UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Mood dependent memory : extension and validation Macaulay, Dawn Leigh 1997

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

Item Metadata


831-ubc_1997-251020.pdf [ 5.63MB ]
JSON: 831-1.0088071.json
JSON-LD: 831-1.0088071-ld.json
RDF/XML (Pretty): 831-1.0088071-rdf.xml
RDF/JSON: 831-1.0088071-rdf.json
Turtle: 831-1.0088071-turtle.txt
N-Triples: 831-1.0088071-rdf-ntriples.txt
Original Record: 831-1.0088071-source.json
Full Text

Full Text

MOOD DEPENDENT M E M O R Y : EXTENSION A N D V A L I D A T I O N by D A W N L E I G H M A C A U L A Y B. Sc., The University of Toronto, 1987 ~ M A . , The University of Alberta, 1990 A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF THE REQUIREMENTS FOR THE D E G R E E OF DOCTOR OF PHILOSOPHY in THE F A C U L T Y OF GRADUATE STUDIES Department of Psychology We accept this thesis as conforming to the required standard T H ^ U N I V E R S I T Y OF BRITISH C O L U M B I A April 1997 © Dawn Leigh Macaulay, 1997 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. * ' .(.,, ••• Department of fey cliolo^ The University of British Columbia Vancouver, Canada Date Q r y ^ o l T ; DE-6 (2/88) Abstract Recent studies have indicated stronger mood dependent memory (MDM) effects when subjects generate both the to-be-remembered events and the cues used in their retrieval, and when subjects who experience similar moods are compared to subjects who report a change in both pleasure and arousal, rather than others who report a change in pleasure alone. The first study was undertaken to experimentally evaluate the relationship between mood change and memory performance. Specifically, four mood manipulations were developed to compare memory performance in subjects assigned to no mood change, change in pleasure, change in arousal, or change in both pleasure and arousal. Subjects generated autobiographical events in response to neutral nouns, and were tested for free recall and recognition of these events two days later. Results demonstrated greater memory disruption after change in both pleasure and arousal than after change in pleasure alone. However, one-dimensional change in arousal led to as much memory disruption as did change in both dimensions. Separate A N O V A s compared each mood in combination with each other mood and indicated that M D M was more strongly supported in some combinations than others. Further, the pattern in free recall was different from the pattern in recognition indicating that M D M effects may vary interactively across moods and tasks. The second study was undertaken to validate prior research by employing subjects who experience large, but naturally occurring, changes in mood. Patients with rapid-cycling Bipolar Disorder performed a variety of tasks during study sessions and were tested for implicit and explicit memory during test sessions. Significant M D M effects were supported in free recall of autobiographical events and inkblot recognition, but not in implicit category production of autobiographical event probes, picture fragment completion, explicit nor implicit memory for letter associates. Results validate M D M as a real world phenomenon that impacts on explicit memory performance in subjects whose moods are not manipulated. Discussion centers on prospects for future studies that consider a wider range of manipulated moods and tasks, theories of emotion, and other clinical groups in the investigation of M D M . iv Table of Contents Abstract ii Table of Contents iv List of Tables vi List of Figures vii Acknowledgements viii Chapter 1. Introduction 1 1.1 History 4 1.2 Current Directions 6 1.2.1 Nature of the Manipulated Mood 7 1.2.2 Nature of the Study Task 12 1.2.3 Nature of the Memory Test 14 1.2.4 Nature of the Differences in Mood 18 1.3 Summary 19 Chapter 2. Experiment 1: Investigation of Moods that Vary on Pleasure and Arousal 21 2.1 Issue to be Resolved 21 2.2 Method 23 2.2.1 Mood Induction Development 23 2.2.2 Design 24 2.2.3 Study Session 24 2.2.4 Test Session 26 2.3 Subjects 27 2.4 Results . 28 2.4.1 Response to Mood Manipulations 28 Affect Grid Ratings 28 PANAS Ratings 32 Time to Achieve Induced Moods 34 Genuineness Ratings 34 Summary 35 2.4.2 Autobiographical Event Generation 36 2.4.3 Free Recall Performance 37 2.4.4 Recognition Performance 42 2.4.5 Summary of Results 46 2.5 Discussion 48 V Chapter 3. Experiment 2 Investigation of Mood Dependence in Bipolar Affective Disorder 55 3.1 Issue to be Resolved 55 3.2 Method 57 3.2.1 Subj ect Recruitment 5 7 3.2.2 Design 58 3.2.3 Study Session 58 3.2.4 Test Session 59 3.3 Subjects 61 3.3.1 Medication Status and Diagnoses 62 3.4 Results 64 3.4.1 Mood Measures 64 3.4.2 Results of Study Session Tasks 65 Autobiographical Event Generation 65 Inkblot and Picture Ratings 67 Latency of Letter Associations 67 Summary of Study Session Results 68 3.4.3 Results of Test Session Tasks 68 Free Recall of Autobiographical Event Generation Probes 69 Category Production Priming 71 Inkblot Recognition 73 Picture Fragment Priming 74 Explicit Memory for Letter Associates 75 Implicit Memory for Letter Associates 76 Summary of Mood Dependent Memory Effects 77 3.4.4 Summary of Results 78 3.5 Discussion 79 Chapter 4. General Discussion 87 References 91 Appendix I Musical Selections 97 Appendix II M C M Analysis of Event Ratings 99 Appendix III Example of Picture Fragment Stimuli 1 0 3 vi List of Tables Table Title Page 1 Recall of Items as a Function of Pleasure/Arousal Change 19 2 Study Session Pleasure Ratings as a Function of Induced Mood 28 3 Test Session Pleasure Ratings as a Function of Induced Mood 29 4 Study Session Arousal Ratings as a Function of Induced Mood 30 5 Test Session Arousal Ratings as a Function of Induced Mood 31 6 Study Session Positive Affect Ratings as a Function of Induced Mood 32 7 Test Session Positive Affect Ratings as a Function of Induced Mood 33 8 Study Session Negative Affect Ratings as a Function of Induced Mood 33 9 Test Session Negative Affect Ratings as a Function of Induced Mood 34 10 Time to Reach Mood Criteria 34 11 Genuineness Ratings as a Function of Induced Mood and Session 35 12 Percent of Event Types Generated as a Function of Induced Mood 36 13a Percent of Targets Free Recalled as a Function of Induced Mood 37 at Study and Test 13b Free Recall as a Function of Pleasure/Arousal Change 37 13c Free Recall as a Function of Induced Mood at Study and Test 38 14 Absolute Differences between Study and Test Ratings of Pleasure 40 and Arousal as a Function of Induced Mood across Sessions 15 Free Recall for Subjects Meeting More Stringent Study Pleasure Criterion 41 16a % Hits - % False Alarms as a Function of Pleasure/Arousal Change 42 16b % Hits - % False Alarms as a Function of Induced Mood 43 at Study and Test 17 d' as a Function of Induced Mood at Study and Test. 44 18 Recognition of A l l Generated Items % Hits - % False Alarms 45 as a Function of Induced Mood at Study and Test 19 Medication Status 63 20 Pleasure and Arousal Ratings as a Function of Study/Test Condition and Session 64 21 P A and N A Ratings as a Function of Study/Test Condition and Session 64 22 Percent of Event Types Generated as a Function of Mood 65 23 Free Recall as a Function of Study/Test Mood 69 24 Category Production Priming as a Function of Study/Test Mood 71 25 Inkblot Recognition as a Function of Study/Test Mood 73 26 Picture Fragment Priming as a Function of Study/Test Mood 74 27 Recall of Letter Associates as a Function of Study/Test Mood 75 28 Re-Association to Letter as a Function of Study/Test Mood 76 29 Summary of M D M Effects across Analyses and Tasks 77 30 Percent of Events Generated, Generation Latency, Rated Intensity, 99 Importance and Vividness as a Function of Mood and Event Type 31 Percent of Events Generated, Generation Latency, Rated Intensity, 101 Importance and Vividness as a Function of Mood and Event Type vii List of Figures Figure Title Page 1 The Affect Grid 9 2 Pleasure and Arousal Ratings After Mood Induction: Session 1 11 3 Eich and Metcalfe (1989), Experiment #1 13 4 Eich et al. (1994), Experiment #1 13 5 Category Production 17 6 Relationship between Affect Grid and PANAS Scales 21 7 Quadrants on the Affect Grid 25 8 Relationship among M D M effects in Free Recall 48 9 Relationship among M D M effects in Recognition 49 Acknowledgments I would like to thank the members of my committee, Dr. Stephen Schertzer, Dr. Rod Wong, Dr. Jim Russell, and Dr. Peter Suedfeld, for their comments and for making the last stage of my doctoral studies a pleasant one. I would especially like to thank my research supervisor Dr. Eric Eich for the opportunity to pursue these studies which have allowed me to grow as a researcher, as well as for his continuous support and encouragement. Special thanks are also due to Dr. Raymond Lam and Ms. Arvinder Grewall, at U B C Mood Disorders Clinic, who made the clinical study possible. I would like to thank the research assistants who worked on the laboratory investigation, Andrea Bull, Susan Carsky, Dawn Layzell, Adam Margesson, Kirsten Campbell, and Reiko West. And to my comrades in arms Becky Mills, Monica Mori, and Maja Lange, go my thanks for their support. My mother, sister and grandmother I thank for the encouragement to continue to pursue my dreams and for their belief that I could achieve them. I would like send my love and thanks for all his understanding and patience to Jordan Elliott who has been my rock and who has seen me through the many highs and lows of my career and life. Finally, I would like to dedicate this dissertation to the memory of my father, John Edwin Macaulay. C H A P T E R 1. INTRODUCTION Emotions are inexorably mingled with our life experiences. Joys and sorrows define pivotal moments in our personal stories and mark major milestones. Remembering these triumphs or tragedies may later trigger in us the re-experience of emotions we felt during the occurrence of such events. Experiencing a strong mood also may remind us of other circumstances in which we experienced similar emotions. The connection between emotion and memory poses puzzles about how the two are connected and under what circumstances one will or will not be influenced by the other. Although the interaction between emotion and thought has been viewed as important from the time of William James, experimental evaluation of the relationship has a relatively short past which has focused on two emotions, namely elation and depression. One emotional effect is mood dependent memory (MDM). Recollection of information to which one is exposed during a strong emotional reaction is often poor later when the mood is no longer experienced. In other words, attempts to recall information when experiencing a mood unlike that experienced during the original event may be less successful than attempts made during a similar mood. This memory effect has been hypothesized to play a causal role in the memory losses associated with some clinical disorders such as dissociative identity disorder and psychogenic fugue (Schacter & Kihlstrom, 1989). A second emotional effect is mood congruent memory (MCM). Investigations have shown that, under certain circumstances, one's current mood has the tendency to influence the valence or tone of materials attended to and remembered (Blaney, 1986). Generally, unpleasant moods enhance memory for negative material and pleasant moods enhance memory for positive material. Depression has long been associated with just such a negative memory bias (Beck, 1967). The "negative cognitive loop" caused by M C M has been hypothesized to maintain depressed mood once it has begun (Ingram, 1984). Thus, mood and its associated effects on Mood Dependent Memory 2 memory have been implicated in significant clinical problems as well as in more mundane aspects of everyday life. Although mood congruent effects on memory have proved to be reliable and robust, the same cannot be said for mood dependent effects. In this dissertation I aim to validate, replicate, and