UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Spatial association learning by rufous hummingbirds (Selasphorus rufus) Wilhelmson, Christianne Elly 1999

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

Item Metadata

Download

Media
831-ubc_1999-0239.pdf [ 3.45MB ]
Metadata
JSON: 831-1.0099317.json
JSON-LD: 831-1.0099317-ld.json
RDF/XML (Pretty): 831-1.0099317-rdf.xml
RDF/JSON: 831-1.0099317-rdf.json
Turtle: 831-1.0099317-turtle.txt
N-Triples: 831-1.0099317-rdf-ntriples.txt
Original Record: 831-1.0099317-source.json
Full Text
831-1.0099317-fulltext.txt
Citation
831-1.0099317.ris

Full Text

SPATIAL ASSOCIATION LEARNING B Y RUFOUS HUMMINGBIRDS (Selasphorus rufus) by CHRISTIANNE E L L Y WILHELMSON B A . , University of Ottawa, 1987 B.Sc, Trent University, 1993 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR T H E DEGREE OF MASTER OF SCIENCE in THE F A C U L T Y OF G R A D U A T E STUDIES (Department of Zoology) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April, 1999 © Christianne Elly Wilhelmson, 1999 in p resen t ing this thesis in partial fu l f i lment of the requ i rements for an a d v a n c e d d e g r e e at the Univers i ty of Brit ish C o l u m b i a , I agree that the Library shall m a k e it f reely avai lable f o r re fe rence and study. I further agree that p e r m i s s i o n for ex tens ive c o p y i n g o f this thesis fo r scho lar ly p u r p o s e s may b e granted by the h e a d of my d e p a r t m e n t o r by his o r her representat ives . It is u n d e r s t o o d that c o p y i n g o r p u b l i c a t i o n of this thesis for f inancial ga in shall not be a l l o w e d w i t h o u t my wr i t ten p e r m i s s i o n . D e p a r t m e n t of ~^ZOQIQjy T h e Un ivers i ty of Brit ish C o l u m b i a V a n c o u v e r , C a n a d a D * e J f ) n / /(e. /??<7 D E - 6 (2/88) Abstract Foraging theory provides a framework for understanding why animals make certain foraging decisions yet provides few insights into how these decisions are made. Psychological studies provide understanding of cognitive mechanisms but without the needed ecological context. This thesis continues the study of spatial association memory in rufous hummingbirds (Selasphorus rufus) in an attempt to understand its possible utility in the wild. In laboratory settings, rufous hummingbirds make associations between two spatially separate objects (a cue and a reward/response site) using spatial association memory. In Chapter Two, I investigated the ability of rufous hummingbirds to switch from using spatial memory to spatial association memory when visible cues remained reliable indicators of food location, even when the food location was changed repeatedly. Regardless of how often birds obtained food from the same location, they quickly used spatial association memory to relocate the reward when it was moved. Birds initially used spatial memory to locate the reward but once this failed they used the available light cue to learn the spatial association. Once birds were using spatial association memory, the spatial memory of the rewarding location was not reinforced in memory as they relied only on the association to locate the reward. In Chapter Three, I found evidence that the components of spatial association memory (the cue and the reward/response site) can be utilized separately from the association itself. Once birds learned a spatial association, a spatial memory of the currently and previously rewarding location remained in the absence of the light cue and this spatial memory was strengthened when birds fed repeatedly from the same location. ii This supports Brown (1992) who suggested that hummingbirds make spatial associations according to Gestalt theory, in which both the association and its parts are presumed to be perceived and remembered. I uncovered this information through analysis of intertrial activity, previously presumed to be independent of within trial behavior. Spatial association memory is more complex and flexible than expected and has characteristics to be useful in the wild, possibly to adjust quickly to profitability changes in flowers and patches. iii Table of Contents Abstract ii Table of Contents iv List of Tables v List of Figures vi Acknowledgements vii Chapter One: General Introduction 1 Chapter Two: The effects of repeated foraging experience on spatial association memory 6 Methods 10 Subjects 10 Experimental Room 10 Training 12 Experimental Procedures 13 Treatments 14 Performance 15 Results and Discussion 15 Initial use of spatial memory 15 No effect of reinforcing a single location as rewarding on foraging success 18 What precipitated the rapid switch to using spatial association? 20 Alternation of spatial memory and spatial association memory 24 Overall performance 28 Effects of intertrial periods 30 Conclusion 32 Chapter Three: Relevance of unrewarded behaviors to spatial association memory 34 Methods 36 Results and Discussion 38 Continued foraging flights during intertrial periods 38 Intertrial foraging: decreased success locating rewarding feeders 41 Focus on currently and previously rewarding locations 48 Effect of intertrial activity on within-trial performance 57 Conclusion 59 Chapter Four: General Conclusions: 61 Literature cited 65 iv List of Tables Table 1. Location of errors after the first switch in each treatment on Day 1 (n=4 for each treatment) 17 Table 2. Number of birds going to currently and previously rewarding location after each of the first 5 switches (first 6 trials) in Treatment 1 on Day 1 21 v List of Figures Figure 1. Experimental panel used in all treatments. Lights (small circles, top row) were 3 cm above each feeder (larger circles, bottom row) and the feeders were 11 cm apart. The background was dark green, and around each feeder was a red label. The L E D light was the stimulus and the feeder was both the response site and the reward site. The filled circle and light rays indicated a lit LED. Which cue-feeder pair was active varied as described in the text 11 Figure 2. Number of birds committing an error in the first trial after a switch on Day 1. Includes only first three switches 16 Figure 3. Mean number of trials needed to relocate the reward after the switch. Includes only birds that committed an error on first trial after switch 19 Figure 4. Mean number of error committed by birds in first 110 trials on Day 1. * = first time location of reward was changed. Error bars are 1 SE 29 Figure 5. Mean number of foraging flights per inter-trial period on Day 1. Treatments 1 and 30, n = 4; Treatments 10 and 20, n = 3. Error bars are +1 SE 39 Figure 6. Location of first probe within and between trials in relation to currently and previously rewarding location. Filled circles indicate performance within trials. Open circles indicate visits to the currently rewarding location between trials. Squares indicate visits to the previously rewarding locations between trials. Intertrial averages are based on those trials in each block where birds engaged in foraging flights 42 Figure 7 Proportion of visits to currently rewarding location in relation to time from trial's end (light off) to first probe in all treatments 45 Figure 8. Mean proportion of first visits to currently and all previously rewarding locations between trials. Treatment 10 = 169 trials; Treatment 20 and 30 = 119 trials. Error bars = 1 SE 50 Figure 9. Location of errors between trials in relation to currently rewarding location ( = 0). Each distance change represents one feeder location ( + = to the right of currently rewarding feeder, - = to the left of currently rewarding feeder) .... 52 Figure 10. Mean proportion of first probes at currently rewarding location in first 10 intertrials after switch (all switches). Treatment = 10 switches; Treatment 20 = 50 switches; Treatment 30 = 3 switches. Error bars = 1 SE 54 Figure 11. Mean proportion of first probes at previously rewarding location in first 10 intertrials after switch (all switches). Treatment = 10 switches; Treatment 20 = 50 switches; Treatment 30 = 3 switches. Error bars = 1 SE 55 vi Acknowledgements I thank my supervisor, Lee Gass, for challenging me to discover my strengths and for his genuine commitment to my success. I thank my supervisory committee: Michael Whitlock, David Shackleton and Don Wilkie for their encouragement and enthusiastic participation in all phases of my work. I sincerely thank past and present Vivarium inhabitants including: Gayle Brown, James Thompson, Doug Armstrong, Gordon Mclntyre, Marc Roberts, Janet Moore, Lara Chatters and Beth Shields. They offered invaluable experience and insight for which I am very grateful. I would especially like to thank Beth for rescuing me from my lab solitude and giving me a wonderful, fresh perspective on everything! I thank the staff of the Zoology Computing Unit (ZCU) including Lance Bailey, Alistair Blachford and Jens Haeuesser; their help was invaluable and allowed me to turn my ideas into reality. I thank Don Brandys for his patient help with our aging hardware and to Bruce Gillespie for all his assistance. I sincerely thank Ernest Keeley for all his help with my statistical analysis. Thanks to all my friends for their unyielding faith and support. In particular, I want to thank Krista De Groot, Bea Beisner and Heather Ferguson; I wouldn't have succeeded without their friendship. I would also like to thank my family. My brother, Michael and sister-in-law Verena, for their friendship and love. My parents, Peter and Anita, for their unwavering support, love and faith; thanks for my roots and my wings! And finally, I want to thank 'my' birds for teaching me patience; sometimes the most interesting discoveries come when you just sit still. vii CHAPTER ONE GENERAL INTRODUCTION Rufous hummingbirds (Selasphorus rufus) are proficient foragers able to locate rewarding flowers and return to them as long as they remain profitable (Gass & Sutherland 1985, Miller et al. 1985, Calder 1993). A very high metabolism and an energetically expensive mode of flight mean they need to feed as often as every 4-5 minutes, depending on the environmental conditions (Gass 1978, Gill 1987). This extraordinary task is complicated by floral species composition that varies in space and time, with different species flowering at different times on the same territory during a season (Gass et al. 1976, Gass 1979). It is also complicated by complex patterns of nectar availability resulting from interactions between production by plants (Hixon & Carpenter 1988) and removal by the forager and its competitors (Wolf 1978, Gill & Wolf 1979, Paton & Carpenter 1984, Armstrong et al. 1987). Therefore, individuals must continually assess their food supply both spatially and temporally. A rewarding site in the morning may not be rewarding in the afternoon and returning to a recently depleted site is energetically costly. This is critical at all times but especially when birds arrive with depleted fat stores at stopover sites during migration (Carpenter et al. 1981, Carpenter et al. 1983). Here, they must quickly assess the distribution and quality of the flowers while competing with others of their own species for territories and with other species for food (Gass et al. 1976, Kodric-Brown & Brown 1978, Gass 1979, Hixon et al. 1983). Hummingbirds successfully forage under these circumstances by relying on strong, flexible learning and memory abilities that allow them to acquire and retain information and adjust quickly when conditions change. Mobile animals can obtain useful information from associating elements in their environment to reward locations using spatial memory (Gallistel 1989, 1990). Spatial 1 memory is well-studied in birds that cache or hoard their food, such as corvids and parids (Balda & Kamil 1988, Herz et al. 1994, Clayton & Krebs 1994, Bednekoff & Balda 1996). It is also seen in female bats that locate their young in large colonies (McCracken 1993), in bees that locate their nests (Robinson & Dyer 1993) and in foraging hummingbirds (Cole et al. 1982, Gass & Sutherland 1985, Mcintyre 1995, Healy & Hurley 1995, Hurley 1996). Spatial memory is an involved process because animals must learn both relevant landmarks and reward locations and the spatial relationships among them. Spatial memory is considered to be specific to one set of landmarks or a location and not to be transferable to others. However, an animal can create a sort of map while foraging by learning relationships between many objects in its environment. It can then use spatial memory to forage freely and successfully within the area of the map as long as the reward locations remain constant. Spatial memory is not strong in all animals because, among other things, it has limited utility in changing environments. As long as the information remains stable in the spatial distribution of its food, an animal can continue to successfully use the associations it has made to specific locations. However, if the information used to remember a location changes or the location is no longer profitable, an animal must relearn new information to forage successfully. Animals can also learn and remember associations between two or more spatially separate objects (a cue and reward) using spatial association memory (Moore 1973, Bowe 1984, Brown & Gass 1993, Brown 1994, Thompson 1994, Mcintyre 1995). For example, in a lab, primates associated a rewarding foodwell with a pointer located anywhere from 1 - 6 inches (2.54 - 15.24 cm) away (Davis 1974). Because the visual cue and the food are separate in space in this paradigm, this task requires animals to integrate several kinds of information: the location of the food, the visual cue that signals availability of the food, and 2 the spatial relationship between them. Spatial association memory differs from spatial memory because the association learned is not anchored to a particular place (Brown & Gass 1993). Once learned, it is portable and can be used in any location that provides the necessary information. Rufous hummingbirds can use spatial memory and spatial association memory to forage. They use spatial memory to locate food based on information such as the complexity of the spatial distribution of their food (Sutherland & Gass 1995) or edges of rewarding patches (Mclntyre 1995). These different patterns may be used at different times, for example edges in the morning (Paton & Carpenter 1984), because birds themselves are affecting the distribution of the profitable flowers through their own foraging. In the lab, hummingbirds have associated a lit red L E D light with a reward 10 cm away (Brown 1992, Brown & Gass 1993). In the field, they might be able to use spatial association to learn to visit flowers preferentially by correlating quality or quantity of nectar to a visible floral trait (Melendez-Ackerman 1997a,b), such as petal pattern, which is in a different spatial location (for e.g., the pattern is on the tip of the petal while the food is in the receptacle of the flower). Spatial association memory might be more useful than spatial memory when the location of food changes. Changes can occur through shifts in nectar production in plants, depletion by residents and intruders or shifts in territory boundaries. If a location is no longer rewarding, the animal has only to transfer the association it has learned to a new location (Mclntyre 1995). Spatial association memory has been suggested as the mechanism behind the ability of food storing birds to relocate their caches (Shettleworth 1985). Mobile animals, such as bees (Gori 1989) and vultures (Houston 1983) can also obtain useful information from associating elements in their environment that are spatially separate. 