INFLUENCE OF RETROACTIVE INTERFERENCE ON THE CONTEXT SHIFT EFFECT

Transcription

1 INFLUENCE OF RETROACTIVE INTERFERENCE ON THE CONTEXT SHIFT EFFECT A thesis submitted to Kent State University in partial fulfillment of the requirements for the degree of Master of Arts By Erin Marie Fleming August 2010

2 Thesis written by Erin Marie Fleming B.A., Kenyon College, 2007 M.A., Kent State University, 2010 Approved by David C. Riccio, Ph.D., Advisor Maria S Zaragoza, Ph. D., Chair, Department of Psychology Timothy Moerland, Ph.D., Dean, College of Arts and Sciences ii

3 TABLE OF CONTENTS LIST OF FIGURES iv ACKNOWLEDGEMENTS v INTRODUCTION...1 EXPERIMENT 1 9 METHOD..10 Subjects.10 Materials 10 Procedure...11 RESULTS.. 15 DISCUSSION 17 EXPERIMENT METHOD..21 Subjects. 21 Materials 21 Procedure...22 RESULTS..26 DISCUSSION 32 GENERAL DISCUSSION 34 REFERENCES..43 iii

4 LIST OF FIGURES TABLE 1 TABLE 2 FIGURE 1 FIGURE 2 FIGURE 3 Study design of experiment Study design of experiment Mean total time on white (in seconds) on day of testing for each group in experiment 1 16 Mean total time on white (in seconds) on day of testing for single long exposure groups in experiment 2 30 Mean total time on white (in seconds) on day of testing for each group in experiment 2, including collapsed single long exposure groups.31 iv

5 ACKNOWLEDGEMENTS I would like to thank all of those people who made the completion of this project possible. Thank you goes first to my thesis advisor, Dr. David Riccio, for all of his guidance throughout the research for this project as well as outside of this project. Without his patience and input I would not be where I am to him and I am eternally grateful to him. I would like to thank my fellow graduate students, Christie Bartholomew, Amber Chenoweth, and Patrick Cullen for their support and guidance throughout this project. I would also like to thank my committee members: Dr. Stephen Fountain, Dr. Beth G. Wildman, and Dr. John Dunlosky for their valuable comments and review of my thesis. I would like to especially thank my family for all of their love and support throughout this process. Without them I would not have been able to make it through the long hours and hard work and I appreciate them more than they will ever know. v

6 Introduction Learning, and memory for learned information, involves much more than just the specific material that is being learned. Memory, in general, includes a number of attributes that are tied to a representation of some past information that can serve as cues to better aid the retrieval of that memory (Spear, 1978; Underwood, 1969). These attributes would include not only information specific to the stimulus and response that bring about a given outcome, but also the contextual (background) cues that were present at the time of learning. More specifically, stimulus attributes are features or characteristics of the training stimuli or peripheral stimuli in the environment. The peripheral stimuli have also been called contextual cues and are stimuli other than the training stimulus that are present at the time of training (Spear & Riccio, 1994). These contextual cues include the room or environment, lighting, visual cues, odors, sounds, and even internal states and emotions. Studies investigating learning and memory using animal models have focused on the specific response and what the cues associated with that response predict, rather than focusing on all of the stimuli or context present during the learning episode. For example, earlier studies looked at Pavlovian conditioning pairing some unconditioned stimulus with a conditioned response. In doing so, they focus only on that conditioned response and the unconditioned stimulus that predicts that response and not the other cues that are present during the learning episode, such as the background environment and other cues 1

7 2 from the environment other than the unconditioned stimulus. One way in which researchers have investigated the influence of these other stimuli has been by looking at what is known as stimulus generalization within the realm of associative learning. Stimulus generalization involves the control of other stimuli over responding and focuses on those cues that are not specific predictive cues, such as the unconditioned stimulus. The importance of these stimuli, and stimulus generalization in particular, has been noted by R. Shepard (1987) as being as fundamental to the learning process as the principle of gravity is to physics. Because the stimuli and contexts at the time of retrieval are rarely ever exactly the same as they were at the time of training, stimulus generalization provides the basis for the transfer of learning to new situations. One aspect of stimulus attributes and their involvement in learning is that their effects change over time. It has been found that more stimuli come to control a response over time rather than just one specific cue. The weakening of control by a specific stimulus that originally predicted a response is reflected in a flattening of the generalization gradient and occurs not only for conditional or discriminative stimuli but also contextual cues (Zhou and Riccio, 1996). Perkins and Weyant (1958) interpret this flattening of the generalization gradient as the forgetting of stimulus attributes or features. The strength of the learned response is basically unchanged over time but comes to be expressed to a wider range of cues that were not associated with the original learning episode. What is most interesting about the finding that the generalization gradient flattens over time is that there need not be any discrimination training involved for the context

