Context dependent memory: The role of environmental cues

Transcription

1 Ashland University From the SelectedWorks of Mitchell Metzger, PhD 2002 Context dependent memory: The role of environmental cues Mitchell M. Metzger, Ashland University Available at:

2 chapter 2 Context Dependent Memory: The Role of Environmental Cues Mitchell M. Metzger Pennsylvania State University - Shenango campus On the evening before classes begin, you attend a party at a friend s house to celebrate (or mourn) the start of a new semester. While at the party, you briefly meet several acquaintances of your friend, and after these short-lived encounters, you leave to get a good night s sleep before your morning classes. The next day as you enter the lecture room for your first psychology class, you are surprised that a fellow classmate begins talking to you as if she has met you before, yet you have no conceivable notion of who she might be. After a few uncomfortable moments while you pretend that you really do know this person, she mentions something about last night s party, and the memory of your meeting now comes rushing back. A scenario similar to this has surely happened to each of us at one time or another. To what can we attribute our failed memory for an event that happened so recently? Did we forget this person because we failed to properly encode the memory of our meeting? Was our forgetting due to a problem with the storage of memory? Or, was our absentmindedness the result of a retrieval deficit because of a lack of appropriate retrieval cues? Within any learning situation, there are a number of stimuli that are present during that particular episode. For example, in classical conditioning, the conditioned stimuli (CS) and unconditioned stimuli (UCS) are of primary interest. In operant conditioning, the relationships between our behavior and the stimuli or consequences that follow them are important. With observational learning, learning occurs by viewing a particular behavior in the presence of a number of distinct stimuli. One thing that each of

3 the above forms of learning have in common is that they all occur in the presence of specific contextual stimuli, or context. According to Domjan (1993), contextual stimuli are background stimuli in the presence of which more discrete stimuli may be presented. In other words, if you are reading this textbook in your dorm room on a Wednesday evening, the contextual stimuli would include the physical properties of your immediate environment. Such stimuli might be your desk and other furniture, paintings and posters hanging on the wall, the color of the paint and carpet in your room, and the overpowering aroma of raspberry that comes from a burning candle your roommate insists on lighting every evening. These stimuli all provide a context for the learning episode that you are currently experiencing, that is, learning the material presented in this chapter. As we will see in the following pages, contextual stimuli play a vital role in learning and the expression of memory, even though you may (incorrectly) assume that these stimuli are irrelevant to a learning situation. Put quite simply, why should the context where you learn something make a difference in how easily and accurately you remember that material? Getting back to the original scenario at the beginning of this chapter, is it possible that the mismatch of contextual cues that were present during the initial encounter with your new friend be the reason why your memory failed at your subsequent meeting in the lecture room? As we will see, there is a long line of research that supports the idea that contextual cues are important for the expression of memory, as studies using both animals and humans have revealed the powerful effects of this important class of stimuli. Before beginning our discussion of context, we must first acknowledge the difference between two very different types of contextual cues, external and internal. External context, which is the focus of this chapter, refers to the environmental cues that are associated with a learning episode. In other words, external context refers to the environmental setting where learning and memory occurs. Internal context refers to an internal state that may also become associated with a learning episode. For example, sleepiness, changes in mood, and the influence of certain drugs produce internal states that may also become associated with a learning episode. This chapter will focus entirely on external context, as the discussion of internal context is reserved for a later chapter on State Dependent Retention (see Chapter 3). It has long been known that environmental cues serve as important factors in the expression of memory. Carr (1925) stated that memories can only be remembered in the presence of stimuli with which they were originally associated. As an example, he stated that subjects who memorized material in a certain room and in the presence of a given experimenter make poorer records when tested for recall by another person in a different room than they do when tested under the conditions that obtained when the material was memorized (pg. 251). In a separate account, McGeogh (1942) hypothesized that retention decrements (forgetting) occurred because of three reasons, one of which was the presence of altered stimulating conditions. As can be seen from these two examples, the research on contextual influence and its effect on memory has long been of interest to psychologists. What these examples illustrate is that memory is disrupted when tested in the absence of environmental cues that were present at the initial learning episode. We will refer to this phenomenon as the context shift effect, the disruption of memory when testing occurs in the presence of altered external stimuli. You might be asking yourself why a simple change in context could have such a drastic effect on the retrieval of memory. In a their account of memory processes, Spear

