UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Being in the right place at the right time : time of day discrimination by pigeons, Columba Livia Saksida, Lisa Marie 1993

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

Item Metadata


831-ubc_1993_fall_saksida_lisa.pdf [ 3.15MB ]
JSON: 831-1.0086292.json
JSON-LD: 831-1.0086292-ld.json
RDF/XML (Pretty): 831-1.0086292-rdf.xml
RDF/JSON: 831-1.0086292-rdf.json
Turtle: 831-1.0086292-turtle.txt
N-Triples: 831-1.0086292-rdf-ntriples.txt
Original Record: 831-1.0086292-source.json
Full Text

Full Text

BEING IN THE RIGHT PLACE AT THE RIGHT TIME: TIME OFDAY DISCRIMINATION BY PIGEONS, COLUMBA LIVIA.byLISA MARIE SAKSIDAB. Sc., The University of Western Ontario, 1991.A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THEREQUIREMENTS FOR THE DEGREE OFMASTER OF ARTSinTHE FACULTY OF GRADUATE STUDIES(Department of Psychology)We accept this thesis as conforming to the required standardTHE UNIVERSITY OF BRITISH COLUMBIAJune 1993© Lisa Marie Saksida, 1993.In 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.(Signature)Department of  ?syclictuDelThe University of British ColumbiaVancouver, CanadaDate ^30 TiALy ) /993DE-6 (2/88)ABSTRACTPigeons were trained to discriminate between four keys, one of which provided food in themornings, another which provided food in the afternoons, and two which never providedfood. Three experiments were performed to determine if pigeons could track foodavailability over a 24-hr period. In Experiment 1, food was available at one place (peckingkey) in the mornings and at a different place in the afternoons. Although the length oftime between sessions was much greater than in previous studies (e.g., Wilkie & Willson,1992), all subjects appeared to demonstrate time-place associative learning . In order torule out the possibility of an alternation strategy, in Experiment 2 only morning or onlyafternoon sessions were given. Subjects maintained well above chance performance,showing that they were not simply alternating between the two rewarded locations. InExperiment 3, the length of time between morning and afternoon sessions was varied. Theresults indicated an increase in errors as the inter-session interval decreased, whichprovides further support for a timing as opposed to an alternation mechanism. Experiment4 was designed to investigate mechanisms underlying the timing behaviour. Lights-ontime was shifted back by 6 hr and no decrease in performance was found during the firstsession following this phase shift. This finding rules out an interval timer and suggeststhat a circadian type of timing mechanism, with a self-sustaining oscillator, mediates time-place learning over a period of 24 hr. Further support for this notion was found inExperiment 5 in which subjects were tested in constant dim light. In that experimentsubjects' continued correct responding provides additional support for a self-sustainingcircadian timer.iiTABLE OF CONTENTSABSTRACT^ iiTABLE OF CONTENTS^ iiiLIST OF FIGURES vACKNOWLEDGEMENTS^ viiINTRODUCTION^ 1TIMING WITH A CIRCADIAN MECHANISM^ 2The Circadian Model^ 6TIMING WITH AN INTERVAL CLOCK^ 10Animal Timing Procedures  11The Internal Clock Model^ 18The Information-Processing Explanation^ 19Properties of the Clock 27The Connectionist Explanation^ 27The Behavioural Explanation. 29SUMMARY OF TIMING SYSTEMS^ 30TIME-PLACE LEARNING^ 33Extension of paradigm 34Underlying mechanisms 35GENERAL METHODS^ 37EXPERIMENT 1: TRAINING 38METHOD^ 38RESULTS AND DISCUSSION^ 39EXPERIMENT 2: AM ONLY / PM ONLY 44METHOD^ 44RESULTS AND DISCUSSION^ 45111ivEXPERIMENT 3: LATE A.M./EARLY P.M^ 48METHOD^ 48RESULTS AND DISCUSSION 48EXPERIMENT 4: PHOTOPERIOD^ 53METHOD^ 53RESULTS AND DISCUSSION 54EXPERIMENT 5: DIM LIGHT^ 61METHOD^ 61RESULTS AND DISCUSSION 62GENERAL DISCUSSION^ 65REFERENCES^ 68LIST OF FIGURESFigure 1.^An information-processing model of timing^ 20Figure 2.^Acquisition of the 24-hr time-place task^ 40Figure 3.^Mean percent of pecks per key during the first and the final block ofbaseline sessions^ 42Figure 4.^A.M. Only / P.M. Only: Percent of pecks per key for both sessionsfollowing the morning-only probes and the sessions following theafternoon-only probes^ 46Figure 5.^Late A.M. / Early P.M.: Performance of all subjects in morning sessionsas compared to inter-session intervals^ 49Figure 6.^Late A.M. / Early P.M.: Performance of all subjects in afternoon sessionsas compared to inter-session interval^ 51Figure 7.^Baseline performance of all subjects after being moved from the colony tothe new room^ 55Figure 8.^Percent of pecks per key for each bird on the day immediately followingthe phase shift^ 57Figure 9.^Percent of pecks per key for all birds after the phase shift was extended fora period of 6 days^ 59viFigure 10.^Percent of pecks per key for all birds under the dim lightcondition^ 63ACKNOWLEDGEMENTSFirst I would like to thank my advisor Don Wilkie for his advice and support overthe past two years. I'd also like to thank the other members of my committee, CathyRankin and Peter Graf, for their time during a summer that has been extremely hectic forboth of them. Finally, I'd like to thank Anita Lee for helping me to figure out the best wayto pick up a pigeon, and Piers Samson, Anna Blasiak, and Anita for their invaluable helpwith testing.viiINTRODUCTIONThe ability to represent the temporal structure of experience is essential in orderfor an animal to base its behaviour on what may be predicted to occur. To behaveoptimally in particular situations, an animal must be able to represent the temporalstructure of events that occur within its daily routine. Gallistel (1990) has distinguishedtwo types of timing by coining the term phase sense to refer to the ability to anticipateevents that recur at a fixed time of the day-night cycle, and interval sense to describe theability to respond to something that comes a fixed amount of time after an event thatoccurs at varying points within the day-night cycle. Church (1984) assumes that thesetypes of timing are mediated by different mechanisms: Phase sense is governed byendogenous oscillators that run constantly, whereas interval sense is controlled bystopwatch-like timers that may be stopped, started, and reset by the occurrence of events.In fact, most short-term timing behaviour (less than 1 hr) has been explained by astopwatch-like mechanism (Church,1978), whereas most long-term timing (24 hr) hasbeen accounted for by circadian rhythms (Aschoff, 1989).Time-place learning can be considered to be a special case of timing behaviourcombining knowledge of time of day with knowledge of spatial location. Manyobservational studies (e.g., Daan, 1981; Daan & Koene, 1981) have suggested thatanimals appear to have the ability to keep track of time of day, that is, to anticipate eventswhich occur on a regular basis, such as restricted feeding. Few researchers, however,have investigated animals' ability to associate a specific place with food availability, letalone the particular timing mechanism that underlies the behaviour.I will begin this thesis by examining the work that has been done in the area oftiming in general; first, timing using circadian rhythms, and second, timing of shorterintervals using an internal clock. Then I will review the research that has been donespecifically with respect to time-place learning. In addition, I will present new workrelating to the mechanisms underlying time-place learning.1TIMING WITH A CIRCADIAN MECHANISM.Most animals and plants display activity episodes which reoccur every 24 hr, evenunder conditions in which all periodically occurring cues are ruled out. In fact, almost allorganisms more advanced than bacteria have continuous biological rhythms (e.g.,activity/rest, body temperature, evoked responses) which occur within a day (Jacklet,1985). These daily rhythms seem to be generated from within the organism and are calledcircadian rhythms.One example of how ubiquitous circadian mechaniams are results from a study inwhich the fungus Neurospora was included in a space shuttle flight (see Sulzman, Ellman,Fuller, Moor-Ede, & Wasser, 1984). The mission involved circumnavigation of the earthwith a periodicity of 1.5 hr for several days. On Earth Neurospora cultures producevegetative spores every 21 to 22 hr. In space, that periodicity was maintained in spite ofthe fact that the solar day was only 1.5 hr in length. This result strongly suggests that anendogenous circadian pacemaker controls circadian rhythms.The central pacemaker which controls the rhythms regulates the mechanism bydelivering a series of pulses in a consistent pattern. A physiological pacemaker has beenlocated in the mammalian suprachiasmatic nuclei (Moore, 1982; Turek, 1983), the avianpineal organ (see Norgren, 1990), in the optic lobes of certain insects, and in the eyes ofsome gastropods (Jacklet, 1985). In fact, multicellular organisms contain not just one, butmany endogenous oscillators with periods ranging from seconds to years, with all types ofintermediate intervals represented (see Aschoff, 1981 and Farner, 1985 for reviewsdemonstrating endogenous oscillators at each of these periods). Gallistel (1990) suggeststhat when a system contains a group of oscillators with widely differing periods it canrepresent times of occurence of events by recording momentary states, or phases, of theoscillatory processes. This provides us with the physiological grounding of a potentialtiming mechanism.23In order to be able to anticipate recurrences at the same phases in later cycles theinternal cycle must be synchronized with external cycles (e.g., day/night cycle, lunar cycle,tidal cycle). No physical oscillation is perfectly periodic and no two physical oscillationshave the same period, so if the external oscillator had no effect on physiologicaloscillations with similar periods then the two oscillations (external and internal) would notbe synchronized. From the phase of the internal oscillator one would not be able topredict the phase of the external oscillator therefore the internal oscillator would be of nouse in predicting regular events in the external world. However, endogenous oscillatorsdo respond to some periodically recurring extrinsic events in such a way as to maintain afixed phase relation between the two processes. This process is called entrainment.The presence of an endogenous self-sustaining oscillatory system allows animals tosynchronize their activity with cycles in the environment, as well as with conspecifics,because these oscillators are entrained by signals from extrinsic oscillators or from otheroscillators within the same organism. Roberts (1965) performed an experiment whichillustrates this phenomenon quite clearly. Using a cockroach as a subject, Roberts made acontinuous recording of its movement and then cut that record into 24-hr segments.During the first 20 days of recording, the cockroach showed an abrupt onset of activityshortly after the lights went out every day, which is a normal occurrence in nocturnalanimals. Recurring bouts of activity happened throughout the night, and activity wasfairly low during the day. At first glance it might appear as though these activities wereexogenous, that is, that darkness leads directly to activity and light suppresses it. In thenext stage of the experiment, Roberts demonstrated that this was not the case. On Day 20he painted over the eyes of the cockroach with nail polish so that it could no longer detectlight. In spite of this, the roach still showed strong daily variations in activity, but theonset of activity was no longer maintained in a fixed phase relation to the light/dark cycle.Instead, daily activity was determined by an endogenous oscillator with a period of 23.5hr. Because this period was shorter than that of the extrinsic oscillator (the light/dark4cycle), and because the extrinsic oscillator could no longer entrain the endogenousoscillator, activity began 30 min earlier each day. After 24 days, the onset of activitycorrelated with lights-on time rather than lights-off. Subjective dusk now corresponded toobjective dawn. At this point it was clear that the onset of activity at dusk was notmerely a response to the lights being extinguished. On Day 50, Roberts then peeled offthe nail polish. During the first eight days after the polish was removed, activity begansome time after the lights were turned off From this, Roberts concluded that onset ofactivity was still being controlled by the endogenous oscillator, which was running behindthe external light/dark cycle. But as soon as the cockroach's sight was returned, theendogenous oscillator again became susceptible to entrainment of the light/dark cycle.From continuous recordings, it could be seen that the onset of activity drift became fasterthan the free run drift in order to "catch up" with the extrinsic oscillations. From theseresults, it was obvious that the transition from light to dark was not eliciting andsuppressing the activity, but when the transition occurred a few hours too late with respectto the endogenous cycle, it transiently accelerated the endogenous oscillation. In thisexperiment, daily phase advances gradually brought the internal clock into the correctphase relation -- it caught up with the extrinsic cycle on Day 58. At this point, theendogenously timed activity onset once again coincided with lights out.Pittendrigh (1980) provided further information about the nature of theoscillations underlying timing behaviour. In this experiment, he used hamsters that weremaintained in constant darkness, and he measured running wheel activity. At thebeginning of the experiment, activity onset seemed to be controlled by an endogenousoscillator running with a period of 23.8 hr. Continuous recordings showed a steady butslow drift of activity onset. When the free running period had been established,Pittendrigh inserted timing signals consisting of 15-min light pulses at different phases ofthe endogenous cycle. The first pulse was given early in the nightly activity bout, that is,early in the subjective night, at the phase of the endogenous oscillation that corresponded5with the dark phase of the extrinsic cycle under natural (entrained) conditions. Pittendrighfound that if the pulse was delivered early in the subjective night, it created a phase delayof one-nineteenth of a cycle, that is, during the next 24-hr period activity occurred 1 hr 15min later than it would have if the pulse hadn't been given. In the days following thephase shift caused by the pulse, the endogenous oscillator continued to drift at the samerate as before the shift, the only difference being that the phase was altered. A secondlight pulse given later in the subjective night led to a similar delay, but a third pulse, whichwas given even later, led instead to a small phase advance. The onset of the next activitybout occured sooner that it would have with no signal. Finally, pulses given farther intothe subjective night caused a large phase advance. From