@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix dc: . @prefix skos: . vivo:departmentOrSchool "Arts, Faculty of"@en, "Psychology, Department of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Willson, Robert James"@en ; dcterms:issued "2008-12-17T19:28:26Z"@en, "1992"@en ; vivo:relatedDegree "Doctor of Philosophy - PhD"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description """A series of six experiments was conducted examining spatial memory in pigeons. Spatial memory in this species has traditionally been characterized as poor in relation to other avian species, and this has led to speculations about adaptive specializations in spatial memory systems. The results from the experiments conducted in the present thesis in which novel procedures were used to study pigeons’ spatial memory suggest that this characterization has been a result of the experimental procedures used to assess spatial memory, rather than an inherent lack of ability. The procedures that were employed in the current research are also novel in that they are consistent with a theoretical memory distinction proposed by Gaffan (1974) between recognition and associative memory. Recognition memory tasks are procedures in which only the to-be-remembered stimulus is presented during the study phase of a trial. The subject must subsequently discriminate between that stimulus and novel or unfamiliar stimuli during the retention test. Associative memory tasks are procedures in which all stimuli to be presented during the retention test are also presented during the study phase of the trial. In these tasks the subject must identify the target from among the presented set of stimuli and remember its identity when subsequently reexposed to the same stimulus set during the retention test. In the current research, pigeons performed well with extended retention intervals when tested on an associative memory task but not when tested on a recognition memory task. The birds exposed to four spatial locations and given a brief interval to ascertain which one of the four locations was rewarded showed excellent retention for the rewarded location, for periods of up to 72 hr. This level of retention is far greater than that observed in previous spatial memory tasks with this species. The procedures that were employed here are somewhat similar to procedures that have been used to study spatial memory in food-storing birds (Brodbeck, Burack & Shettleworth, in press). Although comparisons across experiments and procedures must be viewed with caution, the present results suggest that under some conditions, pigeons apparently perform better than food-storers on this type of task. As such the findings from the current research program have important implications in relation to the issue of adaptive specializations in memory systems and for the comparative study of spatial memory."""@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/3009?expand=metadata"@en ; dcterms:extent "2803118 bytes"@en ; dc:format "application/pdf"@en ; skos:note "SPATIAL ASSOCIATIVE )IEMORY IN PIGEONSbyROBERT JAMES WILLSONB.Sc. (Honours), Dalhousie University, 1982M.A., The University of British Columbia, 1988A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF PSYCHOLOGYWe accept this thesis as conformingto the required standard:THE UNIVERSITY OF BRITISH COLUMBIAJANUARY 1992Robert James Willson, 1992In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.________________________________Department of 1;; i-.The University of British ColumbiaVancouver, CanadaDate ;7DE-6 (2188)ABSTRACTA series of six experiments was conducted examining spatial memoryin pigeons. Spatial memory in this species has traditionally beencharacterized as poor in relation to other avian species, and this hasled to speculations about adaptive specializations in spatial memorysystems. The results from the experiments conducted in the presentthesis in which novel procedures were used to study pigeons’ spatialmemory suggest that this characterization has been a result of theexperimental procedures used to assess spatial memory, rather than aninherent lack of ability. The procedures that were employed in thecurrent research are also novel in that they are consistent with atheoretical memory distinction proposed by Gaffan (1974) betweenrecognition and associative memory. Recognition memory tasks areprocedures in which only the to-be-remembered stimulus is presentedduring the study phase of a trial. The subject must subsequentlydiscriminate between that stimulus and novel or unfamiliar stimuliduring the retention test. Associative memory tasks are procedures inwhich all stimuli to be presented during the retention test are alsopresented during the study phase of the trial. In these tasks thesubject must identify the target from among the presented set of stimuliand remember its identity when subsequently reexposed to the samestimulus set during the retention test.In the current research, pigeons performed well with extendedretention intervals when tested on an associative memory task but notUwhen tested on a recognition memory task. The birds exposed to fourspatial locations and given a brief interval to ascertain which one ofthe four locations was rewarded showed excellent retention for therewarded location, for periods of up to 72 hr. This level of retentionis far greater than that observed in previous spatial memory tasks withthis species. The procedures that were employed here are somewhatsimilar to procedures that have been used to study spatial memory infood-storing birds (Brodbeck, Burack & Shettleworth, in press).Although comparisons across experiments and procedures must beviewed with caution, the present results suggest that under someconditions, pigeons apparently perform better than food-storers on thistype of task. As such the findings from the current research programhave important implications in relation to the issue of adaptivespecializations in memory systems and for the comparative study ofspatial memory.UITABLE OF CONTENTSABSTRACTTABLE OF CONTENTS ivLIST OF FIGURES viACKNOWLEDGEMENTS ix1cognition 392123242526273030323335363840414244454849505153EXPERIMENT 1MethodSubjectsApparatusProcedureResultsDiscussion62626263636575INTRODUCTIONSome genera1comments on the study of comparativeA Brief Review of Spatial Memory ResearchComparative Spatial CognitionSpatial Memory Research in PigeonsBond, Cook and Lamb (1981)Wilkie, Spetch and Chew (1981)Olson and Maki (1983)Roberts and Van Veldhuizen (1985)Spetch and ColleaguesSpetch and Edwards (1986)Spetch and Edwards (1988)Spetch and Honig (1988)Spetch (1990)Dale (1988)Roberts (1988)Zentall, Steirn and Jackson-Smith (1990)Wilkie and ColleaguesWilkie and Summers (1982), Wilkie (l983a)Wilkie (l983b)Wilkie (1984, 1986)Wilkie and Kennedy, 1987; Wilkie, Wilisonand Lee ( 1990)Wilkie (1989)(1988, 1989, in press)ChengPigeon Spatial Memory: A SummaryWillson & Wilkie (1991)lvEXPERIMENT2.78Method 80Subjects and Apparatus 80Procedure 81Results and Discussion 82EXPERIMENT 3 90Method 91Subjects and Apparatus 91Procedure 91Results and Discussion 92EXPERIMENT 4 99Method 103Subjects 103Apparatus 103Procedure 104Results and Discussion 108EXPERIMENT 5 . . . .114Method . . . .117Subjects and Apparatus . . . .117Procedure . . . . 118Results . . . .118Discussion . . . .129EXPERIMENT 6 132Method 133Subjects and Apparatus 133Procedure 133Results 134Discussion 138GENERAL DISCUSSION 138The Associative memory hypothesis revisited 143Willson and Wilkie (1991) revisited 146Zentall, Steirn and Jackson-Smith (1990) revisited 148Comparisons with the study of associative memory in foodstoring birds 150Ecology and comparative cognition 152Conclusions 156REFERENCES 159VLIST OF FIGURESFigure 1. An overhead schematic diagram of the apparatus usedin Willson and Wilkie (1991). Numbers indicate peckingkey positions 55Figure 2. The retention interval data from Wilison and Wilkie(1991), Experiment 2. All data points, except the finalone for each subject, represent blocks in which criterionwas met (DR > 0.70) 60Figure 3. Average discrimination ratio for each subject onSample/Distractor (DR_SD) vs Sample Alone (DR_S) trialsin Experiment 1 67Figure 4. Average percent correct first choices for each subject onSample/Distractor (PC_SD) vs Sample Alone (PCS) trialsin Experiment 1 70Figure 5. A comparison of the group means for each trial type andeach measure of performance from Experiment 1. Symbolsare the same as in Fig. 3 and 4 73Figure 6. A comparison of the mean performance on the four trialtypes from Experiment 2 for each subject. Both the DRand PC data are presented. The group average is alsoshown 83Figure 7. Mean performance at each retention interval for eachsubject in Experiment 3. The group average is alsoshown. Note that the x-axis is a log scale 93viFigure 8. Mean performance on each of the memory intervaldistributions used in Experiment for each subject. Thegroup average is also shown. The first block ofintervals is called DR1, the second block, DR2, and soforth 96Figure 9. Two consecutive sessions are shown schematically. Theblack bar represents a period (60-sec in length),initiated by the first keypeck, during which food wasunavailable. The crosshatched bar represent the key onwhich food was available on a VI 30-sec schedule for thelast 16 mm of the session (key 3 in the Day N-l sessionand key 1 in the Day N session). During this period foodwas not available on the keys represented by grey bars.. .105Figure 10. DRs for subjects in Experiment 4. The first,crosshatched bar is the DR for the final 16 mm of eachsession. This terminal DR is a measure of how wellsubjects located and exploited the profitable key. TheDRs represented by the grey and black bars are DRs fromthe first mm of each session. These initial DRs measuresubjects’ tendency to respond on a key that was rewardedduring the previous session. The grey bars representsessions separated by 24 hours, the black bar, sessionsseparated by 72 hours 109vuFigure 11. Initial DRs for the five blocks of 10 sessions thatcomposed Experiment 4 112Figure 12. Initial DRs calculated for the key that was profitable2 days earlier in Experiment 4 115Figure 13. Initial DRs, averaged over all 80 sessions in Experiment5 119Figure 14. Terminal DRs for each of the eight 2-mm blocks thatcomposed a session. The data are from the last 25sessions of Experiment 5 121Figure 15. Terminal DRs for three selected sessions from the 25sessions in Experiment 5 in which responding wasrecorded in eight 2-mm blocks 124Figure 16. Initial DRs for the eight blocks of 10 sessions thatcomposed Experiment 5 127Figure 17. Initial DRs calculated for the key that was profitable 2days earlier in Experiment 5 130Figure 18. Initial DRs for each session duration employed inExperiment 6 135vii’1.ACKNOWLEDGEMENTSFirst and foremost, my heart-felt thanks to Don Wilkie, my friendand advisor who, through his support and encouragement, made this entireendeavor much easier than it could have been. Thanks also to Anita andKevin Lee, a dynamic sister-brother combination that, each in their ownway, often helped me retain my sanity and kept the lab from sliding intochaos. Their help with data collection and analysis is alsoappreciated. Vern Honig and Marcia Spetch deserve mention because they,more than anybody else, were largely responsible (in a very positivewayt) for my decision to pursue this goal. Sara Shettleworth helpedprovide a bright light at the end of the tunnel and Dave Brodbeck helpprovide many fond memories and interesting discussions along the way.Jim Blackburn, Dave Mumby, Emma Wood and Steve Wicks did also. Manythanks to both the faculty and students (past and present) in theBiopsychology area, who provided a variety of perspectives and manylively discussions on a wide breadth of topics, both academic and nonacademic. Thanks also to Chris LaLonde, who, although he doesn’t quitesee the central importance of biopsychological research (being moreconcerned with developmental issues), was a willing and activeparticipant in many late night discussions (and to Laurie for putting upwith us). Finally my thanks to two very special people, Kate Banks andNarly Golestani, who, albeit for very different reasons, played (and Ihope will continue to play) important roles in my life. Thanks.rlxINTRODUCTIONThis thesis is about spatial memory in pigeons. This species hastypically been characterized as having inferior spatial memory abilitiesin relation to other animals. The results from the current series ofexperiments suggest that this poor characterization is unjustified andthat the level of performance obtained depends upon the nature of thetest used to assess performance. Pigeons are apparently very sensitiveto task variables and this dissertation identifies certain taskvariables that are important in mediating performance on spatial memorytasks. A theoretical memory distinction previously unexplored inpigeons, between recognition memory and associative memory (Gaffan,1974) provides a useful framework for classifying the procedures usedhere. It also provides a potential bridge between the current work andwork done with food-storing birds (i.e., Brodbeck, Burack &Shettleworth, in press) and has important implications for the issue ofadaptive specializations in memory. As such, the approach adopted hereis to paint a fairly broad picture of where the current research fitsinto the previously existing framework on the comparative study ofspatial memory. It begins with a brief discussion of what comparativecognition is, how the study of comparative cognition has beenapproached, potential problems associated with the study of comparativecognition and how the procedures and findings from the current work helpto alleviate those problems.1Spatial Memory in PigeonsThe working-associative-reference memory distinction firstelaborated by Honig (1978) has been the dominant theoretical frameworkfor the study of spatial memory in animals. However, this frameworkdoes not apply well to the procedures employed in the current research.An alternative framework, the recognition-associative memory distinctionof Gaffan (1974), mentioned previously, is presented and the resultsfrom the current work are interpreted within that framework.This is followed by a brief review of spatial memory research, ingeneral, and a fairly detailed review of spatial memory research inpigeons. This information is presented to show that the spatial memoryabilities of pigeons and other species are, when examined closely, moresimilar than is typically presented in the literature.The procedures and theoretical positions adopted here have grownout of previous work conducted in this laboratory and that work isreviewed in detail below (Willson, 1988; Willson & Wilkie, 1991). Inthat work it was discovered that discrimination training between sampleand distractor locations enhanced spatial memory performance. The firstthree experiments of this thesis are primarily concerned with examiningthe mechanisms by which this effect is mediated. It was suggested thatthe discrimination training enhanced attention to the location of thesample and the distractor, an effect that has been observed previouslyin delayed matching to sample with colour and shape stimuli (lJrcuioli &Callender, 1989). The findings from the current research are notentirely consistent with that theoretical position and a reevaluation of2Spatial Memory in Pigeonsthose results in terms of a distinction between recognition andassociative memory is presented.The final three experiments employ a novel associative memoryprocedure for examining spatial memory in pigeons. The main strength ofthe procedure is that it is conceptually simple and offers the subjectflexibility in terms of how to solve the task. In brief, the pigeon isexposed to four illuminated pecking keys and must determine each daywhich key provides reinforcement. Memory for the location of thepreviously rewarded key is assessed during an initial unrewarded periodat the start of each session. The birds show a strong tendency to beginresponding each day at the location that had been rewarded during theprevious session, even when that session occurred 72 hr previously.This level of performance is far beyond that previously seen with thisspecies and may be comparable or better than the performance of food-storing birds on a similar task (cf., Brodbeck et al, in press).The findings, presented briefly above, are discussed in relationto previous work on spatial memory in pigeons as well as in relation tothe issue of adaptive specializations.Some general comments on the study of comparative cognitionIn a general sense, this thesis concerns aspects of comparativecognition. Comparative cognition has been defined as “...the study ofthe minds of organisms. Mind Is the set of cognitive structures,processes, skills, and representations that intervene between experience3Spatial Memory in Pigeonsand behavior.” (Roitblat, 1987 PP. 1-2). Mind in this instance impliesnothing about consciousness, nor does it refer to any non-physicalentity. It is intended only as a convenient term for summarizing agroup of biological structures and processes that are not directlyobservable and must therefore be inferred from the behavior of theorganism.Comparative cognition occupies a somewhat unique position in thebiological sciences because it arose out of very diverse parentdisciplines. The roots of comparative cognition can be found incomparative psychology, ethology and behavioral ecology, cognitivescience, and neuroscience. Scientific endeavors within comparativecognition have often served more than one master. Recent work withfood-storing birds, corvids (i.e., crows, Clark’s Nutcrackers) andparids (i.e., chickadees, tits) nicely illustrates this point (seeKrebs, 1990, for an excellent review of this work).Food-storing birds collect food (generally seeds and/or insects)during times of abundance, cache these food-items and subsequentlyrecover these caches hours, days, weeks, even months later. Cacherecovery is mediated by spatial memory. From the perspective ofcomparative psychology this phenomenal memory performance is of interestbecause it contrasts sharply with the observed performance oftraditional laboratory species such as the rat. Efforts have centeredon examining the similarities and differences between food-storing birdsand other animals (see Sherry, 1984; Shettleworth, 1990). From the4Spatial Memory in Pigeonsperspective of ethology and behavioral ecology, food-storing birds havebeen studied in relation to Optimal Foraging Theory (Krebs, Ryan &Charnov, 1974) and in relation to the issue of adaptive specializations(Krebs, 1990). From the perspective of cognitive science, food-storingbirds have been studied in relation to the theoretical issue of multipleversus unitary memory systems and attempts to generate criteria fordistinguishing between those alternatives (see Sherry & Schacter, 1987).From the perspective of neuroscience, food-storing birds offer a uniqueopportunity for examining correlations between brain structure andbehavior. Members of the corvidae and paridae families differ in thedegree to which they depend on stored food and these differences indegree have been found to correlate well with differences in the size ofthe avian hippocampus in those species. The more reliant the species onstored food the larger the hippocampus (relative to brain and body size,reviewed in Krebs, 1990).Two distinct but related approaches to the study of comparativecognitiàn exist. One, the synthetic approach (Domjan & Galef, 1983;Kamil, 1987), advocates combining the methodology and theory ofpsychology with the insights of behavioral ecologists and ethologistsinto how animals use learning and memory in nature. Closely relatedspecies differing in some aspect of ecology are compared on tasksthought to reflect the abilities that they use in nature, anddifferences are attributed to the differences in ecology. In reality,this approach has been rather difficult to apply, given that it is5Spatial Memory in Pigeonsprobably impossible to find closely related species that differ in onlyone or a limited number of aspects of their ecologies. Any inferencesdrawn must be interpreted cautiously. Nonetheless, work comparingscoring and non-storing corvids (Balda & ICamil, 1989; Olson, 1991) andcomparing storing and non-storing parids (Krebs, Healy & Shettleworth,1990; Shettleworth, Krebs, Healy & Thomas, 1990) has provided promising,if somewhat complex, results at both the behavioural and anatomicallevel.The more traditional approach to comparative work (Bitterman,1975) advocates comparing very different species on traditional learningtasks. This is the most dominant approach adopted within comparativecognition but its utility has been questioned (Macphail, 1982) on thegrounds that it has, thus far, shed very little light on the issue ofspecies differences in cognitive abilities. Although Macphail does notadvocate rejecting the traditional approach, he suggests that anydifferences that have been detected are probably attributable todifferences in sensory or motor ability (Macphail, 1987).One reason for this failure to detect cognitive differences isthat the comparisons needed to make inferences about species differenceswithin this approach has generally not been made. Comparisons betweenthe performance of different species on a single task severely limit thetypes of conclusions that can be drawn, usually to the extent ofconcluding that species A is better than species 8 on task C.Comparisons across several related tasks fare no better, unless the6Spatial Memory in Pigeonsdifferences between species vary systematically as a function of task.If variation exists, a comparison of task demands can provide insightinto the cognitive processes underlying behavior. An especially strongcase can be built by comparing species on related tasks on which thedifferences in performance are reversed (i.e., species A is better ontask C, species B is better on task D). This approach, calledsystematic variation, (Macphail, 1987) has been underutilized withincomparative cognition.Work on pigeon spatial memory, reviewed below, has not adopted thesystematic variation approach, although more recent work has progressedin this direction (i.e., Spetch & Edwards, 1986; Wilison & Wilkie,1991). The conclusions advanced to date concerning pigeon spatialmemory have been limited to the form “pigeons are apparently inferior tomany other species on spatial memory tasks”. Reasons for their apparentinferiority have often been couched in terms of an appeal to some aspectof pigeons’ foraging ecology (see Bond, Cook & Lamb, 1981). Analternative explanation, initially proposed by Macphail (1982, 1987), isthat procedural variables and/or sensory-motor ability differencesaccount for all apparent differences in cognitive ability betweenspecies. In the specific case of pigeon spatial memory the weakness ofthis explanation is that attempts to design procedures on which pigeonsperform at levels comparable to other species have, thus far, met withonly limited success.7Spatial Memory in PigeonsThe position adopted in this thesis is not as extreme as theposition adopted by !1acphail (1982, 1987). Specialized cognitiveabilities do exist (i.e., song learning). However, there are alsocognitive abilities that are general. Spatial memory falls into thelatter category. This hypothesis is at odds with the recent suggestionthat spatial memory in food-storing birds is an example of an adaptivespecialization (i.e., Urebs, 1990) but the findings from the currentwork showing levels of performance in pigeons far better than theirperformance on more traditional spatial memory tasks suggest thatspatial memory as a general process is tenable.The first section of this thesis builds upon previous work fromour laboratory (Wilison, 1988; Willson & Wilkie, 1991) examining spatialmemory in the pigeon. In that work we demonstrated that pigeons’ weremuch more proficient on some spatial memory tasks than previouslybelieved. These experiments will be reviewed in detail below. Severalquestions remained unanswered in that research program and one of thegoals of the experiments described here is to address these unresolvedissues. Specifically, the issue of the mechanism by which the improvedperformance (relative to previous work) observed was mediated isexplored. It is concluded that the abilities tested by the proceduresemployed here mostly closely resemble recognition and associative memorytasks. It is also concluded that the improved performance observedfollowing discrimination training is partially the result of enhancedattention to the relevant sample dimension (i.e., spatial location) but8Spatial Memory in Pigeonsthat this effect works in conjunction with the formation of associationsbetween particular locations and the presence or absence of food.This initial section begins with a a brief, general review of thestudy of spatial memory. The study of spatial memory in pigeons is thendiscussed in some detail, as are the results of our previous work. Thesix experiments that compose this thesis are then presented. The firstthree focus on the mechanisms that mediate the enhanced performanceobserved previously. The latter three experiments focus on a novelprocedure for studying spatial memory and data are presented thatsuggest that the spatial memory abilities of pigeons are far better thanpreviously thought. The thesis concludes with a discussion of theresults and their implications for the comparative study of spatialmemory and the issue of adaptive specializations in learning and memory.A Brief Review of Spatial Memory ResearchThe use of mazes to assess animal intelligence dates back to Small(1901) who used a miniature version of the Hampton Court maze to examinemental processes in the rat. Small used a maze because of its apparentsimilarity to the rat’s natural environment (i.e., a series ofinterconnected burrows). He believed that one could accurately