UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Pigeons’ memory for event duration Spetch, Marcia Louise 1981

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-UBC_1982_A1 S68.pdf [ 4.88MB ]
Metadata
JSON: 831-1.0095636.json
JSON-LD: 831-1.0095636-ld.json
RDF/XML (Pretty): 831-1.0095636-rdf.xml
RDF/JSON: 831-1.0095636-rdf.json
Turtle: 831-1.0095636-turtle.txt
N-Triples: 831-1.0095636-rdf-ntriples.txt
Original Record: 831-1.0095636-source.json
Full Text
831-1.0095636-fulltext.txt
Citation
831-1.0095636.ris

Full Text

PIGEONS' MEMORY FOR EVENT DURATION by MARCIA LOUISE SPETCH B.A. The University of B r i t i s h Columbia, 1977 M.A. The University of B r i t i s h Columbia, 1979 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Doctor of Philosophy i n THE FACULTY OF GRADUATE STUDIES Department of Psychology We accept t h i s thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA October 1981 (c)Marcia Louise Spetch, 1981 In presenting t h i s thesis i n p a r t i a l f u l f i l m e n t of the requirements for an advanced degree at the University of B r i t i s h Columbia, I agree that the Library s h a l l make i t f r e e l y a v a i l a b l e for reference and study. I further agree that permission for extensive copying of t h i s thesis for s c h o l a r l y purposes may be granted by the head of my department or by h i s or her representatives. It i s understood that copying or p u b l i c a t i o n of t h i s thesis for f i n a n c i a l gain s h a l l not be allowed without my written permission. Department of The University of B r i t i s h Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1W5 Date (VJ- A£} / f ?/ i i ABSTRACT Pigeon's working memory for event duration was i n v e s t i -gated using variations of the delayed matching to sample procedure. When a retention i n t e r v a l of variable length was interposed between the sample and comparison s t i m u l i , pigeons responded as though a long-duration sample had been short after retention i n t e r v a l s of 10 sec or greater. This "choose short" e f f e c t occurred r e l i a b l y i n each subject,.regardless of whether the subject was naive or experienced, whether the sample durations were represented by food-access or l i g h t , or whether a two- or three-choice procedure was used. In order to account for these findings, a "subjective shortening" model of memory for event duration was proposed. According to the model, the choose short e f f e c t i s produced by a discrepancy between a r e l a t i v e l y s t a t i c reference memory of the sample durations and a dynamic working memory of the sample durations that "shortens" over the retention i n t e r v a l . This discrepancy produces the tendency to respond as though the long sample was short, A number of predictions, derived from the subjective shortening model, were confirmed i n subsequent experiments. F i r s t , after a long retention i n t e r v a l , the point of subjec-t i v e equality between the short and long samples shifted to a longer duration. Second, stepwise increases i n the reten-t i o n i n t e r v a l produced a temporary choose short e f f e c t , whereas stepwise decreases i n the retention i n t e r v a l produced a temporary choose long e f f e c t . Third, with extended t r a i n i n g at a given retention i n t e r v a l , the choose short and choose long effects diminished and o v e r a l l accuracy improved. These results provided strong support for the subjective shortening model, whereas they could not be interpreted re a d i l y within the context of other conceptuali-zations of working memory processes. i v TABLE OF CONTENTS ABSTRACT . . . ...... . . . i i LIST OF FIGURES v i i ACKNOWLEDGEMENTS , • . • . ix INTRODUCTION 1 The concept of animal memory 1 Early views 1 Challenges to early views 5 Current trends i n animal memory research , 8 The concept of working memory 9 The relationship between working memory and reference memory . . 10 Procedures used to study, working memory . 10 Current views of working memory i n animals 12 Trace decay theory 12 Temporal discrimination hypothesis . . 13 Information processing theory 14 Coding views of working memory 15 Variables thought to a f f e c t working memory ., 16 Delay between sample and comparison stimuli ... 16 Exposure to the sample 17 I n t e r t r i a l i n t e r v a l 18 Interfering stimuli 18 V "Forget" cues, , , ............... 20 Stimulus serving .as the sample 21 Rationale and purpose of the present investigations , 23 PART I: A DEMONSTRATION OF THE "CHOOSE SHORT" PHENOMENON . 25 Experiment 1 25 Method . . 26 Results 28 Discussion ,. 34 Experiment 2 38 Method 38 Results 39 Discussion 45 Experiment 3 4 7 Method 47 Results. 49 Discussion 55 PART I I : TESTS OF THE SUBJECTIVE SHORTENING MODEL 56 Experiment 4 61 Method 62 Results , 64 Discussion , , 69 Experiment 5 70 v i Method 73 Results 77 Discussion 90 GENERAL DISCUSSION . . . . , 95 I Summary of the results and th e i r methodological implications ....... 95 II The subjective shortening model and i t s implications for theories of working memory i n animals ......... .. 98 REFERENCES , ., I l l v i i LIST OF FIGURES 1. Overall matching accuracy on food access and l i g h t t r i a l s as a function of delay for birds i n Experiment 1 ,,.,.«.,...................... 31 2. Percentage of correct choices a f t e r short and long duration samples as a function of delay for birds i n Experiment 1 ,. 33 3. Percentage of correct choices after short and long duration samples during baseline t r a i n i n g for birds i n Experiment 2 41 4. Percentage of correct choices after short and long duration samples as a function of delay for birds i n Experiment 2 , 44 '5. Percentage of t r i a l s on which the birds chose "short," "medium," and "long" af t e r each sample duration during the l a s t ten session of baseline t r a i n i n g i n Experiment 3 . 51 6. Percentage of "short," "medium," and "long" choices after each sample duration at the three delays for birds i n Experiment 3 53 7. Pro b a b i l i t y of a "long" choice as a function of sample duration at each of the three delays for birds i n Experiment 4 , . 66 v i i i 8. Point of subjective equality (PSE) as a function of delay for birds i n Experiment 4 ......... r .......... 68 9a. Percentage of correct choices after short and long sample durations for Birds 7 and 8 during each phase of Experiment 5 80 9b. Percentage of correct choices after short and long sample durations for Birds 9, 10, and 11 during each phase of Experiment 5 82 10. Mean percentage of correct choices after short and long sample durations during the f i r s t session aft e r each delay change i n Experiment 5 84 11. Overall accuracy during blocks of three consecutive sessions for each b i r d i n Experiment 5 88 12. Percentage of "short" choices during the three 0-sec sample tests for each b i r d i n Experiment 5 92 ix ACKNOWLEDGEMENTS I would l i k e to thank Dr. Don Wilkie for his support and guidance during the conduct of the research, and through-out my graduate t r a i n i n g , I also wish to thank Drs, R. Tees and R. Corteen for t h e i r helpful comments and suggestions. F i n a l l y , the assistance of R, Summers, C. Rayner, and D. T r e i t i s g r a t e f u l l y acknowledged. 1 INTRODUCTION The concept of animal memory It i s d i f f i c u l t to imagine how an animal without a capacity to remember past events could e a s i l y return to a source of food or water, fi n d i t s way to a nest or through a burrow system, avoid cues previously associated with dangerous or noxious events, or i n fact show any changes i n behaviour as a r e s u l t of i t s past experience. U n t i l recently, however, the study of animal memory has been systematically ignored by most behavioural s c i e n t i s t s . Early views. Two major factors may have contributed to the reluctance of early psychologists to study animal memory processes. F i r s t , the p o s s i b i l i t y that animals possess higher-order processing capacities, such as memory, had been viewed with skepticism since the time of A r i s t o t l e (cf. Ruggiero & Flagg, 19761. Early investigators therefore were reluctant to attribute any apparent retention a b i l i t i e s i n animals to memory processes. For example, Hunter (1913) studied the re-tentive a b i l i t i e s of ra t s , dogs, racoons, and children, using a delayed response procedure. Each subject was exposed to a set of three s p a t i a l locations, one of which was illuminated b r i e f l y . Following a delay period, the subject was permitted to choose one of the locations and was rewarded with food for choosing the location that previously had been illuminated. 2 Hunter found that rats responded correctly after a maximum delay of 10 sec, dogs after 5 min, racoons after 25 sec, and children a f t e r 50 sec or 25 min, depending upon t h e i r age. Hunter, however, was by no means ready to attribute these a b i l i t i e s to memory processes. He attributed the performance of the rats and dogs to th e i r use of physical orientation towards the correct location during the delay i n t e r v a l . A l -though overt orientation was not used by racoons, Hunter s t i l l did not attr i b u t e t h e i r performance to memory, but instead to "sensory thought." In a l l l i k e l i h o o d , "sensory thought" would not be viewed today as a more parsimonious explanation of retention a b i l i t i e s than "memory." Neverthe-l e s s , Hunter viewed "sensory thought" as a process intermediate to the high-level cognitive and memorial capacities of humans, and the low-level orienting capacities of dogs and ra t s . Following Hunter's lead, most researchers i n the 1920s and 30s viewed the delayed response paradigm mainly as a pro-cedure for comparing species; only a few investigators during thi s period used the delayed response procedure to investigate animal memory processes (e.g., Tinklepaugh, 1928). Accordingly, considerable research was directed toward categorizing species i n terms of whether or not they could bridge a delay without the use of bodily orientation, and toward ordering species i n terms of the maximum delay at which they could perform correct-l y . Both areas of research proved to be unproductive. F i r s t , 3 the results of several studies (e.g., Maier & Schneirla, 1935; Ladieu, 1944) did not support Hunter's conclusion that "lower animals" depend upon bodily orientation to bridge a retention i n t e r v a l ; even rats were found to delay responding success-f u l l y when orientation to the goal during the delay was e l i m i -nated. Second, attempts to order species i n terms of th e i r retention capacities yielded inconclusive r e s u l t s ; i t soon became clear that procedural variables were far more important i n determining the l i m i t s of retention than was the type of species studied (cf. Roberts & Grant, 1976). A second factor that served to impede the study of animal memory processes during the f i r s t half of the 20th century was that most of the major behavioural theorists (e.g., Guthrie, H u l l , Pavlov, Skinner, Spence, Tolman) showed l i t t l e i n t e r e s t i n memory processes. None of these figures conducted any research on memory, and.none included the construct of memory as a major part of t h e i r theories. Although Tolman's (1932) cognitive theory of behaviour included the construct of "expectancy" — a construct that may imply a form of represen-t a t i o n a l memory (cf. Ruggiero & Flagg, 1976) — Tolman himself did not include the construct of memory i n his theory of behaviour. Most learning theorists (e.g., Guthrie, H u l l , Pavlov, Spence) were interested primarily i n the acq u i s i t i o n of learned responses. B a s i c a l l y , they hypothesized that the 4 acq u i s i t i o n of a response was due to the development of associations, either between conditioned and unconditioned st i m u l i (e.g., Pavlov, 1927), or between st i m u l i and responses (e.g., H u l l , 1943). Performance of a learned response was thought to be determined by the strength of these associations and by the degree to which the stimulus conditions of o r i g i n a l t r a i n i n g had been reinstated. Furthermore, learning was often thought to be permanent unless interfered with by new learning (e.g., Guthrie, 1935). Memory therefore was viewed as an unnecessary hypothetical construct because the p r i n c i p l e s of learning could be used to account for performance after a retention i n t e r v a l (cf. Bolles, 1976). The results of experiments that were designed to test these learning theories.did not o f f e r a serious challenge to the view that memory was an unnecessary construct. In fact, the design of most learning experiments tended to obscure memory phenomena. For example, the interstimulus and i n t e r -t r i a l i n t e r v a l s used i n these experiments were generally short enough that "forgetting" rarely occurred. In the rare instan-ces i n which the design of the experiment permitted the eff e c t s of forgetting to occur ( i . e . , when there was a substantial i n t e r v a l between t r a i n i n g and testing) the fact that some retention occurred was emphasized; the fact that substantial forgetting occurred was ignored (cf, Bolles, 1976). Although animal memory was viewed by many early learning 5 theorists as an unimportant contruct, t h i s view could not be shared by those researchers interested i n the physiological basis of memory (e.g., Lashley, 1950). The nature of research on the physiological basis of memory demanded the use of animals as subjects, and thus the t a c i t assumption that animals had s i g n i f i c a n t memory capacities was a necessary part of the rationale for t h i s research. However, apart from making t h i s assumption, psychologists interested i n the b i o l o g i c a l basis of memory made l i t t l e attempt to formalize global theories of animal memory processes (.cf. Spear, 1978) . Challenges to the early views. Although learning theor-i s t s believed that learned associations persisted over time, they maintained that any substantial delay between the events to be associated would prevent learning (e.g., H u l l , 1943; Spence, 1947). Because learning sometimes did occur when the events to be associated were separated by a short delay, some of the early theorists (e.g., H u l l , 1943; Pavlov, 1927) postu-lated a minimal, memory-like process. Both Hull and Pavlov conceptualized t h i s process as.a trace or a f t e r - e f f e c t of the physical stimulus that persisted for a few seconds aft e r the of f s e t of the stimulus, Spence (1947) attributed any learning that occurred when there was a delay between a stimulus and a response to immediate secondary reinforcement. None of these learning theorists believed that associative learning involved any substantial memory mechanism. 6 Over the years, however, data gradually began to accumu-la t e that could not be interpreted e a s i l y i n the absence of a more substantial associative memory mechanism. For example, Petrinovich and Bolles (1957) found that rats could learn to alternate t h e i r responding for reward between two d i f f e r e n t s p a t i a l locations with i n t e r t r i a l i n t e r v a l s of up to 4 hours. They concluded that the rats were able to remember where they had gone on the previous t r i a l and then choose the opposite side. Capaldi (1967) also postulated an associative memory mechanism to account for the results of his reward alternation studies (e.g., Capaldi & Cogan, 1963). In these studies rats that alternately received reward and nonreward i n a runway developed appropriate alternations i n running speeds: they ran faster on rewarded t r i a l s than on nonrewarded t r i a l s even with i n t e r t r i a l i n t e r v a l s of up to 24 hours (Capaldi & Spivey, 1964). Capaldi concluded that a memory of reward or nonreward on the preceding t r i a l could act as a stimulus to control responding on the following t r i a l . Although Capaldi did not have a theory of memory processes per se, he was one of the f i r s t theorists to have the construct of animal memory as an ess e n t i a l component of a theory of learning. A s i g n i f i c a n t impetus for the postulation of an associa-t i v e memory mechanism i n animals developed out of the work of Garcia and his associates (e.g., Garcia, Ervin, & Koelling, 7 1966). They found that rats exposed to a novel taste and sub-sequently made sick by an i n j e c t i o n of a toxic substance, l a t e r showed an aversion to the novel taste. This "conditioned food aversion" was surprising i n view of the fact that i t could occur with delays of up to 24 hours between taste and sickness (Etscorn & Stephens, 1973), The t r a d i t i o n a l b e l i e f that learning required close temporal contiguity between the events to be associated was seriously challenged by these r e s u l t s . The most parsimonious explanation was that the rats must be able to remember novel tastes for long periods of time i n order to associate them with subsequent i l l n e s s (cf. Bolles, 1976). This interpretation was elaborated into a theory of associative memory by Revusky (.1971) . Another convincing example of long-delay associative learning was provided by Lett (1973; 19 75). Lett trained rats to go to one arm of a T-maze, but did not reward them for a correct choice i n the goal box. Instead, she removed the animals from the apparatus for i n t e r v a l s up to 1 hour, and then placed them back i n the s t a r t box where they received reward. The fact that rats were able to learn t h i s task implied an associative memory mechanism capable of bridging a long delay between the response and reward. P a r a l l e l with research into long-delay associative con-d i t i o n i n g was a growing i n t e r e s t i n the study of long-term retention of learned responses i n animals (e.g., Gleitman & 8 Holmes, 1967; Gleitman & Jung, 1963; Kamin, 1957; Gabriel, 1967). Some of thi s work