these results it can be seen thatwhen the timing signal comes at certain phases of the cycle, it slows down; when it comesduring other phases, it speeds up. The responsiveness of the internal oscillator allows it tostay in a fixed phase relation with the external day/night cycle.In sum, endogenous oscillators can be influenced by extrinsic inputs in two ways.First, transient extrinsic inputs such as timing signals can shift the phase of the oscillator toa different part of the cycle. Second, tonic (slowly changing) extrinsic inputs maymodulate the period of the endogenous oscillator. For example, the free-running period ofmost circadian oscillators can be lengthened by a steady level of environmentalillumunation. The periodic factors which lead to entrainment are termed zeitgebers. Themost powerful zeitgeber seems to be the daily transition between light and dark (Aschoff,1989). Weaker zeitgebers, however, such as temperature (Ostheim, 1992), social cues(Gwinner, 1966; Marimuthu, Subbaraj, & Chandrashekaran, 1981), and feeding time (Abe& Sugimoto, 1987; Mistlberger & Rechtschaffen, 1984), have also been demonstrated.The Circadian ModelThe major component of the avian circadian system is the pineal gland, whichfunctions as the main oscillator in the system (Menaker & Zimmerman, 1976). Evidencefor another population of oscillators located outside the pineal in the suprachiasmaticnuclei (SCN) has also been found (Takahashi & Menaker, 1979). The basic model that isused to describe these physiological findings was first presented by Gwinner (1978) andsuggests that the pineal contains a self-sustaining oscillator that drives a population ofweakly coupled, self-sustaining oscillators that in turn control locomotor activity. Theavian circadian system seems to be set up in a manner somewhat opposite to themammalian system. In the rat the principal oscillator is considered to be the SCN.However, the fact that some circadian rhythms persist after the SCN are ablated (e.g.,food-anticipatory activity and temperature) has led some researchers to conclude that theSCN may be only one of several pacemakers in a multiple oscillator system, hierarchicallyorganized with the SCN in the dominant position (Rusak, 1977).The timing signal that holds the oscillator in a fixed phase relation to anotheroscillator does not have to have the same period as the oscillator that it entrainsEndogenous oscillators with different periods may maintain a fixed phase relation throughan exchange of timing signals. For example, many rodents have ultradian (shorter thandaily) activity rhythms with a period of 1.5 - 3 hr. In voles the rhythm phase has a fixedrelation to the circadian rhythm which is maintained even when the circadian oscillation isfree-running (see Gallistel, 1990). It has not been conclusively shown that ultradianrhythms depend on distinct endogenous oscillators. It is possible that each activity bout istriggered when the circadian oscillation reaches a specific phase in its' cycle. However,some evidence has been collected for separate oscillators ( Daan & Aschoff, 1981). Insome rodents ultradian rhythms are not phase locked to the circadian rhythm (Cowcroft,671954). If shorter rhythms were produced by trigger points spaced 2.4 hr apart in acircadian oscillation, the rhythym would always be phase locked to that circadian rhythm.If the rhythms were produced by separate oscillators, however, it may or may not be phaselocked, depending on whether or not it received an effective timing signal from thecircadian oscillator. Even in voles, where the shorter rhythm is obviously phase locked tothe longer one, sometimes a drift is seen in the relation between the ultradian and circadiancycles (Aschoff, 1984). This is most readily explained by the presence of two oscillators.Other evidence for the existence of multiple oscillators is derived from the fact thatthe period of the short-term rhythm is systematically affected by nutritional factors thathave no effect on the period of the circadian oscillation. Factors which increase theamount of food that voles must eat (e.g., lactation or high cellulose in the diet) shorten theperiod of the ultradian oscillation but not the period of the circadian oscillation (Daan &Aschoff, 1981). This can be easily explained if it is assumed that ultradian rhythms aredriven by individual endogenous oscillators. The phenomenon is very difficult to explain,however, by different trigger phases of the same circadian oscillation. Finally, it mightappear that the nutritional effect on ultradian rhythms is being driven by filling andemptying of the stomach or some other metabolic cycle but this has been shown not to bethe case because the rhythm persists even when the vole has no access to food or water(Daan & Slopsema, 1978).The most complete conceptual model of circadian rhythmicity which is based onthe principle that a multi-oscillator system drives periodic behaviour is that of Gallistel(1990). He suggests that the way in which variables underlying any physical oscillationvary with time make it possible to specify any phase (moment) within a period ofoscillation by recording the values of the two variables at that time. If passing time iscontrolled by an oscillatory process, then specific moments can be marked by twointerconnected quantities that describe the state of the oscillatory process at a given time.8This system only works over the period of one oscillation. If the sine and cosinevalues of the oscillation are being used to specify time, then they are unique only for timeintervals that are integer multiples of the period of oscillation. As a result, memory formoments based on the recorded values of components of oscillations does not distinguishbetween corresponding moments in different periods of the longest oscillation contributingto the record. This may not matter to some animals. The honeybee, for example, has alife span of 3 weeks in the summer and thus may not need to distinguish between days.Longer lived animals, however, do have the ability to distinguish days. Birds canrecognize something that happens every second day (Caubsiens & Edwards, unpublishedmanuscript; see Gallistel, 1990). Also, endogenous oscillators with a period of a yearhave been shown (Gwinner, 1981; Farner, 1985). This suggests that the animal's temporalrecord must incorporate coordinates that come from a slower oscillation whose periodencompasses several cycles of the faster oscillation.Based on the above theory, only one oscillatory process with a long enough periodis necessary to record all potential time coordinates. In reality , however, it is not likelythat the mechanism which reads time records will be able to distinguish reliably betweensimilar recorded quantities. In other words, the system will not be able to distinguishmoments separated by seconds if it records moments by storing circannual oscillatoryvalues. Thus a system based on several different oscillators with different orders ofmagnitude is necessary in order to span a large interval but also record moments in therealm of seconds.When entrained to 24 hr, circadian rhythms can potentially be used as a clock thatis essential for many animal behaviours such as sychronization of activities amongindividuals, adjustment of activities to environmental conditions, and the measurement oftemporal intervals (Aschoff, 1989). One type of behaviour whose dependence oncircadian rhythms is well documented is the phenomenon of food-anticipatory activity(FAA), which is apparent in many animals when they are put on a restricted feeding9schedule. When animals are limited to one meal at a fixed time of day, increased arousalduring the hours before feeding is usually observed. This has been detected in manyparadigms, including wheel-running (Bolles & deLorge, 1962; Richter, 1922),unreinforced lever pressing (Boulos & Terman, 1980), activity directed toward an emptyfeeding dish (Birch, Burnstein, & Russell, 1958; Mistlberger & Rusak, 1988), andgeneralized cage activity (Mistlberger & Rechtschaffen, 1984). Explanations based onexternal cues triggering FAA have been ruled out because rats and other speciesdemonstrate the same behaviour in artificial environments in which there is no variation inlight, temperature, or sound that could be a signal for immediate food access.The main piece of evidence for circadian rhythms being involved in the control ofFAA comes from a study which showed that rats fail to anticipate daily meals that areseparated by 19 hr or 29 hr (Bolles & deLorge, 1962). This demonstrates that whenfeeding schedules differ greatly from 24 hr anticipatory activity does not occur, andsuggests that FAA is not based on interval timers that can be reset to any arbitraryinterval. It also coordinates with the idea of entrainment, which allows only for minormodifications of the timing mechanism by external cues. Physical oscillators are limited asfar as the maximum phase shift that can occur in one cycle is concerned.A second bit of evidence involves food deprivation. FAA can last for a week whenpreviously restricted rats are fed ad libitum (Edmonds & Adler, 1977; Aschoff, von Goetz,& Honma, 1983; Honma, von Goetz, & Aschoff, 1983; Gibbs, 1979; Stephan, 1983)however, usually it disappears by the second day after the switch (Gibbs, 1979; Coleman,Harper, Clarke, & Armstrong, 1982; Rosenwasser, Pelchat, & Adler, 1984; Stephan,Swann, & Sisk, 1979). On the other hand, if a rat demonstrating FAA is food deprivedfor at least 2 days FAA persists (Bolles & Moot, 1973; Stephan et al., 1979; Clark &Coleman, 1986; Ottenweller, Tapp, & Natelson, 1990; Mistlberger, 1992b). If 3-7 dayblocks of food deprivation are repeated after one or more weeks of ad libitum feeding,FAA reappears at the usual time of day, even 50 days after the last feeding schedule10(Bolles & Moot, 1973; Stephan et al., 1979; Coleman et al., 1982; Clarke & Coleman,1986; Ottenweller et al., 1990). These observations are incompatible with an interval typeof timer, but can work with an entrainable oscillator model if the coupling between thefood-entrained oscillator and behaviour is gated by the animal's motivational state.A third example of how FAA fits in with a food-entrainable oscillator modelinvolves succeeding activity. Some rats who demonstrate FAA on 24 hr restricted feedingshow a second bout of activity about 3-6 hr after mealtime (Aschoff et al., 1983; Honmaet al., 1983; Stephan et al., 1979). This succeeding activity can be explained in theoscillator model as the trailing end of the active phase of the food-entrained oscillator.This is further supported by the observation that durations of FAA and succeeding activityare negatively correlated across days and feeding cycles of different periodicities so thatthe combined daily duration is preserved (Aschoff et al., 1983; Honma et al., 1983).This area has been extensively researched and I have mentioned only a few of thestudies that have been done. Although the exact mechanisms underlying FAA are still amatter of debate (see Mistlberger, in preparation, for a full review), it provides a goodexample of how one behaviour may be controlled by a circadian mechanism.TIMING WITH AN INTERVAL CLOCKMany instances of temporal discrimination of short intervals have beendemonstrated in animals. Stubbs (1968) showed that pigeons can be trained to make oneresponse if a particular stimulus is longer than a criterion, and another response if thestimulus is shorter than the criterion. When the results were plotted, it could be seen thatthe percentage of responses of one kind varied as a function of stimulus duration. Inaddition to being sensitive to duration of the stimulus, animals are also sensitive toduration of the response. In 1973, Platt, Kuch, and Bitgood demonstrated that responseduration can be altered by differential reinforcement. When a rat receives food if it pushes11a lever longer than a criterion duration, the median response duration increases as a powerfunction of the duration of the criterion. Other examples of timing of short intervalsinclude: If reinforcement is available only if the animal witholds responding longer than acriterion time, the distribution of inter-response times (IRTs) is closely related to theminimum reinforced IRT (Kramer & Rifling, 1970), an aversive event presented at regularintervals produces temporal conditioning (LaBarbera & Church, 1974), and when food ispresented a fixed amount of time after signal onset, the animal's response rate increases asthe time to the next reinforcement approaches (Pavlov, 1927). The results of these, andmany other studies (see Anger, 1963; Gibbon, 1972; Libby & Church, 1974; Dews, 1970)suggest that animals can learn to adjust their behaviour to a specific time interval. Fromthese basic results, it became quite clear that some sort of internal clock was regulatinganimal timing behaviour. This term began as a metaphor, but as research moved forwardcharacteristics of the clock were discovered, and now the concept is no longer thought ofas a metaphor, but as something that truly exists (Church, 1978). A great deal of thework that has been done in this area is the contribution of Church, Gibbon, Meck, andRoberts. They have collaborated on a series of experiments using various proceduresgeared toward developing a model of how elapsed intervals are estimated, remembered,and compared.Animal Timing ProceduresThe peak procedure. This was originally described by Catania (1970), but has morerecently been used by Church, Miller, Gibbon, and Meck (1988). During a peakprocedure session, a rat is placed in a lever box and exposed to two trial types; food andnonfood. During a food (training) trial, the rat is exposed to a white noise signal. Thefirst lever response after 20 s is followed by food, and subsequently the end of the signal.The critical trials are the non-food trials, during which the rat receives the white noise12signal for 120 s but never receives food. The time of each response beginning fromstimulus onset is recorded throughout the session, and this gives an indication of when therat is expecting food, based on the training trials.The data in this paradigm are obtained from the nonrewarded trials. During thesetrials, the mean probability of response increases to a maximum near the time that food issometimes received, and then decreases in a fairly symmetrical fashion. The peak inlatency of responding usually differs from the rewarded latency by a multiplicative factor,the value of which varies between animals but is constant within an individual. Thisfinding led to the development of scalar timing theory (Gibbon, 1977; also see Gibbon,1991) which states that the remembered duration of an elapsed interval consists of theexperienced duration multiplied by a scalar factor, which varies between animals. Asecond observation is that performance on individual trials, unlike the mean functions, ischaracterized by an abrupt change from a state of low responding to