8 3 shift to occur. Specifically, after learning information in one context it has been found that performance is disrupted if testing occurs in a context that is different from that in which they had the original learning episode (Rosas and Bouton, 1997; Smith, 1979; Borovsky and Rovee-Collier, 1990). This context shift effect, which did not include discriminative training, then may diminish over time resulting in the subjects performing equally well in both contexts (McAllister and McAllister, 1963; Rosas and Bouten, 1997; Zhou and Riccio, 1996). The study done by Zhou and Riccio (1996) showed just how important stimulus attributes are in learning and memory retrieval. Zhou and Riccio noted that it is difficult to quantitatively compare aspects of the context and stimulus attributes, such as trying to compare various sounds to various odors. However, a qualitative assessment is possible in that if a change in one aspect of the context does disrupt performance, then it is reasonable to infer that the stimulus was part of the information attended to and encoded by the subject. They wanted to utilize this qualitative assessment by altering both the proximal cues, which are cues associated with the conditioning apparatus, as well as distal cues, which are cues related to the more general background environment. By doing this, they could examine whether both distal and proximal stimuli exert control over performance and investigate the effects of changing more than one element of a context. They also wanted to determine whether the loss of the context shift effect over time would be differential (i.e. are distal attributes forgotten more rapidly than proximal?). Zhou and Riccio s experiment (1996) consisted of two different rooms and two

9 4 different passive avoidance boxes. All rats were trained in Room A in Box A but where the rats were tested varied. They were either tested in the same room and box, different room, different box, or both a different room and different box. The rats were also tested in different time increments, with half being tested 24 hours after training and the other half being tested 2 weeks after training. Using these manipulations they found that at a short delay of 1 day altering either the distal or proximal cues resulted in a severe performance decrement. However, after a 2 week delay it didn t matter which type of cue was altered, distal or proximal, there was no performance decrement. As long as only either the distal or proximal cues were altered, and not both, the generalization gradient flattened at a long delay and all contexts were treated as the original training context. This study illustrates how important the effect of stimulus attributes is on learning at longer intervals and indicates that there are ways that the parameters can be changed to alter typical learning behavior. Without any specific treatments between original learning episode and testing but by merely altering the length of time between train and test and what type of cues are present during testing can reduce the usual performance decrement we would see with the context shift effect. Another very important example of the implications of forgetting stimulus attributes comes from studies investigating the reinstatement of fear. If the UCS is presented outside of the training situation after the conditioned fear response has been extinguished the original fear memory will return (Rescorla and Heth, 1975). Subsequent research indicates that the reinstatement effect may depend on associative processes of generalization between the training and reinstating conditions similar to what is seen with

10 5 the generalization gradient. MacArdy and Riccio (1995) showed that a reinstatement treatment in a context different from the original training context was ineffective when administered at a short interval (either immediately or 1 day after extinction) but became effective at a longer interval (7 days after extinction). These findings are consistent with both the generalization account and the principle of forgetting of attributes. A more recent study (McAllister and McAllister, 2006) compared non-contingent and contingent UCS exposure in several contextual conditions either 1 or 9 days after extinction and confirmed the previous findings. More importantly, the authors noted that the forgetting of detailed characteristics of contexts would make the representations of the original conditioning situations less distinguishable from the post extinction shock situation (p.48). This brings us to the purpose of the present study which is to investigate possible determinants of forgetting of stimulus attributes. The finding that a flatter generalization gradient indicates that conditioned responses occur to a wider array of stimuli may be interpreted as meaning that some of the specific stimulus attributes of the learning episode may have been forgotten. This phenomenon of forgetting of stimulus attributes, in which perceptually distinguishable stimuli become more functionally interchangeable over a retention interval and result in an increase in total responding, contrasts with the performance deficits typically referred to as forgetting (Riccio, Rabinowitz, and Axelrod, 1994). While most research has been done focusing on decrements in performance associated with forgetting of specific learned responses, the present study is concerned with the effects of forgetting of stimulus attributes. The research herein should

11 6 give valuable information on some aspects of the forgetting of stimulus attributes that can lead to a theoretical understanding of the phenomenon. One important memory phenomenon that may play a role in the forgetting of stimulus attributes is retroactive interference. Retroactive interference occurs when new information is presented after the activation of a memory and conflicts with that memory resulting in forgetting (Spear & Riccio, 1994). In a general retroactive interference design, the interfering information is encountered during the retention interval of the original learned information (Hulse, et al., 1975). The difference in the amount of retention of the first task reflects the influence of the learning of the second task. Experimentally produced forgetting occurs when the control group, which receives no other information after the first learning episode, retains more than the experimental group, which receives other information during the retention interval. The first theories of retroactive interference focused on competition between original and interfering information (McGeoch, 1932) as well as intrusion of the original information during the learning of the information encountered during the retention interval which results in an unlearning of the original information (Melton & Irwin, 1941). More modern views of retroactive interference focus on a combination of associative memory and cognition to gain a more detailed understanding of interference. This new way of investigating interference resulted in retroactive interference being more applicable to many more aspects of memory than originally thought. These applications ranged from its significant role in everyday memory through misinformation effects (Loftus & Palmer, 1974) to changes in the effects of retroactive interference from a