assessan animal’s intelligence only through the use of apparatuses andprocedures resembling the natural environment and the problems that ananimal might encounter there. This idea, supported by many researchers9Spatial Memory in Pigeonstoday, temporarily disappeared under the onslaught of the behaviorists’search for general principles of learning through the use of arbitrarilychosen stimuli, responses and testing environments, so that today themain contribution made by Small’s research was the introduction of theuse of mazes into comparative psychology, as well as the use of thecomparative psychologists’ favorite subject, the albino rat.The use of mazes (and rats) in the study of learning quicklyproliferated. Researchers, such as E.C. Tolman, who favoured a molarview of behavior as flexible and purposive, used mazes of variousconfigurations to attack the molecular, reflex-based behaviorism of Hulland his associates. Tolman and his colleagues demonstrated such diversephenomena as latent learning (Tolman & Honzik, l930a,b), place learning(Tolman, Ritchie, & Kalish, 1946b) and short cut learning (Tolman,Ritchie & Kalish, l946a), phenomena that were hard to reconcile with astrict S-R view of behavior. Tolman (1932) also introduced the idea ofthe “cognitive map”, an idea that will be discussed in some detailbelow.Researchers who supported the molecular S-R view of behavior alsoused mazes but their emphasis was on the automatic nature of mazerunning in well trained rats. For example, in one experiment rats weretrained to run a maze by making a series of left-hand turns. One of theleft-hand arms was subsequently blocked and the opposite arm madecorrect. Every subject initially failed to notice the change and “everyone of the rats banged his nose into the end of the blind alley with10Spatial Memory in Pigeonsconsiderable violence” (Gingerelli, 1929, p. 255). Similarly, welltrained rats ran over a pile of food placed in the middle of an alley,seemingly oblivious to its existence (Stoltz & Lott, 1964). Thesefindings were hard to reconcile with the flexible view of behavioradvocated by Tolman, and the molecular S-R view of Hull and hiscontemporaries (Hull, 1943) came to dominate until well into the 1960’s.With the advent of the automated operant chamber, mazes and spatiallearning tasks fell out of general use.However, in the last 15 years the use of mazes and spatiallearning tasks has once again come to the fore in animal learning andideas first expressed by Tolman, such as the cognitive map, havereceived considerable attention. This reorientation was the result of avariety of factors. One such factor was the recognition of thedistinction between “knowing” and “doing”; for example animals oftenhave knowledge that is not manifested in overt behavior (e.g., Wilkie &Masson, 1976). A second factor was the introduction of tasks such asdelayed matching to sample (DMTS - Blough, 1959) that could be solvedonly by responding on the basis of information that was not longerphysically present and thus could be explained only by reference to sometype of representation. A final factor was the rise of a more cognitiveorientation in the study of human learning and this influence slowlyspread to the study of animal learning. Many of the tasks devised toexamine learning in animals during the 60’s and 70’s were meant asanalogs of tasks used to study learning and memory in humans.11Spatial Memory in PigeonsAn example of this was the radial maze introduced by Olton andSamuelson (1976). The wagon-wheel shaped, eight-armed apparatus thatthey used to study short-term memory in the rat was initially viewed asa spatial analog to the learning of word lists in human cognitivepsychology. In their procedure a rat was allowed to explore the mazefreely until it had found all of the food hidden at the distal end ofeach arm. All arms were baited with a piece of food at the beginning ofa trial. Olton and Samuelson’s subjects proved to be very proficient atthe task, quickly learning to visit each arm once without repetition.Various control manipulations ensured that the rats’ performance wasbased on memory, not non-memorial strategies such as response algorithmsor odor trails.Since that seminal paper, work on spatial memory has concentratedon specifying its characteristics both within and between species.Considerable debate has arisen from both of these approaches and eachwill be discussed in turn.Much of the debate about the characteristics of spatial memory hascentered on how spatial information is represented in memory. Olton(1978, 1979) proposed that spatial memory was best conceptualized as aworking memory with the following characteristics: 1) the capacity islarge but limited, 2) accuracy declines as a function of memory load, 3)information does not decay over time, or at least very slowly, 4) thereare no primacy or recency effects, 5) spatial memory can be reset at the12Spatial Memory in Pigeonsend of a test, and finally 6) storage is in the form of a list in whicheach place is individually represented.However, several of these assumptions have been questioned,specifically the idea that spatial memory can be reset and that thenature of the representation is list-like. Evidence against the idea ofresetting come from experiments showing proactive interference (Wright,Urcuioli & Sands, 1986) which will be examined first.Spatial memory has proven highly resistant to interference, butrecent experiments by Roberts and Dale (1981) have demonstrated thatproactive interference can occur under some conditions. Their subjectsreceived five trials per day on an eight arm radial maze. Roberts andDale examined error patterns within the five trials. The rats nevermade errors on their first or second choices of any trial. However, theprobability of making errors on the third through fifth choices clearlydiffered between the second and subsequent trials. On the first trialperformance remained error free until after the fifth choice. Onsubsequent trials performance started to decline after the secondchoice, a finding consistent with the presence of proactive interferenceand incompatible with the notion of a resetting mechanism.In an additional experiment Roberts and Dale showed that imposinga retention interval between a forced and free choice had differentialeffects depending upon whether the rats were being given their first,second, or third trial of the day. Accuracy was much lower following aretention interval on the second or third trial of the day relative to13Spatial Memory in Pigeonsaccuracy on the first trial, a finding consistent with the presence ofproactive interference and inconsistent with the notion of a resettingmechanism.As a final bit of evidence against the notion of a resettabieworking memory, Dale and Roberts noted that their subjects tended toavoid alleys chosen during the latter part of trial N-i when makingtheir initial choices on trial N. If the rats were resetting workingmemory at the end of each trial initial choices on trial n should havebeen uncorrelated with choices on trial N-l. Clearly the rats wereremembering their final choices on trial N-l and avoiding those arms atthe beginning of trial N (see also Roitbiat & Harley, 1988).The list-like nature of spatial memory has also been questioned.The notion of a cognitive map, an idea first proposed by Tolman (1948),has received considerable support. This evidence will be reviewed belowbut some initial comments about what constitutes a cognitive map are inorder. A detailed examination of this issue is beyond the scope of thepresent discussion but a clarification of the terminology is warranted.Tolman’s choice of the term “cognitive map” was somewhatunfortunate because it implied the existence of a map per se. Tolmanwas using the concept as an analogy but it nonetheless led to images ofcartographic maps located in the brain and also to images of homunculiporing over these maps. A better terminology would have been to referto the process of cognitive mapping, a process rather than a structure.In essence, mapping refers to the application of a lawful system for14Spatial Memory in Pigeonsrepresenting information. In the cartographic sense, mapping refers tothe process of transforming an object set (physical space) into arepresentational set via a mapping function. The set definitionestablishes relations between places and the function specifies therelationships in physical space that will be maintained in therepresentational space (Downs, 1979). Note that mapping does not referto the medium in which the representation is expressed. Maps are theend product of mapping but the physical medium of the map is largelyindependent of the mapping process. For example, one could use the samemapping system to draw a map on a piece of paper, a blackboard or on thebeach. Of course the physical characteristic of the map can influencethe accuracy and durability of the representation but this isindependent of the rules by which the representation is constructed.For the most part investigations of cognitive mapping havefocussed on trying to specify the mapping function rather than themedium in which the mapping is expressed (see O’Keefe & Nadel, 1978 fora notable exception) and the term cognitive map has been used as ashorthand for how animals represent their physical world. As notedearlier the idea of a cognitive map was first introduced by Tolman andhe and his associates conducted several experiments aimed atdemonstrating that animals possessed far richer representations ofphysical space than suggested by the strict S-R theorists. Thephenomena of shortcut learning (Tolman, Ritchie & Kalish, l946a) andplace learning (Tolman, Ritchie & Kalish, 1946b) were hard to reconcile15Spatial Memory in Pigeonswith S-R accounts of maze learning. They are also hard to reconcilewith the list-like nature of spatial memory proposed by Olton (1978).However, these experiments proved to be hard to replicate (Gentry, Brown& Kaplan, 1947; Gentry, Brown & Lee, 1948) and have been justifiablycriticized on logical grounds (Olton, 1979) so will not be describedhere. However, as is sometimes the case in science, ideas initiallyrejected as untenable later turn out to be well founded. Such was thecase for the idea of cognitive mapping. More recent work has providedconvincing demonstrations of shortcut learning and place learning (foran excellent review of these issues see Gallistel 1990, especiallyChapter 5).Menzel (1973, 1978), working with chimpanzees, demonstrated thephenomenon of shortcut learning in the following way. In one experimenta handler carried a chimp around a large enclosure and allowed the chimpto watch from close range while an experimenter hid food in 18locations. The route taken on this food hiding expedition wasconvoluted and often recrossed paths already taken. Once all the foodwas hidden, the chimp was returned to its home cage and then released.The questions of interest were: 1) would the chimp remember where thefood had been hidden and 2) if so, how would the chimp go aboutharvesting the hidden food? The answer to both questions was clear.The chimps remembered about 2/3rds of the hiding places and furthermoredid not follow paths similar to those used during food-hiding, butminimized the distance travelled in collecting the food from the16Spatial Memory in Pigeonsremembered locations (see MacDonald & Wilkie, 1990 for a similarexperiment with New World monkeys). These results are consistent withthe notion of a cognitive mapping process in which the relationshipsbetween important locations are preserved (the nature of therelationships that are preserved will be addressed briefly below) andare hard to reconcile with a list-like representation of importantlocations as proposed by Olton (1978).Place learning was demonstrated in an ingenious experiment byMorris (1981). His apparatus, the “water maze”, consisted of a circularswimming pool filled with cool, opaque water. A single location in thepooi contained a small, submerged platform and his subjects, rats, werereleased into the pool and allowed to swim around until they locatedthis hidden platform. Swimming, especially in cool water, is mildlyaversive to rats and the platform provided a means of escaping thewater. After a few trials the rats swam directly to the position of theplatform regardless of whether they were released from the originaltraining position or, as was demonstrated in subsequent transfer tests,from a novel position; even if the platform was moved to a new location.When they failed to find the platform in its expected location theyspent some time searching near that location before expanding the searcharea. Clearly, the rats learned a place rather than a response and thatplace was defined in terms of the relationships between environmentalcues outside of the pool (since there were no local cues present).17Spatial Memory in PigeonsWork by Gould (1984, 1986, 1987) with bees and by Suzucki,Augerinos and Black (1980) with rats also supports the notion ofcognitive mapping. Previous work with bees has shown that bees capturedas they leave the hive and transported in darkness to a feeding sitehundreds of yards away and out of sight of the hive can, when released,fly straight back to the hive (Gould, 1984). In an additionalexperiment, foragers were transported to a feeding site that was eitherin the middle of a lake or on the far shore. After feeding they wereallowed to return to the hive and dance. Dancers that had been fed onthe far shore of the lake successfully recruited other bees. Dancersthat had been fed in the middle of the lake did not, suggesting that thepotential recruits extract some information about location from thedance and make judgements on it about the suitability of the advertisedsite (Gould, 1984). Although suggestive of cognitive mapping, thesefindings can also be explained by reference to. “route-specific memory”(Wehner, 1981).A more convincing demonstration of cognitive mapping comes from anexperiment in which Gould (1986) trained individual bees to forage at afeeding station (A). The bees quickly learned to fly straight from thehive to the feeding station. Gould subsequently captured these foragersas they left the hive and transported them, in darkness, to another site(B). When released, the bees headed directly towards Site A, theiroriginal destination, even though it was not visible and the bees hadnever flown between Sites A and B before. Various control conditions18Spatial Memory in Pigeonswere run that insured that the observed performance could not beexplained by anything other than cognitive mapping.In a series of experiments, Suzuki, Augerinos and Black (1980)examined the role of extramaze cues in spatial memory. They trainedrats on an eight-arm radial maze that was enclosed in a largecylindrical chamber on whose walls extramaze cues could be mounted.Extraniaze cues facilitated performance. Rats trained without extramazecues tended to respond in stereotyped response chains. Transposition ofthe extramaze cues disrupted performance but rotation of the entireextraniaze cue array did not. Suzuki et al concluded that their subjectsused a map-like representation of the extramaze cues to remember thearms on the maze that they had previously visited.More recent work has focussed on the mapping function (i.e., Cheng& Gallistel, 1984; Gallistel, 1990); the types of relationships in thephysical environment that are maintained during cognitive mapping.Although it seems highly unlikely that animals would form incorrectrepresentations of space, their representations could vary incompleteness.For example, Cheng and Gallistel (1984) have described a mappingfunction in which only a small subset of Euclidean properties aremaintained. In this system important locations are assumed to lie at apoint that is an intersection of several straight lines, each of whichhas two other distinct landmarks on it, one on each side of theimportant location. The animal represents the fact that the important19Spatial Memory in Pigeonslocation lies at the intersection of lines between the pairs oflandmarks. The animal arrives at the goal by attempting to positionitself between the various pairs of landmarks. In this system orderproperties of space are represented, but metric properties of space suchas distance and angular separation are not. Cheng and Gallistel havesuggested that this might be the type of representation that diggerwasps use to find their burrows (Thorpe, 1950).Additional work by Cheng and Gallistel also demonstrated that ratsrepresent metric properties of space. Rats were trained in an X-shapedmaze in which the distal end of each arm was baited with food. One armcontained 18 food pellets, another six, another one and the last wasempty. Each arm was associated with a unique landmark. The subjectsquickly learned to visit the 18 pellet arm first followed by the sixpellet arm, etc. Cheng and Gallistel then performed a series of affinetransformations. Affine transformations are transformations thatmaintain all non-metric properties of space. This was accomplished bymoving each landmark to a neighboring corner. This transformationseverely disrupted performance suggesting that the rats wererepresenting metric properties of space.In summary, current conceptions of spatial memory suggest thatspatial information is represented in “map-like form” and that manyanimals represent metric as well as relational properties of space.Locations are encoded in terms of nearby landmarks. Spatial memory is20Spatial Memory in Pigeonshighly resistant to interference effects and has a large, but probablylimited capacity.Comparative Spatial CognitionThe vast majority of spatial memory research has been conductedwith rats. However, examining spatial memory in other species isimportant. If we assume, reasonably, that some form of spatial memorydeveloped early on in evolutionary history, its evolutionary history islikely to have been highly convoluted with many false starts,divergences and convergences. We can learn much from studying thevariations. An example should help clarify this point.Consider birds and bats. Both can fly. However, the means bywhich they accomplish flight are very different. This is a clearexample of convergent evolution; different mechanisms for solving thesame problem. From a structural point of view comparing the two is likecomparing apples and oranges. The wing structures are very different.However, from a functional point of view comparing the two can be veryuseful because both mechanisms accomplish the same goal. They get theanimal off the ground and keep it aloft. If we are interested in flightper se we can learn a lot by comparing the similarities and differences.We can learn what aspects of the ability are general and what aspectsare specific to the particular structures involved. The same can besaid of any cognitive ability, including spatial memory.21Spatial Memory in PigeonsResearch on spatial memory has focussed on examining differencesand similarities between species (these comparisons have until recentlybeen accomplished by examining a series of single species experiments).Even a brief survey of the range and number of species tested would bebeyond the scope of the present thesis. People (Aadland, Beatty & Maki,1985), birds (Balda, 1980), fish (Roitbiat, Tham & Golub, 1982), otherrodents (Wilkie & Slobin, 1981) and even insects (Gould, 1984) have allbeen tested on a variety of different spatial memory tasks and the listcontinues to grow. To summarize briefly, these various experiments haverevealed that animals form representations of space and that thestrength or completeness of those representations vary (see Gallistel,1990, ch. 1-6, for a good overview of much of this research). Inaddition, these comparisons have revealed striking differences in thecapacity and durability with which spatial information is represented.On the one hand, we have the rather remarkable spatial memory ofthe food-storing birds mentioned previously. Their ability to rememberhundreds of cache sites for extended periods of time has frequently beencontrasted with the abilities of other species (Sherry, 1984; Sherry &Schacter, 1987; Shettleworth, 1985) and has led to speculation that thisability may be an adaptive specialization. As mentioned previously, theexistence of adaptive specializations has proved to be rather difficultto pin down at the behavioral level and to date the sparse experimentalevidence that does exist has been messy (see Krebs, 1990).22Spatial Memory in PigeonsOn the other hand, we have another favorite laboratory species,the pigeon. The spatial memory abilities of pigeons, when tested withlaboratory procedures, are apparently rather limited (see Bond, Cook &Lamb, 1981). This contrasts sharply with the remarkable spatialabilities demonstrated by this species outside the laboratory (i.e.,homing). Since this paradox is likely an artifact of the testingprocedures in the laboratory, these procedures will be examined in somedetail in the next section.SPATIAL MEMORY RESEARCH IN PIGEONSThe approach adopted in this section is to examine the majorexperiments on spatial memory in pigeons, individually and in somedetail, in a roughly chronological order. This previous work can bebroadly classified as maze studies and delayed matching of key locationstudies (DMKL, see Wilkie & Summers, 1982). Both lines of research arerelevant to the present thesis and will be examined separately. Thereare many apparent contradictions in the data but in a later section themajor findings are summarized and some tentative reasons for theseapparent differences in pigeOns’ spatial memory abilities are discussed.Readers interested in only the main themes of the research can safelyproceed to this later section (p. 51). Work from our own laboratory onpigeon spatial memory, upon which the current research is based isexamined separately in a later section.23Spatial Memory in PigeonsBOND, COOK AND LAMB (1981)One of the first systematic examinations of spatial memory inpigeons was conducted by Bond, Cook and Lamb (1981). They compared theperformance of pigeons and rats on a radial arm maze. An ideapredominant at the time, and still predominant (see Roitblat, 1987), wasthat an animal’s foraging ecology would influence the evolution of itscognitive abilities. One variation on this theme was what Bond et altermed the “resource-distribution hypothesis”. Olton and Schlosberg(1978) had suggested that animals for whom food resources were diffuselydistributed, irregularly available and easily depleted shouldsystematically avoid recently exploited resources and that these animalsshould demonstrate proficient working memory for recently exploitedlocations (Bond et al called this event memory). The successfulperformance of rats on the radial maze reflected this “win-shift”tendency. Individuals for whom food resources were concentrated anddependable should tend to return to the same site, to use a “win-stay”strategy and poor working memory because the problem can be solvedsolely through the use of reference memory.There was abundant evidence that some animals could use a “winshift” strategy (Gill & Wolf, 1977; Kamil, 1978; Olton & Schlosberg,1978) and that this ability was mediated by memory. However, evidencefor poor working memory in a species that exploited dependable andabundant food sources was lacking. Bond et al examined the foraging24Spatial Memory in Pigeonsecology of the domestic pigeon and concluded that the pigeon seemed alikely candidate to exhibit this tendency.Pigeons feed in locations that are not readily depleted in asingle foraging bout, are gregarious foragers that use the presence ofother birds as a signal for the occurrence of locally abundant foodresources and use traditional feeding sites (Goodwin, 1967).Furthermore, field observations had suggested excellent reference memoryfor previously profitable sites even after absences of a year or more(Levi, 1974). Subsequent work by Vaughan and Green (1984), mentionedpreviously, has confirmed that pigeons have excellent referencememories. For these reasons, Bond et al decided to compare rats’ andpigeons’ performance on the radial arm maze.The results of Bond et al’s experiment at least partiallyconfirmed their hypothesis. Pigeons were far inferior to rats. Infact, the performance of the pigeons was so poor that the authors wereunable to rule out the possibility that the pigeons were using some non-memorial strategy to solve the radial maze task. They recognized thattheir experimental design was weak and considered the possibility thattask variables or other ecological variables could also have beenresponsible for the observed differences between rats and pigeons.WIllIE, SPETCH AND CHEW (1981)At about this time, Wilkie, Spetch and Chew (1981) published areport on short-term memory for location in the ring-dove, a species25Spatial Memory in Pigeonsclosely related to the pigeon. Their apparatus was a modified radialmaze, consisting of two parallel levels of seven arms. The levels wereseparated by a vertical distance of 30 cm and the entrances to adjacentarms on a given level were also 30 cm apart. The arms of the mazeconsisted of tubes 8 cm in diameter and 14 cm long. There was a perchat the entrance to each arm but food was visible only once an arm wasactually entered. At the beginning of each session all arms were baitedwith a small amount of food and sessions continued until a subject hadmade 14 choices. A choice was defined as landing on a perch in front ofa tube.Although initial training was extensive (approximately 100sessions), at asymptote the bird performed at levels far above chance.On average, the birds were correct on 80% of their 14 choices (chancewas 64.5%). Interestingly, Wilkie et al also considered the role offoraging ecology in the evolution of spatial abilities and concludedthat the conditions were quite conducive to the evolution of aproficient spatial working memory.OLSON AND MAKI (1983)An additional experiment questioning