one of highresponding and finally to another state of low responding . This is typically known as the"break-run-break" pattern of response on a single trial (Schneider, 1969). Finally, themean functions of different times of reinforcement (e.g., if time of reward is changed from20 s to 12 s after noise onset) are very similar when time is shown as a proportion of timeof reinforcement. This highlights the fact that the animal's rate of responding isdetermined by the ratio between the remembered reward latency and its measure of theinterval so far elapsed in the current trial. In sum, what determines the rat's rate ofresponding is its memory of the elapsed interval before the reward, not the interval thatwas actually experienced (Church & Broadbent, 1990).Generalization. In a generalization experiment (e.g., Church & Gibbon, 1982), a stimulusis turned on for a variable duration which is to be judged by the animal. If the stimulus ison for the correct (rewarded) duration, a response leads to food. Otherwise, no reward isobtained. Results from generalization experiments show that the probablilty of response13varies as a function of the duration of the signal, that is, there is a peak in response at therewarded duration. This paradigm provides additional support for the scalar properties oftiming because the probability of response is determined by the ratio between the signalduration and the remembered duration of rewarded signals .Bisection. During a bisection experiment (see Church & Deluty, 1976; 1977) the animalhears or sees a signal of variable duration, as in the previous generalization procedure, buttwo levers are inserted into the box at the end of the signal. Pressing on one of the leversis rewarded if the signal was the shortest in the range of signals used, pressing on the otherlever is rewarded if the signal was the longest, and there is no reward if the signal was ofintermediate duration. The data consist of the probability of responding on the "long"lever as a function of the duration of the preresponse signal. Experiments using thisparadigm have shown that responding is based on duration ratios, as opposed todifferences in the subjective temporal quantities, because the point of indifference(intermediate signal duration at which the long lever is chosen on half of the trials) is thelogarithmic midpoint between the shortest and the longest intervals.Estimation. A typical estimation procedure involves a delayed symbolic matching-to-sample (DSMTS) task (e.g., Spetch, 1987; Spetch & Rusak, 1989; Spetch & Wilkie,1983). In this task, pigeons are reinforced for pecking at one stimulus (e.g., a red key)after a short (e.g., 2 s) sample presentation and at another stimulus (e.g., a green key)after a longer presentation of the same stimulus (e.g., 10 s). Retention of the length of thesample is tested by varying the delay between the beginning of the sample and thepresentation of the choice stimuli. Results from this procedure have shown that pigeonsdo store duration information analogically, as opposed to simply a categorical coding oflong or short. Spetch and Sinha (1989) concluded that pigeons retain an analogical, andnot a categorical, memory of the temporal properties of the sample in a DSMTS task.Pigeons do seem to use different temporal coding strategies, however, depending on thetask. In a successive matching-to-sample task, there are 2 different test stimuli but onlyone is presented following the sample on each trial. Pecks to one of the test stimuli arereinforced for a short duration, whereas pecks to the other are only reinforced for a longduration. Grant and Spetch (1981) used this type of task and found that pigeonsemployed a prospective, categorical strategy.Subdivision. Another question regarding timing is whether animals deal with the intervalto be timed as a whole, or do they subdivide it into parts? Rhythmic behaviours during atiming session are often observed (Church, 1978) and there is potential that the behaviourcould be some sort of oscillator used to drive the internal clock. Deluty and Church(1976; see Church, 1978) performed an experiment to discover if this was indeed the case.A loud white noise was pulsed during the light signal to be timed. It was assumed that theanimal would have trouble maintaining a rhythm different from the pulse rate. Using anestimation procedure, rats were trained to press the left lever when the light stimulus had aduration of 2 s, and the right lever if the light duration was longer (2.5, 3.0, 3.5, 4.0 s).Half of the stimuli were delivered at the standard 2 s duration, and the other half werepresented at one of the comparison durations. During different phases of the experiment,the noise pulses were given either at regular 0.5 s intervals, random 0.5 s intervals, or nonoise was presented at all. The results provided no evidence that the pulses increased, orthat the random noise interfered with, the accuracy of timing. In other words, it seems asthough rats do not subdivide when timing an interval.Time-left. In this paradigm, the animal has to judge from moment to moment which oftwo response options will pay off in a shorter amount of time (see Gibbon & Church,1981). At the start of a trial one option will pay off sooner than the other one but at somepoint this reverses. In the pigeon version of the task, a trial begins with the illumination of1415two response keys (red and white). Continuous responding on the red key leads to areward at time T + 30 s. This 30 s interval is known as the standard interval (S). T israndomly determined from trial to trial by choosing with equal likelihood from 6alternatives ranging from 1/6 of S (5 s) to 11/6 of S (55 s). If the bird only pecks at thered key, it realizes that T for a given trial has timed out when one peck turns the keygreen. At that point, the bird loses the option of responding on the white key (it goesblank) so it must finish the trial on the red key and it gets a reward 30 s later. The whitekey, on the other hand, provides a reward at 60 - T s where 60 s is the comparison interval(C). If the bird only pecks the white key then it realizes that T is timed out when one peckextinguishes the red key and thus white becomes the only option. The bird must finish thetrial on this key, and wait 60 - T s before it is rewarded. Whether the bird must finish onthe red or the white key is determined by which key it pecks first after T has timed out.The best approach to take in this situation would be to peck the red key for thefirst 30 s because if T times out while the bird is pecking red it will receive a reward after30 s. If the bird was pecking white, the reward would be obtained after 60 - T s, which isa longer wait than 30 s as long as the time elapsed (T) is less than 30 s. When more than30 s have elapsed, however, the best approach would be to switch to the white keybecause the reward would occur in a shorter amount of time. Basically, in order toperform well on this task, the animal must compute the difference between theremembered duration which corresponds to C and the time elapsed since trial onset. Bothpigeons and rats are able to do this. The probability of responding to white starts low andrises during the trial whereas the probablilty of responding to red starts high and ends low.The point of indifference (latency at which the probability of responding white is equal toresponding red) is a linear function of the difference between C and S.Timing with interruptions. Church and Roberts, in a series of two studies, looked at theconcept of stopping the timing mechanism (Church & Roberts, 1975; Roberts & Church,161978). In the first experiment (Church & Roberts, 1975) rats were trained to press a leveron a discrete 1-min fixed interval schedule. The lever was inserted into the box, and foodbecame available with the first press after 1 min had passed. By the end of training, asteep temporal gradient was developed, that is, the response rate increased as the time ofthe next reinforcement approached. In the critical part of the study, the fixed intervalswere periodically interrupted by a 15 s break, during which the lever was withdrawn andwhite noise was played. The rationale was that if they continued timing during the break,the clock setting would increase during that period of time, whereas if the clock stoppedduring the break, then the clock setting would remain the same. The rats were dividedinto two groups -- a run group, for whom time ran during the break (if there were 40 sbefore food priming when the break began, there would be 25s to the next priming whenthe break ended), and a stop group (if 40 s were left before food priming when the breakbegan, there would be 40 s to the next priming when the break ended). The final resultsdemonstrated that the two groups were indistinguishable before the break, but differedafterward. Initially, all subjects stopped their clocks during the break. Eventually,however, subjects in the run group learned to run their clocks during the interval. Thisprovides strong evidence for flexibility of timing, depending on the situation.Absolute timing units. Two studies by Roberts and Church (1978) suggested that timingunits are absolute and that the internal clock times "up". In this study, rats were trainedon a discrete 30 s FI schedule and a 60 s FI schedule, with a light signalling the shorterinterval and a noise indicating the longer interval. In the critical part of the test, the ratscontinued to receive the same types of trials. In addition, however, on one-third of thetrials the 30 s light signal was shifted to the 60 s noise signal. The shift could occur 6, 12,18, 24, or 30 s after trial onset. On shifted trials, food was primed at a time appropriatefor an absolute up timer for half of the rats (absolute group), and for a proportional uptimer for the remaining rats (proportional group). The rationale was that if rats used the17same internal clock to time the light and the noise, whether they were timing in absolute orproportional units, and whether they were timing up or timing down could be inferredfrom the response rate during the trials which were shifted to the 60 s noise signal. As anexample, consider the case in which, after 12 s of light (30 s signal) the light would beextinguished and replaced by noise (60 s signal). There are four possible responses, andfour corresponding timing mechanisms, that could occur with this type of trial. If the ratwas using absolute up timing, 12 s have passed so performance would be equivalent tothat of a rat on plain FI 60 after 12 s. If an absolute down timer was being used, 18 swould be left on the timer therefore performance would be similar to that of a rat on FI 60after 42 s. A third possible mechanism would be a proportional up timer, and in this case40% of the total time has passed, so the rat would perform as though 24 s (40%) of theFI60 schedule had passed. Finally, using a proportional down timer, 60% of the total timewould be left, consequently the rat would behave as thought 36 s were left. The questionhere is: Does the performance of a rat switched from a FI 30 schedule to a FI 60 scheduleafter 12 s correspond to the behaviour of a rat on FI 60 after 12, 24, or 42 s?The results of their study led them to the conclusion that the response ratio in the60 s signal was a function of the time since the trial began. It did not matter if the trialbegan with a 60 s signal, or if it was shifted after 6, 12, 18, or 24 s. To predict responseratio in the 60 s signal, they needed to know only the time since the trial began; it did notmatter how much of the total time was comprised of either the 30 s or the 60 s signal.Roberts and Church (1978) concluded four things about the nature of the internalclock from this study. First, it provides evidence for a single clock. If separate clockswere used to keep track of the 30 s visual signal and the 60 s auditory signal, then thelength of time that the animal spent exposed to the visual signal prior to the shift shouldhave no effect on behaviour. If one clock was used, then the response ratio in the 60 ssignal should be related to the time spent in the 30 s signal, and it was. This conclusionhas since been modified (see Meck & Church, 1984). Second, the clock advances as a18function of absolute, not proportional, time. In this case, the response ratio was a functionof the duration of the total signal. Since it did not matter what proportion of thecombined signal was the 30s signal, this study provides no evidence for differential ratesof timing. Third, the clock times up. The curves after a shift and the curves of trials inwhich a shift had not occurred were very similar. A rat shifted from FI 30 to FI 60 after12 s has a response ratio similar to that of a rat having had an FI 60 schedule for 12 s. Ifthe clock was timing down as a function of time, the clock would not be explained by asimple shift in criterion. Finally, when the experiment was extended for 40 3hr sessions,the rats in the absolute timing group continued to act in the same manner, whereas the ratsin the proportional group shifted their behaviour. By the end of the experiment, theylearned to deal with time in proportional units. This provides evidence for flexibility oftiming behaviour, which will be dicussed in more detail later.The Internal Clock Model The above studies have led to the more rigorous development of an internal clockmodel. An internal clock advances as a function of time from or during a well-definedevent, and the animal adopts a temporal criterion and response rate which relates theprobablility of a particular response to the clock setting and the criterion (Church, 1978).In an estimation procedure (e.g., generalization or estimation), the signal starts the clock.When the signal terminates, the pigeon reads the value of the clock and makes a decisionto respond either on the short or the long lever based on the relationship between theclock setting and the criterion. If the clock setting is less than the criterion, then thepigeon chooses short, whereas if the clock setting is longer than the criterion, the pigeonchooses long (Spetch & Sinha, 1989). In a production procedure (e.g., peak procedure),the signal starts the clock and the animal continually makes decisions about whether or notto respond, again based on the clock setting and the criterion (Church et al., 1988).The Information-Processing ExplanationThe idea of treating humans as "information-processing machines" began withBroadbent (1958). He suggested that the machine consists of modules and theirinterconnections, and that the task of the researcher is to reverse-engineer the system.Based on observations of the input and the resulting output, inferences can be made abouthow the machine works. His basic model was made up of a selective filter, limited channeland a long-term store where the parts were connected serially such that informationmoved from the senses into central processing and