12 7 developmental stand-point (Harris, 1975). It is apparent that retroactive interference has been studied extensively within general memory mechanisms as a reason for why performance declines or why forgetting occurs. However, there is little, if any, investigation into the effects of retroactive interference on the forgetting of stimulus attributes. While it is not known at this point whether stimulus memory is susceptible to retroactive interference, retroactive interference for context has been indirectly suggested in a study of contextual reminders in college students by Smith (1979). In this study, Smith found that reminders of the original context in which the information was learned resulted in an alleviation of the usual performance impairment seen with the context shift effect. However, Smith also discovered that the effectiveness of the reminder was reduced if the students were exposed to several different contexts in between training and testing. Thus, it appears as though the additional contextual stimuli experienced in between training and testing interfered with the memory for the original context which made reinstatement less effective. When studying retroactive interference of stimulus attributes the use of the usual index of performance loss does not apply. With the majority of studies on retroactive interference, retroactive interference results in forgetting of the originally learned information as indicated by a decrease in performance. To investigate retroactive interference of stimulus attributes, on the other hand, improved performance will indicate that retroactive interference of stimulus attributes has occurred. Retroactive interference from stimulus attributes would interfere with the memory for stimulus attributes from the

13 8 original learning episode and the learned information will be retrieved just as well in any novel context. At the basis of this is the idea that stimulus attributes will become more functionally interchangeable and performance will be similar in all contexts.

14 Experiment 1 Previous research has shown that many different aspects of learning can interfere with previously acquired learning and result in a deficit in responding. Experiment 1 investigated the effects of context, and stimulus attributes in particular, on the memory for a single learning episode. This experiment was based on the approach used by Smith (1979) but where Smith assessed interference indirectly in terms of impaired reinstatement of the target memory, the current experiment evaluated the flattening of the contextual gradient as an index of forgetting of stimulus attributes and as caused by retroactive interference. The flattening of the contextual gradient was evaluated by changes in the context shift effect. Experiment 1 tested the hypothesis that exposure to novel stimuli during the retention interval will serve as sources of retroactive interference which will result in a more rapid loss of the context shift effect when compared with retention controls. 9

15 METHOD Subjects Subjects were 40 female Long-Evans rats approximately 180 days old at start of the experiment. Animals were singly housed in the animal colony at Kent State University in a rectangular plastic transparent plastic box (45.7 (L) x 25.4 (W) x 20.3 (H) cm). Animals were on a 15/9 light/dark cycle and were provided ad lib food and water throughout the experiment. All handling and experiments took place during the light portion at the same time each day. Materials Five different rooms and passive avoidance chambers were used in the experiment. Context A was a 1.83 x 2.74 m room painted white. Posters were placed on each wall to provide visual cues. Long-Evans rat species are known to have adequate vision perception so they can distinguish between varying visual contextual stimuli. The room was illuminated by a standard 40w bulb. Context A had no odor present, but had white noise for background noise. Context B was a 1.83 x 2.74 m room painted white. Posters were also placed on each wall to provide different visual cues. To provide extra differing visual stimuli, context B was illuminated with a 25w red light and no odor present. Context C was a 1.83 x 2.74 m room painted white with posters placed on each 10

16 11 wall and normal lighting. Context C also had the odor of lemon present which was distributed in the air by a cotton ball soaked in 50mL of lemon cooking extract and placed on a paper plate below the passive avoidance apparatus. Context D was a 1.83 x 2.74 m room painted white with posters placed on each wall and normal lighting. Context D also had the odor of vanilla present which was distributed in the air by a cotton ball soaked in 50mL of vanilla cooking extract and placed on a paper plate below the passive avoidance apparatus. Context E was a 1.83 x 2.74 m room painted white with posters placed on each wall and normal lighting. Context E also had the odor of peppermint present which was distributed in the air by a cotton ball soaked in 50mL of peppermint cooking extract. Training, exposure, and testing were conducted in a (L) x (W) x (H) cm black-white Plexiglas shuttle box with a grid floor (2mm grids spaced at 1mm apart center to center). The two compartments were equal in size and divided by a guillotine door that was manually lifted and lowered by a pulley system controlled by the experimenter. The white compartment consisted of a transparent Plexiglas lid with white Plexiglas walls. The black compartment consisted of a transparent Plexiglas lid with black Plexiglas walls. Footshock was delivered through the grid floor via a constant current AC shock generator (Model 5806 Lafayette Instruments Co., Lafayette, IN). Procedure The plan of this experiment was to condition fear in rats either with or without exposure to multiple novel contexts during the retention interval between training and testing. The Ss were tested for their avoidance of the shock compartment with no shock

