other behaviours such as displaying to females, perching, chasing away intruders, hawking insects or feeding from natural flowers. To our knowledge, this is the first time that this ability has been demonstrated in free-living animals. We were also able to compare the birds’ learning of two different context-specifying rules, one based on visual cues and the other based on order. Although birds were able to learn both, they found it harder to identify the rewarded flower when they had to learn the sequence in which it occurred than when the context was cued visually, and performances in these two conditions were characterized by different patterns of errors. In the Visual Condition birds made What errors, Where errors and Which errors with equal frequencies, which suggests that they found all three features equally easy to remember. However, in the Sequential Condition birds made relatively more Which errors than any other kind of error. This might suggest that it was easier for them to learn to use a visual rather than a sequential context in order to distinguish between events. This comparative difficulty of learning a sequencebased rule is somewhat surprising, considering that sequential visiting of flowers is the essence of trap lining, which is the way in which at least some species of hummingbirds are supposed to forage (Feinsinger & Colwell, 1978; Gill, 1988). It is possible that sequence-based rules are simply more cognitively demanding: an individual must not only learn the rule but also remember what has gone before, whereas in the visual cue scenario, the bird need only remember the rule. Although confirming this suggestion requires further experimental investigation, there is some evidence that temporal order may be harder to learn than visual context information because the two types of memory appear to be underpinned by different brain regions (Albasser, Amin, Lin, Iordanova, & Aggleton, 2012; Iordanova et al., 2009). Overall, in this study we found that hummingbirds can combine information about what was rewarded where and in which context
S. A. Jelbert et al. / Animal Behaviour 89 (2014) 185e189
in response to stimuli presented in their natural environment, after relatively few exposures. The ability to combine different types of information in memory is one of the prerequisites for forming episodic-like memories (Clayton, Griffiths, et al., 2001; Eacott & Easton, 2010), which involve remembering the What-WhereWhen, or What-Where-Which occasion, of single events. In our task, birds learned all three of the components that make up an episodic memory, albeit over multiple exposures. Although in this experiment it is likely that the hummingbirds used semantic, rulelearning memory to remember the identities of rewarded flowers (Tulving & Markowitsch, 1998), our findings raise the possibility that hummingbirds may be able to form combined memories for single events. If so, rufous hummingbirds could provide a model for investigating the role of episodic-like memories in natural foraging situations. Acknowledgments We thank the Universities of St Andrews (S.A.J. and S.D.H.) and Lethbridge (T.A.H.), the Natural Sciences and Engineering Research Council of Canada (T.A.H.) and NERC (R.E.S.M.) for funding. We also appreciate the comments made on the manuscript by two anonymous referees. References Albasser, M. M., Amin, E., Lin, T.-C. E., Iordanova, M. D., & Aggleton, J. P. (2012). Evidence that the rat hippocampus has contrasting roles in object recognition memory and object recency memory. Behavioral Neuroscience, 126, 659e669. Browning, P. G. F., Easton, A., Buckley, M. J., & Gaffan, D. (2005). The role of prefrontal cortex in object-in-place learning in monkeys. European Journal of Neuroscience, 22, 3281e3291. Clayton, N. S., Griffiths, D. P., Emery, N. J., & Dickinson, A. (2001). Elements of episodicelike memory in animals. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 356, 1483e1491. Clayton, N. S., Yu, K. S., & Dickinson, A. (2001). Scrub jays (Aphelocoma coerulescens) form integrated memories of the multiple features of caching episodes. Journal of Experimental Psychology: Animal Behavior Processes, 27, 17e29. Eacott, M. J., & Easton, A. (2010). Episodic memory in animals: remembering which occasion. Neuropsychologia, 48, 2273e2280. Eacott, M. J., & Easton, A. (2012). Remembering the past and thinking about the future: is it really about time? Learning and Motivation, 43, 200e208. Eacott, M. J., & Gaffan, E. A. (2005). The roles of perirhinal cortex, postrhinal cortex, and the fornix in memory for objects, contexts, and events in the rat. Quarterly Journal of Experimental Psychology Section B: Comparative and Physiological Psychology, 58, 202e217.