eventually to motor output.The information processing idea has been extended to animal cognition andtraditionally has been used to account for the various properties of the internal clock.Gibbon, Church, and Meck (1984) present a theory in which the basic parts of the modelinclude a pacemaker, a switch, an accumulator, working memory, reference memory, anda comparator (see Fig. 1).Pacemaker. The pacemaker is an internal mechanism that generates pulses. The meanrate of the pacemaker seems to be able to be controlled by drug, dietary, or environmentalmanipulations. For example, methamphetamine has been shown to produce a leftwardshift in the psychophysical function that relates the probablility of a "long" response tosignal duration (Maricq, Roberts, & Church, 1981; Meck, 1983). In other words, in thispsychological model methamphetamine increases the speed of the pacemaker.Haloperidol, on the other hand, causes a rightward shift in the same function (Maricq &Church, 1983) and thus seems to decrease pacemaker speed. These findings suggest thatthe pacemaker rate varies with the effective level of dopamine in the system, becausedrugs that increase the release or decrease reuptake of dopamine cause a leftward shift in19Figure 1: An information-processing model of timing.20SwitchPacemaker2122the function, whereas drugs that inhibit dopamine lead to a rightward shift in the samefunction.Diet also has been shown to affect pacemaker speed. A typical rat diet is high incarbohydrates. Prefeeding with sucrose, however, seems to decrease pacemaker speedwhereas prefeeding with protein increases speed (Meck & Church, 1983).Finally, environmental stimuli can affect pacemaker speed. Wilkie (1987) showedthat the perceived duration of a dim light was shorter than that of a bright light of equallength. This suggests that stimulus intensity can affect pacemaker speed. Also, Meck(1983) showed that stress resulting from footshock tended to increase the speed of thepacemaker.It is interesting that the pacemaker can be influenced by so many variables becausein order to discriminate small time differences the pacemaker should be relatively rapid andstable. Gibbon and his colleagues (1984) account for this fact by suggesting that themeasurement of the pacemaker and temporal judgement should be affected by the samemanipulations by an equivalent amount, and moment-to-moment variation in themeasurement of the pacemaker should correlate with moment-to-moment variation intemporal judgement.Switch. The function of the switch is to gate pulses from the pacemaker to theaccumulator. Its main property is that it can operate in various modes. For example,consider a stimulus that is on for a duration a, off for b, and on for c. In this situation, theswitch can gate pulses in different ways, depending on the mode. If the system is in "run"mode, the duration of pulses is a+b+c (the interval is timed from signal onset to the endof the trial). If the system is in "stop" mode, however, the duration of pulses is only a+c(duration is timed only when the stimulus is on). As mentioned previously, rats initiallyadopt stop mode (Church & Roberts, 1975; Roberts & Church, 1978). In this case, thepacemaker puts pulses into the accumulator during the first stimulus segment, it pauses23during the gap, and then adds pulses again after the gap. This has been shown in both thebisection procedure (Church, 1978; Roberts & Church, 1978) and the peak procedure(Roberts, 1981; Meck, Church, & Olton, 1984).The switch can also operate in "event" mode, where the switch is closed for someshort, fixed period of time after each stimulus onset. In this case, the animal uses itsinternal clock as a counter (Meck & Church, 1983; Church & Meck, 1982). Even whentemporal cues are controlled, the animal can classify a stimulus sequence by number ofelements (Meck & Church, 1983; Fernandes & Church, 1982; Davis & Memmot, 1982).In sum, when a signal is composed of several stimuli, the switch can close at theonset of a stimulus and open at either the beginning of each stimulus (event mode), theend of each stimulus (stop mode), or the end of the trial (run mode). With differentialreinforcement, the animal seems to be able to use any mode and the switch mode can betransferred from one modality to another (Church & Meck, 1984; Meck & Church, 1982;Gibbon & Church, 1981; Roberts, 1982).Accumulator. The purpose of the accumulator is to hold the sum of the pulses: It is an"up" counter that increments in arithmetic units in an absolute manner. Evidence from thetime-left procedure suggests that when the time left in an elapsing interval is equal to thestandard interval, that animals are indifferent in choosing between the intervals (Gibbon &Church, 1981).Working Memory. Working memory, as defined for this particular model, storesinformation about the current trial in the absence of the signal. One property of workingmemory is that it can be reset quickly. For example, in the bisection procedure rats aretrained to make a right response after a 2 s signal and a left response after an 8 s signal.On test trials, a preset signal is presented after which comes an interval with no chance torespond, which in turn is followed by another signal duration. The animal can learn to use24the second signal without interference from the first signal (Church, 1980). Workingmemory is used only if there is an interval between the end of the stimulus and the chanceto respond; otherwise the sensory store (accumulator) is used. When there is a retentioninterval between time of exposure and opportunity to respond, there is a decrement inperformance that can be attributed to decay in working memory (Church, 1980; Kraemer,1985). Church (1980) suggests that this decay is on a dimension orthogonal to time. Thisview has been challenged, however, by Spetch and Wilkie (1983) who provide evidencefor an analogical encoding of the temporal properties of a sample. They further suggestthat subjective shortening in working memory may be due to a gradual resetting of theinternal clock, which argues against Church's (1980) suggestion that working memory canbe reset quickly. Further evidence against a quickly resetable working memorycomponent of the internal clock is provided by Wilkie (1988). This study found proactiveeffects (produced by intertrial manipulations) in pigeons' timing behaviour. In otherwords, new information presented between trials was carried over to a succeeding trialwhere it then influenced classification of event duration. This implies that workingmemory is not cleared at the end of each trial, and therefore that a decay type of process ismore likely to control the emptying of working memoryReference Memory. Reference memory permanently stores information about previoustrials and their consequences. Various drugs and lesions can affect this memory constant.For example, physostigmine produces a leftward shift and atropine produces a rightwardshift (Meck, 1983). Thus it seems that the level of acetylcholine affects the memoryconstant. In addition, lesions of the frontal cortex produce a rightward shift in thefunction or, in other words, increase the remembered time of reinforcement (Maricq,1978). Diet also contributes -- choline administered prior to a testing session leads to aleftward shift in the response function (Meck & Church, 1983).25Comparator. The comparator determines the response on the basis of a decision rulewhich involves a comparison between a value in the accumulator or working memory witha value from reference memory. As a result, it must contain two time values and aresponse rule.Simultaneous Temporal Processing. One additional property of this timing system thatshould be mentioned is the fact that animals must have more than a single timing processeven though initially it appeared as though there was only one clock because the sameclock could be used for different modalities (Meck, Church, & Olton, 1984; Church &Meck, 1982; Meck & Church, 1983; Church & Meck, 1981; Meck & Church, 1982a;Meck & Church, 1982b; Roberts, 1982). More recently, however, rats have been shownto time two signals simultaneously without inteference. Meck & Church (1984) put ratson a 50-s peak procedure signalled by light, with a 1-s noise signal presented every 10 s(except at the 50 s point). The resulting performance of subjects fit into a scalar model inwhich the same model was used for the 50-s and the 10-s timing process and the sameparameter values were used for the 50-s function whether or not the 10-s signal waspresent. Rats have also been shown to time and count simultaneously without interference(Meck & Church, 1983). These results suggest that animals must have many switch-accumulator parts to handle all of these tasks.In sum, there are several components making up the psychological process usedfor timing intervals, each with various properties.(1) Pacemaker: Mean rate can be manipulated by drugs, diet, or stress.(2) Switch: Has a latency to operate and can work in several modes: Run, stop, andreset.(3) Accumulator: Times up, in absolute aritmetic units.(4) Working Memory: Can be reset by events or after lesions in the fimbria fornix whenthere is a gap in the signal.(5) Reference Memory: Transfer from accumulator to reference memory is done with amultiplicative constant which is affected by drugs, lesions, and individual differences.(6) Comparator: Uses the ratio between the value in the accumulator (or workingmemory) and reference memory.The number of pulses in the accumulator is equivalent to the rate of pulses persecond multiplied by the number of seconds that the switch from pacemaker toaccumulator has been closed. The number of pulses in reference memory on a single trialequals the number of pulses in the accumulator at the time of reinforcement times amemory constant. Finally, reference memory consists of distributions of such values.Decisions are based on the similarity of the value in the accumulator to a random sampleof a single value from reference memory. On some trials, the animal may attend to thestimulus and this information processing model will apply, whereas on other trials theanimal may not attend to the stimulus and will respond at some fixed level (Church &Gibbon, 1982; Heineman, Avin, Sullivan, & Chase, 1969). This model allows you toidentify the effects of individual variables. For example, the operation of the switch can bechanged by differential reinforcement; during a gap in the signal it can be made to stayclosed or remain open (Church, 1978). This accounts for the "run" and the "stop" mode,where the value in the accumulator either continues to increase or stops during theinterval. Finally, this internal clock model can operate in either event mode, where arelatively fixed value is added to the accumulator when a signal of variable duration isbegun (Meck & Church, 1983), or in reset mode, where the value in the accumulator isinitialized to zero at signal onset (Church, 1978).26Properties of the ClockSundials, stopwatches, digital watches and computer clocks have many differentproperties, but at the same time they have only one commonality -- they change with timein a regular way. Clocks may operate at different rates, some are more accurate thanothers, some are continuous and some are discrete, some are cyclical and some onlychange in one direction, some are driven by a specific external event, and some depend onan internal mechanism, and finally, some time up and some time down.Results from the studies discussed above suggest that the properties of the internalclock correspond closely to those of a stopwatch. Some of the properties of a stopwatchare:(1) Initializable.(2) Can start timing at a constant rate.(3) If stopped, can hold a constant value indefinitely.(4) Can be restarted from the constant value or(5) Can be restarted from the initial value.(6) Can be read while timing or when stopped.Most often, this is the analogy that is used to understand the properties of theinternal clock.The Connectionist ExplanationRecently it has been suggested by Church and Broadbent (1990) that there areseveral problems with the information-processing model that may be better explained witha connectionist interpretation. They discuss three main problems, all based on the fact thatthe model invokes some cognitive abilities that are difficult to interpret with knownbiological mechanisms. First, there is no known biological correlate for the accumulation2728of time, retention of a distribution of values, or random sampling from memory. Thesecond problem has to do with capacity. Storing information learned on successive trialsrequires storing an increasingly large number of values. As a result, the hardware musteither be replicated, or previous values must be forgotten as new ones are added. Third,in order to remember different times, multiple distributions of values must be maintainedand kept separate. However, animals do not seem to have any trouble handling manytrials or differential times of reinforcement, nor do they appear to lose all past training(Meck & Church, 1984).These problems with the information-processing approach have led to thedevelopment of a basic connectionist model of timing (Church & Broadbent, 1990). Theconnectionist model is similar to the information-processing model in that the maincharacteristics of the system are still derived from the relationship between input andoutput. The main differences, however, are that many processes can occursimultaneously, and the representations are not dependent on the state of any particularelement. Advantages of this type of model include the fact that it is characteristic of thenervous system of animals, the same system of elements can contain information aboutdifferent events, and performance is less likely to be disrupted by modification of theinternal components.The structure of the model is very similar to that of the information-processingmodel with three main differences. The pacemaker is expanded into a set of pacemakers,or oscillators, with different periods. The accumulator is replaced by a set of statusindicators, one for each oscillator. These record not just the number of cycles as theaccumulator would, but also the phase of the oscillator. Finally, the sample distributionsthat make up working and reference memory in the information-processing model arereplaced by matrices of connection weights such that any given time is stored throughoutthe matrix, instead of being held as a scalar number. This is an advantage because adistributed memory matrix can store information about an infinite number of samples of a29value whereas a distribution of values must increase in size to accomodate more samples.Mathematically, times are represented as vectors and memories as matrices of connectionstrengths between units of the time vectors.The connectionist variation on the information-processing theme does add to themodel both in terms of biological feasibility and correlation with other