3 Spatial association memory may be an intermediate mechanism used when a food location becomes unprofitable and spatial memory is no longer reliable. Recent research has begun to identify underlying characteristics of spatial association learning and memory in rufous hummingbirds. Brown (1992) corroborated, using hummingbirds, the conclusions from rat, pigeon and primate studies that speed and accuracy of learning spatial associations decreases with increasing distance between the cue and the reward. She also showed that performance decreases with decreasing distance between the reward sites, demonstrating that spatial association memory is sensitive to contextual information beyond the cue-feeder pair and its spatial relationship. Thompson (1994) found that spatial association memory was independent of global referents, such as edges of a panel, and that birds did not use the geometry of feeders on a panel to help associate a visual cue and a reward. He also found that this association was not anchored to or aided by the particular spatial location of the reward. In principle, by using a general mechanism such as spatial association memory, hummingbirds have a portable and flexible tool allowing adaptation to changing environments (Thompson 1994). However, Mcintyre (1995) found that increasing foraging experience at one location resulted in greater reliance on spatial memory and decreased success locating the rewarding feeder if its location changed. This effect was weaker in the presence of a reliable visual cue using spatial association memory but was still evident. The ability to quickly learn and remember information is important, but holding on to this information too long (i.e. not being able to put it aside and re-learn quickly when things change) would be counter-productive as habitats vary from year to year and from day to day. Many questions still remain concerning the use of spatial memory and spatial association memory in changing environments. In Chapter Two, I investigated the ability of 4 rufous hummingbirds to switch from using spatial memory to spatial association memory when visible cues remained reliable indicators of food location even when the food location was changed repeatedly. My goal was to replicate in a changing environment Mclntyre's findings (1995) of greater reliance on spatial memory when the association between food and a location was reinforced repeatedly. In Chapter Three, I investigated whether hummingbird behavior between trials can reveal information about or even influence the spatial association memory tasks they learn and demonstrate within trials. 5 CHAPTER TWO THE EFFECTS OF REPEATED FORAGING EXPERIENCE ON SPATIAL ASSOCIATION MEMORY Just as the ability to learn and remember depends on both neurological capacity and foraging ecology, so should the ability to use given mechanisms in given situations. In order to navigate and orient through their environments, mobile animals use spatially based mechanisms such as spatial memory. With spatial memory, they learn and remember spatial relationships between multiple elements, including landmarks, celestial cues, and goal locations, and use this information to guide their actions. Research on birds (Balda & Kamil 1988, Sutherland & Gass 1995, Bednekoff & Balda 1996, Hurley & Healy 1996), ants (Werhner et al. 1996), bees (Robinson & Dyer 1993), bats (McCracken 1993) and fish (Warburton 1990) has revealed many uses of spatial memory, including foraging, territoriality and returning to young and nests. Though the phenomenon has not been well studied in the wild, animals can also use spatial association memory to obtain food. With spatial association memory, animals learn the relationship between cues such as the color or shape of objects and spatially separate reward locations (Bowe 1984, Brown & Gass 1993). Unlike associative memory, where the association is between cues and rewards that are contiguous spatially or temporally, spatial association memory incorporates a spatial component into what is being learned. Both spatial memory and spatial association memory allow animals to make good decisions with the types of information available. It is clear that when both spatial information and visual cues are available, many animals rely on spatial memory first and use types of associative memory only when the former proves unreliable (Menzel 1985, Clayton & Krebs 1994, Brodbeck 1994, Brodbeck & Shettleworth 1995, Hurley & Healy 1996, Garber & Paciulli 1997). However, it is not clear 6 that animals could switch as easily between spatial memory and spatial association memory. Spatial association memory differs from associative memory as it is a combination of spatial memory and an associative task. It requires learning to associate going to a location with obtaining a reward (associative) and learning the location of that reward in relation to the cue (spatial). Because spatial memory and spatial association memory each have a spatial component, it is possible that they compete in some way, perhaps during memory formation (Mackintosh 1983, Clayton & Krebs 1994), resulting in reduced or delayed success foraging. However, if animals could use spatial memory or spatial association memory with equal success, they might use spatial association memory when spatial memory fails. Therefore, spatial association memory's utility in the wild may depend on the state of the environment. If animals can switch between these mechanisms easily, then they could use spatial association memory to experience continued success even when profitability in their environment changed. This flexibility may require animals to learn simultaneously all information necessary to use both mechanisms successfully, regardless of which mechanism they are using at the time. Rufous hummingbirds (Selasphorus rufus) can use spatial memory (Sutherland & Gass 1995, Healy & Hurley 1995, Hurley & Healy 1996) to avoid previously visited flowers (Calder 1993) or return to profitable flowers or patches. They might also be able to use spatial association memory (Brown 1992, Brown & Gass 1993, Mclntyre 1995) for the same purposes. A territorial hummingbird can experience periods of stability in its environment during the time that it holds its territory because plants are sessile and the location of food remains constant during that time. Its territory may be bound by the spatial relationships between elements such as trees and rocks, whose spatial location remains constant. These remembered relationships then create a type of mental map by which the bird navigates. 7 Therefore, during this period, birds should have a well developed ability to learn and remember places, including the location of food (Wolf 1969), and successfully forage using spatial memory (Vander Wall 1982, Shettleworth 1983) or spatial association memory. However, within that map profitable locations can and do change through shifts in nectar production by plants (Hixon & Carpenter 1988), depletion of the nectar by intruders (Gill & Wolf 1979), the resident's own foraging (Wolf 1978, Paton & Carpenter 1984, Armstrong et al. 1987) and shifts in territory boundaries, which can change daily (Gass 1978). Because it is a portable and flexible mechanism (Brown & Gass 1993, Mcintyre 1995), spatial association memory could be used when the location of food changes. By 'portable' I mean that if a location is no longer rewarding, an animal can successfully use the association it has learned in a different physical location; even one it has never been to before. For example, if a bird's territory contains two differently colored flowers and it is familiar with both, then when one flower is no longer producing a profitable amount of nectar, the bird could use the color of the second flower to quickly find new profitable locations (Heinrich 1979). Even if hummingbirds can use either spatial memory or spatial association memory and switch between them, their success finding food using either can be affected by the distance between visual cues and rewards (Brown 1993), the complexity of the spatial distribution of their food (Sutherland & Gass 1995) and how frequently an association between a particular location and food has been reinforced. Mcintyre (1995) tested the latter effect when birds could use only spatial memory (no cues for association were available) and when they could use both spatial memory and spatial association memory. The longer a bird had fed from one location, the longer it took to stop using it when it was no longer rewarding and switch to another, suggesting that birds had come to rely primarily on spatial memory to find the reward. This effect was stronger when birds had no access to visual cues and were 8 using only spatial memory. The effect was also magnified the longer they received food from any one location. These results indicate that spatial association memory allowed them to adjust to the change in the reward's location more readily than when they could use only spatial memory. However, Mcintyre (1995) changed the location of the reward only once in each of his treatments. Therefore, he could not determine whether spatial memory would be equally favoured in environments of differing stability (when the location of food changed at different frequencies). The objective of my study was to test the ability of rufous hummingbirds to switch from using spatial memory to using spatial association memory when visible cues remained a reliable indicator of where to find food, but the location of the reward was changed repeatedly. I tested birds in four treatments in which food remained at the same location for different periods of time. As with associative memory, I predicted that spatial association memory would be used only when spatial memory alone proved unreliable. I expected birds in the most unstable environment (in which profitability shifted most frequently) to switch to spatial association memory quickly and continue to use it throughout the experiment. I also expected birds in the most stable environment to rely on spatial memory for longer before switching to spatial association memory, thus 'paying for' their reliance on spatial memory when the spatial structure of their environment broke down. Once birds learned to associate food with a single location, I predicted that the longer this association had been repeatedly reinforced, the more likely a sudden change in the location of the reward and its visual cue would result in difficulty relocating the reward. Finally, I expected birds to switch between spatial memory and spatial association memory, using the latter only when the location of the reward initially changed and returning to using spatial memory when the location remained stable for some extended period of time. 9 M E T H O D S Subjects I used 16 wild-caught rufous hummingbirds (Selasphorus rufus; 5 adult females, 9 adult males, 1 juvenile male and 1 juvenile female). The birds were caught in May and June 1997 at three sites in the lower mainland of British Columbia: Sumas Mountain near Abbotsford, Nairn Falls Park near Pemberton, and Rosewall Park north of Qualicum Beach on Vancouver Island. Ten birds had no experience with experiments in our lab and seven had experienced a similar experiment in June and July 1997. All birds were housed at the Small Mammal Unit of the University of British Columbia Animal Care facility and were transferred to the Zoology Vivarium prior to training for the experiment and remained there until they had completed it. They were housed in individual cages (60 x 60 x 60 cm) in a controlled photoperiod environment (14 L: 10 D). They were fed ad libitum Nektar Plus (Nekton USA), a commercial hummingbird nectar, supplemented with soy protein (Vege Fuel by Twin Lab, 7% w/w) during the entire period at the Vivarium except during experimental trials, when they were fed a 20% sucrose solution. Experimental Room I placed birds individually in one of two experimental rooms, similar to those described in Brown & Gass (1993) and Thompson (1994). At one end of the room (1.2 m wide x 2.5 m long x 2.5 m high) was a dark green aluminum panel (106 cm x 27 cm), whose base was 131 cm from the floor. The panel housed eight red L E D lights (5 mm diameter) and eight feeders (a 3 mm hole with a red circular label surrounding it). Each light was 3 cm directly above a feeder and feeders were 11 cm apart in a horizontal row (Fig. 1). Each room had a 141 cm high "telescoping" perch (3.5 cm wide) approximately 180 cm from the panel, 10 o o o o o o o © © 0 © © 0 © © Figure 1: Experimental panel used in all treatments. Lights (small circles, top row) were 3 cm above each feeder (larger circles, bottom row), and the feeders were 11 cm apart. The background was dark green and around each feeder was a red label. The LED light was the stimulus, and the feeder was both the response site and the reward site. The filled circle and light rays indicate a lit LED. Which cue-feeder pair was active varied as described in the text. 11 placing the bird at eye level with the center of the row of feeders. Each room was lit with three bare 40W incandescent bulbs: one above the perch and two above the panel. A computer monitored and controlled all hardware, including photocells, solenoid valves (General Valve Corporation, Series 3), lights and buzzers. It recorded and responded to arrivals and departures at the feeder panel and perch. Each feeder was monitored by a photocell (attached to the back of the panel) which transmitted 'arrival' and 'departure' signals to the computer. Each feeder was supplied with sucrose solution from a central reservoir via a solenoid valve. When a bird probed the feeder designated as rewarding, 2 ul of sucrose solution was released within 10 ms. Training I used a training protocol modified from Brown (1992), Thompson (1994), Chatters (1996) and Moore (1997). Healthy birds (> 3.5 g) were first cage trained. I placed feeders, covered by a flat green plastic mask and with a red circle around the spout, against the cage wall to accustom birds to feeding near a wall (see Wells 1993b). These colors simulated those of the experimental panel. Within a few days of successfully feeding from the masked feeder, I moved the bird to an experimental room with the panel covered. Typically, hummingbirds hover near the ceiling when first released into a new room. To speed up how quickly they perched and to reduce the stress of the transfer, I extended the perch to its full height, with the perching area widened. A few minutes after the bird first landed, I began gradually to narrow the perch, lower it to its final height and lower the feeder until it was directly in front of the covered panel. I then removed the feeder, exposed one light-feeder pair near the center of the panel, and illuminated the light. When the bird had fed successfully (several foraging flights of at least three or four probes) from the panel, I slowly moved the perch back from the 12 experimental panel to its final position in the room (180 cm from the panel wall). I then exposed the entire array and moved the location of the reward one position to the right on the panel. When the bird had successfully located and fed (again, several foraging flights of at least three or four probes) from this feeder, I covered the panel, returned the cage feeder, and left the bird in the room overnight. Between 0600 - 0700 the next morning (Day 1 of the experiment) I uncovered the panel and ran a training program. I illuminated a light cue and allowed the bird four feeding bouts (2-12 probes per bout) at the feeder below it. I repeated this at each of the locations in a particular complex sequence. This gave the bird experience feeding from all locations and introduced it to the utility of the light cue. Once this training was completed (mean time 2 hours), I began the experiment. Experimental Procedures This was a 2 day trial based experiment, based on Brown (1992) and Thompson (1994). During each trial, only one feeder (the rewarding feeder) provided food when probed. Once the bird was perched, a soft buzzer sounded to get its attention and the trial began. An L E D