the generality of Bond etal’s findings was presented by Olson and Maki (1983), who demonstratedthat pigeons could perform well on a delayed alternation task. Thistype of task requires both an accurate working memory and the use of a“win-shift” strategy. Olson and Maki trained pigeons to perform a26Spatial Memory in Pigeonsdelayed alternation task on a T-maze. Each trial began with a forcedchoice in which the subjects were allowed to enter the unblocked arm ofthe maze and consume a small amount of grain. They were then removedfrom the maze, returned to the start box and then given a free choicebetween the two arms. Choice of the novel arm was reinforced but choiceof the familiar arm was not.Pigeons acquired this task quickly, and their accurate respondingdepended upon the use of extramaze cues. Furthermore, they couldtolerate delays between the forced and free choices of up to 16 mm.Also of interest was the fact that the pigeons readily learned a “winshift” version of the task but not a “win-stay”. These same authorsfailed to find good performance by pigeons on a radial arm maze(unpublished observations described in Olson & Maki, 1983).ROBERTS AND VAN VELDHUIZEN (1985)Roberts and Van Veidhuizen’s (1985) experiment was motivated inlarge part by the contradiction between the results of the threeexperiments described previously. They reasoned that pigeons’performance on the radial maze might be improved if initial trainingtook the form of a gradual introduction to the requirements of the task.Their subjects were initially trained with only two arms of the mazeavailable. The number of available arms was gradually increased untilthe birds were choosing freely amongst all eight arms (only four armswere used in some experiments).27Spatial Memory in PigeonsThe results of a series of seven experiments were reported. Inthe first experiment it was demonstrated that pigeons could remembervisits to four arms, a level of performance better than that suggestedby the research of Bond et al. In the second experiment, pigeons wereallowed to choose freely amongst all eight arms and Roberts & VanVeldhuizen concluded that the pigeons performed at a level comparable tothat seen in rats. The third experiment demonstrated the phenomenon ofproactive interference (P1). When tested with massed trials, the birdsaccuracy on later trials was inferior to their accuracy on earliertrials. Roberts and Dale (1981) had previously demonstrated thisphenomenon in rats.In their fourth experiment, Roberts and Van Veldhuizeninvestigated the effects of imposing a retention interval between fourforced choices and an opportunity to choose freely between all eightarms. Performance declined following a delay but was still above chanceat the longest retention interval tested (5 mm). In the fifthexperiment, the question of whether the observed radial maze performancewas based on the use of intramaze or extramaze cues was investigated.Performance dropped when intramaze cues were removed but still remainedwell above chance. The authors concluded that the pigeons were usingboth intramaze and extramaze cues. The final two experiments examinedreference memory. In one experiment only four of the arms of the mazewere baited but the same arms were baited each day. The pigeons quicklylearned to visit only the baited arms and rarely re-visited an arm28Spatial Memory in Pigeonsduring a trial once it had been visited. In the last experiment, fourarms were again baited each day but the arms differed in terms of howmuch grain they contained at the beginning of the session. One alleyalways contained 9 g of grain, another 3 g, another 0.5 g and the lastwas empty. The birds quickly learned to visit the alleys in a sequencecorresponding to the amount of grain located there. Initial choiceswere most often directed to the 9-g alley, second choices were usuallydirected towards the 3-g alley, etc. The now unbaited alley was rarelyvisited. These results clearly indicated that the pigeons quicklyformed a reference memory of which alleys did and did not contain foodand, furthermore, that they could also learn associations betweenparticular locations and particular amounts of food. Similar findingshad been observed with rats (Hulse & O’Leary, 1982; Olton & Papas,1979).Roberts and Van Veldhuizen concluded that spatial memory wassimilar in pigeons and rats, differing only in capacity (perhaps) anddurability. To explain the necessity of special training the authorsappealed to the notion of “preparedness to learn” (Selignian, 1970). The“resource-distribution” hypothesis, mentioned above, is an example ofthis type of hypothesis (i.e., because of evolutionary pressure certainanimals are predisposed to learn particular tasks). Roberts and VanVeidhuizen suggested that the confining nature of radial mazes and theirstrictly defined pathways might initially inhibit pigeons’ performance,29Spatial Memory in Pigeonsgiven that pigeon foraging behavior has evolved to cope with foraging inopen, unconfined spaces.SPETCH AND COLLEAGUESSpetch and her colleagues have conducted a number of studiesexamining spatial memory in pigeons using a modified version of theradial maze. The “walking maze” consists of a number of feedingstations in a large room. The pigeon’s task is to visit each feedingstation and collect the grain contained there. The path between feedingstations is not constrained in any way and the feeding stations can beset up in any configuration. This open-field type of arrangement wasdesigned to be more similar to the pigeons’ natural foragingenvironment.Spetch and Edwards (1986)Spetch and Edwards’ first attempt to study pigeons’ spatial memoryin an open-field type of environment was a qualified success. Theirinitial apparatus consisted of eight wall-mounted feeders and the birds’task was to visit each feeder by flying up to a perch mounted on thewall. The birds learned the task quickly and performed above chancelevels of accuracy without special training but the procedure wasproblematic. The subjects often took considerable amounts of timebetween choices, in effect imposing their own retention intervals.Baseline performance, although above chance, was low. These factors30Spatial Memory in Pigeonsmade it rather difficult to make some inferences about what the birdswere doing and how they were doing it (i.e., how temporally persistentwas their working memory?). Goodwin (1967, 1983) had noted that pigeonsare predominantly ground feeders and that the tendency to go to theground to feed might be innate (Goodwin, 1954). These observationssuggested to Spetch and Edwards that a ground level open-fieldarrangement of feeders might be a better set-up for examining spatialmemory in pigeons. This “walking maze” open-field arrangementalsoafforded the authors a greater degree of flexibility in terms of thespatial organization of the testing environment. Homer (1984) hadpreviously shown that the structure of a maze can influence bothaccuracy and the tendency to use non-memorial strategies on spatialmemory tasks. Spetch and Edwards wished to see if pigeons would show asimilar tendency.The results of their second experiment showed clearly that thebirds were using spatial memory to solve the open-field task and thatperformance was largely controlled by extramaze cues. The configurationof the maze (they compared circular and linear arrangements of feeders)did influence the pigeons’ patterns of responding: when the feeders wererelatively far apart the birds showed a tendency to minimize traveldistance (Menzel, 1973, 1978) but this tendency could not completelyaccount for the observed level of performance.31Spatial Memory In PigeonsSpetch and Edwards (1988)Spetch and Edwards also used a version of the “walking maze” toexamine pigeons’ use of global and local cues in spatial memory. Globalcues refer to those cues that are macroscopic or large-scale features ofthe environment (Gallistel, 1990). Local cues are cues particular tospecific places within the environment. This distinction is similar tothe extramaze-intramaze cue distinction used previously. The subjectswere initially trained in a situation in which both global and localcues could be used to locate a baited feeder between two unbaitecifeeders. The middle feeder was always baited and the three feeders werealways located in the same room position. In this arrangement, the twounbaited feeders served as local cues and the features of the testingroom served as global cues.In a series of subsequent unreinforced tests, the testingenvironment was manipulated in various ways to examine the birds’reliance on global and local cues. In one test, global and local cueswere pitted against each other by shifting the complete array of feederslaterally. After the shift, one of the end feeders was in the correctroom position (as defined by global cues) but the middle feeder wasstill in the correct location based on the configuration of the localcues. The pigeons showed a strong tendency to choose the middle feeder,suggesting that their choice behavior was controlled by the local cues.In another test, local cues were eliminated by removing one of the endfeeders. Nonetheless the birds nearly always chose the feeder located32Spatial Memory in Pigeonsin the correct room position, a finding that suggests that the birdscould also respond based on the configuration of global cues. In afinal test, global cues were eliminated by moving the test array to anovel room position. The pigeons’ strong tendency to visit the middlefeeder on these tests could be explained only by the use of local cues.On the basis of these results, Spetch and Edwards concluded that pigeonsencode both global and local cues but that the type of cue used to guidebehavior is situation specific with redundant cues being hierarchicallyorganized.Spetch and Honig (1988)Spetch and Honig also conducted a series of experiments using the“walking maze”. Previous work with rats had suggested that mazeperformance was best under conditions in which the animal was allowed tob.form a cognitive map of the testing environment (Suzucki et al, 1980).Rearranging the positions of the available cues within the testingenvironment between trials inhibits (but does not eliminate) accurateperformance of the radial maze task. Spetch and Honig examined whethera similar phenomenon would be observed with pigeons. They were alsointerested in how temporally persistent spatial information was inpigeons’ working memory. Previous work (Olson & Maki, 1983; Roberts &Van Veidhuizen, 1985; Spetch & Edwards, 1986) had examined this issuewith different apparatuses (Spetch and Edwards used the “flight” maze)but it had not been examined using the “walking maze”.33Spatial Memory in PigeonsIn one experiment, they compared two groups of pigeons. For onegroup, the features of the testing environment remained constant fromtrial to trial (the Constant group). For the other group, the featuresof the environment were rearranged between trials (The Variable group).It was not possible to rearrange all features of the environment (i.e.,walls, doors, windows, etc.), so only a subset of those features (i.e.,those that were portable) was rearranged. The Constant group clearlyperformed better than the Variable group. Honig and Spetch concludedthat an intact cognitive map facilitated performance on the workingmemory problem although it was not necessary for performance. They alsoconsidered the possibility that the observed differences might have beendue to differences in the “richness” of the map. Because it wasimpossible to manipulate all features of the testing environment, it ispossible that the performance of the Variable group was also mediated byan intact cognitive map of the environment, albeit one based on fewercues.In their second experiment, Spetch and Honig examined the temporalpersistence of spatial information on the walking maze task. Trialsbegan with four forced choices. Once a bird had visited all fouravailable sites it was removed to a small holding cage. Following adelay of up to 2 hr, the pigeon was released into the testingenvironment and allowed to choose freely between all eight feedingsites. On average, the birds performed well with delays of up to 32 mmbut performed at chance levels when tested with a delay of 2 hr.34Spatial Memory in PigeonsSpetch (1990)Roberts and Van Veidhuizen (1985) suggested that pigeons’performance on spatial memory tasks might be influenced by memory load.Dale (1988, see below) has reiterated this point. Memory load refers tothe number of items currently held in working memory. The suggestion isthat temporal persistence and memory load are negatively correlated.The greater the memory load the lower the temporal persistence and viceversa. Roberts and Van Veldhuizen cited this as the main factorunderlying the difference in temporal persistence observed in the workof Olson and Maki (1983) and their owti research. In the former case thememory load was assumed to be one (i.e., one arm of a T-maze) andtemporal persistence was at least 16 mm. In the latter case, thememory load was assumed to be four (i.e., four forced choices on aneight arm radial maze) and performance dropped to chance levels after adelay of only 5 mm. Spetch directly tested this hypothesis by varyingthe number of forced choices given to pigeons on the walking maze priorto a retention interval.She found no effect of memory load. Memory for previously visitedsites declined as a function of increasing retention interval (RI) butthe forgetting function was similar under all memory load conditions(two, four, or six forced choices prior to the RI). In all conditionsperformance was still above chance levels with a 60-mm RI. In anadditional experiment she demonstrated that the observed performance was35Spatial Memory in Pigeonsdue to spatial memory and that control of responding was based on theglobal cues inherent to the testing environment.DALE (1988)Dale has also examined spatial memory in pigeons using a four armradial maze constructed of chicken wire. His choice of this particularapparatus was founded on an interest in four aspects of pigeons’performance on maze tasks.The inferior performance of pigeons on the radial maze relative totheir performance on the walking maze has often been attributed to theconstrained nature of available response paths in the radial maze (i.e.,subjects must return to the center platform between choices, Spetch &Edwards, 1986). Dale reasoned that another potentially confoundingfactor was the availability of extramaze cues. Radial mazes, at leastthose used with pigeons, restrict the view of the surroundingenvironment. By constructing his maze of chicken wire, Dale removedthis restriction but left the response paths constrained.Dale was also interested in the memory load hypothesis (seeprevious section). He reasoned that performance on a four arm mazeshould fall somewhere between that observed in a T-maze (Olson & Maki,1983) and on an eight arm radial maze (Roberts & Van Veldhuizen, 1985)if this hypothesis had merit. His study preceded that of Spetch (1990)so he was unaware of her results.36Spatial Memory in PigeonsSpetch and her colleagues had examined the importance of intramazeand extramaze cues in pigeon spatial memory in a variety of ways (Spetch& Edwards, 1986, 1988; Spetch & Honig, 1988; Spetch, 1990) but thenature of the possible transformations that they had examined wasconstrained by the nature of the apparatus that they used (i.e., it wasdifficult to manipulate extramaze cues). Dale used a rotation procedureto manipulate the relationship between intramaze and extramaze. In thisprocedure once the subjects have acquired the task, the maze is rotatedin relation to the surround. This manipulation has been usedsuccessfully to investigate the relative importance of intramaze andextramaze cues in rats (Dale & Innis, 1986; Suzucki et al, 1980).As a final manipulation, Dale examined whether pigeons couldreexamine spatial memory following an error and then respondappropriately. Previous research had revealed that errors in memoryparadigms are not always due to forgetting (Roitblat & Scopatz, 1983;Wilkie & Spetch, 1981).The procedure that he used was fairly straightforward. Pigeonswere given three forced choices, removed from the apparatus and,following a delay, given two opportunities to choose the remaining arm.In some conditions the maze was rotated by 90 degrees.The subjects acquired the task quickly and performed well withdelays of up to 5 mm. Accuracy on second choices was above chancelevels with RI’s of up to 30 mm. The results of the maze rotationmanipulation suggested that the pigeons responded on the basis of room37Spatial Memory in Pigeons(i.e., extramaze) cues. Following the rotation, the pigeons generallywent to the arm that was “correct” relative to the room cues rather thanthe arm that was “correct” relative to the maze (i.e, the “correct”place rather than the “correct” arm).Dale concluded that spatial memory in pigeons was similar to thatof rats but that rats seemed better able to remember events in spatialmemory in the face of extended delays. Furthermore, he concluded thatmemory load could influence performance (recall that he was unaware ofSpetch’s (1990) results). In addition, he concluded that pigeons couldreexamine spatial memory and respond appropriately based on thatreexamination.ROBERTS (1988)Roberts (1988), using a modified version of the walking maze, hasexamined pigeon spatial memory in a simulated patchy environment. Hediscussed his results in relation to Optimal Foraging Theory (OFT) aswell as memory, but the aspects of his data relating to OFT areirrelevant to the present thesis and are therefore not discussed. Heset up four “patches” within a large room and allowed his subjects toexplore freely. Patches consisted of circular arrangements of eightfeeding stations and were differentiated both by their location withinthe room and by the amount of food that they contained at the beginningof a session. The amount of food available per patch was varied bymanipulating either the amount of food contained per feeder or by38Spatial Memory in Pigeonsvarying the proportion of feeders within a patch that were baited. Thelocations of the patches was held constant from session to session, aswas the amount of food that they contained at the beginning of thesession.Roberts found good evidence of memory for patch characteristicsunder some conditions but not others. When food density per patch wasmanipulated by varying the amount of food contained in each feeder, thepigeons tended to visit the richer patches first and to spend more timethere. They also tended to visit all patches before making repeatvisits. When food density was varied by manipulating the proportion ofbaited feeders per patch, the birds responded similarly but withoutvisiting the richest patch first. Initial patch choice was unrelated tofood density.Even under conditions in which the pigeons demonstrated accuratememory for patch characteristics, they showed little evidence for memoryof visits to feeders within patches. When re-visiting a patch, thebirds were as likely to visit depleted feeders as to visit full ones.Roberts concluded that the birds exhibited excellent reference andworking memory for patches but little working memory for visits withinpatches. He suggested that this poor within-patch working memory mayhave been a result of the similarity between patches. Although thepatches were spatially distinct, each feeder within a patch wasphysically identical (in fact, all 32 feeders were physicallyidentical). Thus the discrimination between feeders within a patch was39Spatial Memory in Pigeonsprobably extremely difficult, and visits to other patches may haveinterfered with the ability to recall visits within a particular patch(proactive interference). Although Spetch and Edwards (1986) haddemonstrated that pigeons could perform well on the walking maze witheight physically identical feeders, Roberts concluded that thecombination of physical similarity and proactive interference wasprobably responsible for the pigeons poor memory performance withinpatches.ZENTALL, STEIRN AND JACKSON-SMITH (1990)Zentall and his colleagues have developed an operant analog to theradial maze. In their task pigeons are confronted with an array of fiveilluminated pecking keys. For some birds, this array was linear (L),for others it was a two-dimensional matrix (H). For some birds all fivekeys were illuminated with white light (W) and for other birds each keywas illuminated with a distinctive hue (H). To receive reinforcement, apigeon had to peck five times consecutively at the same key. These fivepecks constituted a “choice”. Only the first choice of a particular keywas reinforced. The birds were allowed to go through the sequence ofkeys in a self-determined sequence. Once they had chosen all five keys,the chamber was darkened and after a brief intertrial interval (ITI),another trial began. Zentall et al interpolated delays at variouspoints in the choice sequence and manipulated the number of responsesthat constituted a choice.40Spatial Memory in PigeonsAll subjects learned the task but birds in the K group and the Hgroups learned faster. Somewhat surprisingly, birds for whom the keyswere in a matrix and distinctively coloured did not learn faster thanbirds for whom either space or colour were distinct. The number ofresponses that constituted a choice did not affect the speed at whichthe birds acquired the task. It did, however, affect how well thepigeons remembered their previous choices. When 20 responsesconstituted a choice, there was little evidence of forgetting, even withdelays of up to an 1 hr. The point at which the delay was interpolatedinto the response sequence also affected performance. Delaysinterpolated between the second and third choice or between the thirdand fourth choice were more detrimental than delays interpolated betweenthe first and second choice or between the fourth and fifth choice.Zentall et al interpreted this result in terms of flexible coding. Ifthe delay occurred before the mid-point of the trial the pigeon encodedprevious choices (a retrospective strategy), after the mid-point thepigeon encoded choices yet to be made (prospective coding).WILKIE AND COLLEAGUESUp to this point the discussion of pigeon spatial memory hasprimarily focussed on maze or maze analog studies of various types.Memory for non-spatial stimuli in pigeons has typically been studiedusing operant procedures such as delayed matching to sample (DMTS).Wilkie and his colleagues have conducted an extensive series of41Spatial Memory in Pigeonsexperiments examining pigeon spatial memory using an operant proceduredelayed matching of key location (DMKL, Wilkie & Summers, 1982; seealso, Smith, Attwood, & Nieorowski, 1982) that is a spatial version ofDMTS. In this procedure, the pigeon faces a small, 3 X 3 matrix ofpecking keys. At the start of a trial one of these keys (the sample) isilluminated briefly, usually for about 2 sec. The key is thenextinguished and following a delay it is re-illuminated together withanother randomly chosen key from the array (the distractor). These keysremain lit until the subject responds. If the pigeon chooses thesample, the keys extinguish and the bird is rewarded with a brief accessto grain. If the bird chooses the distractor, both keys extinguish andan intertrial interval (ITI) begins.Wilkie and Summers (1982), Wilkie (1983a)In their initial study of DMXL, Wilkie and Summers examinedacquisition, sample duration, retention and memory load. Acquisitionwas rapid. Most of their subjects reached asymptotic levels ofperformance (80-90% correct) within 30 sessions (36 trials per session).However, levels of performance were better than chance (50%) much soonerthan this, in some cases during the very first session. There was noeffect of sample location. All positions in the nine key array werematched with equal accuracy. There was, however, an effect ofdistractor location. Trials on which distractors appeared close to thesample location resulted in lower levels of performance than trials on42Spatial Memory in Pigeonswhich the distractor was distant from the sample. Performance wasabovechance on both trial types. Accuracy was also effected by the number ofdistractors illuminated during the test phase of a trial.In one experiment, sample presentation was followed byillumination of the sample plus one randomly chosen distractor or bytheillumination of the entire nine key array. Accuracy on the latter trialtype was significantly lower than on the former, although still abovechance. As in the previous experiment, distractors close to the samplelocation were more likely to be chosen than distractors far from thesample location.In another experiment, Wilkie and Summers systematically variedsample and delay duration. Grant (1976) had previously demonstratedthat longer sample durations improved retention in delayed matching withcolour stimuli. The same effect was found in DNKL. Longer samples(maximum — 2 sec.) were remembered better than short (minimum — 0.2sec.). With 1 sec samples, performance dropped to chance levelsfollowing a delay of only 8 sec between sample presentation andsample/distractor presentation. In a final experiment, Wilkie andSummers varied the number of sample locations presented at the start ofeach trial. On some trials, only one sample location was presented. Onother trials, three sample locations were presented in either a linear(row, column or diagonal) or random relationship. Following a 1 secdelay, a sample and one distractor were illuminated. On three sampletrials, the test sample was chosen randomly from the set of three43Spatial Memory in Pigeonspresented initially. Accuracy was better on one sample trials relativeto three sample trials but above chance on both. Performance on linearthree sample trials was better than on random three sample trials.Within the range of linear three sample trials, row and column sampleswere remembered better