theories of timingand time-place learning. For example, the idea of multiple storage pacemakers supportsGallistel's (1990) theory that timing is controlled through the interaction of manyphysiological oscillators. Also, the idea that working memory operates as a matrix ofconnection weights eliminates the problem of infinite capacity. Finally, computersimulations using the connectionist model produce very realistic data. This version of themodel is still new and preliminary, but seems to be a viable path for future research.The Behavioural ExplanationAlthough most of the literature focus on the internal clock model, one majoralternative model has been developed by Killeen and Fetterman (1988). They suggest thatbehaviour is the mediator of temporal control. Animals do not make temporaljudgements, that is, they do not separate stimuli to be timed from other events in theirenvironment and then compare them in some way. Instead, the stimuli within timingexperiments elicit various behaviours, called adjunctive behaviours, and they do so interms of the stimulus's average temporal proximity to reinforcement. When a temporaljudgement is necessary, the animal makes different responses depending on the adjunctivebehaviour they were engaged in at that time. For example, if an animal was pacing when achoice was called for, it would choose the response "short"; if it was pecking the wall itwould choose "long"; and if it was doing neither of those things it would select thealternative that was most often associated with reward in the context of the ongoingbehaviour. They suggest that this is a model of an uncomplicated organism that merely30responds in one way if it is interrupted while doing one thing, and in another way if it isdoing some other thing. Their model does not require a representation in memory becausethe animal is just making simple conditional discriminations. The only essential factor isthat adjunctive behaviours correlated with the response increase in a cumulative mannerover time.According to Killeen and Fetterman, transitions between adjunctive behaviours arecaused by pulses from a biological oscillator within an internal clock system. The counterconsists of the animal's ability to use action states (behaviours such as pacing the side wall)that are correlated with various successive states. They have some observational supportfor this idea. In a study by Church, Getty, and Lerner (1976) they noticed that thestereotyped behaviours of the two subjects that were better at the discrimination weremore rapid than those of the subjects with poorer discrimination. Also, the behaviourduring the signal was the same as the behaviour during the ITI only for the subjects withpoorer discrimination. Time will tell if this will be an effective computation for theinternal clock model.SUMMARY OF TIMING SYSTEMSTwo basic timing systems have been postulated: One based on circadianmechanisms, and the other based on an interval clock model. Usually circadianmechanisms are used to explain timing over the course of a day, whereas interval modelsare used to explain timing of intervals on the order of seconds or minutes. Although theyhave been discussed separately, my intention is not to pit them against each other. It ismost likely, in fact, that these systems work together. Both timers could be part of onemultiple oscillator system, the difference being that they are each working at differentlevels. This multiplicity of timing systems within a single organism may account for theflexibility that seems to be inherent in the timing behaviour of many animals.31The flexibility in timing systems suggests that animals have the capacity to adjustto a changing environment, and predict the likelihood of future events, which is essential inorder to survive. One of the ways in which animals may be predicting these events isconnected to classical conditioning. A theoretical connection between timing and classicalconditioning has been established (Gibbon & Balsam, 1981; Roberts, 1983; Roberts &Holder, 1984a), and in turn classical conditioning has been postulated to be essential forthe optimum fitness of an animal (Hollis, 1982). Initially it was suggested that themechanism used for time discrimination is identical to that used for classical conditioning(Gibbon & Balsam, 1981). Other researchers (Roberts, 1983; Roberts & Holder, 1984a)suggested that the mechanisms of classical conditioning and timing have similar functions;both help the animal to predict future events. Classical conditioning predicts what willhappen, whereas timing predicts when it will happen.The empirical connection between timing and classical conditioning wasestablished by Roberts and Holder (1984b, 1985), when they showed that treatments thatwere meant to change the associative strength (signal value) of a stimulus (a classicalconditioning procedure) also changed the effective duration of the stimulus whenmeasured with a time discrimination procedure. Duration was measured using varioustasks, light or sound was used as the conditioned stimulus (CS) and food was used as theunconditioned stimulus (US). Their work contributed three main findings. First, after CS-alone (extinction) trials, forward-pairing trials increased the effective duration of the CS.Second, after forward-pairing trials, extinction trials decreased the effective duration ofthe CS. Third, when one stimulus (CS+) was followed by food and another (CS-) wasnot, the effective duration of the CS+ was greater than the effective duration of the CS-.These changes in effective duration seemed to be due to changes in timing of the stimulusby the internal clock. This suggests that a decision is made whether or not to time astimulus and that decision depends on the same variables that determine signal value, thatis, the same variables that determine the strength of conditional response.32Further experiments (Holder & Roberts, 1985) provided a more direct measure ofsignal value to determine whether changes in timing and signal value could be explainedwith the same mechanisms. Their study suggested that indeed there is a commonmechanism for signal value and timing because under various treatments the two variableschanged in the same direction, were near zero at the same times, and had similar timecourses. Their proposed common mechanism is esentially a decision-maker which decidesif a stimulus has signal value. The decision about signal value then influences the timingthat is used in the time discrimination. The selectivity of timing in these studies providesexcellent support for the idea that the function of the time discrimination mechanism is tohelp the animal predict when important events will happen. In order to predict when foodwill be available, the animal must be able to time signals for food and at the same time nottime stimuli that are not signals, evidence for which is suggested by this research.Further support for this idea is provided by Hollis (1982, 1984) who establishes abiological function for classical conditioning. Signaled presentations of food, rivals,predators, and mates can lead to anticipatory behaviour (Farris, 1967; Hearst & Jenkins,1971). Hollis (1982) suggests that the function of the conditioned response is to allow theanimal to better deal with the forthcoming unconditioned stimulus. The CR is apreparatory response which Hollis calls prefiguring. These anticipatory CRs function tooptimize the individual's interactions with predators, rivals, mates, and food. Empiricalevidence for this idea is presented in Hollis (1984) where it is demonstrated that thesignaled presence of territory intruders leads to a conditioned response consisting ofanticipatory aggressive behaviour in adult male blue gouramis. In a natural habitat, thisclassical conditioning may give the individual territorial male means with which to increasehis likelihood of successful territorial defense, which ultimately confers a reproductiveadvantage on that individual.TIME-PLACE LEARNINGThe utility of posessing a timing mechanism becomes apparent when one looks atthe behaviour, and survival tactics, of animals in natural settings. Usually, timing is usedin conjunction with spatial memory in order to predict when and where food sources arelikely to be most plentiful. Accordingly, animals' behaviour tends to be influenced jointlyby spatial and temporal control (e.g., Staddon, 1983). The ability of birds to learnassociations between time of day and specific spatial locations of food sources hasgarnered much observational evidence (e.g., Daan, 1981; Daan & Koene, 1981). Severalnaturalistic studies have shown that wild birds are able to adjust their behaviour tochanging patterns of food availability. These studies do not, however, demonstratewhether the birds are using a time-place association to exploit food, or whether they aremerely responding directly to changing food availability.Most experimental work in this area has been done with invertebrates, namelybees. In an early study, Beling (1929) showed that bees can be trained to visit a certainfeeding station near the hive at a specific time of day. On test days, when no food wasprovided, the bees showed peaks of visiting the feeding site in anticipation of and duringthe appropriate time period. Koltermann (1974) demonstrated that this ability was arobust phenomenon when he trained bees to go to one specific site 19 times per day.Few researchers, however, have investigated animals' ability to associate a specificplace with food availability, as opposed to associating one place with different times. Inone such investigation, Biebach, Gordijn, and Krebs (1989) tested garden warblers, Sylviaborin, in a chamber consisting of a living area surrounded by four rooms, each with afeeder. Every day, food was available intermittently in a particular room for 3 hr. Foodwas available for 20 s, after which the bird had to return to the central area for 280 s,before food again became available in the adjacent room. The birds quickly learned to go3334to the correct room and avoid the incorrect rooms. On test days, although all four feederswere available, the birds continued to visit the appropriate rooms at the proper times.A second study, which used pigeons as subjects, demonstrated that these birds canlearn time-place associations over periods of 1 hr (Wilkie & Willson, 1992). In their studyit was shown that the birds could associate a specific pecking key with a particular 15-mininterval during a 1-hr session. The birds were tested in large Plexiglas chambers with onepecking key and feeder mounted on each of the four walls. Subjects were able to see avariety of spatial cues in the room through the transparent walls of the boxes. Subjectsreceived 60 min sessions during which each key provided food rewards for a 15-minperiod. In other words, Key 1 was rewarded for the first 15-min period of the hour, Key 2was rewarded for the second 15 min, and so on. During probe sessions in which the firstand the last 5 min of each 15 min interval were not rewarded, it was demonstrated that thepigeons were tracking food availability over time rather than merely responding tofeedings as cues for food availability.The goal of the present research is two-fold. The first objective is to extend theWilkie and Willson (1992) paradigm over longer periods, namely morning and afternoon.The second aim is to investigate the timing mechanisms underlying the behaviour ofvisiting certain places at certain times of the day.Extension of paradigmThree experiments were performed to determine if pigeons could track foodavailability over a 24-hr period. Ultimately, the reason for this was to enable us to get atthe question of mechanisms underlying the timing behaviour, but first it had to beestablished that the birds could time over such an extended period. Previously, the longestinterval over which pigeons have been shown to track food availability in an experimentalsetting was 1 hr. For this reason, a demonstration of pigeons timing over a 24-hr period inthe laboroatory would be a remarkable finding in itself. Experiment 1 was performed to35determine whether pigeons could actually remember over such an extended period oftimein a laboratory setting. In that experiment, food was available at one place (peckingkey) in the mornings, and at a different place in the afternoons. In Experiment 2, onlymorning trials or only afternoon trials were given, in order to rule out the possibility of analternation strategy. That is, if the birds were alternating their response over trials, thenthey should peck the afternoon key in the session following the morning session,irrespective of time of day. Thus, elimination of an afternoon session should cause themto peck the inappropriate key during the following morning's session. Finally, inExperiment 3, the length of time between morning and afternoon sessions was varied,again to verify a timing as opposed to an alternation strategy.Underlying mechanismsThe main properties of the two potential mechanisms underlying the different typesof timing are as follows. The circadian timing system is based on the phase of anendogenous oscillator (Gallistel, 1990). This oscillator is continuous, self-sustaining, andentrainable. Because it is self-sustaining, it does not require the presence of external cuesfor the timing behaviour (i.e., learned time-place associations) to be maintained. The roleof external cues is to act as "zeitgebers" which entrain the oscillator. In other words, if acircadian system is at work, the behaviour will persist in the absence of zeitgebers. Overtime, however, it may not adhere to a 24-hr period or schedule. Interval (stopwatch)timing, on the other hand, has been intensely investigated in the timing of short intervals(e.g., Gibbon & Church, 1984; S. Roberts, 1981). Basically, these timers are discrete andcan be stopped, started, reset, or restarted by various external events. Since stopwatchtimers are greatly influenced by external events, alterations in these cues will immediatelyaffect any type of behaviour associated with the interval clock. For example, a stopwatchtiming mechanism is reset every day by the lights coming on. This ensures that the animalwill start timing daily events from the same point every day. If lights-on time is shifted,36however, the stopwatch clock is reset for a different time and events will be predicted tooccur on the basis of the new starting point. Accordingly, if a stopwatch timer is beingused then the routine behaviour of the bird should be shifted to correspond with the newreset time.One approach to empirically distinguishing these two possibilities is to manipulatethe external cue, or zeitgeber, and observe the effects on behaviour. According to work inthe field of circadian rhythms, possible zeitgebers include temperature changes, socialsignals, and feeding. However, the most salient zeitgeber to which a cycle may beentrained is the light-dark cycle (Aschoff, 1989). Following from