light was illuminated above a randomly pre-selected rewarding location 0.5 seconds later. Birds were free to fly and probe feeders any time but received food only during trials and only at the feeder cued by the light. The light was turned off and the trial ended if: a) the bird had probed the rewarding location a total of 12 times, not necessarily consecutively; b) the bird returned to the perch after probing any feeder; or c) the bird probed any feeder then neither returned to the perch nor probed another feeder for 15 seconds. This period was defined as a trial whether or not the bird obtained food. Once the light was turned off, a 2 13 minute intertrial period began. This was done in an attempt to keep the bird motivated by providing sufficient food to maintain itself but not allowing it to feed to satiation. The sequence of rewarding locations was determined prior to the experiment. Using numbers to represent the eight feeder locations, I used a random number generator to create a sequence of 32 digits with each feeder location represented four times, with the constraint that locations were never repeated twice in a row. The sequence of rewarding locations was identical for all birds. If a bird completed the sequence of 32 during the test, I used the same series again. Birds remained in the experimental room with the panel covered and its cage feeder in place for the duration of training and the experiment (three days: one day training, two days experiment). Treatments I used four birds in each of four treatments differing only in the number of successive trials on which the reward remained at the same location. In Treatment 1, the location of the reward changed with each trial to generate a baseline measure of performance and behavior under conditions that permit no spatial memory. In the three other treatments, the rewarding location remained the same for some number of successive trials before it changed to a new position (Treatment 10: 10 trials; Treatment 20: 20 trials; Treatment 30: 30 trials). Each day of the experiment ended with the completion of a series of trials (1, 10, 20 or 30 trials, depending on the treatment), as no maximum or minimum number of trials was set a priori for either day. On each day, the experiment was run as long as the bird continued to respond to trials and feed. Treatments were fully randomized and experimental rooms were equally represented in each treatment. Each bird was used in only one treatment. At least half of the birds in each treatment were naive to all experimental procedures. 14 Performance My measure of performance was based on the first feeder probed each trial. This provided the clearest indications of the type of information birds were using to locate the reward. For this analysis, I ignored subsequent probes at any feeder, though birds could probe any feeder and could receive food (to a maximum of twelve probes) if they located the reward before the end of the trial. I defined first probes at rewarding locations as 'correct'. First probes at any other feeder were 'errors'. Chance performance was one correct trial in eight. RESULTS AND DISCUSSION Initial use of spatial memory As expected, rufous hummingbirds in this study used spatial memory before spatial association memory, even in the presence of a reliable visual cue. All birds in Treatments 1 and 20 and two of four birds in Treatments 10 and 30 erred in the first trial after the location of the reward was changed (first switch) (Fig. 2). This indicates that they had not yet learned the spatial association needed to locate the reward. Of the twelve birds that erred at the first switch, nine went to the previously rewarding location and one went one feeder away from it (Table 1). The probability that nine birds would err at the previously rewarding location by chance is extremely low (binomial probability analysis: p (9) = 0.0000). The location of their first probe indicates birds had been using spatial memory, not spatial association memory, to find the reward. These results corroborate other studies that found food-storing birds relied on spatial information before visual cues such as color and pattern (Brodbeck 1994, Clayton & Krebs 1994) to make foraging decisions. It also corroborates Healy & Hurley (1996) who found that rufous hummingbirds relied on location before any visual cues, such as color, to relocate 15 W TJ • Treatment 1 • Treatment 10 • Treatment 20 • Treatment 30 1st 2nd 3rd Switch in location of rewarding feeder and visual cue Figure 2: Number of birds committing an error in the first trial after switches on Day 1. Includes only first three switches. 16 Table 1: Location of errors after the first switch in each treatment on Day 1 (n = 4 for each treatment) Treatment Birds making error At old Closer to old Closer to new Inconclusive Treatment 1 4 3 0 0 1 Treatment 10 2 1 1 0 0 Treatment 20 4 3 0 1 0 Treatment 30 2 2 0 0 0 TOTAL 12 9 1 1 1 17 rewarding flowers. Mobile animals that are known to require good spatial memory for navigation, territoriality and foraging include bees (Menzel 1985), nutcrackers (Balda & Kamil 1992), black-capped chickadees (Brodbeck 1994, Brodbeck & Shettleworth 1995), marsh tits and jays (Clayton & Krebs 1994), primates (Garber & Paciulli 1997), and pigs (Mendl et al. 1997). Rufous hummingbirds also require strong spatial memory to help them avoid previously visited flowers or return to profitable flowers or patches (Gass & Sutherland 1985). All mobile animals use spatial information to move through their environments, however, these animals have evolved the ability to combine their perception of their spatial environment with the ability to remember spatial relationships. One reason they might have learned to rely on spatial information before other kinds is because unlike more object-specific cues, it is less likely to change over time (Brodbeck 1994). No effect of reinforcing a single location as rewarding on foraging success In contrast to evidence that reliance on spatial memory leads to errors when conditions change, my birds switched quickly from spatial memory to spatial association memory, avoiding these errors. Birds in this study began using spatial association memory quickly regardless of how often they had obtained a reward from a single location. Most birds that erred after the first switch relocated the reward within only a few trials, with little or no searching (Fig. 3). By the time the reward was switched for the second time, only three of the sixteen birds erred, two in Treatment 1 and one in Treatment 20. This was a significant decrease (contingency analysis, p < 0.001) from the first switch, indicating birds were now using spatial association memory rather than spatial memory. Three birds in Treatment 1 continued to commit errors after several more switches, with most probes being near the currently rewarding location. This also indicated a shift towards using spatial association 18 6 5 4 w .2 3 2 + 1 0 a Treatment 10 • Treatment 20 rj Treatment 30 1st 2nd 3rd Change in location of rewarding feeder and visual cue Figure 3: Mean number of trials needed to relocate the reward after the switch. This includes only birds that committed an error on the first trial after a switch. Only the first three switches on Day 1 are reported. 19 memory. However, two of the three birds that erred after the second switch went to the previously rewarding location, indicating that they were still using primarily spatial memory. Only one of the sixteen birds (in Treatment 1) made an error after the third switch and it was closer to the rewarding location, indicating that all birds had switched to using spatial association memory to locate the reward. Only one bird in Treatment 10 erred in response to subsequent switches on Day 1, erring at each of the fourth through the seventh switches. What precipitated the rapid switch to using spatial association memory? There are several possible explanations for birds switching rapidly from spatial memory to spatial association memory in this experiment. One is that the switch was precipitated by the experience of failing to obtain food using spatial memory. The impact of not obtaining food using spatial memory may have acted in the same way as other types of rapid learning or one-trial learning, such as taste aversion. Animals in taste aversion studies learn to associate eating a particular type of food with subsequent illness, often after only one negative experience (Hintzman 1978). Once the association is made, they will avoid eating that food even without further negative reinforcement. Birds here may have quickly associated using spatial memory with not finding the reward, then switched to spatial association memory. Most birds found the reward within a few trials after the first switch and few birds made errors in response to subsequent switches, regardless of treatment (Figs 2-3). How one unrewarding experience changed behavior is most evident in Treatment 1. At the first switch (after trial 1), all 4 birds committed an error. Three erred at the previously rewarding location, four feeder positions to the left of the reward (Table 2) indicating a reliance on spatial memory. The fourth bird's error was inconclusive, two feeder positions to the left of the reward. After the second switch, two birds erred with only one going to the 20 Table 2: Number of birds going to currently and previously rewarding location after each of the first 5 switches (first 6 trials) in Treatment 1 on Day 1. Swi tch B i rd s mak ing an er ror B i rd s g o i n g to p rev i ou s l y reward ing locat ion F e e d e r p o s i t i o n s away f r o m cur rent l y reward ing loca t ion B i rd s g o i n g to cur rent ly reward ing locat ion 1 (trial 2) 4 3 4 0 2 (trial 3) 2 1 5 2 3 (trial 4) 1 0 na 3 4 (trial 5) 2 1 2 2 5 (trial 6) 2 0 na 2 21 previously rewarding location. Although some birds continued to make occasional errors afterwards, most were at or near the currently rewarding location indicating a complete switch to spatial association memory by only the fifth switch (Table 2). The failure to obtain a reward likely acted as a signal to switch from spatial memory to spatial association memory. If this were true, it would indicate that recent reliability of a mechanism was more important than how long it had been reliable. This result is consistent with other studies comparing spatial and associative memory where animals quickly stopped visiting locations when they became unrewarding (Clayton & Krebs 1994, Brodbeck 1994). The simple spatial association memory problem presented to the birds may also have contributed to the swift learning of the spatial association. Associating a cue and reward becomes more difficult with increasing distance between the two. Rufous hummingbirds can make this association when either 1.57 cm or 10 cm separate cue and reward but have greater difficulty with the latter distance (Brown 1992). In my study, only 3 cm separated the cue and reward making it a simple problem. If the association had been more difficult to make, it is possible that it would have taken the birds longer to switch to spatial association memory. A more difficult problem would have resulted in more errors being committed as the birds tried to learn the association. This would have likely resulted in a longer transition period and a longer reliance on spatial memory. It is also possible that the transition to using spatial association memory was made easier because birds had used spatial memory successfully while still having visual access to the light cue. Birds began using spatial association memory so quickly that it seemed they had already learned the association between the light and the reward before ever actually using it. After making an error on the first probe of the trial after the first switch, birds in Treatments 10 and 30 took a mean of 1.5 and 2 trials, respectively, to relocate the reward on 22 the first probe. Once birds stopped going to the old rewarding location, they went directly to the new location of the reward on the next trial. Birds in Treatment 20 took slightly longer, a mean of 3.3 trials (Fig. 3), but the difference between treatments was not significant (Kruskall-Wallis, H c = 0.5122, p » 0.05). The latter mean was skewed by the performance of one bird that took 8 trials; the other three took 1 or 2 trials. It is possible that the quick switch to spatial association memory when spatial memory failed was because while birds seemed to be relying only on spatial memory to find the reward, they were 'registering' both spatial relationships and visual cues. Clayton & Krebs (1994) also found that marsh tits and jays switched quickly between spatial memory and associative memory, with no decrease in performance or searching at unrewarding sites. In both studies, it indicates that using spatial memory does not interfere with using associative memory, but rather that these animals are processing all stimuli present and not using selective attention (processing certain stimuli while ignoring others; Kahneman 1973, Klosterhalfen et al. 1978, Wickens 1984, Anderson 1985). Animals might be able to do this if spatial tasks such as spatial memory and associative tasks (which could include spatial association memory) are processed in different parts of the hippocampus. Hippocampal aspiration in black-capped chickadees resulted in reduced success locating a reward using spatial memory but not when using visual cues (Sherry & Vaccarino 1989). Processing different types of information in different parts of the brain might reduce or eliminate conflict between mechanisms, increasing the likelihood of using them simultaneously. Also, if spatial memory and spatial association memory are automatic, requiring no attention on the part of the individual (Keele 1973), they could function effectively at the same time if they are being used for the same overall action (Gass 1985). This would also allow for processing of different types of information simultaneously. When a hummingbird arrives at a new patch, it 23 might be processing both spatial information (the location of flower and other objects) and local information (color or patterns on a flower) at the same time. Eventually, it might associate a particular color with a profitable flower and then use this information to make associations without relying solely on spatial information. This would support the hypothesis that animals, including chipmunks and other mammals use spatial memory, visual cues and random exploration in a very integrated way when foraging (Vander Wall 1991). The ease of switching to spatial association memory regardless of treatment may also be because what we call "spatial association memory" has two components: spatial memory and associative memory. Increasing experience feeding at one location can affect spatial memory (Mclntyre 1995). However, in spatial association memory, the stronger element is the association between the light cue and food and not between a location and food, as in spatial memory. Because the association between the light and food is not anchored to a single location, the location is not reinforced. Therefore, increasing experience feeding at any location would not affect the birds' ability to abandon that site when it became unrewarding. Birds had access to the light cue at all times and if they were processing its significance on some level, even though they were using spatial memory, location itself would never have been strongly reinforced. Alternation of spatial memory and spatial association memory This study provides some evidence that hummingbirds can return to using spatial memory once they have begun to learn a spatial association. However, they did not alternate between mechanisms for as many trials as I expected them to. I speculated that the utility of spatial association memory in the wild is in locating new food sources when the nectar