than diagonal samples.In an additional experiment, Wilkie (1983a) demonstrated that theobserved level of matching performance in DMKL generalized to matchingof locations that the pigeons had never encountered previously. Hetrained pigeons on the DMKL task using only a subset of the ninepossible location sample stimuli. He then ran a series of transfertests in which all nine sample locations served as samples. Four of hisfive subjects performed as well with novel locations as sample stimulias they did with familiar sample stimuli, a result suggesting that thebirds had learned a generalized rule of the form “choose the locationthat matches the sample”.Vilkie (1983b)Wilkie next turned his attention to the issue of retroactiveinterference. Retroactive interference refers to the potentiallydisrupting effects of behaviors or stimuli that occur between the to-beremembered stimulus and tests for memory of that stimulus. DMTS can bedisrupted when events intervene between sample presentation and testing(Cook, 1980), but radial maze performance in rats is quite insensitiveto such interference, occurring only when the intervening stimuli are44Spatial Memory in Pigeonshighly similar to the to-be-remembered stimuli (Roberts, 1981). Wilkiefound a similar effect in DMKL. Changes in illumination during aretention interval did not affect matching accuracy. However,presenting an additional spatial stimulus did. If the interpolatedstimulus matched the sample location, accuracy was enhanced relative totrials in which no interpolated stimulus was presented. If theinterpolated stimulus matched the (as yet unseen) distractor stimulus,accuracy declined sharply to chance levels. An irrelevant stimuluspresented during the retention interval also disrupted performance butnot as severely as presenting the distractor. Additional manipulationsensured that the subjects were not simply remembering the last locationseen.Wilkie (1984, 1986)Wilkie has also addressed the issue of proactive interference(P1). The presence of proactive interference in spatial memory arguesagainst the notion of a resettable memory (see the discussion of spatialmemory above). Robert and Dale (1981) had demonstrated that P1 occurson spatial tasks under some conditions with rats. Roberts and VanVeidhuizen (1985) had also demonstrated P1 in pigeons on the radialmaze. Wilkie conducted a series of experiments to see if P1 alsooccurred in DMKL.In one series of experiments, Wilkie (1984) manipulated ITI and RIlength. He found that matching accuracy declined as a function of45Spatial Memory in Pigeonsdecreasing ITI and as a function of increasing RI. Although thisfinding was consistent with the presence of P1, Wilkiewas cautious inhis interpretation because these data were also explicable by an appealto differential expectancies of reward during trials andintertrialintervals. Shortening the ITI and lengthening the RI both reduce thedifference in expectations. In other words, the pigeonshave a highexpectation of reward during a trial and a low expectation of rewardduring the ITI. Increasing the RI decreases the expectation of rewardduring the trial. Decreasing the ITI increases the expectation ofreward during the ITI. Wilkie favoured this latter interpretation ofhis results.In two additional experiments he demonstrated that presenting non-contingent reinforcement during the ITI resulted in a decrease inmatching accuracy (cf., Gamzu & Williams, 1971) but that presentingresponse contingent reinforcement facilitated performanceirrespectiveof the reinforced locations’ relationship to the trial stimuli.Reinforced sample presentations (the sample from the next scheduledtrial), distractor presentations (the distractor from the next scheduledtrial) or an irrelevant location all facilitated performance. Nonreinforced sample presentations during the ITI disrupted matchingperformance but non-reinforced distractor or irrelevant locationpresentations had no consistent effects. Wilkie interpreted theseresults as being more consistent with the differential reward46Spatial Memory in Pigeonsexpectancies explanation and concluded that in some respects DMKLresembled classical conditioning.In another series of experiments, Wilkie (1986) again examined theissue of P1. One manifestation of P1 is that performance on latertrials within a session should be poorer than performance on earliertrials. Wilkie found no evidence for this type of deterioration withinsessions. Another characteristic of P1 is that it exerts strongereffects when a small potential set of sample stimuli is used (see Wrightet al, 1986) and when the potentially interfering event is recent.Wilkie systematically varied the size of the sample set and found noevidence of stronger P1 with smaller sample sets. In fact, his subjectsperformed better with smaller sample sets, a finding opposite to whatwould be expected if P1 was exerting an influence. He also examined theeffects of sample recency on matching accuracy. On some trials, thesample location from trial N-l served as the distractor on trial N. Onother trials, the distractor was chosen from trials other than N-l.There was no difference in performance between these two types oftrials. The pigeons performed equally well whether the distractor hadbeen the sample on trial N-l or from an earlier trial. Wilkie concludedthat DMKL performance might be immune to P1 and that the main reason forthis immunity is probably the speed at which location information isforgotten in this paradigm.47Spatial Memory in PigeonsWilkie and Kennedy, 1987; Wilkie, Vilison and Lee ( 1990)Wilkie and Kennedy (1987) developed a computer simulation ofpigeons’ performance on the DMKL task. The model was based on sixassumptions: 1) The key matrix is stored in a stable, map-likerepresentation. 2) Working memory is assumed to involve an attentionfocus or “pointer” that moves (“drifts”) over the surface of the mapmatrix. 3) This pointer migrates towards the representation of thesample location while the sample key is illuminated. 4) The pointerwanders randomly when the sample is not illuminated. 5) Duringacquisition, the pigeon learns the rule “choose the key location whoserepresentation is closest to the location of the pointer” and 6) Betweentrials the pointer is positioned randomly on the key matrixrepresentation.Wilkie and Kennedy’s computer simulation of this “drift” model ofpigeon short-term memory for spatial location (see also, Roitblat,1984a,b, 1987) yielded results that were in accord with the majority ofprevious results obtained with the DMKL procedure. Furthermore, severalpredictions were derived from the simulation. Wilkie, Willson and Lee(1990) tested some of these predictions.The “drift” model predicts that accuracy on trials during whichtwo sample locations are presented simultaneously should vary as afunction of the distance between the two samples. Accuracy should behigher on trials with adjacent samples relative to trials with nonadjacent trials. In the former case, the pointer will stop close to the48Spatial Memory in Pigeonsboundary of the two keys. In the latter case, the pointer should stopsomewhere between the two sample keys. It follows from thisthatpigeons should make more errors when an incorrect key fallsbetween twocorrect keys than when it is adjacent to a correct key. This predictionwas confirmed. Pigeons were much more likely to respond incorrectly toa distractor key located between two sample keys. Matchingaccuracy onthis type of trial was actually below chance, a finding that was alsopredicted by the “drift” model. On linear three-sample trials (i.e.,the sample defines a row, column or diagonal), the “drift” modelpredicts enhanced accuracy when the middle sample is presented as thetest sample relative to trials on which one of the end samples ispresented. This prediction was also confirmed.Wilkie (1989)The “drift” model assumes that pigeons represent the DMICL keymatrix in a map-like form. A recent experiment by Wilkie (1989)supports this idea. Cheng and Gallistel (1984), mentioned previously,have suggested that rats represent Euclidean properties of space such asdistance and direction. Their conclusions were based on several studiesin which space was transformed in various ways. In DMKL these physicaltransformations are difficult to achieve so Wilkie used multidimensionalscaling techniques (MDS) to make inferences about the structure of thismap-like representation. MDS refers to a set of related computational49Spatial Memory in Pigeonstechniques capable of constructing a map strictly from distanceinfo rniat ion.Wilkie assumed that confusions between locations were a good indexof psychological distance. Easily confused locations arepsychologically close. He used confusion scores as distances in anMDS program and recovered a reasonable approximation of the key matrixas output. He tested several different geometries (e.g., city-block)but discovered that a two-dimensional Euclidean geometry produced thebest fit to the data.CHENG (1988, 1989, IN PRESS)Cheng conducted a series of experiments examining pigeons’ use oflandmarks to find a hidden goal. His task is similar to the MorrisWater maze (Morris, 1981) described previously. He trained pigeons tosearch for a hidden food-well in a small rectangular arena. The arenacontained one or more distinct landmarks. To ensure that the within-arena landmarks came to control the pigeons’ search behavior rather thanthe features of the surrounding room, Cheng shifted the arena within theroom each trial. Once the pigeons had acquired the task varioustransformations of the within-arena landmark were performed. Displacingthe landmark(s) in the horizontal plane caused the pigeons to shifttheir searching behavior by a distance comparable to the size of theshift. Vertical displacements or changing the size of the landmark(s)did not disrupt searching. When two landmarks were present, the pigeons50Spatial Memory in Pigeonsrelied more heavily on landmarks closer to the hidden food-well (i.e.,shifting a near landmark caused a greater shift in searching behaviorthan moving a far landmark.Cheng proposed a vector sum model to account for his results.According to this hypothesis, the pigeon encodes a number of landmark-to-goal vectors. When searching for the hidden food-well again, thepigeon adds to each landmark-goal vector the corresponding vector fromits present location to the landmark in question to generate navigationvectors. The pigeon then moves to a location determined by a weightedaverage of the navigation vectors. Near landmarks are weighted moreheavily than far landmarks.Similar models have been proposed for bees (Cartwright & Collett,1983, 1987), rats (Cheng, 1986; Cheng & Gallistel, 1984) and gerbils(Collett, Cartwright & Smith, 1986). Rats, pigeons and gerbils allencode metric properties of space. Bees apparently do not. Landmarkdisplacement experiments have also been conducted with Clark’ sNütcrackers with similar results but no formal model of landmark use hasyet been offered for this species (Vander Wall, 1982).PIGEON SPATIAL MEMORY: A SUMMARYAlthough initial work on spatial memory in pigeons led to thesuggestion that their spatial memory abilities were limited or even nonexistent (Bond et al, 1981) more recent work has painted a somewhatbrighter picture (Spetch & Edwards, 1986, 1988; Spetch & Honig, 1988;51Spatial Memory in PigeonsSpetch, 1990). Spatial memory in pigeons appears to be very similar tothat observed in other species such as the rat, differingonly in termsof temporal persistence and, perhaps, capacity.Pigeons’ performance on spatial tasks is very sensitive toprocedural variables. They perform best on tasks that provide a wealthof landmarks and relatively unconstrained pathways between goals. Thespatial distinctiveness of the to-be-remembered locations is alsoimportant. Spatially distinct locations are easily recalled, even inthe face of extended delays (Spetch & Honig, 1988). Locations that arenot spatially distinct (Wilkie & Summers, 1982) can be remembered onlyfor brief periods of time. Space is represented in a map-likeform inwhich Euclidean properties of space are maintained (Cheng, 1988, 1989,in press; Wilkie, 1989). “Richer” maps, that is, maps in which morefeatures and relationships between features are represented, supporthigher levels of performance, both in terms of accuracy and temporalpersistence, than impoverished maps (Spetch & Honig, 1988; Wilkie &Summers, 1982).The data bearing on the capacity issue are equivocal. Studiespurporting to show capacity constraints have typically confounded memoryload with other factors (Dale, 1988; Roberts & Van Veldhuizen, 1985) ortested subjects under suboptimal conditions (Roberts, 1988). Studiesfailing to show capacity constraints have probably not employedprocedures that come close to straining pigeons’ capacity (Spetch,1990).52Spatial Memory in PigeonsIt has been suggested that animals such as the rat and food-storing birds learn associations between locations and the presence orabsence of food (Shettleworth, 1985; Staddon, 1983). The same isapparently true of pigeons. However, the sensitivity of pigeons toprocedural variables on spatial tasks makes comparisons between tasksproblematic so it is unclear how readily one can generalize acrosstasks. This is especially true of comparisons between maze studies andDMKL studies. Wilison and Wilkie (1991) have addressed this issue bymodifying the DMKL procedure in various ways to produce levels ofperformance that are comparable to that observed in maze procedures.WILLSON & WILKIE (1991)Attempts to answer the question “How should radial mazeprocedures, as used to study spatial memory in rats, be modified for thestudy of spatial memory in pigeons?” have led to a much greaterunderstanding of spatial memory in pigeons. In a recent series ofexperiments, Willson and Wilkie (1991) asked a similar question inregards to DMKL. They asked: Is it possible to modify the DMXLprocedure in such a way as to attain levels of retention comparable tothat seen with other spatial memory tasks?One potentially limiting factor in the DMKL procedure was theapparatus. Sample locations are presented within a small matrix of keyson an opaque, vertical surface, in a dimly lit room. Wilison and Wilkie53Spatial Memory in Pigeonsreasoned that making the potential sample locations more spatiallydistinct, both by increasing the distance between potential samples andby providing prominent landmarks, would improve retention. Theapparatus that they developed is pictured in Figure 1. It wasconstructed of Plexiglas and a single key was mounted on each of 10sides of the apparatus. A feeder was mounted on the eleventh side.Although proactive interference has not been conclusivelydemonstrated in the DMKL procedure Wilison and Wilkie sought to reducethe possibility that it might manifest if the retention of spatialinformation in this paradigm was improved. The best way to avoid P1 isto employ trial-unique stimuli or at least not repeat stimuli within asession. Because the potential number of possible sample locations waslimited (10) the authors conducted only one trial per day. Within ablock of 10 sessions each key served as the sample once and as thedistractor once. The same key was never used on consecutive days. On agiven trial, the sample and distractor were always equidistant from thefeeder to prevent key biases. Willson and Wilkie also modified the DMKLprocedure in four other ways.Increasing sample duration has been shown to improve matchingaccuracy on the DMKL procedure (Wilkie & Summers, 1982). Matchingaccuracy also improves when responding to the sample is reinforced(Wilkie, 1983c). Both of these features were incorporated into theprocedure. Willson and Wilkie used a 15-mm sample presentation and54Spatial Memory in PigeonsFigure 1. A schematic diagram of the apparatus used inWilison and Wilkie (1991). Numbers indicatepecking key positions.55Spatial Memory in PigeonsHt I87A 8N44I / F180cm 3 102 1 8D_____p EA. Hendecagonal Skinner box E. DoorB. Grain Dispenser F. Tvo-vay mirrorC. Vooden base 9. Cement columnD. Table H. Comector box56Spatial Memory in Pigeonsreinforced responding to the sample on a variable interval (VI) 30-secschedule.A third procedural change was made in light of the fact thaterrors on memory tasks are not always due to forgetting (Brown & Cook,1986; Dale, 1988; Devenport, 1989; Roitbiat & Harley, 1988; Wilkie &Spetch, 1981). Because single incorrect responses may or may notreflect forgetting, the authors allowed the pigeons to make severalchoices on a trial. During retention tests, both the sample and adistractor were illuminated for 1 mm. If a subject made more responsesto the sample during this period, the distractor extinguished andresponding to the sample was reinforced on a VI 30-sec schedule for 5mm. If the subject made more responses to the distractor, both keysextinguished and the trial ended. If a subject made an equal number ofresponses to the sample and distractor during the test, the nextresponse determined the trial outcome.The final modification was made in light of a finding by Urcuioliand Callender (1989). These investigators gave their pigeons off-baseline training to discriminate between the stimuli that served assamples in a DMTS task. Pigeons that received discrimination trainingacquired DMTS faster than pigeons who did not. Urcuioli and Callenderinterpreted this finding as stemming from enhanced attention to thesample stimuli produced by differential reinforcement. Previous workfrom our laboratory (Willson, unpublished data) had suggested that themodifications described previously were not always sufficient to enhance57Spatial Memory in Pigeonsretention of the sample location. Thus we incorporated discriminationtraining into some conditions in the hope that it would enhanceattention to the sample location.In their first experiment, Willson and Wilkie tested six pigeonsin two conditions. In one condition, the Sample Alone condition, a keywas lit and responding to that key was reinforced on a VI 30-secschedule for 15 mm. This was followed by a delay and then theillumination of the sample and a distractor. In the other condition,the Sample/Distractor condition, two keys were lit. Responding to one,the sample, but not the other, the distractor, was reinforced.Following the delay both of these keys were re-illuminated. Subjectsreceived 30 daily trials on each condition and the delay was always 30sec. The order in which the subjects received the conditions wascounterbalanced.Somewhat surprisingly, performance on the Sample Alone conditionwas near chance and did not improve over the course of training.However, performance on the Sample/Distractor condition was excellent,better than 80% for all subjects. Furthermore, this excellentperformance was present from the first day of training. Willson andWilkie observed that performance on the Sample Alone condition wassubject to P1. Performance tended to be better as the number of daysbetween key repetitions increased. A regression equation fit computedfor the data intersected chance at 2.53 days. Thus it seemed that thesubjects in the Sample Alone condition could remember the location of58Spatial Memory in Pigeonsthe sample stimulus but found it difficult to discriminate betweencurrent and previous presentations of that stimulus. No evidence for P1was found on the Sainpie/Distractor condition. Wilison and Wilkieconcluded that the improved performance observed followingSample/Distractor training was likely due to enhanced attention to theSample (and Distractor) stimuli produced by differential reinforcement.In a second experiment, Willson and Wilkie varied the length ofthe retention interval (RI) between Sample/Distractor training and theretention test. All six subjects performed well with RI’s of up to 30mm but most subjects did much better than this, and one subjectdemonstrated above chance performance following an RI of 24 hr. Thesedata are presented in Figure 2.Thus the answer to the question “Is it possible to modify the DNXLprocedure in such a way as to attain levels of retention comparable tothat seen with other spatial memory tasks?” is apparently “yes.However, it remains unclear exactly what modifications mediated thisimproved performance. The discrimination training between sample anddistractor is apparently critical, but the degree to which the otherfactors (i.e., the physical layout of the apparatus, only a single trialper day, etc.) contributed is unclear. The purpose of the presentresearch was to examine pigeons’ performance on a modified version ofthe DNKL procedure using the original DMKL apparatus in a multiple trialper day procedure.59Spatial Memory in PigeonsFigure 2. The retention interval data from Wilison andWilkie (1991), Experiment 2. All data points,except the final one for each subject, representblocks in which criterion was met (DR > 0.70).60Spatial Memory in Pigeons0.S•0.80.7•00.6zo UBIROlI— 05.0.4 001R04o0.3 x0.20.I• V.. V V’V•’tV V10.001 0.01 0.1 1 10 24RETCNTION INTERVAL (HRS).- — — —oIIx61Spatial Memory in PigeonsEXPERIMENT 1The purpose of the first experiment was to replicate the findingsfrom our previous work using the traditional DMKL apparatus and morethan one trial per day. An additional aim was to compare the effects ofSample/Distractor training vs Sample Alone training within sessions.Each bird received four daily trials. Two trials began with theillumination of a randomly selected key. On the other two trials both asample and a distractor were illuminated. Only responses to the samplewere reinforced. This initial portion of the trial lasted 10 mm. Thekeys were then turned off and a retention interval began that lastedeither 5 or 30 sec. Thus the four trial types were: Sample Alone-5 sec,Sample Alone-30 sec, Sample/Distractor-5 sec, and Sample/Distractor-30sec. The order in which the trial types occurred was randomized.METHODSubjectsFour King pigeons who had varied experimental histories instandard operant chambers were used. All subjects were maintained atapproximately 90% of their free feeding weight by mixed grain obtainedduring the experimental session and from occasional post-sessionsupplements when necessary. Vitamin enriched water, health grit andcrushed oyster shell were available ad libitum in the large, plastic62Spa cial Memory in Pigeonscoated, wire mesh home cages. Subjects were tested 5 days per weekduring the middle portion of the light cycle that was matched to naturalsunrise and sunset times.ApparatusThe apparatus consisted of a panel that could be attached to theside of a subjects’ home cage. A subject was transported, in its homecage, to the small dark testing room and the panel was attached to theside of the cage. Each cage was equipped with a door that could beopened to allow the subject access to the panel. Each panel wascomposed of a standard grain feeder, a houselight, and a square 3 X 3matrix of pecking keys. Behind each key was a microswitch, whichdetected pecks having a force greater than .15 N, and a red light-emitting diode. The keys were about 3 cm in diameter and were mountedabout 5 cm apart, center to center. Data collection and experimentalcontrol were carried out by a Data General NOVA 3 computer operatingunder RDOS and the MANX programing language (Gilbert & Rice, 1979).ProcedurePreliminary Training. Because all subjects had previously been trainedto respond to illuminated pecking keys, relatively little preliminarytraining was necessary. All birds were shaped until they wereresponding at a steady rate on all nine keys on a VI 30-sec schedule.The experiment proper began on the following day.Trials. There were four daily trials. On each trial, a pair of peckingkeys was chosen at random. One key was designated as the sample, the63Spatial Memory in Pigeonsother as the distractor. Subjects received two trials on which bothkeys were illuminated throughout the trial (the Sample/Distractorcondition) and two trials on which only the sample was illuminatedduring the initial phase (the Sample Alone condition). Each of thesetrial types was paired with a short retention interval (5 sec) and along retention interval (30 sec). There was a 2-mm ITI between trials.The houselight was illuminated during trials but the ITI was spent indarkness.Initial Phase. In the Sample/Distractor condition both keys wereilluminated and remained lit for 10 mm. Pecks to the sample werereinforced on a VI 30-sec schedule with 4 sec of access to mixed grain.Pecks to the distractor had no scheduled consequences. Following 10 mmof exposure, the next peck to the sample produced a final reinforcement(to ensure that the sample was the last key responded to prior to theRI). The pecking keys extinguished after this final reinforcement andremained dark for the duration of the retention interval (RI), thatlasted for either 5 sec or 30 sec. The RI was timed from the end of thehopper presentation. The Sample Alone condition was identical in allrespects except that the distractor key was not illuminated. Allsubjects were tested on these conditions for 30 days.Test Phase. Following the RI, the sample and distractor wereilluminated and remained lit for 1 mm. Pecks during this period wererecorded, as was the location of the pigeon’s initial choice. If thesubject had made more responses to the sample by the end of the 1-mmperiod, the distractor was turned off and pecks to the sample produced 4sec of access to mixed grain on a VI 30-sec schedule for 5 mm. Thiswas followed by an ITI of 2 mm. If the pigeon made more responses to64Spatial Memory in Pigeonsthe distractor, both keys were turned off and 5 ruin were added to theITI. [If a subject happened to make an equal number of responses toboth keys, the first response following the end of the 1-ruin test perioddetermined whether or not the trial continued.]Data Analysis. Two performance measures were computed for each subject.A discrimination ratio (DR) based on responding during the 1-ruin testperiod was calculated for each trial type and