was directed at assessing the empiri-c a l basis for the widespread assumption of the d u r a b i l i t y of learning i n animals, and with the interference theory of for -getting i n animals (cf. Gleitman, 1971) . In addition, other work i n thi s area was concerned with apparent nonmonotonic retention functions (e.g., the "Kamin" and "incubation" effects; cf. Spear, 197 8), and with the factors that produced these functions. This l i n e of research further served to promote int e r e s t i n animal memory processes (see Spear, 1978, for a detailed review of t h i s work). Resistance to the study of animal memory abated further during the early 19 70s as the popularity of t r a d i t i o n a l learning theories declined (Seligman, 1970) and in t e r e s t i n cognitive approaches to the study of animal behaviour increased (e.g., Bolles, 1975; Jarrard, 1971; Premack, 1971). These trends provided a more supportive climate for the systematic study of memory processes i n animals. Consequently, research on animal memory has accelerated dramatically during the past decade, and i s today a major f i e l d of study i n animal behaviour. Current trends i n animal memory research During the past decade, research on animal memory has followed three general l i n e s . One l i n e of research has been directed toward the study of long-term retention of acquired 9 responses. This retention of learned responses has been referred to alternately as long-term retention (Spear, 1978), long-term memory (cf. Spear, 1978), retentive memory (Revusky, 1971) , or reference memory (Honig, 1978). Other research has been concerned with long-delay learning; i . e . , where there i s a delay between the events to be associated. The memory process thought to be involved i n thi s type of learning has been c a l l e d associative memory (Honig, 1978; Revusky, 1971). A t h i r d l i n e of research has been concerned with the retention of s p e c i f i c events over a short delay when associative learning already i s complete. This kind of retention has been c a l l e d short-term retention (Riley, Cook, & Lamb, i n press), short-term memory (Roberts & Grant, 1976), or working memory (Honig, 1978) . In th i s thesis, the terminology of Honig (.1978) w i l l be used: the three types of memory described above w i l l be r e f e r -red to as reference memory, associative memory, and working memory. The remainder of thi s thesis w i l l be concerned primarily with the study of working memory. The concept of working memory Honig (1978) has described working memory i n terms of a group of procedures that involve delayed conditional d i s c r i m i -nations. Working memory i s thought to be required whenever an animal's discriminative responding i s based upon a conditional 10 stimulus that terminates before the response can be executed, and that varies from t r i a l to t r i a l - Consequently, i n order to respond c o r r e c t l y , the animal must remember which conditional stimulus had been present at the beginning of the t r i a l , and avoid "confusion" with the memory of sti m u l i present on pre-vious t r i a l s (Honig, 197 8). The rel a t i o n s h i p between working memory and reference  memory. According to Honig (1978), reference memory refers to the long-term maintenance of learned associations. Because most working memory tasks involve the ac q u i s i t i o n and mainte-nance of conditional discriminations i n reference memory, a stable reference memory i s necessary for stable performance on these tasks. Thus, as Honig (1978) has pointed out, i t i s important to be aware of the p o s s i b i l i t y that changes i n per-formance on a working memory task sometimes may r e f l e c t changes i n reference memory, as well as changes i n working memory. Procedures used to study working memory. There are a number of procedures that can be used to study working memory i n animals, including the delayed response task (Hunter, 1913), reward substitution tasks (Tinklepaugh, 19 28), delayed alterna-t i o n tasks (Capaldi, 1967), r a d i a l arm maze tasks (Olton, 1978) , advance key procedures (Honig, 1978), and the delayed paired comparisons task (Shimp & M o f f i t t , 1977). Recently, the delayed matching to sample task and i t s variations have found wide use among researchers interested i n working memory. 11 These delayed matching procedures have advantages over some of the e a r l i e r procedures (e.g., Hunter, 1913) i n that they f a c i l i t a t e precise control over experimental s t i m u l i , and ea s i l y eliminate the p o s s i b i l i t y of simple orientation towards the location of the correct stimulus during the delay. In the delayed matching to sample (DMTS) task (e.g., Blough, 1959) t r i a l s consist of three components: presentation of a sample stimulus; a delay ( i . e . , retention) i n t e r v a l ; and presentation of two or more comparison s t i m u l i , one of which phy s i c a l l y matches the sample. Choice of the comparison that matches the sample results i n reinforcement. The stimuli used as the samples are alternated randomly over t r i a l s . Consequently, i n order to choose the correct comparison stimulus, the animal must remember which sample stimulus had been presented at the beginning of the t r i a l . Delayed oddity from sample tasks (e.g., Lydersen, Perkins, & Chairez, 1977) are i d e n t i c a l to DMTS tasks except that the subject i s rewarded for choosing the comparison that does not match the sample. Another v a r i a t i o n of the DMTS task i s c a l l e d delayed symbolic matching to sample (DSMTS; e.g., Wilkie, 1978). This task d i f f e r s from the DMTS and the delayed oddity tasks i n that none of the comparison stimuli i s i d e n t i c a l to the sample s t i m u l i ; the rel a t i o n s h i p between the sample and the correct comparison i s ar b i t r a r y . F i n a l l y , i n successive matching to sample tasks (e.g., Nelson & Wasserman, 1978)., presentation of the sample i s followed after a delay by 12 a single test stimulus that either does or does not match the sample. Reinforcement i s available only when the test stimulus matches the previously presented sample. These tasks hereafter w i l l be referred to c o l l e c t i v e l y as delayed matching tasks. Current views of working memory i n animals There are a number of questions that might be asked about the process of working memory i n animals. For example, how i s information stored? How much information can be stored? What i s the form or content of the memory? How i s i t organized? Is information maintained a c t i v e l y or passively over a delay? If i t i s maintained passively, how i s i t retrieved? What are the sources of forgetting? In recent years, a number of views or hypotheses have been proposed that address some of these issues. Before describ-ing each of these views, i t i s important to note that these views do not always address a l l of the possible questions that could be asked about working memory. Because these views often focus upon d i f f e r e n t aspects of working memory, they cannot always be contrasted e a s i l y . Trace decay theory. Roberts and Grant (Roberts, 19 72; Roberts & Grant, 1976) have developed a theory of working memory based on the concept of a stimulus trace. These i n v e s t i -gators hypothesized that presentation of a sample stimulus 13 generates an in t e r n a l trace of the sample that i s strengthened gradually during the exposure to the sample, and that decays gradually i n i t s absence. Thus, the storage of sample informa-ti o n i s a simple function of exposure time: longer exposure times increase the strength of the memory trace. What i s stored ( i . e . , the "content" of the memory) i s a representation that i s isomorphic with the sample. Loss of the stored repre-sentation i s thought to correspond either to a passive decay of the i n i t i a l trace, or to interference from competing traces (Roberts & Grant, 1976). This trace decay theory subsequently has been modified to include an active rehearsal process that serves to maintain the sample trace during the delay (Grant & Roberts, 1976). According to t h i s revised trace theory, loss of stored information could be produced by interference with t h i s active rehearsal process, as well as by simple decay of the trace, or interference from competing traces. Temporal discrimination hypothesis. D'Amato (.1973) and D'Amato and Worsham (.1974) have developed an hypothesis of working memory that i s based on a temporal discrimination process, rather than on a storage process. Using the DMTS task as an i l l u s t r a t i o n , D'Amato conceptualized the problem for the subject as involving a decision about which of the compari-son stimuli has most recently served as the sample. In t h i s conceptualization, the "storage" and "content" of the memory are reduced to a process of temporal discrimination, and 14 "forgetting" i s attributed to discrimination f a i l u r e rather than to a loss of stored information. For example, the dele-terious e f f e c t s of long delays ( i . e . , retention intervals) on matching accuracy are thought to occur because a discrimination of which of the comparison stimuli was l a s t seen as the sample i s more d i f f i c u l t . Although D'Amato's basic view i s that the "storage" of memories can be reduced to a process of discrimination, he does invoke an "internal representation" process i n order to account for delayed symbolic matching. The addition of t h i s l a t t e r process was necessary because the comparison stimuli i n sym-b o l i c matching are never seen as samples, and thus i t would be impossible to solve t h i s task on the basis of t h e i r r e l a t i v e recency. Therefore, the c r i t i c a l discrimination i n symbolic matching tasks i s between the r e l a t i v e recency of "internal representations" of the sample set (D'Amato, 1973). Information processing theory. Wagner (Wagner, 1978; Wagner, Rudy, & Whitlow, 1973) has proposed a model of animal memory that i s based i n part upon Atkinson and S h i f f r i n ' s (1968) two-process theory of human memory. B a s i c a l l y , Wagner's model assumes that there i s a short-term storage buffer i n which a limited amount of information can be maintained through "active rehearsal," According to t h i s model, working memory i s l i m i t e d by a number of factors, including the size of the short-term storage buffer, and the capacity of the rehearsal 15 mechanism. As new items enter the short-term store, they may displace old ones from the store, and rehearsal of one item may decrease the a b i l i t y to rehearse another item. The "content" of memory i s conceptualized as an activated repre-sentation of the stimulus i n the short-term store; "forgetting" i s an i n a c t i v a t i o n of the stored representation due to either a passive decay process, f a i l u r e of rehearsal, or displacement by a new item entering the short-term store. Coding views of working memory. Recently, a number of "coding" processes have been invoked to account for working memory phenomena (Carter & Werner, 1978; Farthing, Wagner, Gilmour, & Waxman, 1977; Honig & Wasserman, 1981; Riley et a l . , i n press; Roitblat, 1980). "Codes" have been conceptualized variously as a stimulus trace (Farthing et a l . , 1977; Riley et a l . , i n press), an abstract representation of the sample stimulus (Farthing et a l . , 1977), a representation of the test stimulus (Roitblat,. 19.80), a covert or overt mediating response (Carter & Werner, 1978), a response decision, or in s t r u c t i o n (Honig, 1978; Honig & Wasserman, 1981; Riley et a l . , i n press), a representation of anticipated reinforcement (Spetch, Wilkie, & Skelton, 1981), or, i n fa c t , as any "transformation of the sample stimulus that allows the organism to function appropri-ately at the time of the t e s t " (Riley et a l . , i n press, p. 7). As can be seen, the coding hypotheses as a group do not make up an i n t e r n a l l y consistent "theory" of working memory. 16 Individually, however, d i f f e r e n t coding hypotheses represent d i s t i n c t conceptualizations of the content or form of memory. Variables thought to a f f e c t working memory There are a number of factors that a f f e c t performance on delayed matching tasks. In general, most of these factors have been assumed to af f e c t working memory (Roberts & Grant, 1976; Farthing et a l . , 1977; Roitblat, 198.0; Cook, 1980; but see Wilkie & Spetch, 1978; i n press). These factors and th e i r e f f e cts on matching accuracy w i l l be b r i e f l y described i n the following sections, along with a b r i e f account of how these eff e c t s have been interpreted within the context of current theories of working memory. Delay between sample and comparison s t i m u l i . One widely studied variable i n delayed matching tasks i s the duration of the delay, or retention i n t e r v a l , between the sample and com-parison stimuli.- Generally, matching accuracy has been found to decrease systematically with increases i n the delay, and to approach chance l e v e l within a few seconds (e.g., Blough, 1959; Roberts & Grant, 1976), although some instances of above-chance matching accuracy with delays of 25 sec . (wilkie,. 1978), or even 60 sec (Grant, 1976) have been reported. According to the trace decay model of working memory (Roberts & Grant, 1976), the decline i n matching accuracy over the delay i s due to a simple decay process, which i s thought 17 to occur as a negatively accelerated function of time since the sample o f f s e t . Wagner's (;1978) information processing model also postulates a trace decay mechanism which can account for the e f f e c t of delays. An alternative interpretation of the e f f e c t of delay i s provided by the temporal discrimination hypothesis (D'Amato, 1973). According to t h i s view, performance a f t e r a delay depends upon the animals' a b i l i t y to make a temporal discrimina-tion between the r e l a t i v e recency of the samples. As the delay i n t e r v a l increases, the r a t i o describing the r e l a t i v e recency of the samples decreases, which leads to poorer temporal d i s -crimination. A s i m i l a r d i s c r i m i n a t i o n - l i k e process has been proposed within the context of cer t a i n versions of the coding view to account for decreased matching accuracy at longer delays. Roitblat (1980), for example, suggested that pigeons maintain a coded representation of the correct choice stimulus during the delay, and that as the delay increases, t h i s representation increasingly becomes "confused" with other choice s t i m u l i . Exposure to the sample. Delayed matching by pigeons improves with increases i n presentation time of the sample (e.g., Nelson & Wasserman, 1978; Roberts & Grant, 1974; 1976) and with increases i n the number of responses required to terminate the sample (e.g., Roberts & Grant, 1976; Wilkie & Spetch, 1978) . 18 According to the trace decay theory, increases i n the duration of exposure to the sample are thought to strengthen the memory trace and thereby improve matching accuracy (e.g., Roberts & Grant, 1976). The e f f e c t of sample duration can also be interpreted i n terms of coding processes. For example, a number of investigators (e.g., Farthing et a l . , 1977; R o i t b l a i t , 1980) have suggested that a coded representation of the sample i s formed gradually during exposures to the sample. Increases i n the sample duration improve the l i k e l i h o o d that t h i s coded representation w i l l be developed f u l l y , thereby increasing the pr o b a b i l i t y of an accurate choice. I n t e r t r i a l i n t e r v a l . Delayed matching accuracy improves with increases i n the i n t e r t r i a l i n t e r v a l (Jarrard & Moise, 1971; Herman, 1975; Maki, Moe, & Bierley, 1977; Nelson & Wasserman, 1978). Lengthening the i n t e r t r i a l i n t e r v a l improves performance, according to the trace decay theory, because i t allows more time for the decay of competing stimulus traces from the previous t r i a l s (Roberts & Grant, 1976). The f a c i l i t a . t i v e e f f e c t of increases i n the i n t e r t r i a l i n t e r v a l i s also consis-tent with the temporal discrimination hypothesis (D'Amato, 1973). Temporal discrimination i s f a c i l i t a t e d with longer i n t e r t r i a l i n t e r v a l s because the r e l a t i v e recency r a t i o of the samples i s increased. Interfering s t i m u l i . Delayed matching performance i s 19 adversely affected by the presentation of certain extraneous " i n t e r f e r i n g " stimuli (e.g., ambient illumination) during the delay i n t e r v a l (e.g., Cook, 1980; Roberts & Grant, 1976; Wilkie, Summers, & Spetch, 1981), Cook (1980) has demonstrated that the degree of interference produced by extraneous delay i n t e r v a l s t i m u l i i s related d i r e c t l y to the degree of stimulus change, and Wilkie et a l . (19 81)-. have shown that the degree to which stimuli disrupt delayed matching accuracy i s correlated p o s i t i v e l y with the degree to which these s t i m u l i disrupt simple operant responding. Most views of working memory can account for the deleter-ious e f f e c t s of some extraneous delay i n t e r v a l s t i m u l i . For example, views that postulate active rehearsal mechanism (e.g., Grant & Roberts, 1976; Farthing et a l . , 1977; Roitblat, 1980; Wagner, 1978) can account for the i n t e r f e r i n g effects of extraneous delay i n t e r v a l stimuli by assuming that these stimuli disrupt the rehearsal process. A l t e r n a t i v e l y , the simple trace decay theory (Roberts & Grant, 1976) dealt with i n t e r -ference e f f e c t s by assuming that the delay i n t e r v a l stimuli generate a memory trace that competes or interferes with the trace of the sample. D'Amato (1973) interpreted the deleterious e f f e c t of delay i n t e r v a l illumination on delayed matching per-formance i n