in a current source became unprofitable. Though optimal diet theory predicts that one food type 24 will either always or never be taken upon encounter (zero-one rule, reviewed in Stephens & Krebs 1986), partial preference is likely for many animals including hummingbirds (see Krebs & McCleery 1984 for reasons). In a hummingbird territory, the profitability of flowers varies from full to completely empty. This profitability can be affected by, among other things, the bird's own foraging during the day and seasonal changes in the nectar production of different floral species. Given this, hummingbirds should be more likely to use daily experience at different flowers, within and between species, to decide where to forage rather than a simple 'all or nothing' rule. In general, they learn and remember the most profitable locations and where they have already foraged that day to make their 'most optimal choices' at each moment. In such a complex situation, an animal would benefit from being able to use several mechanisms to help find and remember the most profitable locations. Birds could use spatial association memory as profitability changes over time and use spatial memory as long as the profitability of going to particular flowers or patches remains stable. In this study, even though location did provide reliable information about the reward's current location in the trials between switches, once birds fully learned the spatial association, most used spatial association memory for the duration of the experiment (both days). They likely did not return to using only spatial memory when the location of the reward remained the same because once they used spatial association memory consistently, it never failed to result in finding the reward. If it had led to errors, birds might have returned to using spatial memory because the map of their foraging environment had not changed since they had last used it successfully. The birds' reliance on spatial association memory once learned is further supported by a pilot test on 2 birds at the end of Day 2. After they had completed their second day of the experiment (in Treatment 10 and 30, respectively), I ran 5 extra trials changing the location of 25 the cue and reward with each trial (as in Treatment 1). The bird in Treatment 10 went to the reward first on each trial, while the bird in Treatment 30 went to the reward 4 times of 5, making an error only on the first trial. The location of the error was 3 feeders away from the reward and I cannot conclude from it whether the bird was using spatial memory or spatial association memory. Even when the reward remained at the same location for a consistent number of trials between switches over 2 days, birds continued to find it after an unpredictable alteration in the tempo of change. This indicated they were still relying only on spatial association memory. The behavior of one bird in Treatment 10 also indicates that at least some birds can switch back and forth between spatial memory and spatial association memory after learning the spatial association. This bird successfully relocated the reward after switches 2 and 3 but then made an error on switches 4 through 7. On switch 2, the reward was moved by 55 cm (5 feeder places) from its previous location, while on switch 3 it was moved by 22 cm (2 feeder places). Locating the reward by chance in two consecutive switches under these circumstances is unlikely (probability = 1.5%). Therefore I'm confident the bird was using spatial association memory then. If the bird had in fact been using spatial memory, we would have expected errors near the initial location of the reward but this was not the case. However, the location of the errors in later switches indicated that this bird switched back to spatial memory. Three of the four errors it committed after switches 4 though 7 were at or near the old location of the reward (the fourth was neither near the old nor new location of the reward). Also, the bird took a mean of three trials to relocate the reward after these four switches and on two occasions after locating the reward this bird revisited the old location of the reward indicating no clear use of spatial association memory. 26 It is possible that after switch 3, something put the reliability of the visual cue in doubt. If spatial memory was abandoned when it became unreliable, the same could apply to spatial association memory. I can't say with any confidence what might have placed the reliability of the light cue in doubt. But a momentary failure of a pump to supply food when the feeder was probed could, in principle, have caused it. When this happens, a bird will quickly start avoiding that location (Thompson 1994, pers. obs.). However, all indications are that the equipment performed properly, and there is no evidence that this bird in this study failed to get food at rewarding feeders, as it did not avoid any one location. Also, birds sometimes probe repeatedly at an unrewarding feeder during one bout, but they rarely do this on consecutive trials (pers. obs.). That this bird returned to rewarding locations and probed multiple times on each trial indicates it was obtaining food during these trials. The behavior of this bird indicates not only that birds can switch quickly from spatial memory to spatial association memory, but that they can also quickly switch back. It also indicates that the use of either spatial association memory or spatial memory is influenced by its most recent reliability rather than how long it had been reliable earlier. If this is true, then an experiment in which the significance of the light was unclear (for example, sometimes it meant food was available and sometimes that it wasn't available) should result in birds abandoning spatial association memory more quickly. Finally, there is some evidence that birds can rely on both mechanisms while learning the spatial association. After locating the reward successfully, three birds in Treatment 10 and two in Treatment 20 probed at the previously rewarding location first in subsequent trials. After one or two trials of this unrewarding behavior they focused on the new location of the reward. This indicated that although they were now using spatial association memory, they had not yet stopped applying the information learned while using only spatial memory. I 27 observed no birds in Treatment 30 behaving this way. Since I did not observe this behavior after later switches, this indicated that as the experiment progressed, birds had completely switched to using spatial association memory. Overall performance Because birds switched so quickly from spatial memory to spatial association memory, I did not find the expected difference in performance between treatments. In fact, overall performance was high from the beginning of the experiment in all treatments. Only five of the sixteen birds made more than four errors in the first ten trials of Day 1 in any treatment (Fig. 4). After trial 50, few birds made any errors at all (Fig. 4). There was no significant difference between treatments in the number of errors committed, except between trials 11 -20 (residuals indicate homogeneity of variances - repeated measures A N O V A using 10 trial blocks as the repeated measure: F 3 , . 5 = 3.80, p = 0.0398) and 21 - 30 (F 3 ,15 = 3.80, p = 0.0307); (Fig. 4). The significant difference between Treatments 1 and 20 between trials 11 -20 (Tukey Test: q ) 2 , 4 = 5.74) was due to the high performance in Treatment 20. No birds in this treatment erred between trials 11-20. Similarly, the significant difference between Treatment 20 and 30 in trials 21 - 30 (Tukey Test: q 12,4 = 4.5) was due to the very low performance in Treatment 20. All birds in this treatment except one committed between 6-8 errors in trials 21-30, which coincided with the first switch in location of the reward. Nor were there any differences between treatments in the rate of improvement in the first 110 trials (comparison of slopes in Fig. 4; ANCOVA, p > 0.21). This was also the case in the first 50 trials when most birds were still committing errors (ANCOVA, p > 0.18). However, performance in Treatment 20 was clearly unique between trials 21-30 (Fig. 4). In 28 </> o UJ Treatment 1 Treatment 10 Treatment 20 Treatment 30 - - - - C h a n c e 1-10 11-20 21-30 31-40 41-50 51-60 61-70 71-80 81-90 91-100 101-110 Trials in blocks of ten Figure 4: Mean number of errors committed by birds in first 110 trials on Day 1. * = first time location of reward changed for each treatment. Error bars = +1 SE. 29 contrast with Treatments 1,10 and 30, where the number of errors progressively decreased, in Treatment 20 there was a significant increase in errors between trials 21-30 (polynomial analysis; p = 0.0321) followed by a significant decrease in errors (p = 0.035) between trials 31 - 40. This indicates that the switch decreased the bird's effectiveness at locating the reward on the first probe of a trial using spatial association memory. Why Treatment 20 birds were nearly perfect (mean 19.5 correct out of 20 trials) in the first series of trials is unclear. However, there is no reason to believe that initial success locating the reward was related to the treatment itself. Until the first switch, the rule for all birds was the same: food is at one location and the location is cued by light. If high performance had been related to treatment, then we would expect even higher performance level in Treatment 30 birds, which was not the case. Because there was no indication during training that they would be the 'fastest learners', the reason those birds were so successful is likely chance. The quick learners just happened to be assigned to the same treatment. Of the 16 birds in this study, 11 were consistently locating the reward within 5 trials (i.e. they were 'fast-learners'). We would then expect that within each treatment, 2.75 birds should have been 'fast-learners'. But because birds were randomly assigned to a treatment, four of these birds were assigned by chance to Treatment 20. By the third switch, all birds in other treatments were doing as well as those in Treatment 20. Effects of intertrial periods I believe that the results of this study did not replicate those of Mclntyre (1995) because of differences in experimental design. My experiment was structured around trials during which birds could obtain food, separated by 2 minute intertrial periods when they could not, although birds could fly to the panel and probe feeders at any time. The duration of 30 any trial depended on how quickly the bird responded at the beginning of a trial and how long it remained at the panel, but the duration of the intertrial periods was the same for all birds. Mclntyre, on the other hand, allowed birds to obtain food a maximum of 3 to 30 times during a period of 30 to 300 minutes, depending on the treatment, with no defined intertrial period. His birds could always obtain a reward at one of the feeders when they flew to the panel. Perhaps the birds' activities during intertrial periods in my study account for the difference in our results. It is possible that during this time my birds obtained information about the reliability of a location. Mclntyre's birds never experienced not receiving food at a 'rewarding' location. My birds did, since a location could be rewarding before and after an intertrial period yet unrewarding during. Therefore, a bird could learn that 'this location is sometimes reliable' in my experiment but not in Mclntyre's. If the birds were starting to process the association between the light cue and reward, as well as getting conflicting information on the reliability of spatial memory, this may have been why they stopped using it so quickly regardless of how long a location had been rewarding during trials. It is also possible that the intertrial period increased hunger in my birds. They could not satiate themselves on one foraging trip because they could obtain only a maximum of 24 of of 20% sucrose solution within any trial. This maximum food consumption and the birds flight activity during intertrial periods when they could not get food resulted in birds remaining motivated to look for food until late in the day. It is possible that hunger influenced birds to stop using spatial memory more quickly when it failed to locate the reward, as they were highly motivated to find food. If this were so, the effect of an increasing intertrial time could be tested with several more difficult spatial association problems. The prediction is that increasing hunger would motivate birds to switch even more quickly from spatial memory to spatial association memory. 31 C O N C L U S I O N The results of this study indicate that rufous hummingbirds, like many mobile animals, rely on spatial information before visual cues when making decisions. It also provides evidence that they can switch quickly to other mechanisms, such as spatial association memory, if required. There is also some evidence that birds can return to using spatial memory after learning spatial association memory. Therefore the switch to spatial association memory in this study was likely "temporary", only lasting until the birds had a reason to stop using it but in this experiment they never did. I cannot conclude unequivocally why the switch from spatial memory to spatial association memory occurred so quickly. However, one or a combination of the failure to obtain food using spatial memory, the easy spatial association memory problem, and having access to the light cue when using spatial memory may have contributed. The latter could be clarified with a study requiring birds to switch between two distinct tasks (spatial memory and spatial association memory). In the current study the visual cue was always a reliable source of information and birds were likely learning this even while they still were using spatial memory. After spatial memory failed after early shifts in profitability, birds switched to spatial association memory and continued to use it as it never failed. If birds had to learn to locate the reward using only spatial memory, with no light cue present, how quickly would they learn to relocate the reward if the light cue were suddenly introduced? And what would happen if birds had to switch back and forth between these two tasks. It might provide a clearer answer to the questions raised in this study. Rufous hummingbirds exhibit the flexibility to switch back and forth between two cognitive mechanisms. Without means to adjust, sudden change can be detrimental to an animal's survival. If a food source is rewarding for a period of time but then its quality changes, the ability to quickly avoid this location and use other information to locate food is 32 useful. Hummingbirds abandon locations quickly when they no longer reward them, because if it takes them many negative experiences to abandon a location, it could result in a high net energy loss. In general, the ability to use many types of environmental information to locate new food source quickly should maintain successful and efficient foraging. With increasing evidence that rufous hummingbirds can efficiently perform spatial association memory tasks, we now require studies that look at spatial association memory in the wild. This would not only provide more information about the mechanism itself but also more importantly provide evidence for how and why this species has evolved the ability to use it. This study also provides some indication that intertrial activity might influence the mechanisms and behavior within trials. I examine this more closely in Chapter Three. 