extend our current knowledge of M D M , via two dissimilar approaches. The first study is a laboratory investigation which focuses on specific predictions about how the nature of mood change impacts on the production of mood dependence. Laboratory investigations of mood dependence require two sessions during which subjects are assigned to experience, typically, either happy or sad moods. During the initial session, subjects are exposed to information while experiencing such a mood. Later, some subjects re-experience similar moods while others experience dissimilar moods, and all attempt to remember information presented in the initial session. Typically, studies comprise a 2 x 2 design in which study mood (happy or sad) is crossed with test mood (happy or sad). If subjects assigned to matching moods remember more information than do subjects assigned to mismatching moods, the experiment demonstrates mood dependence. Though past research has been mixed, current researchers have begun to make progress. One group of investigators has demonstrated and replicated M D M (Eich & Metcalfe, 1989; Eich, Macaulay, & Ryan, 1994). And, though the effects have been reliable, there remain a number of issues surrounding their occurrence. One such issue concerns the quality of the mood change for subjects in different moods. The publications presented a total of seven studies in which M D M was demonstrated. A l l subjects assigned to mismatching moods were required to show a large change in pleasure; some reported an associated change in arousal while others did not. A median split broke mismatched mood subjects into two groups; those reporting change in pleasure alone, versus those reporting change in both pleasure and arousal. Subjects assigned to matching moods recalled most, while subjects who reported change in pleasure alone recalled a moderate amount, and while subjects who reported change in both pleasure and arousal recalled Mood Dependent Memory 3 fewest items. Although this pattern held over seven studies, it is correlational only and suggests the need for further investigation. In the current investigation, subjects were assigned to one of four moods; either delight, sadness, calmness or anxiety. During the experience of one of these moods, subjects were exposed to some information. Later, they returned to the laboratory and experienced a second mood, either matching or mismatching before recalling that information. Crossing these four moods allows examination of M D M in moods that differ in systematic ways. Different mood combinations will allow comparison of memory effects when subjects were assigned to experience moods differing on pleasure, on arousal, or on both dimensions. As covered in detail later, theories of emotion can be interpreted to provide opposing predictions about which combinations of moods should be "most" different. The purpose of this study is to determine i f the previously reported pattern is replicable and if memory performance is predictable from the literature on the nature of emotion. The second study is a more exploratory investigation of clinical subjects suffering from rapid-cycling Bipolar Depression. The purpose of this study is to determine if M D M effects are evident in populations in which strong, but naturally-occurring, changes in mood state are experienced. If M D M is to be viewed as a primary cause of amnesias reported in certain clinical populations, such as psychogenic fugue or dissociative disorder, then memory impairment should also be associated with the extreme moods and mood shifts experienced by manic-depressive patients. Although many researchers have argued that the effects may explain the amnesias associated with certain clinical disorders (Schacter & Kihlstrom, 1989; Nissen, Ross, Willingham, Mackenzie, & Schacter, 1988), there remains only one direct investigation of mood dependence in a clinical population (Weingartner, Miller & Murphy, 1977). Participants were followed through periods of mood stability in mania or hypomania and depression as well as through periods of mood change. By employing tasks which vary on certain characteristics, such as the implicit or explicit nature of the memory demands, the contributions of these characteristics to the production of M D M may be clarified. Examination Mood Dependent Memory 4 of task performance in individuals whose moods are naturally occurring also provides an opportunity to validate and extend findings that have previously been restricted to the laboratory. 1.1 HISTORY OF MOOD DEPENDENT MEMORY RESEARCH Bower and his colleagues are credited with establishing the experimental investigation of the connection between emotion and cognition (Macaulay, Ryan, & Eich, 1993). For almost two decades, Bower has focused on understanding both mood congruent and mood dependent retrieval. Although experimentally distinct, the cognitive mechanism which supports these memory effects may be one and the same. In both cases, according to Bower (1992), moods are represented as nodes in the associative network of knowledge. Mood nodes are connected or associated to certain other concepts in the network. When moods are experienced, these mood nodes are activated. This activation spreads along the associated pathways, sending some subthreshold amount of energy to associated concepts. In accordance with this view, mood congruence is evident because these concepts, given the residual subthreshold activation, are easier than other non-associated concepts to bring to mind. Mood dependence is simply the extension of these ideas to include the creation of new connections between a mood node and previously non-associated concepts through co-occurrence during the study session. The later experience of a similar mood will cause "spreading activation" through the network which will render associated concepts closer to threshold. Again, these mood matching thoughts will require less additional activation to reach their threshold than will other non-mood related events. Thus, they should be easier to bring to mind during a matching mood than during a mismatching mood. Mood dependence was first demonstrated in a laboratory when hypnotized subjects learned 2 lists, one while happy and the other while sad. Later, subjects re-experienced one mood and were asked to remember both previous lists. The experience of either happiness or sadness during memory testing led to greater recall of the list learned in that mood than in the opposite mood (Bower, Monteiro & Gilligan, 1978). Other labs reported M D M and extended the Mood Dependent Memory 5 original reports to include moods induced via non-hypnotic methods (Leight & Ellis, 1981; Schare, Lisman, & Spear, 1984), and to include other populations such as children (Bartlett & Santrock, 1979; Bartlett, Burleson, & Santrock, 1982). At the time, M D M and M C M seemed equally reliable and well understood. This positive sense was conveyed by Bower's (1981) suggestion that the M D M paradigm might eventually serve as a tool to investigate the relationship among emotions. He discussed Plutchik's (1980) emotion theory which describes the relationship among emotions in a circumplex model. Some emotions, such as fear and anger, are considered to be opposites and thus lie directly across from each other in the model. Other emotions, such as contempt, are seen as blends of two emotions, like disgust and anger, and are located between those emotions. Bower reasoned that mood dependence may be an independent means of validating claims about relative distance between emotions. Plutchik's model could be taken to mean that the experience of a nominally different mood at test would not disrupt memory performance if it was close to the study mood on the circumplex, say anger and disgust. However, the experience of a test mood which opposes the study mood on the circumplex may show most memory disruption. At the time, M D M was an emerging field of study, with its major investigator envisioning it as a future tool to contribute to our understanding of the relations among emotions. However, by the mid 1980's, researchers began reporting difficulty in demonstrating the basic phenomenon. Blaney (1986) reviewed 20 studies of which only 7 supported Bower's earlier results. He concluded that mood dependent memory was an unreliable effect. Since then, other researchers have also published failures to replicate (e.g., Bower & Mayer, 1985; Wetzler, 1985; Johnson & Klinger, 1988; Haaga, 1989). Bower (1987) noted that in most studies, the materials to be remembered comprised a list of neutral probe words. He reasoned that learning such a list while concurrently experiencing a mood may provide only a weak link in memory between the mood and materials, i f one existed at all. He further hypothesized that M D M may be encouraged if the materials to be remembered are somehow deeply or causally connected to the experience of that mood. The hypothesis was operationalized by having mood induction instructions serve as Mood Dependent Memory 6 memory test targets. Bower and Mayer (1989) reported six studies in which they attempted to produce M D M . The only study to demonstrate the effect employed the causal belongingness methodology but was immediately followed by a failure to replicate. In the end, they exclaimed: "We have hit an impasse in our experiments on M D R (mood dependent retrieval). We have searched in vain for some variable that we consider "significant" which would distinguish among experimental arrangements that produced positive versus negative results on MDR" (Bower & Mayer 1989, pp. 153-154). 1.2 CURRENT DIRECTIONS Major memory theories, such as Hull's (1943) stimulus-drive approach, Tulving's (1983) encoding specificity principle, and Morris, Bransford, and Franks' (1977) transfer-appropriate processing approach, all predict mood dependent effects on memory. Each theory predicts that the greater the similarity between learning and testing situations, the greater the probability that previously learned information will be reproduced. This consensus makes it important to understand the conditions under which mood dependence will and will not be demonstrated. Furthermore, many psychologists consider the mood dependent aspect of memory to play a causal role in production of significant memory disturbances associated with clinical disorders such as traumatic amnesia, psychogenic fugue, dissociative identity disorder, and depression (Bower, 1981; Schacter & Kihlstrom, 1989; Nissen et al., 1988; Teasdale 1983). If so, it may be possible to duplicate these memory disruptions and amnesias, although to more modest degree, in the laboratory. Thus, from a clinical perspective, it is also important for us to understand basic mood dependent effects. Finally, given the theoretical connection between mood congruent and mood dependent effects it is unclear why one effect would be consistently demonstrated across numerous laboratories and numerous induction techniques whereas the other effect is demonstrated only sporadically (see Blaney, 1986). Thus, from the perspective of mood research in general, the puzzle of mood dependence requires solution. Mood Dependent Memory 7 Although the investigation of mood dependence has been said to reveal more questions than answers (Kihlstrom, 1989), some progress now has been made. Eich and his colleagues have successfully produced and replicated mood dependent effects (Eich & Metcalfe, 1989; Eich et al., 1994). These successful demonstrations have allowed specification of four critical features which must be addressed in any study of mood dependence (Eich, 1995). 