Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Do hummingbirds use contextual information when performing spatial association tasks? Thompson, James 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1994-0632.pdf [ 3.82MB ]
Metadata
JSON: 831-1.0087398.json
JSON-LD: 831-1.0087398-ld.json
RDF/XML (Pretty): 831-1.0087398-rdf.xml
RDF/JSON: 831-1.0087398-rdf.json
Turtle: 831-1.0087398-turtle.txt
N-Triples: 831-1.0087398-rdf-ntriples.txt
Original Record: 831-1.0087398-source.json
Full Text
831-1.0087398-fulltext.txt
Citation
831-1.0087398.ris

Full Text

DO HUMMINGBIRDS USE CONTEXTUAL INFORMATION WHEN PERFORMING SPATIAL ASSOCIATION TASKS? by JAMES THOMPSON B.Sc, The University of Alberta, 1991 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (Department of Zoology) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA July 1994 © James Thompson, 1994 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Ubrary shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of ZlPO^<y^^ The University of British Columbia Vancouver, Canada Date DE-6 (2/88) ABSTRACT One factor hindering the flow of information between the related disciplines of psychology and behavioural ecology is their different foci. Psychologists have searched for general principles governing animal learning and behaviour, while behavioural ecologists have concentrated more on the adaptive significance of behaviour. Focussing investigation on how animals solve particular problems of ecological significance can provide a common frame of reference for both disciplines. This thesis explores the kinds of information rufous hummingbirds (Selasphorus rufus) use to solve spatial association problems. The spatial association paradigm differs from other associative tasks in that subjects must associate a lit cue in an array of several cues with a rewarding feeder in an array of several feeders that are spatially separated from the cue array. In four experiments hummingbirds learned a baseline spatial association task and then performed it in a treatment in which the spatial context of the task was altered. Each experiment precluded the use of a particular kind of information. Experiments 1 and 2 indicated that the birds were using a simple behavioural rule such as "fly to the feeder nearest the lit cue" to perform the spatial association task. This behavioural mechanism was independent of changes in the 11 spatial context of the task, including the geometry of the arrays, on two spatial scales. Experiment 3 showed that hummingbirds could learn a spatial association task in which the correct feeder was one of two equally close to the cue and that performance on this task was independent of the associated elements occurring in a coherent geometric array of other elements. Experiment 4 showed that, contrary to the predictions of associative learning theory, hummingbirds attend to both feeders nearest the lit cue when performing a task like that in Experiment 3. Their performance on this task, however, was strongly linked to the orientation of the cue and two nearest feeders with respect to some global referent. The birds exhibited a large amount of sign-tracking in the above experiments. Analysis of when they sign-tracked in the context of their performance in the tasks suggested that they were approaching the cue as part of their strategy for accurately locating the correct feeder after they had learned the association. Ill Table of Contents Abstract ii Table of Contents iv List of Tables vi List of Figures vii Acknowledgments x Chapter 1: General Introduction 1 Chapter 2: Do Hummingbirds Use Spatially Differentiated Information When Performing Spatial Association Tasks?5 Experiment 1 10 Methods 10 Subjects 10 Experimental Environment 11 Baseline Treatment 12 Restricted Visibility Treatment 14 Training 14 Testing and Data Collection 15 Analysis of Video Records 17 Results 17 Discussion 20 Experiment 2 22 Methods 24 Subjects 24 Experimental Environment 24 Experimental Design 25 Training 25 Testing and Data Collection 27 Results 28 Discussion 31 A Behvioural Model For Spatial Association 34 Experiment 3 35 Methods 38 Subjects 38 Experimental Environment 38 Experimental Design 39 Testing and Data Collection 39 Results 40 Discussion 42 Experiment 4 43 Methods 46 Sub j ects 46 IV Experimental Environment 46 Experimental Design 47 Testing and Data Collection 50 Results 50 Switch 1: Baseline to DISROT 1 50 Switch 2: DISROT 1 to DISROT 2 55 Discussion 58 Mental Rotation Abilities 58 Orientations versus Vectors 61 General Discussion 64 Chapter 3: Sign-Tracking By Hummingbirds Performing A Spatial Association Task 66 Methods 69 General Protocol 69 Experimental Manipulations 71 Video Records 71 Results 73 Acquisition of Sign-Tracking 73 Extinction of Sign-Tracking 74 Discussion 76 Chapter 4: General Conclusions 86 Literature Cited 89 List of Tables Table 1. Number of incorrect first visits (errors) after approaching the lit cue in the HEX treatment. Numbers in parentheses indicate the total number of visits to that light that were followed by visits to a feeder..32 Table 2. Number of trials in which birds approached the lit cue in the last 5 trials of baseline and the first 5 trials in each orientation in DISROT 1. Refer to Figure 12 for orientation designations 77 VI List of Figures Figure 1. (a) Baseline and (b) restricted visibility (top view) arrays. Cues (solid circles) were separated from feeders by 11 cm, and adjacent cue-feeder pairs were also separated by 11 cm. Cues in the RV treatment were 11 cm behind the panel (horizontal dashed line). Dashed lines converging on the single horizontal bar represent the imaginary rays along which each cue could be seen from the perch 13 Figure 2. (a) Mean (±SE; n=6) proportion correct first choices in blocks of 10 trials in baseline and RV treatments (separated by vertical dashed line). The horizontal dashed line indicates chance performance. Negative numbers indicate trials before the switch, (b) Mean (±SE; n=4) proportion of recorded trials in which birds approached the lit cue before probing any feeders in baseline and RV treatments 18 Figure 3. (a) Baseline and (b) hexagonal treatment arrays. Cues (solid circles) were separated from feeders by 3 cm. Adjacent cue-feeder pairs in baseline were separated by 11 cm. Adjacent cues in the HEX treatment were separated by 11 cm, adjacent feeders by 14 cm. Opposite cues were 22 cm apart. Numbers identify cue-feeder pairs (see text) 26 Figure 4. (a) Mean (±SE; n=5) proportion correct first choices in blocks of 10 trials in baseline and HEX treatments (separated by vertical dashed line). Horizontal dashed line indicates chance performance. Negative numbers indicate trials before the switch, (b) Mean (±SE; n=5) proportion of recorded trials in which birds approached the lit cue in baseline and HEX treatments 29 Figure 5. (a) Baseline and (b) disaggregated arrays. Adjacent cues (solid circles) and adjacent feeders were separated by 9 cm. The two feeders closest to a cue were both 9 cm from it and the rightmost was always the correct one. Patches in the disaggregated array were separated by 80.5 cm horizontally and 34.4 cm vertically (cue to cue) 37 Figure 6. (a) Mean (±SE; n=5) proportion correct first choices in blocks of 10 trials in baseline and DIS treatments (separated by vertical solid line). Negative numbers indicate trials before the switch, (b) Mean (±SE; n=5) proportion of recorded trials in which birds approached the lit cue in baseline and DIS treatments 41 Vll Figure 7. (a) Baseline, (b) DISROT 1, and (c) DISROT 2 arrays. Dimensions are identical to Experiment 3 (Fig. 5). Three birds received the arrangements shown at left and three received those shown at right. Letters identify each orientation (see text) 49 Figure 8. (a) Mean (±SE; n=6) proportion correct first choices in blocks of 10 trials in baseline and DISROT 1 treatments (separated by vertical solid line). Negative numbers indicate trials before the first switch. (b) Mean (±SE; n=6) proportion of recorded trials in which birds approached the lit cue in baseline and DISROT 1 treatments 51 Figure 9. (a) Number of visits to the correct patch out of the first four times the cue there was lit in DISROT 1 of Experiment 4. (b) Number of correct first visits in first four positive trials at each orientation in DISROT 1 of Experiment 4 53 Figure 10. (a) Mean (±SE; n=6) proportion correct first choices in blocks of 10 trials in DISROT 1 and DISROT 2 treatments (separated by vertical solid line). Negative numbers indicate trials before the second switch. (b) Mean (±SE; n=6) proportion of recorded trials in which birds approached the lit cue in baseline and DISROT 2 treatments 56 Figure 11. Number of correct first visits in the first 10 trials of DISROT 2 versus in the last 10 trials of DISROT 1 for each orientation. Each point represents a pair of observations for one individual. Diagonal line represents perfect transfer of performance between the two treatments; points above the line indicate performance increased in DISROT 2, points below indicate performance dropped in DISROT 2 57 Figure 12. Baseline treatment arrays for Experiments 1-4. Experiment 1: baseline and restricted visibility arrays. Experiment 2: baseline and hexagonal arrays. Experiment 3: baseline and disaggregated arrays. Experiment 4: DISROT 1 arrays; 3 birds were switched into each arrangement and the baseline treatment was identical to Experiment 3 72 Figure 13. Cumulative approaches to the lit cue versus cumulative correct first visits in baseline treatments of Experiments 1,3, and 4 for naive individuals with continuous video records. Symbols are plotted every 10 data points and different symbols represent different individuals. The solid line represents perfect correlation between the two measures. (a) Baseline treatment of Experiment 1 (3 individuals). (b) Baseline treatment of Experiment 3 (2 individuals). Vlll Note the large number of symbols for one individual. (c) Baseline treatment of Experiment 4 (4 individuals) 75 Figure 14. Cumulative approaches to the lit cue versus cumulative correct first visits for individuals with continuous video records in orientations A, B, C, and D of DISROT 1 (Experiment 4). Symbols are plotted every 4 data points and different symbols represent different individuals. The solid line represents perfect correlation between the two measures 78 IX Acknowledgments I gratefully thank my research supervisor, Lee Gass, for his enthusiastic support of all my endeavours, research or otherwise. The members of my supervisory committee, Don Wilkie, Jamie Smith, and Dolph Schluter, provided valuable advice and stimulating criticism at all stages of this study. I also thank Gayle Brown without whose intellectual and technical achievements this study would not have been possible. Finally, I thank my family for their support and friends for all the good memories. This research was supported by an NSERC operating grant to Lee Gass. Personal funding was provided by a University Graduate Fellowship and teaching assistantships. X CHAPTER 1 GENERAL INTRODUCTION Recently, there have been calls for a unification of the often indistinguishable disciplines of behavioural ecology and comparative psychology (Kamil and Yoerg 1982; Johnston 1985; Gould and Marler 1987). Proponents argue that combining what we learn from rigorously controlled experiments focussing on general learning mechanisms (traditional psychological research) together with knowledge about the adaptive significance of changes in the interactions between animals and their environments (behavioural ecology) leads to a more complete understanding of animal behaviour. A stumbling block in this integrative effort is the task specificity of many psychological theories. In an effort to uncover general laws and principles that govern animal learning, much of traditional psychological research was conducted on relatively few species within strictly specified paradigms designed to minimize the effect of any species specific predispositions (Shettleworth 1989; Timberlake and Lucas 1989). As a result, theories derived from such research tend to be specific to particular methodologies and difficult to apply in behavioural ecology, where the focus is usually on the animals themselves (a very wide variety of species), often in their natural environments (Johnston 1985; Shettleworth 1989, 1993). One way to re-focus questions about behaviour to make them more useful to both disciplines is to ask how as well as whether animals perform ecologically relevant tasks (Miller 1985). Investigating how animals accomplish particular tasks forces consideration of the species specific behavioural repertoire subjects bring to them. Studying animals to learn about the animals themselves provides a common frame of reference for all animal behaviourists, regardless of discipline. We cannot assume that behavioural mechanisms revealed in a semi-naturalistic laboratory task are necessarily those animals use to solve the same problem in nature (Shettleworth 1989), but such efforts can, at the very least, broaden our understanding of animals' capabilities and challenge us to discover their adaptive significance. Tasks requiring animals to use spatial relationships among elements in their environments are especially amenable to this kind of questioning because we can control the kinds, number, and distribution of elements available to them. Brown (Brown and Gass 1993; Brown in press) used this approach to demonstrate that hummingbirds can learn to associate a lit LED cue with a rewarding feeder up to 11 cm away. A unique feature of Brown's paradigm is that she presented a horizontal array of several cues directly above a similar array of feeders to her birds. The location of the lit cue within the array, and hence of the rewarding feeder, changed randomly every trial. The birds readily learned to respond correctly at a different location every trial, and Brown concluded that hummingbirds learn a "spatial association" in which they use the lit cue as a determinant not only of when to respond, but also where. The purpose of this thesis was to determine how hummingbirds accomplish the spatial association task and, more generally, how they use information available in the spatial relationships among elements in their environment. In each of four experiments (Chapter 2) I tested specific hypotheses concerning the kinds of information hummingbirds use by presenting a spatial association task to them in a baseline treatment and then changing the spatial arrangement of the cues and feeders to eliminate the use of a particular kind of information. Other investigators (e.g. Nadel and Winner 1980; Honey et al. 1990) have demonstrated that changing the context (the visual and/or olfactory environment) in which a task occurs can greatly affect its performance. Outlining hypotheses about specific mechanisms a priori allowed me to infer how such changes might affect hummingbirds' ability to perform the spatial association task. I videotaped the experiments and quantified some aspects of cue directed behaviour. Such behaviour is often observed in other paradigms that employ spatially localized cues as predictors of reward and is called sign-tracking (Hearst and Jenkins 1974; Tomie et al. 1989). Sign-tracking is a phenomenon that seems to combine elements of both Pavlovian and instrumental conditioning (two types of learning that psychologists traditionally studied separately) and has thus generated a large amount of