33 CHAPTER THREE RELEVANCE OF UNREWARDED BEHAVIORS TO SPATIAL ASSOCIATION MEMORY In the study reported in Chapter Two, I examined the performance of rufous hummingbirds (Selasphorus rufus) during trials when a light cued the availability and location of food on a feeder array. In that experiment, trials lasted a mean of 14.4 seconds (± 0.7) and were separated by 2 minute intertrial periods when no cue was lit nor food available but birds were still free to probe feeders. Since the computer recorded continuously, I studied behavior only in the 12% of the time I considered most directly relevant to my hypothesis. This "bout-structured" experimental design is common in psychology and ethology, where we often focus on the specific activities predicted by our hypotheses during short, structured periods of time (D'Amato 1970). We often ignore behavior outside these periods because it seems extraneous and can be difficult to relate to the mechanisms we are studying. In classical conditioning experiments, for example, if the animal remains in the experimental environment, it is standard to exclude behavior between trials from the analysis because the conditioned and unconditioned stimuli (cue and reward) are not present then. However, since we impose the experimental structure on the study animal, it is unclear whether we can reasonably assume that all relevant components of the behavior being studied occur only within the confines of trials. Learning is the acquisition of visual, auditory and other information (Hintzman 1978, Carlson 1993) as well as changes in behavior that occur through experience. How information is acquired can affect what changes occur. For example, temporal or spatial separations between elements can affect how easily animals learn associations (Polidora & Fletcher 1964, Rescorla 1980, Mackintosh 1983, Yaginuma & Iwai 1986, Brown & Gass 34 1993). There are three parts in a learned association: the cue, the reward site and the response site. However, it is the separation between the cue and the response site that influences how well the association is learned (Gibbon et al. 1988, Pinel et al. 1986). Another result of this separation may be that animals also learn and remember the stimulus and reward/response site components used in the association. One way to determine whether the memory of the components is maintained after the association is learned is to examine parts of the experiment described in Chapter Two when one of the elements is not present; that is, between trials. The objective of this study is to determine whether what birds do between trials could reveal information about or even influence what they learn within trials. Since the majority of activity occurred outside trial periods, it seemed plausible to hypothesize that some of it might be relevant. In our laboratory, we have always recorded and referred to between-trial activity but no one has ever analyzed it quantitatively. I knew that birds continued to probe feeders between trials in my experiment, even though at no time during these periods did they ever receive food. I focused my analysis on the frequency of foraging flights and the location of first probes at the feeder array between trials. Hummingbirds learn quickly to forage where food is available and avoid unprofitable locations (Gass & Montgomerie 1981). Therefore, I predicted that when they learned that food was available only at a location cued by a light, they would also learn that if there was no light cue anywhere, there was no food anywhere either. Given this, I expected that the number of foraging flights during the intertrial period would decrease towards zero as success during trials increased towards perfection. I also predicted that since birds were using spatial association memory to locate the reward during trials, the location of the first probes would be random in the absence of the light. Finally, I predicted that because birds would learn the 35 structure of the experiment, within-trial behavior would not be influenced by between-trial activity and vice-versa. METHODS Data were collected as described in Chapter Two. This was a 2 day trial-based experiment, based on Brown (1992) and Thompson (1994). During each trial, only one feeder out of 8 (the rewarding feeder) provided food when probed. Once the bird was perched, a soft buzzer sounded to get its attention and the trial began. An L E D light was illuminated above a randomly pre-selected rewarding location 0.5 s later. Birds were free to fly and probe any of the feeders any time but received food only during trials and only at the feeder cued by the light. The light was turned off and the trial ended if: a) the bird had probed the rewarding location a total of 12 times, not necessarily consecutively; b) the bird returned to the perch after probing any feeder; or c) the bird probed any feeder then neither returned to the perch nor probed another feeder for 15 seconds. This period was defined as a trial whether or not the bird obtained food. Once the light was turned off, the 2 minute intertrial period began. It ended when a buzzer sounded and the light was illuminated 2 minutes later. During this experiment, all departures and arrivals of 16 rufous hummingbirds at a perch and at any of 8 feeders within a feeder array were monitored continuously by a computer. I used four treatments with the experimental variable being the number of successive trials on which the reward remained at the same location. Within trials, a red light cue signaled both the availability and location of a reward. In Treatment 1, the location of the reward changed with each trial to generate a baseline measure of performance and behavior under conditions that permit no spatial memory. In the three other treatments, the rewarding location remained the same for some number of successive trials before it changed to a new 36 position (Treatment 10: 10 trials; Treatment 20: 20 trials; Treatment 30: 30 trials). I set no maximum or minimum number of trials a priori. In Chapter Two, I examined activity only within trials. Here, I examine data collected during the same experiment but focus on activity during the 2 minute intertrial periods when no cue was present or food available during Day 1. I had used four birds in each treatment but due to mechanical difficulties, I excluded two birds from this analysis (one each from Treatments 10 and 20). Since trials began only when a bird probed at the panel after the light was lit, I excluded any activity between the ends of intertrial periods and the beginnings of trials. I defined "activity" as a bird leaving then returning to the perch. I defined a "foraging flight" as a bird leaving the perch, probing at least one feeder, then returning to the perch, even though the attempt never resulted in obtaining food. I defined a "non-foraging flight" as a bird leaving then returning to the perch without probing a feeder. By definition, birds could engage in only one foraging flight and no non-foraging flights within trials. However, between trials they could engage in multiple foraging and non-foraging flights. To indicate what information birds were using, I considered only the location of the first probe of the first foraging flight when the intertrial period began. I defined the rewarding location during the current series of trials as "currently rewarding" and defined the rewarding location in the previous series of trials as "previously rewarding". To explore any changes in the number of foraging flights over the course of the day, I included all foraging flights during the intertrial periods in my analysis. 37 RESULTS AND DISCUSSION Continued foraging flights during intertrial periods Contrary to expectation, the birds' behavior indicated that they behaved no differently during intertrial periods than during trials. They continued to engage in foraging flights between trials throughout Day 1. The mean number of foraging flights per intertrial, averaged over all intertrials, ranged from 1.1 + 0.06 in Treatment 10 to 1.9 ± 0.08 in Treatment 1. This is comparable to within trials where by definition, only one foraging flight could be made during any trial. Individual birds made between 0 and 11 foraging flights per intertrial, though the numbers varied extensively in all treatments (Fig. 5). Also contrary to my expectations, the mean number of foraging flights per intertrial remained statistically constant throughout the day in all treatments (repeated measures A N O V A , first and last 30 trials as blocks, p > 0.05 in each treatment). Assuming that the sole motivation for foraging flights between trials was to get food, the continued foraging effort indicates that birds never learned that absence of a light cue signals absence of food, as predicted. Whether a light was lit or not, birds continued to fly to the panel between trials, likely expecting to obtain food using spatial and visual information that had been useful only moments ago (the time between the light being extinguished at trial's end and the first intertrial foraging flight, the iag time', ranged anywhere from 1 to 120 seconds but the mode was only 21 seconds and the mean was 55.8 seconds; see below). One reason birds did not associate intertrials with not obtaining food could be that the failure to get food between trials might not have been costly enough for this aspect of the protocol to be worth learning. We often assume that animals behave optimally by never expending more energy than they intake. However, animals can behave in a way not 38 SZ O ) o> c o> ra o TTreatment 1 —I—I—I—I—I—I—I—I—I—I— 10 20 30 4 0 50 60 70 80 90 100 110 120 130 . c Treatment 10 0 10 20 30 4 0 50 60 70 80 90 100 110 120 130 140 150 160 Treatment 20 10 20 30 40 50 60 70 80 90 100 110 Treatm ent 30 a t c a 10 20 30 4 0 50 60 70 Intertrials 80 90 100 110 Figure 5: Mean number of foraging flights per inter-trial period on Day 1. In Treatments 1 and 30, n = 4; In Treatments 10 and 20, n = 3. Error bars are +1 SE. 39 considered optimal and the currency of 'optimal behavior' may vary with viewpoint (Gass 1985, Roberts 1991). For example, behavior that is seen as suboptimal might be an animal gathering information, and its utility may be delayed for many reasons (Oaten 1977, Lima 1984). If non-optimal behaviors can persist, then a behavior that costs little but might provide a benefit would likely not be abandoned quickly. I had assumed that birds would learn when food was available (during trials) and when it was not (between trials) because the lit cue visually differentiated these time periods. However, this presumed a level of cognition of the experimental structure that was not essential to the birds' survival in this environment. Energetically, it was not necessary that they learn that flying to the panel during intertrial periods would never result in getting food, as they obtained enough food during trials to sustain these short but unrewarded flights. If the food had been unavailable for a longer period of time, if the cost of foraging then had been greater or if the outcome of the foraging effort during trials had been less rewarding, it is possible that the imposed structure of the experiment might have become relevant to the birds, and between-trial foraging trips ceased. Suboptimal behavior, which will persist if its cost is low, can interfere with learning (Reebs 1993). Since there was no clear indication of recognition of the difference between and within trials, it is possible that birds' behavior in intertrials was similar to or influenced by what they learned within trials. Birds may have continued to fly to the panel between trials because one of the two tasks they had learned and integrated within trials when using spatial association could still be performed at trial's end. Within trials, birds learned to fly to a feeder below a light in response to it being lit and a buzzer sounding (time from light on until the first feeder probed, mode = 2 seconds in all treatments). Responding to the light is a classical conditioning response. A conditioned stimulus (a cue) presented during trials elicits a conditioned response 40 (fly to panel) because it is associated with an unconditioned stimulus (food); (Carlson 1993). The second task birds learned quickly was that to obtain a reward, they had to probe the particular feeder cued by the light. This is an operant conditioning response. A particular behavior is performed (probe a particular feeder) immediately preceding an event (obtaining food) (Carlson 1993). If birds remembered these two components separately, they may have been able to perform them independently of each other. For example, birds could fly to the panel in the absence of the light cue between trials because they had learned that flying to the panel and probing a feeder was necessary to obtain food. As the light cue was not part of the operantly conditioned response, its absence would not hinder the memory of this task. Therefore, birds could still fly to the panel with the expectation of obtaining food, but with a reduced likelihood of locating the rewarding feeder. Intertrial foraging: decreased success locating rewarding feeders Birds located the currently rewarding location between trials (i.e. the feeder that was cued by the light and provided food in the current series) even when they couldn't use spatial association memory, but with less success than within trials. Birds located the correct feeder on the first probe of trials in 89% - 93% of all trials, depending on treatment (Chapter 2). However, when no light cue was available between trials, only 22% - 50% of their first probes were at the currently rewarding location, depending on the treatment. Though this is a much higher success level than expected by chance, this was a large drop in performance relative to within trials (Fig. 6). In Treatments 1 and 30, there was a significant drop in performance in all but the first series (residuals indicate homogeneity of variance - repeated measures A N O V A using 10 trial blocks as the repeated measure: Treatment 10, 1 s t series F 2 , n = 1.23, p> 0.05; Treatment 30 1 s t series F 2 , n = 1.74, p> 0.05). In Treatment 10, there was a 41 Treatment 1 • • • ,o o -•- • • B - . • -^••1 — I 5 6 7 8 9 10 Blocks of 10 trials i i 12 13 14 Treatment 10 f - l - ' l I " I F ^ - h 7 8 9 10 1 1 Blocks of 10 trials — | — - f - - — i i -12 13 ' 14 15 16 17 Treatment 20 .2 Z 0 8 + S £ 0.6 + 9- O 0.4 0.2 0 2 o 3 4 Blocks of 20 trials A Treatment 30 # rtion * * 0.8 -u £ 0.6 -9 . - -O . . Propo O 0.4 -° 0 . 2 -0 -1 • " " f " — 1 2 3 Blocks of 30 trials 1 1 4 Figure 6: Location of first probe within and between trials in relation to currently and previously rewarding location. Filled circles indicate performance within trials. Open circles indicate visits to the currently rewarding location between trials. Squares indicate visits to the previously rewarding locations between trials. Solid line indicates chance level (0.125). Intertrial averages are based on those trials in each block in which birds engaged in foraging flights. 