1.2.1 NATURE OF THE MANIPULATED MOOD The most fundamental element in investigation of mood dependence is the experience of two moods which are similar for some subjects but dissimilar for others. The large majority of investigators have used mood manipulation techniques to achieve this end. Mood manipulation has been criticized on the basis of demand characteristics and the possible production of "mood" effects via means other than mood (Hasher, Rose, Zacks, Sanft & Doren, 1985). As Ellis and Ashbrook (1989) point out, the cardinal advantage of mood induction is to allow causal statements about effects produced by that manipulation. However, this is no guarantee that such a manipulation influences mood alone. Each manipulation is associated with certain advantages and weaknesses. The issue, then, becomes one of ensuring that a mood induction meets key requirements. One key feature in M D M research is that produced moods should be relatively intense. If target materials are associated with a particular emotion in memory, one can assume that there will be graded response generalization around that emotion cue (Bower, 1992). A later mismatching mood may fall in that range of graded response unless induced moods are sufficiently intense to reduce this possible overlap. Thus, detection of mood effects should be improved by induction of intense moods. In fact, there is some evidence to suggest that M D M may become evident i f and only if one experiences some requisite intensity of mood change (Ucros, 1989). A second feature is to ensure that each and every subject experiences the assigned mood state. Although mood groups may differ on rated mood after an induction, some proportion of subjects may not be experiencing the mood, which would impede detection of mood effects. Mood Dependent Memory 8 Many researchers employ a set of visual analog scales (VAS) to serve as a manipulation check, and require subjects to report a difference of at least 10% from pre- to post-induction measures for the manipulation to be considered successful (Martin, 1990). A problem is that, though researchers may measure numerous attributes, typically a single scale is used for evaluation. The overall quality of mood may not be captured by this measurement approach. For example, a person could begin a study indicating no depression and high elation on two separate scales. After induction, ratings of slight depression and moderate elation on these two scales would meet the 10% change criterion for induced depression. Subjects reporting moods characterized by moderate elation are unlikely to show effects of sadness and their inclusion would only serve to diminish any mood dependent effect. A third feature is that moods should be durable enough to withstand performance of a cognitive task. Though many researchers measure mood before subjects begin a task, few measure it upon task completion. Eich and Metcalfe (1989) reported a pilot study in which the Velten (1968) technique, which consists of self-statements that begin as neutral and become increasingly elated or depressed, produced significant group differences on immediately measured mood. However, moods measured upon completion of a short cognitive task did not differ by induction condition. Chartier and Ranieri (1989) measured mood before the Velten induction and every 6 min. thereafter for 30 min. They reported that with no intervening task the influence of both depression and elation inductions had ended in 6 to 12 min. A fourth feature is that produced moods should be genuine. Although many researchers have discussed the problem of demand characteristics in mood manipulation techniques (Polivy & Doyle, 1980; Buchwald, Strack, & Coyne, 1981), there are few published data that speak to the issue of genuineness. Bogus feedback techniques have been employed by social cognition researchers to produce moods while reducing possible demand characteristics. Subjects are given predetermined feedback about their performance on personality tests (Forgas & Bower, 1987), analogies tests (Forgas, Bower, & Moylan, 1990), or computer games (Isen, Clark, Shalker, & Karp, 1978). An advantage of these approaches is that subjects are not actively Mood Dependent Memory 9 attempting to achieve any particular mood state. Subjects are not aware of the true purpose of the tasks and simply react, presumably with authentic mood change, to the feedback. Although these manipulations avoid an artificial situation, they also have drawbacks. There are ethical concerns about manipulating mood without prior informed consent and thus a limit on the intensity of mood that experimenters can attempt to induce with this method. Chartier and Ranieri (1989) reported that failure feedback produced less intense mood change than did the Velten technique; however, with no intervening task, the mood was maintained for nearly 30 min. In contrast, success feedback produced mood change which abated after only 6 min. With these features in mind, Eich and his colleagues have developed a mood manipulation, versions of which have been introduced in Eich and Metcalfe (1989) and Eich et al. (1994). The technique, dubbed LIRIC , has a number of components that together distinguish it from other manipulations. First, it requires subjects to Z/sten to appropriate selections of classical music. Second, it involves personal Reflection, meaning that subjects contemplate previous autobiographical occurrences, or their own constructed imaginings, to aid development of a particular mood. Third, it is applied on an /deographic basis, meaning that success of mood induction for any subject will be determined by that individual subject's response. Finally, it is Criterion based, meaning that a certain intensity of mood must be achieved before the manipulation is considered successful. The LIRIC mood induction has been an important advance over other techniques and has played a significant role in the production of reliable mood dependent effects. In a typical LIRIC study, subjects are advised that their current mood will be periodically assessed by means of the matrix in Figure l~an adaptation of the Affect Grid developed by Russell, Weiss, and Mendelsohn Figure 1. The Affect Grid. Extremely High Arousal c CO CO CO _QJ Q . c j>. cu E CD 5 •R LU 1 Extremely Low Arousal Mood Dependent Memory 10 (1989). The matrix was designed to measure two principle components of current emotional experience, namely pleasure (horizontal axis) and arousal (vertical axis). Subjects are advised that, beginning on the left, a mark in one of the columns connotes a mood that is extremely, very, moderately, or slightly unpleasant, neutral in the center, and slightly, moderately, very and extremely pleasant on the right. Beginning at the top, arousal is indicated by the placement of a mark in one of the rows ranging from extremely, very, moderately, and slightly high arousal, to neutral in the center, and to slightly, moderately, very and extremely low arousal. Prior to mood induction, subjects make their baseline mark, and are apprised that they will listen to selections of classical music that should help them develop a pleasant (or unpleasant) mood. It is stressed that because music alone cannot create the desired state, they should concentrate on ideas or images relating to real-life incidents that make them feel pleasant (or unpleasant). Subjects are advised that to aid them in achieving and maintaining as intense a state as possible, the mood appropriate music will continue to play in the background for the entire session once it started. Subjects are seated in a comfortable lounge chair, bordered by stereo speakers, and are left in the room with the music and their thoughts. Experimenters return to the room on 5 min. intervals and subjects rate current mood on a clean copy of the Affect Grid. Once a matrix is marked, experimenters determine i f subjects are ready to go on with the rest of the experiment which occurs when pleasant mood (P) subjects mark any of the squares in the two right-most columns of the mood matrix or when unpleasant mood (U) subjects check any of the squares in the two left-most columns. Subjects are not told that the start of testing is contingent on their achieving a criterion level of pleasure or displeasure, lest they try to rush matters by rating their mood as being more extreme than it really is. The LIRIC technique seems to work well. Across a number of recent studies, 88% of subjects succeeded in attaining the criterion level of mood in both of two separate sessions. The success rate for women (88%) was comparable to that for men (86%). Marks on the Affect Grid are translated into two 9-point scores indicating degree of pleasure and arousal. The scores range from 4 to -4 on each dimension (e.g., extremely pleasant Mood Dependent Memory 11 to extremely unpleasant). Figure 2 presents typical scores at the beginning of a session (BSL), once the pleasure criterion was reached (CRT) and after the performance of the cognitive task (PST). Pleasure ratings taken when subjects reached the induction criterion show an impressive group difference, a difference that must be achieved by definition of that criterion. Pleasure ratings of the P group were readily maintained for the period of the task, whereas those of the U group had a greater tendency to dissipate. However, analyses indicated that pleasure ratings differed reliably between groups post-task. Though the criterion restricted pleasure alone, arousal ratings also differed substantially across groups at criterion, a difference that remained reliable post-task. Mood ratings gathered in the second session duplicated these patterns. Large group differences were evident in both pleasure and arousal ratings when subjects achieve the pleasure criterion. Although differences were diminished post-task, they remained reliably different across groups. In two recent experiments, subjects were asked to evaluate the extent that their induced moods resembled real moods. Subjects rated their moods on an 11 point scale on which 0 indicated extreme artificiality, 5 indicated moderate genuineness and 10 indicated extreme genuineness. Although P moods achieved higher genuineness ratings, (P and U mean ratings of 8.3 and 7.4, respectively), the large majority of subjects in both mood conditions rated the moods as more than moderately genuine. The L I R I C induction creates intense moods in both happy and sad groups ensuring large differences on pleasure ratings. Arousal differences are also produced, although they are not Figure 2. Pleasure and Arousal Ratings. Session 1 4 3 2 -4 ' BSL CRT • Pleas-P • Pleas-U M Arous-P • Arous-U Mood Dependent Memory 12 required by the technique. Importantly, the ideographic application of the induction ensures that every successful participant experiences intense moods. Further, the moods are durable enough that, though performance of a task diminishes the moods, the differences between groups remain significant post-task. Finally, subjects perceive the moods as resembling moods they experience in real life, rating them as moderately genuine. As such, the LIRIC procedure demonstrates all four attributes which form the necessary foundation on which to establish the investigation of mood dependent effects. 