4 Figure Mean rectal body temperature of rats receiving brief cold exposure on Days 1, 3, 5, 7, and 9. On Days 2, 4, 6, 8, and 10, all subjects were exposed to a context consisting of altered proximal and distal cues. On Day 11, Group DIFF received the cold treatment in the contexts used for the even numbered days, while the Group SAME received the cold treatment in the original adaptation context. From Kissinger and Riccio, and Riccio (1994) identified two reasons why context shift effects occur, or why memory is poor when tested in the presence of altered stimuli. Whenever we recall something, we recruit the use of retrieval cues, stimuli that are related to an experience that facilitates the recall of other information related to that experience (Domjan, 1993). In classical conditioning, for instance, the conditioned and unconditioned stimuli act as retrieval cues, but context also serves in this capacity. By removing the context associated with the initial learning episode (shifting or altering the original learning context), the environmental retrieval cues are lost. In essence, this may produce a scenario where there may be insufficient cues present to facilitate the recall of that particular memory. In other words, by changing the context we lose the ability to use the contextual stimuli as retrieval cues to aid in the expression of memory. A second possibility why the context shift effect occurs is that a change in context may, in fact, facilitate the recollection of other memories that may then interfere with the retrieval of the memory we are intending to recall. Whether the context shift effect occurs because of a lack of appropriate cues or because of the arousal of competing memories is not important for our discussion here. What remains clear is that alterations in context can create drastic disruptions in the ability to remember past experience. The second question you may be asking yourself is how do researchers know that changing the context during a retention test specifically affects memory, and is not affecting something else? Isn t it possible that the organism remembers how to perform a previously-learned act but simply chooses not to when placed in a different context? Could it be that the new (novel) environment in which the organism is tested creates a distraction that keeps the subject from performing the previously learned task? These are all legitimate questions, and data collected from a variety of studies provides us with information that allows us to rule out such interpretations. As one example, Kissinger and Riccio (1995) demonstrated a robust context shift effect for hypothermia tolerance,

5 even when the training context and the shifted testing context were presented equally during tolerance acquisition. In short, Kissinger and Riccio (1995) exposed rats to cold water (to induce hypothermia) every other day in one distinct context. On alternate days, the animals were exposed to a different context for an equal amount of time. After hypothermia tolerance had been established (shown by an increase in core body temperature over repeated trials), the animals were given exposure to cold water in the alternate context. Even when subjects had previously been exposed to the shifted context prior to the test (therefore reducing it s novelty), the context shift effect was still observed (See Figure 2.1). That is, those animals tested in a context different from that of tolerance acquisition showed a disruption in tolerance, while those tested in the original context retained their tolerance. The Kissinger and Riccio (1995) data, along with data from several other studies, successfully demonstrated that the novelty of stimuli in the altered context cannot account for the memory deficits that accompany testing in changed stimulus conditions (for a review of these concepts see Spear and Riccio, 1994). In a recent review of the literature on context dependent memory in humans, Smith and Vela (2001) stated that, by most accounts, the context shift effect is more robust in animals than humans. While this may be true, there is an impressive amount of data indicating that contextual stimuli play a vital role in the expression of memory in humans as well. In this next section, we will discuss several experiments in detail which have manipulated contextual stimuli and, therefore, produced memory deficits. We will begin our discussion of these studies with some experiments from the animal literature, and follow that discussion with several examples from the human literature. REVIEW OF THE CONTEXT SHIFT EFFECT IN ANIMALS Countless laboratories have demonstrated that changing environmental stimuli between training and testing has a disruptive effect on the memory performance of animals. Importantly, these observations of the context shift effect are not limited to a handful of experimental procedures, but are rather seen in a variety of paradigms and experimental situations. For instance, the context shift effect has been observed in studies of drug tolerance (Siegel, 1975); hypothermia tolerance (Harrod, Metzger, Stempowski, and Riccio, in press; Metzger, Harrod, Kissinger, and Riccio, 1998; Riccio, MacArdy, and Kissinger, 1991); appetitive runway tasks (Gisquet-Verrier and Alexinski, 1986); shock motivated avoidance procedures (McAllister and McAllister, 1963); and operant discrimination procedures (Thomas and Lopez, 1962). This wide range of experimental paradigms, ranging from aversive (using negative or painful stimuli) to appetitive (using positive or pleasant stimuli) procedures, demonstrate the robustness of the context shift effect, and further highlight the importance of context on the expression of many types of memory. For purposes of this chapter, we will now discuss in more detail the results of several of these previously mentioned studies. Our first example of the context shift effect comes from a study by Perkins and Weyant (1958) that examined the role of a change in context over an extended testing interval. Perkins and Weyant were interested in whether a delay in testing might affect the magnitude of the context shift effect. While the effect of delay on the context shift effect is a separate question that will be addressed later in this chapter (also see Riccio, Rabinowitz, and Axelrod, 1994 for a review), the results of the immediate testing conditions from Perkins and Weyant s study are relevant for the current discussion. In short, food deprived rats were trained to run on an elevated runway to receive food