42 significant drop in performance in all but the first two series (repeated measures A N O V A using 10 trial blocks as the repeated measure: 1st series Fi,s = 1.42, 2 n d series F^s = 4.70, p> 0.05). In Treatment 20, there was a significant drop in performance, however, it was in the 1st and 3r d series, but not in the 2 n d , when birds' within-trial performance was poor (repeated measures A N O V A using 10 trial blocks as the repeated measure: 2 n d series F i 6 = 4.04, p> 0.05). The consistent drop in performance from within to between trials indicates that the light cue was necessary for locating the reward consistently (in > 89% of trials) but was not necessary for locating it at a level far above chance (in > 22% of trials; chance = 12.5%). These results also indicate that the proportion of intertrials on which birds probed at the currently rewarding location varied with treatment (Fig. 6). One reason for this drop might be because birds could no longer use the light cue to confirm their target destination during the flight. There is strong evidence that birds decide where to probe first before leaving their perch (Brown 1992, Thompson 1994, unpublished data) but use the light cue to reconfirm their target (Thompson 1994). Between trials in this experiment, birds could not confirm their target and so were less accurate than within trials. Thompson (1994) found that after birds learned a spatial association problem, their performance decreased significantly but remained well above chance when the visibility of the light cue was restricted upon departure from the perch. He hypothesized that birds were using the light cue not only to make and use the spatial association but also to confirm their destination as they flew towards the panel. Comparably, Polidora & Fletcher (1964) hypothesized that the ability to sample a stimulus prior to initiating an instrumental response increases the ability to make an association. In Thompson's study, birds had access to the light cue while on the perch, but not while flying to the panel. Birds in my study did not have access to the light at any time between trials. However, in both cases, without access to the 43 cue to confirm their destination en route, birds might have relied on their spatial memory of where they last saw the light to make the association. That birds located the currently rewarding location with equal frequency whether the light had gone out 120 seconds beforehand or had just been extinguished (Fig. 7) indicates that two minutes was not long enough for that memory to decay. However, the decrease in success locating the most recently rewarding feeder may be because the memory of the light's location was weak, resulting in difficulty using the association. The reduction in success probing the most recently rewarding locations between trials might also be related to the mechanism used to locate the reward within trials. When hummingbirds learn a spatial association, they also use spatial memory in some way (Brown 1992, Mcintyre 1995). These results seem to indicate that the spatial memory of locations is weaker in the absence of cues when learned as part of a spatial association than when learned using spatial memory alone. With spatial memory, animals obtain useful information from associating elements in their environment to a single reward location without necessarily obtaining the specific 'cues' to the profitability of that location (Cole et al. 1982, Gass & Sutherland 1985, Gallistel 1989, 1990, Healy & Hurley 1995, Hurley 1996). With spatial association memory, animals do not have to remember a particular location as rewarding (Bowe 1984, Brown & Gass 1993, Mcintyre 1995) because they use the association to guide them to the rewarding location. As the light cue and the feeder are separate in space by definition, spatial association memory has several components: the location of the food, the light cue that signals availability of the food, and the spatial relationship between them (Brown & Gass 1993). In the absence of the light, birds may remember only part of the association: the most recent location of the reward. In this case, when a rewarding location is visited using spatial association memory and then the cue 44 0.03 -, » 0.02 <0 o ™ a. ~ o.Oi -| 1 A _ 20 Treatment 1 A A A A A A N\N\ A A A, A A A _ 40 60 80 100 120 0.03 o C -2 0.02 o .2 O 0) o C 0.01 A 20 A l\ /WV\ AA 40 60 80 Treatment 10 1 M . 100 120 0.03 o C -2 0.02 o .2 o 0) O C 0.01 Treatment 20 100 120 0.03 o It) = - 0.02 o .2 2 * 20 40 60 Seconds 80 Treatment 30 100 120 Figure 7: Proportion of all visits to currently rewarding location in relation to time from trial's end (light off) until first probe in all treatments. 45 is removed, the memory of the location could be less reliable than if the feeder were visited using spatial memory alone. If the spatial memory component were as reliable as the spatial association memory itself, then success locating the currently rewarding location within and between trials would have been the same. If birds had no spatial memory of the currently rewarding location, I would have expected the first probes during intertrials to be random, but this was not the case. This suggests that within spatial association memory, location of the reward/response site is a separate component from the location of the light or the spatial relationship between them, but is not as strong a memory as the association itself. Spatial memory and spatial association memory may remain separate in memory after an association has been learned because the components of spatial associations are being processed individually due to their temporal and spatial separation. Temporal separation of cue and response hinders the ability to learn some associations (Rescorla 1980, Mackintosh 1983), because a time lag interferes with forming the association. In this study, birds saw the light and heard the buzzer, then flew to the panel and probed feeders. The time between the light onset and the first probe ranged from 1 second to 50 minutes (with a mode of approximately 2 seconds in all treatments). The longer lags occurred late in the day when we often see many aspects of the birds' behavior slow down. In Treatments 1,10 and 30, birds first probed a feeder less than 2.5 seconds after the trial started in 23% - 27% of all trials. In Treatment 20, the lag was less than 2.5 seconds in only 12% of trials. Even such brief lag times between seeing the cue and flying to the panel to probe a feeder may have resulted in the memory of these two components remaining separate in memory even after the association was learned. The elements of a recently learned association might remain in memory separately because it is a two step process to see a light and a spatially separate feeder. Such 46 discontiguous elements are believed to be processed separately because animals have difficulty paying attention to spatially separate items at the same time (Polidora & Fletcher 1964). This might explain why birds had greater difficulty making a spatial association with increasing distance between cues and feeders (Brown 1994). This of course does not take into account the effect of distance between cue/feeder pairs on the ability to make spatial associations. The ability to perceive these two components could be hindered by the spatial separation of the cue-feeder pairs. According to Gestalt theory and Brown (1994), decreasing distance between cue-feeder pairs results in increasing difficulty learning a spatial association. However, in this study the pairs were far enough apart not to interfere with the perception of a cue and its associated feeder. If the light cue and the most recently rewarding locations do remain in memory as separate elements, this would support Brown's (1994) interpretation of spatial association memory. She suggested that spatial association follows the underlying principle behind Gestalt theory that the 'whole' is greater than the sum of its parts (Rock & Palmer 1990). For example, a line composed of individual discontiguous elements has a different perceptual property than the elements themselves. This is because the "parts" interact with one another in the visual field, allowing the perception of a "whole". In this theory, the line or the "emergent property" has qualities that are not apparent in its parts, except in the eyes of observers (Rock & Palmer 1990). This is true even though we still see the parts at the same time as we see the whole. The idea that parts can be perceived as a whole is based on the laws of grouping (Wertheimer 1950, Rock & Palmer 1990). In this experiment, birds may have learned the association (the "whole"), but may also have remembered each of the parts of that association, including the currently rewarding location. But in that case, I suggest that the 47 memory of the location would be less reliable because it was learned within spatial association memory. In summary, these results indicate that birds can locate recently rewarding locations which they had visited using spatial association memory even in the absence of cues. However, I cannot say with certainty how they did this. The drop in performance from within trials to between trials indicates that the absence of the light cue (and conceivably the buzzer) made a difference. But it is unclear whether this was because birds had lost their ability to check their destination en route, remembered the location where they last made a spatial association and this memory was weaker, or remembered the most recently rewarding location visited. Regardless, there is some evidence that learning a spatial association results in learning and remembering both the association itself and its components (the spatial memory of both the cue and the reward/response site). Birds may be capable of using the memory of these components when they cannot use the association itself but the result is a short-term decrease in foraging success. The focus on the most recently rewarding locations is also evidence that what birds learned within trials affected their behavior between trials. Focus on current and previous rewarding locations The memory of a rewarding location learned using spatial association memory may be more susceptible to being reinforced as rewarding when visited in repeated trials than the association itself. In Chapter Two, I concluded that the presence of the light cue inhibited the reinforcement of a particular location as rewarding whether birds fed at a location for 10, 20 or 30 trials in succession. However, I found that between trials in Treatments 10, birds were equally likely to visit the feeders that provided food during the current series of trials and the previous series. These locations accounted for 40% of all intertrial visits (22% at the currently rewarding location, 18% at the previously rewarding location). I found the same 48 trend in Treatment 20, where these locations accounted for 43% of all intertrial visits (23% at the currently rewarding location, 20% at the previously rewarding location); (Fig. 8). The focus on these two locations is well above chance, as birds visited 25% of the feeder locations in > 40% of intertrials. However, in Treatment 30 birds went to the currently rewarding location of the reward in 50% of intertrials and to the previously rewarding location in only 7% of intertrials (Fig. 8). Between trials, there was a significant difference between treatments in visits to the currently (data arcsine V transformed - residuals indicate homogeneity of variance; A N O V A , F 2 ,9 = 10.37, p = 0.0081) and previously rewarding location (ANOVA, F 2 ,9 = 5.58, p= 0.0356). There was a significant difference between Treatment 30 and 10 (Tukey, q 7,3 = 5.52) and between Treatment 30 and 20 (Tukey, q 7,3 = 5.26) due to the high proportion of probes at the currently rewarding location in Treatment 30. There was also significant difference between Treatment 30 and 20 (Tukey, q 7,3 = 4.39) at the previously rewarding location for the same reason. The distribution of errors between trials also shows that Treatment 30 birds concentrated their effort on the currently rewarding feeder more than birds in other treatments. Hummingbirds in spatial association studies demonstrate that they have learned by visiting the currently rewarding feeder in the majority of trials. Brown (1992) found that as birds learned a spatial association, the majority of any errors they made were clustered immediately beside the rewarding feeder, even as the location of the reward was moved. Therefore, errors don't necessarily indicate that something has not been learned but rather can reveal a lack of accuracy in implementing what has been learned (Brown 1992). During intertrials in my study, birds could have visited any of eight feeders, yet the majority of their visits (from 34% of all visits in Treatment 1 to 78% of all visits in Treatment 30) were centered on the currently rewarding location. In Treatment 30, 53% of all errors were immediately to the right or left of 49 0.6 -(A 0.5 -<D £1 O V . Q . 0.4 -tn 0.3 -c o ort 0.2 -a o Q. 0.1 -0.0 -X i to H i Treatment 10 Treatment 20 I I Treatment 30 Chance i L i . Currently rewarding Previously rewarding Rewarding 2 series ago Rewarding 3 series ago Rewarding 4 series ago Figure 8: Mean proportion of first visits to currently and all previously rewarding locations between trials. Treatment 10 = 169 trials; Treatments 20 and 30 = 119 trials. Solid line indicates chance level (0.125). Error bars = 1 SE. 50 the currently rewarding location, in a very leptokurtic distribution (Fig. 9; j2= 11.00). The distribution of errors in Treatment 30 also resembled most closely the distribution normally seen within trials (Brown 1992, unpublished data). In Treatment 1, only 22% of errors were at these locations with the distribution being only slightly mesokurtic (Fig. 9; jz= -0.35). In Treatment 10, 33% of errors were immediately beside the currently rewarding location, while in Treatment 20, 37% of errors were at these locations. The distribution of errors in both these treatments were similarly leptokurtic (Fig. 9; Treatment 10, 72= 3.62, Treatment 20, 72 = 4.20). Birds in Treatment 30 may have focused on the currently rewarding location between trials because their spatial memory of the rewarding location was affected by experience within trials. If the cue and the rewarding location are separate in memory, based on Gestalt theory (Rock & Palmer 1990), then the memory of the currently rewarding location could be retrieved using spatial memory alone (i.e. between trials). We know that repeated foraging at the same location does not reinforce a particular location as rewarding when using spatial association memory (Brown 1992, Chapter 2) but does when using spatial memory (Sutherland & Gass 1995, Mcintyre 1995). Based on where birds probed in this study, there is evidence that when they had to use spatial memory between trials, the spatial memory of the currently rewarding location (i.e. the feeder that was cued by the light and provided food in the current series) had been strengthened with increasing amount of positive reinforcement received at that location within trials. Birds in Treatments 10 and 20 showed no strong preference for the currently rewarding location during intertrials, but birds in Treatment 30 did. Therefore, I conclude that going to the same rewarding location more than 20 times in succession using spatial association memory results in the spatial memory of the location being reinforced as rewarding. It also results in a weakening of the memory of other locations 51 0.60 0.50 0.40 0.30 0.20 0.10 0.00 Trea tmen t 1 o S _ Q . 0 J_LB 0 1 T r e a t m e n t 10 0 l 0.60 0.50 0.40 0.30 0.20 0.10 0.00 T r e a t m e n t 20 i i -7 -6 -5 -4 -3 -2 2 3 4 5 6 7 0.60 0.50 0.40 0.30 0.20 0.10 0.00 ri T r e a t m e n t 30 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 Distance from currently rewarding feeder Fig 9: Location of errors between trials in relation to currently rewarding location ( = 0). Each distance change represents one feeder location ( + = to the right of currently rewarding feeder, - = to the left of currently rewarding feeder). 