1.2.2 NATURE OF THE STUDY TASK A second area of consideration in investigation of mood dependence is the nature of the study task. Eich and Metcalfe (1989) noted that tasks which seem most sensitive to mood are those like word association and hypothesized that events arising from internal processing would be more closely associated to mood than would be events arising from perception of the external world. Whereas internally generated events should generally be better recalled than those arising externally—the standard generation effect (Slamecka & Graf, 1978)—their recall may also be more disrupted by a change in mood. Subjects were induced into pleasant or unpleasant moods via the LIRIC technique. The experimenter orally presented subjects with a category name, an exemplar, and either a second exemplar, which subjects simply repeated aloud (e.g., ICE CREAM FLAVORS: CHOCOLATE -VANILLA), or the first letter of a second exemplar, which subjects used to generate an appropriate response aloud (e.g., MUSICAL INSTRUMENTS: DRUM- G). Two or three days later, after induction of a matching or mismatching mood, subjects were asked to remember the words they said aloud in the previous session. Subjects who experienced similar moods in the Test session recalled more generated words than did subjects whose moods were dissimilar (Figure 3). This pattern was also evident, although more weakly so, among recall of repeated words. Thus, the first experiment provided evidence of mood dependent recall among both generated and repeated words. Mood Dependent Memory 13 Figure 3. Eich & Metcalfe, (1989, Exp. 1) 35 30 = 2 5 8 20 CD a. 15 * 10 5 0 P / P P / U U / P U / U • Generate • Repeat Considering the four experiments as a whole, mood dependent effects were consistently evident among recall of targets that had been generated by subjects. That is, for generated words, experiencing a mood that matched the mood at study was a significant advantage to subjects' memory performance. In stark contrast to reports by other investigators, Eich & Metcalfe (1989) were reporting reliable mood dependent effects. These results have recently been replicated in an independent laboratory (Beck & McBee, 1995) lending further support to the hypothesis that items arising as a result of internal processing are more deeply connected to mood and thus are more susceptible to mood dependent effects. Recent investigations have focused on the recollection, description, and rating of autobiographical events, a task which is designed to amplify the connection between target materials and subject mood by increasing reliance on internal reasoning and thought (Eich et al., 1994). Subjects, experiencing pleasant or unpleasant moods, were required to recollected events from their personal past in response to common-word probes such as SHIP and STREET. After describing the event aloud, subjects rated it along several dimensions including its intensity, importance, and vividness. In the test session, following induction of matching or mismatching moods, subjects were asked to recall as many of the events or probe words from the first session as possible. A l l three studies provided consistent evidence of Mood Dependent Memory 14 mood dependent memory (Figure 4). Specifically, in each case, subjects who experienced matching moods recalled 7-9% more of the total target events than did subjects whose study and test session moods did not match. Eich (1995) has proposed the "do-it-yourself hypothesis, a view of M D M research which states, in part, that events generated through internal mental operations such as reasoning, imagination, or interpretation may be more closely connected to or deeply coloured by one's current mood than are events whose processing is guided by external sources. By this view, the demonstration of mood dependence requires the selection of study tasks which encourage deep connection between the mood and the materials. The later experience of a different mood will then dissimilarly impact upon reasoning, imagination or interpretation which will make remembering the target materials more difficult. 1.2.3 NATURE OF THE MEMORY TEST A third consideration in the study of mood dependence is the nature of memory tests. In each of 4 experiments, Eich and Metcalfe (1989) followed free recall with tests of recognition memory. Subjects were tested for recognition of words they repeated or generated via the category cues and exemplars. Although free recall of generated words was mood dependent in all four experiments, and although free recall of repeated words was more weakly, but nonetheless significantly, mood dependent in three experiments, recognition memory provided no evidence of mood dependence. Thus, like study tasks, memory tests differ in their sensitivity to changes in mood state. In an attempt to replicate this finding, Beck and McBee (1995) followed free recall with recognition. Their data indicated mood dependent recognition which was somewhat stronger for repeated than generated words. One weakness with their approach was inclusion of previously recalled items in the recognition data; items that are highly likely to be recognized. Given that recall of targets was mood dependent, subjects with matching moods may have had an easier time recognizing more previously recalled targets than subjects with mismatching moods. That Mood Dependent Memory 15 is to say, inclusion of previously recalled materials may have disproportionately benefited performance of matching mood subjects. Thus, the effect of mood on recognition performance requires re-examination after Beck and McBee's (1995) report. A different feature to investigate in mood dependence is the implicit or explicit nature of the memory task. Graf and Schacter (1985) coined the term implicit memory to describe situations where a prior experience enhances task performance without the necessity of subjects becoming consciously aware of that prior experience. An important aspect of the implicit/explicit distinction is the intention of subjects. In explicit tasks, like recall or recognition, subjects make purposeful and conscious efforts to remember particular prior episodes and may use various strategies and available cues to achieve this aim. In implicit tasks, subjects are asked to perform to the best of their ability but performance does not require active remembering of any specific experience. Rather, memory is inferred on implicit tasks by comparison of performance on previously exposed material to performance on new material. Enhanced performance for previously studied materials indicates memory for that exposure and is otherwise known as priming. A typical implicit memory task is word-identification. Subjects study a list of words by rating them on some attribute like pleasantness or whether they contain the letter "e". Later, subjects are presented with words on a computer screen. Presentation rate is selected so that subjects are able to read new words aloud only 30-40% of the time. The comparison of interest is between performance on new and previously rated words. Enhanced identification of previously viewed words reflects priming or implicit memory for those words although subjects may be unaware of the previous exposure to some of the targets. To date, virtually all research on mood dependence has been carried out with tasks such as free recall and recognition, in which subjects must consciously remember past events. The effect of emotion on implicit tasks is of particular interest in that it may provide clues to the mechanisms mediating mood dependent effects. One view holds that mood encourages elaboration of information in mood-relevant ways, thereby establishing associations between the Mood Dependent Memory 16 original learning episode and the mood. These associations allow a similar mood to become a salient cue during subsequent retrieval. Mood might thus be employed by subjects as one of a number of cues during conscious retrieval. If so, one would not expect mood to influence implicit memory, because elaboration and conscious retrieval strategies typically affect performance on explicit but not implicit tasks (Graf, Mandler & Hayde, 1982). A n alternate view is that mood exerts its influence because it is incorporated along with other relevant aspects of a target event into a unitary memory representation. The later experience of a similar mood acts as partial re-presentation of the prior event. Reprocessing of this partial information may enable redintegration of the entire memory representation and allow same mood events to come to mind more readily (Graf & Ryan, 1990; Horowitz & Prytulak, 1969; Mandler, 1988). By this view, performance on implicit tests of memory should be enhanced by the presence of similar moods. Another reason to investigate mood dependence in implicit memory revolves around demand characteristics. A recurring criticism is that mood effects could be explained by subject response to experimental demand. One published study investigated the production of "mood" effects under conditions of simulated mood (Perrig & Perrig, 1988). Subjects simulating moods duplicated M C M but not M D M effects. Nevertheless, subjects may infer that being in a different mood during memory testing "should" be associated with poor memory performance and act appropriately. This explanation is readily ruled out with employment of implicit memory tasks in which subjects are unaware that their memory is being tested. Implicit tasks are often designed to mask the fact that information previously encountered in the experiment could aid performance. They are typically presented as tests of general knowledge or skill in which subjects are instructed to use the first reasonable solution or to perform as quickly as possible. If subjects are not aware that they are performing memory tests, then response to experimental demand is not a reasonable explanation of mood dependent results. Thus, detection of mood dependence on implicit memory tasks would indicate that demand played little or no role in their production. Mood Dependent Memory 17 Initial investigations of M D M in implicit memory were focused on adaptations of the typical implicit memory task described above (see Macaulay et al., 1993). In each case, subjects, experiencing pleasant or unpleasant moods, were exposed to visual images, such as slides of common objects. In the next session, subjects, who were experiencing a matching or mismatching mood, were shown a larger number of stimuli which were presented in a way that initially rendered the items indistinguishable, such as by setting the slide projector out of focus. These obscuring influences were removed in equal gradual steps to the point at which subjects could correctly identify the stimulus. In each of three studies, subjects more quickly identified previously viewed objects than new ones, showing significant priming. However, in no case was this priming mood dependent. That is, having matching or mismatching moods made little difference to the speed with which subjects could perform the task. In the prior studies, experimenters provided more and more information until, finally, the correct answer was presented. These tasks encouraged subjects to focus on the external environment which eventually specified the correct response. The "do-it-yourself hypothesis suggests that tasks which encourage reliance on the subject's own internal resources and decision making processes during memory testing should be more susceptible to mood's influence than tasks which encourage reliance on materials provided by the experimenter. Implicit tasks of this type have been more promising but, as yet, have provided only mixed support for M D M . Recent investigations of word-stem completion and category production priming have each yielded one positive demonstration of the effect. For both tasks, the positive M D M outcome occurred when subjects had studied items (e.g., DESK) via autobiographical event generation during the experience of pleasant or unpleasant moods—another Mood Dependent Memory 18 generation during the experience of pleasant or unpleasant moods-another indication of study task importance. At test, subjects were provided with three-letter word-stems (e.g., DES-) which were to be solved with the first word that came to mind, or category cues (e.g., PIECES OF FURNITURE) which were to be solved by providing six category exemplars. Subjects produced studied items reliably more often than expected by chance, revealing significant priming, and this priming was greater when study and test moods matched (Figure 5). Although category production has been investigated in two subsequent studies, neither replicated the effect. It appears more important to consider specific processes underlying task performance and the overlap of such processes at study and test rather than the general implicit/explicit nature of the memory task. Visual tasks appear relatively insensitive to mood change. This may be contributed to by experimenter presentation, in contrast to subject generation, of the materials, and by the presentation of direct copies of visual materials at test. Implicit tasks which rely upon subjects generating memories at study and on subjects generating candidates at test have shown more promise. However, significant mood dependent priming remains to be replicated. 