interest from psychologists. Studies of sign-tracking are typically conducted under an autoshaping paradigm which is very different from the one in which the hummingbirds sign-tracked. In Chapter 3 I analyzed the hummingbirds' sign-tracking in the spatial association tasks to evaluate whether current theories about the phenomenon adequately account for its expression in these experiments. CHAPTER 2 Do Hummingbirds Use Spatially Differentiated Information When Performing Spatial Association Tasks? Rescorla (1980) identified three questions that frame the study of any learning process: Under what conditions does learning occur? What does the animal actually learn? How is this learning expressed in its behaviour? The answers to these questions could greatly improve the predictive capability of species specific behavioural models by enhancing our understanding of the information particular species use to guide their behaviour (Janetos and Cole 1981; Dill 1983; Krebs and McCleery 1984; Kamil et al. 1993). Defining exactly what information animals use while they perform tasks requiring the use of spatial relationships among elements in their environment can be especially problematic. A rat attempting to re-locate a platform submerged in an opaque liquid (Morris water maze; Morris 1981), for example, has spatially differentiated information available to it on several spatial scales. Information is present in the arrangement of objects in space from a local (e.g. the location of the platform in relation to the walls of the tank) to a more global scale (e.g. the orientation of the tank in the experimental room). Only by further experimental investigation can we reveal what information the rat uses to navigate while in the tank. In general, the question to be answered is: Does the ability to perform tasks requiring the use of spatial relationships depend on the spatial context in which they are learned? Spatial memory and orientation tasks are relatively simple to study in this context. If animals somehow encode the spatial organization of certain elements in their environment with respect to some goal (e.g. a food cache or burrow), then the goal can be thought of as occurring in a spatial context. By manipulating this context we can determine if, and in what way, subsequent goal rediscovery depends on the initial conditions under which its location was learned. This approach has demonstrated that, for many animal species, spatial memory depends strongly upon the spatial context under which the task was originally learned (see Gallistel 1989 for review). Related studies focus on factors governing which elements animals choose as landmarks (Cheng and Sherry 1992; Spetch and Cheng 1992). Another kind of task that requires the use of spatially differentiated information is spatial association (Brown and Gass 1993). Under this paradigm, stimuli are paired with response sites spatially separated from them. When stimulus and response sites are contiguous, we generally assume that animals learning the association between them encode both the stimulus and the reward (Rescorla and Holland 1982; Rescorla 1985). With spatially separated elements, however, animals must associate the stimulus not only with the availability of food but also with its location. The requirement that animals seek reward at a site spatially separated from the stimulus is similar to an instrumental learning task. Animals performing such tasks are thought to associate their responses to the stimulus with the outcome of those actions (Colwill and Rescorla 1986; Rescorla and Colwill 1989). Both stimulus-reward and response-outcome associations are specific to the spatial association task itself, but are they the only information animals attend to in performing this task? Scavengers using circling vultures as an indication that food is available below them (Rabenold 1983), for example, should look for carrion wherever vultures circle. That is, in this case the scavengers' ability to learn and manifest this association should be insensitive to details of where it occurs; it should be "portable". In general, we might expect that once learned, performance in a spatial association task should be unaffected by changing the spatial context in which the associated pair of elements occurs. But how robust is this portability? To find out, the spatial context could be perturbed at several spatial scales as with spatial memory tasks, and in several ways. Thus, remembering that performance in a spatial association task is the combination of learning the association and manifesting it, we can ask: Are spatial association abilities independent of changes in the spatial structure of the environment over all spatial scales, or is contextual information at a particular scale essential (e.g. the orientation of associated elements with respect to a global referent such as the pull of gravity)? Spatial association has historically been difficult to demonstrate (Bowe 1984), but Brown (Brown and Gass 1993; in press) has shown recently that hummingbirds exhibit remarkable abilities at such a task in a laboratory setting. In Brown's experiments, hummingbirds learned to associate a lit, red LED (light emitting diode) cue with a rewarding feeder below it in horizontal arrays of six adjacent evenly spaced cue-feeder pairs. She used a trial based procedure in which the location of the lit cue and hence of the rewarding feeder changed randomly between trials. Brown identified several factors that influenced the learning of this task, but her experiments did not address whether hummingbirds reguire some spatial structure in the environment other than the specific relationship between a lit cue and its feeder to perform the task once it is learned. To answer this question, I performed several experiments in which I tested the possibility that in manifesting a learned spatial association (i.e. in using behavioural rules that facilitate recovery of the reward) hummingbirds use information available in the spatial context in which the task occurs. In each of four experiments, I allowed hummingbirds to learn a spatial association task in a baseline treatment and then eliminated one kind of information on a particular spatial scale by 8 changing the spatial context of the task. I used two different spatial association tasks and manipulated the spatial context on three different spatial scales. In the first spatial association task (Experiments 1 and 2), the feeder closest to the lit cue was rewarding (similar to Brown and Gass 1993). In Experiment 1 I restricted the visibility of the cues (a local scale manipulation) to determine if the birds required the relationship between the cues and feeders to be visible at all times after having learned the spatial association. In Experiment 2, I hypothesized that birds could be using features of the experimental environment (e.g. the edges of the panel containing the cues and feeders, the walls of the experimental room, or the pull of gravity) to aid their performance of the task. Thus, I altered the orientation of the associated elements (cue-feeder pairs) with respect to these features to determine if they were used as non task-specific global cues. In the second spatial association task (Experiments 3 and 4), only one of two feeders equidistant from each cue was rewarding. In Experiment 3, I asked whether grouping all cues and feeders into a visually coherent array facilitates manifestation of spatial association. Experiment 4 was similar to Experiment 2 in that I altered the orientation of sets of elements (a cue and two feeders equidistant from it) with respect to non task-specific features of the experimental environment. EXPERIMENT 1 It is well documented that animals approach cues that signal rewards (Tomie et al. 1989) and hummingbirds are no exception (G. S. Brown personal communication). The hummingbirds in Brown's experiments typically approached and hovered in front of the lit cue before visiting any feeders. Once at the cue, however, they must fly to the correct feeder below in order to receive the food. Is the cue required for this phase of the response also, acting perhaps as a beacon or landmark allowing the birds to navigate by dead reckoning (Gallistel 1989)? I tested this possibility by allowing hummingbirds to learn a spatial association task in a baseline treatment and then switching them to a restricted visibility (RV) treatment. The RV treatment allowed birds to approach the lit cue, but made it impossible for them to use it as a beacon to navigate from there to the correct feeder. If hummingbirds require a constant spatial context on a local scale in order to manifest the association, the RV treatment should greatly impair their performance. Methods Subjects Six adult rufous hummingbirds (Selasphorus rufus) were used in the study. The five males and one female were captured on Sumas Mountain, British Columbia in May and June of 1992 and participated in experiments from September to 10 November of the same year. The birds were housed individually in wire mesh cages (60 x 60 x 60 cipa) in a large room at the UBC Animal Care facility under a constant photoperiod (15h light:9h dark). On weekdays the birds were fed a commercial hummingbird food (Nektar Plus, Nekton USA Inc.) supplemented with soy protein (6% w/w) ad libitum from a commercial hummingbird feeder. On weekends they received 20% (w/w) sucrose solution. Three days before it was used in an experiment, each bird was moved to the laboratory where it was housed in a holding cage under identical conditions until the experiment began. Experimental Environment 1 used exactly the same experimental rooms and hardware as Brown (Brown and Gass 1993). I used two experimental rooms (1.3 X 2.5 x 2.5m high) each lit with three 40W incandescent bulbs. Part of one of the end walls in each room was a flat green aluminum panel (106 x 61 cm high) which housed an array of six red LEDs (4 mm in diameter) and feeders. The feeders were holes (3 mm in diameter) surrounded by fluorescent orange Avery labels (19 mm in diameter). Behind each feeder (outside the experimental room) was a photocell and attached to it was a sucrose reservoir. When a bird probed a rewarding feeder with its bill, triggering the photocell, a solenoid pump (General Valve Corporation, Series 3) delivered 2iLil of 20% sucrose (w/w) into the reservoir. Each room contained a perch (180 11 cm from the panel and 153 cm above the floor) equipped with a photocell, a video camera (Sony Handycam CCD-VlOl) on a tripod located just behind and below the perch, and the camera's remote control which was attached to a wall. A computer controlled all the hardware, turning the LEDs and cameras on and off, delivering food, and monitoring the photocells. Occasionally, the relay that started the camera misfired resulting in a trial not being recorded, and sometimes the videotape ran out during a trial (I changed the tape during the following intertrial period), thus excluding it from the video record. Both these occurrences together always accounted for fewer than 5 trials (2.7% of total) being "lost" for any particular bird. Baseline Treatment The baseline treatment allowed the birds to learn the spatial association task and generated measures of performance and behaviour to which I could compare the results in the restricted visibility treatment. Six cue-feeder pairs were arranged in a rectangular array with each cue separated from a feeder directly below it by 11 cm (Fig. la). At this distance, birds in Brown's (unpublished data) experiments often stopped and looked at the lit cue as they navigated from it to the feeder. Adjacent cue-feeder pairs were also separated by 11 cm because Brown (in press) found that her birds learned the task relatively quickly when 12 (a) • • • • • • \ \ I / / / (b) Figure 1. (a) Baseline and (b) restricted visibility (top view) arrays. Cues (solid circles) were separated from feeders by 11 cm, and adjacent cue-feeder pairs were also separated by 11 cm. Cues in the RV treatment were 11 cm behind the panel (horizontal dashed line), Dashed lines converging on the single horizontal bar represent the imaginary rays along which each cue could be seen from the perch. 13 adjacent cue-feeder pairs were widely spaced. Thus the baseline treatment was designed to produce rapid learning, strongly biasing them to their preferred strategy for manifesting the spatial association. Restricted Visibility Treatment The restricted visibility (RV) treatment was identical to baseline except that cues were moved from the panel to a Plexiglas holder 11 cm behind it. To see a particular LED, birds had to be within its narrow cone of emitted light and, more importantly, on the same horizontal plane. The LEDs were arranged so that all could be seen from the perch (Fig. lb). I draped a large sheet of opaque black plastic behind the cue array to exclude ambient light from the laboratory and ensure that only lit cues could be seen from inside the experimental room. Training All birds received training prior to experiments. When in holding cages at the laboratory, birds fed from feeders covered by flat green plastic masks, each with a hole surrounded by an orange Avery label through which the food delivery tube of the feeder protruded. I transferred a bird and its masked feeder into an experimental room with the cue and feeder arrays covered approximately three days after it arrived at the laboratory. Two days later I removed the feeder, uncovered one cue-feeder pair, and allowed the bird 14 to feed at will from the panel feeder. After the bird had fed several times, I re-covered the array. The following day (day 1 of the experiment), after the bird had fed from a panel feeder, I lit the cue above it and the bird was again free to feed at will. After the bird had fed several times, I removed all the covers from the array and the bird could visit all feeders, but it received food from only the feeder directly below the lit cue. I then ran a training program on the computer that lit each of the six cues once in a random order and allowed the bird four feeding bouts at the feeder below the lit cue before its cue was extinguished and a new one lit. This training sequence typically lasted 2 hours and testing began immediately afterward. Testing and Data Collection I followed a trial based testing procedure identical to that of Brown (in press). Five seconds before the start of each trial, the computer turned the video camera on. At the start of each trial, if the bird was perched, a soft buzzer behind the panel sounded for 0.5 s to focus its attention on the panel. If the bird was not perched, the trial started as soon as it returned. As the buzzer sounded, the LED above the rewarding feeder came on; the same random sequence of rewarding feeders was used for all birds. During trials birds could trigger the rewarding feeder and obtain 2iLil of 20% sucrose solution up to 12 times. They 15 could visit any feeder in any sequence but always received food at the rewarding feeder (up to the 12 visit maximum). A trial ended if the bird made the maximum number of visits to the rewarding feeder, if it probed any feeder and then perched, or if it was off its perch for more than 15 seconds. At the end of a trial the computer turned the cue and camera off and a two minute intertrial period began. The birds frequently visited the feeders and cues during the intertrial period but received no food. A course of trials typically lasted three to four hours and birds stayed in the experimental room with their regular feeders overnight if they were to be tested the following day. Based on Brown's (in press) performance data at this separation distance, I gave the birds 90 trials on each treatment, but to guard against poor performance and hence lack of food causing the birds' condition to deteriorate, I gave them only 60 trials per day. This meant that birds began the RV treatment after 30 baseline treatment trials on day 2 of the experiment. After that, I stopped the experiment, entered the experimental room and removed the cues from the panel. I then left the experimental room, placed the cues in the holder and immediately restarted the experiment. The entire switch procedure took between 5 and 10 minutes. 