17 12 present 48 hours after training in either the original training context or a novel context. All subjects were handled for 3 days prior to training. Handling consisted of physical contact between experimenter and subject for 5 minutes at the same time each day 3 days in a row. There were a total of ten rats in each group in the experiment. Rats were randomly assigned to the four groups. Group Same and Shift were the control groups that received no contextual exposure in between training and testing. The Multiple Context Exposure (MCE) Same and MCE Shift groups were the experimental groups that received exposure to novel contexts between training and testing (See Table 1). Given the usual degree of variability among rats and past experience with similar studies ten rats was an acceptable amount in each group and allowed adequate power to reject the null hypothesis if there was an experimental effect. This experiment used Pavlovian differential fear conditioning in a passive avoidance chamber. This type of conditioning involves pairing a neutral cue with a foot shock and results in learning of fear of the originally neutral cue. In this study, the neutral cue was the black compartment of the passive avoidance apparatus. During training, the subject was placed in the white compartment of the apparatus. After 10 seconds, the guillotine door was opened, and the subject was free to cross to the black compartment of the apparatus. Once the subject crossed to the black compartment the guillotine door was closed and the subject received 2 brief 1 second inescapable foot shocks of 0.5 ma intensity 5 seconds apart. Cross latency scores were calculated which involved recording how long it took the rats to cross from the white to the black side once the guillotine door was opened. After training was complete, the rats were then returned to their home cages.

18 13 24 hours after training, the subjects in the experimental group experienced multiple context exposure and the subjects in the control groups received handling. The multiple context exposure consisted of the mere exposure to 3 different contexts for 10 minutes in each context. The order of the contextual exposures was randomized and the contexts used were Contexts C, D, and E. During exposure, the subjects were placed on the white side of a passive avoidance chamber with the guillotine door closed. The guillotine door remained closed during the duration of the exposure session so there was no exposure to the black compartment in an attempt to avoid extinction. The rats remained in the context for 10 minutes and were then transported immediately to the next context where the exposure process was repeated. Subjects in the control groups received handling 24 hours after training. Handling consisted of physical contact between experimenter and rat in the colony for 30 minutes (the total amount of time the experimental groups received multiple exposures). Testing occurred 24 hours after the multiple context exposure session and took place in either the training context (groups Same and MCE Same) or a novel context (groups Shift, MCE Shift). Testing involved behavioral observations over a period of 10 minutes during which no foot shock was administered. The behavioral observations included latency scores, or how long it took the subjects to cross from the white compartment to the black compartment, as well as total time on white scores which was simply the total amount of time the rat spent in the white compartment of the apparatus. Maintained memory of stimulus attributes was reflected in short latency scores and/or lower total time on white scores.

19 14 Group 1: Same Group 2: Shift Group 3: Multiple Context Exposure Same Group 4: Multiple Context Exposure Shift Day One Day Two Day Three Train in Context Handled for 30 Test in Context A A minutes Train in Context Handled for 30 Test in Context B A minutes Train in Context 10 minute exposure to Test in Context A A Context C, 10 minute exposure to Context D, 10 minute exposure to Train in Context A Context E 10 minute exposure to Context C, 10 minute exposure to Context D, 10 minute exposure to Context E Test in Context B Table 1. Study design for experiment 1. The six groups are listed in the first column and the days of the experiment are listed in the second, third, and fourth columns. This table indicates what procedure each group underwent each day.

20 15 RESULTS Training latencies were evaluated in a one-way ANOVA to determine whether all animals exhibited comparable cross latencies, which indicates rats natural preference of the black compartment at training. The ANOVA showed there were no significant differences among the training latencies for the 4 groups (F(3,36)= 1.381, p=.26). Testing cross latencies were evaluated in a one-way ANOVA and used to determine retention of original fear memory where shorter cross latencies would indicate strong fear and, thus, retention of the information from the training episode. The ANOVA showed that there was a statistically significant difference among the testing cross latencies for the 4 groups (F(3,36)=2.901, p<.05)(same (M=332.20, SD=285.94), Shift (M=216.40, SD=259.86), MCE A (M=83.00, SD=92.91, MCE B (M=102.60, SD=158.64)). A post-hoc Tukey test was run to investigate where the differences were between the groups but it yielded no statistically significant results at p<.05. Mean total time on white during testing was evaluated using a one-way ANOVA to see whether the memory for stimulus attributes was affected by the multiple context exposure treatment. The ANOVA showed that there was a marginally statistically significant difference among the total time on white scores for the 4 groups (F(3,36)=2.768, p=.056). A post-hoc Tukey test for group showed that there was a difference between the total time on white scores for the group that was trained and tested in the same context with no experimental manipulation on the second day (Same) and the group that was trained in one context and tested in a different context after being exposed

21 16 to multiple contexts during the retention interval (Multiple Context Exposure Shift) with p<.05. Group Multiple Context Exposure Shift showed little fear in the novel context, with total time on white scores below half of the total time of testing, while Same group exhibited a large amount fear with total time on white scores showing that they spent very near to the entire ten minutes in the white compartment. The mean total time on white scores for each group are shown in figure 1. Figure 1. Mean total time on white (TTW) scores (±SEM) in experiment 1 for the control groups, Same and Shift, and the experimental groups, Multiple Context Exposure Same (MCE Same) and Multiple Context Exposure Shift (MCE Shift), at retention test. The Same group that was trained in one context and tested in the same context with no experimental manipulation during the retention interval had significantly different TTW scores when compared with the Multiple Context Exposure Shift group that was trained in one context and tested in a context different form the training context after receiving exposure to multiple contexts during the retention interval.