1.2.4 NATURE OF THE DIFFERENCES IN MOOD A fourth consideration in investigation of mood dependence is the nature of the mood at test compared to the mood at study. The majority of studies investigating mood dependence have examined happy versus sad moods. Therefore, investigations typically involve comparison of subjects reporting pleasant feelings and high levels of energy to subjects reporting feelings of displeasure and low levels of energy. Although this is typical of moods reported after mood induction, it is not always so. As mentioned earlier (pp. 10-11), the LIRIC technique is designed to require a particular intensity of pleasure or displeasure before the induction is considered successful, whereas arousal is free to vary. Although people typically report a corresponding rise in arousal with an increase in pleasure, or a corresponding fall in arousal with a decrease in pleasure, some people report little difference in arousal levels between pleasant and unpleasant moods. Mood Dependent Memory 19 Eich & Metcalfe (1989) and Eich et al. (1994) reported 7 studies, summarized in Table 1, in which subjects generated materials which would later serve as targets in surprise tests of free recall. Subjects were divided into 3 groups—one in which subjects experienced similar moods in both sessions, one in which subjects assigned to mismatching moods reported large differences in pleasure but little difference in arousal, and one in which mismatching mood subjects reported large differences on both pleasure and arousal. Planned comparisons revealed that in each study, subjects experiencing differences on both pleasure and arousal remembered significantly fewer items than matched mood subjects. Subjects who experienced a large change in pleasure but little change in arousal remembered an amount that fell at or between these extremes in every case. These studies support the proposal that experience of moods differing on both pleasure and arousal produces a larger memory disruption than does the experience of moods which differ on pleasure alone. One interpretation of this trend is that change in both pleasure and arousal reflects a greater psychological distance between the two moods, and as predicted by Bower (1981), a greater tendency to disrupt memory performance. 1.3 SUMMARY The nature of the manipulated mood, the nature of the study task, the nature of the memory test, and the nature of the differences between moods all contribute to the circumstances which will and will not demonstrate mood dependent effects. Manipulated moods should be I Table 1 Recall of Generated Items as a Function of Pleasure/Arousal Change Pleasure/Arousal Change % Recall Eich & Metcalfe, (1989) Category Generation Gen x 1 Gen x 3 Small/Small 21 36 Large/Small 16 33 Large/Large 9 26 Eich et al. (1994) Autobiographical Generation Small/Small 38 Large/Small 31 Large/Large 29 Mood Dependent Memory 20 intense, durable, genuine and experienced by each subject to improve the probability of demonstrating mood effects. The LIRIC induction technique readily ensures each of these concerns. However, the LIRIC induction has until now been available for only happy and sad moods. Mood dependent effects have been demonstrated consistently when event generation has been employed as the study task. Mood dependence also has been consistently demonstrated when memory testing consists of free recall. However, given new reports of mood dependence in recognition, controversy remains. Although there is good reason to investigate the production of mood dependence in implicit tasks, the results thus far are equivocal. Priming has been consistently demonstrated across many different implicit tasks, but M D M is most promising in tasks which rely upon subject generation of responses at test. The effect has been demonstrated, but remains to be replicated, when subjects study materials via event generation and are tested for memory of the materials via implicit category production. Finally, the nature of the differences between the induced moods may prove to be an important consideration. Specifically, correlational data show that subjects who report a difference in both pleasure and arousal between study and test tend to show larger memory disruption than do subjects who report a difference on pleasure alone. Mood Dependent Memory 21 C H A P T E R 2. EXPERIMENT 1: INVESTIGATION OF MOODS THAT VARY ON PLEASURE AND AROUSAL 2.1 ISSUES TO BE RESOLVED An issue remaining to be resolved is the contribution of pleasure and arousal change to production of mood dependence. Although correlational data indicate that subjects who experience change in both pleasure and arousal between study and test tend to show greater memory disruption than subjects who experience change in pleasure alone, this needs to be borne out by independent manipulation of pleasure and arousal. The previously reported effects may be confounded with success of the mood manipulation. That is, subjects reporting pleasant and high arousal moods in one session crossed with unpleasant and low arousal moods in another may simply have had a more realistic mood experience than were subjects whose moods differed on pleasure but not arousal. Unfortunately, no data exist regarding genuineness of the moods in these prior studies. Factor analytic studies of self-reported moods consistently reveal two underlying factors. The interpretation that provides the basis of the Affect Grid is that mood varies along the dimensions of pleasure and arousal (Russell, 1980; Russell et al., 1989). A different interpretation of the factors is that pleasant and high arousal states and unpleasant and low arousal states fall along a single dimension (Watson & Tellegen, 1985). By this view, happy and sad moods lie at the extremes of the positive affect (PA) dimension and represent high and low levels of P A respectively. The second dimension is labeled negative affect (NA) and ranges from unpleasant and agitated (high NA) to pleasant and calm (low NA) states. These two Figure 6. Relationship between Affect Grid and PANAS Scale High NA High Arousal High PA Low Pleasure \ / . High Pleasure \ \ \ / / Low PA Low Arousal LowNA Mood Dependent Memory 22 interpretations of emotional experience consider different combinations of moods to vary on both fundamental dimensions of emotion (see Figure 6). Whereas Russell's interpretation would describe happy-sad and anxious-calm mood combinations as differing on both fundamental dimensions, in Watson and Tellegen's interpretation, happy-calm and anxious-sad combinations would differ on both dimensions. As such, these views can be taken further to make different predictions about which moods possess greater psychological distinctiveness and, according to Bower's (1981) proposal, about which study-test mood combinations would produce the greatest M D M disruption. Other theorists believe there are evolutionarily basic emotions that form the foundation for all emotional experience. According to Johnson-Laird and Oatley (1988), the 5 basic emotions of fear, joy, sadness, anger and disgust all serve to prepare a person for certain types of interaction and to ready different prototypical action patterns. Where happiness is related to the establishment and maintenance of attachments, sadness is related to the loss of such attachments. Fear prepares us for escape, anger prepares us to fight, and disgust prepares us to reject someone or something. These prototypical action patterns may be associated with different underlying attention or memory processes (Macaulay et al., 1993). So, while fear may encourage us to intently monitor our environment for threat (Mathews & Eysenck, 1987), sadness may encourage us to focus inward as we reflect upon our loss (Pyszczynski, Hamilton, Herring & Greenberg, 1989). Different basic cognitive capacities may be altered by the experience of these moods. By this view, joy, sadness, and anxiety all represent different fundamental emotions and may thus impact differently on basic cognitive capacities. However, delighted and calm moods are both based on the fundamental mood joy, and therefore would be considered equivalent. Working from the interpretation of emotion supported by Russell et al. (1989), one would predict mood dependence to be most likely in this study between delighted-sad and anxious-calm moods as they differ on both basic dimensions of pleasure and arousal. If one holds Watson and Tellegen's (1985) view of emotion, mood dependence would be predicted for moods differing on P A and N A . Thus the delighted-calm and anxious-sad mood combinations should show the Mood Dependent Memory 23 largest memory effects. Finally, if one holds the view supported by Johnson-Laird and Oatley (1988), one would expect mood dependence on all but the delighted-calm mood combination. Direct comparison of a variety of moods in this study is required to clarify the importance of mood distinctiveness in production of mood dependence. Further, this investigation should prove informative about the nature of mood itself. The information will be valuable given that it derives from objective analysis of memory performance in individuals reporting particular moods, an avenue that, although previously suggested, has not been previously pursued. Another issue to be resolved is the apparent contradiction in the current literature regarding M D M in recognition. While Eich and Metcalfe (1989) did not, Beck and McBee (1995) did report significant dependent effects in recognition. The principle distinction between studies may be selection of different scoring methods, which could account for the apparent contradiction in results. Whereas inclusion of previously recalled items may suggest dependence (a la Beck & McBee, 1995), their removal should eliminate the effects (a la Eich & Metcalfe, 1989), demonstrating the importance of scoring method and replicating both previous investigations of mood dependent recognition. 2.2 M E T H O D 2.2.1 Mood Induction Development Procedures and music based on the standard L I R I C mood manipulation were developed to induce four moods that vary on pleasure and arousal. The induced moods were labeled anxious (unpleasant and high arousal), delighted (pleasant and high arousal), calm (pleasant and low arousal) and sad (unpleasant and low arousal). Musical selections were chosen not only on the valence of the emotional response they evoked, but also on the "energy" they possessed. "Delighted" music evokes pleasant feelings but is also associated with high levels of energy, portrayed by rhythm or tempo. Similarly, "calm" pieces evoke pleasant feelings and are associated with low levels of energy, conveyed by slow tempo or certain instruments such as the cello or harp. These same principles were applied in the selection of "anxious" and "sad" music. Mood Dependent Memory 24 Music to induce each mood was recorded onto four tapes (for complete listings, see Appendix I). Each side of the tapes contained different selections of music appropriate for a given mood to ensure that subjects assigned to experience the same mood on two occasions would not hear the same music. Four sets of instructions were used in conjunction with the music to aid in modification of both pleasure and arousal. Specifically, subjects were given examples of scenarios that might create the desired mood. For example, subjects assigned to the delighted mood were to select events, like falling in love or playing volleyball with friends, which could inspire in them pleasant and excited feelings, while subjects in the calm mood were asked to ponder pleasant scenarios like floating on a raft in the sun or sleepily lounging in bed on a Saturday morning. 