52 as rewarding (Fig. 8). I had expected to see this effect within trials but it is apparent only between trials, when the cue is absent and the bird has to guess randomly or use spatial memory. The high proportion of first intertrial probes at the currently rewarding location in Treatment 30 was not an artifact of the amount of experience within the current series. I could have concluded that one reason for the difference between Treatments 10/20 and Treatment 30 was simply that the rewarding location remained at the same place longer in the latter treatment. However, when I examined only the first 10 intertrials after each switch, the trend was the same. Birds in Treatment 30 focused primarily on the currently rewarding location while birds in Treatments 10/20 focused on both the current and previously rewarding locations (Fig. 8). In the first 10 intertrials after the switch, 36% of probes in Treatment 30 were at the currently rewarding location, while only 16% of first probes in Treatment 10 and 12% in Treatment 20 were at the currently rewarding location (Fig. 10); (Kruskall-Wallis, H c = 4.73, p > 0.05). Birds focused most often on the previously rewarding location in Treatments 20 and 30 (Fig. 11). Birds went to the previously rewarding location in 36% of the first 10 intertrials in Treatment 20 and in 28% in Treatment 30 (Fig. 11). In Treatment 10, only 13% of first probes were at the previously rewarding location (Kruskall-Wallis, H c = 6.50, p < 0.05). The birds' focus on the currently and previously rewarding feeders in intertrials right after the switch in Treatment 30 compared to Treatments 10 and 20 indicates that these results are not due to differences in the number of successful foraging trips within the current series. More likely it resulted from differences in the number of successive trials the rewarding feeder was at the same location in previous series. A similar trend can also be seen within trials where there was a decrease in the lag time between the cue being lit and the bird probing at the panel; this may be related to birds learning the spatial association. The 53 Treatment 10 Treatment 20 Treatment 30 Figure 10: Mean proportion of first probes at currently rewarding location in first 10 intertrials after switch (includes all switches). Treatment 10 = 16 switches; Treatment 20 = 5 switches; Treatment 30 = 3 switches. Error bars - 1 SE. 54 0.6 CO 0.5 -o n o Q. 0.4 -+-< (0 i _ iff 0.3 -£ O orti 0.2 --Q. o CL 0.1 Pllili Treatment 10 Treatment 20 Treatment 30 Figure 11: Mean proportion of first probes at previously rewarding location in first 10 intertrials after switch (includes all switches). Treatment 10 = 16 switches; Treatment 20 = 5 switches; Treatment 30 = 3 switches. Error bars = 1 SE. 55 proportion of trials with response time < 2.5 seconds seems to increase over time in all Treatments. Though the increases were small in Treatment 10 and 20, it is similar in form to that seen in Treatment 30, which showed the greatest increase. Since trial numbers differed from treatment to treatment, this indicates the response is related to learning the spatial association and not simply an artifact of having more experience at one location. One explanation for the difference in the number of visits to the currently rewarding location between Treatments 10/20 and Treatment 30 is that the capacity of the birds' memory window is too small to hold more than a certain number of foraging experiences. Under the marginal value theorem, foragers should move to another patch once the profitability of the current patch falls below a certain value estimated by averaging instantaneous harvest rates (Charnov 1976). However, the mechanisms by which this decision is made are not clear. One possibility is the use of a memory window. A memory window is a time period over which harvest information is averaged (Krebs et al. 1974, Cuthill et al. 1990, Valone 1992). We assume that this window is a series of slots containing information gathered over a period of time (Valone 1992). As new information is added to the window, older information is pushed out and the mean is updated (Cowie 1977). Therefore, a forager bases its decision on whether to stay in a patch or leave on its most recent foraging experience by averaging successful foraging experiences over a certain period of time. Memory window models also theorize that when a forager enters a new patch, the memory window is reinitialized using the harvest rate of the mean patch in the environment (Valone 1992). The size of the window varies with the distribution of resources in the habitat, species, and travel time between patches (Valone 1992). Moderate size windows are preferred in many environments because small windows are more acutely influenced by stochastic variations in profitability within patches and large windows contain outdated information (Inoue 1983a, b, Valone & Giraldeau 1993). 56 If rufous hummingbirds use a memory window to decide when to leave patches and they used something like it in this experiment, these results indicate that this window may hold somewhere around 30 successful foraging experiences when the travel time is small. In Treatments 10/20, when there were more than 10 but fewer than 30 successful foraging experiences per series (i.e. trials), the memory window slots may have been dominated by visits to the currently rewarding location, while the remainder of the window was filled with the experience of the previously rewarding location. The similarity in focus on the currently and previously rewarding locations in Treatments 10/20 may be because the switch to focusing on the currently rewarding location is seen only after 20 to 30 positive foraging experiences. In Treatment 30, the memory window was fully occupied by memories of the successful foraging trips at the currently rewarding location, leaving little room for the experience of visiting the previously rewarding location. The evidence is found in the focus of birds in this treatment on the currently rewarding location. If the memory window model is correct, this would explain the differences between Treatment 10/20 and Treatment 30. Effect of intertrial activity on within-trial performance Success in locating rewards within trials may have been improved by 'practice' behavior between trials. As much as birds had learned to respond to particular cued locations within trials, they also learned to fly to the panel to get food. As I did not remove birds from the experimental room during intertrial periods, it is possible that within-trial performance (which was high) was influenced by interaction with the experimental environment during these unrewarding periods. In some bout-structured cognitive studies, animals are placed in neutral environments during a retention period (e.g. Balda & Kamil 1988, Mendl 1997) after being exposed to particular problems. A benefit from the experimenter's perspective of 57 removing the animal is that it allows what was learned to remain untainted by further experience in the room, which otherwise could disrupt what has been learned (Menzel 1979, Lewis 1986), allowing the researcher to obtain a clear answer to the question. However, could the problem being learned within trials be reinforced positively between trials, even without a reward, simply by going through the motions of the task? If it can, the birds in my study might have reinforced what they were learning within trials by flying to the panel, probing at feeders and returning to the perch between trials. This may have allowed them to be more proficient at learning the spatial association memory (and spatial memory) problem within trials, perhaps because they were more "comfortable" with their entire environment and the experimental protocol. This interpretation also suggests that the association between cue, location and food was not being formed only within trials. Though I have no strong evidence that birds learned that there was no food available during intertrials, there is some evidence that what occurred within trials affected behavior between trials. Birds engaged in 360 intertrial foraging flights in Treatment 1, 284 in Treatment 10 and 313 in Treatment 30. However, Treatment 20 birds engaged in only 159 intertrial foraging flights. The low number of flights in Treatment 20 might be due to the high level of performance in the initial series of trials resulting in reduced overall effort during intertrials. However, the greatest proportion of Treatment 20 intertrial flights occurred in the 2 n d series when birds were having the greatest difficulty locating rewards during trials. The greatest proportion of intertrial foraging flights in Treatments 10 and 30 occurred in the 1st series. Treatment 20 birds' low performance within trials in the 2 n d series indicated they had not learned the spatial association and therefore had more difficulty finding food. This may have resulted in an increased effort to obtain food and more foraging flights during the 2 n d series of intertrials. These results further confirm how inter-trial and within-trial behavior are 58 linked and should not necessarily be separated in analysis. To test whether intertrial behavior did increase performance within trials it would be necessary to deny access to the panel between trials in one treatment. A moderately difficult spatial association memory task would ensure that differences in performance would be measurable. However, taking hummingbirds out of the room is not a viable option in multiple trial experiments as the disturbance and stress on the birds would be counterproductive to the study. If the panel could be covered between trials without disturbing the birds (see Gass & Sutherland 1985) or if the birds could move between two rooms, such a study could be possible and we could begin to answer these questions. CONCLUSION Intertrial behavior in bout-structured experiments can help reveal the processes being studied within trials. In bout-structured studies, animals are removed from the experimental environment during intertrial or retention periods. This prevents their obtaining additional information about their environment or reinforcing their most recent experiences by interacting with their surroundings. In my study, birds remained in the experimental room during intertrial periods and could continue "foraging" but without getting food. The results indicate that their experiences during this time may have both influenced how they learned within trials and revealed normally hidden components of spatial association memory, the mechanism being studied within trials. This would indicate that learning is not solely a result of associating immediate experiences but is influenced by experience not directly related to the task. The experiment in Chapter Two was not designed to answer questions related to intertrial behavior. But it seems clear that what birds do between trials is not irrelevant to 59 what I wanted to study within trials. The activity of birds within these periods provides some information about spatial association memory that was not apparent within trials. Any confirmation of the hypotheses presented in this chapter would need a study designed to answer these questions specifically. In future, the behavior of birds between trials should not be ignored as it has been in the past (Brown 1992, Thompson 1994). With this new perspective on intertrial behavior, we should examine more closely the purpose and necessity for intertrial periods in spatial association studies. Though there are disadvantages to not having these periods (see Brown 1992, Mclntyre 1995), they might not always be necessary. Depending on what exactly is being studied in the experiment, intertrial periods could be excluded or designed in such a way as to reduce influence on within-trial behavior. The clearest way to reduce this influence is by removing the animal from the experimental room during intertrials. However, if an intertrial period where animals must remain in the experimental environment is necessary, we should consider both the effects on within-trial behavior and what we could learn between trials about behavior within trials. We now have evidence that within trials and between trials are not the independent time periods we have always assumed. 60 C H A P T E R F O U R G E N E R A L CONCLUSIONS This study of spatial association learning in rufous hummingbirds is part of a i continuing effort to learn more about cognitive behavior within an ecological context. Foraging theory provides us with many explanations for and predictions about foraging decisions but few insights into how these decisions are made. Psychological studies reveal a lot about the cognitive mechanisms that animals could use to make foraging decisions, yet we are still learning how or if these mechanisms are used in the wild. My study revealed a greater complexity than expected about spatial association memory use in rufous hummingbirds and provides more context to how it could be used by this and related species. This study suggests that the components of a spatial association memory (visual cue and reward/response site) can be remembered independently of the association itself. The birds' focus on the currently rewarding location between trials in all treatments, when no cue was available, indicates that once a spatial association is learned, a spatial memory of the reward/response site remains. This supports Brown (1992) who suggested that the way hummingbirds make spatial associations follows Gestalt theory, in which both the association and its parts are presumed to be each perceived and remembered. Further evidence for the use of a spatial memory of the reward/response site is that birds in Treatment 30 focused primarily on the currently rewarding location while birds in Treatments 10 and 20 focused on the currently and previously rewarding locations equally. Since repeated foraging experience at a particular location reinforces the spatial memory of that location (Mclntyre 1995) but the rewarding location is not reinforced when using spatial association memory (Chapter Two), I conclude that spatial memory was likely being used between trials and it persisted for at least 61 two series of trials. I speculated in Chapter Two on other reasons why the current location of the reward/response site was remembered though not as well between trials as within-trials. Further studies are necessary to more confidently support or reject any of these suggested hypotheses. My study also revealed that there is a possible impact on learning when the study animal remains in the experimental environment during bout-structured experiments. As I discussed in Chapter Three, there is evidence that within-trial activity influenced between-trial behavior and that between-trial behavior may also have affected within-trial performance. The assumption that animals will reveal what and how they are learning only during imposed periods of the experimental protocol is unfounded. My work suggests that animals register information about their environments at all times and therefore they are learning even when they are not participating in 'the experiment'. Birds in this study used information gathered within trials to continue foraging for food between trials, never behaving as if they had learned that no reward was available between trials. That aspect of the experimental structure was perhaps not relevant enough to learn or their behavior was part of reinforcing what they had learned within trials in spite of the lack of food reinforcement. Regardless, the result of their continued activity between trials was revealing aspects of spatial association memory that were not apparent within trials. This information could not have been uncovered had I ignored these time periods. Including behavior outside the structure of the experiment in our analysis challenges our assumptions of how experimental protocols are perceived and learned by study animals and suggests closer scrutiny of the protocol when designing bout-structured experiments. This study also suggests that hummingbirds in the wild could use spatial association memory at times when their environment is changing. Though we have learned much about 62 this mechanism in laboratory studies, we have no quantitative evidence for when or even whether hummingbirds use spatial association memory in the wild. Based on what we do know, I suggest it could be used to adjust to changes in profitability of flowers or patches. To test this, I created an environment in which the profitability of a location suddenly changed after a fixed number of trials. Though temporal changes in profitability do not occur in the wild as predictably as they did during this experiment, the results indicated that spatial association memory has the characteristics to be useful when a new rewarding location must be found. It could be used at times of change because it is not anchored to a particular location, the spatial location of the reward/response site is not reinforced with repeated foraging experience and the spatial memory of where a bird recently obtained food could be used to improve success over random foraging. Also, my study provides evidence that birds can quickly learn the cues to use in a spatial association even when they are using another mechanism such as spatial memory to forage. This ability would allow them to quickly change from spatial memory to spatial association memory and back, with little energetic cost. This type of flexibility would seem essential for animals whose food source is continually changing and is possibly why rufous hummingbirds show the ability to use spatial association memory in the lab. This study has revealed as much about the way we study spatial learning in rufous hummingbirds as it has about spatial association memory. I believe that spatial association memory is a complex mechanism whose usefulness is its versatility. It seems to work in conjunction with other mechanisms such as spatial memory allowing animals the ability to respond to changes in their environment. There is evidence in this study that the components of spatial association memory (such as the visual cue) can be processed prior to actually being used and that it is a reliable mechanism at times of stability (within series of trials). Also, 63 even in the absence of the visual cue, the spatial memory of the reward/response site visited using spatial association memory can provide enough information to locate recently rewarding locations more efficiently than through random foraging. I am certain there is more that can still be learned about spatial association memory in laboratory studies. However, I believe that concurrent field studies with hummingbirds and other species which could confirm the use of spatial association memory in floral or patch selection would provide an even greater ecological context to our continuing study of this cognitive mechanism. 