22 DISCUSSION The results from experiment 1 did not support the hypothesis that multiple context exposures serve as sources of retroactive interference and result in a more rapid loss of the context shift effect. Rats that received multiple context exposures during the retention interval and were tested in a context that was different from the training context showed significantly less fear to the black compartment when compared with rats trained and tested in the same context with no experimental manipulation during the retention interval. The results also showed that there was no statistically significant difference between the group that was trained and tested in the same context with no experimental manipulation during the retention and the group that was tested in a context that was different from the training context with no experimental manipulation during the retention interval. These results go against the usual findings of the context shift effect that has been shown in many previous studies. While the overall cross latency scores at test were statistically significantly different, the post-hoc analyses were not statistically significant thus giving no information as to where the differences were. The discussion of the results will be based on the total time on white scores at testing. Many researchers that use passive avoidance training in their research tend to rely more heavily on the total time on white scores than the cross latency scores at test. In a recent article by Santucci & Cardiello (2004) they posit that animals may have short cross latencies with long total time on white scores, indicating fear and a strong memory for initial learning, because they were not reminded 17

23 18 of the aversive event until they enter the black compartment. Thus, the rats may remember their initial training, as indexed by longer total time on white scores, but they may not activate that memory until they cross to the compartment in which the aversive event occurred. Because of this, the total time one white scores will be discussed herein as they are a more accurate indication of fear than cross latencies at test. Based on the hypothesis that multiple context exposures should produce retroactive interference resulting in a more rapid loss of the context shift effect, there should not have been a significant difference in total time on white scores between the Same group and the Multiple Context Exposure Shift group. By having a statistically significant difference for total time on white scores between these groups there was not the amount of generalization to stimulus attributes that the hypothesis predicted. The first explanation for why this may have occurred has to do with extinction. The way in which the multiple context exposure groups received their multiple exposures may have resulted in extinction of their original learning. Because they are placed in the white compartment of a passive avoidance apparatus in each context in between training and testing, they may have extinguished the memory that they should fear the black side of the apparatus. However, this seems unlikely not only because they were never exposed to the black compartment but also because if this were the case there would have been no significant difference between the two Multiple Context Exposure groups because they received the same exposure procedure. A second possible reason why there was a significant difference between Same and Multiple Context Exposure Shift is that being placed in the white compartment of the

24 19 passive avoidance apparatus, even in novel contexts, may have acted as a reminder treatment of the original training, similar to what was found in Rescorla and Heth s study (1975). This explanation makes more sense than the extinction reasoning because it can explain why there is no difference between the multiple context exposure group that was tested in the same context that they were trained in (Multiple Context Exposure Same) and group Same but there is a difference between group Same and the Multiple Context Exposure Shift group. If the animals were receiving a reminder treatment with the current exposure procedure, then the memory for the original learning would be strengthened and they would not be affected by the multiple contexts. Because of this, the Multiple Context Exposure group tested in the same context they were trained (MCE Same) in would show a strong fear memory and the Multiple Context Exposure group tested in a different context from the training context (MCE Shift) would show a strong memory for stimulus attributes and, thus, distinguish between the training context and the novel testing context and act with no fear. This is exactly what happened, based on the results of this study, and is a major confound. A major issue with the current results is the absence of the usual context shift effect. Based on previous studies a statistically significant difference was expected between the Same group and the Shift group that would show the context shift effect that has been found in many studies carried out before this one. The reason for the lack of the context shift is unclear.

25 Experiment 2 In order to test the original hypothesis from experiment 1 without confounds that may have arisen due to the exposure procedure that procedure needed to be altered. Experiment 2 was run to test the hypothesis that exposure to novel stimuli during the retention interval will serve as a source of retroactive interference, resulting in a more rapid loss of the context shift effect when compared with retention controls. To test this hypothesis, major changes were made in the way that the exposure procedure was done to avoid any possible extinction or reinstatement effects. Extra control groups were also added that would show whether any effect is due to the multiple contexts or just a single novel contextual exposure over the same amount of time. 20