6 reinforcement. Essentially, during the training procedure the rats were reinforced on some trials and not reinforced on other trials. As expected, during the course of training, the subjects began to run more quickly on reinforced rather than non-reinforced trials. After reaching a pre-determined training level, one-half of the rats were given a memory test in the original training apparatus, and the other one-half of rats were given the retention test in a different training apparatus. The only difference between the original apparatus and the shifted apparatus was the color of the walls of the runway (black walls rather than white). As you might predict, when subjects were tested in the runway that was the same runway in which they were trained, they ran quickly (which was an indication of good memory for training). However, Perkins and Weyant also observed that performance on the runway task was disrupted (slow running speed) when subjects were tested in a different apparatus from which they were trained. This result indicated that subjects tested in the different apparatus were able to distinguish the altered stimulus condition, and in essence, a change in context (changing the color of the runway) resulted in a performance decrement (poorer memory) for these subjects. While the Perkins and Weyant (1958) study demonstrated that context can affect performance for appetitive stimuli, a study by McAllister and McAllister (1963) demonstrated that contextual stimuli are also important in paradigms that use aversive stimuli. In their study, rats were trained to escape from a fearful stimulus. In the first phase, rats were conditioned to fear the presentation of a light by associating the light with a footshock. That is, whenever the light was presented, the rats would receive a mild footshock soon afterward. (As you might expect, it doesn t take animals very long to learn this relationship, as they will begin to show fearful behaviors (crouching and freezing) after a few number of trials.) After a fear response to the light was established, the rats were then required to jump a hurdle to escape from the fearful stimulus. In essence, the training apparatus was divided by a small barrier, and when the light was presented, subjects could escape from the footshock by jumping the hurdle to the other side of the apparatus. Similar to Perkins and Weyant (1958), after training was completed, the McAllister s implemented a contextual change condition to see if a change in context would affect the expression of what their subjects had previously learned. The conditions in this experiment consisted of testing subjects in a similar, yet different, apparatus than what was used during initial conditioning, or testing subjects in the same apparatus that was used during training. Consistent with the results of Perkins and Weyant (1958), the McAllister s reported that subjects given testing trials in the altered apparatus performed more poorly than those subjects tested in the identical apparatus that was used during training. That is, subjects tested in the original training apparatus jumped the hurdle to escape from the light more quickly than those animals tested in the shifted apparatus. (These results were later replicated and extended by Rohrbaugh and Riccio, 1968). One of the more powerful examples of a memory decrement resulting from a change in context occurred in the study of drug tolerance. Shepherd Siegel has accumulated a mountain of data suggesting morphine tolerance is, partly, a learned response that depends on contextual cues. Siegel s model of drug tolerance posits that tolerance to morphine is the result of classical conditioning, where the contextual cues present during drug administration serve as the conditioned stimulus, and the effect of the drug acts as the unconditioned stimulus. His model states that, provided that drug administration reliably occurs in the presence of consistent contextual cues, tolerance will

7 develop. However, if tolerance is acquired and the administration of the drug then occurs in the absence of the predictive environmental cues (the CS is removed), tolerance to morphine will be disrupted. In a striking display of the predictive power of his model, Siegel (1975) gave daily administration of morphine to rats in the same contextual surroundings, and tolerance was recorded by measuring the sensitivity of the subjects to a painful stimulus. After tolerance to the morphine had been established, Siegel tested onehalf of the subjects in the presence of the same contextual cues, and tested the other onehalf of subjects in different environmental conditions (they were moved to a different room for testing). As, I m sure, you ve already guessed, those subjects tested in the same environment retained their tolerance to the drug, while those tested in a different environment demonstrated a significant loss of tolerance. Importantly, the results obtained by Siegel (1975) have been replicated when different procedures and paradigms are used to measure tolerance to morphine, including procedures using hot plates, tail flick assessment, flinching/jumping assessment, and analgesia assessment Insert Siegel (1989) Figure 1 about here (Siegel, 1989). This wide variety of procedural manipulations attests to the robustness of Siegel s original findings. Similar to the results from drug tolerance research, the effect of contextual change on other forms of tolerance (e.g., hypothermia tolerance) has also been reported (Riccio et al., 1991; Metzger et al., 1998). Siegel s (1975) results are similar to those of Perkins and Weyant (1958) and McAllister and McAllister (1963) in that they all demonstrate changes in context can disrupt performance for a previously learned task. However, as you may have noticed, there is a subtle difference between the Siegel (1975) study and those described earlier. The changes in context implemented by Perkins and Weyant (1958) and McAllister and McAllister (1963) consisted of changing the apparatus that was used during testing, while the Siegel (1975) context change consisted of changing the room where testing occurred. That is, when the training apparatus is changed, proximal cues (those associated with the training apparatus) are altered, and when the environmental room is changed, distal cues (those associated with the training environment, or room) are altered. Interestingly, a recent study examining these stimuli confirmed that when proximal and distal cues are altered independently of each other, changes in both types of cues are associated with disruptions of learned tolerance (Kissinger and Riccio, 1995). As both types of contextual changes result in disruptions of memory, it can be assumed that both proximal contextual cues and distal contextual cues are important for the expression of memory. REVIEW OF THE CONTEXT SHIFT EFFECT IN HUMANS As can be seen with the above examples, there are many instances in the animal literature which highlight the importance of contextual cues in the expression of previously learned behavior. While the research on humans has not shown quite as robust of an effect of context on memory, there are a number of studies which show that, under certain circumstances, context plays an important role in memory processes. One such example utilized one of the most drastic contextual changes ever implemented in the study of context-dependent memory. Godden and Baddeley (1975) set out to examine