averaged over blocks of 10sessions. A DR is a ratio of correct responses to total responses.Thus if the subject responds only to the sample, the DR will be 1.0. Ifthe subject responds to the sample and distractor equally often, the DRwill be 0.50, and so forth. The proportion of correct first responseswas also computed for each trial type and averaged over blocks of 10days. To increase statistical power, two factors, each containing twolevels, (Sample Alone vs Sample/Distractor and Short RI vs Long RI) werecombined to form one factor (Trial Type) with four levels.RESULTSFigure 3 shows the DRs for each subject. All subjects exhibited asimilar pattern of results. Performance was much higher on theSample/Distractor trials ( fl — .751 ) than on the Sample Alone trials( H — .517 ). Three out of four subjects remained at chance level onthe Sample Alone trials and the performance of the one subject that wasabove chance dropped on the long retention interval trials. There wasno effect of retention interval on the Sample/Distractor trials; allsubjects were well above chance levels of performance at both the longand short retention interval ( H — .748 vs M — .755, respectively).65Spatial Memory in PigeonsThe percent correct first choices data from each subject aredepicted in Figure 4. Although the pattern of results is similar tothat of the DRs, there is clearly much more variability and the overalllevel of performance is lower. The only trials on which all subjectswere clearly above chance were the Sample/Distractor trials with a shortretention interval ( H — .71).The mean performance on each trial type and retention interval forboth the DR measure and percent correct measure are shown in Figure 5.The differences discussed above can be seen clearly and were confirmedthrough statistical analyses. A 3 (Blocks of 10 sessions) by 4 (TrialTypes) by 2 (Measures) analysis of variance revealed both a significanteffect of trial type and measure [ F(3,27) — 16.43, p < .001, andF(l,9) — 34.200, p < .0004, respectivelyl. Post-hoc Newman-Keulscomparisons revealed that performance on the two Sample Alone trials didnot differ nor did performance on the two Sample/Distractor trials, thusthere was no effect of retention interval. However, performance onSample Alone trials was significantly lower than performance onSample/Distractortrials. Performance as measured by the DR measure was significantlyhigher than as measured by the percent correct 1st choice measure. (AllNewman-Keuls comparisons were conducted with alpha — .05).There was no effect of block [ F(2,9) — 0.1293, p > .87];performance during the initial block of sessions was the same as inlater blocks of sessions, nor were any of the interactions significantall p’s > .38].66Spatial Memory in PigeonsFigure 3. Average discrimination ratio for each subject onSample/Distractor (DR SD) vs Sample Alone (DRS) trials inExperiment 1.67Spatial Memory in PigeonsBird 1100.820.6I.0.2A DRSOGo0 10 20 30 40Reterdon Inteve ISeCIBird 21.0 I08.20.6A OGo I .c_s0 10 20 30 40Ret,b Wter Eeecl68Spatial Memory in PigeonsBird 31.0 I IGe02* ce_so0.0 •c.s0 10 20 30 40Rotenbon rterv teedBird 41.0Ge____________________9‘S0402Go I I I ._s0 10 20 30 40Reten hter leed69Spatial Memory in PigeonsFigure 4. Average percent correct first choices for each subject onSample/Distractor (PC_SD) vs Sample Alone (PC_S) trials inExperiment 1.70Spatial Memory in PigeonsBird 11.00.81G8_8DG0 I •_s0 10 20 30 40Retenon IItg r%dBird 21.0 IP6Q402Go I I I0 10 20 30 40teed71Spatial Memory in PigeonsBird 31.0A PC300 •po_sO 10 20 30 40fletena k,terv (aeciBd4A PO8DGo I I I •Pc_so io 20 30 4072Spatial Memory in PigeonsFigure 5. A comparison of the group means for each trial type and eachmeasure of performance from Experiment 1. Symbols are thesame as in Fig. 3 and 4.73Spatial Memory in Pigeons1.0AverageC-)a)004-Ca)0G)000I0.00.80.60.2Retention Interval Iseci40• PCSD• Pc_SA DR_SDI DR_S0 10 20 3074Spatial Memory in PigeonsDISCUSSIONThese findings clearly replicate the findings of Wilison andWilkie (1991). The facilitative effect of the Sample/Distractor trialsis clearly evident on both short and long RI trials. It is also clearthat the DR provided a higher, more stable measure of performancerelative to the percentage of correct first choices. In comparison toperformance on the traditional DMKL task (where performance typicallydrops to chance at an 8-sec RI, Wilkie & Summers, 1882), the subjectsperformed better on Sample/Distractor trials at the 30-sec RI based onthe DRs ( M — .666), but not on the percentage of correct firstresponses ( M — .436).Also of interest is the fact that performance on Sample Alonetrials remained at approximately chance levels during the experiment.One possibility is that the subjects failed to acquire the Sample Alonetask; they received only 30 Sample Alone trials during the experiment.However, pigeons typically acquire the DMKL task quickly, sometimeswithin the very first session (Wilkie & Summers, 1982). Furthermore, inrelated research, pigeons have been tested on the Sample Alone conditionfor hundreds of trials with no improvement in performance (Willson,unpublished data). Clearly, extending the sample presentation andreinforcing the sample stimulus do not necessarily lead to improvedmatching performance.The fact that pigeons fail to perform well on Sample Alone trialsin the present protocol is also relevant to the issue of what mechanism75Spatial Memory in Pigeonsfacilitates performance on Sample/Distractor trials. Wilison and Wilkie(1991) suggested that the observed levels of performance were due toenhanced attention to the location of the sample. The present findingssuggest that attentional enhancement, if it occurs, is fairlytransitory, operating only within trials, a finding not completely inaccord with the findings of Urcuioli and Callender (1989) whodemonstrated that off-baseline discrimination training facilitates theacquisition of matching and concluded that the mechanism underlying theeffect was attentional. Discrimination training enhances attention tothe stimulus dimensions relevant to the matching task and thusfacilitates acquisition. Willson and Wilkie hypothesized that a similarmechanism mediated performance on Sample/Distractor trials in theirresearch.However, an important difference between the two experiments ishow the discrimination training was conducted. In Urcuioli &Callender’s research, discrimination training occurred off-baseline(i.e., independent of the matching procedure). Such was not the case inWillson and Wilkie’s experiment. Discrimination training was anintegral part of the matching procedure, in effect, on-baseline. In thepresent research, discrimination training occurred in both contexts. OnSample/Distractor trials discrimination training was an integral part ofthe trial. However, the discrimination training inherent toSample/Distractor trials was in essence off-baseline discriminationtraining on the Sample Alone trials. Therefore, performance on Sample76Spatial Memory in PigeonsAlone trials should have been better than the levels of performanceobtained in Wilison and Wilkie (1991), if the discrimination trainingwas enhancing attention to the relevant sample dimension, location. Itclearly was not facilitating attention to location in general but mayhave enhanced attention to the particular locations used within a trial.If discrimination training had been enhancing attention to location in ageneral sense, performance on the Sample Alone task following trainingon the Sample/Distractor task in Willson and Wilkie’ s originalexperiment should have been better than performance on the Sample Alonetask prior to Sample/Distractor training. There was no such effect.A more plausible explanation is that the excellent performanceobserved following Sample/Distractor training resulted from differentialconditioning to the stimuli presented during the initial part of thetrial. The pigeons learned that one stimulus was associated withreinforcement, the other with extinction. Following the RI, thesubjects continued to approach the S+ and avoid the S-. On Sample Alonetrials, the pigeon also learned that one stimulus was associated withreinforcement but made errors during the test because the distractorstimulus had also been associated with reinforcement in the relativelyrecent past. This explanation is consistent with the P1 data fromWillson and Wilkie (1991). Recall that their subjects performed betteron the Sample Alone task when the number of days between repetitions ofthe trial stimuli increased.77Spatial Memory in PigeonsAlthough this explanation is plausible, the data upon which it isbased are weak because the order in which the trial types occurred wasrandomized. Sample Alone training did not consistently followSample/Distractor training and any effect of Sample/Distractor trainingon Sample Alone trials would have been attenuated as a result. Thepurpose of the second experiment was to test the attentional anddifferential conditioning hypotheses more directly.EXPERIMENT 2One approach to testing the attentional vs differentialconditioning hypotheses would be to train the subjects on theSample/Distractor condition first, then test them on Sample Alonetrials. This procedure was implicit in the training received by some ofthe subjects in Willson and Wilkie’s (1991) first experiment. Recallthat half of the subjects received Sample/Distractor training prior toSample Alone training. It follows from the attentional hypothesis thatthese subjects should have performed better on Sample Alone trials thansubjects that had not received Sample/Distractor training. As mentionedabove, no difference was apparent and this lack of an order effectargues against the Attentional hypothesis.However, another possibility is that this procedure was notsensitive enough to differentiate between these two hypotheses. If theeffects of Sample/Distractor are not long-lasting and its enhancing78Spatial Memory in Pigeonseffects therefore deteriorate, performance on Sample Alone trials woulddecrease during the course of the 30 days of testing. Two of the threesubjects that received Sample/Distractor training prior to Sample Alonetraining performed best during the first block of 10 Sample Alonetrials. Although far from conclusive, this result does suggest that amore sensitive test might be required to differentiate between theAttentional and Differential Conditioning hypotheses.Therefore the approach taken in the present experiment was to testthese two hypotheses using a within-trials procedure. Subjects receivedfour daily trials. One trial was identical to the Sample Alonecondition in the previous experiment. The other three trials began withSample/Distractor training but differed in terms of the stimulipresented during the retention test. One trial was identical to theSample/Distractor condition in the previous experiment. On the criticaltrials, one of the stimuli presented during Sample/Distractor trainingwas replaced with another stimulus during the test portion of the trial.On one trial, the distractor was replaced (Novel S-), on the other, thesample was replaced (Novel S+). Although these stimuli were notstrictly novel (having potentially been used on previous trials) theywere novel with respect to the trial in progress. The differentialconditioning and attentional hypotheses make clearly differentpredictions on these trials.The differential conditioning hypothesis predicts that performanceon Novel S- trials will be identical to that observed on Sample Alone79Spatial Memory in Pigeonstrials. The testing conditions are identical in the two cases.Therefore, the pigeons should confuse the S+ from the current trial witha stimulus that had (potentially) served as an S+ on a previous trialbut now served as the distractor. Performance on Novel S+ is a bitharder to predict but should also be similar to performance on SampleAlone trials for reasons similar to those outlined above. Presumablythe attenuated performance would arise as a result of confusion betweentwo stimuli recently associated with extinction.The enhanced attention hypothesis predicts that performance on theNovel S- and Novel S+ trials will be similar to performance onSample/Distractor trials. The discrimination training enhancesattention to the location of the sample stimulus (and presumably to thelocation of the distractor). Thus on Novel S- trials, the pigeon shouldcontinue to approach the S+ and on Novel S+ trials should avoid the 5-.METHODSubjects and ApparatusThree of the four subjects from Exp. I and one new subjectparticipated in the present study. Although naive with respect to thepresent procedures, this new subject had participated in previous DMiCLresearch and required no preliminary training. All housing conditionswere identical to those of the previous experiment. The same apparatuswas also used.80Spatial Memory in PigeonsProcedureBecause all birds responded to illuminated pecking keyspreliminary training was unnecessary. There were four daily trials. Onone trial (Sample Alone), a randomly selected key was illuminated andremained lit for 10 mm. Responses to this key were reinforced on a VI30-sec schedule with 4 sec of access to mixed grain. On the other threetrials a pair of randomly selected keys was illuminated and remained litfor 10 mm. Responses to one key (S+) but not the other (S.) werereinforced on a VI 30-sec schedule. On all four trials, the 10-mmsample exposure ended with a final reinforcement (response contingent)and a brief, 30-sec RI began.Following the RI a pair of keys was illuminated and remained litfor one mm. On Sample Alone trials this pair consisted of the sample(S+) and a randomly selected distractor (S-). On Sample/Distractortrials, this pair consisted of the sample (S+) and distractor (S.) fromthe initial portion of the trial. On Novel S-f trials, the pairconsisted of the distractor CS-) from the initial phase of the trial anda randomly selected key (S-f) and on Novel S- trials, the pair consistedof the sample (S+) from the initial phase and a randomly selected key(S-). Responses to each member of the pair were recorded separatelyduring the 1-mm test. If the subject made more responses to S+ duringthe test, S- was extinguished and responses to S+ were reinforced on aVI 30-sec schedule for an additional 5-mm period followed by a 2-mmITI that was spent in darkness. If the subject made more responses to81Spatial Memory in PigeonsS-, both keys were extinguished and 5 mm were added to theITI. If asubject made an equal number of responses to S+ and S- during the testperiod the next response determined the trial outcome. Theorder inwhich the four trial types occurred was random. The measuresofperformance computed in Exp. 1 were also computed for this experiment.RESULTS AND DISCUSSIONThe data from the present experiment are shown in Figure 6.Performance as measured by the DRs was again better, on average, thanthe percentage of correct first choices (PC), but the pattern of resultswas similar in both cases. All subjects performed best on theSample/Distractor trials relative to the other three trial types (DR —.727, PC — .642). Performance on the Novel S+ and Novel S- trials wasbetter than on the Sample Alone trials but did not differ from eachother (DR — .60, PC — .59 and DR — .57, PC — .57, respectively)Performance on the Sample Alone trials was at chance (DR — .50, PC —.44).A 12 (Blocks of 5 sessions) by 4 (Trial Types) by 2 (Measures)analysis of variance revealed a significant effect of measure [ F(l,36)— 8.1157, p — .008 1 and a significant effect of trial type [ F(3,108) —16.7686, p — 0 ). There was no significant effect of block [ F(1l,36) —0.8429, p — .60 ], nor were any of the interactions significant.Post hocNewman-Keuls comparisons revealed that performance on Sample/Distractortrials was better than on the other three trial types. Performance on82Spatial Memory in PigeonsFigure 6. A comparison of the mean performance on the four trial typesfrom Experiment 2 for each subject. Both the DR and PC dataare presented. The group average is also shown.83Bird 1 SpatialMemory in PigeonsI:s sCdUoBird 2101a8•06.& 0.402I0.0 IDRIIS S.D S_NO F’&DCd84Spatial Memory in Pigeons1.002106IP20.01.0jOBBird 3S_D S...ND r8_DBird 4••••S S_D S_ND _DCd85DiscriminationRatkorProportknCorrectQQ00P0to0)(00Co C,)obCD0 ID> CD -, CC)CDm I— mSpatial Memory in Pigeonsthe Novel S+ trials and Novel S- trials was better than on Sample Alonetrials but didn’t differ from each other. (All comparisons wereconducted with alpha — .05)The results of the present experiment are not entirely consistentwith either the differential conditioning hypothesis or the enhancedattention hypothesis. Although performance on Novel S- and Novel S+trials was better than that observed on Sample Alone trials (as would bepredicted by the enhanced attention hypothesis) it was not as good asperformance on Sample/Distractor trials (as predicted by thedifferential conditioning hypothesis). Thus neither account, by itself,can explain the present findings. However, aspects of both can accountfor the results. An example will be used to illustrate.Assume that the total amount of associative value available in aconditioning situation can vary between -l and 1. All potentiallypredictive stimuli compete for this total value. Pairings withreinforcement increase the value of a particular stimulus and pairingswith non-reinforcement decrease its value (cf., Rescorla & Wagner,1972). In the absence of conditioning this value “drifts” towards 0.The total amount of associative value that accrues to a particularstimulus will depend on not only its own predictive value but also thepredictive value of other stimuli present in the conditioning situation.As conditioning progresses the subject learns that some stimuli are morepredictive of reinforcement than others and learns to direct itsattention towards those stimuli.87Spatial Memory in PigeonsFor example, during the initial part of a Sample Alone trial,attention is directed towards both the illuminated stimulus and itslocation. Both of the cues predict reinforcement to an equal extent sothe associative value of both increase towards a maximum of 0.5 (half ofthe total associative value available). On Sample/Distractor trials,one location exclusively predicts reinforcement and its associativevalue increases towards 1. The other location predicts non-reinforcement and its associative value decreases towards -1. Theilluminated stimuli on Sample/Distractor trials are associated equallywith reinforcement and non-reinforcement and their associative valueremains near 0. Stimuli from previous trials have an associative valueof close to 0 because they have been associated in the past with bothreinforcement and non-reinforcement, and because associative value tendsto drift towards 0 in the absence of further conditioning.If we assume that the 10-mm sample period is sufficient time forthe associative values to attain their maxima and that the associativevalue do not decrease much during the RI, the following conditionsprevail during the test portion of the trial:88Spatial Memory in PigeonsSTIMULUS TYPETRAINING TESTTRIAL TYPEPS+ VS+ PS- VS- C DSA 0.5 0.5 N/A N/A 0.5 0.0SD 1.0 0.0 -1.0 0.0 1.0 -1.0NS 1.0 0.0 -1.0 0.0 0.0 -1.0ND 1.0 0.0 -1.0 0.0 1.0 0.0where:P — location stimulusV = visual stimulusS+ rewardednon-rewardedC — correctD — distractorSA — Sample AloneSD — Sample/DistractorNS = Novel SampleND — Novel DistractorThus performance on Sample/Distractor trials should be bestbecause the choice involves a discrimination between two stimuli withextreme and opposite associative values. Performance on Sample Alonetrials should be poor because the choice involves a discriminationbetween two stimuli with intermediate associative values. Performanceon Novel S- and Novel S+ trials should fall somewhere between because89Spatial Memory in Pigeonsthe choice is between a stimulus with an extreme associative value andone with an intermediate associative value.That this model can explain the pattern of results obtained in thepresent experiment is no surprise since it was designed with thatpurpose in mind. However, it can also explain the data from Willson andWilkie (1991). This issue will be addressed in the General Discussionbelow.In the next experiment, the issue of temporal persistence isaddressed. Specifically, the question asked is: How long can pigeonsremember the location of the rewarded sample on this modified DMKLprocedure?EXPERIMENT 3The matching accuracy of the subjects on the Sample/Distractortrials in Experiments 1 and 2 did not decrease following an RI. Thepurpose of this third experiment was to explore the degree to whichpigeons could retain spatial information at even longer delays. Itseemed unlikely that performance would equal that observed in Wilisonand Wilkie (1991) due to the lack of visible landmarks and the closeproximity of the response alternatives in the traditional DMKLprocedure. Nonetheless, performance should be superior to thattypically observed in DMKL procedures.90Spatial Memory in PigeonsMETHODSubjects and ApparatusThe three subjects from Experiment 1 who had also been used inExperiment 2 were used in the present study. The apparatus wasidentical to that used in the previous experiments. All aspects of thepigeons care and housing were identical to those in the previousexperiments.ProcedureAll subjects were tested on the Sample/Distractor condition forthree daily trials. At the beginning of a trial, two stimuli wererandomly selected and illuminated, as in the previous experiment, andall other conditions were identical to those of the previous experiment,except as indicated below. The interval between initial training andsubsequent testing assumed one of three values at random. The shortestinterval, which occurred in each session, was 5 sec. The initial valuesof the two other RI’s were 30 sec and 60 sec, respectively. Subjectswere tested on this distribution of retention intervals for 30 sessionsto establish a baseline. The longest retention interval was thenincreased to 5 nun, and the middle RI to 60 sec. This distributionremained in effect for 20 sessions at which time the middle RI wasincreased to 5 mm, and the longest RI to 10 mm. The duration of themiddle and longest RI’s was doubled every 20 sessions until the value of91Spatial Memory in Pigeonsthe longest RI was 40 mm and the middle RI was 20 mm. The experimentended after 20 sessions with the RI’s at these values. Matchingaccuracy was measured by calculating a DR based on responding during thefirst minute of the test phase, as in the previous experiment.RESULTS AND DISCUSSIONFigure 7 shows the retention interval data foreach of the threesubjects and the mean. The forgetting functions for all three subjectsare fairly flat. Only Subject #2 showed a substantial decrease inperformance as the RI was lengthened and the decay assumed an invertedU-shape rather than a linear decrease. Subject #3 actually performedbetter at the longest RI (i.e., 40 mm). Two of the three subjects werestill performing at .70 or better when testing wasterminated and allsubjects were still performing at levels well above chance. An analysisof variance performed on the DRs at each of theseven delays used in thepresent experiment revealed no significant drop inperformance [ F(6,12)—1.0280, p > .45 ]. A one-sample t-test on the DR values from the 40-mm RI confirmed that performance was still significantly better thanchance [ t(2) — 7.28, p < .02 ].The data from each of the memory interval distributions tested aredepicted in Figure 8. Although the data are quitevariable, it is clearthat the memory functions based on the three RI’s tested in each blockof sessions are fairly flat. Recall that each subject received threedaily trials and that the RI for one of those trials was always 5 sec.92Spatial Memory in PigeonsFigure 7. Mean performance at each retention interval for each subjectin Experiment 3. The group average is also shown. Notethat the x-axis is a log scale.93Spatial Memory in PigeonsBird 11.0 I II0.60.5 1 I1 10 100 1000 10000ReteniSon terv teedBird 21.0i. .\\1 10 100 1000 10000Retenbon hiterv Lead94Spatial Memory in PigeonsBird 31.0 I I0706Q51 10 100 1000 10000Reti tervaI (sedAverage1.00906051 10 100 1000 10000Reten Wterva lead95Spatial Memory in PigeonsFigure 8. Mean performance on each of the memory intervaldistributions used in Experiment 3 for each subject.The group average is also shown. The first block ofintervals is called DR1, the second block, DR2, andso forth.96Dba*nI*IcinReI-.I-.rq m000-‘-4miD0-‘ 0 IDiea*nlnellonRatio0000014cD80•-D00§0.•I.k•*-I 0.