26 METHOD Subjects Subjects were 59 female Long-Evans rats approximately 90 days old at start of the experiment. Animals were singly housed in the animal colony at Kent State University in a rectangular plastic transparent plastic box (45.7 (L) x 25.4 (W) x 20.3 (H) cm). Animals were on a 15/9 light/dark cycle and were provided ad lib food and water throughout the experiment. All handling and experiments took place during the light portion at the same time each day. Materials Five different rooms and passive avoidance chambers were used in the experiment. Context A was a 1.83 x 2.74 m room painted white. Posters were placed on each wall to provide visual cues. Long-Evans rat species are known to have adequate vision perception so they can distinguish between varying visual contextual stimuli. The room was illuminated by a standard 40w bulb. Context A had no odor present, but had white noise for background noise. Context B was a 1.83 x 2.74 m room painted white. Posters were also placed on each wall to provide different visual cues. To provide extra differing visual stimuli, context B was illuminated with a 25w red light and no odor present. Context C was a 1.83 x 2.74 m room painted white with posters placed on each wall and normal lighting. Context C also had the odor of lemon present which was distributed in the air by a cotton ball soaked in 50mL of lemon cooking extract and 21

27 22 placed in a plastic dish in the room. The odor was distributed throughout the room for 10 minutes prior to exposure as well as during exposure. Context D was a 1.83 x 2.74 m room painted white with posters placed on each wall and normal lighting. Context D also had the odor of vanilla present which was distributed in the air by a cotton ball soaked in 50mL of vanilla cooking extract and placed in a plastic dish in the room. The odor was distributed throughout the room for 10 minutes prior to exposure as well as during exposure. Context E was a 1.83 x 2.74 m room painted white with posters placed on each wall and normal lighting. Context E also had the odor of peppermint present which was distributed in the air by a cotton ball soaked in 50mL of peppermint cooking extract and placed in a plastic dish in the room. The odor was distributed throughout the room for 10 minutes prior to exposure as well as during exposure. Training and testing were conducted in a (L) x (W) x (H) cm black-white Plexiglas shuttle box with a grid floor (2mm grids spaced at 1mm apart center to center). The two compartments were equal in size and divided by a guillotine door that was manually lifted and lowered by a pulley system controlled by the experimenter. The white compartment consisted of a transparent Plexiglas lid with white Plexiglas walls. The black compartment consisted of a transparent Plexiglas lid with black Plexiglas walls. Footshock was delivered through the grid floor via a constant current AC shock generator (Model 5806 Lafayette Instruments Co., Lafayette, IN). All multiple context exposures and long exposures took place in a clear Plexiglas shoebox measuring (30.5) x (15.2) x (15.2) cm with a clear Plexiglas lid. Procedure

28 23 As with experiment 1, the plan of this experiment was to condition fear in rats either with or without exposure to multiple novel contexts during the retention interval between training and testing. The Ss were tested for their avoidance of the shock compartment with no shock present 48 hours after training in either the training context or a novel context. The procedure for experiment 2 was the same as for the procedure of experiment 1 with a few changes to address exposure issue in experiment 1 and to include the new control groups (See table 2). There were ten groups with 8 rats in the Same group, Multiple Context Exposure Same group, and Multiple Context Exposure Shift group and 6 rats in the Shift group. There were 5 rats in each of the Long Exposure (LE) groups except for the Long Exposure Context D Same group which had a total of 4 rats. Rats were randomly assigned to the ten groups, with group Same, Shift, and all 6 Long Exposure groups the control groups and Multiple Context Exposure Same and Multiple Context Exposure Shift the experimental groups. Given the usual degree of variability among rats and past experience with similar studies this number of rats was an acceptable amount in each group and allowed adequate power to reject the null hypothesis if there was an experimental effect. Control groups were included that controlled for possible effects of duration of exposure in each of the separate contexts used in the exposure treatment (Context C, D, and E). The control groups, groups 5-10 (Table 2), had the same training and testing procedure, but the exposure procedure on day 2 was altered. 24 hours after training the rats in groups 5-10 were placed in one of the 3 contexts for 30 minutes. The Long

29 24 Exposure Context C group was placed in context C for 30 minutes, the Long Exposure Context D group was placed in context D for 30 minutes, and the Long Exposure Context E group was placed in context E for 30 minutes. The rats in the Long Exposure groups were placed in a clear Plexiglas shoebox in the context for the entire duration and were then taken to their home cages until testing which occurred 48 hours after training. Again, the testing procedure was the same for experiment 2 as it was for experiment 1. The Ss in groups Long Exposure Context C Same, Long Exposure Context D Same, and Long Exposure Context E Same were tested in the same context as the training context. The Ss in groups Long Exposure Context C Shift, Long Exposure Context D Shift, and Long Exposure Context E Shift were tested in a novel context that was different from the training context and different from the context they were exposed to on day 2.