8 whether the recall of a previously learned word list would be affected by the conditions in which those words were learned and subsequently recalled. In their study, eighteen subjects from a university scuba diving club were recruited to participate in an experiment. During the experiment, subjects learned a list of 36 unrelated two and three Recall Environment Dry Wet Learning Environment Mean Recall Score SD Mean Recall Score SD Total Dry Wet Total Table Mean number of words recalled as a function of learning and recall environment. From Godden & Baddeley, syllable words, and were required to recall as many words as possible during a subsequent retention test. Godden and Baddeley designed the experiment so that four training/testing conditions would be represented. In one condition, subjects learned the word list on dry land, and were tested for recall on dry land. In another condition, the divers learned the word list underwater, and were tested for recall underwater. In the final two conditions, the environment at recall was different from the environment at learning (learning on dry land and testing underwater, or learning underwater and testing on dry land). The results of the experiment showed a clear effect of context. Importantly, there was no difference in the number of recalled words due to either the environment of learning or testing. That is, subjects in one condition (dry land) did not learn, nor remember, more effectively than subjects in the other condition (underwater), or vice versa. Relevant to our current discussion, Godden and Baddeley found a significant interaction between training and testing environment. Subjects who initially learned the list of words on dry land recalled significantly more words when tested on dry land compared to when they were tested underwater. Likewise, subjects who initially learned the list underwater performed significantly better in the underwater testing condition to when they were tested on dry land. These results clearly demonstrate a context shift effect, as subjects in the altered stimulus conditions at testing performed significantly worse than subjects tested in the same environmental conditions. The Godden and Baddeley (1975) experiment showed that a context shift effect occurred when drastic changes in context were implemented (dry land vs. underwater). Would the context shift effect also be evident in situations where the changes in context were less obvious? The answer to that question appears to be yes, as a series of experiments by Steven Smith (1979) suggest that extreme changes in contextual stimuli are not a necessary requirement to observe the context shift effect in human samples. Smith used the same general procedure as Godden and Baddeley (1975) in that subjects

9 learned lists of words in one context and were then given a recall test in the same or a different context; however, the differences between the contexts in Smith s experiments were much less drastic. In the Smith study, undergraduate psychology students learned and were tested for these word lists in different classrooms and laboratories at a Group Instruction Prior to Free Recall Mean Recall SC None 18.0 DC-C Recall Room B, view photos of Room B 18.8 DC-R Recal Room B, think about Room B 17.2 DC None 12.0 DC-P Recall room at home, think about room at home 9.6 Note: SC = same context, DC=C = different context-cued, DC-R = different contextremember, DC = different context, DC-P = different context-placebo instruction. Table Group designations, pre-recall instructions, and mean word recall. From Smith, university. In the first experiment, Smith observed that when subjects were required to recall the previously learned words in the same context in which they were trained, those subjects average recall was 24 words (out of the initial list of 30). However, when trained in one context (a small room with computer equipment) and then tested for recall in a different context (a large room with pictures, posters, books, tables, and chairs) the performance of these subjects dropped to an average of 18 words (out of the initial list of 30). Taken together, these findings indicate that performance was 25% better when tested in the same context, as compared to when subjects were trained and tested in mismatched contexts. The results of this experiment replicate those of Godden and Baddeley (1975), and further extended the results of contextual shift effects to situations where the perceived differences between the training and testing contexts are much less noticeable. A second experiment in the Smith (1979) study further highlighted the importance of context in memory, and demonstrated that the detrimental effect of a context shift could be reversed if a reinstatement procedure was used. A reinstatement procedure occurs when subjects are exposed to stimuli or cues from the original training episode. In a second experiment, Smith divided undergraduate subjects into one of five experimental conditions. The control group studied a list of words and were given a recall test in the same context where training occurred. The other four groups of subjects were trained in one context and were given the subsequent recall test in a different context; however, some of these subjects were given special instructions prior to the recall test in the shifted context. In one condition, subjects were shown photographs of the original training context in an attempt to reinstate memory for the original context and, therefore, facilitate memory on the recall test. Another group of subjects (who were not shown photographs of the original context) were instructed to first write down as many things as they could remember from the original training context. After attempting to remember stimuli found in the training context, these subjects were told that this procedure might help them better remember the words they had originally learned.