terms of his temporal discrimination hypothesis by assuming that "visual events w i l l appear more recent after an i n t e r v a l spent i n darkness than aft e r a l i k e period f i l l e d with 20 a myriad of v i s u a l perceptions." (D'Amato, 1973, p. 263). "Forget" cues. Several recent studies (Grant, 1981; Maki & Hegvik, 1980; R i l l i n g , Kendrick, & Stonebraker, i n press) have examined the e f f e c t on matching performance of "forget" and "remember" cues presented during the delay i n t e r v a l . In these studies, birds were trained under a delayed matching task i n which the presentation of a "remember" cue during the delay sig n a l l e d that the regular retention test ( i . e . , presen-tation of the comparison stimuli) would occur at the end of the delay, and the presentation of a "forget" cue indicated that the retention test would be cancelled on that t r i a l . To test for the e f f e c t of the forget cues, "surprise" retention tests were conducted on some forget-cue t r i a l s ; matching accur-acy on these t r i a l s has been found consistently to be lower than on remember-cue or no-cue t r i a l s . The location of the forget cue within the delay i n t e r v a l also i s important; the f o r -get cue has a greater detrimental e f f e c t upon matching accuracy i f i t occurs early i n the delay than i f i t occurs l a t e r i n the delay i n t e r v a l (Grant, 1981; R i l l i n g et a l . , i n press). The deleterious effects of forget cues on matching accuracy are inconsistent with a simple trace decay process (Roberts & Grant, 1976), i n which information i s l o s t through a passive decay of the trace, or an interference from competing traces. There i s no obvious reason to assume that forget cues should a f f e c t the decay process, and because the stimuli serving 21 as forget and remember cues generally are counterbalanced, the trace of the forget cues should not compete with a trace of the sample stimulus to any greater extent than would the trace of the remember cue. Grant (1981) has interpreted the e f f e c t of forget cues i n terms of active rehearsal processes. He suggested that the forget cue decreases matching accuracy because i t results i n a cessation of rehearsal of the sample memory. In contrast, R i l l i n g et a l . (in press) have interpreted the effects of forget cues i n terms of a "behavioral-context hypothesis." According to t h i s hypothesis, the animal's behaviour during the delay i n t e r v a l produces stimuli that are necessary for r e t r i e v a l of sample representations during the choice period. They suggested that forget cues disrupt delay i n t e r v a l behaviour and thereby eliminate part of the stimulus context needed to ret r i e v e the sample representation. Stimulus serving as the sample. The stimulus used as the sample also can a f f e c t performance on working memory tasks. For example, Farthing et a l . (1977) found differences between pigeons' memory for colored samples and l i n e - t i l t samples: the "forgetting curves" for the l i n e - t i l t s were steeper and were less affected by sample duration than were the curves for colored samples. In the same vein, Wallace, Steinert, Scobie, and Spear (1980) found that rats remembered an auditory sample better than a v i s u a l sample after delay i n t e r v a l s of 2 and 4 22 sec, even though t h e i r delayed matching performance with these two types of samples was equivalent at the 0-sec delay. These sorts of findings have s i g n i f i c a n t implications for current views of working memory processes. For example, in order for the simple trace decay theory to encompass d i f f e r -ences i n the "forgetting" functions with d i f f e r e n t types of st i m u l i i t would be necessary to make an additional assumption that traces for d i f f e r e n t stimuli decay at d i f f e r e n t rates (cf. Farthing et a l . , 1977). S i m i l a r l y , i n order for coding views to encompass these e f f e c t s , i t must be assumed that "encoding i s more e f f i c i e n t for some types of stimuli than for others" (Farthing et a l . , 1977, p. 528), or that the rehearsal of some codes i s more e f f e c t i v e than the rehearsal of others. The d i f f e r e n t i a l retention of d i f f e r e n t sample stimuli i s also problematic for D'Amato1s temporal discrimination hypothesis. If memory of a sample i s based simply upon a discrimination of i t s r e l a t i v e recency, then there i s no reason to expect that one type of sample stimulus would be remembered better than another. Wagner's information processing model includes an assump-ti o n that some sti m u l i command rehearsal more than others. Evidence consistent with t h i s notion i s provided by demonstra-tions that "surprising" stimuli are remembered better than "expected" stimuli (Terry & Wagner, 1975; Maki, 1979; but see 23 C o l w i l l & Dickinson, 1980). An extention of thi s view could be that other properties of sti m u l i may have a similar e f f e c t upon rehearsal, and thi s could account for the differences i n retention found between d i f f e r e n t samples. Rationale and purpose of the present investigations Clearly, the stimulus or stimulus dimension used i n working memory tasks may have a s i g n i f i c a n t e f f e c t on animals' performance, a finding that has important implications for current views of working memory processes. In the human psychophysical l i t e r a t u r e , a d i s t i n c t i o n has been made between quantitative, or "prothetic" stimulus dimensions, and q u a l i t a -t i v e , or "metathetic" dimensions (Coren, Porac, & Ward, 1979). For prothetic dimensions, changes i n the physical stimulus (e.g., weight) are perceived as a::change in the apparent quantity of the stimulus (e.g., heaviness), whereas for metathetic dimensions, changes i n the physical stimulus (e.g., wavelength) are perceived as a change i n the apparent qu a l i t y of the stimulus (e.g., c o l o r ) . Certain psychophysical laws seem to hold for one of these types of stimulus dimension but not for the other CScharf, 1975). It i s possible that the d i s t i n c t i o n between dimensions that have been c l a s s i f i e d as metathetic and those that have been c l a s s i f i e d as prothetic on the basis of human psychophysical studies also may have significance for the study of working 24 memory i n animals. Almost a l l studies of working memory i n animals have used stimuli that vary along dimensions that are c l a s s i f i e d by humans as metathetic: e.g., d i f f e r e n t colors (Wilkie & Spetch, 1978), l i n e t i l t s (Farthing et a l . , 1977), shapes (Cohen, C a l i s t o , & Lentz, 1981), responses (Maki, Moe, & Bierley, 1977), s p a t i a l locations (Wilkie & Summers, i n press), food or no food (Wilkie, 1978), and f l i c k e r i n g or steady l i g h t (Blough, 1959). L i t t l e i s known about animals' working memory for sti m u l i that are varied along dimensions that are c l a s s i f i e d as prothetic, such as duration. Although numerous studies have been concerned with animals' discrimination of event duration (e.g., Church, Getty, & Lerner, 1976; Elsmore, 1972; Kinchla, 1970; Pe r i k e l , Richelle, & Maurissen, 1974; Reynolds & Catania, 1962; Spetch & Wilkie, 1981; Stubbs, 1968), at the time the present research was i n i t i a t e d there were no published studies of working memory for event duration i n animals. In view of the fact that animals' performance on working memory tasks may depend upon the stimulus used as the sample (Farthing et a l . , 1977), and that working memory for sti m u l i that vary along "prothetic" dimensions had not been studied, a systematic investigation of pigeons' memory for event duration was warranted. 25 PART I: A DEMONSTRATION OF THE "CHOOSE SHORT" PHENOMENON Experiment 1 This experiment was designed to investigate pigeons' working, memory for short and long duration samples over various delay i n t e r v a l s . The procedure used i n t h i s experiment was one in which pigeons were trained to peck a red comparison stimulus following short (2-sec) samples, and a green comparison stimu-lus following long (10-sec) samples. On half the t r i a l s i n each session, the sample consisted of timed access to food, whereas on the remaining t r i a l s the sample was l i g h t presenta-t i o n . The delay between the sample and comparison stimuli was varied within the range of 0 to 20 sec during each session. The use of both food-access and l i g h t samples was i n -cluded i n the present study for two reasons. F i r s t , i t pro-vided a means of assessing the generality of working memory for event duration across these two types of samples. Second, i t would extend previous research (Spetch & Wilkie, 1981) which had shown that pigeons are more accurate i n discriminating the duration of food access than the duration of l i g h t at a 0-sec delay. Because most variables that a f f e c t delayed match-ing performance at a 0-sec delay show a similar e f f e c t at longer delays (cf. Roberts & Grant, 1976; Wilkie & Spetch, 1978) , i t was expected that t h i s superior performance with food access as 26 the sample would be maintained at the longer delay i n t e r v a l s . Method Subjects Three adult S i l v e r King pigeons (Birds 1, 2, and 3) served as the subjects. Each was experienced at discriminating stimulus durations, but none had been exposed previously to delay i n t e r v a l manipulations. Each b i r d was deprived of food u n t i l they were approximately 80% of t h e i r free-feeding weight and maintained at t h i s weight by mixed grain obtained during and after d a i l y experimental sessions. The birds were housed i n large i n d i v i d u a l home cages i n which water and health g r i t were continuously available. Apparatus A BRS-Foringer Model #PS-004 pigeon chamber was used. One wall contained a horizontal row of three p l a s t i c pecking keys; each required a force of .2 N to operate. An Industrial Electronics Engineers' Series 10 stimulus projector was mounted behind each key; these illuminated the keys with a uniform f i e l d of colored l i g h t . A Gerbrands Model #G5610 solenoid-operated feeder that permitted timed access to mixed grain was centered below the keys. Grain presentations were illuminated by a 2.8 W lamp located within the feeder. The houselight consisted of two 2.8 W lamps mounted behind a transparent p l a s t i c r e f l e c t o r above the pecking keys; these lamps provided 27 a di f f u s e illumination of the chamber. Control of experimental conditions and c o l l e c t i o n of data i n th i s and a l l subsequent experiments were performed by a Data General Nova 3 computer. Procedure Baseline Procedure. A v a r i a t i o n of the delayed symbolic matching to sample (DSMTS) paradigm was used. T r i a l s began with the presentation of a sample stimulus for either a short (2 sec) ; or a long (10 sec). duration. Food access (presen-tati o n of the raised illuminated g r a i n - f i l l e d feeder) served as the sample stimulus on half of the t r i a l s ; on the remaining t r i a l s the sample consisted of illumination of the houselight. Sample o f f s e t was followed by illumination of the side pecking keys with a green and a red comparison stimulus; the po s i t i o n of red and green was varied randomly across t r i a l s . Correct choices (a peck at the red comparison key after short samples, or at the green key aft e r long samples) produced a 5-sec .: u". grain r e i n f o r c e r , followed by a 30-sec i n t e r t r i a l i n t e r v a l (ITI). Incorrect choices (red after long samples or green after short samples) terminated both comparison stimuli and i n i t i a t e d the ITI. Each subject was trained under th i s procedure with a 0-second delay between the sample and comparison stimuli u n t i l matching accuracy appeared stable and asymptotic. The ac q u i s i -t i o n data from t h i s phase of the experiment are reported i n a 28 previous study (Spetch & Wilkie, 1981). For the l a s t ten days of t h i s phase, matching accuracy was well above chance on both types of t r i a l s but was higher on food-sample t r i a l s than on light-sample t r i a l s for a l l birds. Delay Manipulations. Variations i n the length of the delay between sample o f f s e t and comparison st i m u l i onset ( i . e . , the retention interval) were conducted within sessions. Three series of delays were arranged. The f i r s t series consisted of delays of 0, 1, 2, 3, 4, and 5 sec; the second series consisted of delays of 0, 5, and 10 sec, and the t h i r d series consisted of delays of 0, 5, and 20 seconds, During each se r i e s , the 0-sec delay occurred on half of the t r i a l s of each session; the longer delays occurred on the remaining t r i a l s with equal p r o b a b i l i t i e s . Each b i r d was tested for 60 sessions under Series 1 and for 40 sessions under each of Series 2 and 3. Results Figure 1 shows the birds' o v e r a l l matching accuracy with food-access and l i g h t samples at each delay. Data points for the 0- and 5-sec delays were averaged over the three series of delays. Consistent with the r e s u l t s of previous studies CSpetch & Wilkie, 1981), o v e r a l l matching accuracy with food access samples was better than with l i g h t samples at the 0-sec delay. However, i t i s clear from Figure 1 t h a t the 29 superior matching accuracy with food as the sample was not maintained over any of the longer delay i n t e r v a l s . These results were confirmed by a p r i o r i orthogonal: comparisons between average matching accuracy on food and l i g h t t r i a l s at each of the eight delay i n t e r v a l s {0 sec: t (.2) =15.67, p<.05; 1 sec: t (2 )=1 .34 , p>.05; 2 sec: t (2 )= .97 , p>.05; 3 sec: t(.2) = . 45 , p>.05; 4 sec: t(.2)=1.75, p>.05; 5 sec: t(.2) = . 95 , p>.05; 10 sec: t (2 )=1 .21 , p>.05; 20 sec: t (2) =.75, p>.05] . Figure 2 shows each bird's matching accuracy aft e r short and long samples. At the short delays, the percentage of correct choices was approximately equal after both short and long samples. However at the longer delays accuracy after short samples was greater than accuracy after long samples. These effects were assessed by a four-way repeated measures analysis of variance on choice accuracy, with the factors being delay, sample type, sample duration, and subjects. This analysis revealed that there was a s i g n i f i c a n t main e f f e c t of delay IF(7 ,14)=12.4 , p<.05] , sample duration IF(1,2)=34.64, p<.05J, and subjects r[F ( 2 , 1 4 ) = 7 0 . 2 4 , p<.05] . In addition, there were s i g n i f i c a n t two-way interactions between delay and sample duration [F(7,14)=4.29, p<.05] , delay and subjects IF(14,14)=3.48, p<.05] , and sample type and subjects IF(2,14)= 6 . 1 1 , p<.05J. F i n a l l y , there were s i g n i f i c a n t three-way i n t e r -actions between delay, sample type, and subjects IF (.14 ,14) =2 . 51 , p<.05] , and between sample type, sample duration, and subjects 30 Figure 1. Overall matching accuracy (percent correct) on food-access and l i g h t t r i a l s at each of the eight delays. Figure 2. Percent correct after short and long samples at each of the eight delays, with food access as the samples ( l e f t panels) and l i g h t as the samples (right panels). BIRD 1 SHORT LONG BIRD 2 BIRD 3 i i i i i i 012345 10 LIGHT T—I—I—I—I—I T 012345 10 DELAY(SEC) 34 [F( 2 ,14)=10.3 , p<.05]. No other effects were s i g n i f i c a n t . The effects of delay and sample duration were analyzed further by a p o s t e r i o r i comparisons (Newman-Keuls, p= .05). These comparisons revealed that accuracy after short samples was s i g n i f i c a n t l y greater than af t e r long samples at the 10-and 20-sec delays, but not at shorter delays. In addition, o v e r a l l matching accuracy was s i g n i f i c a n t l y greater at 0-sec, 1-sec, and 2-sec delays than at either 5-sec, 10-sec, or 20-sec delays, and accuracy af t e r 5- and 10-sec delays was s i g n i f i c a n t l y greater than accuracy after the 20-sec delay. Thus, changes i n the delay i n t e r v a l had three effects on the birds' performance. F i r s t , with delays greater than 0 sec, the s i g n i f i c a n t difference between matching accuracy with food and l i g h t samples disappeared. Second, increases i n the delay led to a decrease i n o v e r a l l matching accuracy. Third, with long delays, the birds showed a strong tendency to choose the comparison stimulus associated with the short samples, as indicated by the s i g n i f i c a n t l y greater accuracy after short than after long samples. This increased tendency to choose the comparison associated with the short sample hereafter w i l l be c a l l e d the "choose short" e f f e c t . Discussion Consistent with the results of previous studies of delayed 35 matching (e.g., Roberts & Grant, 197 6), there was a systematic decrease i n o v e r a l l matching accuracy as a function of the delay for both food- and light-duration samples. However, the expec-tati o n that food durations would be remembered over longer delays than l i g h t durations was not confirmed. Performance with food samples was not better than performance with l i g h t samples at any of the delays except 0-sec: This i s s u f p r i s - . ing i n view of the fact that most variables that improve matching at a 0-sec delay also improve performance at longer delays (e.g., Roberts & Grant, 1976; Wilkie & Spetch, 1978). One inte r p r e t a t i o n of t h i s r e s u l t i s that food durations are not more memorable for pigeons than l i g h t durations i n spite of the fa c t that they are more discriminable. A l t e r n a t i v e l y , food durations might i n fact be more memorable than l i g h t durations, but the increased tendency for birds to choose the comparison associated with the short sample at long delays ( i . e . , the "choose short" effect) may have decreased accuracy for both, thereby masking the differences i n memorability. The choose short e f f e c t found i n Experiment 1 was i n t e r -esting p a r t i c u l a r l y i n view of the f a c t that no comparable re s u l t s have been reported i n other delayed matching studies that have used "metathetic" stimulus dimensions such as l i n e t i l t s and colors. These re s u l t s suggest that duration might be processed and retained d i f f e r e n t l y than other stimulus dimensions. 