-8IISpatial Memory in PigeonsBird 30.9210*08•4064 0R3A *20.5 1 I •c*i1 10 100 1000 10000Retention kterv [sedAverage1.009208* C06 . •DR44AI I I •i1 10 100 1000 10000Retention kiterv [aed98Spatial Memory in PigeonsPerformance on the 5-sec RI trials was highly variable and tendedto decrease as the length of the other RI’s increased. For all threesubjects, performance was lowest on the 5-sec RI trials when the middleand longest RIs were 20 and 40 mm, respectively, althoughthisdifference was not significant [ F(4,8) — 3.3665, p < .07 1. Honig(1987) found a similar effect in a delayed matching to sample paradigmwith color stimuli. In his experiments, performance on a common RIdecreased as the average length of the RI’s increased (e.g., performanceon a 5-sec RI was worse when the 5-sec interval was embedded in adistribution including 5- and 10-sec RI’s relative to a distribution inwhich the 5-sec RI was embedded in a distribution including 1- and5-secRI’s). He attributed the effect to an expectancy of the averageduration of the RI. Such an explanation could also account for thefindings from the present research.EXPERIMENT 4The results from the previous three experiments clearly indicatethat discrimination training facilitates performance on a modifieddelayed matching of key location task. However, it is also clear thatthe observed level of performance is unlikely to reflect the full extentof pigeons’ ability to remember the location of the previously rewardedkey in this type of paradigm. The close proximity of the responsealternatives and the lack of visible landmarks are almost certainly99Spatial Memory in Pigeonsfactors inhibiting performance. Recall that Wilison and Wilkie (1991)observed retention of up to 24 hr under some conditions when they usedsimilar procedures but a larger Plexiglas apparatus from which room cueswere visible.One potential problem with the procedure that they used was thefact that the to-be-remembered location was not the location at whichthe birds actually received food (see Figure 1). Remembering where torespond to get food may be more difficult than remembering a location atwhich food is actually available.Another potential problem with Willson and Wilkie’s procedure wasthat correct responding depended on the use of a win-stay strategy.Bond et al (1981) have suggested that win-stay may be the type ofstrategy that pigeons use when foraging because they typically forage inlocations that are not depleted in a single foraging bout. However,Olson and Maki (1983) demonstrated that pigeons readily learn a win-shift strategy and learn win-shift faster than win-stay. Plowright andShettleworth (1990) make a similar point based on pigeons’ performanceon a two-armed bandit problem (TAB). TAB procedures require that thesubject discover which of two response alternatives provides the highestpayoff (food is typically dispensed on a random ratio schedule). Thevalues of the alternatives (i.e., the number of responses required, onaverage, to receive a reinforcement) typically change either within orbetween sessions and the subject must track the changes to maximizetheir rate of food intake. In Plowright and Shettleworth’s experiment100Spatial Memory in Pigeonsthe values of the response alternatives changed between sessions so thepigeons’ task was to figure out each day which of the two alternativeshad the highest payoff. Plowright and Shettleworth tested a particularmathematical model of choice behavior in this situation and found areasonably good fit to their data. Any deviations from the modeldepended on a strong tendency to shift away from the most recentlyreinforced alternative. Subsequent experiments clearly showed thatshifting was learned more readily than staying and that this tendency toshift depended upon the proximity of the response alternatives. Thepigeons showed a strong tendency to shift when the response alternativeswere widely separated but not when they were close together.However, Shettleworth and Plowright (1989), using a similar TAproblem, found that pigeons exhibited a strong tendency to return topreviously rewarded locations. It seems, then, that pigeons are capableof both staying and shifting. Their strategy depends upon proceduralvariables and the method by which the tendency to stay or shift isassessed. At the molecular level, the level of individual responses,pigeons show a strong tendency to shift away from previously rewardedlocations. At the molar level, the level of overall preferences, theyshow a strong tendency to return to previously rewarded locations.The procedure presented below addresses these potential problemsby assessing spatial memory in a large Plexiglas operant chamber. Apecking key and food hopper are mounted on each wall so the to-beremembered location is also the location at which the pigeons receive101Spatial Memory in Pigeonsfood and it can be encoded in terms of the landmarks external to thebox. The pigeons’ task is to discover, each day, which one of the fourpecking keys provides grain. Retention for the location of thepreviously rewarded key is assessed during an initial 1-mm period atthe start of the session. Reinforcement is not available during thisassessment. In essence, the task, in terms of memory, has no correctanswer. What the birds do during the initial portion of the trial isindependent of what happens during the later part of the trial (i.e.,there is no contingency between responding during the “test” andresponding later in the trial). This flexibility should allow the birdsto exhibit their natural tendency. If they remember the location of thepreviously rewarded site they should either return to that location atthe start of the next session (a win-stay strategy) or avoid thatlocation (a win-shift strategy). Either outcome would indicate that thebirds remembered the previously rewarded site. If the birds cannotremember the location of the previously rewarded key their responsesduring the assessment phase should be distributed randomly or theyshould exhibit some systematic searching strategy (i.e., always startsearching at key 1).102Spatial Memory in PigeonsMETHODSubjectsThe subjects were four King pigeons who had previously served in avariety of different behavioral experiments but who were naive withrespect to the procedures used in the present study. Housing conditionswere identical to those employed in the previous experiments.ApparatusA 3.5 cm-diameter key was mounted on the center of each wall ineach of two large square Plexiglas chambers. Each key was mounted 20 cmabove the floor. Behind each key was a microswitch, which sensed peckshaving a force greater than .15 N, and a 28 VDC #313 lamp covered by ared gelatin filter (about 1.5 cd/rn2). A standard grain feeder wasmounted directly below each key. Each chamber was located in a small(about 2 m x 2 m x 3 rn), well lit testing room. Subjects could see avariety of room cues (window, door, wall posters, etc) through thetransparent Plexiglas walls of the boxes. The floor area of the boxeswere 3600 cm2 and 2025 2 Data collection and experimental controlwere carried out by the MANX language (Gilbert & Rice, 1979) running ona minicomputer.103Spatial Memory in PigeonsProcedureBecause each subject had previous key peck training, nopreliminary training was required. Each subject received 51 sessions oftraining during this experiment. Sessions occurred atapproximately thesame time (a few hours after light onset) each day, 5 days a week.Sessions began with the illumination of the four identical red peckingkeys and lasted for 17 mm following the first response. During allsessions food reward for key pecking was unavailable during an initialperiod that lasted 1 mm. This unreinforced intervalis represented inblack in Figure 9, which shows a schematic of two consecutive sessions.After this initial non-rewarded period, food for keypecking wasavailable according to a VI 30-sec schedule on one of the four keys.The key that provided food, represented by the crosshatched bar inFigure 9, was selected randomly in each session by a computer program.Thus, during each daily session the probability was 0.25 that aparticular key would produce food. The probability that thesame keywould produce food on two consecutive sessions was 0.0625. In summary,the key feature of the procedure was that the location of the profitablekey varied in an unpredictable way from session to session. The basicrequirements of the task were that subjects had to first locate the keythat provided food intermittently and then respond on that key to obtainas many as 32, on average, 5-sec food reinforcements per session.In each session a record was kept of how many times the subjectspecked each of the four keys, during both the initial 1-mm nonrewarded104Spatial Memory in PigeonsFigure 9. Two consecutive sessions are shown schematically. The blackbar represents a period (1 mm in length), initiated by thefirst keypeck, during which food was unavailable. Thecrosshatched bar represent the key on which food wasavailable on a VI 30-sec schedule for the last 16 mm of thesession (key 3 in the Day N-i session and key 1 in the Day Nsession). During this period food was not available on thekeys represented by grey bars.105Spatial Memory in Pigeons0 5 10I.4rutes15fstfree>-UitwooneL I0 5 1510MrutesDay N-i DayNIortreetwo106Spatial Memory in Pigeonsperiod, and the subsequent 16-mm period in which food was available onone of the four keys. Frequencies of pecking the four keys were used toform two discrimination ratios (DR). The first was based on the last 16mm of the session and was the ratio of pecks on the rewarded key to thetotal number of pecks to all keys. For example, in Figure 9, whichshows two consecutive sessions, this DR would be the ratio of pecks onkey 3 to the total pecks on all keys during the session labelled “Day N1”. In the session labelled “Day N” this DR would be the ratio of pecksto key 1 to the total pecks on the other keys. This DR, referred to asthe “terminal” DR, is a measure of the subjects’ ability to locate, andthen exploit, the profitable key within a session.The second DR was calculated from the pecks made during theinitial unrewarded period and was the ratio of pecks made on Day N tothe key that had been rewarded on Day N-l to the total number of pecksmade to all keys on Day N. For example, in Figure 9 this DR for Day Nwould be the ratio of pecks on key 3 (the rewarded key on Day N-i) tothe total number of pecks made in the initial nonrewarded period. ThisDR, referred to as the “initial” DR, is a measure of perseveration ofresponding, from Day N-i to the beginning of Day N, to the location thathad previously provided food. Response perseveration implies thatsubjects remember the location of food from the previous day (avoidingthe previously reinforced key would imply the same thing). Randomresponding would yield DRs of 0.25 in both the initial nonrewardedperiod and in the subsequent rewarded period because there are four107Spatial Memory in Pigeonspossible response alternatives. We also calculated a DR based on thenumber of responses made on Day N to the key that had been rewarded onDay N-2. This DR was a ratio of the pecks to the key that had beenrewarded on Day N-2 to the total responses minus responses to the keythat had been rewarded on Day N-i. This DR was calculated this waybecause it became clear that responses during the initial unrewardedperiod were not randomly distributed, but were biased towards the keythat had been rewarded on the previous day. Under this protocol, randomresponding would yield a DR of .333. Days on which the same key wasrewarded in consecutive sessions were not included in the N-2 analysis.RESULTS AND DISCUSSIONFigure 10 shows initial and terminal DRs for each subject, basedon the last 50 sessions in the experiment. The first, crosshatched, baris the terminal DR (i.e., relative responding on the rewarded key duringthe final 16 mm of a session). This DR is an index of subjects’ability to find, and respond to, the rewarded key. This DR is muchhigher than chance (0.25) for all subjects.The grey and black bars in Figure 10 show the initial DRs (i.e.,relative responding to the key that had been rewarded on the previousday during the initial unrewarded period of each session). The greybars show sessions separated by 24 hr (i.e., Tuesday through Fridaysessions). The black bars show Monday sessions that occurred 72 hrafter the previous session. All subjects responded primarily to the key108Spatial Memory in PigeonsFigure 10. DRs for subjects in Experiment 4. The first, crosshatchedbar is the DR for the final 16 mm of each session. Thisterminal DR is a measure of how well subjects located andexploited the profitable key. The DRs represented by thegrey and black bars are DRs from the first 1 mm of eachsession. These initial DRs measure subjects’ tendency torespond on a key that was rewarded during the previoussession. The grey bars represent sessions separated by 24hours, the black bar, sessions separated by 72 hours.109Spatial Memory in PigeonsCl)0iiC04dcr5C1.00.80.60,40.20.0one two three four meanSubject110Spatial Memory in Pigeonsthat had been rewarded during the previous session, even after 72 hours.Both the 24- and 72-hr DRs were significantly above chance [one-samplet-test of t(3) — 14.0, p — .008 and t (3) — 14.26, p — .007,respectily].The results shown in Figure 10 show that subjects learned thelocation of food, remembered that information over periods of manyhours, and used that information when deciding where to begin searchingfor the profitable key during the next session. This experiment lasted51 sessions and the above results were based on the final 50 sessions.Initial DRs averaged over blocks of 10 sessions were also calculated.In examining these data only one interesting trend was observed - theinitial DRs tended to get smaller over blocks. Figure 11 shows averageddata over the five blocks of 10 sessions. These data suggest thatsubjects were less likely to perseverate from day to day with increasingexposure to the conditions of the experiment. This decrease could be aconsequence of several processes: the subjects’ unlearning of a “winstay” strategy, subjects’ learning that food location varies randomlyfrom day-to-day, or a cumulative proactive memory interference effect.However, caution must be used in interpreting this trend because thedecrease over blocks was not statistically significant when analyzed bya one-factor repeated measures analysis of variance [ F (4,12) — 2.01, p— .156].111Spatial Memory in PigeonsFigure 11. Initial DRs for the five blocks of 10 sessions thatcomposed Experiment 4.112DiscriminationRatio92000-0N))0IIItD 0 0N) C3 OlIII....I-.I.... I-.Qq 0Spatial Memory in PigeonsAn initial DR based on the number of pecks made during Day N tothe key that had been rewarded 2 days before (i.e., on Day N-2) was alsocalculated. These DRs are shown in Figure 12. Again, subjectsresponded more on this key than would be expected by simple randomresponding to three keys, which would lead to a DR of 0.333. The DRsshown in Figure 12 are significantly larger than 0.333 ( one sample t(3) — 5.54, p —.012 ].The results of this experiment show that pigeons tend to begineach session by responding primarily to the key that had been rewardedduring the previous session. They also respond to the key that had beenrewarded two sessions earlier. Such response perseveration implies thatpigeons remember the location of a previously rewarded key over manyhours, in some cases for as long as 96 hr. This level of retention isfar greater than that previously seen with pigeons on any other spatialmemory task.EXPERIMENT 5The main purpose of Experiment 4 was to examine what pigeons woulddo on a spatial task in which the location of a profitable pecking keyvaried randomly from session to session. The results of that experimentunequivocally demonstrated that the birds remembered the location of theprofitable key from previous sessions and used that information indeciding where to begin searching at the beginning of the next session.114Spatial Memory in PigeonsFigure 12. Initial DRs calculated for the key that was profitable 2days earlier in Experiment 4.115Spatial Memory in Pigeons0ctC0CSC-)C’)1.00.80.60.40.20.0one two three four meanSubject116Spatial Memory in PigeonsThey clearly adopted a win-stay strategy and used that strategy toreturn to locations that they had not been exposed to for as long as 72hr. This is far better performance than has been demonstrated in otherspatial memory procedures with this species. In the present experimentthe same procedure was used but a more detailed recording of withinsession responding was obtained. By examining the patterns ofresponding within sessions it should be possible to gain some insightinto the relationship between discovering the profitable key, howquickly the pigeons learn about the location of the profitable key andhow memory for that location is influenced by the speed of discovery andthe degree to which that location is learned. The duration of theexperiment was also increased from 51 to 81 sessions to see ifadditional exposure to the conditions of the experiment would modify thestrength of the pigeons’ tendency to use a win-stay strategy.METHODSubjects and ApparatusThe subjects were four King pigeons maintained under conditionsidentical to those in the previous experiments. As in Experiment 4, allsubjects had previously been trained to peck keys for food reinforcementbut were naive with respect to the present procedures. The apparatusused in Experiment 4 was also used for the present experiment.117Spatial Memory in PigeonsProcedureBecause the subjects had previous key peck training, nopreliminary training was required. The subjects received 81 sessionsidentical to those in Experiment 4. One procedural difference in thisexperiment was that pecking during the final 16 mm was recorded ineight 2-nun blocks during the final 25 sessions of the experiment. Thiswas done in order to examine how quickly subjects were able to find therewarded key during each session.As in Experiment 4, subjects were tested at about the same timeeach day, 5 days per week. Testing was done a few hours after lightonset.RESULTSFigure 13 shows the initial DRs for each subject, averaged overall sessions in the experiment. As in Experiment 4 this DR, whichmeasures subjects’ perseveration at the key that was rewarded during theprevious session, was considerably above chance, one sample t (3) —10.07, p — .0021. Thus at the start of each session subjects respondedmore to the key that had been rewarded during the previous session.In this experiment pecking during the final 16 mm of the last 25sessions was recorded in eight 2-mm blocks. Figure 14 shows averagedDRs for each of these blocks. These data show that subjects quicklydiscovered the rewarded key. Subjects responded to the rewarded key118Spatial Memory in PigeonsFigure 13. Initial DRs, averaged over 80 sessions in Experiment 5.119Spatial Memory in Pigeons0ccC0a5C20ci,0I I I I1.00.80.60.4020.0five six seven eight meanSubject120Spatial Memory in PigeonsFigure 14. Terminal DRs for each of the eight 2-mm blocks thatcomposed a session. The data are from the last 25 sessionsof Experiment 5.121C0DiscriminationRatioI.. mo0or\\)0p 03O0) - 0)Spatial Memory in Pigeonssignificantly above chance levels even at the end of the first 2 mm ofa session [ one sample t (3) = 5.94, p .0095 ].The gradualness of the learning curves shown in Figure 14 maskoccasions in which learning seemed to be “all or none”. Figure 15 showsthree selected sessions for each subject in which there was “all ornothing” responding to the rewarded key.There was a moderate but significant, positive correlation betweenthe terminal DR from session N-i (the measure of learning) and theinitial DR from session N (the measure of retention), calculated fromthe data from the last 25 sessions of the experiment (one sample t (3)— 4.27, p .024 ). The Pearson r values for subjects 1 through 4 were:0.602, 0.489, 0.259, and 0.229, respectively. Thus between 36 and 5% ofthe variance in the initial DRs was accounted for statistically by theterminal DRs from the previous session. The subjects that showed thestrongest tendency to exploit the profitable key almost exclusivelyafter finding it (as indexed by the terminal DRs) also showed the •strongest tendency to begin responding on day N at the key that had beenrewarded on day N-i. The mean terminal DRs from the last 25 sessionsfor subjects 1-4 were: 0.582, 0.675, 0.849, and 0.876, respectively.The mean initial DRs were: 0.420, 0.429, 0.525, and 0.615.In Experiment 4 there was a trend for the initial DRs to decreaseover the 50 sessions that composed that experiment. Figure 16 showscomparable data for the present experiment. Again, there was a tendencyfor initial DRs to decline over sessions. As in Experiment 4,123Spatial Memory in PigeonsFigure 15. Terminal DRs for three selected sessions from the 25sessions in Experiment 5 in which responding was recorded ineight 2-mm blocks.124Spatial Memory in Pigeons1.00.804-0.6C020.01.00.800.6020.0MINUTESMINUTES2 4 6 8 10 12 14 162 4 6 8 10 12 14 16125Spatial Memory in Pigeons0.6 -S.bject 702 -00 I I I I•2 4 6 8 10 12 14 16MINUTES1.0_0.8-0azO.6 -02 &bject80.0 I I I2 4 6 8 10 12 14 16MUTES£26Spatial Memory in PigeonsFigure 16. Initial DRs for the eight blocks of 10 sessions thatcomposed Experiment 5.127DiscriminationRatio000Q0-01\\)0)0IIII0 C)0)-I—IJ_,-IIII-.QqSpatial Memory in Pigeonshowever, this decrease was not statistically significant when analyzedwith a repeated measures analysis of variance, [ F (1,21) = 1.89, p =.1211 1.The results of both Experiment 4 and 5 show that Day N-l learningcarries over to the early part of the Day N session. As in Experiment 4we looked for evidence that learning on Day N-2 also affects respondingon Day N. Figure 17 shows these initial DRs. These DRs aresignificantly above chance (.333) [ one sample t (3) 3.90, p = .03 ].DISCUSSIONThe results from the present study replicate and extend thefindings from Experiment 4. The subjects clearly remembered thelocation of the rewarded key from trial N-i and used that information inin deciding where to begin responding on Day N. The decreasing trend inthe magnitude of this effect observed in Experiment 4 was also evidenthere but was, once again, not significant. The all-or-none pattern ofresponding observed once the birds had discovered the rewarded keysuggests that they had learned that only one key provided access to foodeach session, but it seems unlikely that the birds were learning thatthe location of the rewarded key varied randomly from day-to-day. Thesubjects’ continued use of a win-stay strategy supports this assertion.The decreasing, but non-significant trend in the magnitude of theinitial DRs is most likely due to a build up of proactive interference129Spatial Memory in PigeonsFigure 17. Initial DRs calculated for the key that was profitable 2days earlier in Experiment 5.130Spatial Meniory in Pigeons1.00.80.40.20.00C0CE0Cl)five six seven eight meanSubjects131Spatial Memory in Pigeonsup of proactive interference (P1). The data showing that subjects alsoremember the location of the rewarded key from session N-2 areconsistent with the presence of P1.The fact that the subjects in the present experiment wereresponding to the profitable key at above chance levels even during thefirst 2 mm of the session suggests that the 16 mm sessions used in thefirst two experiments provided more than sufficient time for thesubjects to locate and subsequently remember the profitable key. InExperiment 6 the duration of the sessions was systematically manipulatedin an attempt to discover the relationship between session length andmemory for the location of the profitable key.EXPERIMENT 6The within-session data from the previous experiment showed thatbirds quickly discovered the location of the profitable key. Thissuggests that the birds may have been forming strong associationsbetween particular locations and the presence or absence of food at anearly point in the session. In this experiment the duration of thesessions was systematically shortened in order to examine how muchexposure the birds required to form associations that would persist for24 hr.132Spatial Memory in PigeonsMETHODSubjects and ApparatusThe same subjects and apparatus as used in Experiment 5 were usedin this experiment.ProcedureAfter the completion of Experiment 5 the subjects were given anadditional 20, 16-mm sessions. These sessions were similar to those inExperiments 4 and 5: The first peck each day to one of the fourilluminated pecking keys initiated a 1-mm period in which food wasunavailable and pecking was measured; this was followed by a 16-mmperiod in which food was available intermittently on one key, theidentity of which was chosen at random each day. The subjects weresubsequently given 20 sessions in which the rewarded, second part ofeach session was shortened from 16 mm to 12 mm. After these sessions,the subjects received 20 sessions with each of the following durations:8 mm, 4 mm, and finally, 1 mm. These durations were timed from whenthe subject actually discovered the profitable key (i.e,. following thefirst reinforcement).133Spatial Memory in PigeonsRESULTSFigure 18 shows the initial DRs for each subject, averaged overall sessions of a particular duration. As in Experiments 4 and 5 thisDR, which measures subjects’ perseveration to the key that was rewardedduring the previous session, was considerably above chance (.25) for 16-mm sessions, one sample t (3) 8.21, p .0038. All of the subjects’DRs were above chance during the 12-mm sessions, but because of thelarge variance produced by subject 5’s strong tendency to perseverate tothe key that was rewarded on the previous day during the session, theone sample t-test of performance against chance was not significant, [ t(3) =2.22, p .ll ]. A t-test of performance versus chance with subject5 excluded was significant for these sessions, [ t (2) 11.00, p.0082 1.The initial DRs remained above chance during both 8- and 4-mmsessions, t (3) = 3.57, p =.038, and t (3) = 3.52, p = .039,respectively. Although the group mean DR was apparently above chanceduring the 1-mm sessions ( H = .36 ), the t-test was not significant,[ t (3) = 1.90, p = .15 ].The session duration variable was also analyzed by a repeatedmeasures analysis of variance. This analysis showed no effect ofduration, F (4,12) = 1.103, p = .3999.134Spatial Memory in PigeonsFigure 18. Initial DRs for each session duration employed in Experiment6.135Spatial Memory in PigeonsSitject 51.0 I0.80.60.4O2-0010 5 10 15 20Dt,at’on&ISubject 6100.8060.40.20.00 5 10 15 20Disation136Spatial Memory in PigeonsSubject 71008.206E04 /02CC)0 5 15 2CSubjeDt S1008.2a: 06000 5 10 15 20Du’at’137Spatial Memory in PigeonsDISCUSSIONThe basic finding of Experiment 6 was that pigeons require only ashort exposure to the location of a rewarded pecking key in order toremember it for 24 hr. Two of the four subjects continued to showstrong preferences for the N-l key even with sessions of only 1 mm inwhich they received, on average, only 2 reinforcements. The analysis ofvariance conducted on the session length variable did not detect anysignificant differences in performance in the different session lengths.However, a one-sample t-test showed that performance wasindistinguishable from chance