10 Another group that was tested in a different context were given no instructions prior to the recall test, and subjects in the final condition were told to remember an irrelevant context (one of their rooms at home) prior to the recall test. Smith hypothesized that subjects who were given a reinstatement procedure for the original training context (those shown pictures of it or asked to remember the learning context) would perform significantly better than those who did not mentally reinstate the original training context. Consistent with his predictions, the results indicated that there were no significant differences in the number of words recalled for subjects tested in the same context, subjects shown pictures of the original context, or subjects asked to mentally picture the original context. Furthermore, subjects in each of the above conditions performed significantly better than those subjects given no instructions when tested in the different context or those tested in a different context and asked to remember an irrelevant context. This second experiment by Smith provides two important pieces of information regarding the effects of context on memory. First of all, these results replicate previous reports that a change in context negatively affects recall for previously learned information. Secondly, these results suggest that a context change can be overcome by reinstating memory for the original training context prior to the test in a shifted context. That is, when subjects are exposed to aspects of the original training episode prior to testing in an altered context, the context shift effect is eliminated. (Of course, the extent to which the context shift effect will be reduced by reinstatement procedures will depend on the saliency of the reinstatement cues.) This context reinstatement effect has been shown in the animal literature as well, as many studies have shown that the context shift effect is eliminated when animals are exposed to cues and stimuli from the original training context prior to testing in shifted conditions (see Spear and Riccio, 1994). While the Smith (1979) experiments demonstrated that changes in the training and testing room can produce the context shift effect, other studies have demonstrated that changes in physical room assignment are not necessary for the context shift effect to occur. Balch, Bowman, and Mohler (1992) demonstrated the context shift effect by changing background music between a learning episode and subsequent retention test. In this study, subjects were required to study a list of words in the presence of jazz or classical music that was classified as having either a slow or a fast tempo. After the initial exposure to the word list, participants were given a recall test for the previously studied words in the presence of the same context (same music/tempo) or a changed context (different music/tempo). Consistent with findings from Smith (1979), Balch et al. (1992) reported that subjects tested in the presence of the same type of music they had heard during training performed significantly better than subjects tested in the presence of a different musical selection. In addition, Cann and Ross (1989) reported evidence for similar findings with odors rather than music. In this study, the magnitude of recognition memory was facilitated by the presence of an odor during training and testing. In short, subjects were shown pictures in the presence of a distinctive odor. Some groups were tested for recognition of the pictures in the presence of the same odor while others were tested with a different odor than what was present during initial presentation of the photographs. Consistent with the theme presented in this chapter, participants performed better when the odor at training and testing matched, and performed more poorly (more errors were made) when the odors at training and testing did not match. (A similar finding for the context shift effect using odors was reported by Schab (1990), who used a word recall paradigm). These demonstrations clearly show

11 that environmental context can be manipulated in a variety of ways, and further illustrate that a physical change of room location is not a necessary requirement for the context shift effect to occur. Through the previous examples it has been easy to see the importance of contextual information in the expression of memory. A number of studies from the animal and human literature have highlighted the importance of context, and have shown that memory is detrimentally affected when contextual stimuli are manipulated. One important issue then becomes the question of when contextual stimuli become important. That is, do we see context effects only with adult organisms, or are immature creatures also susceptible to contextual change effects? Several studies from the animal literature have demonstrated that the context shift effect also occurs in pre-adult (infant) animals (see Richardson, Riccio, and McKenney, 1988). In addition to evidence from animal studies, Carolyn Rovee-Collier and her colleagues have collected some impressive data which indicate that human infants are also susceptible to the effects of contextual change. We will now briefly describe one such study that has investigated this phenomenon. Testing the memory of a human infant is no small task. Obviously, we can t expect babies to memorize lists of words so we can test them for recall at a later time. However, Rovee-Collier and her colleagues have found a way around this issue, and have developed an ingenious method for testing memory in small infants. In essence, their subjects are trained to make a kicking response for positive reinforcement by utilizing a stimulus that babies find appealing: a mobile. Infants as young as three months of age learn this response relatively quickly, which Insert Rovee-Collier, Griesler & Earley (1985) Figure 1 about here is accomplished by attaching a ribbon to the child s ankle, and attaching the other end of the ribbon to the mobile. The children learn to make kicking responses as each response will produce movement of the mobile, which is the reinforcing stimulus. The more frequent a child s kicking response, the more reinforcement will be delivered in the form of a moving mobile. Rovee-Collier has demonstrated, through a series of experiments, that three month old infants are very capable of learning this response, and that their memory processes work much like that of an adult. That is, these children s memories are susceptible to forgetting, forgetting can be alleviated by reactivation treatments, and, as we will soon see, infant s memories are vulnerable to contextual influences. In a very creative study, Rovee-Collier, Griesler, and Earley (1985) sought to examine the role of contextual stimuli on memory in three month old subjects. Two separate experiments clearly demonstrated that infants as young as three months of age incorporate contextual stimuli in learning and memory situations, as manipulations of context led to deficits (or enhancements) of memory in both experiments. In these experiments, context was manipulated by altering the appearance of a crib bumper, which was a 61-cm high cloth that lined the inside of the child s crib. The bumpers were distinctive from one another, as one bumper consisted of a striped red and blue pattern, while the other bumper consisted of yellow circles imposed on a green background. In the first experiment, three-month-old infants were first trained on the leg-kicking response in the presence of one of the two crib bumpers. After establishing a baseline of responding, subjects were divided into groups that were given subsequent exposure to the