36 The choose short e f f e c t does not appear to be interpreted e a s i l y within the context of most current views of working memory. For example, according to both the trace decay theory and c e r t a i n coding views of working memory, accuracy after long samples should be better than accuracy a f t e r short samples. Accuracy should be better af t e r long samples, according to c e r t a i n coding views, because longer duration s t i m u l i are more completely coded than shorter duration s t i m u l i (e.g., Roitblat, 1980). Better accuracy af t e r long samples also would be predicted on the basis of the trace decay theory because longer sample st i m u l i should generate a stronger trace than shorter duration s t i m u l i (Roberts & Grant, 1976). Clearly, neither of these views can accomodate e a s i l y the observation that accuracy after short samples was better than accuracy af t e r long samples at long delays. An alternative interpretation of the "choose short" e f f e c t , which w i l l be elaborated upon i n Part II, i s that animals base t h e i r responses on a subjective representation of the sample duration i n working memory that shortens as a function of the delay i n t e r v a l . At a 0-sec. delay, the animals subjective representation of the sample duration s t i l l would be close to the actual sample durations, leading to equally accurate performance a f t e r short and long samples. However, at long delays, the animals' representation of the duration of the long sample would have shortened, and thus would be more 37 similar to the actual duration of the short sample. This would lead to more incorrect choices of the comparison associated with the short sample. Although the subjective duration of the short sample would also become shorter after a long delay, i t would be less s i m i l a r to the actual duration of the long sample, and thus incorrect choices to the comparison associated with the jlong. sample should not be as prevalent as incorrect choices to the comparison associated with the short sample. In view of the potential importance of the choose short e f f e c t for current formulations of animal memory processes, and the paucity of research on animals' working memory for event durations, further systematic investigation of the choose short e f f e c t was warranted. 38 Experiment 2 This experiment was designed to es t a b l i s h the generality of the choose short e f f e c t using naive subjects. In t h i s experiment, pigeons were trained under a procedure s i m i l a r to that used i n Experiment 1, except that each b i r d received only one type of sample stimulus ( l i g h t for two birds, food access for one b i r d ) , and the comparison stimuli associated with the short and long samples d i f f e r e d for the three birds. Accuracy afte r short and long samples then was compared at delays of 0, 5, and 2 0 sec. Method Subjects The subjects were three naive S i l v e r King pigeons (Birds 4, 5, and 6). Deprivation and housing conditions were the same as described i n Experiment 1. Apparatus The apparatus was the same as that described i n Experiment 1 except that for Birds 4 and 5, the side pecking keys were illuminated with yellow and blue l i g h t as the comparison s t i m u l i . Procedure During preliminary sessions, the birds were trained to eat from the raised illuminated grain feeder, and then trained 39 to peck the side pecking keys when illuminated with blue or yellow l i g h t (Birds 4 and 5), or with red or green l i g h t (Bird 6). Next, the birds were trained for several sessions under the baseline (0-sec delay) DSMTS procedure. A l l aspects of t h i s procedure were the same as described i n Experiment 1 except that houselight durations served as the sample stimuli on a l l t r i a l s for Birds 4 and 5, and feeder durations served as the sample st i m u l i on a l l t r i a l s for Bird 6. In addition, blue and yellow l i g h t s served as the comparison st i m u l i for Birds 4 and 5, and t h e i r designation as correct after short and long samples was counterbalanced over these two birds. Red and green l i g h t s served as the comparison st i m u l i for Bird 6. Following baseline t r a i n i n g , the birds received several sessions i n which a variable delay procedure was i n e f f e c t , with 0-sec delays occurring on a random half of the t r i a l s , and 5- and 20-sec delays occurring equally often On the remaining t r i a l s . Results Acquisition Figure 3 shows the birds' matching accuracy after short and long samples during baseline t r a i n i n g . Bird 4 was more accurate after short samples, whereas Bird 5 was s l i g h t l y more accurate a f t e r long samples. For Bird 6, accuracy after short gure 3. Percent correct a f t e r short and long duration samples during blocks of consecutive sessions of baseline t r a i n i n g . B L O C K S O F 10 S E S S I O N S o o sho r t e • 1 ong 6 T 1 1 ' 1 1 1 2 3 4 5 B L O C K S O F 5 S E S S I O N S 42 and long samples was very s i m i l a r . Thus there was no systematic tendency for the birds to choose either one of the comparison stimuli during a c q u i s i t i o n . Delay manipulations Figure 4 shows matching accuracy aft e r short and long samples at the three delays. These data are displayed as two rep l i c a t i o n s , which correspond to the f i r s t and the second half of the t o t a l number of sessions i n which the variable delay procedure was i n e f f e c t . During both r e p l i c a t i o n s the birds showed a strong tendency to choose the comparison associ-ated with short samples at the 20-sec delay. This choose short e f f e c t was not as pronounced at the 5-sec delay, and was not apparent at the 0-sec delay. F i n a l l y , o v e r a l l accuracy decreased as the delay i n t e r v a l was increased. These ef f e c t s were assessed using a four-way repeated measures analysis of variance on choice accuracy, the factors being delay, sample duration, r e p l i c a t i o n s , and subjects. This analysis yielded s i g n i f i c a n t main ef f e c t s of delay [F(2,4)= 15.99, p<.05], sample duration [F (1,2)=90.11, p<.05], and subjects [F(2,4)=15.49, p<.05]. In addition, there was a s i g -n i f i c a n t two-way int e r a c t i o n between delay and sample duration [F(2,4)=18.18, p<.05], and between delay and subjects [F(4,4)= 16.85, p<.05], as well as a s i g n i f i c a n t three-way in t e r a c t i o n between delay, sample duration, and subjects [F(4,4)=7.58, p<.05]. No other effects were s i g n i f i c a n t . gure 4. Percent correct after short and long duration samples at the three delays during the f i r s t half (solid lines) and the second half (dashed lines) of the delay manipul ti o n phase. / 45 The effects of delay and sample duration were analyzed further by a p o s t e r i o r i pairwise comparisons (Newman-Keuls, p=.05) . These analyses showed that accuracy was greater af t e r short samples than after long samples at the 20-sec delay, but not at the 0-sec or 5-sec delay. Moreover, th i s e f f e c t was s i g n i f i c a n t for each b i r d , and during both r e p l i c a t i o n s . F i n a l l y , o v e r a l l accuracy was s i g n i f i c a n t l y higher at the 0-sec delay than at either the 5-sec or 20-sec delay. Thus, each b i r d showed a choose short e f f e c t at the 20-sec delay, and the e f f e c t remained stable across test sessions. Discussion These results r e p l i c a t e and extend those of Experiment 1 i n four ways. F i r s t , they demonstrate that the choose short e f f e c t i s a r e l i a b l e phenomenon, and that i t also occurs i n naive subjects. Second, i n t h i s experiment, each subject was exposed to only one type of sample stimulus (either food or l i g h t durations). Thus, the choose short e f f e c t i s not s p e c i f i c to the more complex task of Experiment 1, which i n -volved both kinds of sample st i m u l i within the same session. Third, the present results suggested that the choose short e f f e c t cannot be attributed e a s i l y to either a color preference, or to a bias present during a c q u i s i t i o n . There was no apparent relationship between the birds' tendency to choose short or 46 long during a c q u i s i t i o n and the choose short e f f e c t they d i s -played during the delay manipulation phase. F i n a l l y , t h i s experiment demonstrated that the choose short e f f e c t does not diminish with extended testing with the variable delay pro-cedure . 47 Experiment 3 This experiment was designed to extend the previous findings by demonstrating the choose short e f f e c t with a procedure that involved more than two sample durations. Two pigeons were trained under a three-choice DSMTS procedure i n which three d i f f e r e n t l i g h t durations served as "short," "medium," and "long" samples, and three d i f f e r e n t key stimuli served as the comparisons. The birds were trained to match each of the three sample st i m u l i to the appropriate comparison. The e f f e c t of the delay i n t e r v a l on choice of the three com-parisons after each sample duration ("short," "medium," and "long") then was examined. Under t h i s procedure, a choose short e f f e c t would be r e f l e c t e d by a selec t i v e increase i n incorrect choices to the "short" comparison, and a decrease i n correct choices to both the "medium" and the "long" com-parisons. Method Subjects and apparatus Birds 1 and 2 from Experiment 1 served as the subjects. The apparatus was the same as that used i n Experiment 1 except that three d i f f e r e n t stimuli (red, yellow, and green l i g h t for Bird 1; red l i g h t , green l i g h t , and a white rectangle on a black background for Bird 2) were presented on the three 48 pecking keys as comparisons. Procedure Baseline procedure. Each t r i a l began with the presenta-t i o n of a houselight sample for one of three durations: 2 sec (short), 6 sec (medium), or 14 sec (long). Immediately follow-ing termination of the sample (0-sec delay), the three pecking keys were illuminated with the three comparison s t i m u l i . For Bird 1, choices of red after short, yellow after medium and green aft e r long were designated correct; for Bird 2 the correct choices were green after short, the rectangle after medium, and red after long. Correct choices were reinforced with 4-sec access to grain followed by a 30-sec ITI; incorrect choices terminated the t r i a l and i n i t i a t e d the ITI. The order of presentation for three samples and the six possible key arrangements of the comparison stimuli was determined randomly. Bird 1 received 50 sessions, and Bird 2 received 30 sessions of baseline t r a i n i n g . Delay manipulations. The variable delay procedure was i n i t i a t e d following baseline t r a i n i n g . On a randomly-determined half of the t r i a l s i n each session the delay was 0-sec; on the remaining t r i a l s a delay of 5-sec or 10-sec occurred, each with an equal p r o b a b i l i t y . A l l other aspects of the procedure were unchanged. Both birds were tested for 20 sessions, 49 Results Baseline performance Figure 5 shows the percentage of t r i a l s on which the birds chose the "short," medium," or "long" comparison keys after each sample duration during the l a s t 10 sessions of baseline t r a i n i n g . Each b i r d was well above chance l e v e l (33.3%) i n choosing the correct comparison after each sample duration. Moreover, aft e r short and long samples, more errors occurred to the "medium" comparison, whereas after medium samples the errors were equally l i k e l y to be a choice of the "short" or "long" comparisons. Thus, both birds' choice of the comparison stimuli c l e a r l y was controlled by the sample duration. Delay manipulations Choices of the three comparison st i m u l i a f t e r each sample duration at the three delays are shown i n Figure 6. Data from a l l 20 sessions are included. Several features of these data should be noted. F i r s t , after both the medium and long samples, choice of the correct comparison decreased over the delays, whereas after the short samples, choice of the correct comparison did not change greatly over the delay. Second, after medium samples, choice of the "short" comparison i n -creased at the 10-sec delay, whereas choice of the "long" comparison did not change substantially at any of the delays. gure 5 . Percentage of t r i a l s on which the birds chose the "short" (S), "medium" (M), or "long" (L) compari-son stimulus after each of the three sample durations during the l a s t 10 sessions of baseline t r a i n i n g . Slashed bars represent correct choices the dashed l i n e indicates chance l e v e l (.33%) for a three-choice procedure. 100' 1 0 < o UJ CD < 50 100 LU O CX UJ CL 5 0 2-sec SAMPLE BIRD 1 6-SPC SAMPLE 14-sec SAMPLE M S M L BIRD 2 M 2 -sec SAMPLE 6-sec SAMPLE 14-sec SAMPLE / - - -_ ^ n i - i n n I—I M L S M L S M L COMPARISON CHOSEN Figure 6. Percentage of t r i a l s on which ' the birds chose the "short" (S), "medium" (;M) , and "long" (L) compari son after each of the three sample durations at the 0-, 5-, and 10-sec delays. Slashed bars represent correct choices. The dashed l i n e indicates chance l e v e l (33%) for a three-choice procedure. lOOn 501 2-sec SAMPLE All sml sml sml 0 5 10 BIRD 1 6-sec SAMPLE kL sml sml sml 0 5 10 14-sec SAMPLE XI sml sml sml 0 5 10 10 Ol BIRD 2 2-sec SAMPLE 501 tL Q0 A6-sec SAMPLE h i 1 i 11 11 J U-sec SAMPLE sml sml sml 0 5 10 sml sml sml COMPARISON CHOSEN 0 5 10 sml sml sml 0 5 10 DELAY ( s e c ) 54 Third, a f t e r long samples, choice of the "short" comparison increased at the 10-sec delay. These results were confirmed by three, three-way repeated measures analyses of variance of choices of the "short," "medium," and "long" comparisons, with the factors being delay, sample duration, and subjects. The analysis of choices of the "short" comparison revealed a s i g n i f i c a n t main e f f e c t of sample duration [F(2,21=46.08, p<.05]. No other effects were s i g n i f i c a n t . The analysis of choices of the "medium" comparison showed a s i g n i f i c a n t sample duration by delay interaction IF(4,4)=14.02, p<.05], but no other s i g n i f i c a n t e f f e c t s . S i m i l a r l y , the analysis of choices of the "long" comparisons showed a s i g n i f i c a n t sample duration by delay interaction [F(4,4)=7.65, p<.05], and no other s i g n i f i c a n t e f f e c t s . Subsequent a p o s t e r i o r i comparisons (Newman-Keuls, p= .05) showed that incorrect choices of the "short" comparison after both medium and long samples occurred more often at the 10-sec delay than at the 0-sec delay, whereas incorrect choices of the "medium" comparison after either short or long samples, and of the "long" comparison a f t e r e i t h e r short or medium samples, did not change s i g n i f i c a n t l y over the three delays. Furthermore, correct choices of both the "medium" and the "long" comparisons decreased s i g n i f i c a n t l y at both of the longer delays; whereas, correct choices to the "short" 55 comparison did not change s i g n i f i c a n t l y as a function of the delay. Thus, under the three-key procedure, increases i n the delay produced an increased tendency to choose the comparison associated with the shortest sample. Discussion These results extend the previous findings by showing that the choose short e f f e c t also occurs i n a three-choice procedure. In addition, the three-choice procedure permitted a more refined analysis of the response tendencies that are the basis of the choose short e f f e c t . The choose short e f f e c t demonstrated i n the previous experiments could have been due to either an increased tendency to choose the "short" comparison, or to an increased tendency to avoid choosing the "long" comparison. Although the two-choice procedure did not provide any means of distinguishing between these two alterna-t i v e s , the three-choice procedure was useful i n t h i s regard. If the choose short e f f e c t was due to an avoidance of choosing the "long" comparison rather than a tendency to choose the "short" comparison, then correct choices of the "medium" comparison would not be expected to decrease over the delay, and incorrect choices of the "short" comparison should not increase s e l e c t i v e l y ; the fact that both of these results were obtained supports the assumption that the choose short e f f e c t does r e f l e c t an increased tendency to "choose short." 