this, in Experiment 4 thelight-dark cycle was manipulated with a phase shift. More specifically, lights-on time wasshifted back by 6 hr, and performance was observed the following day. If the birds'learned time-place associations are being controlled by an interval timing mechanism thenan immediate shift in behaviour should be observed. However, if the behaviour iscontrolled by a self-sustaining oscillator, then the shift should have a minimal effect on thephase of the following cycle, that is, on the next day's performance.In Experiment 5, a further test of an underlying circadian timing mechanism wasperformed. The birds were held in constant dim light, and their performance on the time-place learning task was measured. If a bird is using an interval timer, elimination of lights-on and lights-off should cause errors in performance on the task because the major cuefor resetting the internal clock would be missing. If the bird is using a circadianmechanism, however, performance should continue as usual since this timing system isbased on a self-sustaining oscillator which is relatively independent of changes in externalcues.GENERAL METHODSSubjectsThe subjects were one White King (Allanon) and three Silver King (Silvercloak,Johann, and Jack) pigeons. The birds were maintained at approximately 90% of free-feeding body weight with a grain mixture obtained during testing sessions as well as post-session feedings. Subjects lived in large plastic coated wire mesh cages with ad libitumaccess to grit, oyster shells, and vitamin fortified water. The colony was climatecontrolled (200 C) and had a light/dark cycle matched to natural sunrise/sunset times.Luminance in the colony was approximately 82 cd/m 2 . Data collection occurred between3 January 1992 and 15 September 1992. Sunrise times ranged from 05:50 to 08:08 andsunset times ranged between 16:24 and 20:52 during this period. (Data provided by theAtmospheric Environment Service of Environment Canada).All subjects had previous experience with pecking keys for food reinforcement,but were naive with respect to this experimental procedure.ApparatusThe apparatus was identical to that used by Wilkie and Willson (1992). Briefly,pigeons were tested in one of two large, clear Plexiglas Skinner boxes with one 3.5 cmdiameter pecking key on the centre of each of the four walls. The floor area of the firstbox was 3600 cm2 (Silvercloak and Jack) and the floor area of the second box was 2025cm2 (Johann and Allanon). Behind each key was a microswitch, which recorded peckshaving a force greater than 0.15 N. The keys were illuminated with a 28VDC #313 lightcovered with a red gelatin filter, and each key was mounted above a standard grain feeder.The box was located in a small (2m x 2m) well lit testing room. Subjects were able to seea variety of room cues (window, door, posters, etc.) through the transparent walls of the37chamber. Data collection and experimental control were carried out by the MANXprogramming language (Gilbert & Rice, 1979) running on a minicomputer.General ProcedureAll subjects had previous key pecking experience; therefore no preliminary trainingwas required. Trials began with the illumination of all four keys in the box, and wereinitialized by the first peck made by a subject. After initialization, each trial lasted for 17min. During the first minute of the test, pecks to each key were recorded but did notproduce a food reward. During the remainder of the trial, 5-s access to mixed grain wasavailable on a variable-interval 1-min schedule for pecking the appropriate key (e.g., Key 1in the morning and Key 3 in the afternoon).EXPERIMENT 1: TRAININGMETHODThe initial question that I addressed was whether the pigeons could learn a time-place association over a long interval (24 hr). Subjects were exposed to a discrete trialprocedure in which one of four keys in a large Plexiglas Skinner box was rewarded,depending on the time of day. Response-initiated trials lasted for 17 min, and werepresented twice per day, once in the morning (starting between 9:00 and 10:00) and oncein the afternoon (starting between 15:30 and 16:30), 5 days per week. Altogether, eachsubject received 40 sessions of baseline training. Each bird had a different pair of keysthat provided food at different times in order to control for possible key preferences dueto previous experience with the present apparatus : Silvercloak -- morning key 1,afternoon key 3, Johann -- morning key 2, afternoon key 4, Jack -- morning key 3,afternoon key 1, Allanon -- morning key 4, afternoon key 2. All keys were lit in each3839session, instead of only the two that were rewarded for each bird, so that we couldmeasure potentially different types of error in later experiments.RESULTS AND DISCUSSIONFigure 2 illustrates pecks to the appropriate key for each bird for both morning andafternoon sessions during the first 8 weeks of the experiment. A discrimination ratio wascalculated by comparing the number of pecks to the appropriate key to the total number ofpecks made during the first non-rewarded minute of the session, and this ratio wasconverted to a percentage. Each point on the graph represents the mean of discriminationratios for each 5-day block during the 8-week period. It is important to emphasise thatfood cues were not available to indicate which key was correct.If the subject did not discriminate and simply responded randomly performancewould be 25% (one out of four keys is rewarded). The slope of the lines indicates thatthree of the birds began with a score at a chance or slightly above chance level, but by theend of the training session they were performing well above chance in the range of 80 to95 % correct. One bird, Jack, seemed to have some problems with the task duringmorning sessions. Initially he acquired the task at a rate similar to the rest of the subjects,but around the fourth week his performance suddenly dropped back to 50 % correct. Inthe afternoon, however, Jack reached a level of 80% correct by the eighth week.Figure 3 summarizes the percent of pecks per key in the first minute of eachsession during both the first block and the final block of baseline sessions. Pecks to allfour keys are represented: "Correct" bars represent the percent of pecks made to theappropriate key for the session, "alternate" bars represent the percent of pecks made to thekey appropriate for the alternate session, and "error" bars represent pecks made to theother two keys in the box .An examination of the data in this figure suggests that the birds were indeedcapable of acquiring the morning/afternoon task. That is, all subjects seemed quite40Figure 2: Acquisition of the 24-hr time-place learning task. This graph shows percent ofpecks to the appropriate key for both morning and afternoon sessions for allsubjects.^ SilvercloakLs - Johanne - Jack+ Allanon—A^ MeanBlock 1 2^3^4^5^6^7^8BlocksMorning AcquisitionAfternoon Acquisition—+— Silvercloak/ - 8 - Johann- e- -- Jack+ Allanon—A-- Mean0Block^1^2^3^4^5^6^7^8Blocks100 r808080 //40a_-20jr\41Figure 3: Mean percent of pecks per key during the first and the final block of baselinesessions. "Correct" bars represent pecks made to the appropriate key for thesession, "alternate" bars represent pecks made to the key appropriate for thealternate session, and "error" bars represent pecks made to the 2 keys which arenever rewarded.42Mean Baseline (Block 1)43100806040200^ errorM alternate■ correctA.M.^P.M.TimeMean Baseline (Block 8)100806040200^ error111 alternate■ correctA.M.^P.M.Time44capable of responding to the appropriate key at the proper time of day. A one-wayrepeated measures analysis of variance indicated a significant effect of blocks on percentcorrect for both morning acquisition (F(28)=2.804, p<.05) and afternoon acquisition(F(28)=4.244, p<.01). Also, a one sample t-test comparing final baseline to chance (25%)was significant for both morning (t(3)=3.88,p <.05) and afternoon (t(3)=11.39,p <.001).All subjects appeared to demonstrate time-place associative learning, even thoughthe length of time between trials was much greater than in previous studies illustrating thisphenomenon in pigeons (e.g., 1 hr in Wilkie & Willson, 1992). The birds responded atdifferent places at different times -- they learned that one out of four keys provided food inthe mornings and a different key provided food in the afternoons. At first glance, thisseems to indicate that the pigeons were keeping track of time of day in order to determinewhich was the appropriate key. However, there are other potential explanations for thisability. For instance, a simple alternation strategy is a possibility. Consider the subjectAllanon. For every second trial, Key 4 provides food, whereas for the alternate trials, Key2 is rewarded. He may simply be using an 'every-other-trial-key 2-is-good' type of rule todetermine which key will provide him with a reward. In order to rule out this possibility,several probe trials were conducted in Experiment 2.EXPERIMENT 2: AM ONLY / PM ONLYMETHODIn an attempt to eliminate the possibility of an alternation type of strategy, eighttrials in which only the morning session was presented were inserted twice per weekbetween baseline days. Thus, in one 4 week period birds were given 8 morning onlysessions and 12 baseline sessions. If the birds were merely alternating their response overtrials, then they should automatically peck the afternoon key in the session following themorning session, regardless of the time of day. Thus, elimination of any afternoon session45should cause them to peck the wrong key during the following morning's session.Following these morning only probes, a similar type of test was conducted in which eightmorning sessions were eliminated over a 4-week period. Again, these afternoon onlysessions were given twice per week and were alternated with 12 baseline sessions. Therationale was the same: Missing the morning session should cause the birds to go to theinappropriate key during the afternoon session of the same day, in the case that alternationis being used.RESULTS AND DISCUSSIONFigure 4 illustrates percent of pecks per key for both the sessions following themorning-only probes and the sessions following the afternoon-only probes. This graph hasthe same format as Figure 3, (i.e., pecks to all four keys are represented).All of the birds, with the possible exception of Jack in the morning, seemed topeck the appropriate key in spite of the elimination of the previous session. Subjectsseemed to do slightly better in the mornings, with an average of 80 % of responses to theappropriate key, whereas in the afternoon about 75 % of responses were to the correctkey.Thus, all subjects showed evidence of using a timing as opposed to an alternationstrategy . The subjects responded to the appropriate key in both the mornings and theafternoons, in spite of the fact that the previous session had been eliminated. If a bird wasreceiving rewards for pecking Key 2 in the mornings and Key 4 in the afternoons, thenKey 2 again, an alternation strategy would suggest a high degree of error if one of thesessions was eliminated. The low error rate and high degree of response to the appropriatekey suggest that this was not the case. In order to further strengthen this finding,however, a second type of probe trial was conducted in Experiment 3.Figure 4: Percent of pecks per key for both sessions following the morning-only probesand the sessions following the afternoon-only probes.4647Silvercloak -- AM Only / PM Only^ Johann -- AM Only / PM Only10080604020^ errorSi alternate■ correct10080604020^ error• alterna■ correctA.M.^P.M.^ A.M.^P.M.Time^ TimeJack -- AM Only / PM Only^ Allanon -- AM Only / PM Only10080604020^ errorRI alternate■ correct10080604020^ errorE alternate■ correctA.M.^P .M .^ A.M.^P.M .Time^ TimeEXPERIMENT 3: LATE A.M./EARLY P.MMETHODA second way of ensuring that a timing strategy was being used was to change thespacing between morning and afternoon sessions. If an alternation strategy is being used,then the error rate should be the same as in the baseline sessions, or perhaps less becauseof the shortened retention interval between the two sessions. If a timing strategy is beingused, however, the error rate should increase in relation to how far the probe sessiondeviates from the baseline time of session. Baseline time of sessions was approximately09:30 for the morning sessions and 16:00 for the afternoon sessions. Two separate probetrials were conducted (six sessions each): Probe 1 (first session at 11:00 and secondsession at 15:00) and Probe 2 (first session at 12:30 and second session at 14:00). Thusthe inter-session intervals varied between 6.5 hr (baseline), 4.5 hr (Probe 1) and 1.5 hr(Probe 2).RESULTS AND DISCUSSIONPerformance on Probe 1, Probe 2, and Baseline for morning sessions is illustratedin Figure 5.Two of the birds (Silvercloak and Allanon) tended to perform worse than baselineon both types of probe, as expected if subjects were using the phase of a circadianoscillator as a timer. The other two birds, however, showed different patterns. Johannshowed slightly better performance on Probe 2 than on baseline, and Jack was better atboth probes than baseline. Jack's aberrant results are likely due to the fact that his typicalmorning baseline performance was far below average, as mentioned in Experiment 1. Theresults for the afternoon session are much clearer (see Figure 6).48Figure 5: Performance of all subjects in morning sessions as compared to inter-sessionintervals.