12 mobile after a one or a two week delay. Subjects were further subdivided so that some infants were tested in the same context (the bumper was identical to that of training), while other infants were tested in the presence of a different context (the bumper was different from that of training). Results from the 1-week test indicate that only those subjects who were tested in the presence of the original training bumper retained (remembered) the leg kicking response. That is, for subjects tested in the presence of an altered context, the leg kicking response was significantly less than what was observed during training. In essence, a context shift effect had occurred for subjects tested in the presence of the altered context (the different bumper). Additionally, Rovee-Collier et al. (1985) reported that complete forgetting had occurred after a two week delay between initial training and the retention test. Subjects that were tested two weeks after training had completely forgotten the leg kicking response, regardless of whether contextual stimuli were held constant or altered between training and testing. The results of this experiment demonstrate that even young infants utilize aspects of contextual stimuli in learning and memory processes, as shown by the better performance of the group tested in the same context at the 1-week test. A second experiment in the Rovee-Collier et al. (1985) study further examined the role of context as a reinstatement cue (see our earlier discussion of the Smith (1979) experiments). Rovee-Collier set out to examine whether contextual stimuli could be used to reinstate memory for a forgotten response, as subjects in the earlier condition had forgotten the leg-kicking response after the two week delay was imposed. In the second experiment, a different group of three month old infants were trained on the leg-kicking response, and after a baseline of responding was established, the infants were given a retention test two weeks after the completion of training. Twenty-four hours prior to the retention test, one group of subjects was given a reinstatement (reminder) treatment to the contextual surroundings by exposing the subjects to the original crib bumper that was used during training. A different group of infants was given exposure to a different bumper than what was used during training. (The reinstatement procedure for both groups consisted of three minutes of non-reinforced (no mobile present) exposure to the crib bumper.) After this reminder treatment, the retention test for the kicking response was then administered on the following day. The results indicate that only those subjects exposed to the original training context (the crib bumper used during training) exhibited the leg-kicking response, while those subjects exposed to a different crib bumper (a different context) did not show the leg-kicking response during the test two weeks after initial learning. Much like the results of Smith (1979) with list learning in adults, reinstatement of aspects of the contextual surroundings alleviated forgetting in the threemonth-olds. The results of this experiment show that forgetting in infants, just like that of adults, can be reduced by exposure to aspects of the training context. That is, exposure to the original crib bumper was sufficient to offset the normal forgetting that had occurred over the two week retention interval. These two experiments by Rovee-Collier and her colleagues highlight the importance of contextual stimuli in young infants, and further demonstrate that contextual stimuli can be used to reinstate memory after a response has been forgotten. REVIEW OF THE FORGETTING OF STIMULUS ATTRIBUTES A separate question within the context-dependent memory literature is whether extended time intervals between training and subsequent testing alter the magnitude of