56 PART I I : TESTS OF THE SUBJECTIVE SHORTENING MODEL The results of the f i r s t three experiments demonstrated the r e l i a b i l i t y and generality of the choose short e f f e c t . Experiment 1 demonstrated the choose short e f f e c t with both food-access and l i g h t durations as samples. Experiment 2 extended the generality of t h i s e f f e c t to naive subjects, and ruled out the p o s s i b i l i t y that simple color preferences were the basis of the choose short e f f e c t . F i n a l l y , Experiment 3 extended the generality of the choose short e f f e c t to a more complex procedure that involved three sample durations and three choice s t i m u l i . As mentioned i n Experiment 1, one interpretation of the choose short e f f e c t i s that the duration of events i n pigeons' working memory shortens over time. A model, based on thi s idea of subjective shortening, w i l l now be described i n d e t a i l . This model hereafter w i l l be referred to as the "subjective shortening" model. The subjective shortening model has two principal com-ponents: a working memory and a reference memory. According to t h i s model, pigeons base t h e i r choice of the comparison sti m u l i upon a working memory of the sample duration that shortens as a function of time. Consequently, the representa-t i o n of the sample i n working memory i s shorter a f t e r an extended delay than i t i s after a b r i e f delay. These changes 57 i n the subjective duration of the sample as a function of the delay are the basis of the choose short e f f e c t . The manifesta-t i o n of t h i s choose short e f f e c t , however, also depends upon the second, reference memory, component of the model. During i n i t i a l t r a i n i n g with a 0-sec delay., the pigeons form an association between the sample representations and the correct comparisons; th i s association i s maintained i n reference memory. This reference memory, which i s based upon the 0-sec delay trials, is maintained during the delay phase provided that a considerable portion of the t r i a l s consist of 0-sec delays. The choose short e f f e c t on long delay t r i a l s occurs because of a discrepancy between the working memory representation of the sample and the reference memory representation of the sample. As the delay increases, the working memory of the long sample shortens, becoming less s i m i l a r to the reference memory of the long sample, and more si m i l a r to the reference memory of the short sample. Simple stimulus generalization then would lead to an increased tendency to choose the short comparison. Because the working memory of the short sample also i s shortened after a long delay, i t becomes less s i m i l a r to the reference memory of the long sample. Therefore, stimulus generalization would not be expected to produce an increased tendency to choose the long comparison. Thus, the two com-ponents of the model, working memory and reference memory, combine to produce the choose short e f f e c t . 58 At the time th i s research was i n i t i a t e d , no other i n v e s t i -gations of the effects of delay manipulations on duration d i s -crimination i n animals had been reported. However, two studies of memory for event duration i n animals have since appeared. Church (1980) used a r i g h t - l e f t , lever-choice procedure to test rats' memory for short (2-sec) and long (8-sec) signals over retention i n t e r v a l s of 0, .5, 2, 8, and 32 sec. He found that with delay increases of up to 8 sec, accuracy decreased approximately equally for short and long signals. However, with a 32-sec delay, accuracy was much lower aft e r long signals than after short signals; the rats chose the "short" lever on approximately 65% of the t r i a l s . Church also found that when the signal was omitted, the rats chose the "short" lever with a p r o b a b i l i t y of .79. In a second experiment, Church compared performance aft e r a .5-sec, a 2-sec, and an 8-sec retention i n t e r v a l to perform-ance aft e r a 0-sec retention i n t e r v a l , using a psychophysical procedure. In t h i s study he found no evidence of any changes i n rats' subjective representation of duration as a function of the retention i n t e r v a l . For the most part, Church's results are not inconsistent with those of the present research. F i r s t , Church did demon-strate that after a long retention i n t e r v a l (.32 sec) , rats showed a choose, short e f f e c t . Second, although there was no evidence of a choose .short e f f e c t i n his second experiment, this could have been due to a f a i l u r e to use a s u f f i c i e n t l y long 59 retention i n t e r v a l (e.g., the 32-sec i n t e r v a l used i n his previous experiment). Church, however, interpreted the choose short e f f e c t demonstrated i n his f i r s t experiment i n terms of "biased guessing" rather than subjective shortening. He favored t h i s interpretation because the rats also showed a choose short tendency when the signal was omitted. On the basis of the s i m i l a r i t y between the rats' performance aft e r a 32-sec delay and after a 0-sec s i g n a l , Church argued that a sim i l a r process was the basis of responding i n both cases. S p e c i f i c a l l y , he suggested that the tendency to choose short r e f l e c t e d biased guesses that occur " i f the trace of the sample had faded" ( i . e . , after a 32-sec delay), or " i f the signal was not pre-sented" ( i . e . , on the 0-sec signal t e s t s ) . Nevertheless, the subjective shortening model also can account for the s i m i l a r i t y between responding i n the absence of a signal and responding af t e r long delays. F i r s t , a ten-dency to choose short af t e r long delays should occur because of the subjective shortening of the signal duration. Second, a similar choose short tendency should occur af t e r no signal because of stimulus generalization; the absence of a signal may be more similar to a b r i e f signal than to a longer s i g n a l . Thus, rats should respond after no signal and af t e r long delays i n much the same way: they should choose short. The guessing interpretation of responding after no signals and after a long 60 delay, on the other hand, does not o f f e r any explanation for why guessing should be biased i n the short d i r e c t i o n . The second study investigating animals' memory for event duration was conducted by Cohen et a l . (1981). They tested pigeons' working memory for a .5-sec (short) and a 4-sec (long) sample i n a DSMTS procedure. They varied the delay from 0 sec to 16 sec i n a stepwise fashion i n which each delay was i n e f f e c t for several consecutive sessions. They found no consistent differences between accuracy after short and long samples at any of the delays. Although Cohen et a l . f a i l e d to demonstrate a s i g n i f i c a n t choose short e f f e c t , the experimental procedure they used and the i r method of data c o l l e c t i o n may have obscured t h i s e f f e c t . They used a stepwise rather than a variable delay procedure, and they only reported and analyzed the mean accuracy scores of the l a s t f i v e sessions at each delay. Although the present experiments demonstrated that the choose short e f f e c t i s r e l a t i v e l y stable over t e s t session when a variable delay procedure i s used, i t i s possible that the choose short e f f e c t may decrease aft e r extended t r a i n i n g at a single, constant delay. In fact, t h i s can be predicted d i r e c t l y from the sub-j e c t i v e shortening model (see Experiment 5), The purpose of the following two experiments, was to pro-vide empirical tests of the subjective shortening model i n the l i g h t of the recent studies by Church (1980) and Cohen et a l . (1981). 61 Experiment 4 One method of te s t i n g the subjective shortening model i s to use a psychophysical procedure to compare the point of sub-je c t i v e equality (PSE) across d i f f e r e n t delay i n t e r v a l s (cf. Church, 1980). The PSE i s thought to represent the sample duration that produces an i n t e r n a l representation that i s halfway between the "long" and "short" t r a i n i n g durations. In order to determine the PSE, subjects are exposed to a num-ber of sample durations i n addition to the t r a i n i n g values, and the p r o b a b i l i t y of choosing "long" i s plotted as a function of sample duration. Then the method of least squares can be used to calculate a l i n e a r regression equation r e l a t i n g the p r o b a b i l i t y of choosing long to sample duration. This equa-tion then can be used to derive the sample duration that cor-responds to a p r o b a b i l i t y of .5 of choosing long, which defines the PSE. If the sample durations subjectively become shorter over the delay, then a longer sample duration would be required i n order to maintain a p r o b a b i l i t y of .5 of choosing long. Thus, according to the subjective shortening model, the PSE should s h i f t towards longer durations as the delay i n t e r v a l i s i n -creased . Although Church (1980) f a i l e d to f i n d any s h i f t s i n the PSE as a function of the delay i n t e r v a l , he did not use any 62 delays of longer than 8 sec. The present experiment therefore was designed to compare the PSE at 0-, 5-, and 20-sec delays. Pigeons were trained under a DSMTS procedure with 2- and 10-sec samples, then were tested with various sample durations after delays of 0, 5, and 20 sec. Functions r e l a t i n g the pro b a b i l i t y of choosing long to sample duration then were determined, and the PSE was calculated for each delay i n t e r v a l . A s i g n i f i c a n t increase i n the value of the PSE as a function of increases i n the delay c l e a r l y would support the subjective shortening model. Method Subjects and apparatus These were the same as those used i n Experiment 1. Procedure Baseline sessions. In the present experiment, the birds f i r s t were exposed to a few baseline sessions which involved the same basic procedure as described i n Experiment 1, except that only one type of stimulus (light) served as the 2- and 10-sec samples, and correct choices were reinforced with a pro b a b i l i t y of .75. Generalization t e s t i n g : Series 1. The procedure used i n t h i s condition was sim i l a r to that used i n the baseline sessions, except that within each session, generalization test 63 t r i a l s occurred with a pro b a b i l i t y of .25. On these t e s t t r i a l s , the l i g h t sample was presented for one of three t e s t durations: 4 sec, 6 sec, or 8 sec, each occurring with an equal p r o b a b i l i t y . Following the delay, the usual comparison sti m u l i were presented, A peck to either of the comparison keys terminated the t r i a l and was recorded but never reinforced. On the remaining 75 percent of the t r i a l s , one of the usual 2- and 10-sec samples was presented, each with an equal p r o b a b i l i t y , and correct choices were reinforced with a proba-b i l i t y of .75. Both birds were tested under, t h i s condition for 30 sessions. Generalization t e s t i n g : Series 2. The procedure used i n t h i s condition was i d e n t i c a l to that used for Series 1, except that the test durations of the samples i n th i s series were 6 sec, 12 sec, and 18 sec. Both birds were tested for 30 sessions under th i s condition. Treatment of the data. For each series, the point of subjective equality (PSE) was estimated for each of the three delays i n the following manner. F i r s t , a l i n e a r regression equation r e l a t i n g the p r o b a b i l i t y of a "long" response to sample duration was calculated by the method of least squares for each delay. Second, these equations (of the form y = mx + b) were used to fi n d the sample duration at each delay that corresponded to a pr o b a b i l i t y of .5 of a "long" response, which represents the estimated PSE (cf. Church & Deluty, 1978). 64 In addition, the proportion of t r i a l s on which the birds chose the "long" comparison was plotted as a function of the sample duration for each of the three delays, and the PSE was plotted as a function of delay. For these graphic presenta-tions the data were collapsed over the two ser i e s . Results Figure 7 shows the p r o b a b i l i t y of responses to the "long" comparison as a function of sample duration for the three delay i n t e r v a l s . The delay i n t e r v a l c l e a r l y had a marked e f f e c t upon these functions. The sample duration at which the birds began to choose "long" on over half of the t r i a l s was much longer af t e r a 20-sec delay i n t e r v a l than after a 0-sec or a 5-sec delay. A comparison of the estimated PSE at the three delay i n t e r v a l s corroborated these observations: the PSE was longer at the 20-sec delay than at either the 0-sec or 5-sec delay (see Figure 8). At the 0-, 5-, and 20-sec delays, respectively, the estimated PSE for Bird 1 was 4.8 sec, 5,8 sec, and 17.6 sec during Series 1, and 5.4 sec, 3.9 sec, and 20.6 sec during Series 2. For Bird 2 the estimated PSE with the 0-, 5-, and 20-sec delays, respectively, was 4.4 sec, 5,1 sec, and 20.6 sec during Series 1, and 4,5 sec, 7.6 sec, and 10.8 sec during Series 2. 65 Figure 7. Probab i l i t y of a response to the "long" comparison as a function of sample duration for the 0-sec, 5-sec, and 20-sec retention i n t e r v a l s . RETENTION INTERVAL: 0 1 l 1 1 1 1 : 1 « ~ 2 4 6 8 10 H 18 S A M P L E DURATION ( s e c ) gure 8. Point of subjective equality at the three delays. D E L A Y ( S E C ) 69 A two-way repeated measures analysis of variance per-formed on the PSE data revealed a s i g n i f i c a n t main e f f e c t of delay [F (.2,2) =33 . 8 , p<.05], but not of series [F (1,1) = .64, p> .05] or subjects [F (1,2) = .76, p>.05] . None of the interactions were s i g n i f i c a n t . Subsequent a p o s t e r i o r i comparisons (Newman-Keuls, p=.05) confirmed that the PSE was longer at the 20-sec delay than at either the 0-sec or the 5-sec delay. Discussion The finding that the PSE had shi f t e d towards a much longer duration at the 20-sec delay confirms the prediction that was derived from the subjective shortening model. These results challenge Church's (1980) argument that the subjective duration of a signal does not decrease over delays. The f a i l u r e to f i n d any changes i n the PSE as a function of delay may be r e s t r i c t e d to the shorter delays used i n his study. 70 Experiment 5 Experiment 5 was designed to provide further support for the subjective shortening model. As i n the previous experi-ments, pigeons were trained f i r s t on a 0-sec DSMTS procedure with two durations of l i g h t as the samples. In thi s experi-ment, however, the delay between the sample and the comparison sti m u l i then was manipulated i n a stepwise rather than a variable fashion. The delay was increased f i r s t from 0 sec to 5, 10, then 20 sec, and then was decreased to 5 and 0 sec, with each delay being i n e f f e c t for several consecutive sessions. According to the subjective shortening model, the step-wise delay procedure should produce d i f f e r e n t results from those produced by the variable delay procedure. In order to i l l u s t r a t e how these d i f f e r e n t outcomes are derived from the model, i t may be useful to consider the memory requirements under each procedure. In both procedures, during i n i t i a l t r a i n i n g with a 0-sec delay, i t i s assumed that a reference memory i s formed which contains an association between the representations of the samples at a 0-sec delay and the correct comparison s t i m u l i . When the delay then i s manipulated using the variable delay procedure, a random half of the t r i a l s i n each session s t i l l are composed of 0-sec delays. These 0-sec delay t r i a l s should serve to maintain the reference memory established during i n i t i a l t r a i n i n g . On the remaining t r i a l s , i n which a longer delay i s presented, responding should con-71 tinue to be guided by the previously established association i n reference memory. Thus, the animals' responses are based not only upon t h e i r working memory of the sample, but also upon the relationship between th i s working memory and the i r reference memory. A choose short tendency occurs after long delays because of the discrepancy between t h e i r working memory of a long sample (which has become shorter) and the representa-t i o n of a long sample i n the i r reference memory (which i s based upon the 0-sec delay). In the variable delay procedure, th i s choose short e f f e c t i s stable over time because of the r e l a t i v e s t a b i l i t y of.the reference memory. When the delay i s manipulated using a stepwise procedure, on the other hand, the reference memory should not remain stable over the delay manipulation phase because 0-sec delays are not interspersed with the longer delays. Extended t r a i n i n g with a single delay at a time should r e s u l t i n a change i n the reference memory as a new association between the sample repre-sentations at that delay and the correct comparisons i n learned. Because the stepwise procedure results i n a changing reference memory, thi s procedure was expected to y i e l d a number of unique outcomes. F i r s t , with each increase i n the delay between the sample and the comparisons, the pigeons should show a temporary choose short e f f e c t . The reasons for t h i s temporary choose short e f f e c t are as follows. During the 0-sec delay, the pigeon 72 would have learned to associate the two comparisons with the long and short sample representations, and thi s association would become part of t h e i r reference memory. However, when the delay between the samples and comparisons i s lengthened, the pigeon i n i t i a l l y should base i t s responses on the pre-vously established reference memory. If the subjective representation of the samples i n working memory had shortened as a function of the delay, then the working memory duration of the long sample would be shi f t e d towards the reference memory duration of the short sample, leading to an increased tendency to choose short. Second, with extended t r a i n i n g at a given delay, t h i s choose short e f f e c t should diminish and o v e r a l l accuracy should improve as a new association between the subjective durations of the samples and the comparisons i s formed and maintained i n reference memory. Third, following extensive t r a i n i n g at a given delay, a subsequent decrease i n the delay should r e s u l t i n a temporary tendency to choose the comparison associated with the long sample (a "choose long" e f f e c t ) . The reason for thi s tempor-ary choose long e f f e c t can be i l l u s t r a t e d with the following example. Given that the working memory durations of the samples would have shortened more after a 2Q-sec delay than afte r a 5-sec delay, the subjective representations of the samples afte r a 5-sec delay would be longer r e l a t i v e to those afte r a 20-sec delay. If the animal had formed a reference 73 memory containing an association of the comparison stimuli with the subjective sample durations present at a 20-sec delay, a decrease i n the delay from 20 sec to 5 sec should r e s u l t i n i t i a l l y i n a disruption of the d i s c r i m i n a b i l i t y of the short sample, and hence more incorrect choices of the "long" comparison. In summary, on the basis of the subjective shortening model, i t was predicted that with a stepwise delay procedure: 1) increases i n the delay between the samples and comparisons would lead to an i n i t i a l choose short e f f e c t , 2) decreases i n th i s delay would produce an i n i t i a l choose long e f f e c t , and 3) with extended t r a i n i n g at a given delay, these choose short and choose long effects would diminish and o v e r a l l matching accuracy would improve; furthermore, there should be a posit i v e c o r r e l a t i o n between these two changes. Empirical confirmation of these predictions would provide strong support for the subjective shortening model. Experiment 5 also was designed to assess pigeons' choice behaviour when the sample was omitted ••(.i.e.