2.2.2 Design The experiment consisted of two sessions (Study and Test) which were separated by two days. Study mood (anxious, delighted, calm and sad) was crossed with Test mood (anxious, delighted, calm and sad) conforming to a 4 x 4 design and a total of 16 Study/Test mood conditions. 2.2.3 Study Session At the start of the first session, subjects were shown a copy of the Affect Grid. Experimenters explained the Grid in detail (pp. 9-10), and requested that subjects mark the square which best exemplified current levels of pleasure and arousal. After making their baseline mark, subjects were informed that mood was also to be measured periodically with the PANAS scale (Watson, Clark, & Tellegen, 1988). The scale comprises 20 adjectives, 10 of which relate to positive affect and 10 of which relate to negative affect. To complete this scale, experimenters read each adjective aloud and subjects respond with a number from 1 (not at all or slightly) to 5 (extremely) which indicates the extent to which each adjective describes their current feelings. This scale serves as a check of the mood manipulation. Mood Dependent Memory 25 After completing their baseline PANAS scale, subjects listened to classical music in order to achieve a particular mood state, either anxiety, delight, calmness or sadness. Experimenters left the room but returned periodically to measure subjects progress, and when experimenters thought the time was right, they continued with the rest of the study. Determination of readiness relied upon both pleasure and arousal scores as shown in Figure 7. Subjects were required to endorse moderate or more intense pleasure or displeasure simultaneous with endorsement of moderate or more intense high or low arousal depending upon the assigned mood. For example, subjects assigned to anxious moods were required to endorse both moderate, very, or extreme displeasure and moderate, very, or extremely high arousal. Regardless of pleasure or arousal ratings, all subjects contemplated their selected thoughts in conjunction with the music for a minimum of 10 minutes. If subjects did not report moods in the required quadrant in 40 minutes, they were fully debriefed, paid for their time and dismissed from the experiment. On reaching the requisite levels of pleasure and arousal, subjects were read one of two lists of 12 common unrelated words, such as CITY and KEY. Subjects generated a personal event for each word. Al l 24 probes, plus 4 others used as examples, are common concrete neutral nouns culled from Brown and Ure's (1969) word norms. Subjects were instructed to say OK as soon as they generated an event specific enough that they could describe the event and give an approximate date of occurrence. If a subject failed to generate such an event within 2 min, that probe was skipped and the next one was read. If an Figure 7. Quadrants on the Affect Grid. Mood Dependent Memory 26 event was generated, the experimenter logged the generation latency and asked the subject to continue with the description. Experimenters transcribed where and when the event occurred, what happened and to whom. Next, subjects rated (a) the original emotionality of the event on a scale ranging from +4 (extremely positive) through 0 (neutral) to -4 (extremely negative) (b) the importance of the event at the time it occurred on a scale ranging from 1 (not at all important) to 5 (extremely important) and (c) the vividness of their recollection of the event on a scale ranging from 1 (vague) to 5 (extremely vivid). After generating, recounting and rating 12 events, subjects completed another Affect Grid and PANAS scale. Subjects were asked to rate the genuineness of the mood they experienced, with their responses ranging from 0 (extremely artificial) to 5 (moderately genuine) to 10 (extremely genuine). Subjects were reminded to return in 2 days to complete the study. Subjects in pleasant moods were immediately released whereas subjects who experienced unpleasant moods spent some time chatting with the experimenter and eating cookies until they felt ready to leave. 2.2.4 Test Session Procedures involved in manipulating and measuring mood during the Test session duplicated those at Study. On reporting the requisite ratings for pleasure and arousal, subjects were reminded that 2 days earlier they generated personal events in response to certain probe words. Subjects then spent 5 min trying to recall aloud the probe words they had used for event generation, or, failing that, to recount memories they had previously described. Next, subjects were read aloud all 24 probe words, 12 of which they had used to generate personal events and 12 of which were new. Subjects performed old/new recognition for these words. At the end of the recognition task, subjects again rated current mood on the Affect Grid and PANAS scales. Mood Dependent Memory 27 After being informed that the study was over, subjects rated the genuineness of the mood experienced in the Test session. Subjects were then completely debriefed both orally and in writing and received money for their participation. 2.3 SUBJECTS Participants were recruited from UBC students 17 to 24 years of age. Upon arrival, participants completed the Beck Depression Inventory (Beck, 1967) as a screening measure of depression. Subjects were excluded from the study if their scores fell above 15 or above 10 if that score included any positive response to the question about suicidal ideation (question #9). Eligible subjects were asked to use the music and their selected thoughts with the aim of developing particular moods. Study design called for a total of 160 subjects (16 cells at 10 subjects per cell). To achieve this number, testing was started on a total of 248 subjects, 88 of whom tried but failed to meet the mood criterion. Ten other subjects were dropped from the study when they did not return for their second session (3 women and 3 men) or when the experimenter made an error during testing (2 women and 2 men). The overall success rate (160/248 or 65%) was comparable for women (118/183 or 64%) and men (42/65 or 65%), but was noticeably lower than the rate reported in the introduction. The lower success rate is understandable given that the current study required subjects to report particular levels of both pleasure and arousal whereas previously pleasure alone determined success. Further, the rate is comparable to Clark's (1983) estimation of the success rate for the Velten technique (50-70%). Examination of subjects for whom mood manipulation failed indicates that arousal was the most frequent cause of mood failure. Specifically, 72% of mood manipulation failures were considered failures because, although pleasure ratings were at or above the criterion, arousal ratings did not attain critical levels in 40 min. By comparison, only 12% of mood failures were due to pleasure ratings alone. Neither set of ratings reached required levels in 7% and in 9% of mood failures both pleasure and arousal ratings reached critical intensity but these levels were not Mood Dependent Memory 28 attained simultaneously. Pleasure appears more amenable than arousal to the current manipulation. 2.4 R E S U L T S 2.4.1 RESPONSE T O M O O D MANIPULATIONS Subject responses to the mood manipulations are evaluated via the Affect Grid, PANAS, time to achieve moods, and genuineness ratings. Affect Grid Ratings Affect Grid ratings were obtained prior to mood induction (BSL), every 5 minutes thereafter until subjects reported both pleasure and arousal levels above the mood criteria (CRT), and again upon completion of the cognitive task(s) in each session (PST). Marks made on the Affect Grid were translated into 2 scores that ranged from -4 to +4 reflecting subjective pleasure and arousal. Pleasure Ratings Pleasure ratings obtained throughout Study sessions are reported in Table 2. Analysis of variance (ANOVA) of pleasure ratings obtained at baseline do not differ significantly across groups (F < 1). Pleasure ratings taken once subjects achieved the mood criteria reflect an effect of Study Mood (F(3, 156) = 772.0, p < .001). As required by the pleasure criterion (F(l, 156) = 2340.0, p < .001), subjects in delighted and calm groups (pleasant moods, or P) reported pleasure ratings that differed from ratings reported by subjects in anxious and sad groups (unpleasant moods, or U). This planned comparison was explored with Tukey post-hoc tests that revealed no differences in pleasure ratings within the P or U mood groups. Average intensity of pleasure or displeasure Table 2. Study Session Pleasure Ratings as a Function of Induced Mood. Induced Mood n Delighted 40 Calm 40 Anxious 40 Sad 40 Rating Occasion BSL CRT PST 1.3 1.4 1.1 1.1 2.9 2.4 2.9 2.5 -2.7 -0.4 -2.6 -0.8 Mood Dependent Memory 29 achieved when subjects met the mood criteria also appeared comparable across moods (P mood |2.9| vs. U mood |-2.6| pleasure ratings). That is, all four moods were initially induced to similarly intense pleasure or displeasure. After completion of autobiographical event generation, differences in rated pleasure had diminished to some extent (F(3, 156) = 65.2, p_ < .001) but nevertheless, a planned comparison indicates that subjects in P moods reported pleasure levels that remained significantly different from those reported by subjects in U moods (F(l, 156) = 194.3, p < .001). While Tukey tests revealed no differences within P or U mood groups, pleasure ratings remained more intense in P than in U mood groups post-task (means of |2.4| and |—0.6] respectively). This result is not surprising as subjects may be less motivated to maintain unpleasant feelings. Granted that moods are less intense after task completion, visual inspection of group baseline and post-task averages indicates, nevertheless, that ratings reported post-task differ from pre-induction pleasure ratings (difference between average BSL and PST in P and U groups =1.1 and -1.7, respectively). Test Session pleasure ratings (Table 3) demonstrate precisely the same pattern as the Study Session. Pleasure ratings gathered once subjects achieved the mood criteria displayed a Test Mood effect (F(3, 156) = 802.3, p_ < .001) supported by pleasure ratings which differed between P mood and U mood subjects (F(l, 156) = 2690.3, g < .001), but not within P or U mood groups. Ratings taken upon completing recall and recognition testing revealed diminished (F(3, 156) = 163.3, p_ < .001) but nevertheless convincing differences between the ratings of P and U mood subjects (F(l, 156) = 136.0, p < .001). Tukey tests revealed that post-task pleasure ratings did not differ within P mood groups but that anxious subjects reported less intense displeasure post-task than sad subjects (F(l, 156) = 7.28, p < .05). Average intensity of pleasure post-task appeared greater for P than U mood groups (means of |2.3| and |-1.5|, respectively). However, comparison of baseline and post-task ratings Table 3. Test Session Pleasure Ratings as a Function of Induced Mood. Induced Mood n Delighted 40 Calm 40 Anxious 40 Sad 40 Rating Occasion BSL CRT PST 1.2 2.6 2.2 1.3 2.8 2.4 1.4 -2.6 -1.3 1.4 -2.6 -1.7 Mood Dependent Memory 30 indicate that, after memory testing, subjects continue to report pleasure levels that differ from those reported prior to mood induction (difference between average BSL and PST in P and U groups =1.1 and -2.9, respectively). Unpleasant mood groups' pleasure ratings appeared more intense post-task in the Test session than in the Study session. This result may be due to time spent on the cognitive task. The Study task required 12-25 min where memory tests required 7 min in total (5 min for free recall, and 1-2 min for recognition). Another reason for greater mood dissipation in Study sessions is that event generation involves active participation by subjects in describing and rating their personal memories. The nature of the task may have distracted subjects from maintaining their mood state. Arousal Ratings Study session arousal ratings are reported in Table 4. As expected, baseline ratings of