13 the context shift effect. This has been studied extensively by David Riccio and his colleagues, who have observed that context shift effects become less pronounced over extended time intervals between initial learning and a subsequent test in altered conditions. Riccio has referred to this process as the forgetting of stimulus attributes. In an early paper documenting this phenomenon, Riccio, Richardson, and Ebner (1984) pointed out that context shift effects occur quite reliably when a retention test follows soon after original training; however, when a retention test is administered after a delay period, the context shift effect is greatly reduced or disappears altogether. Why should this occur? Why would a change in context disrupt the expression of memory, but only occur if the retention test occurs soon after the original learning episode? Riccio has argued that, when a retention test in an altered context occurs soon after training, organisms are quite capable of distinguishing between the training context and the altered testing context. When the contexts are distinguishable, the context shift occurs, either because of the lack of appropriate retrieval cues or the activation of competing memories. However, after a long delay between training and testing, organisms forget the specific characteristics of the training context, and their performance (memory) will generalize to other (altered) contexts. That is, while performance in an altered context is disrupted with a short delay, performance will often generalize to altered contexts after a long delay because the subjects have forgotten the specific characteristics of the original learning situation. In essence, while testing in altered contexts is quite disruptive with a short training to testing interval, contextual change effects become less prominent over extended time delays. Riccio and his colleagues have collected a large amount of animal data in support of this hypothesis, from a wide variety of procedural paradigms including studies of morphine tolerance (Feinberg and Riccio, 1990), state dependent retention (MacArdy and Riccio, 1991), reinstatement of extinguished fear responses (MacArdy and Riccio, 1995), pretest reminder cuing (Zhou and Riccio, 1994), the reactivation of inaccessible memories (Morgan, Flint, and Riccio, 1998), and conditioned taste aversion (Metzger, unpublished data). In addition, the findings from other investigations can also be explained with this framework, including data from the delayed tests of Perkins and Weyant (1958) and McAllister and McAllister (1963). Of the above cited examples, the Feinberg and Riccio (1990) report provides a good example for our discussion, as it directly relates to the Siegel experiments that were discussed earlier in this chapter. As in Siegel s experiment, rats in the Feinberg and Riccio study were given four morphine injections, once every other day, to establish tolerance to the drug. (For all animals, the morphine injections were administered in a distinctive environment.) Prior to the test for morphine tolerance, subjects were given another injection of morphine, either in the same context as that of the previous injections, or a different context. In addition, the test for tolerance was administered either two or seven days after training. Consistent with the results of Siegel (1975), those animals tested for morphine tolerance in a different context after a two day delay demonstrated a disruption of tolerance compared to those subjects given the test in the same context in which they were trained. That is, a context shift had occurred for those animals tested in an alternate environment. However, when the test for tolerance occurred after a seven day interval, no context shift effect was observed, as animals tested in the different context demonstrated the same level of tolerance as those animals tested in the same context in which they were trained. After a seven day retention interval, subjects had forgotten the specific characteristics of the training context, and therefore, a

14 change in context failed to produce a context shift effect. As stated by the authors, after a long retention interval the physical change in the context is ineffective because subjects have forgotten the precise features of the environment originally paired with drug administration (pg. 276). While these findings show clear support for the hypothesis that the specific characteristics of a learning context is forgotten over time, the Feinberg and Riccio (1990) investigation also provide insight to why some laboratories Retention Interval 2-Day 7-Day Context Condition Mdn M SE Mdn M SE Same Different Table Median, mean, and standard error of Paw-Lick latency as a function of retention interval and context condition following morphine exposures. From Feinberg and Riccio, may fail to reproduce the context shift effect. These results suggest that the context shift effect is most readily observed when the training to testing interval is relatively short, as delays in testing may produce scenarios where testing in altered contexts would have no disruptive consequence on the expression of memory. It should be noted that the forgetting of stimulus attributes is not limited to animal research, as Riccio et al. (1994) have made a valid argument stating that many phenomenon observed in the human literature can also be explained through this framework. For example, Borovsky and Rovee-Collier (1990) demonstrated that infants (six-month-olds) were susceptible to the context shift effect, as a decrease in leg kicking occurred in the presence of a shifted context (different crib bumper), but the shift effect only occurred when testing with the altered crib bumper occurred soon after training. When subjects were placed in the shifted context after 7 or 14 days, no context shift effect had occurred, as the leg kicking response was as prominent in the shifted context as in the original training context. This finding can well be explained through a forgetting of stimulus attributes interpretation. When a change in the crib bumper occurred soon after training, performance was impaired, as expected. However, when the shift in context occurred either 7 or 14 days later, the infants had apparently forgotten the specific characteristics of the original crib bumper, and their responding generalized to the altered stimulus. This finding from Borovsky and Rovee-Collier (1990) parallels findings from the animal literature, in that a shift in context was only effective in disrupting memory when the context shift occurred soon after the original learning episode. In addition, Riccio et al. (1994) have argued that the forgetting of stimulus attributes model can adequately explain such psychological processes as constructed memory, the sleeper effect, the false fame effect, and others. As we have readily seen with the context shift effect, the forgetting of stimulus attributes appears to be a robust