> 0-sec sample t e s t s ) . I t was expected that pigeons would be more l i k e l y to treat a 0-sec sample as "short" than as "long" because of the process of stimulus generalization. Method Subjects Five naive adult S i l v e r King pigeons (Birds 7, 8, 9, 10, 74 and 11) served as the subjects. Deprivation and housing con-dit i o n s were the same as those described for subjects i n Experiment 1. Apparatus The t e s t chamber for Birds 7, 8, and 9 was i d e n t i c a l to that described i n Experiment 1. For Birds 10 and 11, a BRS/LVE #132-02 light-proof, sound-attenuating t e s t chamber was used. One wall of th i s chamber contained a horizontal array of three pecking keys, each equipped with a microswitch to sense pecks of .2N or greater. An Industrial Electronics Engineers' Series 10 stimulus projector was mounted behind each key; these illuminated the center key with a uniform f i e l d of yellow l i g h t , or with a white square on a dark background, and the side keys with a uniform f i e l d of either red or green l i g h t . A BRS/LVE #114-10 grain feeder containing a 2.8W lamp was mounted below the center key. Procedure Preliminary t r a i n i n g . During a few preliminary sessions, each b i r d was trained to eat from the raised illuminated grain feeder, and then trained to peck the center key when i l l u m i -nated with yellow l i g h t , and the side keys when illuminated with either red or green l i g h t . Baseline condition (0-sec DSMTS). A v a r i a t i o n of the DSMTS procedure, si m i l a r to that described i n Experiment 1, was used. In th i s experiment, t r i a l s began with the presenta-75 ti o n of a yellow l i g h t on the center pecking key; the f i r s t peck to th i s key terminated the yellow l i g h t and produced the sample stimulus, which lasted for either a short (.2-sec) or long (10-sec) duration. For Birds 10 and 11, the sample stimulus was the presence on the center key of a white square on a dark background, whereas for Birds 7, 8, and 9 the sample stimulus was illumination of the houselight. Immediately (0-sec) following the o f f s e t of the sample, the side keys were illuminated with red and green l i g h t , which served as the comparison sti m u l i . . For Birds 10 and 11 the red comparison was designated as.correct after short samples and the green comparison aft e r long samples; for Birds 7, 8, and 9, green was designated correct after short, and red after long samples. A peck to the correct comparison terminated both comparisons and produced 3-sec access to grain followed by a 30-sec ITI; pecks to the incorrect comparison terminated the t r i a l and i n i t i a t e d the ITI. The presentation of the short and long samples, and the arrangement of the red and green comparisons on the side keys occurred i n a mixed and counterbalanced order over t r i a l s i n each session. Each b i r d was trained under th i s condition u n t i l matching accuracy was well above chance and appeared to be stable. Accordingly, Birds 7, 8, 9, 10, and 11 were exposed to this condition for 24, 18, 18, 36, and 24 sessions, respectively. 76 Delay manipulations. During t h i s condition, a l l aspects of the procedure were the same as during baseline except that the delay between the sample and comparison st i m u l i was manipulated over blocks of several sessions. For each b i r d , the delay manipulations were conducted i n three consecutive phases: Phase 1: Increases i n the delay. During t h i s phase, the delay was increased for several sessions to 5 sec, then to 10 sec and/or to 20 sec. Birds 7, 8, and 10 were tested at a l l three of these delays, whereas Bird 9 was tested only at the 5- and 20-sec delays and Bird 11 was tested only at the 5- and 10-sec delays. Birds 7, 8, 9, 10, and 11 were tested at the 5-sec delay for 21, 21, 21, 24, and 27 sessions, respectively. Birds 7, 8, 10, and 11 were tested at the 10-sec delay for 15, 12, 12, and 33 sessions, and Birds 7, 8, 9, and 10 were tested at the 20-sec delay for 18, 21, 12, and 18 sessions, respectively. Phase 2: Decreases i n the delay. During t h i s phase, the delay f i r s t was decreased from 10 sec (Bird 11) or from 20 sec (Birds 7, 8, 9, and 10) to 5 sec for several sessions. Subsequently, the delay was decreased to 0 sec for several sessions. Birds 7, 8, 9, 10, and 11 were tested at the 5-sec delay for 6, 6, 9, 9, and 12 sessions, and at the 0-sec delay for 6, 6, 6, 6, and 9 sessions, respectively. 77 Phase 3: Replication of the increases i n the delay. Following exposure to the 0-sec delay i n Phase 2, the delay was increased again to 5 sec for six sessions (Bird 8) or three sessions (Birds 7, 9, 10, and 11). For Bird 8, the delay subsequently was increased once again to 20 sec for three additional sessions. 0-sec sample te s t s . Following the delay manipulation phase, each of the birds was administered three 0-sec sample test: sessions i n which the sample st i m u l i were not presented; the comparison st i m u l i simply were presented at the beginning of each t r i a l without being preceded by a sample. Each of the three 0-sec sample te s t sessions were separated by a baseline (0-sec delay) session. On the 0-sec sample tests, the birds' choice of the comparisons, which had been designated as correct for short and long samples throughout the experiment, was recorded. These te s t sessions were conducted to determine whether the birds would show a tendency to choose short after a 0-sec sample. Results The results of t h i s experiment confirmed each of the predictions derived from the subjective shortening model. During the f i r s t session following an increase i n the delay to 5 sec, 10 sec, or 20 sec, the birds showed a consistent 78 choose short e f f e c t , whereas during the f i r s t session following a decrease i n the delay to 5 sec, the birds showed a choose long e f f e c t . In addition, with extended t r a i n i n g at each delay the choose short and choose long e f f e c t s diminished and o v e r a l l accuracy improved. F i n a l l y , on 0-sec sample tests, the birds tended to choose the short comparison. These results w i l l be discussed i n d e t a i l i n the following sections. A. Choose short and choose long effects The percentage of correct choices af t e r short and long samples i s shown i n Figures 9a and 9b for each b i r d i n blocks of three consecutive sessions, except for the f i r s t three sessions after each delay change which are shown i n d i v i d u a l l y to f a c i l i t a t e detection of the i n i t i a l choose short and choose long e f f e c t s . Each increase i n the delay resulted i n an i n i t i a l choose short e f f e c t , whereas a decrease i n the delay led to an i n i t i a l choose long e f f e c t . Figure 10 shows the mean per-centages of a l l the birds' correct choices after short and long samples during the f i r s t session af t e r each delay change i n the three phases. Clearly, during Phases 1 and 3 (increases i n the delay), accuracy was higher after short samples than afte r long samples, whereas during Phase 2 (decreases i n the delay), accuracy was higher a f t e r long than aft e r short samples. A p r i o r i one-tailed dependent t - t e s t s , comparing accuracy a f t e r short and long samples, were conducted on the data shown i n Figure 10. These analyses revealed that accuracy after gure 9a. Percent correct a f t e r short and long duration samples during blocks of three consecutive sessions of the experiment, and during the f i r s t three i n d i v i d u a l sessions after each change i n the delay (indicated by the s o l i d v e r t i c a l lines) for Birds 7 and 8. D E L A Y ( S E C ) gure 9b: Percent correct after short and long duration samples during blocks of three consecutive sessions of the experiment, and during the f i r s t three i n d i v i d u a l sessions afte r each change i n the delay (indicated by the s o l i d v e r t i c a l lines), for Birds 9, 10, and 11. 82 DELAY (SEC ) I I u r e 10. Mean p e r c e n t a g e (+ SEM) o f c o r r e c t c h o i c e s a f t e r s h o r t ( s l a s h e d b a r s ) and l o n g (open b a r s ) sample d u r a t i o n s d u r i n g t h e f i r s t s e s s i o n a f t e r e a c h d e l a y c h a n g e . 100 90 p h a s e 1: i n c r e a s e d 60-1 V 80-1 y 7 0 4 ft XA W / 10 s h o r t • l o n g pha se 2 : d e c r e a s e d p h a s e 3: i n c r e a s e d 20 D E L A Y ( S E C ) 85 short samples was s i g n i f i c a n t l y greater than accuracy af t e r long samples during the session following each delay increase i n Phase 1 [5-sec delay: t(.4)=4.41, p<.01; 10-sec delay: t(3) = 2.68, p<.05; 20-sec delay: t(3)=3.4, p<.05], and following the second delay increase to 5 sec i n Phase 3 It (.4) =3.67, p< .05]. In contrast, accuracy after short samples was s i g n i f i -cantly lower than accuracy after long samples during the f i r s t session following a decrease i n the delay to 5 sec i n Phase 2 It(4).=2.67, p<.05]. The lower accuracy after short samples at the 0-sec delay i n Phase 2 was not s i g n i f i c a n t It(4)=1.19], B. E f f e c t of extended t r a i n i n g at a given delay 1. Decreases i n the choose short and choose long e f f e c t s . The choose short and choose long e f f e c t s that were present after the delay changes tended to diminish with extended exposure to each of the delays (Figures 9a and 9b). To test the significance of th i s change with extended t r a i n i n g , "choice r a t i o s " were calculated using the data from the f i r s t and l a s t blocks of three sessions at each delay i n Phase 1 and Phase 2. These rat i o s were calculated by div i d i n g the per-centage of correct choices after short samples by the sum of the percentage of correct choices after both short and long samples. Thus, r a t i o s of greater than .5 indicate higher accuracy after short samples ( i . e . , a choose short e f f e c t ) , and r a t i o s of less than .5 indicate higher accuracy after long samples ( i . e . , a choose long e f f e c t ) . A decrease i n the choose 86 short e f f e c t over t r a i n i n g therefore would be r e f l e c t e d i n a decrease i n the choice r a t i o , whereas a decrease i n the choose long e f f e c t over t r a i n i n g would r e s u l t i n an increase i n the choice r a t i o . For a l l except the 10-sec delay i n Phase 1, the predic-t i o n that the choose short and choose long effects would diminish with extended t r a i n i n g was confirmed by one-tailed dependent measures t-tests comparing the choice r a t i o s for the f i r s t and l a s t block of three sessions. During Phase 1 (delay increases), the choice r a t i o decreased s i g n i f i c a n t l y from the f i r s t to the l a s t block of sessions at the 5-sec delay IM (first)=.626, M Mast) = . 510 ; t (.4) =3.45, p<.05] and at the 20-sec delay IM (first)=.625, M (last)=.502; t(3)=6.28, p<.01], but not at the 10-sec delay IM (first)=.589, M (last)= .515; tC3)=2.21, p>.05]. In contrast, during Phase 2 (delay decreases), the choice r a t i o increased s i g n i f i c a n t l y from the f i r s t to the l a s t block of sessions at both the 5-sec delay [M ( f i r s t ) = . 44.0, M (last) = .515; tt4)=4.17, p<.01J and the 0-sec delay I'M (f irst). = . 476, M (last) = . 511; t(.4)=2.61, p<.05]. 2. Increases i n o v e r a l l accuracy. Figure 11 shows each bird's o v e r a l l matching accuracy for blocks of three consecu-t i v e sessions. As predicted on the basis of the subjective shortening model, o v e r a l l matching accuracy improved with extended t r a i n i n g at each delay during Phase 1 and Phase 2. One-tailed dependent measures t-tests comparing o v e r a l l gure 11. Each bird's o v e r a l l match-ing accuracy during blocks of three consecutive sessions of the experiment. DELAY (sec) 89 accuracy on the f i r s t and l a s t block of three sessions con-firmed that o v e r a l l accuracy increased s i g n i f i c a n t l y from the f i r s t to the. l a s t block of sessions at the 5-sec delay [t(4) = 7.51, p<.01], at the 10-sec delay [t(.3)=3.57, p<.05] and at the 20-sec delay [t(.3)=5.15, p<.01J during Phase 1, as well as at the 5-sec delay It(4)=6.18, p<.01] and the 0-sec delay It(4)=2.65, p<.05] during Phase 2. 3. Relationship between the increases i n accuracy and  the decreases i n the choose short and choose long e f f e c t s . In addition to predicting that extended t r a i n i n g at a given delay would lead to an increase i n o v e r a l l accuracy and a decrease i n the choose short and choose long e f f e c t s , i t was also predicted that these two eff e c t s would be correlated: both changes should begin to occur when the animal begins to acquire a new association between the comparisons and the sub-je c t i v e sample durations present at the p a r t i c u l a r delay. This prediction was tested by determining the c o r r e l a -t i o n between the choice r a t i o and o v e r a l l accuracy at each delay during Phase 1 and Phase 2, using the data from the f i r s t two blocks of sessions for each b i r d . The prediction of a di r e c t relationship between increases i n accuracy and decreases i n the choose short e f f e c t during Phase 1 was confirmed by a s i g n i f i c a n t negative c o r r e l a t i o n (one-tailed, p<,05). between the choice r a t i o and o v e r a l l accuracy at the 5-sec delay (r=-,788), and the 10-sec delay (r=-.649). At the 20-sec 90 delay the c o r r e l a t i o n was negative, but f a i l e d to reach s i g -nificance (r=-.294). . S i m i l a r l y , the prediction that there should be a d i r e c t r elationship between increases i n accuracy and decreases i n the choose long e f f e c t during Phase 2 was confirmed by a s i g n i f i c a n t p o s i t i v e c o r r e l a t i o n between the choice r a t i o and o v e r a l l accuracy at the 5-sec delay (r=.771), and at the 0-sec delay (r=.788). 0-sec Sample Tests Figure 12 shows the percentage of t r i a l s on which the birds chose the "short" comparison during the three 0-sec sample tests. A l l birds showed a consistent tendency to choose the "short" comparison i n the absence of the sample. The results of a dependent t - t e s t (two-tailed) of the mean percentage of "short" and "long" choices confirmed that the birds chose short s i g n i f i c a n t l y more often than they chose long on these 0-sec sample tests [t(4)=5.17, p<.05]. Discussion Each.of the predictions derived from the subjective shortening model was confirmed. F i r s t , each stepwise increase i n the delay led to an i n i t i a l choose short e f f e c t . Second, a decrease i n the delay to 5 sec led to an i n i t i a l choose long e f f e c t . Third, these choose short and choose long e f f e c t s diminished, and o v e r a l l accuracy improved as a function of gure 12. Percentage of t r i a l s on . which the birds chose "short" during the three 0-sec sample tests. <N1 b i r d 7 8 9 10 H O X H Z w u 100-75-5 0 -25-0-1 2 3 1 2 3 1 2 3 1 2 3 0 - S E C S A M P L E T E S T S 93 tr a i n i n g at each delay. Furthermore, these two e f f e c t s were s i g n i f i c a n t l y correlated. The empirical confirmation of these predictions provides strong support for the subjective shorten-ing model. The present results also i l l u s t r a t e the importance of precedural variables i n the study of working memory. Clea r l y , the stepwise delay procedure used i n t h i s experiment produced very d i f f e r e n t outcomes from those produced by the variable delay procedure used i n the previous experiments. Furthermore, the present results showed that, with a stepwise procedure, performance at a given delay i s very d i f f e r e n t during i n i t i a l sessions than during l a t e r sessions at the same delay: the choose short and choose long effects are pronounced during the f i r s t few sessions at a new delay, but disappear af t e r extended t r a i n i n g at that delay. In view of the present re-s u l t s , i t seems l i k e l y that the f a i l u r e of Cohen et a l . (1981) to demonstrate a choose short e f f e c t was due, at l e a s t i n part, to t h e i r f a i l u r e to report and analyze the data from the f i r s t few sessions at each delay. F i n a l l y , the results of the 0-sec sample tests confirmed the expectation that the absence of a sample would be treated more l i k e a short sample than a long sample. Although Church (1980) has interpreted similar findings i n terms of "biased guessing," he does not o f f e r any explanation of why guesses should be biased consistently toward "short." In contrast, 94 the process of stimulus generalization readily can account for the fact that 0-sec samples are treated more like short samples than long samples. 