16 Analysis of Video Records I used a VCR that allowed frame-by-frame tape advancement for analysis of video records. Since each frame of videotape represents 0.033 s, I determined the duration of behaviours by counting the number of frames during which they were exhibited and later converting this to seconds. For each trial I recorded whether the bird approached a cue, how long it stayed there, which feeder it subsequently visited and how long it took to fly from the cue to the feeder. I scored an "approach to the cue" if the bird hovered in front of, or near a cue and a "visit to a feeder" if the bird stopped in front of, or probed it. Results Birds learned the baseline condition quickly and well, although they varied considerably among themselves and 3 performed somewhat erratically even after 90 trials. However, all birds were well above chance by trial 60, three performed perfectly over their final 20 trials and overall they averaged 82% performance over their last 30 trials (Fig. 2a). This performance is similar to Brown's (in press) results at similar cue-feeder and cue-feeder pair separation distances. If the birds were using the lit cues only as beacons to navigate to correct feeders in the baseline treatment, they should have dropped to chance performance after the switch when the beacons were unavailable as navigational aids. 17 TO CO > ^ CO L _ LL O 0) c; o 100 80 60 40 20 -0 -90 I 100 o CO o Q-d < CCS o c .o o a. o i _ CL 80 -60 -40 -20 0 Baseline RV Treatment Figure 2. (a) Mean (±SE; n=6) proportion correct first choices in blocks of 10 trials in baseline and RV treatments (separated by vertical dashed line). The horizontal dashed line indicates chance performance. Negative numbers indicate trials before the switch, (b) Mean (±SE; n=4) proportion of recorded trials in which birds approached the lit cue before probing any feeders in baseline and RV treatments. 18 Overall, performance decreased significantly after the switch but remained well above chance (Fig. 2a; mean performance 30 trials before versus 30 trials after switch; paired t-test: t=-4.38, n=6, P=0.007). Individual performances reveal more, however. Two birds continued fairly high performance post-switch, and another quickly regained 100% performance. Another two birds, who had performed erratically in the baseline treatment, continued to do so after the switch albeit at a slightly lower level. Bird 3 performed erratically before the switch and was performing so poorly by trial 74 after it that I removed him from the experiment. He died 4 weeks later from unknown causes. Thus, while restricted visibility of the cues significantly affected overall performance, the birds performed above chance and in some cases, perfectly. Other measures of behaviour provide insight into how these changes occurred. Analysis of video data (4 of the 6 birds) showed that the birds approached the lit cue significantly less frequently after the switch (Fig 2b; proportion of recorded trials in which birds approached the lit cue in baseline versus RV treatments; paired t-test: t=-4.48, n=4, P=0.02). When birds approached the lit cue under baseline conditions, they typically flew directly from their perch to hover in front of it briefly before dropping to a feeder. Such flights from perch to feeder were significantly longer in duration (1.8 s versus 1.2 s on average) than those in which they approached the feeder 19 directly (time from the perch to the feeder in baseline trials in which birds approached the lit cue versus those in which they did not; paired t-test: t=4.915, n=3, P=0.039). In most RV trials, the birds flew from the perch as if they would approach the cue, but instead began their drop to the feeder approximately 0.5 m from the array without pausing. This alternative approach strategy allowed birds in the RV treatment to arrive at a feeder from the perch approximately 0.45 s sooner, on average, than in baseline trials (mean duration of flights from the perch to a feeder in baseline versus RV treatments; paired t-test: t=-3.23, n=6, P=0.02). There was one other notable change in behaviour after the switch. Part of the switch protocol required me to enter the experimental room, during which time the birds were very agitated and remained so for about half an hour afterward. Although the effects of this are difficult to isolate and quantify, it appeared that agitated birds often made more errors than when they were calm. Discussion Overall performance on the spatial association task dropped significantly in the RV treatment but remained well above chance. Had birds been relying on lit cues solely as beacons, they would not have been able to perform above chance under RV conditions. The fact that they performed well above chance coupled with their greatly reduced amount of cue directed behaviour shows that their ability to locate 20 the correct feeder cannot be explained purely as an inflexible, mechanistic response to the lighting of a cue. But why did their performance drop at all? Over the last 30 trials of the RV treatment, most individuals (all but bird 3) performed at a level higher than that immediately after the switch. This suggests that the new treatment had some temporary performance inhibiting effects. The new appearance of the array after the switch might have induced some neophobia (Greenberg 1984, 1990). The birds in Brown's (unpublished data) only switch experiment with a protocol similar to the present study (from visible guides between cues and feeders to without) also demonstrated an initial decrement in performance to a level lower than that at which their performance eventually stabilized. In addition, my presence in the experimental room during the switch greatly agitated the birds, which also may have had a negative impact on their early post-switch performance. The above effects must be attributed to the experimental protocol, however, rather than to an experimental effect of the RV conditions themselves. An effect of restricting the visibility of the cues unique to that treatment might be that the birds could not "check" their choice of feeder. If they were using a rule such as "fly to the feeder nearest the lit cue", then it is possible that they "checked" this choice en route from the cue to the feeder in order maximize their accuracy. Thus, when the cue was no longer continuously available under RV 21 conditions, the birds may have stopped approaching it because its utility in improving their accuracy when manifesting their rule was diminished. Although their post-switch accuracy was diminished, their foraging flights were shorter, on average, than in baseline. This raises the interesting possibility that the birds were trading off performance (a measure of energy intake) and energy expenditure (in terms of foraging flight duration). In sum, hummingbirds' spatial association ability is fairly robust in the face of changes to the visibility of the cues; a change in the spatial context of the task on a local scale. The associations they learned between the lit cue and the availability of reward and between probing the feeder nearest the lit cue and receiving food probably remained intact. The RV treatment affected their accuracy in manifesting this knowledge, but not enough to prevent performance well above chance. This does not preclude the possibility that they were using information available at a more global scale, however. EXPERIMENT 2 Given that the birds in Experiment 1 did not rely on the information available in the cue at a very local scale, what other information in addition to the relative positions of the cues and feeders could they use to accomplish the spatial association? Cheng and Sherry (1992) showed that pigeons use the perpendicular distance of a goal from an 22 edge when encoding its spatial position. There is also evidence that animals can navigate by following vectors from landmarks (Gallistel 1989; Nadel 1990). For hummingbirds performing the spatial association task, one possibility is that they use some global referent not specific to the task, such as the pull of gravity, to guide their drop from the lit cue (or from the horizontal plane of the lit cue, as in the RV treatment) to the feeder. Under this hypothesis, after having learned the task with the cue-feeder pairs aligned with the pull of gravity, they should be sensitive to re-orienting them in this more global spatial context and their performance should suffer. Brown (unpublished data) addressed this issue indirectly by shifting the horizontal row of feeders exactly one feeder position with respect to the row of cues (thus creating an array of cues and feeders in the outline of a parallelogram). Birds successfully learned both "right" and "left" parallelograms, allowing Brown to conclude that they could learn a spatial association task when the correct feeder was not nearest the cue. Brown's experiment demonstrated that hummingbirds can learn to associate a feeder with a cue located along a vector originating at the cue and at an angle with respect to the pull of gravity. Since all cue-feeder pairs were aligned the same way, however, it provides little evidence that this ability is independent of the orientation of these pairs in this larger context. One way to test this using 23 Brown's paradigm would be to switch birds from a "right" to a "left" parallelogram. It could be argued, however, that the birds would have only to learn one new vector angle with respect to the pull of gravity to accomplish the new task, making detection of any experimental effect difficult. I performed a more robust test in which I changed the orientation of the six cue-feeder pairs in several different ways simultaneously thus precluding the simple use of a rule based on the pull of gravity. Methods Subjects I used the five surviving birds (four males and one female) from Experiment 1 in Experiment 2. Experimentally naive birds were not necessary for this experiment, which examined the portability of foraging strategies rather than the learning of the strategies themselves. They were housed and fed under identical conditions as in Experiment 1. Trials began in January of 1993, 11 weeks after the end of Experiment 1. Experimental Environment I used exactly the same experimental rooms and hardware as in Experiment 1, except for the arrangement of LEDs and feeders. 24 Experimental Design This experiment was designed to test the possibility that the birds were using some global referent (such as the pull of gravity) to aid navigation from the lit cue to the correct feeder. There were two treatments: a baseline treatment, in which the cue-feeder pairs were arranged in a linear array to establish a baseline of performance and behaviour and a hexagonal array treatment (HEX) in which the cues and feeders were at the corners of two concentric hexagons (Fig. 3). The cue-feeder pairs in the baseline treatment were separated by 11 cm as in the baseline of Experiment 1, but the elements of each pair were separated by only 3 cm (the easiest distance in earlier experiments; Brown in press). With this, I hoped to ensure that the birds would attain high performance quickly, thus maximizing my ability to detect any experimental deterioration of performance. The cue-feeder distance was maintained in the HEX treatment. The feeders were at the corners of the outer hexagon thus minimizing any relative spacing effects by actually increasing the distance between adjacent feeders relative to baseline (Brown in press). Training Training prior to the start of experimental trials was identical to that in Experiment 1 but with the following modification: the birds never received food at a feeder 25 (a) • 5 1 • (b) • 4 2 • O 3 O Figure 3. (a) Baseline and (b) hexagonal treatment arrays. Cues (solid circles) were separated from feeders by 3 cm. Adjacent cue-feeder pairs in baseline were separated by 11 cm. Adjacent cues in the HEX treatment were separated by 11 cm, adjacent feeders by 14 cm. Opposite cues were 22 cm apart. Numbers identify cue-feeder pairs (see text). 26 unless the cue above it was lit. With this change, I hoped to increase the speed with which they learned the association between the lit cue and food reward. Thus, when I uncovered a cue-feeder pair for the first time, the cue was lit immediately. I expected that previous experience in Experiment 1 would also facilitate rapid learning in the baseline treatment. Testing and Data Collection In general, methods were identical to those used in Experiment 1. The specific protocol here was somewhat more complex, though, since I had to physically replace panels to perform the switch. Before starting the experiment, I decided that the birds would have to reach a criterion of 80% performance over their last 30 trials (26 out of 30 correct) in the baseline treatment before they began trials in the HEX treatment. All birds reached this criterion within the first 60 trials and so all received the HEX treatment at the beginning of the second day. At the end of day 1, I put the birds back in their holding cages while I changed the panels in the experimental rooms, then immediately returned them to the experimental rooms with their regular feeders. The arrays were covered until the following morning when I removed the regular feeders and began trials immediately thereafter, so each bird's first experience of the hexagonal array was on its 61st trial. 27 I videotaped all birds in both treatments, although technical difficulties resulted in the video record for two birds missing seven trials in one case and two in the other. Video analysis was the same as in Experiment 1. Results The birds learned the baseline spatial association task quickly and well (Fig. 4a). All reached the criterion of 80% correct over their last 30 trials within 60 trials and all performed perfectly in the last 10 trials before the switch. These results are almost identical to those obtained by Brown (in press) under similar conditions. Performance dropped slightly but significantly at first under the HEX treatment (number of correct first visits 10 trials before the switch versus 10 trials after it; paired t-test: t=-3.207, n=5, P=0.033). One of the five birds continued at 100% immediately after the switch, however, and only 2 of the others averaged less than 80% in the first 10 trials after the switch (both averaged 60%). All birds eventually regained perfect performance over at least one block of 10 trials. Video data revealed that all five birds approached the cue in many baseline trials, as in the baseline treatment of Experiment 1. Two approached it in more than 80% of recorded trials, and the other three approached it in less than half of their trials (Fig. 4b). Probably because of the short cue-feeder separation distance, flights in which 28 o O 100 CO > - ^ — I CO LL "c3 40 -CD 20 0 Baseline HEX -60 -40 -20 0 Trial 20 40 60 CO CD x: o CO o CL < m CO o d o o ex o 100 80 -60 -40 -20 0 (b) ( -. _ i - . , -1 * ] 1 Baseline HEX TrGatment Figure 4. (a) Mean (±SE; n=5) proportion correct first choices in blocks of 10 trials in baseline and HEX treatments (separated by vertical dashed line). Horizontal dashed line indicates chance performance. Negative numbers indicate trials before the switch, (b) Mean (±SE; n=5) proportion of recorded trials in which birds approached the lit cue in baseline and HEX treatments. 