64 Literature cited Anderson, J.R. 1985. Cognitive psychology and its implications. New York: W.H. Freeman & Company. Armstrong, D.P., C L . Gass & C D . Sutherland. 1987. Should foragers remember where they've been? Explorations of a simulation model based on the behavior and energetics of territorial hummingbirds. In: Foraging Behavior (Ed. By A . C Kamil, J.R. Krebs & H.R. Pulliam), pp 563 -586. New York: Plenum Press. Balda, R.P. & A.C. Kamil. 1988. The spatial memory of Clark's nutcrackers (Nucifraga columbiana) in an analogue of the radial arm maze. Animal Learning and Behavior. 16(2), 116-122. Balda, R.P. & A.C. Kamil. 1992. Long-term spatial memory in Clark's nutcracker (Nucifraga columbiana). Animal Behaviour. 44, 761-769. Bednekoff, P.A. & R.P. Balda. 1996. Observational spatial memory in Clark's nutcrackers and Mexican jays. Animal Behaviour. 52, 833-839 Bowe, C A . 1984. Spatial relations in animal learning and behavior. The Psychological Record. 34, 181-209. Brodbeck, D.R. 1994. Memory for spatial and location cues: a comparison of a storing and a non-storing species. Animal Learning and Behavior. 22(2), 119-133. Brodbeck, D.R. & S.J. Shettleworth. 1995. Matching location and color of a compound stimulus: comparison of a food-storing and non-storing bird species. Journal of Experimental Psychology (Animal Behavior Processes). 21(1), 64-77. Brown, C S . 1992. Factors affecting spatial association learning by hummingbirds. Ph.D. Thesis, University of British Columbia. Brown, C S . & C L . Gass. 1993. Spatial association learning by hummingbirds. Animal Behaviour. 46. 487-497. Brown, C S . 1994. Spatial association learning by rufous hummingbirds (Selasphorus rufus): Effects of relative spacing among stimuli. Journal of Comparative Psychology. 108(1), 29-35. Calder, W.A. 1993. Rufous Hummingbird (Selasphorus rufus). In: The Birds of North America. No. 53 (Ed by A. Poole & F. Gill), pp 1-20. Philadelphia: The Academy of Natural Sciences; Washington. D . C : The American Ornithologists' Union.. 65 Carlson, N.R. 1993 Psychology: the science of behavior. 4m edition. Toronto: Allyn & Bacon. Carpenter, F.L., M.A. Hixon & D.C. Paton. 1981. Premigratory changes in time budgets of territorial hummingbirds. American Zoologist, 21, 921. Carpenter, F.L., D.C. Paton & M.A. Hixon. 1983. Weight gain and adjustment of feeding territory size in migrant hummingbirds. Proceedings of the National Academy of Science. USA. 80, 7259-7263. Charnov, E.L. 1976b. Optimal foraging: the marginal value theorem. Theoretical Population Biology. 9. 129-136. Chatters, L. 1996. The dynamics of hummingbird flight. MSc Thesis, University of British Columbia. Clayton, N.S. & J.R. Krebs. 1994. Memory for spatial and object-specific cues in food storing and non-storing birds. Journal of Comparative Physiology A. 174, 371-379. Cole, S., F.R. Hainsworth, A.C. Kamil, T. Mercier & L.L . Wolf. 1982. Spatial learning as an adaptation in hummingbirds. Science. 217, 655-657. Cowie, R.J. 1977. Optimal foraging in great tits (Parus major). Nature, 268, 137-139. Cuthill, I.C., A. Kacelnik, J.R. Krebs, P. Haccou & Y. Iwasa. 1990. Starlings exploiting patches: the effect of recent experience on foraging decision. Animal Behaviour. 40. 625-640. D'Amato, M.R. 1970. Experimental Psychology: methodology, psychophysics and learning. Toronto: McGraw-Hill Book Company. Davis, R.T. 1974. Spatial discontiguity. In: Primate Behavior, vol. 3: Monkeys as perceivers (Ed by R.T. Davis), pp 62-72. New York: Academic Press. Gallistel, C R . 1989. Animal cognition: the representation of space, time and number. Annual Review of Psychology. 40, 155-189. Gallistel, C R . 1990 The Organization of Learning. Cambridge, Mass: MIT Press. Garber, P A . & L . M . Paciulli. 1997. Experimental field study of spatial memory and learning in wild capuchin monkeys (Cebus capucinus). Folia Primatologia, 68, 236-253. Gass, C.L., G. Angehr & J. Centa. 1976. Regulation of food supply by feeding territoriality in the rufous hummingbird. Canadian Journal of Zoology, 54, 2046-2053. 66 Gass, C.L. 1978. Rufous hummingbird feeding territoriality in a suboptimal habitat. Canadian Journal of Zoology. 56. 1535-1539. Gass, C.L. 1979. Territory regulation, tenure, and migration in rufous hummingbirds. Canadian Journal of Zoology. 57, 914-923. Gass, C.L. & R.D Montgomerie. 1981. Hummingbird foraging behavior: decision-making and energy regulation. In: Foraging Behavior: ecological, ethological. and psychological approaches (Ed by A.C. Kamil & T.D. Sargent), pp 159-194. New York: Garland STPM Press. Gass, C.L. & G.D. Sutherland. 1985. Specialization by territorial hummingbirds on experimentally enriched patches of flowers: energetic profitability and learning. Canadian Journal of Zoology. 63, 2125-2133. Gass, C.L. 1985. Behavioural foundations of adaptation. In: Perspectives in ethology Vol 6 (Ed by P.P.G. Bateson & P.H. Klopfer). pp 63-107. New York: Plenum Publishing Corporation. Gibbon, J., R.M. Church, S. Fairhurst & A. Kacelnik. 1988. Scalar expectancy theory and choice between delayed reward. Psychological Review. 95(1), 102-114. Gill, F.B. 1987. Ecological fitting: use of floral nectar in Heliconia stilesii daniels by three species of hermit hummingbirds. The Condor. 89, 779-787. Gill F.B. & L.L . Wolf. 1979. Nectar loss by golden-winged sunbirds to competitors. The Auk. 96,448-461. Gori, D.F. 1989. Floral color change in Lupinous argenteus (Fabaceae): Why should plants advertise the location of unrewarding flowers to pollinators? Evolution. 43, 870-881. Healy, S.D. & T.A. Hurly. 1995. Spatial memory in rufous hummingbirds (Selasphorus rufus): a field test. Animal learning and behavior. 23(1), 63-68. Heinrich, B. 1979. Bumblebee Economics. Cambridge, Mass: Harvard University Press. Herz, R.S., L. Zanette & D.F. Sherry. 1994. Spatial cues for cache retrieval by black-capped chickadees. Animal Behaviour. 48, 343-351. Hintzman, D.L. 1978. The psychology of learning and memory. San Francisco: W.H. Freeman & Company. Hixon, M.A., F.L. Carpenter & D.C. Paton. 1983. Territory area, flower density, and time budgeting in hummingbirds: an experimental and theoretical analysis. The American Naturalist. 122(3), 366-391. 67 Hixon, M.A. & F.L. Carpenter. 1988. Distinguishing energy maximizers from time minimizers: a comparative study of two hummingbird species. American Zoologist, 28,913-925. Houston, D.C. 1983. In: Vulture biology and management (Ed by S.R. Wilbur & J.A. Jackson), pp 135-152. Berkeley: University of California Press. Hurly, T.A. 1996. Spatial memory in rufous hummingbirds: memory for rewarded and non-rewarded sites. Animal Behaviour. 51, 177-183. Hurly, T.A. & S.D. Healy. 1996. Memory for flowers in rufous hummingbirds: location or local visual cues? Animal Behaviour. 51,1149-1157. Inoue, T. 1983a. Foraging strategy of a non-omniscient predator in a changing environment (I) model with a data window and absolute criterion. Resource and Population Biology. 25,81-104. Inoue, T. 1983b. Foraging strategy of a non-omniscient predator in a changing environment (II) model with two data windows and a relative comparison criterion. Resource and Population Biology. 25, 81-104. Kahneman, D. 1973. Attention and effort. Englewood Cliffs, NJ: Prentice-Hall. Keele, S.W. 1973. Attention and human performance. Goodyear: Pacific Palisades. Klosterhalfen S., W. Fischer & M.E. Bitterman. 1978. Modification of attention in honeybees. Science, 201, 1241-1243 Kodric-Brown, A. & J.H. Brown. 1978. Influence of economics, interspecific competition, and sexual dimorphism on territoriality of migrant rufous hummingbirds. Ecology. 59(2), 285-296. Krebs, J.R., J.C. Ryan & E.L. Charnov. 1974. Hunting by expectation or optimal foraging? A study of patch use by chickadees. Animal Behaviour. 22, 953-964. Krebs, J.R. & R.H. McCleery. 1984. Optimization in behavioural ecology. In: Behavioural Ecology: an evolutionary approach (Ed by J.R. Krebs & N.B. Davies). pp 91-121. Sunderland MA: Sinauer Associates. Lewis, C A . 1986. Memory constraints and flower choice in Pieis rape. Science, 232, 863-865. Lima, S.L. 1984. Downy woodpecker foraging behavior: Efficient sampling in simple stochastic environments. Ecology, 65(1), 166-174. 68 Mackintosh, N.J. 1983. Conditioning and associative learning. Oxford: Oxford University Press. McCracken, G.F. 1993. Locational memory and female-pup reunions in Mexican free-tailed bat maternity colonies. Animal Behaviour, 45, 811-813. Mclntyre, Gordon D. 1995. Pattern learning and spatial memory in rufous hummingbirds (Selasphorus rufus). MSc Thesis, University of British Columbia. Melendez-Ackerman. E.J. 1997a. Patterns of color and nectar variation across an Ipomopsis (Polemoniaceae) hybrid zone. American Journal of Botany, 84(10), 41-47. Melendez-Ackerman, E.J., D.R. Campbell & N.M. Waser 1997b. Hummingbird behavior and mechanisms of selection on flower color in Ipomopsis. Ecology, 78(8).-2532-2541. Mendl, M . , K. Laughlin & D. Hitchcock. 1997. Pigs in space: spatial memory and its susceptibility to interference. Animal Behaviour. 54: 1491-1508. Menzel, R. 1979. Behavioural access to short-term memory in bees. Nature. 281, 368-369. Menzel, R. 1985. Learning in honeybees in an ecological and behavioral context. In: Experimental Behavioural Ecology and Sociobiology (Ed by B.H. Holldobler & M . Lindauer). pp 55-74. Sunderland: Sinauer Associates. Miller, R.S., S. Tamm, G.D. Sutherland & C L . Gass. 1985. Cues for orientation in hummingbird foraging: color and position. Canadian Journal of Zoology. 63, 18-21. Moore, B.R. 1973. The role of directed Pavlovian reactions in simple instrumental learning in the pigeon. In: Constraints on learning: limitations and predispositions, pp 159-187. London and New York: Academic Press. Moore, J. 1997. Ecomorphology of rufous hummingbirds: an investigation of maneuverability and agility in four age-sex classes. MSc Thesis, University of British Columbia. Oaten, A. 1977. Optimal foraging in patches: a case for stochasticity. Theoretical Population Biology. 12, 263-285. Paton, D.C. & F.L. Carpenter. 1984. Peripheral foraging by territorial rufous hummingbirds: defense by exploitation. Ecology. 65(6), 1808-1819. Pinel, J.P.J., M.J. Mana & D.M. Wilkie. 1986. Postshock learning and conditioned defensive burying. Animal Learning and Behavior, 14(3), 301-304. 69 Polidora, V.J. & H.J. Fletcher. 1964. Analysis of the importance of S-R spatial contiguity for proficient primate discrimination performance. Journal of Comparative and Physiological Psychology. 57(2), 224-230. Reebs, S.G. 1993. A test of time-place learning in cichlid fish. Behavioural Processes. 30, 273-282. Rescorla, R.A. 1980. Pavlovian second-order conditioning: studies in associative learning. Hillsdale, NJ: Lawrence Erlbaum Associates. Roberts, W A . 1991. Testing optimal foraging theory on the radial maze: the role of learning in patch sampling. Animal Learning and Behavior. 19(4), 305-316. Robinson, G.E. & F.C. Dyer. 1993. Plasticity of spatial memory in honeybees: reorientation following colony fission. Animal Behaviour, 46, 311-320. Rock, I. & S. Palmer. 1990. The legacy of Gestalt psychology. Scientific American. December, 84-90. Sherry, D.F. & A.L. Vaccarino. 1989. Hippocampus and memory for food caches in black-capped chickadees. Behavioral Neuroscience. 103(2), 308-318. Shettleworth, S.J. 1983. Memory in food-hoarding birds. Scientific American. 102-110. Shettleworth, S. J. 1985. Food storing by birds: implications for comparative studies of memory. In: Memory systems of the brain. (Ed by N.H. Wienberger, J.L. McGaugh & G. Lynn), pp 231-250. New York: Oxford University Press. Stephens, D.W. & J.R. Krebs. 1986. Foraging Theory. Princeton, NJ: Princeton University Press. Sutherland, G.D. & C L . Gass. 1995. Learning and remembering of spatial patterns by hummingbirds. Animal Behaviour, 50, 1273-1286. Thompson, J. 1994. Do hummingbirds use contextual information when performing spatial association tasks? MSc Thesis, University of British Columbia. Valone, T.J. 1992. Information for patch assessment: a field investigation with black-chinned hummingbirds. Behavioral Ecology. 3, 211-222. Valone, T.J. & L-A. Giraldeau. 1993. Patch estimation by group foragers: what information is used? Animal Behaviour. 45, 721-728. Vander Wall, S.B. 1982. An experimental analysis of cache recovery in Clark's nutcracker. Animal Behaviour. 30, 84-94. 70 Vander Wall, S.B. 1991. Mechanisms of cache recovery by yellow pine chipmunks. Animal Behaviour. 41, 851-863. Warburton, K. 1990. The use of local landmarks by foraging goldfish. Animal Behaviour. 40, 500-505. Wehner, R., B. Michel & P. Antonsen. 1996. Visual navigation in insects: coupling of egocentric and geocentric information. The Journal of Experimental Biology. 199, 129-140. Wells, D.J. 1993b. Ecological correlates of hovering flight of hummingbirds. Journal of Experimental Biology. 178, 59-70. Wertheimer, M . 1950. Laws of organization in perceptual forms. In: A source book of Gestalt psychology (Ed by W.D. Ellis), pp 71-88. New York: Humanities Press. Wickens, C D . 1984. Engineering psychology and human performance. Toronto: Charles E. Merrill Publishing Company. Wolf, L .L . 1969. Female territoriality in a tropical hummingbird. Auk. 86, 490-504. Wolf, L .L . 1978. Aggressive social organization in nectivorous birds. American Zoologist. 18,765-778. Yaginuma, S. & E. Iwai. 1986. Effect of small cue-response separation on pattern discrimination in macaques (Macaca fuscata and M . mulatta). Journal of Comparative Psychology. 100(2), 137-142. 71 

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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0099317/manifest

Comment

Related Items