30 25 Group 1: Same Group 2: Shift Group 3: Multiple Context Exposure Same Group 4: Multiple Context Exposure Shift Group 5: Long Exposure Context C Same Group 6: Long Exposure Context D Same Group 7: Long Exposure Context E Same Group 8: Long Exposure Context C Shift Group 9: Long Exposure Context D Shift Group 10: Long Exposure Context E Shift Day One Day Two Day Three Train in Context Handle for 30 minutes Test in Context A A Train in Context Handle for 30 minutes Test in Context B A Train in Context 10 minute exposure to Test in Context A A Context C, 10 minute exposure to Context D, 10 minute exposure to Train in Context A Train in Context A Train in Context A Train in Context A Train in Context A Train in Context A Train in Context A Context E 10 minute exposure to Context C, 10 minute exposure to Context D, 10 minute exposure to Context E 30 minute exposure to Context C 30 minute exposure to Context D 30 minute exposure to Context E 30 minute exposure to Context C 30 minute exposure to Context D 30 minute exposure to Context E Test in Context B Test in Context A Test in Context A Test in Context A Test in Context B Test in Context B Test in Context B Table 2. Study design for experiment 2. The ten groups are listed in the first column and the days of the experiment are listed in the second, third, and fourth columns. This table indicates what procedure each group underwent each day.

31 RESULTS Training latencies were evaluated in a simple one-way ANOVA to evaluate whether all animals exhibited comparable cross latencies, which indicates rats natural preference of the black compartment at training. The ANOVA showed there were no significant differences among the training latencies for the 10 groups (F(9,49)=.912, p=.523). Testing cross latencies for the long exposure groups were evaluated in a one-way ANOVA and used to determine retention of original fear memory where shorter cross latencies would indicate strong fear and, thus, retention of the information from the training episode. The ANOVA showed that there was a significant difference among the cross latencies for testing for the long exposure groups (F(5,23)=11.556, p<.001). A posthoc Tukey test indicated that the difference was between the group that received long exposure in Context C and was trained and tested in the same context (M=399.75, SD=279.59) and those that were tested in a context different from the training context after they received long exposure either in Context C (M=62.20, SD=51.31), Context D (M=56.20, SD=45.63), or Context E (M=31.60, SD=19.17) with p<.05. A post-hoc Tukey test indicated that there was also a difference between the group that received long exposure in Context D and was trained and tested in the same context (M=325.60, SD=262.15) and those that were tested in a context different from the training context after they received long exposure in Context E (M=31.60, SD=19.17) with p<.05 but was 26

32 27 not statistically significantly different from those groups exposed to Context C or Context D. The post-hoc Tukey test also indicated that there was a difference between the group that received long exposure in Context E and was trained and tested in the same context (M=600.00, SD=.00) and those that were tested in a context different from the training context after they received long exposure in Context C (M=62.20, SD=51.31), Context D (M=56.20, SD=45.63), or Context E (M=31.60, SD=19.66) with p<.001. Total time on white scores for testing for the long exposure groups were first evaluated in a one-way ANOVA. The ANOVA showed that there was a significant difference among the total time on white scores for testing for the long exposure groups (F(5,23)=33.671, p<.001). A post-hoc Tukey test indicated that the difference was between the groups that were tested in the original context (Same) and those groups that were tested in a novel context (Shift) with p<.001. Mean total time on white scores are shown for the long exposure groups in Figure 2. Because there was no significant difference between contexts (Contexts C, D, and E) for total time on white scores, the groups were collapsed into 2 groups, one of which included the Long Exposure groups that were tested in the same context they were trained in and the other which included the Long Exposure groups that were tested in a novel context that was different from the training context. A one-way ANOVA was used to test any differences in the cross latency scores at testing for those groups that received multiple context exposures, those who received long contextual exposures to one context, and those who received no experimental manipulation during the retention interval between training and testing. The ANOVA

33 28 showed that there was a significant difference among the cross latency scores at test for the 6 groups (F(5,53)=6.518, p<.001). A post-hoc Tukey test was done to investigate the differences in means between the 6 groups. The Tukey test showed that a difference between the Shift group that was tested in a context different from the training context with no experimental manipulation during the retention interval (M=65.50, SD=108.82) and the group that received long exposure to a single context and was trained and tested in the same context (M=444.79, SD=233.54) with p<.01. The Tukey test also revealed a difference between the group that received multiple context exposures during the retention interval but were trained and tested in the same context (M=377.50, SD=289.25) and the group that received long exposure to a single context during the retention interval and was tested in a context different from the training context (M=50.00, SD=40.50) with p<.01. The Tukey test revealed no significant differences in cross latency scores during test between the group that received multiple contextual exposures and was tested in a context different from the training context (M=195.25, SD=264.56) and any of the other groups analyzed. A one-way ANOVA was used to test any differences in the total time on white scores during testing among the 6 groups (Same, Shift, Multiple Context Exposure Same, Multiple Context Exposure Shift, Long Exposure Same, and Long Exposure Shift). The ANOVA showed that there was a statistically significant difference for total time on white during training for the 6 groups (F(5,53)=30.139, p<.001). A post-hoc Tukey test showed that the difference was between the Same and Shift groups, Same and Long Exposure Shift groups, Shift and Multiple Context Exposure Same groups, Shift and