29 birds approached the lit cue before visiting the feeder were not significantly longer than direct flights from the perch to the feeder (mean duration of flights from the perch to the feeder in which birds approached the lit cue versus those in which they did not; paired t-test: t=1.710, n=5, P=0.162). As in Experiment 1, the birds were agitated by my presence in the experimental room to uncover the array just before trial 61. All but one of the birds visited the lit cue less in the HEX treatment than in baseline (Fig. 4b), but this trend was insignificant (proportion of trials in which birds approached the lit cue in baseline versus HEX treatments; paired t-test: t=-1.72, n=5, P=0.160). Nor was there a difference in the time it took the birds to reach the feeder from the perch in the baseline and HEX treatments (mean perch to feeder flight durations in baseline versus HEX treatments; paired t-test: t=0.75, n=5, P=0.50). If the birds were using some global referent such as the pull of gravity as an aid in baseline, their post-switch errors should be strongly biased to feeders below the lit cue. This expectation applies only to cue-feeder pairs 1, 2, 4, and 5 since all feeders were below cue 6 and only one was below cue 3 (Fig. 3b). I analyzed only trials in which the birds visited the lit cue prior to probing any feeders because trials in which they did not reveal nothing about the hypothesis that they were flying from the cue along a vector aligned with the pull of gravity. Only at cue 5 was 30 there a large downward bias in their errors (Table 1). However, one bird was responsible for four of the six errors at this cue and this was because the correct feeder failed to deliver food (the photocell that triggered food delivery suddenly stopped working and thus did not deliver food or record visits). The bird promptly learned to ignore that correct feeder, probing feeder 4 instead. In general, then, post-switch errors do not support the hypothesis that birds relied heavily on global referents such as the pull of gravity when navigating from the lit cue to a feeder. Discussion Disrupting the orientation of the cue-feeder pairs with respect to a global scale referent had little if any effect on performance of a spatial association task. The slight, ephemeral drop in some individuals' performance after the switch could be due to neophobia (Greenberg 1984, 1990), to my presence in the experimental room immediately before to the onset of trials agitating the birds, and/or to the fact that they received no training prior to their first trial on the HEX treatment. Any of these factors could have disturbed the birds enough to negatively affect their initial performance in the HEX treatment. The lack of significant changes in other measures also indicates that to the birds, the baseline and HEX treatments were the same, or at least closely related spatial association tasks. 31 Table 1. Number of incorrect first visits (errors) after approaching the lit cue in the HEX treatment. Numbers in parentheses indicate the total number of visits to that light that were followed by visits to a feeder. errors to any feeder below cue errors to any feeder above cue total 1 1 0 0(10) cue 2 1 1 number 4 2(47) 1 1 2(32) 5 0 6 6(50) 32 Several studies have examined the portability of performance in rats and pigeons in a classical conditioning paradigm (Bouton and Peck 1989; Hall and Honey 1989; Honey et al. 1990; Bouton et al. 1993; but see Lovibond et al. 1984; Kaye and Mackintosh 1990). These studies showed that performance in classical, Pavlovian conditioning tasks is impaired when the conditioned stimulus (CS) and food reward (unconditioned stimulus, US) are presented in a context different from the one in which the task was learned. In these cases, "context" typically includes both visual and olfactory features of the environment. The explanation proposed for the context specificity of appetitive conditioning is that the context in which subjects learn the relationship between the CS and US facilitates their retrieval of information about it. Presenting the CS and US in a different context is thus thought to eliminate its facilitative effect, leading to a decrement in performance in the conditioning task. Since these studies do not address instrumental learning, it is unclear whether a response-outcome association would also be affected by such changes. The RV and HEX treatments changed the visual context and this, in addition to the factors mentioned above, may have impaired the birds' ability to manifest the associations they learned in baseline. This explanation, however, does not account for the ability of some individuals to continue to perform at pre-switch levels immediately after it. 33 A Behavioural Model for Spatial Association An interpretation that accounts for perfect transfer of spatial associations between different spatial contexts is that the birds encode the spatial relationship between the lit cue and its feeder as an abstraction. The instrumental part of this task (the response-outcome association) can be thought of as a rule (e.g. "fly to the feeder nearest the lit cue") that allows birds to respond appropriately to elements matching this abstraction. There is a large amount of evidence that animals encode the spatial relationships of elements in their environments into map-like representations (hummingbirds: Armstrong et al. 1987; Sutherland and Gass unpublished data; and see reviews by Gallistel 1989; Nadel 1990). Hummingbirds, already biologically predisposed to forming abstractions of important features in their environment, could apply this ability to the spatial association problem. This model fits the data from Experiments 1 and 2 well. In baseline, the birds could form an abstraction that contains the lit cue and its feeder and learn a rule (e.g. "fly to the feeder nearest the lit cue") allowing them to find the food in the "environment" of the panel. Performance in each trial would depend on the birds recognizing, and responding appropriately to, sets of elements in the panel matching their abstraction. Their above chance performance in the RV treatment of Experiment 1 shows that they were able to do this (i.e. the spatial 34 association was portable), but they adopted a new, faster, but less accurate method of manifesting their rule for finding the food ("fly directly to the feeder nearest the lit cue"). In Experiment 2, the birds recognized elements as matching their abstraction, and applied their response rule ("fly to the feeder nearest the lit cue") equally well in both the baseline and HEX treatments. Can the model account for the effects of the relative spacing between cues and feeders on the speed with which hummingbirds learn the spatial association task (Brown in press)? The model assumes that achieving good performance (i.e. a high percentage of correct first choices in a given number of trials) depends not only on learning that the lit cue indicates that food is available, but also on learning how to find it. If increasing the vertical ("tall" treatments) and decreasing the horizontal ("packed" treatments) separation between the cues and feeders made "the feeder nearest the lit cue" more difficult to discern, then we might expect birds to learn this rule more slowly. This should result in birds in "tall" or "packed" treatments requiring more trials to achieve good performance, which is what Brown found. EXPERIMENT 3 I concluded from Experiments 1 and 2 that performance of a spatial association task by hummingbirds was unaffected by changing the spatial context on local or global scales. 35 I also proposed a model under which the birds encode the lit cue and rewarding feeder as an abstraction. But what of the other cues and feeders in the array? Brown's spatial association paradigm differs from other associative tasks in that instead of one CS and US, there is an array of several of each kind of element. In terms of the number of visible elements, the spatial context of the task is richest at the scale of the array and Brown (in press) has demonstrated that different arrangements of these elements affect the speed with which hummingbirds learn the spatial association task. The purpose of this experiment was to test whether changes at this scale would also significantly affect their performance in the task after it had been learned. In this experiment I used a spatial association task in which the correct feeder was not the one closest to the cue. The cue-feeder array was trapezoidal in outline so that two feeders were equidistant from each cue (Fig. 5). The profitable feeder was always both the rightmost feeder of the two closest to the lit cue and also the feeder at a particular angle from the cue with respect to some global referent such as the pull of gravity. To determine whether having the cue-feeder pairs embedded in a coherent array of elements facilitated their recognition as such by the birds, I disaggregated sets of one cue and two feeders from the array. I predicted that performance would be impaired if the birds' abstraction of, or rule for responding in the 36 • • • • (a) o (b) O Figure 5. (a) Baseline and (b) disaggregated arrays. Adjacent cues (solid circles) and adjacent feeders were separated by 9 cm. The two feeders closest to a cue were both 9 cm from it and the rightmost was always the correct one. Patches in the disaggregated array were separated by 80.5 cm horizontally and 34.4 cm vertically (cue to cue). 37 task required the associated elements to be embedded in a larger geometric structure. Methods Subjects Five adult rufous hummingbirds fSelasphorus rufus) were used in this study. Two were captured on Sumas Mountain, British Columbia in May and June of 1992. Three were captured at the same location in May and June of 1993. All were tested in May and June of 1993. Four were experimentally naive at the start of the experiment, and one had participated in Experiments 1 and 2 (2 months earlier). The housing and maintenance of the birds were identical to those in previous experiments. Experimental Environment I used the same experimental rooms as in Experiments 1 and 2. I attached all the feeders and LED holders for both the baseline and disaggregated treatments to one flat green panel. Elements not in use in a particular treatment were covered with acetate painted the same flat green as the panel and attached to it with double sided tape. The birds gave no indication that they distinguished these covers from the actual surface of the panel. The rest of the hardware was the same as in previous experiments. Technical failures of the video equipment plagued this experiment and numerous trials were lost to the video 38 record. Conclusions based on video data must, therefore, be regarded as tenuous at best. Video analysis was the same as in previous experiments. Experimental Design This experiment was designed to examine the effect of changing the spatial context of the spatial association task by perturbing the spatial structure of the array of cues and feeders. As in previous experiments, a baseline treatment allowed birds to learn the spatial association task and provided me with baseline performance measures. I presented birds with a trapezoidal array of four cues and five feeders (Fig. 5a). Each feeder was 9 cm from both the nearest cue and feeder(s). Thus, a cue and the two nearest feeders formed the corners of an equilateral triangle. The right-most of the two closest feeders was correct. In the disaggregated (DIS) treatment, I covered the baseline array in the center of the panel and uncovered the cue-feeder triangles (again of side 9 cm) in the four corners of the panel (Fig. 5b). Testing and Data Collection The general training, testing, and data collection procedures were identical to those used in Experiment 2, except that the birds received a minimum of 60 and a maximum of 80 trials each day. As in Experiment 2, birds began the DIS treatment after they achieved 80% performance over their 39 last 30 trials in the baseline treatment. All but one reached this criterion on the first day of testing, and thus began the DIS treatment during their second day of trials. To minimize any effect of an overnight decrement in performance, those that reached criterion on the first day also received 10 trials on the baseline treatment before starting the DIS treatment on the second day. To perform the switch, I entered the experimental room, covered the baseline array and uncovered the DIS array. I then left the experimental room and changed some hardware. The entire switch procedure required about 10 minutes and trials resumed immediately afterward. Results Four birds reached criterion within their first 80 trials in the baseline treatment (Fig. 6a). The other required 138 trials to reach criterion and began the DIS array on his third day of trials. Performance dropped slightly but insignificantly after the switch (Fig. 6a; performance 10 trials before versus after the switch; paired t-test: t=-1.554, n=5, P=0.195; mean performance 30 trials before versus after the switch; paired t-test: t=-2.031, n=5, P=0.083). Two birds dropped from perfect performance to 60% in the first 10 trials in the DIS treatment and another dropped from perfect to 90%. The other two birds, performing at 90% and 100% in their last 10 trials before the switch, experienced no decrement. 40 100 CO > 1 LL "o a> i _ 1 ^ o O CO CD x: o CO o D. CI < CO o c .o o d o 100 80 -60 -40 20 0 (b)  J o -o i Baseline DIS Treatment Figure 6. (a) Mean (±SE; n=5) proportion correct first choices in blocks of 10 trials in baseline and DIS treatments (separated by vertical solid line). Negative numbers indicate trials before the switch, (b) Mean (±SE; n=5) proportion of recorded trials in which birds approached the lit cue in baseline and DIS treatments. 