arousal do not differ significantly across moods (F < 1). Upon achieving the mood criteria, arousal ratings reflected an effect of Study Mood (F(3, 156) = 760.6, p < .001). Ratings recorded by subjects in the delighted and anxious groups (high arousal moods, or HI) differed significantly from ratings of subjects in the calm and sad groups (low arousal moods, or LO), as required by the arousal criterion (F(l, 156) = 2194.4, g < .001, means of 2.5 and -2.5 respectively). These means also make obvious that the initial intensity of arousal is comparable across HI and LO groups. Tukey tests indicated that arousal ratings of LO groups did not differ significantly, but revealed a difference between HI mood groups (F(l, 156) = 7.28, p < .05). The anxious mood group reported higher arousal than did the delighted mood group (means of 2.7 and 2.3 Table 4. Study Session Arousal Ratings as a Function of Induced Mood. Induced Rating I Occasion Mood n BSL CRT PST Delighted 40 0.0 2.3 1.9 Calm 40 -0.1 -2.5 -0.6 Anxious 40 0.1 2.7 1.3 Sad 40 0.1 -2.6 -0.6 respectively). Mood Dependent Memory 31 Post-event generation arousal ratings had moved toward neutral in all four groups. Nevertheless, differences between groups remained highly reliable (F(3, 156) = 27.2, p < .001). Specifically, HI mood group subjects reported significantly higher levels of arousal than did L O mood group subjects post-task (F(l, 156) = 78.4, p_ < .001) but subjects did not differ within HI or LO mood groups. L O mood groups appeared to maintain less intense post-task arousal than did HI mood groups (means of |-0.6| and |1.6| respectively). Visual inspection of differences between average baseline and post-task ratings indicates that subjects typically maintained a large distance from baseline ratings in HI moods, while the difference was more moderate among LO moods (difference between average BSL and PST in HI and LO =1.5 and -0.6, respectively). Perhaps the effort involved in event generation renders maintenance of low levels of arousal difficult. Though the greater drop in arousal intensity and the smaller distance from baseline in the L O mood groups is somewhat disappointing, the difference in arousal between HI and LO groups does remain highly significant. Test session arousal ratings are reported in Table 5. Analyses revealed a Test session pattern matching that present in the Study session. Upon achieving the mood criteria, arousal Table 5. ratings showed an effect of Test Mood (F(3, 156) = Test Session Arousal Ratings as a 832.6, p < .001) which was supported by arousal ratings Function of Induced Mood. of HI mood groups differing significantly from L O mood Induced Rating Occasion groups (F ( l , 156) = 2512.6, p < .001). Tukey tests Mood n BSL CRT PST indicated no differences within groups. Ratings gathered Delighted 40 1.2 2.5 2.4 after completion of free recall and recognition revealed Calm 40 Anxious 40 0.9 -2.4 0.9 2.5 -0.8 2.2 that differences in rated arousal had diminished Sad 40 1.1 -2.6 -1.4 somewhat (F(3, 156) = 113.5, p < .001), but that the HI mood group ratings remained significantly different from Mood Dependent Memory 32 those of the LO mood group (F(l, 156) = 57.2, g < .001). Further evaluation with Tukey tests revealed no differences within HI mood groups, but indicated that sad subjects reported lower levels of arousal than calm subjects (F(l, 156) = 6.66, g < .05). At Test, as at Study, post-task means indicate better maintenance of arousal in HI than in LO mood groups (means of |2.3| and |-1.1| respectively). However, unlike the Study session, inspection of differences between average baseline and post-task arousal shows that the L O mood groups maintained a greater distance from their baseline arousal levels as did their HI mood counterparts (difference between average BSL and PST in LO and HI = -2.1 and 1.2, respectively). Similar to the pattern for pleasure ratings, the arousal ratings appear more stable in Test than Study sessions. Time on task may also be influencing arousal ratings and seems the most parsimonious explanation for the greater persistence of both pleasure and arousal components at Test. PAN AS Ratings Although the mood manipulations require subjects to endorse certain quadrants on the Affect Grid, there are no such requirements on the PANAS scale. As such, the PANAS serves as a second mood measure which, if the mood manipulations are effective, will demonstrate high P A scores for subjects in the delighted mood condition, low P A scores for sad subjects, high N A for anxious and low N A for calm subjects. Table 6. Positive Affect Ratings Study Session Positive Affect Ratings as a Function of Induced Mood. Table 6 presents P A scores across induced moods in the Study Session. Baseline ratings of P A did Induced Mood n Rating Occasion BSL CRT PST not differ across moods (F < 1). Upon registering pleasure and arousal ratings in the assigned quadrant, P A ratings demonstrated a strong effect of Study Mood (F(3 Delighted 39/40 30.0 34.2 32.5 Calm 40 29.9 24.6 26.3 Anxious 40 28.2 27.0 25.7 Sad 40 29.2 18.6 21.3 , 155) = 40.8, p < .001). Planned comparisons revealed Mood Dependent Memory 33 Table 7. Test Session Positive Affect Ratings as a Function of Induced Mood. that P A ratings of delighted mood subjects were higher than average P A ratings of subjects in anxious and calm moods, and P A ratings of the sad mood group were lower than the average P A ratings of the anxious and calm groups. Analyses of ratings obtained once subjects completed event generation indicated that although the effect had diminished somewhat, (F(3, 156) = 16.5, p < .001), group means demonstrated the same pattern as those obtained prior to the task. Test session scores of P A are presented in Table 7. Ratings gathered once subjects reported Affect Grid scores above critical levels revealed a strong effect of Test Mood (F(3, 156) = 35.6, p < .001). Planned comparisons indicated that the delighted mood group reported P A ratings higher than, and the sad mood group reported P A ratings lower than, the average of the anxious and calm mood groups. Ratings obtained after free recall and recognition tests showed that the effect of the Test Mood had diminished slightly (F(3, 15) = 30.8, g < .001) but remained significant. The pattern of means was similar to the pattern reported for the Study session. Induced Mood n Delighted 40 Calm 40 Anxious 40 Sad 40 Rating Occasion BSL CRT PST 27.9 31.7 31.2 27.3 20.9 23.2 28.4 25.6 25.9 27.3 16.8 18.2 Negative Affect Ratings Ratings of N A across induced moods in the Study Session are presented in Table 8. Once subjects reported the requisite levels of pleasure and arousal, their N A ratings showed a strong effect of Study Mood (F(3, 155) = 87.5, p < .001). Visual inspection makes obvious that the most striking differences are between P and U mood groups. Yet, planned comparisons revealed that anxious mood subjects reported N A ratings higher than, and calm Table 8. Study Session Negative Affect Ratings as a Function of Induced Mood. Induced Mood n Rating Occasion BSL CRT PST Delighted 39/40 14.6 11.5 11.7 Calm 40 14.2 11.3 11.2 Anxious 40 13.5 26.7 17.1 Sad 40 14.7 20.9 15.4 Mood Dependent Memory 34 Test Session Negative Affect Ratings as a Function of Induced Mood. Induced Mood n Rating Occasion BSL CRT PST Delighted 40 Calm 40 Anxious 40 Sad 40 12.5 11.1 11.7 12.8 10.6 10.8 13.0 22.4 18.2 12.4 18.4 16.4 mood subjects reported N A ratings lower than, the | Table 9. average N A ratings of delighted and sad subjects. Post-task, N A ratings showed the same pattern and a somewhat attenuated but nevertheless reliable effect of Study mood (F(3, 156) = 16.1, p < .001). Ratings of N A across induced Test moods are presented in Table 9. Ratings of N A gathered once subjects reported appropriate Affect Grid ratings (F (3, 156) = 76.7, p < .001) and after recall and recognition testing (F(3, 156) = 31.2, p < .001), reflected Test Mood effects, and a pattern of means pattern similar to that found in the Study Session. Clearly, though P A and N A ratings were both influenced by the mood manipulations, they were not influenced independently. Time to Induce Moods Average time required to induce each mood in each session is shown in Table 10. Analysis of Study Mood, Test Mood and Session revealed no significant effects in time to develop critical pleasure and arousal ratings (all Fs < 1.95, g >. 10). Average induction times range from 13.4 to 17.4 minutes across moods and suggest that the manipulations require comparable lengths of time to be achieved. Genuineness Ratings Upon completion of the tasks in either session, subjects were provided with an 11 point scale on which to evaluate the genuineness of the experimentally induced moods they experienced. Genuineness ratings are reported as a function of Induced mood in Table 11. A N O V A was calculated to consider Study mood, Test mood and Session as factors in Table 10. Time to Reach Mood Criteria Induced Mood nl/n2 Session Study Test Delighted Calm Anxious Sad 40/40 40/40 40/40 40/40 17.4 13.4 17.2 15.8 17.0 17.0 16.6 17.0 Average Time 17.1 15.8 Mood Dependent Memory 35 genuineness ratings. A Session effect indicated that Session 1 scores were lower than Session 2 scores (F(l, 143) = 22.7, p < .001, means of 6.9 and 7.6 respectively). The scores appear slightly lower than those obtained in prior experiments. This may be accounted for by the lower intensity of pleasure or displeasure required by the criterion ('moderate' as opposed to 'very' intense in past studies with genuineness ratings). The main effect was qualified by a significant interaction between Study mood and Session (F(3, 143) = 4.08, p < .01) and a marginal interaction between Test mood and Session (F(3, 143) = 2.09, p < . 10). Analyses indicate that pleasant moods led to somewhat higher genuineness ratings than did unpleasant moods (Session 1 genuineness 7.3 vs 6.6, and Session 2 genuineness 7.8 vs 7.4). Higher genuineness ratings for pleasant and for Test session moods may be contributed to by more successful maintenance of pleasant moods generally and to the shorter tasks performed under Test session moods. The more that the appropriate mood is maintained until task completion, the more evidence subjects will have suggesting that mood induction was successful. Summary Affect Grid ratings indicated that mood manipulations created differences between pleasant and unpleasant moods and between high and low arousal moods that were sustained until the end of the tasks in either session. These differences were extremely large (as ensured by the criteria), and although they were smaller post-task, they remained highly reliable. Examination of the absolute intensity of pleasure and arousal ratings at each rating occasion confirmed that moods were induced to equal intensity initially but ratings of unpleasant and low arousal groups showed stronger tendencies to return to neutral. However, in both sessions, U mood groups Table 11. Genuineness Ratings as a Function of Induced Mood and Session. Session Induced Mood nl/n2 Study Test Delighted 40/40 6.9 7.9 Calm 40/40 7.8 7.7 Anxious 40/39 6.4 7.2 Sad 40/40 6.7 7.5 6.9 7.6 Mood Dependent Memory 36 reported post-task ratings that appeared to differ from baseline pleasure at least as much as, if not more than, those ratings for P mood groups. Interestingly, in arousal, these differences between post-task and baseline scores, appeared greater in HI mood groups at Study, but greater in LO mood groups at Test. The nature of the activities required by the tasks in each session (active description and ratings at Study versus quiet memory search at Test) may have contributed to this effect. Although the mood criteria meant that subjects endorsed certain quadrants on the Affect Grid, there were no such requirements on the PANAS scale. Delighted subjects demonstrated high P A scores while sad mood subjects demonstrated low P A scores. Although in N A ratings, differences between P and U groups were most striking, anxious and calm mood subjects did report high and low N A scores, respectively. Further support for the effectiveness of the mood manipulation is provided by genuineness ratings. The large majority of subjects rated their mood experience as more than moderately genuine, with only small differences between ratings of genuineness for P and U moods. 