(Baseline, 6.5 hr; Probe 1, 4.5 hr; Probe 2, 1.5 hr).49100900 8000 70C0000 6050401-A.M. Performance versus Inter -Session Interval50/*/\   Mean^ Allanon— — Jack-'^ - Johann\   Silvercloak2^3^4^5^6^7Inter-Session Interval (Hrs.)Figure 6: Performance of all subjects in afternoon sessions as compared to inter-sessioninterval (Baseline, 6.5 hr; Probe 1, 4.5 hr; Probe 2, 1.5 hr).51P.M. Performance versus Inter-Session Interval10090//080 •- -^--//0--It0°) 70 --•0_^ Mean60 -  ^AllanonJackJohann50 Silvercloak1^2^3^4^5^6^7Inter -Session Interval (Hrs.)5253Between 80% and 99% of the variance in percent correct could be accounted forby inter-session interval (Silvercloak r 2 = .992; Johann r2=.801; Jack r2=.797; Allanonr2=.895). As predicted, this suggests that performance was worse depending on how farthe interval between testing times deviated from the original testing times, therebyproviding further evidence for a timing strategy.One reason for the variable results for the birds during the morning session may bethe fact that the actual difference in times for the two probe trials (2.5-4.5 hours) was notvery large in relation to the total amount of time that the bird was representing (24 hours).As mentioned previously, the subjects' timing systems seem to be somewhat flexible.Animals must be able to adapt to changes in their environment. Consequently, theirtiming systems seem to be able to accomodate slight discrepancies in time of trialpresentation. All three types of probes in both Experiment 2 and Experiment 3 providestrong evidence that the birds were using a timing strategy, so we moved on to looking atthe possible underlying mechanisms in Experiment 4.EXPERIMENT 4: PHOTOPERIODMETHODOnce it is established that the birds are using a timing strategy, the next step is tolook at the mechanisms underlying the behaviour. One possibility is that the birds areusing an interval or stopwatch timer. A second possibility is that the birds are using acircadian timing mechanism. A main feature of a circadian system is that it is self-sustaining. Therefore, after a sudden shift in photoperiod behaviour will continue as usual,at least for the first cycle following the shift. However, with extended exposure to thenew photoperiod, the circadian cycle will gradually become entrained to the new light-dark cycle, and timing behaviour will shift accordingly. In contrast, an interval timer is54reset daily by lights-on time. Any behaviour based on timing will not persist but insteadwill immediately shift in accordance with the new photoperiod.In this experiment the four birds in the study were moved from the colony to a newroom so that their light-dark cycle could be shifted without affecting the other birds in thecolony. A new baseline measure was taken in order to ensure that the move did not causea change in the birds' performance. Illuminance in the new room approximated that of thecolony (95 cd/m2). After 15 days of adjustment to the new room, lights-on time wasshifted backward by six hours (from 06:00 to 24:00). The subjects were tested as usualfor baseline sessions on the day immediately following the shift (around 09:30 and 16:00)for 6 days following the shift.RESULTS AND DISCUSSIONFigure 7 shows that the birds were not affected by the change in rooms;consequently the shift in light-dark cycle was performed.Figure 8 illustrates the performance of each bird on the day immediately followingthe phase shift. Each of the four birds continued to respond at above chance to theappropriate key (Silvercloak, 70%; Johann, 98%; Jack, 65%; Allanon, 100%). It isinteresting to note that for most birds, correct response actually increased after the phaseshift. Since subjects' behaviour was not deleteriously affected by the phase shift, thesedata are consistent with a circadian based timing system.All four birds maintained their performance in the session following the phase shift.This provides evidence for a circadian, not a stopwatch, timing system. A clock based onan interval system would be reset by a major external cue such as a phase shift. Acircadian system, on the other hand, is run by a self-sustaining oscillator, and thereforewould not be expected initially to lead to a change in behaviour following a phase shift.Finally, Figure 9 shows performance of the birds when the phase shift wasextended for a period of 6 days. The data for individual birds are considerably different.Figure 7: Baseline performance of all subjects after being moved from the colony to thenew room.55P.M.A.M.^ errorIM alternate■ correctA.M. P.M.100 -^ 10080^ errorE alternate■ correct604020080604020010010080806040402020600^ error111 alternate■ correct^ errorIN alternate■ correct56Silvercloak -- New Baseline^Johann -- New BaselineTime^ TimeJack -- New Baseline^ Allanon -- New BaselineA.M.^P.M.^ A.M.^P.M.Time^TimeFigure 8: Percent of pecks per key for each bird on the day immediately following thephase shift.57100100808060404020206000 A.M.^P.M.P.M.A M1:1 errorE alternate■ correct^ errorE alternate■ correctI 11008060402001:1 errorEl alternate■ correctA.M. P.M.^ errorE alternate■ correctA.M.^P.M .10080a)a)o_ 60(f) 40o_20058Silvercloak -- Session Following Phase Shift^Johann -- Session Following Phase ShiftTime^ TimeJack -- Session Following Phase Shift^Allanon -- Session Following Phase ShiftTime^ TimeFigure 9: Percent of pecks per key for all birds after the phase shift was extended for aperiod of 6 days.59100802004060A.M. P.M.^ errorE alternate■ correct^ errorIN alternate■ correctA.M.60Silvercloak -- Phase Shift^Johann -- Phase ShiftTime^TimeJack -- Phase Shift^Marion -- Phase Shift10080604020^ errorM alternate■ correct^ errorE alternate■ correctA.M.^P .M .^ A.M.^PMTime^ Time61Both Johann and Allanon performed at an above chance level, although theirscores did drop noticeably. Silvercloak, however, fell to a level that was basically atchance. He seemed to peck every key equally, including those for which he was neverrewarded. Jack's scores were also greatly influenced by the continued phase shift, but hisstrategy was quite different from Silvercloak's. While his performance in the morning wasstill extremely high, performance in the afternoon for the correct key was actually belowchance. Pecks to the morning-appropriate key in the afternoon, however, approached80%. This suggests that Jack used the strategy of pecking at one key, both in themornings and the afternoons, thereby indicating that he was no longer timingappropriately.The finding that two of the birds were able to continue to respond appropriatelyfor up to 6 days provided further evidence for a circadian based system. First, two of thebirds were able to continue to respond appropriately for up to 6 days, thereby providingfurther evidence for a circadian based system. Also, the fact that these birds scoresdropped quite a bit from their performance during the first session following the shift, incombination with the poor performance by the other two subjects after six days, providesstrong evidence that they are not using an alternation type of strategy. If they werealternating, then performance should either remain the same or improve over the course ofthe 6 day period. Although this experiment provides strong evidence for a circadian-basedtiming mechanism, we performed one more test in Experiment 5 to confirm that this is thecase.EXPERIMENT 5: DIM LIGHTMETHODA second way of testing a circadian timing mechanism is, instead of shifting lights-on time, to eliminate it altogether. As in the previous experiment, absence of the daily62lights-on cue should not affect timing behaviour, at least for the first cycle after thetransition to dim light, if it is based on a self-sustaining mechanism.In this experiment the birds were maintained in constant dim light (14 cd/m 2) andthen were tested as usual for baseline sessions ( 9:30 and 16:00) for 4 days following themove to dim light.RESULTS AND DISCUSSIONPerformance on the first day of the dim light condition is shown in Figure 10.Three subjects (Johann, Jack, and Allanon) performed extremely well.For each of these birds percent of pecks to the correct key were well abovechance, and did not differ from baseline scores. This strongly supports a self-sustaining,circadian-based timing system. Silvercloak, on the other hand, performed above chance inmorning sessions but did not peck accurately in the afternoons, thereby suggesting that hemay have been using an alternate timing strategy. Silvercloak's performance in thisexperiment illustrates the fact that different individuals may use different strategiesdepending on the situation.Figure 10: Percent of pecks per key for all birds under the dim light condition.6310080604020100204060800A.M. P.M .A.M . P.M .^ errorIN alternate■ correct^ error• alternatE■ correct^ errorEl alternate■ correct 0P.M.A.M.P.M .A.M.100Iffl error^ alternal■ correct100806040208060402064Silvercloak -- Dim Light^ Johann -- Dim LightTime^TimeJack -- Dim Light^ Allanon -- Dim LightTime^ Time65GENERAL DISCUSSIONThe first goal of this research was to determine whether pigeons are capable, in thelaboratory, of timing over intervals as long as 24 hr. Previous work has demonstratedtiming capabilities in the order of seconds (e.g., Gibbon & Church, 1984; S. Roberts,1981) or over 1 hr (Wilkie & Willson, 1992). The present work suggests that pigeons areindeed capable of timing over 24 hr and that they appear to be using a true timing strategy,as opposed to a learned pattern such as alternation. When required to peck at a key thatwas reinforced only in the mornings and a different key which was providing food only inthe afternoons, the pigeons learned this discrimination up to at least 80% correct. Wheneither morning or afternoon sessions were omitted, the birds maintaind their level ofresponding. Finally, when the amount of time between sessions was altered the birds'error level increased. These three findings suggest that pigeons can time over an extendedperiod. This ability is important in the natural environment as resources may be variablyavailable over the course of a day or a week, not merely over the course of an hour. Anorganism must be able to integrate timing and behaviour in order to gain optimal resourcesin the real world.Once it was established that subjects were capable of timing over an extendedperiod I was interested in the mechanism underlying the behaviour. The results fromExperiment 4 and 5 suggest that this long term timing is mediated by a circadian clock asopposed to an interval timer. When their light-dark cycle was shifted back by 6 hr thebirds' performance on the day following the shift remained high. In addition, when birdswere kept in constant dim light, performance in the subsequent testing sessions wascomparable to baseline performance. This suggests that their timing involved some sort ofself-sustaining oscillator which would allow for the maintenance of appropriate behaviourin spite of significantly altered external cues. If a stopwatch mechanism was being usedthen a substantial decrease in performance would be expected because the clock should be66reset by the change in lights-on time, and subsequent behaviour should correspond withthat shift. This finding corresponds with Biebach et al.'s (1989) work on garden warblers,where it was also found that a circadian mechanism best explained learning on a time-placetask in which four 3-hr periods were timed. It does not correspond, however, with afinding of stopwatch-like timing in a one hour time-place paradigm in pigeons (Wilkie,Saksida, Samson & Lee; in press). This suggests that session duration, not speciesdifference, is the most important determinant of the timing system used. In addition, thissupports the idea that pigeons can be flexible in their use of timing systems: Pigeonsappear to use a stopwatch-like system for relatively short intervals and a circadian-basedsystem for longer intervals.Further evidence for flexibility of timing was found in the last part of Experiment4. In this part of the experiment, the phase shift was extended for a period of six days.Each bird appeared to use a different strategy to cope with the change in light-dark cycle.This suggests that although pigeons may generally use particular strategies to time overdifferent intervals, they seem to be able to adapt their individual strategies depending onthe situation.Individual differences in timing strategies is a thread that runs through several ofthe experiments in this thesis. In the first 3 experiments, the birds had very comparabledata, with the exception of Jack in the mornings. His accurate performance duringafternoon sessions, however, suggested that he was learning the task and was not using anon-timing strategy such as alternation. His performance in the mornings did improveover the course of the research, however, which again suggests that he was not using adifferent strategy but was just slower in learning the task. Experiments 4 and 5, whichwere designed to investigate the mechanisms underlying the time-place learning behaviour,showed some individual differences that appear to be more likely due to differingstrategies as opposed to differences in acquisition time. The obvious example of thisoccurs in Experiment 4 when the phase shift was extended for a period of several days.67Initially, when the phase shift occured, all four subjects adopted the same circadian-basedtiming strategy. Their performance on the task was not affected by the major change inlight-dark cycle. However, after several days of the lights-on cue not corresponding withreward, all of the birds' behaviour began to change. Two of the subjects, Johann andAllanon, were still going to the keys that had been rewarded before the phase shift. Aconsiderable amount of error ("trying out" other keys), however, did begin to creep intotheir data. Silvercloak, on the other hand, started to peck at all keys equally, even thosefor which he had never been rewarded. Jack took a completely different approach to theproblem. He pecked consistently at one key, both in the morning and in the afternoon,thus he did very well during morning sessions but failed miserably in the afternoons.These differences in the data do not suggest that the birds were using different timingstrategies, because in this part of the experiment none of the birds were timingappropriately for the rewards that were available to them. Instead, they were usingdifferent strategies to solve the problem that although their self-sustaining timingmechanism was still running, the rewards that they had been getting previously were nolonger available. In other words, although a circadian timing mechanism is not governedby external events, it is still affected by them, and will gradually become entrained(synchronised) with the light-dark cycle (see Aschoff, 1989). An interesting follow up toExperiment 4 would be to extend the phase shift for several weeks, and determine whetherthe birds could eventually all return to baseline performance, in spite of the differentstrategies that they might use to get there.68REFERENCESAbe, H. & Sugimoto, S. (1987). Food anticipatory response to restricted food accessbased on the pigeon's biological clock. Animal Learning & Behavior, 15, 353-359.Anger, D. (1963). The role of temporal discriminations in the reinforcement of Sidmanavoidance behavior. Journal of the Experimental Analysis of Behavior, 6, 477-506.Aschoff, J. (1989). Temporal orientation: Circadian clocks in animals and humans.Animal Behaviour, 37, 881-896.Aschoff, J. (1984). Circadian timing. In: J. Gibbon & L. Allan (Eds.), Timing and timeperception, (pp. 442-468). New York: New York Academy of Sciences.Aschoff, J. (Ed.). (1981). Biological rhythms, handbook of behaviouralneurobiology, 4. New York: Plenum Press.Aschoff, J., von Goetz, C., & Honma, K. (1983). Restricted feeding in rats: Effects ofvarying feeding cycles. Zeitschrift fur Tierpsychologie, 63, 91-111.Aschoff, J., Daan, S., & Groos, G. (Eds.). (1982). Vertebrate circadian systems:Structure and physiology. Berlin: Springer-Verlag.Beling, I. (1929). Uber das Zeitgedachtnis der Bienen. Zeitschrift fur vergleichendePhysiologie, 9, 259-238.Biebach, H., Falk, H. & Krebs, J.R. (1991). The effect of constant light and phase shiftson a learned time-place association in garden warblers (Sylvia Bonin): Hourglassor circadian clock? Journal of Biological Rhythms, 6, 353-365.Birch, D., Burnstein, E., & Clark, R. (1958). Response strength as a function of hours offood deprivation under a controlled maintenance schedule. Journal ofComparative Physiology and Psychology, 350-354.Bolles, R.C., & deLorge, J. (1962). The rat's adjustment to a diurnal feeding cycle.Journal of Comparative Physiology and Psychology, 55, 760-762.69Boulos, Z., & Terman, M. (1980). Food availability and daily biological rhythms.Neuroscience and Biobehaviour Review, 4, 119-131.Broadbent, D.E. (1958). Perception and communication. New York: Pergamon Press.Caccamise, D.F., & Morrison, D.W. (1986). Avian communal roosting: Implicationsfor diurnal activity centers. American Naturalist, 128, 191-198.Catania, A.C. (1970). Reinforcement schedules and psychophysical judgements: A studyof some temporal properties of behavior. In: W.N. Schoenfeld (Ed.), The theoryof reinforcement schedules (pp. 1-42). New York: Appleton-Century-Crofts.Church, R.M. (1984). Properties of the internal clock. In: J. Gibbon and L. Allan (Eds.),Timing and time perception, (pp. 567-582). New York: New York Academy ofSciences.Church, R.M. (1980). Short-term memory for time intervals. Learning and Motivation,11,208-219 .Church, R.M. (1978). The internal clock. In: Hulse, S.H., Fowler, H. & Honig, W.K.