41 The birds approached the lit cue in more of their recorded trials in the DIS treatment, although this trend was not significant (Fig. 6b; proportion of recorded trials in which birds approached the lit cue in baseline versus DIS treatments; paired t-test: t=-1.155, n=5, P=0.312). As in previous experiments, birds who were performing well typically flew to the lit cue and then directly to the rewarding feeder. Discussion Disaggregating the array of cues and feeders did not affect the spatial association abilities of the birds in this study. I found no significant changes in any measure of behaviour, which suggests that hummingbirds did not attend to the overall geometry of the array while performing the spatial association task. As in Experiment 2, the results of this study clearly demonstrate the portability across spatial contexts of spatial associations and the behavioural rules that enable hummingbirds to apply them. Given the inhibiting effects on learning of a high spatial density of cue-feeder pairs (Brown in press), the DIS array (due to the lower spatial density of cue-feeder pairs) may even be easier for hummingbirds to learn than the baseline array. Thus, while I found no effect on performance of changing an aspect of spatial context that altered the spatial density of cues and feeders, it is possible that such a change could affect the rate of learning of the 42 spatial association task. I did not test this exper imenta1ly, however. Nor does Experiment 3 address how the birds encode this new spatial association task. In Experiment 2 I proposed that they formed an abstraction of the lit cue and the rewarding feeder and that their ability to recognize that elements on the panel matched this image and respond appropriately was independent of the spatial context of the task on two different scales. Birds in the DIS treatment could have viewed each "patch" of a cue and two feeders as an array similar to that in baseline, but with fewer elements. The results of Experiment 3, however, provide no evidence as to whether birds attended to only the lit cue and rewarding feeder (as suggested in Experiments 1 and 2) or to the lit cue and both nearest feeders. Thus, in Experiment 4 I asked: how do hummingbirds choose which of the two feeders closest to the lit cue to visit? EXPERIMENT 4 The results of Experiments 1 and 2 suggested that the birds learned a simple behavioural rule to manifest the spatial association: "fly to the feeder nearest to the lit cue". Such a rule was not applicable in Experiment 3 because two feeders were equidistant from the lit cue. What behavioural rule did they use to manifest the spatial association under these new conditions? 43 There are three possibilities, each indicative of a different mental process. If, as in Experiments l and 2, they attend only to the lit cue and its feeder, then the most parsimonious hypothesis is that they fly from the lit cue along a vector at some angle with respect to some global referent to the feeder below. Note that although the spatial association formed in this case is the same as in Experiments 1 and 2, the response-outcome association (i.e. the behavioural rule) is not. Experiment 3 did not exclude the possibility that birds included both nearest feeders in their abstraction, however, and this generates two additional, related possibilities for rules by which they could manifest the spatial association. If birds attend to both nearest feeders, then they could use a rule such as: "fly to the rightmost of the two feeders closest to the lit cue". However, like the vector strategy mentioned above, the concept of "rightmost" also depends on the orientation of the two feeders with respect to some global referent. Under this hypothesis, then, the abstraction that the birds formed in Experiment 3 would include the cue and two closest feeders and be oriented such that the feeders were below the cue. The final possibility is that birds form an abstraction consisting of the cue and two nearest feeders that is not anchored in some larger spatial framework. If this were the case, as in Experiment 2, birds could recognize a set of one cue and two feeders as matching their abstraction regardless 44 of its orientation with respect to a global referent. Pigeons have demonstrated mental rotation abilities (e.g. Neiworth and Rilling 1987) which suggests that they may form abstract representations of elements in their environment that are not anchored in a spatial context. If hummingbirds possessed such abilities, then they could rotate their abstraction to match the orientation of the task at hand (or vice versa), then apply the "rightmost" rule to locate the correct feeder. Thus, while the behavioural rule is the same as under the previous hypothesis, its application would be independent of the orientation of the task with respect to global referents. I distinguished the first two hypotheses from the third by switching hummingbirds from a baseline treatment identical to that in Experiment 3 into one in which each of the four sets of one cue and two feeders was rotated by a different multiple of 90 degrees in addition to being disaggregated. Performance should be significantly disrupted if they cannot perform the mental rotation since the angle of the vector connecting the cue and correct feeder, with respect to the pull of gravity, is different from that in baseline in 3 of 4 cases. After 70 trials in the first rotated treatment, birds began a second in which the triangles were again rotated to distinguish among the first two hypotheses. Birds flying along a vector at a particular angle with respect to a global referent (the first hypothesis above) would have to 45 learn and then anchor four rules for locating the correct feeder (one for each triangle) to four different points in space in the experimental room. By changing the rule required at each of the four corners of the panel, I expected the performance of birds using the above strategy to drop. On the other hand, if the orientation of the cue and feeders was their sole determinant of rule choice (i.e. during the first rotated treatment they had formed 4 abstractions with a different rule for each) they should "follow" the orientations to their new locations and experience no significant decrement in performance. Methods Subjects Six adult rufous hummingbirds (Selasphorus rufus) were used in this study. Four were caught on Sumas Mountain, British Columbia in May and June of 1992. The other two were caught at the same location in May and June of 1993. One bird had participated in Experiments 1 and 2 (2 months earlier), and the other five were experimentally naive at the start of trials. Housing and maintenance were identical to those used in previous experiments. Experimental Environment I used the same experimental rooms as in Experiments 1, 2, and 3. As in Experiment 3, I attached all the feeders and LED holders for the baseline and both disaggregated, 46 rotated treatments to one flat green panel. Those elements not in use in a particular treatment were covered with acetate painted the same colour as the panel and attached to it with double sided tape. The birds gave no indication that they distinguished these covers from the actual surface of the panel. The rest of the hardware was the same as in previous experiments. Unlike in Experiment 3, most trials were recorded on videotape. Most unrecorded trials were at the end of treatments, well after behavioural norms had been established. Videotape analysis was the same as in previous experiments. Experimental Design The first part of this experiment was designed to determine whether birds performing the baseline spatial association task could mentally rotate abstractions of sets of one cue and two feeders in order to apply their behavioural rule. The feature of this strategy that distinguishes it from the other two proposed above is that birds using it should, as in Experiment 2, be able to perform the task regardless of the orientation of the cue and two nearest feeders with respect to some global referent. To test this, birds received a baseline treatment identical to that used in Experiment 3 immediately followed by a disaggregated, rotated treatment (DISROT 1) similar to the DIS treatment of Experiment 3. The "patches" were 47 rotated by 0, 90, 180, and 270 degrees (Fig. 7b) thus creating four distinct orientations of otherwise identical sets of a cue and two feeders. I expected good performance in all orientations if birds were using the mental rotation strategy, but only in orientation A (oriented the same as in baseline) if they were not. The second part of this experiment was designed to test whether the birds learned to fly from the lit cue at a particular angle with respect to some global referent, or whether they chose the "rightmost" of the two feeders closest to the lit cue. The former of these two possibilities implies that they encoded only the lit cue and rewarding feeder into their abstraction. In this case, in order for birds to perform well in DISROT 1, they would have to anchor each of four rules (one for each orientation) to the location of the lit cue on the panel since they would have no other way of determining which rule to use in a given trial. The latter possibility implies that they encoded the lit cue and both nearest feeders into their abstraction. Under this hypothesis, birds in DISROT 1 would have to learn four abstractions and a behavioural rule for each, but the rules would be specific to the orientations themselves, rather than to locations on the panel. After 70 trials in DISROT 1, birds began a treatment (DISROT 2) in which I changed the arrangement of the four different orientations for each bird. I expected birds using rules anchored to a particular place to perform poorly 48 (a) • • • • o o o o o o • c O • A o o o o o o B • D • (b) o o B A o o D • • c O O o o o (c) o o o o Figure 7. (a) Baseline, (b) DISROT 1, and (c) DISROT 2 arrays. Dimensions are identical to Experiment 3 (Fig, 5) . Three birds received the arrangements shown at left and three received those shown at right. Letters identify each orientation (see text). 49 because the vector along which they had learned to fly at that place would lead them to the incorrect feeder, or to no feeder at all. Those using rules specific to the orientations, on the other hand, should continue to perform at pre-switch levels because they could "follow" a particular orientation to its new location on the panel. Testing and Data Collection Training, testing, and data collection procedures were identical to those in Experiment 3. Birds began DISROT 2 after 70 trials in DISROT 1. I assumed that after 70 trials, their behaviour and performance at each orientation would be stable enough that I could detect any changes by comparing cumulative performance at each orientation in DISROT 1 to that in DISROT 2. Results Switch 1: Baseline to DISROT 1 As in Experiment 3, five birds reached criterion on baseline during the first experimental day, and one required an extra day of baseline trials (150 trials total) before being switched to DISROT 1. After the switch, overall performance dropped significantly (Fig. 8a; mean performance 30 trials before versus after the switch; paired t-test: t=-12.538, n=6, P=0.00). Performance varied among orientations, however, so I analyzed performance at each orientation separately. 50 100 CO -4—' c» > -t—' CO LL "o a> i _ o O 80 60 40 20 0 -60 Baseline CO 0) o 03 o Q. < CO c o o CI o CL 100 ]0 60 40 20 -0 Baseline DISROT 1 Treatment Figure 8. (a) Mean (±SE; n=6) proportion correct first choices in blocks of 10 trials in baseline and DISROT 1 treatments (separated by vertical solid line). Negative numbers indicate trials before the first switch. (b) Mean (±SE; n=6) proportion of recorded trials in which birds approached the lit cue in baseline and DISROT 1 treatments. 51 For each orientation, I asked: How motivated were birds to feed in the patch and how well did they perform once there? I hypothesized that if it appeared to the birds that they could apply whatever rule they were using in baseline at a patch with a particular orientation, then they should be well motivated to feed there. A measure of this motivation is the number of times they visited the correct patch the first four times its cue was lit. I also wished to determine if patterns of errors would indicate that they were applying some cognitive process in common in all patches. Hence, I examined the outcome of the first four trials at a patch in which it was correct and was visited before any others (henceforth termed "positive trials"). Four of the six birds went to the correct patch in 3 or more of the first four trials for orientation A ("rightside-up" orientation; refer to Fig. 7b for orientation designations; Fig. 9a). Of the first four positive trials in patches of this orientation, five birds chose the correct feeder 3 or more times (Fig. 9b). Thus, as expected from Experiment 3, birds had little trouble performing the spatial association at the "rightside-up" orientation even though here it was presented simultaneously with three other orientations. Individuals were less motivated to forage at orientation B (the "upside-down" orientation), visiting the correct patch only about half the time, on average, in the first 4 trials in which the cue was lit. Only four birds 52 4 JZ o -4—' CD CL - 1 — ' O CD i _ o O _o CO "co o o 2 1 > 0 A B C Orientation D t5 4 CO Q -o O O CO > O CD 1— O O 2 [-1 h 0 (b) 1 ( -< > < > ( » A B C D Orientation Figure 9. (a) Number of visits to the correct patch out of the first four times the cue there was lit in DISROT 1 of Experiment 4. (b) Number of correct first visits in first four positive trials at each orientation in DISROT 1 of Experiment 4. 53 had four or more positive trials in this orientation, but of those that did, most performed well. The first positive trial of four of the six individuals in this orientation was incorrect; that is, they flew up and right from the cue instead of up and to the left as they should have done in this "upside-down" patch. Birds were more motivated to forage in orientation C ("sideways-right") than in the "upside-down" orientation, with an average of three positive trials out of the first four times its cue was lit. Their performance on positive trials was poor, however; all of the first four positive trials in this patch for four of the six birds were incorrect. Finally, birds were least motivated to visit orientation D ("sideways-left"), generally visiting the correct patch only once or twice in the first four times its cue was lit. Individuals performed at chance (50%), on average, in their first four positive trials in this orientation. In contrast to the clear difference in performance between baseline and DISROT 1, birds approached the lit cue in about the same proportion of trials in both treatments (Fig. 8b; proportion of recorded trials in which birds approached the lit cue in baseline versus DISROT 1 treatments; paired t-test: t=-0.520, n=6, P=0.625). Note that I found a similar trend in Experiment 3 despite the 54 large difference in performance between the DIS and DISROT 1 treatments. Switch 2: DISROT 1 to DISROT 2 After 70 trials in DISROT 1, birds received a new treatment (DISROT 2) in which the orientations were "moved" to a different corner of the panel. Overall performance in DISROT 2 was not significantly different from that in DISROT 1 (Fig. 10a; correct first visits in the 10 trials before versus 10 trials after switch 2; paired t-test: t=1.581, n=6, P=0.175; mean performance in the 30 trials before versus 30 trials after the switch; paired t-test: t=-1.419, n=6, P=0.215). To estimate the portability of the rule(s) the birds were using in the latter stages of DISROT 1, I plotted each birds' performance in the first 10 trials after switch 2 against the last 10 trials before it for each orientation (Fig. 11). Performance in a particular orientation should remain at pre-switch levels or improve in DISROT 2 if the rule the birds used there was portable. Five of the six birds either maintained their pre-switch performance (four birds) or improved after the switch in orientation A. In both orientations B and C, three birds improved, one remained nearly constant and two did worse. Finally, in orientation D only two birds improved after the switch, one remained fairly constant and three performed poorly relative to pre-switch performance. 