at the shortest session length (i.e., 1mm). However, it should be noted that there was considerableindividual differences in performance during the 1-mm sessions;subjects 6 and 8 were considerably above chance during these very shortsessions. Although there was a non-significant trend for the magnitudeof the initial DRs to decrease over the course of Experiments 4 and 5,this trend was nonetheless confounded with decreases in session lengthin the present experiment. It is possible, therefore, that the nonsignificant t-test on the 1-mm session data was at least partially dueto this confound. Although speculative, this suggestion does seemplausible.GENERAL DISCUSSIONTaken together, the experiments reported here clearly indicatethat pigeons can retain spatial information for periods of time much138Spatial Memory in Pigeonslonger than previously demonstrated. The spatial distinctiveness of theto-be-remembered locations and the presence of visible landmarks wasclearly important. Performance was much better in the last threeexperiments, in which spatially distinct locations were used andlandmarks were available, than in the first three experiments, in whichthey were not. Discrimination training between profitable andnonprofitable sites was also very important. The results from the firsttwo experiments clearly showed that performance in the absence ofdiscrimination training remained at approximately chance levels. It wassuggested that pigeons in the present task learn associations betweenlocations and the presence or absence of food. The poor performance onSample Alone trials was explained in terms of the pigeons’ inability todiscriminate between locations recently associated with food andlocations associated with food in the more remote past. The resultsfrom Experiments 4 and 5 showing that pigeons remember profitablelocations from previous sessions support this assertion, albeitindirectly.Another possibility, one that has not been considered up to thispoint, is that Sample Alone trials and Sample/Distractor trials (and, byextension the trials in Experiments 4-6) are distinctly different tasksthat test different aspects of memory. Gaffan (1974) has distinguishedbetween recognition memory and associative memory tasks. Recognitionmemory tasks entail presentation of only the to-be-remembered stimuli inthe study phase of a trial. During the test phase of the trial a139Spatial Memory in Pigeonssubject must discriminate between a familiar stimulus and a novelstimulus. Delayed matching and non-matching to sample are examples of arecognition memory task (as is Sample Alone training in the presentcontext). In contrast, on associative memory tasks all of the stimulito be discriminated later are present during the study phase of thetrial. The subject must remember which of the previously viewed stimuliis important. Shettleworth and her colleagues (Brodbeck, Burack &Shettleworth, in press; Shettleworth, 1985) have argued that foodstoring is an associative memory task because birds often visitpotential storage sites without storing food. They do, however,preferential return to sites where food was stored. TheSample/Distractor training employed in Experiments 1-3 and the procedureused in Experiments 4-6 in the present research also fit the definitionof an associative memory task because the pigeon “visits” (i.e.,responds to) locations that do and do not provide food butpreferentially return to the sites associated with food. The parallelsbetween associative memory tasks that have been used with food-storingbirds and the present procedures will be explored in more depth below.Several results in addition to the excellent retention observed inthe current series of experiments also deserve comment. It is clearfrom the results of Experiments 4-6 that the birds were adopting a winstay strategy. In other words, they were returning to locations thathad provided food in the past. Taken at face value, this findingcontrasts with the findings of Plowright and Shettleworth (1990) who140Spatial Memory in Pigeonsobserved that pigeons are predisposed to shift (win-shift or lose-shift). As mentioned above, these findings are not contradictory whenconsidered in their appropriate context. The tendency to stay wasconsidered in terms of an overall preference. Single responses were notconsidered. Plowright and Shettleworth (1990) examined individualchoices and observed a tendency to shift. Related research(Shettleworth & Plowright, 1989) that considered overall preferencesfound a tendency to return to previously profitable locations. Theconclusion offered depends upon the level of analysis. Severalnaturalistic studies have observed a similar phenomenon (Gass &Sutherland, 1985; Smith & Sweatman, 1974; Zach & Falls, 1976).For example, Gass and Sutherland observed hununingbirds foraging onpatches of nectar-producing flowers. In some conditions, the authorsenhanced patches by adding sugar water to the flowers. The hummingbirdsquickly discovered and exploited these “enriched” patches and, at thestart of foraging the next day, returned to the previously enrichedpatches.An additional finding from the present research that deservescomment is the failure to observe enhanced performance followingreinforced and extended sample presentation during Sample Alonetraining. This finding, when taken at face value, appears to contradictprevious work showing that increasing the duration of the samplepresentation and reinforcing the sample improve matching performance(Wilkie & Summers, 1982; Wilkie, 1983). If increasing sample duration141Spatial Memory in Pigeonsand reinforcing the sample are considered as techniques for increasingthe informational value of the sample stimulus, the present results arenot surprising. Presumably increasing the sample duration improvesmatching performance by providing the subject with ample opportunity toencode the identity of the sample. Reinforcing the sample presumablyimproves matching performance by identifying the sample as a stimulusassociated with reinforcement. At some point, increasing the sampleduration further and providing additional pairings of the sample withfood does not add any further information. The session durationmanipulations in Experiment 6 suggest that a 1 minute sample and twopairings (on average) of the sample with reinforcement are sufficient toreach this limit. However, this explanation must be viewed with cautionsince there were other factors in addition to extended samplepresentation and reinforcement present during that experiment (i.e.,discrimination training). An additional factor that suggests a cautiousacceptance of this explanation is that sample duration and reinforcingthe sample were not manipulated independently so it is unclear to whatdegree each of those manipulations contributed to the observedperformance.It has long been recognized that correct performance on memorytasks does not necessarily reflect intact memory. Animals can respondcorrectly through the use of non-memorial strategies or by chance (seeOlton & Samuelson, 1976). An often overlooked but complementary pointis that errors on memory tasks are not necessarily due to forgetting142Spatial Memory in Pigeons(Dale, 1988; Devenport, 1989; Roitbiat, 1980; Roitbiat & Harley, 1989;Wilkie & Spetch, 1981; Wilison & Wilkie, 1991). The findings from thepresent research add to that body of research. Performance on thepresent tasks, as measured by the percent correct first choices, wasworse than performance as measured by the discrimination ratios. It wasalso much more variable. The processes that determine performance onmemory tasks are complex and offering animals an opportunity to makeseveral “choices” during memory tasks provides a far better measure ofmemory than single responses.THE ASSOCIATIVE MEMORY HYPOTHESIS REVISITEDThe hypothesis that has been presented to explain the results fromthe current series of experiments rests on the idea that pigeons learnassociations between locations and the presence or absence of food.Ignoring for the moment the possibility that Sample Alone trials testrecognition memory whereas Sample/Distractor training trials testassociative memory, the poor performance on Sample Alone trials is aresult of a failure to discriminate between two stimuli that are weaklyassociated with food. On Sample/Distractor trials the excellentperformance observed is based on a discrimination between a stimulusstrongly associated with reinforcement and a stimulus stronglyassociated with non-reinforcement. The intermediate performance onNovel S+ and Novel S- trials implicate a possible attentionalenhancement effect of discrimination training that strengthens the143Spatial Memory in Pigeonsassociation between particular locations and the presence or absence offood. Thus the level of performance on Novel S+ and Novel S- is theresult of a discrimination between a stimulus strongly associated withreinforcement (or non-reinforcement) and a stimulus weakly associatedwith reinforcement (or non-reinforcment). The excellent performance ofthe subjects in Experiments 4-6 is based on a mechanism similar to thatmediating performance on Sample/Distractor trials except that thediscrimination is between a location that has recently been associatedwith reinforcement and three locations that have recently beenassociated with non-reinforcement.If Sample Alone training is considered as a recognition memorytask, the only conclusion that we can draw from the present research isthat recognition memory is not as temporally persistent as associativememory but that discrimination training can sometimes improveperformance on recognition memory tasks. For example, DMTS is arecognition task and Urcuioli & Callender (1989) have shown thatdiscrimination training facillitates performance on DMTS tasks. NovelS+ and Novel S- trials also fit the definition of a recognition memorytask. The stimuli presented during the study phase are not identicalthose presented during the test phase. The birds’ task on the test isto discriminate between a familiar stimulus and a novel stimulus. Inthis case, however, there are two potential sample stimuli.Wilkie and Summer (1982) showed that matching accuracy on DMKLtasks typically declines as the number of sample stimuli presented144Spatial Memory in Pigeonsduring the study phase increases. The opposite result, an increase inperformance, was obtained in the present work. Clearly, thediscrimination training during the study phase of the trial had someenhancing effect. Based on previous research by Urcuioli and Callender(1989) the most likely explanation is enhanced attention to the locationof the sample stimuli.An important function of hypotheses is to organize an existingbody of data and generate testable predictions from their synthesis.The associative memory hypothesis offered here adequately accounts forthe present results and for the results from several previousexperiments. The approach adopted in the next section is to apply theideas developed here to our own previous work (Willson & Wilkie, 1991)and then extend the analysis to other paradigms that have been used toexamine spatial memory in pigeons. The present work will then becompared to work on associative memory in food-storing birds.Comparisons across experiments and across species are somewhatproblematic because of procedural differences but there are interestingparallels between previous work and the current work that deservecomment. There are interesting differences as well and thesedifferences suggest several fruitful avenues for future comparativework. That issue will be addressed in a later section.145Spatial Memory in PigeonsWillson and Wilkie (1991) revisitedWilison and Wilkie found results very similar to those reportedhere. This should not be surprising given that one of the main purposesof the current research program was to adapt the procedure that theydeveloped for use with the traditional DMKL apparatus and multiple dailytrials. An additional aim was to see if it was possible to improve uponthe spatial memory performance that Wilkie and Summers (1982) observed.The current work replicates and extends the major findings from thatprevious research and these findings have been interpreted above asexpressing associative memory. Several other aspects of their data arealso consistent with the associative memory hypothesis.Recall that Willson and Wilkie found evidence of proactiveinterference during the Sample Alone training. The associative memoryhypothesis suggests that these P1 effects were largely a result of thetraining protocol. The stimuli that they used were pairs of keys thatwere equidistant from the feeder. Each member of the pair served as thesample on one trial and the distractor on another within each block of10 sessions. The same key pair was never used on consecutive days.Consider the following hypothetical case during the first block of10 trials. On trial N-2, key 1 serves as the sample and key 10 servesas the distractor (see Figure 1). Key 1 becomes associated withreinforcement. Key 10 remains fairly neutral because it has never beenexplicitly paired with reinforcement or non-reinforcement. During thetest, the pigeon should correctly choose Key 1 because it has recently146Spatial Memory in Pigeonsbeen paired with reinforcement. On trial N, key 10 serves as the sampleand becomes associated with reinforcement. Key 1 retains a fairly highassociative value because it was recently paired with reinforcement andhas not been explicitly paired with non-reinforcement. It will havelost some associative strength due to the tendency of associative valueto “drift” towards zero but will remain quite high. During the test,the pigeon must decide between Key 10, which has most recently beenassociated with reinforcement, and Key 1, which has also been associatedwith reinforcement. This is a more difficult discrimination and errorsare much more likely to occur. When more days intervene between the useand re-use of a particular key pair, the discrimination becomes easierbecause the associative value of the distractor has “drifted” furthertowards zero.Two other predictions follow from this account. First,performance during the first 5 days of Sample Alone training should havebeen better than performance in subsequent blocks. Two out of threebirds showed this trend. Second, performance on Sample Alone trialsfollowing Sample/Distractor training should have been worse thanperformance on Sample Alone trials prior to Sample/Distractor trials.This is a counterintuitive prediction in light of the hypothesizedattentional enhancement function of discrimination training. However,it was also hypothesized that discrimination training results in ahigher associative value for a particular location relative to SampleAlone training because of enhanced attention to the location of the147Spatial Memory in Pigeonssample. Therefore, on Sample Alone trials following Sample/Distractortraining the pigeon must choose between two stimuli of high associativevalue. This trend was apparent in Wilison and Wilkie’s data but was notsignificant, probably because of the small number of subjects in eachcondition.Zentall, Steirn and Jackson-Smith (1990) revisitedThe associative memory hypothesis can also account for thefindings of Zentall et al. They tested pigeons on an operant analog ofthe radial arm maze. Two of their findings seems especially importantin relation to the hypothesis under consideration here. Recall thatthey found that a delay interpolated between the second and thirdchoices or between the third and fourth choice was more detrimental toperformance than a delay after the first choice or before the lastchoice. They interpreted their results in terms of flexible coding;early responses were controlled by retrospective strategy, and laterresponses were controlled by a prospective strategy. The associativememory hypothesis suggests a mechanism by which this change in choicecriterion could be implemented. Early in the response sequence thepigeons respond by avoiding locations that have recently been associatedwith food. Later responses are controlled by avoiding locations thathave not recently been associated with food.In Zentall et al’s procedure, the first choice was always correct(assuming that the bird fulfilled the response requirement before148Spatial Memory in Pigeonsswitching) and the chosen stimulus was reinforced. Thus in order torespond correctly on the second choice the pigeon must simply avoid thelocation that has most recently been associated with food. On the thirdchoice, assuming that the bird has responded correctly on the first twochoices, the bird must avoid two locations that have recently beenassociated with reinforcement and the probability of making an errorincreases (due to the increased possibility of confusing visited andnon-visited sites). Following the third choice, the bird’s responsecriterion shifts from on avoiding locations recently associated withreinforcement to approaching locations that have not recently beenassociated with reinforcement. There are two such locations availableand the probability of making an error remains high (due to potentialconfusions between visited and non-visited locations). On the fifthchoice, correct responding depends on approaching the only location thathas not been associated with reinforcement so the probability of makingan error is low.Zentall et al also found evidence of P1. Performance on latertrails within a session was worse than on earlier trials. This P1likely results from the increasing similarity in the value of the to-bediscriminated locations on successive trials within the session.Associative value is assumed to drift towards zero during ITI’s butnonetheless remains high for all of the stimuli at the beginning of atrial later in the session.149Spatial Memory in PigeonsComparisons with the study of associative memory in food-storing birdsComparisons between experiments and across procedures areproblematic because of differences in the conditions under which theexperiments were run. Small details that are seemingly insignificantfrom the point of view of the experimenter can have profound influenceson behaviour. Comparisons between species across experiments areespecially prone to these problems. However, the procedure employed inExperiments 4-6 is functionally similar to one used by Brodbeck, Burakand Shettleworth (in press) to study one-trial associative memory inBlack-Capped Chickadees, a food-storing member of the parid family.In one of Brodbeck et al’s experiments (Experiment 1), chickadeesreleased into a large indoor aviary encountered three feeders, one ofwhich contained a peanut. Once the chickadee had discovered the baitedfeeder it was allowed to peck at the peanut for 30 sec. The lights werethen turned out and the bird returned to its cage. During the retentioninterval, which lasted 5 mm, the feeders were covered with Post-itnotes to conceal the peanut and the bird was re-admitted to the aviaryand allowed to search for the now hidden bait. Visits to feeders wererecorded and the bird was allowed to search until it discovered andconsumed the peanut. For one group of chickadees the locations of thefeeders and the location of the peanut varied randomly from trial totrial (the unique group). For another group, the location of thefeeders remained constant but the location of the peanut varied amongthe three feeders.150Spatial Memory in PigeonsOn being readmitted to the aviary, birds in the unique grouprelocated the hidden peanut in fewer responses than would be expected bychance and continued to do so throughout the course of the experiment.At the start of the experiment, the birds in the non-unique group alsoperformed above chance but their performance deteriorated to chancelevels with repeated testing. An examination of their response patternsrevealed that they were systematically visiting the feeders during therecovery phase (i.e., used a response algorithm).The procedure used in the present research is most similar to thatexperienced by the non-unique group in Brodbeck et al’s study. Ignoringfor the moment the fact that the temporal parameters of the two studieswere very different and that the method of assessing retention was alsodifferent, the pigeons in the present experiments performed better thanthe chickadees in Brodbeck et al’s experiment. The pigeons in thepresent research clearly remembered the location of the previouslyprofitable key even after 72 hr and with initial exposure to the correctkey of only 4 mm (two birds performed above chance levels with exposureof only 1 mm). The chickadees performed at chance levels with a delayof 5 mm. It is possible, but unlikely, that the chickadees might haveperformed better with a longer retention interval. However, even ifsuch were the case, unpublished research from our laboratory suggeststhat the preference for previously rewarded keys gets stronger withshorter retention intervals (Willson, Wilkie & Lee, unpublished data).151Spatial Memory in PigeonsOf course, the differences between the the procedures used hereand those used by Brodbeck et al (in press) makes these points extremelyspeculative at best. However, the data do suggest several promisingavenues for further research. It is possible that the longer retentionof pigeons on the non-unique task reflects a well developed ability to“relabel” potential feeding sites based on current experience that isabsent or not so well developed in chickadees. If such were the caseone would expect that spatial memory in chickadees would be more proneto P1 than in pigeons. Another important issue is the speed with whichassociations between locations and the presence or absence of food areformed. Food-storing birds clearly form this type of associationquickly, especially with unique locations. The results of the presentexperiments suggest that pigeons also form these associations quickly,but it remains unclear how they would perform with unique locations.These points are explored further in a later section.ECOLOGY A1D COMPARATIVE COGNITIONThe idea that an animal’s cognitive capacities have been shaped byevolution and the demands of the natural environment is well accepted incomparative cognition. In the late 50’s and early 60’s researchers, ina sense, rediscovered Small’s (1901) assertion that an animal’scognitive capacities can be understood only in relation to the problemsthat an it faces in the natural environment. This rediscovery wasprecipitated by a growing realization that animals were not the tabula152Spatial Memory in Pigeonsrasa that some behaviorists proposed. A growing body of evidencesuggested that animals came into the laboratory with strongpredispositions to behave in particular ways in response to particularproblems (Bolles, 1970; Breland & Ereland, 1961; Hinde & StevensonHinde, 1973). The notion of biological constraints on learning is stillan issue today (Rozin & Schull, 1988; Sherry & Schacter, 1987), althoughthe phenomena are generally considered in terms of “adaptivespecializations” in learning and memory rather than constraints.Unfortunately, this approach has often been misapplied. Appealsto an animal’s ecology have been used to justify the existence or lackof a particular skill as measured by a particular procedure (see Bond etal, 1981). However, statements such as “Species A can’t do task Bbecause its environment isn’t conducive to the evolution of thenecessary cognitive skills” add nothing to our knowledge aboutcognition. Functional accounts of behavior can be of use by definingthe problems that an animal might face in the natural environment.Similarly, the knowledge about particular cognitive abilities that wegain in the laboratory can increase our understanding of how animalscope with the demands of the natural environment (Cheverton, Kacelnik &Krebs, 1985; Shettleworth, 1988).A brief examination of the foraging behaviour of pigeons providesa good illustration of this approach. Pigeons are gregarious and foragein a limited number of traditional feeding sites that are notnecessarily depleted in a single foraging bout (Goodwin, 1967; Levi,153Spatial Memory in Pigeons1974). They also tend to revisit the same sites on a daily basis (Levi,1974; Murton, Coombs & Thearle, 1972). They can, however, discover andexploit new sources of food when they become available and also seem tobe sensitive to the availability of food resources as a function of timeof day (Murton, Coombs & Thearle, 1972).The above description suggests that the foraging behavior ofpigeons should be influenced by the presence or absence of conspecifics(i.e., the presence of foraging conspecifics is a reliable cue for aprofitable feeding site), the temporal availability of food and,perhaps, recent experience with particular foraging sites (and manyother factors).Lefebvre and Palameta (1988) have demonstrated that foragingbehavior is influenced by social factors and recent work from ourlaboratory has demonstrated that pigeons can learn time-placeassociations (Wilkie & Willson, in press). The results of the presentresearch suggest that recent experience with a profitable feeding sitecould potentially influence where a pigeon decides to begin foraging.At the present time, our knowledge of these potential determinants offoraging behavior is limited and the issues require additionalattention. Our knowledge of how these factors interact is non-existentand this problem offers a productive area for further research, both interms of understanding the mechanisms that determine foraging in thewild and in understanding the cognitive abilities of pigeons.154Spatial Memory in PigeonsTaken at face value, the above analysis suggests that foraging inpigeons and foraging by food-storing birds such as chickadees representdifferent problems and therefore that the cognitive skills used bypigeons and food-storers during foraging are likely to be different.However, this interpretation must be viewed with caution. The idea thatforaging is determined by different cognitive