(Eds.) Cognitive processes in animal behavior. (pp. 277-310) New Jersey:Erlbaum.Church, R.M., & Broadbent, H.A. (1990). Alternative representations of time, number,and rate. Cognition, 37, 55-81.Church, R.M., & Deluty, M.Z. (1977). The bisection of temporal intervals. Journal ofExperimental Psychology: Animal Behaviour Processes, 3, 216-228.Church, R.M., & Deluty, M.Z. (1976). Scaling of time by rats. Paper presented at theAnnual Meeting of the Psychonomic Society, St. Louis, Missouri.Church, R.M., & Gibbon, J. (1982). Temporal generalization. Journal of ExperimentalPsychology: Animal Behavior Processes, 8, 165-186.Church, R.M., Miller, K.D., Gibbon, J., & Meck, W.H. (1988). Symmetrical andasymmetrical sources of variance in temporal generalization. Paper presented atthe Annual Meeting of the Psychonomic Society, Chicago, November, 1988.70Church, R.M., & Meck, W.H. (1984). The numerical asttributes of stimuli. In: H.L.Roitblat, T.G. Bever, & H.S. Terrace (Eds.) Animal cognition (pp. 445-464).Hillsdale, NJ: Erlbaum.Church. R.M., & Roberts, S. (1975). Control of an internal clock Paper presented atthe Annual Meeting of the Psychonomic Society, Denver, Colorado.Cowcroft, P. (1954). The daily cycle of activity in British shrews. Proceedings of theZoological Society (London), 123, 715-729.Daan, S. (1981). Adaptive daily strategies in behavior. In: J. Aschoff (Ed.), Biologicalrhythms, (pp. 275-298). New York: Plenum.Daan, S., & Aschoff, J. (1981). Short-term rhythms in activity. In: J. Aschoff (Ed.),Biological rhythms (pp. 491-499). New York: Plenum.Daan, S., & Koene, P. (1981). On the timing of foraging flights by oystercatchers,Haematopus ostralegus, on tidal mudflats. Netherlands Journal of SeaResearch, 15, 1-22.Daan, S., & Slopsema, S. (1978). Short term rhythms in foraging behavior of thecommon vole, Microtus avarilis. Journal of Comparative Physiology, 127, 215-227.Davis, H., & Memmot, J. (1982). Counting behavior in animals: A critical evaluation.Psychological Bulletin, 92, 547-591.Dews, P.B. (1970). The theory of fixed-interval responding. In W.N. Schoenfeld (Ed.)The theory of reinforcement schedules (pp. 43-61). New York: Appleton-Century-Crofts.Edmonds, S.C. & Adler, N.T. (1977). Food and light as entrainers of circadian runningactivity in the rat. Physiology & Behavior, 18, 915-919.Eiserer, L.A. (1984). Communal roosting in birds. Bird Behaviour, 5, 61-80.Farner, D.S. (1985). Annual rhythms. Annual Review of Physiology, 47, 65-82.71Farris, H.E. (1967). Classical conditioning of courting behavior in the Japanese quail,Coturnix coturnix japonica. Journal of the Experimental Analysis of Behavior,10, 213-217.Fernandes, D.M., & Church, R.M. (1982). Discrimination of number and sequentialevents by rats. Animal Learning & Behavior, 10, 171-176.Gallistel, C.R. (1990). The organization of learning. Cambridge, Mass: MIT Press.Gibbon, J. (1991). Origins of scalar timing. Learning and Motivation, 22, 3-38.Gibbon, J. (1977). Scalar expectancy theory and Weber's Law in animal timing.Psychological Review, 84, 279-325.Gibbon, J. (1972). Timing and discrimination of shock density in avoidance.Psychological Review, 79, 68-92.Gibbon, J., & Balsam, P.D. (1981). Spreading association in time. In: C.M. Locurto,H.S. Terrace, & J. Gibbon (Eds.), Autoshaping and conditioning theory (pp. 219-253). New York: Academic Press.Gibbon, J., & Church, R.M. (1984). Sources of variance in an information processingtheory of timing. In: H.L. Roitblat, T.G. Bever, & H.S.Terrence (Eds.) Animalcognition (pp. 465-488). Hillsdale, N.J. : Erlbaum.Gibbon, J., & Church, R.M. (1981). Time left: Linear versus logarithmic subjectivetime. Journal of Experimental Psychology: Animal Behavior Processes, 7, 87-107.Gibbon, J., Church, R.M., & Meck, W.H. (1984). Scalar timing in memory. In: J.Gibbon & L. Allan (Eds.), Timing and time tprception (pp. 52-77). New YorkAcademy of Sciences.Gilbert, S.G., & Rice, D.C. (1979). NOVA SKED II: A behavioural notation languageutilizing the Data General Corporation real-time disk operating system.Behavioral Research Methods & Instrumentation, 11, 71-73.72Grant, D.S., & Spetch, M.L. (1981). Pigeons' memory for event duration: Differencesbetween choice and successive matching tasks. Learning and Motivation, 22,180-199.Gwinner, E. (1981). Circannual systems. In: J. Aschoff (Ed.), Biological rhythms, (pp.391-410). New York: Plenum.Gwinner, E. (1978). Effects of pinealectomy on circadian locomotor activity rhythms inEuropean starlings, Sturnus vulgaris. Journal of Comparative Physiology, 126,123-129.Gwinner, E. (1966). Entrainment of a circadian rhythm in birds by species-specific songcycles (Ayes, Fringillidae: Carduelis spinus, Serinus serinus). Experientia, 22,765.Hearst, E., & Jenkins, H.M. (1974). Sign tracking: The stimulus reinforcer relation anddirected action. Austin, Texas: Psychonomic Society.Heineman, E.G., Avin, E., Sullivan, M.A., & Chase, S. (1969). Analysis of stimulusgeneralization with a psychophysical method. Journal of ExperimentalPsychology, 80, 215-224.Hoffman, K. (1960). Experimental manipulation of the orientational clock in birds. ColdSpring Harbour Symposium on Quantitative Biology, 25, 379-387.Holder, M.D., & Roberts, S. (1985). Comparison of timing and classical conditioning.Journal of Experimental Psychology: Animal Behavior Processes, 11, 172-193.Hollis, K.L. (1984). The biological function of Pavlovian conditioning: The best defenseis a good offense. Journal of Experimental psychology: Animal BehaviorProcesses, 10, 413-425.Hollis, K.L. (1982). Pavlovian conditioning of signal-centered action patterns andautonomic behavior: A biological analysis of function. Advances in the Study ofBehavior, 112, 1-64.73Honma, K., von Goetz, C., & Aschoff, J. (1983). Effects of restricted daily feeding onfreerunning circadian rhythms in rats. Physiology and behaviour, 30, 905-913.Jacklet, J.W. (1985). Neurobiology of circadian rhythm generators. Trends inNeurosciences, 8, 69-73.Killeen, P.R., & Fetterman, J.G. (1988). A behavioral theory of timing. PsychologicalReview, 95, 274-295.Kleber, E. (1935). Hat das Zeitgedachtnis der Bienen biologische Bedeutung? Z. vergl.Physiol.,22, 221-262.Koltermann, R. (1974). Periodicity in the activity and learning performance of the honeybee. In: L.B. Browne (Ed.) The experimental analysis of insect behavior. (pp.218-226). Berlin: Springer.Kramer, T.J., & Rilling, M. (1970). Differential reinforcement of low rates: A selectivecritique. Psychological Bulletin, 74, 224-254.LaBarbera, J.D., & Church, R.M. (1974). Magnitude of fear as a function of expectedtime to an aversive event. Animal Learning & Behavior, 2, 199-202.Lang, J.W. (1976). Amphibious behavior of Alligator mississippiensis: Roles of acircadian rhythm and light. Science, 191, 575-577.Libby, M.E., & Church, R.M. (1974). Timing of avoidance responses by rats. Journal ofthe Experimental Analysis of Behavior, 22, 513-517.Maricq, A.V., & Church, R.M. (1983). The differential effects of haloperidol andmethamphetamine on time estimation in the rat. Psychopharmacology, 79, 10-15.Maricq, A.V., Roberts, S., & Church, R.M. (1981). Methamphetamine and timeestimation. Journal of Experimental Psychology: Animal Behavior Processes, 7,18-30.Maricq, A.V. (1978). Some effects of lesions on the prefrontal cortex on timing behaviorin the rat. Unpublished honours thesis, Brown University.74Marimuthu, G., Subbaraj, R., & Chandrashekaran, M.K. (1981). Social entrainment ofthe circadian rhythm in the flight activity of the microchiropteran bat Hipposiderosspeoris. Behavioural Ecology and Sociobiology, 8, 147-150.Meck, W.H. (1983). Selective adjustment of the speed of internal clock and memoryprocesses. Journal of Experimental Psychology: Animal Behavior Processes, 9,171-201.Meck, W.H., & Church, R.M. (1984). Simultaneous temporal processing. Journal ofExperimental Psychology: Animal Behavior Processes, 10, 1-29.Meck, W.H., Church, R.M., & Olton, D.S. (1984). Hippocampus, time, and memory.Behavioral Neuroscience, 98, 3-22.Meck, W.H., & Church, R.M. (1983). A mode control model of counting and timingprocesses. Journal of Experimental Psychology: Animal Behavior Processes, 9,320-334.Meck, W.H., & Church, R.M. (1982a). Abstraction of temporal attributes. Journal ofExperimental Psychology: Animal Behavior Processes, 8, 226-243.Meck, W.H.,& Church, R.M. (1982b). Discrimination of intertrial intervals in cross-modal transfer of duration. Bulletin of the Psychonomic Society, 19, 234-236.Menaker, M., & Zimmerman, N. (1976). Role of the pineal in the circadian system ofbirds. American Zoologist, 16, 45-55.Mistlberger, R. (1993, in prep.). Circadian food-antivipatory activity: Formal modelsand psychological mechanisms.Mistlberger, R. E., & Rusak, B. (1988). Food-anticipatory circadian rhythms in rats withparaventricular and lateral hypothalamic ablations. Journal of Biological Rhythms,3, 277-291.Mistlberger, R. E.,& Rechtschaffen, A. (1984). Recovery of anticipatory activity torestricted feeding in rats with ventromedial hypothalamic lesions. Physiology &Behaviour, 334, 227-235.75Moore, R.Y. (1982). The suprachiasmatic nucleus and the organization of a circadiansystem. Trends in Neurosciences, 5, 404-407.Norgren, R.B. (1990). Neural basis of avian circadian rhythms. Bird Behaviour, 8, 57-66.Ostheim, J. (1992). Coping with food-limited conditions: Feeding behavior, temperaturepreference, and nocturnal hypothermia in pigeons. Physiology & Behavior, 51,353-361.Pavlov, I.P. (1927). Conditioned reflexes (translated by G.V. Anrep). London: OxfordUniversity Press.Pittendrigh, C.S. (1980). Some functional aspects of circadian pacemakers. In M. Suda,0. Hayaishi, & H. Nakagawa (Ed.), Biological rhythms and their centralmechanism (pp. 3-12). New York: Elsevier.Pittendrigh, C. S. (1954). On temperature independence in the clock system controllingemergence time in Drosophila. Proceedings of the National Academy of ScienceU.S.A., 40, 1018-1029.Platt, J.R., Kuch, D.O., & Bitgood, S.C. (1973). Rats' lever-press duration aspsychophysical judgements of time. Journal of the Experimental Analysis ofBehavior, 19, 239-250.Richter, C.P. (1922). A behaviouristic study of the activity of the rat. ComparativePsychology Monographs, 1, 1-55.Roberts, S. (1983). Properties and function of an internal clock. In: R.L. Mellgren(Ed.), Animal cognition and behavior (pp. 345-397). Amsterdam: North-Holland.Roberts, S. (1982). Cross-modal use of an internal clock. Journal of ExperimentalPsychology: Animal Behavior Processes, 8, 2-22.Roberts, S. (1981). Isolation of an internal clock. Journal of Experimental Psychology:Animal Behavior Processes, 7, 242-268.76Roberts, S., & Church, R.M. (1978). Control of an internal clock. Journal ofExperimental Psychology: Animal Behavior Processe, 4, 318-337.Roberts, S., & Holder, M.D. (1985). Effect of classical conditioning on an internalclock. Journal of Experimental Psychology: Animal Behavior Processes, 11,194-214.Roberts, S., & Holder, M.D. (1984a). The function of time discrimination and classicalconditioning. In: J. Gibbon and L. Allan (Eds.), Timing and time perception (pp.228-241). New York: New York Academy of Sciences.Roberts, S., & Holder, M.D. (1984b). What starts an internal clock? Journal ofExperimental Psychology: Animal Behavior Processes, 10, 273-296.Roberts, S.K. (1965). Photoreception and entrainment of cockroach activity rhythms.Science, 148, 958-960.Rusak, B. (1977). The role of the suprachiasmatic nuclei in the generation of circadianrhythms in the golden hamster, Mesocricetus auratus. Journal of ComparativePhysiology, 118, 145-164.Schneider, B.A. (1969). A two-state analysis of fixed-interval responding in the pigeon.Journal of the Experimental Analysis of Behavior, 12, 677-687.Spetch, M.L. (1987). Systematic errors in pigeons' memory for event duration:Interaction between training and test delay. Animal Learning & Behavior, 15,1-5.Spetch, M.L., & Rusak, B. (1989). Pigeons' memory for event duration: Intertrialinterval and delay effects. Animal Learning & Behavior, 17, 147-156.Spetch, M.L., & Sinha, S. S. (1989). Proactive effects in pigeons' memory for eventduration: Evidence for analogical retention. Journal of Experimental Psychology:Animal Behavior Processes, 15, 347-357.77Spetch, M.L., & Wilkie, D.M. (1983). Subjective shortening: A model of pigeons'memory for event duration. Journal of Experimental Psychology: AnimalBehavior Processes, 9, 14-30.Staddon, J. E. R. (1983). Adaptive behavior and learning. Cambridge: CambridgeUniversity Press.Stubbs, D.A. (1968). The discrimination of stimulus duration by pigeons. Journal of theExperimental Analysis of Behavior, 11, 223-238.Sulzman, F., Ellman, D., Fuller, C., Moore-Ede, M., & Wasser, G. (1984). Neurosporacircadian rhythms in space: A reexamination of the endogenous-exogenousquestion. Science, 225, 232-234.Takahashi, J.S., & Menaker, M. (1979). Brain mechanisms in avian circadian systems.In: M. Suda, 0. Hayaishi, & H. Nakagawa (Eds.) Biological rhythms and theircentral mechanism. (pp. 95-109). Amsterdam: Elsevier.Turek, F.W. (1983). Neurobiology of circadian rhythms in mammals, Bioscience, 33,439-444.Wilkie, D.M. (1988). Proactive effects in pigeons' timing behavior: Implications for aninternal-clock model. Animal Learning & Behavior, 16, 132-136.Wilkie, D.M. (1987). Stimulus intensity affects pigeons' timing behavior: Implicationsfor an internal clock model. Animal Learning & Behavior, 15, 35-39.Wilkie, D.M., Saksida, L.M., Samson, P., & Lee, A. (In press). Properties of time-place learing by pigeons, Columba livia.Wilkie, D.M., & Willson, R.J. (1992). Time-place learning by pigeons, Columba livia.Journal of the Experimental Analysis of Behavior, 57, 145-158.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items