55 100 CO - » — ' CO > - 1 — ' CO L_ LL "o CD i - _ O O RO 60 40 20 0 - 8 0 DISROT 2 40 60 CO CD JZ o CO o \ Q. Q. < CD c o o CL o CL 100 80 60 -40 20 0 Baseline DISROT 2 Treatment Figure 10. (a) Mean (±SE; n=6) proportion correct first choices in blocks of 10 trials in DISROT 1 and DISROT treatments (separated by vertical solid line). Negative numbers indicate trials before the second switch. (b) Mean (±SE; n=6) proportion of recorded trials in which birds approached the lit cue in baseline and DISROT 2 treatments. 56 o a. Q O 2 4 6 Correct First Visits in DSROT 1 10 o 02 Q o O 2 4 6 8 10 Correct First Visits in DISROT 1 O Q o O 8 X 6 - X 4 - / -II 2 1^ y " / ^ n \^ 1 1 — • 1 1 2 4 6 8 Correct First Visits in DISROT 1 10 o cr Q o O 2 4 6 8 10 Correct First Visits in DISROT 1 Figure 11. Number of correct first visits in the first 10 trials of DISROT 2 versus in the last 10 trials of DISROT 1 for each orientation. Each point represents a pair of observations for one individual. Diagonal line represents perfect transfer of performance between the two treatments; points above the line indicate performance increased in DISROT 2, points below indicate performance dropped in DISROT 2. 57 As in DISROT 1, the birds approached the lit cue in about the same proportion of trials as in baseline (Fig. 10b; proportion of recorded trials in which birds approached the lit cue in baseline versus DISROT 2 treatments; paired t-test: t=-1.607, n=6, P=0.169). Thus, despite the new appearance of the panel, birds typically visited the lit cue before visiting any feeders. Discussion Mental Rotation Abilities This part of the experiment determined whether the huinmingbirds could mentally rotate an abstraction that included the lit cue and both nearest feeders in order to locate rewarding feeders in a spatial association task. The fact that performance dropped significantly under DISROT indicates that they had not been using a rotatable geometric abstraction. This supports the model proposed to explain the results of Experiments 1 and 2 under which birds use a simple distance rule (e.g. "fly to the feeder nearest the lit cue") to locate the correct feeder. Such a rule was not applicable in Experiments 3 and 4 and the rules the birds used in these experiments, unlike those in Experiments 1 and 2, included information contained in the relationship of the cues and feeders to some global referent. They could use this information in two ways: (1) fly from the lit cue at an angle with respect to a global referent, or (2) fly to the rightmost of the two feeders 58 (e.g. with respect to the imaginary line parallel to the pull of gravity between them) nearest the lit cue. The critical difference between these behavioural rules is that the latter implies that birds attended to both of the feeders nearest the lit cue when attempting to manifest the spatial association in Experiments 3 and 4; a process not predicted by traditional associative learning theory, which assumes animals encode only the stimulus and reward (Rescorla 1985). Three results from DISROT 1 lend support to both hypotheses. First, several birds initially stopped approaching the lit cue in orientations B, C, and D while their performance on those orientations was poor. I proposed in Experiment 1 that they do not approach the lit cue unless they are using it as an aid in manifesting their behavioural rule. In the context of DISROT 1, this could indicate that they were learning new ways to manifest the association in each of these patches (i.e. learning three new rules). Second, in orientation D, where there were no feeders to the right of the cue, this could explain the low number of visits to the patch and the apparently random selection of feeders when there. That the birds initially went to the patch at all indicates that they expected to find food somewhere in the patch when the cue was lit and this shows that the association between the lit cue and the availability of food remained intact. Lastly, when foraging in orientation C ("sideways-right"), most birds (4 of 6) 59 were strongly biased to the (incorrect) feeder below and to the right of the cue. These results support the plausibility of both hypotheses but do not distinguish between them. Initial errors in orientation B ("upside-down"), on the other hand, support the hypothesis that they were flying to the rightmost of the two feeders closest to the cue with respect to a global referent. If they were using this rule in baseline, then they should initially apply it in DISROT 1 wherever two feeders in a patch are oriented the same as in baseline (i.e. lying along a horizontal line and equidistant from the lit cue). These requirements are met in orientation B, and birds were initially strongly biased to the rightmost feeder which is the incorrect one in this orientation. The fact that there was a feeder below and to the right of the cue in orientation C could have made it appear more similar to the abstraction that birds formed in baseline and thus could account for their higher motivation to forage there than in orientation B. In addition to the influences mentioned above, learning in other patches may also have biased birds' choices in a particular patch. For example, if a bird learned to find food in orientation B after the first switch (by flying up and to the left of the cue) before finding food in orientation D ("sideways-left"), this might have biased its choice of feeder in that patch to the one above and to the left of the cue. Thus, differences in the number of trials 60 birds took to discover food in a particular patch and the effect of this on their choices in other patches could account for some of the variability among birds in performance in the orientations. Orientations versus Vectors Although birds performed well in orientation A in DISROT 1, their initial performance in orientations B, C, and D was poor. Most, however, eventually discovered the correct feeder in all the patches and achieved consistent performance. Their behaviour suggested that after finding food, they learned new rules for manifesting the spatial association. Were these rules anchored to specific locations on the panel, or to specific orientations? The fact that birds suffered little decrement in performance after the second switch, from DISROT 1 to DISROT 2, maintaining their performance much better than they had after the first switch, strongly suggests that the birds learned orientation specific foraging rules. This implies that they encoded the cue and both nearest feeders into an abstraction for each orientation that was anchored with respect to a global orientation but was portable with respect to a global location referent. In effect, they learned three additional spatial association tasks in DISROT after a latency period in which they attempted to apply the one rule they had learned in baseline. While this model may 61 explain the general trends of performance, it cannot account for two puzzling results. The first is that some birds' performance in some orientations improved dramatically immediately after the second switch, in some cases going from 0% before the switch to 60% (number of correct first visits in blocks of 10 trials) or better in the first 10 trials after it (Fig. 11). Studies of the effect of changing context on latent inhibition may provide some insight into such results. These studies compared subjects' acquisition of a conditioned response (CR) to a CS to which they had previous unrewarded exposure to that of subjects that had no prior exposure to the CS. Subjects that had previous experience with the CS acquired the CS-US association more slowly than control groups (Hall and Honey 1989). Changing the context in which the CS and US were presented for the subjects with previous exposure to the CS ameliorated this latent inhibition effect. Birds could gain unrewarded exposure to the CS in this study by responding to the lit cue in a particular patch (and hence a particular spatial context) but performing poorly there. Changing the spatial context of that orientation in DISROT 2 by moving it to a new location both in the room and relative to other orientations, may have allowed them to quickly learn the spatial association in that orientation in its new location. Another possibility that could reinforce this effect is that the birds remembered previous success at a particular 62 location on the panel. Thus, moving an orientation in DISROT 2 in which a bird had performed poorly in DISROT 1 to a location where it had experienced success previously could have increased its expectation of finding food and hence increased its motivation to look for it. Similarly, previous experience at a particular location may also explain the second puzzling result whereby performance at a particular orientation sometimes dropped greatly in the first 10 trials of DISROT 2. If previous unsuccessful experiences in a location were influencing these individuals' foraging choices, then their errors should reflect that they were avoiding that location. Performance dropped by more than 20% after the switch in five instances (Fig. 11), two of which fit the pattern described above. The other three cases are more difficult to explain. In one case, performance declined in orientation B after it was moved to where C had been. Using the rule for correct performance in orientation C ("up and to the right") in B would lead a bird to the unrewarding feeder and most of the errors in this case were to the wrong feeder in the correct patch. In the remaining two cases, performance declined in orientations that were moved to locations where the bird had been successful previously, and I cannot account for them. The conclusion that the birds attended to both nearest feeders when manifesting the spatial association implies that there may be a 180 degree "window of rotation" in which 63 the angle of the vector connecting the cue and correct feeder with respect to a global referent could be changed without affecting the relationship of the two feeders to each other in a larger spatial framework. This suggestion could be tested by switching birds from a baseline treatment into one in which the sets of one cue and two feeders were rotated from their baseline orientation by less than 90 degrees in each direction. GENERAL DISCUSSION To my knowledge, no studies have specifically tested which aspects of a particular context are important to animals performing spatial association tasks, and in what way. The results from the present experiments suggest that hummingbirds associate the lit cue and rewarding feeder, and that this association is portable across many different spatial contexts. This conclusion is important because it provides evidence that the birds do not learn some simple set of behaviours triggered by the presentation of the lit cue, but rather more abstract associations between cue and feeder that allow more flexible and more generally applicable behavioural responses to changing conditions. Spatial association tasks are somewhat unique in the study of learning because they are an amalgam of two learning processes (associative and instrumental) usually studied separately. This is reflected in the models I proposed to account for the results of the present 64 experiments. I suggested that hummingbirds encode not only the lit cue and rewarding feeder, but also aspects of the spatial context they use when manifesting the spatial association. Thus, depending on the task, the association hummingbirds form between their responses and the outcome of these actions can be context specific because they use information available in the relationship of feeders to some global referent to help them locate the correct one. Such a result demonstrates the danger inherent in broad hypotheses that account for certain abilities across experimental paradigms. 65 CHAPTER 3 Sign-Tracking By Hummingbirds Performing A Spatial Association Task When animals approach stimuli that predict the availability of food this is called sign-tracking (Hearst and Jenkins 1974). "Sign-tracking" is used to describe all kinds of cue directed behaviour, but the phenomenon is typically studied using pigeons or rats in an autoshaping paradigm in which a stimulus is presented for a few seconds and is immediately followed by the delivery of food at a nearby station (e.g. Brown and Jenkins 1968). Under these conditions, individuals of both species come to approach and contact the stimulus when it is presented and during the short interval prior to the delivery of food; that is, they sign-track. Slightly different experimental apparatus are used for each species in deference to their inherent foraging predispositions (Timberlake and Lucas 1989): pigeons come to peck at a keylight lit immediately prior to the delivery of food at an illuminated hopper, and rats come to press a lever inserted into the experimental chamber just prior to food delivery. Such behaviour is interesting because a voluntary skeletal response seems to be acquired in a manner similar to the way reflexive, Pavlovian responses are acquired (Tomie et al. 1989). Prior to the discovery of fairly rapidly acquired sign-tracking, such responses had to be 66 painstakingly conditioned in an instrumental learning paradigm (e.g. Breland and Breland 1961). Explanations for sign-tracking based soley on Pavlovian or instrumental learning theory have failed to fully account for its expression. In a recent review, however, Tomie et al. (1989) suggested that components from both instrumental and Pavlovian learning function in the acquisition and maintenance of sign-tracking. They proposed that the acquisition of sign-tracking is best explained as a Pavlovian conditioned response (CR) to the reliable, positive contingency between the conditioned stimulus (CS) and the unconditioned stimulus (US; usually a food reward). In this view, the CR is a subset of behaviours elicited by the US that is also appropriate to the CS. Since foraging behaviours differ among species (pigeons peck for grain, rats often manipulate food items before ingesting them) and within species with respect to what is being ingested (pecking at food versus drinking water) this explanation accounts for the species specific nature of the CR and its dependence within a species on the nature of the US (Jenkins and Moore 1973; Timberlake and Lucas 1989). They also suggested that the maintenance of sign-tracking, on the other hand, is best explained by treating it as an instrumental response to the CS; a positive association between approaching a stimulus and the outcome of that action (acquiring food). 67 Combining elements from Pavlovian and instrumental learning theory seems to successfully explain exisiting descriptions of sign-tracking, but does it explain cue directed behaviour in other paradigms? Brown (in press) presented hummingbirds with a horizontal array of six cues and a similar array of feeders 3 cm or 11 cm below it, and their task was to associate a lit cue with the rewarding feeder directly below it. Hummingbirds approached lit cues consistently in this spatial association task, although Brown did not quantify the response (G. S. Brown personal communication). Brown used stimuli that reliably predicted not only the concurrent availability of food, but also its location, unlike in traditional autoshaping experiments. Given this non-traditional methodology, can current explanations adequately account for hummingbirds' apparent sign-tracking behaviour in spatial association tasks? Before this question can be answered, the cue directed behaviour of the hummingbirds must be quantified to determine whether it qualifies as sign-tracking. If so, then the conditions under which it is acquired, maintained and extinguished should be similar to those assumed by existing explanations. If these theories cannot fully account for sign-tracking by hummingbirds in the spatial association paradigm, a new explanation will be needed to account for the phenomenon in both traditional and non-traditional paradigms. 