95 GENERAL DISCUSSION The f i r s t section of the general discussion i s comprised of a summary and discussion of the res u l t s of the present investigations, and t h e i r implications for the experimental study of working memory i n animals. The second section con-s i s t s of a discussion of the implications of the present research for theories and models of working memory processes. I Summary of the results and t h e i r  methodological implications The general purpose of the experiments i n Part I was to investigate pigeons' working memory for event duration. In these experiments, v a r i a t i o n s of the DSMTS procedure were used in which event durations served as the sample s t i m u l i . The f i r s t experiment revealed that when a variable delay was interposed between the sample and the comparison s t i m u l i , pigeons showed a strong tendency to choose the comparison stimulus associated with the short sample at 10- or 20-sec delays, but not at shorter delays. This choose short e f f e c t was found to occur with both food and l i g h t durations as the sample s t i m u l i (Experiments 1 and 2), with both naive and experienced subjects (Experiments 1 and 2), and with both a two-choice (Experiments 1 and 2) and a three-choice procedure (Experiment 3). Furthermore, i n addition to being s t a t i s t i -96 c a l l y r e l i a b l e , the choose short e f f e c t occurred i n every subject i n each experiment. Experiments 4 and 5 extended the results of the f i r s t three experiments by showing that the choose short e f f e c t was re f l e c t e d i n a s h i f t i n the point of subjective equality after a 20-sec delay (Experiment 4), and that the choose short e f f e c t also occurred after stepwise increases i n the delay (Experiment 5). In addition, the f i f t h experiment demonstrated that a choose long e f f e c t occurred after a stepwise decrease i n the delay. Both the choose short e f f e c t and the choose long e f f e c t diminished with extended exposure to a given delay. F i n a l l y , Experiment 5 also demonstrated that af t e r a 0-sec sample the birds tended to choose the "short" comparison. The present experiments constitute the f i r s t systematic investigation of pigeons' working memory for event duration. Although previous research has shown that the type of stimulus used as the sample may have a s i g n i f i c a n t e f f e c t on performance on working memory tasks (e.g., Farthing et a l . , 1977), no phenomena similar to the choose short and choose long e f f e c t s have been reported to occur when more commonly-used s t i m u l i , such as colors or l i n e t i l t s , serve as the sample. Thus, memory for event duration may d i f f e r i n important ways from memory for other types of stimulus dimensions (cf. Church, 1980). The results of the present investigations have a number 97 of methodological implications for the study of working memory. F i r s t , the present results suggest that the routine practice of analyzing and reporting o v e r a l l accuracy as opposed to accuracy after each sample i n delayed matching tasks may some-times obscure important phenomena .(cf. Sidman, 1980) . Had the present data been analyzed i n terms of o v e r a l l accuracy alone, the choose short and choose long e f f e c t s would not have been detected. Second, the re s u l t s of Experiment 1 suggest that chance l e v e l o v e r a l l accuracy i s not necessarily i n d i c a t i v e of random choice behaviour. In a two-choice procedure, o v e r a l l accuracy may fluctuate around chance l e v e l either because the subjects are choosing randomly or because they are choosing one sample more often than the other (e.g., the choose short e f f e c t ) . Third, and most importantly, the present results i l l u s t r a t e that the procedure used to manipulate the delay may have a profound e f f e c t upon,the outcome of studies of working memory. In the present experiments, the variable delay procedure c l e a r l y yielded d i f f e r e n t r e s u l t s from those obtained with the stepwise delay procedure. Under the variable delay procedure, the choose short e f f e c t was stable over test sessions, whereas under the stepwise delay procedure the choose short e f f e c t occurred only during the f i r s t few sessions following a delay increase. In addition, stepwise decreases i n the delay produced a temporary choose long e f f e c t , which 98 had never been observed under the variable delay procedure. These results suggest that the use of only one type of delay manipulation procedure or the practice of reporting only data from the l a s t few sessions (e.g., Cohen et a l . , 1981; Wilkie & Spetch, 1978) may sometimes lead to inconsistent or erroneous conclusions about the processes of working memory i n animals. II The subjective shortening model and i t s implications  for theories of working memory i n animals In order to account for the choose short e f f e c t observed i n the f i r s t experiment, a model of memory for event duration was proposed. This "subjective shortening model" consists of two es s e n t i a l components: working memory and reference memory. According to the model, a reference memory of the sample durations and th e i r association with the comparison sti m u l i i s established during i n i t i a l t r a i n i n g . This reference memory, once established, remains r e l a t i v e l y stable within and between t r i a l s provided that a substantial proportion of the t r i a l s consist of the delay i n t e r v a l at which the animal was o r i g i n a l l y trained. On the other hand, the working memory of the sample undergoes a systematic change within the t r i a l when there i s a delay between the sample and comparison s t i m u l i : the remembered duration of the sample shortens over the delay. It i s the discrepancy between the reference memory of the sample 99 and the working memory of the sample that produces the choose short e f f e c t . As the working memory of the long sample shortens, i t becomes more similar to the reference memory of the short sample, thereby producing a tendency to respond as though a long sample was short. In a variable delay procedure that includes a substantial proportion of t r i a l s containing the delay used i n o r i g i n a l t r a i n i n g (e.g., 0-sec), reference memory should be r e l a t i v e l y stable. Consequently, the d i s -crepancy between the reference memory of the sample and the working memory of the sample a f t e r a long delay should be main-tained over test sessions. The res u l t s of the f i r s t three experiments were completely consistent with t h i s subjective shortening model. A number of predictions derived from the subjective shortening model were tested i n Experiments 4 and 5. Experi-ment 4 tested the prediction that the point of subjective equality would s h i f t to a longer sample duration as the delay between the sample and comparisons was increased. This out-come was predicted on the basis of the model i n the following way. If the subjective duration of the sample i n working memory shortens over the delay, then aft e r a long delay the sample duration that would be treated as the subjective mid-point between the short and long samples i n reference memory would be longer. This prediction was confirmed by the results of Experiment 4, which showed that the point of subjective 100 equality was sh i f t e d to a s i g n i f i c a n t l y longer duration after a 20-sec delay. Thus, Experiment 4 provided further support for the subjective shortening model. It was Experiment 5, however, that provided the most stringent test of the subjective shortening model. This Experiment tested a number of predictions derived from the model concerning the outcome of stepwise manipulations of the delay. The p i v o t a l aspect of the model from which the pre-dictions were derived was the interplay between working memory and reference memory. According to the model, the s t a b i l i t y of a previously established reference memory depends upon the proportion of t r i a l s that consist of the delay used during the i n i t i a l t r a i n i n g . When the delay i s changed to a value that remains constant on a l l t r i a l s for several sessions, as i n a stepwise procedure, the reference memory of o r i g i n a l t r a i n i n g should not be maintained. Instead, a new reference memory that i s based upon the working memory of the samples at that constant delay gradually should develop. As a r e s u l t , the discrepancy between the working.memory and the reference memory of the samples gradually should diminish. Moreover, once a reference memory has been established at a long delay, a subsequent decrease i n the delay should produce a temporary discrepancy between working memory and reference memory that i s i n the opposite d i r e c t i o n to that produced by an increase i n the delay. 101 Thus, i t was predicted that stepwise increases i n the delay would produce a temporary choose short e f f e c t and that aft e r extended t r a i n i n g at a given delay the choose short e f f e c t would diminish and o v e r a l l accuracy would improve. Furthermore, stepwise decreases i n the delay would produce a temporary choose long e f f e c t , which also would diminish aft e r extended t r a i n i n g at a given delay. Each of these predictions was confirmed by the results of Experiment 5. Because none of these effects could have been predicted simply on the basis of the results of the previous experiments, which had revealed only a stable choose short e f f e c t , t h i s experi-ment represented the most rigorous test of the predictive power of the subjective shortening model. The present r e s u l t s do not appear to be e a s i l y i n t e r -pretable within the context of other current views of working memory processes. D'Amato's (1973) temporal discrimination hypothesis assumes that performance on delayed matching tasks i s based upon a discrimination of the r e l a t i v e recency of the samples. This view does not provide any basis for predicting that one stimulus would be chosen more often than the other, given that both samples occur equally often i n the session. Thus, a temporal discrimination hypothesis of working memory cannot read i l y accommodate any of the present r e s u l t s . It i s also d i f f i c u l t to imagine how coding views of 102 working memory can deal with the present findings. According to some coding views (e.g., Farthing et a l . , 1977), the encod-ing process occurs gradually during exposure to the sample. Long duration samples are assumed to be remembered better than short duration samples because long samples are encoded better than short samples. Thus, th i s coding view would lead to the prediction that accuracy af t e r long samples would be higher than accuracy after short samples. Clearly, the choose short e f f e c t i s not consistent with t h i s coding view of working memory. Furthermore, there i s no reason to assume that a coded representation of one stimulus duration should change systematically into a coded representation of another stimulus duration over a delay i n t e r v a l . Thus, without such an ad hoc assumption, the present results could not be predicted e a s i l y on the basis of coding views of working memory. Wagner's information processing model does contain an assumption that some stimuli are maintained i n short-term memory longer than other stimuli because they "command" rehearsal to a greater degree. Although i t would be possible to assume that short stimuli are rehearsed more e f f e c t i v e l y than long s t i m u l i , several features of the present data suggest that d i f f e r e n t i a l rehearsal i s not the basis of the choose short e f f e c t . F i r s t , the choose short e f f e c t often was characterized by below chance l e v e l accuracy af t e r long samples. Because f a i l u r e to rehearse the sample should at worst produce 103 chance l e v e l accuracy, these r e s u l t s suggest that the long samples were not forgotten, but instead were remembered as being short. The second feature of the present data that argues against a d i f f e r e n t i a l rehearsal interpretation of the present findings was the occurrence of a choose long e f f e c t i n Experiment 5. Explaining why short samples should be re-hearsed better af t e r delay increases but long samples after delay decreases, would constitute a serious challenge for a d i f f e r e n t i a l rehearsal interpretation of the present r e s u l t s . Roberts and Grant's (.19.76) trace decay theory also cannot e a s i l y accommodate the present r e s u l t s . According to t h e i r theory, longer duration samples should produce a stronger trace and therefore be remembered better than short duration samples. Accordingly, accuracy after long samples should be better than accuracy a f t e r short samples (cf. Cohen et a l . , 1981). The choose short e f f e c t c l e a r l y i s inconsistent with t h i s expectation. Although the choose short e f f e c t might be explained by a modified trace decay theory, i n which the process of subjective shortening i s conceptualized as a decay of the sample trace along the duration dimension, trace decay theory s t i l l could not accommodate the results of Experiment 5 unless i t also included a reference memory component sim i l a r to that of the subjective shortening model. The r e s u l t s of Experiment 5 are also problematic for a 104 "biased guessing" interpretation of the choose short e f f e c t , such as that proposed by Church (1980). I t i s not immediately obvious why guesses should be biased towards short after the delay i s increased, become progressively less biased during extended exposure to a given delay, and then be biased towards long af t e r the delay i s decreased. In short, the subjective shortening model provides the best f i t to the present r e s u l t s . None of the other conceptuali-zations of working memory processes, as they are currently formulated, can accommodate the present findings. There are a number of possible reasons why the res u l t s of the present investigations cannot be interpreted e a s i l y within the context of other current views of working memory. One p o s s i b i l i t y i s that other current views do not represent accurate descriptions of the processes involved i n working memory. However, i t seems unlikel y that a l l of these views are completely inaccurate, given that each of them seems to account well for some working memory phenomena. A second p o s s i b i l i t y i s that the processes involved i n working memory for event duration may be d i f f e r e n t from those involved i n memory for the more commonly studied stimulus dimensions such as wavelength or l i n e orientation. Thus, the present r e s u l t s may represent a l i m i t a t i o n on the generality of other models of working memory, rather than a challenge to t h e i r i n t e r n a l 105 v a l i d i t y . In fact, i t may be necessary to develop d i f f e r e n t models to account for the processes of working memory for d i f f e r e n t types of stimulus dimensions or d i f f e r e n t types of memory tasks (cf. Honig, i n press; Riley et a l . , i n press). In the human psychophysical l i t e r a t u r e , c e r t a i n stimulus dimensions have been c l a s s i f i e d as metathetic because subjects perceive a change i n the physical stimulus as a change i n i t s qua l i t y , whereas other stimulus dimensions have been c l a s s i f i e d as prothetic because subjects perceive a change i n the physical stimulus as a change i n i t s quantity. Although i t may never be possible to determine whether an animal subject perceives a q u a l i t a t i v e or a quantitative change i n a physical stimulus, the d i s t i n c t i o n between dimensions that have been c l a s s i f i e d as metathetic and those that have been c l a s s i f i e d as prothetic on the basis of human reports may have some u t i l i t y for the development of models of animal memory processes. For example, i t i s possible that c e r t a i n models of animal memory may best describe the processes involved i n memory for "metathetic" s t i m u l i , and.that other models may be needed to describe memory for "prothetic" s t i m u l i . The subjective shortening model appears to describe the processes involved i n pigeons' memory for event duration, a dimension that has been c l a s s i f i e d as prothetic. Although i t seems u n l i k e l y that a process of subjective shortening -is 106 involved i n working memory for sti m u l i that are varied along a metathetic continuum, such as color, i t i s possible that subjective shortening may be involved i n memory for stimuli that are varied along other prothetic dimensions, such as size, or length. Whether or not the subjective shortening model w i l l provide an adequate description of memory for any stimulus dimensions other than duration remains to be deter-mined. The concept of subjective shortening i n memory i s not en t i r e l y new (cf. Frankenhaeuser, 1959; Ornstein, 1969). For example, Frankenhaeuser (19 59) believed that memory of a time i n t e r v a l depended upon retention of the stimulus events that f i l l e d the i n t e r v a l ; i f any of these stimuli were f o r -gotten, the i n t e r v a l i t s e l f would be remembered as being shorter. In support of t h i s view, Frankenhaeuser reported that human subjects' estimates of past time were consistently smaller than the present time estimates on which they were based. Furthermore, she reported that "a close correspondence was found between amount of time retained and number of stimuli retained." (Frankenhaeuser, 1959, p. 121). The idea that memory of time undergoes systematic change or d i s t o r t i o n has also been discussed i n r e l a t i o n to a phenome-non reported i n the human psychophysical l i t e r a t u r e c a l l e d the "time-order-error" (cf. A l l a n , 1979). When two sti m u l i are 107 presented successively to a subject whose task i s to compare them along a p a r t i c u l a r prothetic dimension (e.g., weight, duration, loudness), i