15 phenomenon, as it has been observed across multiple experimental settings utilizing a wide variety of experimental procedures and paradigms. MECHANISMS OF CONTEXT EFFECTS It has long been known that there are specific brain regions that are involved in the regulation of encoding, storing, and retrieval processes. While several specific brain regions have been implicated in memory function (e.g., the hippocampus, amygdala, and entorhinal cortex to name a few), the hippocampus is one region that has received considerable attention in the expression of contextual memory, or memory for context. However, as you will soon see, while there may be data suggesting that the hippocampus is involved with the expression of memory for contextual stimuli, there are also conflicting data which prevent us from drawing specific conclusions. For example, Kim and Fanselow (1992) used a classical conditioning procedure in which they trained rats to fear a tone by pairing the tone with footshock, all in the presence of specific contextual stimuli. After a conditioned response was established, Kim and Fanselow removed the hippocampus of some subjects and reported an interesting finding. As expected, control subjects (who did not have their hippocampus removed), demonstrated fear responses in the presence of either the tone or the context where training took place. However, the subjects with hippocampal lesions retained fear to the tone after surgery, but had completely lost their fear to the context. That is, these rats failed to show fearful responses in the presence of the contextual stimuli where training had taken place. The fact that they retained fear to the tone suggested that memory for context was specifically impaired in these animals. In other words, Kim and Fanselow (1992) demonstrated that the hippocampus was important for the expression of memory for context, but not as critical for the expression of memory for the tone presented during training. Similar results have been obtained by other laboratories (Aguado, Hall, Harrington, and Symonds, 1998), and others have demonstrated that the hippocampus is critical for contextual memory retrieval using extinction and latent inhibition paradigms (Corcoran and Maren, 2001; Holt and Maren, 1999; Maren and Holt, 2000). However, while memory deficits are common following hippocampal damage, a number of studies have failed to find memory deficits for contextual stimuli following damage to the hippocampal region (Hall, Purves, and Bonardi, 1996; Holland, Lamoureux, Hans, and Gallagher, 1999; Winocur, 1997). Conflicting results such as these prompted an article by Holland and Bouton (1999), who specifically addressed the role of the hippocampus in contextual representations. Basically, Holland and Bouton cautioned that we should not assume that the hippocampus is specifically involved in every aspect of contextual memory, as that assumption is based on the results of several studies which suggest that interpretation. As they succinctly stated, contextual learning encompasses a variety of changes in learning and performance, some of which involve the hippocampus and others which do not (pg. 195). Holland and Bouton (1999) suggest that current models of hippocampal involvement in contextual memory need further elaboration and development to account for these conflicting findings. While it might be tempting to assume that the hippocampus is the spot where contextual memory representations are manipulated, evidence from a number of studies demonstrating a lack of hippocampal involvement in contextual memory lends support to the Holland and Bouton argument.

16 SUMMARY, APPLICATION, AND CONCLUSIONS As we have seen throughout the examples provided in this chapter, contextual stimuli play an important role in the expression of previously learned behaviors. Many studies have demonstrated that memory is disrupted when testing occurs in the presence of stimuli that were not present during initial training (that is, a context shift effect occurs). Importantly, this phenomenon has been observed extensively in both the animal and human literature, and is shown in a wide variety of paradigms and procedures. The extensiveness and robustness of the context shift effect clearly shows that this phenomenon is reliable, under a wide variety of scenarios, further attesting to the importance of external context in memory processes. In addition to changes in context reliably disrupting memory, we have seen that exposure to contextual stimuli can also offset forgetting. An experiment by Smith (1979) showed that reinstatement of the training context alleviated the context shift effect, in that subjects who were shown pictures of the original context or were asked to remember the original training context prior to testing in a different environment performed comparably to those trained and tested in the same context. Under different circumstances, Rovee-Collier et al. (1985) demonstrated that exposure to aspects of the training context alleviated forgetting in three month old infants. After a two week delay (when forgetting of the leg kicking response was shown to be complete), those subjects that were given exposure to the training context prior to the test retained the leg kicking response. Given the relative importance of context in forgetting and memory processes, what do these findings mean to you? Clearly, memory works optimally when the context at retrieval is identical to the context at encoding. But what about situations where it is not possible or feasible to have a matched context at encoding and retrieval? For instance, does this mean that you should spend your time studying for an exam in the room where you are to take the test? If contextual effects are that powerful and robust, it would seem that you should do exactly that, to reduce the possibility of a context shift effect when you take your next exam. Before jumping to such conclusions, however, we need to examine a study by Saufley, Otaka, and Bavaresco (1985), who sought to examine the importance of context in a non-laboratory situation: The college classroom. With the plethora of information confirming that contextual stimuli were critical for the expression of memory, Saufley et al. (1985) examined data from thousands of college students who were required to take mid-term examinations in rooms different from the regular lecture room where they normally took notes. That is, while contextual effects have been largely documented in controlled laboratory settings, would the same effects be manifested in a more natural environment away from the laboratory? Saufley and his colleagues compared data from more than 4,500 students at the University of California, Berkley, who were enrolled in a variety of undergraduate courses (psychology, computer science, physical anthropology, and introductory biology). At the university, because of crowded conditions and room size constraints, it was often necessary to have some students take their exams in rooms different from the lecture room. The question of interest was whether the change in context created a situation where memory was impaired because of altered stimulus conditions. That is, would the students being tested in a different classroom experience a context shift effect and, therefore, would their performance on the exam suffer as a result? If data from the laboratory also applied to