68 I performed several experiments similar to Brown's using methods that focused on the portability of spatial associations across spatial contexts. I tested several hypotheses concerning the kinds of information hummingbirds use to perform two different spatial association tasks by suddenly altering the spatial context in which the tasks were performed. Each "switch" eliminated the possibility of using one kind of information and I expected the birds' performance to drop if this change affected their ability to perform the association. During these experiments I quantified the acquisition, maintenance and extinction of cue directed behaviours by hummingbirds. In this chapter, I evaluate them as examples of sign-tracking in the context of other aspects of their behaviour. This will serve as a measure of the general applicability of currently accepted explanations for the phenomenon. Methods General Protocol I performed several experiments to evaluate the portability of spatial associations across spatial contexts (Experiments 1 through 4 in Chapter 2). The general testing procedures were the same for all experiments and are described in detail in Chapter 2, but I shall briefly review them here as they bear directly on the sign-tracking issue. I presented hummingbirds with either four cues (4 mm diameter red LEDs) and five or eight feeders (3 mm holes 69 surrounded by 19 mm florescent orange Avery labels) or six cues and six feeders mounted on a flat green panel on one wall of an experimental room. The separation distances between the cues and feeders ranged from 3 cm to 11 cm in different experiments. Birds received approximiately three hours of training with the panel feeders and lit cues, involving approximately 30 visits to the feeders before the start of an experiment. At the start of each experimental trial, a computer turned a video camera on and then simultaneously sounded a soft buzzer for 0.5 s, designated a feeder at random as rewarding, and lit the cue associated with it. Birds left the perch from 0.01 to 2850 s after the cue was lit (mode approximately 0.5 s), but typically took approximately 1 s to fly from the perch to the panel where they could probe any feeder in any sequence. Whenever birds probed the correct feeder during trials, up to a maximum of 12 times, 2/xl of 20% (w/w) sucrose solution was delivered into it immediately. The cue remained lit until the bird had probed the correct feeder the maximum number of times, perched after probing any feeder, or was away from the array for more than 15 s. Each trial was separated from the next by a two minute intertrial period. If the bird was not perched at the end of this period, the start of the next trial was delayed until it perched. 70 Experimental Manipulations In each experiment, after birds had reached a criterion of 80% correct first visits to the rewarding feeder over their last 30 trials in a baseline treatment designed to allow rapid learning of a spatial association task, I altered the spatial context of the task in a particular way. In Experiment 1 the baseline treatment was identical to one of Brown's (in press) in that the feeder nearest and directly beneath the lit cue was rewarding. Immediately after birds reached criterion in this treatment, I restricted the visibility of the cues by placing them behind the panel, outside the experimental room (Fig. 12). In this treatment, birds could not see the cues without looking at them through the holes in the panel from which they had been removed, and cues were arranged so that all were visible from the perch. For the "switch" treatments in Experiments 2, 3, and 4, I rearranged sets of cues and feeders on the panel (Fig. 12). The spatial association task in the identical baseline treatments of Experiments 3 and 4 differed from that in Experiments 1 and 2 in that there was an extra feeder, the arrays were organized trapezoidally, and the correct feeder was always the rightmost of the two nearest the lit cue (Fig. 12). Video Records I recorded most of most birds' trials on videotape (4 of 6 birds in Experiment 1 and all birds in Experiments 2, 71 o o o o o o Experiment l \ \ 1 / / / ~\~\~i r~~r \ \ \ I I / \ \ \ I I / \ \ 11 / / ^^ \\\i I / W l / / / o o o o o o Experiment 2 • • • • o o o o o Experiment 4 Experiment 3 DISROT 1 o » HEX Disaggregated Figure 12. Baseline treatment arrays for Experiments 1-4, Experiment 1: baseline and restricted visibility arrays. Experiment 2: baseline and hexagonal arrays, Experiment 3: baseline and disaggregated arrays. Experiment 4: DISROT 1 arrays; 3 birds were switched into each arrangement and the baseline treatment was identical to Experiment 3. 72 3, and 4; 3194 of 3401 trials in all). Technical problems resulted in the video record for some individuals being discontinuous and I excluded such records from quantitative analyses in which I examined the time course of sign-tracking. Very few trials for any particular bird were "lost", however, and I included discontinuous video records in analyses of the amount of sign-tracking within individuals in different treatments. Since I was interested in how hummingbirds acquire and use sign-tracking, I included only experimentally naive birds in quantitative analyses. This included all birds in Experiment 1, none in Experiment 2, and all but one in both Experiments 3 (four of five) and 4 (five of six). I analyzed video records on a VCR capable of frame-by-frame advancement of the tape, allowing me to determine the duration of a particular behaviour by multiplying the number of frames in which it occurred by 0.033 s/frame. I considered that a bird "approached a cue" if it hovered in front of or near it before probing any feeder in that trial. I scored a "visit to a feeder" if the bird hovered in front of or probed a feeder. Results Acquisition of Sign-Tracking Experimentally naive birds approached the lit cue in 668 out of 1267 (52.7%) recorded baseline trials in Experiments l, 3, and 4. Typically, they flew from the 73 perch to the panel, hovered briefly in front of the lit cue, and then flew to the rewarding feeder, occasionally stopping to hover briefly en route. Animals in other paradigms often direct appetitive behaviours toward cues (Tomie et al. 1989), but I observed no behaviour that indicated that the hummingbirds were attempting to probe lit cues as if they were feeders, although the resolution of the camera did not allow me to determine whether they attempted to lick it. However, birds seldom approached close enough to contact cues with their tongues. The subjective impression of all observers of videotapes was that birds only looked at cues and made no attempt to manipulate them. Six of nine naive birds with continuous video records consistently sign-tracked only after reaching consistent performance (Fig. 13). After a variable duration of inconsistent but improving performance, birds typically performed correctly for several trials in a row and then began to sign-track in most trials; afterward, however, sign-tracking and good performance were strongly correlated for most birds. Thus baseline conditions induced a large amount of sign-tracking, which was especially consistent when performance in the treatment was high. Extinction of Sign-Tracking Birds sign-tracked in significantly fewer trials in the restricted visibility (RV) treatment of Experiment 1 than in baseline (proportion of recorded trials in which birds 74 Experiment 4 30 60 90 CUMULATIVE CORRECT Figure 13. Cumulative approaches to the lit cue versus cumulative correct first visits in baseline treatments of Experiments 1,3, and 4 for naive individuals with continuous video records. Symbols are plotted every 10 data points and different symbols represent different individuals. The solid line represents perfect correlation between the two measures. (a) Baseline treatment of Experiment 1 (3 individuals). (b) Baseline treatment of Experiment 3 (2 individuals). Note the large the large number of symbols for one individual. (c) Baseline treatment of Experiment 4 (4 individuals). 75 approached the lit cue in baseline versus RV treatments; paired t-test: t=-4.48, n=4, P=0.02). Their performance remained well above chance, albeit at a lower level than in baseline. Immediately after the new treatment began, they altered their approach to the panel. They flew from the perch at approximately the level of the cues (that were visible through holes in the panel), then began their drop to a feeder approximately 0.5 m from the panel without pausing to hover. As in other treatments, I observed no appetitive behaviours directed at the cues; the birds did not probe their bills through any holes in the panel other than feeders. After sign-tracking in nearly every trial toward the end of the baseline treatment, birds initially stopped sign-tracking in orientations B, C, and D in the DISROT 1 treatment of Experiment 4 (Table 2). As in baseline, however, consistent approaches to the lit cue were usually coupled with good performance (Fig. 14); this coupling is striking because of the great variation in performance among orientations. The most common exceptions to this were individuals who consistently sign-tracked for between 3 and 11 of 20 trials in orientations C (2 individuals) and D (3 individuals) despite performing poorly there (Fig. 14). Discussion The spatial association paradigm, despite being different than those in which sign-tracking has been 76 Table 2. Number of trials in which birds approached the lit cue in the last 5 trials of baseline and the first 5 trials in each orientation in DISROT 1. Refer to Figure 12 for orientation designations. bird 8 18 52 56 basel 5 5 5 4 ine A 3 3 5 0 Orientation B 4 3 0 2 C 2 0 3 3 D 2 1 1 3 77 10 20 CUMULATIVE CORRECT 30 10 20 CUMULATIVE CORRECT 30 Figure 14. Cumulative approaches to the lit cue versus cumulative correct first visits for individuals with continuous video records in orientations A, B, C, and D of DISROT 1 (Experiment 4). Symbols are plotted every 4 data points and different symbols represent different individuals. The solid line represents perfect correlation between the two measures. 78 described and interpreted previously, engendered a large amount of sign-tracking by hummingbirds. In this case, stimuli were highly localized in space and predicted both the availability and location of food rather than only the availability of food as in other paradigms. This, coupled with the short CS exposure time relative to the length of the intertrial period, provided conditions well suited to sign-tracking (Tomie et al. 1989). Note that unlike the case in autoshaping paradigms in which food always follows CS presentation, the birds' behaviour determined the outcome of trials in these experiments. That is, they could hover in front of any cue and probe any feeder, but they received food from only one of four, six, or eight feeders whether or not they sign-tracked. Thus, they could (and occasionally did) impose a negative CS-US contingency by visiting the lit cue, then an incorrect feeder, and then perching (thus extinguishing the cue and ending the trial). They could also greatly extend the duration of unrewarded exposure to the CS by staying on the perch or otherwise visiting no feeders until long after the cue was lit. These two occurrences were most often observed while performance was poor and, by current accounts, should inhibit the expression of sign-tracking because they interfere with learning the instrumental response (Tomie et al. 1989). In this view, good performance (i.e. a reliable pairing of the CS and US) would allow the instrumental association between approaching the 79 cue and obtaining reward to form, resulting in the maintenance of consistent sign-tracking. This could be reflected in its expression lagging behind that of good performance in these experiments. Thus, despite being derived from paradigms which employ a fixed positive CS-US contingency, current explanations for the maintenance of sign-tracking provide a plausible framework for interpreting the temporal patterns of sign-tracking exhibited in these experiments. Those explanations also predict that some amount of appetitive behaviour be directed at the cue, but this was not the case in these experiments, however. Birds performing well could have used the buzzer and/or the lit cue as indicators that food was available (or the sound of the camera starting five seconds before the buzzer as an indicator that food would soon be available), but only the lit cue provided information about its location; this is fundamental to the spatial association paradigm. In general, animals do not sign-track towards diffuse auditory cues like the buzzer in my experiments (Tomie et al. 1989). This suggests that the hummingbirds sign-tracked toward a stimulus that they used as an indicator of the location rather than the availability of food. Current theories suggest that in traditional sign-tracking paradigms, the CS predicts the availability of food and that this is reflected in the species specific, anticipatory appetitive nature of sign-tracking (Tomie et al. 1989, Timberlake and Lucas 1989) . If the hummingbirds 80 were responding to the cue as part of their process for locating the correct feeder, then appetitive behaviour directed toward the cue would be inappropriate. Indeed, I observed no behaviour that could be interpreted as appetitive; hummingbirds never probed cues, nor did they ever appear to lick them. In addition, the US was available to them from the time the cue was lit, and they often probed the correct feeder without having approached the lit cue. Furthermore, they could perform well above chance in the restricted visibility treatment of Experiment 1 in which they almost never sign-tracked. These facts together suggest that the US had very little influence on the behaviours expressed while sign-tracking. A more appropriate model for their sign-tracking is one in which birds associate the lit cue with the discovery of food (rather than with food itself). That is, hummingbirds sign-track toward a cue that reliably predicts the location of food, rather than just its availability. Could other paradigms in which just one cue is reliably paired with one response site be considered special cases of the spatial association paradigm? The task in these paradigms could be learned initially as a spatial association in which subjects associate the cue (e.g. one keylight for pigeons) not only with the availability of food, but also with the location of the single food hopper. Despite the fact that the location of the food hopper never changes, it has been reasonable to presume that animals in 81 these paradigms may continue to sign-track because they are not motivated to change this inexpensive behaviour. Recently, however, Silva et al. (1992) found that pigeons will abandon sign-tracking in favour of goal-tracking (waiting by the hopper for the delivery of food) if the keylight and hopper are separated by a large enough distance (that is, if the cost of sign-tracking is high). This cost could have been either the additional time