mechanisms is not reallythe issue (for example, social factors are unlikely to affect foragingby food-storers in the same ways that they influence pigeon foraging)nor is the potential existence of adaptive specializations in learningand memory at issue. At a functional level pigeons and food-storersface a similar problem while foraging, namely that of returning tolocations where they expect to find food. For food-storers this is alocation where the bird has created its own food source (i.e., a cache).For pigeons this is a location where they have encountered food in thepast. However, previous research with food-storers has shown thatmemory for stored versus encountered food is similar in food-storingbirds (Shettleworth, Krebs, Healy & Thomas, 1990), so the problems facedby pigeons and food-storers are in fact functionally similar.The results of the present research suggest that pigeons canreturn to previously profitable locations in the face of extended delaysand that the temporal persistence of their memory for that location isfar more resilient than suggested by previous laboratory research. Themechanism by which they accomplish this feat is proposed to be theformation of an association between particular locations and the155Spatial Memory in Pigeonspresence or absence of food. A similar mechanism has been proposed forfood-storing birds (Brodbeck et al, in press; Shettleworth, 1985). Thistentative parallel between species suggests several promising avenuesfor further research. For example, do food-storing birds formassociations between locations and the presence or absence of food morerapidly and effectively than pigeons? When trained under equivalentconditions are such associations better retained by food-storers? Arefood-storers more prone to proactive interference than pigeons, and ifso why? Are associative memory processes for spatial and non-spatialstimuli similar in these species? The number of potential questions isalmost infinite but answers to any subset of those questions wouldgreatly increase our understanding of cognition and would also bear onthe issue of whether the memory performance of food-storing birds is anexample of an adaptive specialization in learning and memory. Answersto these questions will depend upon direct comparisons between specieson a variety of tasks. Devising appropriate procedures to do so willnot be an easy task but a detailed analysis of the task demands ofexperimental procedures currently being used to study cognition wouldcertainly provide a solid place to start.CONCLUSIONSThe traditional approach to the study of comparative cognition,the approach that has been adopted here, has been justifiably criticized(Macphail, 1982) on the grounds that we have learned very little about156Spatial Memory in Pigeonsspecies differences using this approach. It was argued in theintroduction that this lack of progress was largely a result of amisunderstanding about the types of information needed to makeinferences about species differences within this approach. In order tomake strong inferences about species differences it is necessary tocompare performance on a variety of tasks in which the relativedifferences between species vary (systematic variation, Macphail, 1982,1987). By analyzing differences in task demands it is possible to makeinferences about differences in cognitive ability.Comparisons between pigeons and other species on spatial memorytasks have consistently demonstrated that pigeons are apparentlyinferior to many other species. More recent work has painted a somewhatless bleak picture of pigeons’ spatial memory abilities (see Spetch,1990; Wilison & Wilkie, 1991) but they still appear to be lessproficient on spatial tasks. Spatial memory encompasses a vast range ofpotential processes. The range of spatial problems that animals face inthe natural environment argue against the idea of an all-encompassingspatial memory.Some spatial tasks performed by animals in the natural environmentare unique to the ecology of the animal in question. Homing in pigeonsand the food-storing behavior of some corvids and parids are obviousexamples. However, it is possible, by analyzing the components of theseapparently unique problems, to devise procedures for making validcomparisons. For example, the use of landmarks, an ability clearly157Spatial Memory in Pigeonsimplicated in pigeon homing (Fuller, Kowalski & Wiltschko, 1983), hasbeen studied in a variety of species. The results of the current seriesof experiments suggest that spatial associative memory, a cognitiveability thought to be critical in food-storing behavior, could also be aproductive avenue for further comparative research.The present research has been interpreted within the traditionalcomparative framework. This is a somewhat misleading interpretation,given that only one species was tested and the testing procedures werenovel. There are clearly interesting parallels between the proceduresused here and similar procedures used with food-storing birds (Brodbecket al, in press). The possibility that pigeons may be better than foodstorers on spatial associative memory tasks under some conditionssuggests it is probably premature to conclude that the phenomenal memoryperformance of food-storing birds represents an adaptive specializationin memory. However, any conclusions must remain speculative until thetwo species are compared under identical conditions.158Spatial Memory in PigeonsREFERENCESAadland, J. Beatty, W.M., & Maki, R.H. (1985). Spatial memory ofchildren and adults as assessed in the radial arm maze.Developmental Psychobiology, 18, 163-172.Balda, R.P. (1980). Recovery of cached seeds by a captive Nucifragacaryocatactes. Zeitschrift fur Tierpsycholgie, 52, 331-346.Balda, R.P., & Kamil, A.C. (1989). A comparative study of cacherecovery by three corvid species. Animal Behaviour, 38, 486-495.Bitterman, M.E. (1975). The comparative analysis of learning. Science,188, 699-709.Blough, D.S. (1959). Delayed matching in the pigeon. Journal of theExperimental Analysis of Behavior, 2, 151-160.Bolles, R.C. (1970). Species-specific defense reactions and avoidancelearning. Psychological Review, 77, 32-48.Bond,A.B., Cook, R.G., & Lamb, M.R. (1981). Spatial memory andperformance of rats and pigeons in the radial-arm maze. AnimalLearning & Behavior, 9, 575-580.Breland, K., & Breland, M. (1961). The misbehavior of organisms.American Psychologist, 16, 681-684.Brodbeck, D.R., Burack, 0.R., & Shettleworth, S.J. (in press). Onetrial associative memory in Black-Capped Chickadees. Journal ofExperimental Psychology: Animal Behavior Processes.Spatial Memory in PigeonsBrown, M.F., & Cook, R.G. (1986). Within-trial dynamics of radial armmaze performance in rats. Learning and Motivation, 17, 190-205.Cartwright, B.A., & Collett, T.S. (1983). Landmark learning in bees.Journal of Comparative Physiology A, 151, 521-543.Cartwright, B.A., & Collett, T.S. (1987). Landmark maps for honeybees.Biological Cybernetics, 57, 85-97.Cheng, K. (1986). A purely geometric module in the rat’s spatialrepresentation. Cognition, 23, 149-178.Cheng, K. (1988). Some psychophysics of pigeon’s landmark use. Journalof Comparative Physiology A ,162, 815-826.Cheng, K. (1989). The vector sum model of pigeon landmark use. Journalof Experimental Psychology: Animal Behavior Processes, 15, 366-375.Cheng, K. (in press). Three psychophysical principles in the processingof spatial and temporal information. In W.K. Honig, & J.G.Fetterman (Eds.), Cognitive aspects of stimulus control.Hillsdale, NJ: Erlbaum.Cheng, K., & Gallistel, C.R. (1984). Testing the geometric power of ananimal’s spatial representation. In H.L. Roitblat, T.G. Bever,and H.S. Terrace (Eds.), Animal Cognition (pp. 409-424).Hillsdale, NJ: Erlbaum.160Spatial Memory in PigeonsCheverton, J., Kalcelnik, A., & Krebs, J.R. (1985). Optimal foraging:constraints and currencies. In B. Holledobler and M. Lindauer(Eds.), Experimental Behavioral Ecology and Sociobiology (pp. 109-126). New York: Verlag.Collett, T.S., Cartwright, B.A., & Smith, R.C. (1986). Landmarklearning and visio-spatial memories in gerbils. Journal ofComparative Physiology A, 158, 835-851.Cook, R.G. (1980). Retroactive interference in pigeon short-term memoryby a reduction in ambient illumination. Journal of ExperimentalPsychology: Animal Behavior Processes, 6, 326-338.Dale, R.H.I. (1988). Spatial memory in pigeons on a four-arm radialmaze. Canadian Journal of Psychology, 42, 78-83.Dale, R.H.I., & Innis, N.K. (1986). Interactions between responsestereotypy and memory strategies on the eight-arm radial maze.Behavioural Brain Research, 19, 17-25.Devenport, L. (1989). Sampling and contextual change. Learning andMotivation, 20, 97-114.Domjan, 11., & Galef, B. (1983). Biological constraints on instrumentaland classical conditioning: retrospect and prospect. AnimalLearning & Behavior, 11, 151-161.Downs, R.M. (1979). On the nature of cognitive maps. The Behavioraland Brain Sciences, 2, 499-500.161Spatial Memory in PigeonsFuller, E., Kowalski, V., & Wiltschko, R. (1983). Orientation of homingpigeons: compass orientation vs. pilotting by landmarks. Journalof Comparative Physiology, 153, 55-58.Gaffan, D. (1974). Recognition impaired and association intact in thememory of monkeys after transection of the fornix. Journal ofComparative and Physiological Psychology, 86, 1100-1109.Gallistel, C.R. (1990). The organization of learning. Cambridge, MA:MIT Press.Gass, C.L., & Sutherland, G.D. (1985). Specialization by territorialhununingbirds on experimentally enriched patches of flowers:Energetic profitability and learning. Canadian Journal ofZoology, 63, 2125-2133.Gentry, G., Brown, W.L., & Kaplan, S.J. (1947). An experimentalanalysis of the spatial location hypothesis in learning. Journalof Comparative and Physiological Psychology, 40, 309-322.Gentry, C., Brown, W.L., & Lee, H. (1948). Spatial location in thelearning of a multiple-T maze. Journal of Comparative andPhysiological Psychology, 41, 312-318.Gilbert, S.C., & Rice, D.C. (1979). NOVA SKED II: A behavioral notationlanguage utilizing the Data General real-time disk operatingsytem. Behavior Research Methods & Instrumentation, 11, 71-73.Gill, F.B., & Wolf, L.L. (1977). Nonrandom foraging by sunbirds in apatchy environment. Ecology, 58, 1284-1296.162Spatial Memory In PigeonsGingerelli, J.A. (1927). Preliminary experiments on the causal factorsin animal learning. Journal of Comparative Psychology, 9, 245-274.Goodwin, D. (1954). Notes on feral pigeons. Avicultural Magazine, 60,190-213.Goodwin, D. (1967). Pigeons and doves of the world. British Museum(Natural History), Ithica, NY: Cornell University Press.Goodwin, D. (1983). Behaviour. In M. Abs (Ed.), Physiology andbehaviour of the pigeon (pp. 285-308). London: Academic Press.Gould, J.L. (1984). Natural history of honey bee learning. In P.Marler, & H.S. Terrace (Eds.), The Biology of Learning (pp. 141-180). Berlin: Springer-Verlag.Gould, J.L. (1986). The locale map of Honey Bees: Do insects havecognitive maps? ScIence, 232, 861-863.Gould, J.L. (1987). Landmark learning in honeybees. Animal Behaviour,35, 26-34.Grant, D.S. (1976). Effect of sample presentation time on long-delaymatching in the pigeon. Learning and Motivation, 7, 580-590.Hinde, R.A., & Stevenson-Hinde, J. (1973). Constraints on learning.London: Academic Press.Honig, W.K. (1978). Studies of working memory in the pigeon. In S.H.Hulse, H. Fowler, & W.K. Honig (Eds.) Cognitive processes inanimal behavior (pp. 211-248). Hillsdale, N.J. :Erlbaum.163Spatial Memory in PigeonsHonig, W.K. (1981). Working memory and the temporal map. In N.E.Spear, & R.R. Miller, (Eds.), Information processing in animals:Memory mechanisms (pp. 167-198). Hilisdale, N.J. :Erlbaum.Honig, W.K. (1987). Memory interval distribution effects in pigeons.Animal Learning & Behavior, 15, 6-14.Homer, J. (1984). The effect of maze structure upon the performance ofa multiple-goal task. Animal Learning & Behavior, 12, 55-61.Hull, C.L. (1943). Principles of behavior, New York:Appleton-CenturyCroft.Hulse, S.H., & O’Leary, D.K. (1982). Serial pattern learning: Teachingan alphabet to rats. Journal of Experimental Psychology: AnimalBehavior Processes, 8, 260-273.Kamil, A.C. (1978). Systematic foraging by a nectar-feeding bird, theAmakihi (Loxops virens). Journal of Comparative and PhysiologicalPsychology, 92, 388-396.Kamil, A.C., Krebs, J.R.,& Pulliam, H.R.(1987). Foraging Behavior. NewYork, NY:Plenum Press.Kamil, A.C. (1987). A synthetic approach to the study of animalintelligence. Nebraska Symposium on Motivation, 1987, 257-308.Krebs, J.R. (1990). Food-storing birds: Adaptive specialization inbrain and behaviour? Philosophical Transaction of the RoyalSociety Series B, 329, 55-62.164Spatial Memory in PigeonsKrebs, J.R., Healy, S.D., & Shettleworth, S.J. (1990). Spatial memoryof paridae: comparison of a storing and a non-storing species, thecoal tit, Parus ater, and the great tit, P. major. AnimalBehaviour, 39, 1127-1137.Krebs, J.R., Ryan, J.C., & Charnov, E.L. (1974). Hunting by expectationor optimal foraging? A study of patch use by chickadees. AnimalBehaviour, 22, 953-964.Lefebvre, L.., & Palameta, B. (1988). Mechanisms, ecology, and populationdiffusion of socially learned, food-finding behavior in feralpigeons. In T.R. Zentall & B.C. Galef Jr. (Eds.) SocialLearning: Psychological and Biological Perspectives (pp.141-164).Hillsdale, NJ: Erlbaum.Levi, W.M. (1974). The pigeon. Sumpter, SC: Levi.MacDonald, S.E., & Wilkie, D.M. (1990). Yellow-nosed monkeys’(Ceropithecus ascanius whitesidel) spatial memory in a simulatedforaging environment. Journal of Comparative Psychology, 104,382-387.Macphail, E.M. (1982). Brain and intelligence in vertebrates. Oxford:Clarendon Press.Macphail, E.M. (1987). The comparative psychology of intelligence. TheBehavioral and brain sciences, 10, 645-695.Menzel, E.W. (1973). Chimpanzee spatial memory organization. Science,822, 943-945.165Spatial Memory in PigeonsMenzel, E.W. (1978). Cognitive mapping in chimpanzees. In S.H. Hulse,H. Fowler, & W.K. Honig (Eds.) Cognitive processes in animalbehavior (pp. 341-373). Hilisdale, N.J. :Erlbaum.Morris, R.G.M (1981). Spatial localization does not require thepresence of local cues. Learning and Motivation, 12, 239-260.Murton, R.K., Coombs, C.F.B., & Thearle, R.J.P. (1972). Ecologicalstudies of the feral pigeon Columba livia. II. Flock behavior andsocial organization. Journal of Applied Ecology, 9, 875-889.O’Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive map.Oxford: Clarendon Press.Olson, D.J. (1991). Species differences in spatial memory among Clark’sNutcrackers, Scrub Jays and Pigeons. Journal of ExperimentalPsychology: Animal Behavior Processes, 17, 363-376.Olson, D.J., & Maki, W.S. (1983). Characteristics of spatial memory inpigeons. Journal of Experimental Psychology: Animal BehaviorProcesses, 9, 266-280.Olton, D.S. (1978). Characteristics of spatial memory. In S.H. Hulse,H. Fowler, & W.K. Honig (Eds.) Cognitive processes in animalbehavior (pp. 341-373). Hillsdale, N.J. :Erlbaum.Olton, D.S. (1979). Mazes, maps and memory, American Psychologist, 34,583-596.Olton, D.S., Becker, J.T., & Handelmann, G.E. (1979). Hippocampus,space and memory. The Behavioral and Brain Sciences, 2, 313-365.166Spatial Memory in PigeonsOlton, D.S., & Pappas, B. (1979). Spatial memory and hippocampalfunction. Neuropsychologia, 17, 669-682.Olton, D.S., & Samuelson, R.J. (1976). Remembrance of places passed:Spatial memory in rats. Journal of Experimental Psychology:Animal Behavior Processes, 2, 97-116.Olton, D.S., & Schlosberg, P. (1978). Food searching strategies inyoung rats: win-shift predominates over win-stay. Journal ofComparative and Physiological Psychology, 92, 609-618.Plowright, C.M.S., & Shettleworth, S.J. (1990). The role of shifting inchoice behavior of pigeons on a two-armed bandit. BehaviouralProcesses, 21, 157-178.Rescorla, R.A., & Wagner, A.G. (1972). A theory of Pavlovianconditioning: Variations in the effectiveness of reinforcement andnonreinforcement. In A.H. Black and W.F. Prokasy (Eds.),Classical conditioning II (pp. 64-99). New York: Appleton-Century-Croft.Roberts, W.A. (1981). Retroactive inhibition in rat spatial memory.Animal Learning & Behavior, 9, 566-574.Roberts, W.A. (1988). Foraging and spatial memory in pigeons (Columbalivia). Journal of Comparative Psychology, 102, 108-117.Roberts, W.A., & Dale, R.H.I. (1981). Remembrance of places lasts:Proactive inhibition and patterns of choice in rat spatial memory.Learning and Motivation, 12, 261-281.167Spatial Memory in PigeonsRoberts, W. A., & Van Veidhuizen, N. (1985). Spatial memory in pigeonson the radial-arm maze. Journal of Experimental Psychology:Animal Behavior Processes, 11, 241-260.Roitbiat, H.L. (1980). Codes and coding processes in pigeon short-termmemory. Animal Learning & Behavior, 8, 341-351.Roitbiat, H.L. (1984a). Representation in pigeon working memory. InH.L. Roitbiat, T.G. Bever, & H.S. Terrace (Eds.), Animal Cognition(pp. 79-98), Hillsdale, N.J.: Erlbaum.Roitblat, H.L. (l984b). Pigeon working memory: Models for delayedmatching-to-sample performance. In M.L. Commons, A.R. Wagner, &R.J. Herrnstein (Eds.), 79-97. Quantitative analyses of behavior:Discrimination processes, Vol. 4. Cambridge, MA: Ballinger, 161181.Roitbiat, H.L. (1987). Introduction to Comparative Cognition. NewYork, NY: W.H. Freeman and Company.Roitblat, H.L., & Harley, R.E. (1988). Spatial delayed matching tosample performance by rats: Learning, memory and proactiveinterference. Journal of Experimental Psychology: Animal BehaviorProcesses, 14, 71-82.Roitblat, H.L., & Scopatz, R.A. (1983). Sequential effects in delayedmatching-to-sample performance. Journal of ExperimentalPsychology: Animal Behavior Processes, 9, 202-221.168Spatial Memory in PigeonsRoitbiat, H.L., Tham, W., & Golub, L. (1982). Performance of Betta4splendis in a radial arm maze. Animal Learning & Behavior, 10,108-114.Rozin, P.,, & Schull, J. (1988). The adaptive-evolutionary point of viewin experimental psychology. In R.C. Atkinson, R.C. Herrnstein, G.Lindzey and R.D. Luce (Eds.), Handbook of Experimental Psychology(pp. 503-546). New York: Wiley-Interscience.Seligman, M.E.P: (1970). On the generality of the laws of learning.Psychological Review, 77, 406-418.Sherry, D.F. (1984). What food storing birds remember. CanadianJournal of Psychology, 38, 304-321.Sherry, D.F., & Schacter, D.L. (1987). The evolution of multiple memorysystems. Psychological Review, 94, 1-16.Shettleworth, S.J. (1985). Food storing by birds: Implications forcomparative studies of memory. In N.H. Wienberger, J.L. Mccaughand C. Lynch (Eds.) Memory systems of the brain (pp. 231-250), NewYork: The Guilford Press.Shettleworth, S.J. (1990). Spatial memory in food-storing birds.Philosophical Transaction of the Royal Society Series B, 329, 143-151.Shettleworth, S.J., Krebs, J.R., Healy, S.D., & Thomas, G.M. (1990).Spatial memory of food-storing tits (Parus ater and P.atricapillus): Comparison of storing and non-storing tasks.Journal of Comparative Psychology, 104, 71-81.169Spatial Memory in PigeonsShettleworth, S.J., é Plowright, C.M.S. (1989). Time horizons ofpigeons on a two-armed bandit. Animal Behaviour, 37, 610-633.Small, W.S. (1901). Experimental study of the mental processes of therat. American Journal of Psychology, 11, 133-165.Smith, J.N.M., & Sweatman, H.P. (1974). Food searching behavior oftitmice in patchy environments. Ecology, 55, 1216-1232.Smith, J.P., Attwood, J.C., & Niedorowski, L. (1982). Delayed choiceresponding by pigeons when the correct response is not predictablefrom the sample stimulus. Journal of the Experimental Analysis ofBehavior, 37, 57-63.Spetch, M. L. (1990). Further studies of pigeons’ spatial workingmemory in the open-field task. Animal Learning & Behavior, 18,332-340.Spetch, M.L., & Edwards, C.A. (1986). Spatial memory in pigeons(Columba livia) in an open-field feeding environment. Journal ofComparative Psychology, 100, 266-278.Spetch, M.L., & Edwards, C.A. (1988). Pigeons’ (Columba livia) use ofglobal and local cues for spatial memory. Animal Behaviour, 36,293-295.Spetch, M.L., & Honig, W.K. (1988). Characteristics of pigeons’ spatialmemory in an open-field task. Animal Learning & Behavior, 16,123-131.Staddon, J.E.R. (1983). Adaptive behavior and learning, Cambridge:Cambridge University Press.170Spatial Memory in PigeonsStoltz, S.B., & Lott, D.F. (1964). Establishment in rats of apersistent response producing a net loss of reinforcement. -Journal of Comparative and Physiological Psychology, 57, 147-149.Suzucki, S., Augerinos, C., & Black, A.H. (1980). Stimulus control ofspatial behavior on the eight-arm radial maze in rats. Learningand Motivation, 11, 1-18.Thorpe, W.H. (1950). A note on detour behaviour with Ammophilapubescens Curt. Behaviour, 2, 257-264.Thorpe, W.H. (1963). Learning and Instinct in Animals. London:Metheun.Tolman, E.C. (1932). Purposive behavior in animals and man. New York:Appleton-Century-Crofts.Tolman, E.C. (1948). Cognitive maps in rats and men. PsychologicalReview, 55, 189-208.Tolman, E.C., & Honzik, C.H. (1930a). Degrees of hunger, reward andnon-reward, and maze-learning in rats. University of CaliforniaPublications in Psychology, 4, 237-256.Tolman, E.G., & Honzik, C.H. (l930b). Introduction and removal ofreward, and maze performance in rats. University of CaliforniaPublications in Psychology, 4, 257-275.Tolman, E.G., Ritchie, B.F., & Kalish, D. (l946a). Studies of spatiallearning. I. Orientation and the short-cut. Journal ofExperimental Psychology, 36, 13-24.171Spatial Memory in PigeonsTolman, E.G., Ritchie, B.F., & Kalish, D. (1946b). Studies of spatialI.learning. II. Place learning vs response learning. Journal ofExperimental Psychology, 36, 221-229.Urcuioli, P.J., & Callender, J. (1989). Attentional enhancement inmatching-to-sample: Facilitation in matching acquisition bysample-discrimination training. Animal Learning & Behavior, 17,361-367.Vander Wall, S.B. (1982). An experimental analysis of cache recovery inClark’s nutcracker. Animal Behaviour, 30, 84-94.Vaughan, W., & Greene, S.L. (1984). Pigeon visual memory capacity.Journal of Experimental Psychology: Animal Behavior Processes, 10,256- 271.Wehner, R. (1981). Spatial vision in arthropods. In H. Autrum (Ed.),Handbook of sensory physiology (pp. 287-616), Berlin: Springer-Verlag.Wilkie, D.M. (1983a). Pigeons’ spatial memory: II. Acquisition ofdelayed matching of key location and transfer to new locations.Journal of the Experimental Analysis of Behavior, 39, 69-76.Wilkie, D.M. (l983b). Pigeons’ spatial memory: III. Effects ofdistractors on delayed matching of key-location. Journal of theExperimental Analysis of Behavior, 40, 143-151.Wilkie, D.M. (1983c). Reinforcement for pecking the sample facilitatespigeons’ delayed matching to sample. Behavior Analysis Letters,3, 311-316.172Spatial Memory in PigeonsWilkie, D.M. (1984). Pigeons’ spatial memory: IV. Effects of intertrialinterval manipulations on delayed matching of key-location.Canadian Journal of Psychology, 38, 178-195.Wilkie, D.M. (1986). Pigeons’ spatial memory: V. Proactive interferencein the delayed matching of key-location paradigm occurs only underlimited conditions. Animal Learning & Behavior, 14, 257-266.Wilkie, D.M. (1989). Evidence that pigeons represent Euclideanproperties of space. Journal of Experimental Psychology: AnimalBehavior Processes, 15, 114-123.Wilkie, D.M., & Kennedy, D.J. (1987). Computer simulation of pigeons’performance on a spatial matching task. Behavioural Processes,14, 105-122.Wilkie, D.M., & Masson, M.E. (1976). Attention in the pigeon: Areevaluation. Journal of the Experimental Analysis of Behavior,26, 207-212.Wilkie, D.M., & Slobin, P. (1983). Gerbils in space: Performance on a17-arm radial maze. Journal of the Experimental Analysis ofBehavior, 40, 301-312.Wilkie, D.M., & Spetch, M.L. (1981). Pigeons’ delayed matching tosample errors are not always due to forgetting. BehavioralAnalysis Letters, 1, 317-323.Wilkie, D.M., Spetch, M.L., & Chew, L. (1981). The ring dove’s shortterni memory capacity for spatial information. Animal Behaviour,29, 639-641.173Spatial Memory in PigeonsWilkie, D.M., & Summers, R.J. (1982). Pigeons spatial memory: Factorsaffecting delayed matching of key location. Journal of theExperimental Analysis of Behavior, 37, 45-56.Wilkie, D.M., & Willson, R.J. (in press). Time-place learning bypigeons, Columba livia. Journal of the Experimental Analysis ofBehavior.Wilkie, D.M., Willson, R.J., & Lee, K. (1990). Further support for the“drift” model of pigeons’ short-term memory for spatial location.Behavioural Processes, 22, 113-120.Wilison, R.J. (1988). Evidence that pigeons are not lost in space:Pigeons perform well at long retention intervals on a modifieddelayed matching of key location task. Unpublished Mastersthesis: The University of British Columbia.Willson, R.J., & Wilkie, D.M. (1991). Discrimination trainingfacilitates pigeons’ performance on one-trial-per-day delayedmatching of key location. Journal of the Experimental Analysis ofBehavior, 55, 201-212.Wright, A.A., Urcuioli, P.J., & Sands, S.F. (1986). Proactiveinterference in animal memory. In D.F. Kendrick, M.E. Ruling, &M.R. Denny (Eds.), Theories of animal memory (pp.101-125).Hilisdale, NJ: Erlbaum.Zach, R., & Falls, J.B. (1976a). Ovenbirds (Aves:Parulidae) huntingbehavior in a patchy environment: an experimental study. CanadianJournal of Zoology, 54, 1863-1879.174Spatial Memory in PigeonsZentall, T.R., Steirn, J.N., & Jackson-Smith, P. (1990). Memorystrategies in pigeons’ performance of a radial-arm maze analogtask. Journal of Experimental Psychology: Animal BehaviorProcesses, 16, 358-371.175"@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "1992-05"@en ; edm:isShownAt "10.14288/1.0098873"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Psychology"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use."@en ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Spatial associative memory in pigeons"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/3009"@en .