t i s often found that the order of pre-sentation of the stimuli affects subjects' judgements of equality. For example, when two st i m u l i longer than 1 sec and of equal duration are presented successively, subjects often judge the f i r s t stimulus as shorter than the second (e.g., Hellstrom, 1977; Woodrow, 1935). Kohler (1923) has suggested that t h i s type of error occurs because the second stimulus i s being compared to a "faded" trace of the f i r s t stimulus, leading to an underestimation of the f i r s t stimulus. In support of his fading trace theory, Kohler (19 23) found that the degree to which the f i r s t stimulus was underestimated increased as a function of the length of the interstimulus i n t e r v a l . The idea that remembered durations may shorten over time also has been mentioned i n the animal memory l i t e r a t u r e (Church, 1980; Honig, i n press). Church (1980) discussed t h i s idea i n terms of an "internal clock" mechanism, which he assumes i s the basis of rat s ' a b i l i t y to time event durations. He suggested that shortening of subjective durations could occur i f the int e r n a l clock was reset gradually toward zero during a retention i n t e r v a l : a f t e r a long retention i n t e r v a l the rats would respond on the basis of a "shortened" ( i . e . , 108 p a r t i a l l y reset) clock setting. Church, however, dismissed the idea that the i n t e r n a l clock i s reset gradually and con-cluded that the subjective duration of events i n rats' memory does not shorten over time. Honig (in press) also mentioned the p o s s i b i l i t y that the remembered duration of p r i o r s t i m u l i might be "foreshortened." He suggested that t h i s process of foreshortening might not have been observed i n Church's study because the rats may have solved the task by remembering a coded response decision rather than the duration of the stimulus. I t i s important to note that the subjective shortening model proposed i n the present investigations i s an attempt to describe the processes involved i n memory for event dura-ti o n rather than an attempt to explain the s p e c i f i c mechanisms responsible for subjective shortening. There are several possible mechanisms that could produce subjective shortening, such as the gradual resetting of an i n t e r n a l clock (cf. Church, 1980), decay of a stimulus trace along the time dimension (cf. Roberts & Grant, 1976), or forgetting of events that f i l l the i n t e r v a l (cf. Frankenhaeuser, 1959; Ornstein, 1969). The subjective shortening model, as i t i s presently formulated, does not specify which, i f any, of these possible mechanisms may underly the subjective shortening process. F i n a l l y , the two-process framework of the subjective shortening model, which emphasizes the relationship between 109 working memory and reference memory, may have general u t i l i t y for the development of models of working memory i n animals. The two-process framework i s derived from the idea that per-formance on a working memory task i s determined not only by the processes of working memory but also by the content of reference memory (cf. Honig, 1978; i n press). Changes i n reference memory may a f f e c t performance on a working memory task, and changes i n working memory may gradually a l t e r the content of reference memory. This i n t e r a c t i v e , two-process framework was an esse n t i a l aspect of the subjective shortening model; without i t , many of the results of Experiment 5 could not have been predicted. Furthermore, the emphasis on thi s i n t e r a c t i v e r e l a t i o n s h i p between working memory and reference memory d i f f e r e n t i a t e s the subjective shortening model from most other models of working memory. The fac t that single-process models of working memory, such as trace decay, cannot e a s i l y accommodate the present data does not necessarily mean that they are incorrect, but rather that they may be incom-plete. For example, trace decay theory might e a s i l y accommo-date the present data by adopting a two-process in t e r a c t i v e framework. The shortening of time i n working memory could be conceived of as a process of decay along the time dimension; then, by assuming that the subject responds on the basis of the s i m i l a r i t y between a working memory trace of the sample 110 and a representation of the sample i n reference memory, the results of the present experiments could be predicted. Thus, an awareness of the ways i n which reference memory and working memory may interac t might f a c i l i t a t e the development of more viable models of working memory i n animals. I l l REFERENCES Alla n , L. G. The perception of time. Perception and  Psychophysics, 1979., 26^ , 340-354. Atkinson, R. C , & S h i f f r i n , R. M. Human memory: A proposed system and i t s control processes. In K. W. Spence & J. T. Spence (eds.), The psychology of learning and  motivation. V o l . 2, New York: Academic Press, 1968. Blough, D. S. Delayed matching i n the pigeon. Journal of the Experimental Analysis of Behavior, 1959, 2_, 151-160. Bolles, R, C. Learning, motivation and cognition. In W. K. Estes (ed.), Handbook of learning and cognitive processes. Vol. 1. H i l l s d a l e , N.J.: Lawrence Erlbaum, 1975. Bolles, R. C. Some relationships between learning and memory. In D. L. Medin, W. A. Roberts, & R. T. David (eds.), Processes of animal memory. H i l l s d a l e , N.J.: Lawrence Erlbaum, 1976. Capaldi, E. J . A sequential hypothesis of instrumental learning. In K. W. Spence & J, T. Spence (eds,), The  Psychology of learning and motivation. Vol. 1. New York: Academic Press, 19 67. Capaldi, E. J . , & Cogan, D. Magnitude of reward and d i f f e r e n -t i a l stimulus consequences. Psychological Reports, 1963, 13_, 85-86. Capaldi, E. J . , & Spivey, J . E. I n t e r t r i a l reinforcement and after e f f e c t s at 24-hour i n t e r v a l s . Psychonomic Science, 1964, 1, 181-182. Carter, D. E., & Werner, T. J . Complex learning and informa-t i o n processing by pigeons: A c r i t i c a l analysis. Journal of the Experimental Analysis of Behavior, 1978, 29, 565-601. Church, R. M. Short-term memory for time i n t e r v a l s . Learning  and Motivation, 1980, .11, 208-219. Church, R. M., Getty, D, J., & Lerner, N. D. Duration discrim-ination by rats . Journal of Experimental Psychology: Animal Behavior. Processes, 1976, 2, 303-312. 112 Cohen, S. L. , C a l i s t o , G. , & Lentz, B. E. Comparisons of sample st i m u l i i n delayed symbolic matching-to-sample: Some res u l t s and implications. The Psychological  Record, 1981, 31, 77-93. C o l w i l l , R. M. & Dickinson, A. Short-term retention of "surprising" events by pigeons. Quarterly Journal of  Experimental Psychology, 1980, 3_2, 539-556. Cook, R. G. Retroactive interference i n pigeon short-term memory by a reduction of ambient illumination. Journal  of Experimental Psychology: Animal Behavior Processes, 1980, 6, 326-338. Coren, S., Porac, C., & Ward, L. M. Sensation and perception. New York: Academic Press, 1979, D'Amato, M. R. Delayed matching and short-term memory i n monkeys. In G. H. Bower (ed.), The psychology of  learning and motivation: Advances i n research and  theory. Vol. 7. New York: Academic Press, 1973. D'Amato, M. R., & Worsham, R. W. Retrieval cues and short-term memory i n capuchin monkeys, Journal of Comparative  and Physiological Psychology, 1974, 86_, 274-282. Elsmore, T. F. Duration discrimination: Effects of probabil-i t y of stimulus presentation. Journal of the Experimen- t a l Analysis of Behavior, 1972, !L8, 465-469. Etscorn, F., & Stephens, R. Establishment of conditioned taste aversions with a 24-hour CS-US i n t e r v a l . Physiological Psychology, 1973, 1, 251-253. Farthing, G. W., Wagner, J . M., Gilmour, S., & Waxman, H. M. Short-term memory and information processing i n pigeons. Learning and Motivation, 1977, 8^, 520-532. Frankenhaeuser, M. Estimation of time. Stockholm: Almquist & Wiksel, 1959. Gabriel, M. Intersession delay i n avoidance extinction to t r a i n i n g and generalization s t i m u l i . Psychonomic  Science, 1967, 9, 243-244. Garcia, J., Ervin, F. R., & Koelling, R. A. Learning with prolonged delay of reinforcement. Psychonomic Science 1966, 5, 121-122. 113 Gleitman, H. Forgetting of long-term memories i n animals. In W. K. Honig & P. H. R. James (eds.), Animal memory. New York: Academic Press, 1971. Gleitman, H., & Holmes, P. Retention of incompletely learned CER i n rats . Psychonomic Science, 1967, 1_, 19-20. Gleitman, H., & Jung, L. Retention i n ra t s : The e f f e c t of proactive interference. Science, 1963, 142, 1683-1684. Grant, D. S. E f f e c t of sample presentation time on long-delay matching i n the pigeon. Learning and Motivation, 1976, 7, 580-590. Grant, D. S. Stimulus control of information processing i n pigeon short-term memory. Learning and Motivation, 1981, 12, 19-39. Grant, D. S., & Roberts, W. A. Sources of retroactive i n -h i b i t i o n i n pigeon short-term memory. Journal of  Experimental Psychology: Animal Behavior Processes, 1976, 2, 1-16. Guthrie, E. R. The psychology of learning. New York: Harper, 1935. Hellstrom, A. Time-errors are perceptual. Psychological  Research, 1977, 39_, 345-388. Herman, L. M. Interference and auditory short-term memory i n the bottle-nosed dolphin. Animal Learning &  Behavior, 1975, 3_, 43-48. Honig, W. K. Studies of working memory i n the pigeon. In S. H. Hulse, H. Fowler, & W. K. Honig (eds.), Cognitive  processes i n animal behavior. H i l l s d a l e , N.J.: Erlbaum, 1978. Honig, W. K. Working memory and the temporal map. In N. E. Spear & R. R. M i l l e r (eds.), Information processing i n  animals: Memory mechanisms. H i l l s d a l e , N.J.: Lawrence Erlbaum, i n press. Honig, W, K., & Wasserman, E. A, Performance of pigeons on delayed simple and conditional discriminations under equivalent t r a i n i n g procedures. Learning and Motivation, 1981, 12_, 149-170. 114 H u l l , C. L. Princ i p l e s of behavior. New York: Appleton-Century-Crofts , 1943. Hunter, W. S. The delayed reaction i n animals and children. Behavior Monographs, 1913, 2, 1-86. Jarrard, L. E. (ed..). Cognitive processes of nonhuman primates. New York: Academic Press, 1971. Jarrard, L. E., & Moise, S. L. Short-term memory i n the mon-key. In L. E. Jarrard (ed.), Cognitive processes of  nonhuman primates. New York: Academic Press, 1971, Kamin, L. J. The retention of an incompletely learned avoidance response. Journal.of Comparative and  Physiological Psychology, 1957, 50, 457-460. Kinchla, J . Discrimination of two auditory durations by pigeons. Perception and Psychophysics, 1970, 8_, 299-307. Kohler, W. Zur Theorie des Suksessiuvergleichs und der Z e i t f e h l e r . Psychologie Forschung, 1923, 4_, 115-175. Ladieu, G. The e f f e c t of length of the delay i n t e r v a l upon delayed alternation i n the albino r a t . Journal of  Comparative Psychology, 1944, 37, 273-286. Lashley, K. S. In search of the engram. Symposia of the  Society for Experimental Biology, 1950, 4_, 454-582. Lett, B. T. Delayed reward learning: Disproof of the t r a d i t i o n a l theory. Learning and Motivation, 1973, 4_, 237-246. Lett, B. T. Long delay learning i n the T-maze, Learning  and Motivation, 1975, 6_, 80-90. Lydersen, T., Perkins, D., & Chairez, H. Effects of fi x e d -r a t i o sample and choice response requirements upon oddity matching. Journal of the Experimental Analysis  of Behavior, 1977, 25_, 97-101. Maier, N. R. F., & Schneirla, T. C. P r i n c i p l e s of animal  psychology. New York: McGraw-Hill, 1935. 9 115 Maki, W. S., J r . Sustained attention: Stimulus control determined by the schedule of cue production i n t e r -acting with cue duration. Animal Learning & Behavior, 1975, 3, 312-316. Maki, W. S., J r . Pigeons' short-term memories for surprising vs. expected reinforcement and nonreinforcement. Animal Learning & Behavior; 1979, 7_, 31-37. Maki, W. S., J r . , & Hegvik, D. K. Directed forgetting i n pigeons. Animal Learning & Behavior, 1980, 8_, 567-574. Maki, W. S., J r . , Moe, J. C , & Bierley, C. M. Short-term memory for s t i m u l i , responses, and rein f o r c e r s . Journal  of Experimental Psychology: Animal Behavior Processes, 1977, 3, 156-177. Nelson, K. R., & Wasserman, E. A. Temporal factors influencing the pigeon's successive matching-to-sample performance: Sample duration, i n t e r t r i a l i n t e r v a l , and retention i n t e r v a l . Journal of the Experimental Analysis of Behavior, 1978, 30_, 153-162. Olton, D. S. Characteristics of s p a t i a l memory. In S. H. Hulse, H. Fowler, & W. K. Honig (eds.), Cognitive  processes i n animal behavior. H i l l s d a l e , N.J.: Erlbaum, 1978. Ornstein, R. E. On the experience of time. London: Penguin, 1969. Pavlov, I. P. Conditioned reflexes. (Translated by G. V. Anrep.) London: Oxford University Press, 1927. Pe r i k e l , J . J., Richelle, M., & Maurissen, J. Control of keypecking by the duration of a v i s u a l stimulus. Journal of the Experimental Analysis of Behavior, 1974, 2J2, 131-134. Petrinovieh, L., & Bolles, R. C. Delayed alternation: Evidence for symbolic processes i n the rat. Journal  of Comparative and Physiological Psychology, 1957, 50, 363-365. Premack, D. Language i n a chimpanzie? Science, 1971, 172, 808-822. 116 Reynolds, G. S., & Catania, A, C. Temporal discrimination i n pigeons. Science, 1962, 135, 314-315. Revusky, S. The role of interference i n association over a delay. In W. K. Honig & P. H. R. James (eds.)., Animal memory. New York: Academic Press, 19 71. Riley, D. A., Cook, R. G,, & Lamb, M. R. A c l a s s i f i c a t i o n and analysis of short term retention codes i n pigeons. In G. H. Bower (ed.). , The psychology of learning and  motivation. New York: Academic Press, i n press. R i l l i n g , M., Kendrick, D. F., & Stonebraker, T. B. Stimulus control of forgetting: A behavioral analysis. In M. L. Commons, A. R. Wagner, & R. J. Herrnstein (eds.), Quantitative studies of operant behavior: Acquisition . Ballinger Publishing Co., i n press. Roberts, W. A. Short-term memory i n the pigeon: Effects of re p e t i t i o n and spacing. Journal of Experimental  Psychology, 1972, 94, 74-83. Roberts, W. A., & Grant, D. S. Short-term memory i n the pigeon with presentation time precisely controlled. Learning and Motivation, 1974, 5, 393-408. Roberts, W. A. & Grant, D. S. Studies of short-term memory i n the pigeon using delayed matching to sample. In D. L. Medin, W. A. Roberts, & R. T. Davis (eds.), Processes of animal memory. H i l l s d a l e , N.J.: Erlbaum, 1976. Roitblat, H. L. Codes and coding processes i n pigeon short-term memory. Animal Learning & Behavior, 1980, 8_, 341-351. Ruggiero, F. T. & Flagg, S. F. Do animals have memory? In D. L. Medin, W. A. Roberts, & R. T. David (eds.), Processes of animal memory. H i l l s d a l e , N.J.: Erlbaum, 1976. Scharf, B. (ed.) Experimental sensory psychology. Glenview, 111.: Scott, Foresman and Co., 1975. Seligman, M. E. P. On the generality of the laws of learning. Psychological Review, 1970, 406-418. 117 Shimp, C. P. & M o f f i t t , M. Short-term memory i n the pigeon: Delayed-pair-comparison procedure and some r e s u l t s . Journal of the Experimental Analysis of Behavior, 1977, 28, 13-25. Sidman, M. A note on the measurement of conditional discrim-ination. Journal of the Experimental Analysis of Behavior, 1980, 33, 285-289. Spear, N. E. The processing of memories: Forgetting and retention. H i l l s d a l e , N.J.: Lawrence, Erlbaum, 1978. Spence, K. W. The role of secondary reinforcement i n delayed reward learning. Psychological Review, 1947, 5_4_, 1-8. Spetch, M. L. & Wilkie, D. M. Duration discrimination i s better with food access as the signal than with l i g h t as the s i g n a l . Learning and Motivation, 1981, 12, 40-64. Spetch, M. L., Wilkie, D. M., & Skelton, R. W. Control of pigeons 1 pecking topography by a schedule of alternating food and water reward. Animal Learning & Behavior, 1981, 9, 223-229. Stubbs, A. The discrimination of stimulus duration by pigeons. Journal of the Experimental Analysis of Behavior, 1968, 11, 223-238. Terry, W. S. & Wagner, A. R. Short-term memory for "surpris-ing" vs. "expected", unconditioned s t i m u l i i n Pavlovian conditioning. Journal of Experimental Psychology:  Animal Behavior Processes, 1975, 1_, '122-133. Tinklepaugh, O. L. An experimental study of representative factors i n monkeys. Journal of Comparative Psychology, 1928, 8, 197-236. Tolman, E. C. Purposive behavior i n animals and men. New York: Appleton-Century-Crofts, 1932. Wagner, A. R. Expectancies and the priming of STM. In S. H. Hulse, H. Fowler, & W. K. Honig (eds.), Cognitive  processes i n animal behavior. H i l l s d a l e , N.J.: Erlbaum, 1978. Wagner, A. R., Rudy, J. W., & Whitlow, J. W. Rehearsal i n animal conditioning. Journal of Experimental Psychology, 1973, 97, 407-426. 118 Wallace, J . , Steinert, P. A., Scobie, S. R., & Spear, N. E. Stimulus modality and short-term memory i n ra t s . Animal Learning & Behavior, 1980, 8_, 10-16. Wilkie, D. M. Delayed sybolic matching to sample i n the pigeon. The Psychological Record, 1978, 28_, 463-469. Wilkie, D. M. & Spetch, M, L. The e f f e c t of sample and comparison r a t i o schedules on delayed matching to sample i n the pigeon. Animal Learning & Behavior, 1978, 6, 273-278. Wilkie, D. M. & Spetch, M. L. Pigeons' delayed-matching-to-sample errors are not always due to forgetting. Behaviour Analysis Letters, i n press. Wilkie, D. M. & Summers, R. J . Pigeons' s p a t i a l memory: Factors af f e c t i n g delayed matching of key locations. Journal of the Experimental Analysis of Behavior, i n press. Wilkie, D. M., Summers, R. J., & Spetch, M. L. The e f f e c t of delay i n t e r v a l stimuli on delayed symbolic matching to sample i n the pigeon. Journal of the Experimental  Analysis of Behavior, 1981, 35, 153-160. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0095636/manifest

Comment

Related Items