and energy required to sign-track or lost opportunity, because sign-tracking at these distances precluded pigeons from obtaining the reward, which was available only for a short time. If, as hypothesized above, sign-tracking initially functioned in the process of goal location but was no longer needed because the hopper was easy to find did not change location, its extinction could simply be a reflection of the pigeons being motivated to abandon a redundant behaviour. The instances in which sign-tracking was extinguished also support the hypothesis that it is expressed as part of the birds' process of locating the correct feeder. Performance declined in the restricted visibility treatment of Experiment 1, although it remained well above chance, and the birds rarely approached, and never probed, the holes through which they could see the cues. As mentioned previously, the role of the cue as a predictor of the availability of food is not as critical as its role in predicting the location of the reward. That the birds stopped sign-tracking in the restricted visibility treatment 82 could indicate that the lit cue had lost some of its utility in this role; that it was of lower quality compared to that in baseline. In Chapter 2, I argued that the birds used the lit cue to aid in manifesting a behavioural rule such as "fly to the feeder nearest the lit cue". I also argued from the fact that in baseline they often paused briefly to look at the lit cue in transit to the feeder that they were using the lit cue to check their progress from it to the nearest feeder. Restricting its visibility made this impossible and they stopped sign-tracking in favour of the less accurate, but faster strategy of flying directly from the perch to a feeder. The birds could have stopped sign-tracking in this treatment because it was a costly strategy that was no longer a beneficial component of their mechanism for locating the correct feeder. The fact that they also stopped approaching the lit cue in the first few trials in orientations in the DISROT 1 treatment of Experiment 4 that were dissimilar to baseline also supports the above hypothesis. This dissimilarity prevented birds who were using a rule learned in baseline (such as "go to the rightmost of the two feeders nearest the lit cue") from performing well. Thus the temporary cessation of sign-tracking could reflect the birds learning that their old rule was inappropriate in three of four orientations, abandoning it and the sign-tracking that helped them manifest it, then learning three new rules and re-acquiring sign-tracking in these orientations. This 83 process was hindered by sets of elements that appeared similar to baseline, such as the (incorrect) feeder below and to the right of the cue in orientation C. In this case, some birds continued to apply their baseline rule in spite of the fact that it resulted in poor performance. One prediction of the hypothesis that hummingbirds use sign-tracking as an aid to performance is that the easier the task, the less sign-tracking should be observed. Brown (Brown and Gass 1993; Brown in press) found that hummingbirds learned a spatial association task most easily when the separation between the lit cue and the rewarding feeder was small (3 cm), which suggests that the smaller the separation distance, the easier the spatial association task. In the baseline treatment of Experiment 2, the cues and feeders were separated by only 3 cm (cf. 11 cm in Experiment 1, and 9 cm in Experiments 3 and 4) and the birds sign-tracked least in this treatment. However, this comparison is confounded by the fact that all the birds in Experiment 2 had participated in Experiment 1 two months earlier and thus were not naive to the spatial association task. In conclusion, the maintenance of the consistent sign-tracking exhibited by hummingbirds in spatial association tasks can be adequately accounted for by existing theories derived from other paradigms. I suggest, however, that we closely examine the nature of the task when attempting to predict and later explain the behaviours animals exhibit 84 while sign-tracking. More importantly, this study demonstrates the power of theories that, like the spatial association task, relax the traditional distinctions between instrumental and Pavlovian learning and combine components from each. The generality of such theories will be useful as we attempt to apply what we learn about animals' behaviour in the laboratory to their behaviour in their natural habitats. 85 CHAPTER 4 GENERAL CONCLUSIONS In interpreting Experiments 1 and 2 of Chapter 2, I proposed a behavioural mechanism that hummingbirds use to solve a spatial association problem in which feeders are aligned with cues in a simple way. This mechanism, best phrased as a behavioural rule such as "fly to the feeder nearest the lit cue", is independent of the alignment of the cue-feeder pairs with respect to global referents and independent of the continuous visibility of the cues. Any new behavioural model should account not only for the present results, but also for previous ones. Brown (in press, unpublished data) demonstrated the effect of several factors on the learning of the spatial association task. Does the behavioural model mentioned above account for these results also? One prediction of the model is that birds learning such a rule should do so more slowly with an arrangement of cues and feeders in which it is difficult to determine which feeder is nearest to a particular cue. Brown (in press) showed that birds learned more slowly when she reduced the distance between adjacent cue-feeder pairs and/or increased the distance between the cues and feeders. Such treatments should have no effect on birds using a rule based on some global referent, or on using the lit cue as a beacon. In addition. Brown also predicted that if tightly packed and/or widely spaced cue-feeder arrays were 86 interfering with the birds' ability to discern the feeder nearest the lit cue, then providing a visible connection between the two elements should ameliorate this effect, and she found this as well (Brown, unpublished data). Thus the behavioural model I proposed to account for the results of Experiments 1 and 2 also provides a plausible explanation for how all the factors demonstrated to affect the learning of spatial association tasks can act. In Experiments 3 and 4, two feeders were equally close to each cue, precluding the use of the simple rule proposed in Experiments 1 and 2, and under these conditions the birds chose the rewarding feeder with reference to a non task specific global referent. Further investigation showed that they can index behavioural rules by the orientation of these sets of one cue and two feeders with respect to a global referent. These results provide insights into how hummingbirds behave in given situations; especially how they may mentally represent important features in their environment. Such insights are an important first step in understanding how these normally territorial birds "see" their environment, form behavioural rules for living in it, and use these rules in daily life. In Chapter 3 by integrating analysis of sign-tracking (approaching and/or contacting a stimulus that is predictive of the delivery and/or location of food) with performance in the spatial association tasks I was able to provide an explanation for this phenomenon, and to my knowledge this is 87 new. Most work to date has focused on factors that affect the expression of sign-tracking so as to properly classify it as an example of one kind of learning or another (Tomie et al. 1989). By stepping outside of the traditional autoshaping paradigm, I have been able to suggest a utility for sign-tracking and this should stimulate further research into this interesting phenomenon. In this thesis I have revealed possible behavioural mechanisms used by hummingbirds to perform spatial association tasks. The birds learned rules specific to the tasks themselves, but documenting the kinds of rules they learned and the kinds of information they used to learn them provides a basis for proposing rules they might use in other situations. Investigations that explore or model hummingbirds' behaviour will benefit from a clearer understanding of how they are predisposed to solve certain kinds of problems, and I suggest that this conclusion will apply to many other animal species as well. 88 Literature Cited Armstrong, D. P., Gass C. L., & G. D. Sutherland. 1987. Should foragers remember where they've been? Explorations of a simulation model based on the behaviour and energetics of territorial hummingbirds. In: Foraging Behaviour (Ed. by A. C. Kamil, J. R. Krebs, & H. R. Pulliam), pp. 563-586. New York: Plenum Press. Bouton, M. E. & Peck, C. A. 1989. Context effects on conditioning, extinction, and reinstatement in an appetitive conditioning preparation. Anim. Learn. Behav., 17, 188-198. Bouton, M. E., Rosengard, C , Achenblach, G. G., Peck, C. A., & Brooks, D. C. 1993. Effects of contextual conditioning and unconditional stimulus presentation on performance in appetitive conditioning. Q. J. Exp. Psychol., 46B, 63-95. Bowe, C. A. 1984. Spatial relations in animal learning and behavior. Psychol. Rec, 34, 181-209. Breland, K. & Breland, M. 1961. The misbehavior of organisms. Am. Psychol., 16, 681-684. Brown, P. L. & Jenkins, H. M. 1968. Autoshaping of the pigeon's key-peck. J. Exp. Anal. Behav., 11, 1-8. Brown, G. S. 1994. Spatial association learning by rufous hummingbirds (Selasphorus rufus): effects of relative spacing among stimuli. J. Comp. Psychol., in press. Brown, G. S. & Gass, C. L. 1993. Spatial association learning by hummingbirds. Anim. Behav., 46, 487-497. Cheng, K. & Sherry, D. F. 1992. Landmark-based spatial memory in birds (Parus atricappilus and Columbia livia): The use of edges and distances to represent spatial positions. J. Comp. Psychol., 106, 331-341. Colwill, R. M. & Rescorla, R. A. 1986. Associative structures in instrumental learning. In: The Psvchology of Learning and Motivation (Ed. by G. H. Bower), vol. 20, pp. 55-104. New York: Academic Press. Dill, L. 1983. Adaptive flexibility in the foraging behavior of fishes. Can. J. Fish. Aquat. Sci., 40, 398-408. Gallistel, C. R. 1989. Animal cognition: the representation of space, time, and number. Annu. Rev. Psychol., 40, 155-189. 89 Gould, J. L. & Marler, P. 1987. Learning by instinct. Sci. Am., 256, 74-85. Greenberg, R. 1984. Neophobia and the foraging site selection of a neotropical migrant bird: an experimental study. Proc. Nat. Acad. Sci. USA, 81, 3778-3780. Greenberg, R. 1990. Feeding neophobia and ecological plasticity: a test of the hypothesis with captive sparrows. Anim. Behav., 39, 375-379. Hall, G. & Honey, R. C. 1989. Contextual effects in conditioning, latent inhibition, and habituation: associative and retrieval functions of contextual cues. J. Exp. Psychol. Anim. Behav. Processes, 15, 232-241. Hearst, B. & Jenkins, H. M. 1974. Sign tracking: the stimulus-reinforcer relation and directed action. Monograph of the Psychonomic Society, Austin, Texas. Honey, R. C., Wilis, A., & Hall, G. 1990. Context specificity in pigeon autoshaping. Learn. Motiv., 21, 125-136. Janetos, A. C. & Cole, B. J. 1981. Imperfectly optimal animals. Behav. Ecol. Sociobiol., 9, 203-209. Jenkins, H. M. & Moore, B. R. 1973. The form of the autoshaped response with food or water reinforcers. J. Exp. Anal. Behav., 20, 163-181. Johnston, T. D. 1985. Introduction: conceptual issues in the ecological study of learning. In: Issues in the Ecological Studv of Learning (Ed. by T. D. Johnston & A. T. Pietrewicz), pp. 1-24. Hillsdale, New Jersey: Lawrence Earlbaum Associates. Kamil, A. C. & Yoerg, S. I. 1982. Learning and foraging behavior. In: Perspectives in Ethology (Ed. by P. G. Bateson & P. Klopfer), vol. 5, pp. 325-364. Plenum Press. Kamil, A. C , Misthal, R. L., & Stevens, D. W. 1993. Failure of simple optimal foraging models to predict residence time when patch quality is uncertain. Behav. Ecol., 4, 350-363. Kaye, H. & Mackintosh, N. J. 1990. A change in context can enhance performance of an aversive but not of an appetitive conditioned response. Q. J. Exp. Psychol., 42B, 113-134. 90 Krebs, J. R. & McCleery, R. H. 1984. Optimization in behavioural ecology. In: Behavioural Ecology; An Evolutionary Approach (Ed. by J. R. Krebs & N. B. Davies), pp. 91-121. Oxford: Blackwell Scientific Publications. Lovibond, P. F., Preston, G. C., & Mackintosh, N. J. 1984. Context specificity of conditioning, extinction, and latent inhibition. J. Exp. Psychol. Anim. Behav. Processes, 10, 360-375. Miller, D. B. 1985. Methodological issues in the ecological study of learning. In: Issues in the Ecological Study of Learning (Ed. by T. D. Johnston & A. T. Pietrewicz), pp. 73-95. Hillsdale, New Jersey: Lawrence Earlbaum Associates. Morris, R. G. M. 1981. Spatial localisation does not require the presence of local cues. Learn. Motiv., 12, 239-260. Nadel, L. & Winner, J. 1980. Context and conditioning: a place for space. Physiol. Psychol., 8, 218-228. Nadel, L. 1990. Varieties of spatial cognition. In: The Behavioural and Neurological Bases of Higher Cognitive Functions (Ed. by A. Diamond), vol. 608, pp. 613-636. Ann. NY Acad. Sci. Neiworth, J. J. & Rilling, M. E. 1987. A method for studying imagery in animals. J. Exp. Psychol. Anim. Behav. Processes, 13, 203-214. Rabenold, P. R. 1983. The communal roost in black and turkey vultures — an information center? In: Vulture Biology and Management (Ed. by S. R. Wilbur & J. A. Jackson), pp. 135-152. Berkeley: University of California Press. Rescorla, R. A. 1980. Pavlovian Second-Order Conditioning: Studies In Associative Learning. 120 pp. Hillsdale, New Jersey: Lawrence Earlbaum Associates. Rescorla, R. A. 1985. Associationism in animal learning. In: Perspectives on Learning and Memory (Ed. by L. G. Nilsson & T. Archer), pp. 39-61. Hillsdale, New Jersey: Lawrence Earlbaum Associates. Rescorla, R. A. & Colwill, R. M. 1989. Associations with anticipated and obtained outcomes in instrumental learning. Anim. Learn. Behav., 17, 291-303. 91 Rescorla, R. A. & Holland, P. C. 1982. Behavioural studies of associative learning in animals. Annu. Rev. Psychol., 33, 265-308. Shettleworth, S. J. 1989. Animals foraging in the lab: problems and promises. J. Exp. Psychol. Anim. Behav. Processes, 15(1), 81-87. Shettleworth, S. J. 1993. Varieties of learning and memory in animals. J. Exp. Psychol. Anim. Behav. Processes, 19(1), 5-14. Silva, F. J., Silva, K. M., & Pear, J. J. 1992. Sign-versus goal-tracking: effects of conditioned-stimulus-to-unconditioned-stimulus distance. J. Exp. Anal. Behav. 57, 17-31. Spetch, M. L. & Cheng, K. 1992. Landmark use by pigeons in a touch screen spatial search task. Anim. Learn. Behav., 20, 281-292. Timberlake, W. & Lucas, G. A. 1989. Behaviour systems and learning: from misbehaviour to general principles. In: Contemporarv Learning Theories: Instrumental Conditioning Theory and the Impact of Biological Constraints on Learning (Ed. by S. B. Klein & R. R. Mowrer), pp. 237-275. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Tomie, A., Brooks, W., & Zito, B. 1989. Sign-tracking: the search for reward. In: Contemporary Learning Theories; Instrumental Conditioning Theory and the Impact of Biological Constraints on Learning (Ed. by S. B. Klein & R. R. Mowrer), pp. 191-223. Hillsdale, New Jersey: Lawrence Erlbaum Associates. 92 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0087398/manifest

Comment

Related Items