189
Feinsinger, P., & Colwell, R. K. (1978). Community organization among neotropical nectar-feeding birds. American Zoologist, 18, 779e795. Gaffan, D. (1994). Scene-specific memory for objects: a model of episodic memory impairment in monkeys with fornix transection. Journal of Cognitive Neuroscience, 6, 305e320. Gass, C. L., & Sutherland, G. D. (1985). Specialization by territorial hummingbirds on experimentally enriched patches of flowers: energetic profitability and learning. Canadian Journal of Zoology, 63, 2125e2133. Gill, F. B. (1988). Trapline foraging by hermit hummingbirds: competition for an undefended, renewable resource. Ecology, 69, 1933e1942. González-Gómez, P. L., Bozinovic, F., & Vásquez, R. A. (2011). Elements of episodiclike memory in free-living hummingbirds, energetic consequences. Animal Behaviour, 81, 1257e1262. Healy, S. D., & Hurly, T. A. (1995). Spatial memory in rufous hummingbirds (Selasphorus rufus): a field-test. Animal Learning and Behavior, 23, 63e68. Healy, S. D., & Hurly, T. A. (2003). Cognitive ecology: foraging in hummingbirds as a model system. Advances in the Study of Behavior, 32, 325e359. Henderson, J., Hurly, T. A., Bateson, M., & Healy, S. D. (2006). Timing in free-living rufous hummingbirds, Selasphorus rufus. Current Biology, 16, 512e515. Hurly, T. A. (1996). Spatial memory in rufous hummingbirds: memory for rewarded and non-rewarded sites. Animal Behaviour, 51, 177e183. Iordanova, M. D., Burnett, D. J., Aggleton, J. P., Good, M., & Honey, R. C. (2009). The role of the hippocampus in mnemonic integration and retrieval: complementary evidence from lesion and inactivation studies. European Journal of Neuroscience, 30, 2177e2189. Iordanova, M. D., Good, M. A., & Honey, R. C. (2008). Configural learning without reinforcement: integrated memories for correlates of what, where, and when. The Quarterly Journal of Experimental Psychology, 61, 1785e1792. Marshall, R. E. S., Hurly, T. A., & Healy, S. D. (2012). Do a flower’s features help hummingbirds to learn its contents and refill rate? Animal Behaviour, 83, 1163e 1169. Marshall, R. E. S., Hurly, T. A., Sturgeon, J., Shuker, D. M., & Healy, S. D. (2013). What, where and when: deconstructing memory. Proceedings of the Royal Society B, 280. Mumby, D. G., Gaskin, S., Glenn, M. J., Schramek, T. E., & Lehmann, H. (2002). Hippocampal damage and exploratory preferences in rats: memory for objects, places, and contexts. Learning and Memory, 9, 49e57. Norman, G., & Eacott, M. J. (2005). Dissociable effects of lesions to the perirhinal cortex and the postrhinal cortex on memory for context and objects in rats. Behavioral Neuroscience, 119, 557e566. Pearce, J. M. (1994). Similarity and discrimination: a selective review and a connectionist model. Psychological Review, 101, 587e607. Penick, S., & Solomon, P. R. (1991). Hippocampus, context, and conditioning. Behavioral Neuroscience, 105, 611e617. Pickens, C. L., & Holland, P. C. (2004). Conditioning and cognition. Neuroscience and Biobehavioral Reviews, 28, 651e661. Rudy, J. W. (2009). Context representations, context functions, and the parahippocampal-hippocampal system. Learning and Memory, 16, 573e585. Tulving, E. (1972). Episodic and semantic memory. In E. Tulving, & W. Donaldson (Eds.), Organisation of memory (pp. 381e403). New York: Academic Press. Tulving, E., & Markowitsch, H. J. (1998). Episodic and declarative memory: role of the hippocampus. Hippocampus, 8, 198e204.