34 29 Multiple Context Exposure Shift groups, Shift and Long Exposure Same groups, and Long Exposure Same and Long Exposure Shift groups (p<.001). The context shift was found in the following conditions: group Same, which was trained and tested in the same context with no experimental manipulation on day two, and group Shift, which was trained in one context and tested in a different context with no experimental manipulation on day two, shows the usual context shift effect; group Same and the Long Exposure Shift group, which was trained in one context and tested in another with exposure to one context for an extended period of time on day two; group Shift, which was trained in one context and tested in another with no experimental manipulation on day two, and Long Exposure Same group, which was trained and tested in the same context with exposure to a single context for a prolonged period of time on day two; the Long Exposure groups based on the statistical significance between Long Exposure Same group and Long Exposure Shift group; the Multiple Context Exposure Shift group, which were exposed to multiple contexts on the second day and tested in a novel context, and the Shift group, with the Multiple Context Exposure Shift group showing similar amounts of fear as the Same group, which received no experimental treatment and was trained and tested in the same context. Mean total time on white scores are shown for all groups, including collapsed Long Exposure groups, in Figure 3.

35 30 Figure 2. Mean total time on white (TTW) scores (±SEM) for Long Exposure (LE) groups in experiment 2 at retention test. The three Same groups (LE Context C Same, LE Context D Same, & LE Context E Same) that were trained and tested in the same context with exposure to a single context during the retention interval were significantly different from the three Shift groups (LE Context C Shift, LE Context D Shift, & LE Context E Shift) that were trained in one context and tested in a novel context with exposure to a single context during the retention interval.

36 31 Figure 3. Mean total time on white (TTW) scores (±SEM) for all groups in experiment two at retention test, including collapsed Long Exposure (LE) groups. The Same and Shift groups were statistically significantly different which indicates the context shift effect occurred. The Multiple Context Exposure groups (MCE Same & MCE Shift) were not statistically significant from one another nor from the Same group indicating retroactive interference from the multiple contexts which resulted in a flattening of the generalization gradient. The Long Exposure Same and Long Exposure Shift groups were statistically significant from one another which shows the usual context shift effect comparable to the normal context shift effect seen with the Same and Shift groups comparison. This indicates that retroactive interference effects occurred because of the exposure to multiple contexts and was not just time dependent nor was there anything particularly salient about any one of the contexts used during exposure.

37 DISCUSSION The results from experiment 2 support the hypothesis that exposure to novel stimuli during the retention interval serve as a source of retroactive interference resulting in a more rapid loss of the context shift effect. The groups that were exposed to multiple contexts between training and testing showed an alleviation of the decrement in performing usually seen when being tested in a novel context, based on the total time on white scores. This indicates that the multiple exposures that these groups experienced acted as retroactive interference for the memory for stimulus attributes from the original training context and, because of the retroactive interference, the rats failed to distinguish the training context and the novel testing context. The results also showed that interference was not an issue of the amount of time in a novel context, as can be seen by the total time on white scores for the long exposure control groups. Because the long exposure control groups, that were exposed to only one novel context for the same amount of time the multiple exposure groups received exposure, did not show a flattening of the stimulus generalization gradient, it can be inferred that the alleviation of the performance decrement in the multiple context exposure groups was due to having been exposed to multiple contexts and experiencing multiple types of stimuli during the retention interval. The long exposure groups used in this study not only allows us to control for any effect duration of time exposed may have had on the forgetting of stimulus attributes but also controlled for the possibility that one 32

38 33 context may have proven to be particularly effective at serving as retroactive interference. Because none of the contexts were statistically significantly different from one another we can deduce that none of the contexts were particularly salient on their own and the retroactive interference effect found was a result from a combination of multiple contexts. This is important in understanding some of the determinants of forgetting of stimulus attributes and what can serve to interfere retroactively with an encoded learning episode and the contextual cues that go along with it.

39 GENERAL DISCUSSION The results of experiment 1 and 2 have given some insight into what can affect the forgetting of stimulus attributes. With experiment 1, the projected results were not obtained, but the results still give some idea of what can affect memory for stimulus attributes. If the results were due to a reminder effect, as postulated, then this indicates that perhaps the usual flattening of the stimulus generalization gradient can be avoided by using reminder treatments even in novel contexts. While previous studies have investigated in part, they used footshock reminders in new contexts to serve as a reminder of the original training (McAllister and McAllister, 2006; Rescorla and Heth, 1975). Zhou & Riccio (1994) showed that you can use a reminder treatment to alleviate the context shift effect but their reminder treatment involved exposure to the training context. The results from the first experiment may indicate that a footshock nor exposure to the original training stimulus is not necessary for a reminder treatment to be effective, but perhaps just being placed in a similar apparatus without exposure to the feared side can provide enough cues to serve as a reminder of original training. The results from the second experiment give some insight into what can affect the forgetting of stimulus attributes. The finding that being exposed to multiple contexts without any training or apparatus exposure can interfere with the memory for stimulus attributes from the original training context further reinforces the importance of context in learning and memory and indicates that contextual stimuli are subject to retroactive 34