2.4.2 AUTOBIOGRAPHICAL EVENT GENERATION To examine the Event Generation task for M C M effects, events were divided into positive, neutral, and negative groups, and are reported as a function of Study session mood in Table 12. Analyses revealed a main effect of Event Type (F(2, 312) = 179.4, g < .001) such that a greater proportion of generated events were rated as positive than negative and more were rated as negative than neutral (means of 57.2, 29.5 and 13.2 respectively). A significant Study Mood by Event Type interaction indicated mood congruence (F(6, 312) = 15.8, p < .001). Specifically, subjects in the P Table 12. Percent of Event Types Generated as a Function of Induced Mood Induced Event Gen Mood Type % Delighted pos 70.3 [40] neut 11.7 neg 17.9 Calm pos 66.4 [40] neut 12.9 neg 20.6 Anxious pos 40.2 [40] neut 15.4 neg 44.3 Sad pos 51.8 [40] neut 12.7 neg 35.4 Mood Dependent Memory 37 mood groups rated more of their memories as positive and fewer as negative than did their U mood counterparts (P mood subjects pos vs. neg = 68.4 and 19.2; U mood subjects pos vs. neg = 46.0 and 39.8). Tukey post-hoc tests revealed no differences within P or U mood groups. Subject ratings of generated events were collected, but to maintain a focus on the principle concern of this study which is M D M performance, interested readers are directed to Appendix II which contains detailed M C M analysis of these ratings. Overall the analyses are consistent with those reported in Eich et al (1994) and suggest that the mood manipulations produced effects that were sufficiently potent to influence the way that subjects performed the Event Generation task. 2.4.3 FREE RECALL PERFORMANCE The percentage of words remembered as a function of induced Study and Test moods is presented in Table 13 a. An overall A N O V A revealed only a significant effect of Study Mood (F(3, 144) = 3.26 g < .05). Tukey contrasts indicate that subjects who performed event generation under sad moods (60.2) recalled more words later (regardless of induced mood at Test) than did their anxious counterparts (50.8). The recall of neither group differed significantly from subjects assigned to delighted (55.3), or calm (52.5) Study moods. The primary purpose of this investigation was evaluation of one- versus two-dimensional mood change on free recall. Table 13b presents the free recall data translated Table 13 a. Percent of Targets Free Recalled as a Function of Induced Mood at Study and Test. Test Mood Study Delighted Calm Anxious Sad Mood (SD) (SD) (SD) (SD) Delighted 58.3 55.8 51.7 55.5 (14.7) (13.1) (7.7) (16.0) Calm 53.3 56.7 50.8 49.2 (14.8) (13.5) (13.9) (10.7) Anxious 57.5 45.0 55.6 45.0 (17.3) (16.8) (10.0) (14.8) Sad 57.5 60.0 53.3 70.0 (12.7) (15.6) (19.3) (15.8) Table 13b. Free Recall as a Function of Pleasure/Arousal Change Change in Pleasure/Arousaln % Small/Small 40 60.1 Large/Small 40 54.6 Small/Large 40 51.9 Large/Large 40 52.2 Mood Dependent Memory 38 into divisions comparable to those from Eich & Metcalfe (1989) and Eich et al., (1994). A N O V A revealed a significant effect of assigned pleasure and arousal change on free recall (F(3, 156) = 3.73, g < .05). Planned comparisons revealed a significant difference between subjects who experienced matching moods and subjects who experienced change on both pleasure and arousal Table 13 c. (F(l, 156) = 5.82, p < .02), but only a trend between matching Free Recall as a Function of mood subjects and those assigned to experience change on Induced Mood at Study and Test pleasure alone (F(l, 156) = 2.87, g < .09), replicating the prior Study/Test Mood % pattern. However, moods that varied on arousal alone produced as much memory disruption as did moods that varied Arousal Change: on both pleasure and arousal (F(l, 156) = 6.34, p < .02). Delighted/Delighted 58.3 Delighted/Calm 55.8 As previously planned, A N O V A was performed on 2 x 2 Calm/Delighted 53.3 Calm/Calm 56.7 matrices of Study by Test moods in order to clarify how each Anxious/Anxious 55.6 mood interacts with each other mood. Table 13c recasts the Anxious/Sad 45.0 Sad/Anxious 53.3 free recall data in summaries for each A N O V A . Moods in the Sad/Sad 70.0 resulting six matrices vary on arousal, pleasure or both Pleasure Change: dimensions. Each A N O V A was analyzed for three effects, Delighted/Delighted 58.3 Delighted/Anxious 51.7 Study mood, Test mood and their interaction. This approach Anxious/Delighted 57.5 Anxious/Anxious 55.6 allows evaluation of the strength of the mood dependent effect Calm/Calm 56.7 (indicated by a significant interaction) for each pair of moods. Calm/Sad 49.2 Sad/Calm 60.0 Considering first the mood combinations that vary on Sad/Sad 70.0 arousal, means in the delighted and calm mood combination fell Pleasure and Arousal Change: in an M D M pattern, but analysis revealed neither significant Delighted/Delighted 58.3 Delighted/Sad 55.5 main effects nor interaction (all Fs < 1). The anxious and sad Sad/Delighted 57.5 Sad/Sad 70.0 mood combination was analyzed to reveal a significant Study Calm/Calm 56.7 mood effect (F (1, 36) = 5.48, g < .025) such that event Calm/Anxious 50.8 Anxious/Calm 45.0 generation gerformed while in a sad mood led to greater Anxious/Anxious 55.6 recollection than did an anxious mood (means of 61.6 and 50.3). Mood Dependent Memory 39 The main effect was qualified by a significant interaction (F(l, 36) = 7.89, g < .01). Subjects recalled more in matching than mismatching moods, indicating mood dependence. Turning to moods that vary on pleasure, analysis of the delighted and anxious mood combination revealed neither significant main effects nor interaction (all Fs < 1.1), and was not mood dependent. The calm and sad mood combination showed a main effect of Study mood (F(l,36) = 7.38,g<.01) such that subjects who generated events while sad remembered more than subjects who did so in a calm mood (means of 65.0 and 53.0). The main effect was qualified by a trend indicating a mood dependent interaction between Study and Test moods (F(l , 36) = 3.87, g< .06). The calm and anxious mood combination represents change along both pleasure and arousal. Its analysis revealed no main effects (both Fs < 1) but a trend indicating mood dependence (F(l,36) = 3.58, g < .07). The delighted and sad mood combination likewise represents change along both gleasure and arousal and it corresgonds most closely to the tygical investigation of mood degendent memory via "happy" and "sad" moods. Analyses reveal no significant main effects (both Fs < 2.1, g > . 15) but a weak trend toward interaction between Study and Test moods (F(l, 36) = 2.65, g < . 11) indicating mood degendence. Although trends are tygically mentioned at a=.10, in this study overall patterns are of interest and there is a clear division between the F values of the delighted-sad combination (interaction F = 2.65) and those of the delighted-anxious and delighted-calm combinations that show greater variance within- than between-groups (both interaction Fs < 1). The impact of matched or mismatched reported moods—rather than assigned mood condition—on recall performance, is worth investigation. Calculation of a correlation for that purpose required a single value which could reflect the intensity of mood maintained in either session as well as differences or similarities in mood across sessions. Mood intensity was reflected by averaging pre- and post-task pleasure and arousal ratings for each session. The absolute differences between these session averages (presented in Table 14) were summed to Mood Dependent Memory 40 provide an overall estimate of subject mood. This Table 14. indicator was marginally negatively related to recall Absolute Differences between Study and Test Pleasure and Arousal as a Function performance (r(158) = -0.15, p < .06). of Induced Moods across Sessions. Comparison of mood change scores in Abs Abs delighted-calm and delighted-anxious combinations (the Induced Mood Study/Test Diff Pleas Diff Arous two combinations that did not yield significant M D M ) to those of other combinations is particularly Arousal Change: interesting. If mismatched mood subjects were to Delighted/Delighted Delighted/Calm 0.45 0.60 0.55 3.25 report very small differences in mood, M D M would be Calm/Delighted Calm/Calm 0.75 0.60 4.20 0.55 attenuated, which may explain why the effects do not Anxious/Anxious 0.75 0.70 exist for these combinations. Referring to Table 14, the Anxious/Sad Sad/Anxious 0.85 0.70 3.45 4.40 mood change estimates across sessions in the delighted-Sad/Sad 0.45 0.65 calm and delighted-anxious mood combinations appear Pleasure Change: substantially smaller that those reported by subjects Delighted/Delighted Delighted/Anxious 0.45 5.05 0.55 0.60 assigned to moods that vary in both pleasure and Anxious/Delighted Anxious/ Anxious 3.60 0.75 0.65 0.70 arousal. However, they are comparable to other one- Calm/Calm 0.60 0.55 dimensional mood change conditions, namely the Calm/Sad Sad/Calm 4.65 4.30 0.85 0.95 anxious-sad and calm-sad combinations, both of which Sad/Sad 0.45 0.65 show significant dependent effects. This comparability Pleasure and Arousal Change: in mood change indicates that some other factor must Delighted/Delighted Delighted/Sad 0.45 5.60 0.55 5.00 be responsible for the negligible evidence of M D M in Sad/Delighted Sad/Sad 4.30 0.45 4.85 0.65 delighted-calm and delighted-anxious mood Calm/Calm 0.60 0.55 combinations. Calm/Anxious Anxious/Calm 4.20 4.35 3.90 3.35 Induced moods should possess certain features Anxious/Anxious 0.75 0.70 in order to encourage M D M effects. One such feature is mood intensity. In six of seven prior studies that produced positive M D M , the standard L I R I C mood manipulation was not considered successful until subjects reported "very" or "extremely" Mood Dependent Memory 41 Table 15. pleasant or unpleasant moods. In this study, as in one prior Free Recall for Subjects Meeting More Stringent demonstration, subjects were required to achieve only Pleasure Criterion at Study moderate pleasure intensity. It seemed reasonable to examine Study/Test Mood n % the patterns of recall performance for subjects who met this more stringent criterion, on either pleasure or arousal ratings. Arousal Change: Selection of subjects on the basis of this arousal criterion Delighted/Delighted Delighted/Calm 9 4 62.0 47.9 proved unworkable. Selecting on either Session arousal scores Calm/Delighted Calm/Calm 6 8 48.6 57.3 left numerous empty cells or cells which were represented by Anxious/Anxious 6 55.6 fewer than three subjects. Selecting on more intense pleasure Anxious/Sad Sad/Anxious 8 7 45.8 54.8 scores in the Study session left 96 subjects whose scores are Sad/Sad 5 78.3 presented in Table 15. Pleasure Change: This selection process revealed M D M in all Delighted/Delighted Delighted/Anxious 9 8 62.0 53.1 combinations that had shown it before, in fact, now they all Anxious/Delighted Anxious/Anxious 3 6 58.3 55.6 reached conventional significance. The results also changed for Calm/Calm 8 57.3 the delighted-calm mood combination. Interestingly, selection Calm/Sad Sad/Calm 7 4 47.6 52.1 according to more intense pleasure ratings did not change the Sad/Sad 5 78.3 outcome in the delighted-anxious moods, that vary on pleasure, Pleasure and Arousal Change: but revealed significant M D M (F(l, 23) = 5.42, g < .03), in the Delighted/Delighted Delighted/Sad 9 7 62.0 57.1 delighted-calm combination, that varies on arousal. Sad/Delighted Sad/Sad 4 5 56.3 78.3 Another important aspect of induced moods, is relative Calm/Calm 8 57.3 durability. Inclusion of subjects whose manipulated moods had Calm/Anxious Anxious/Calm 3 7 41.7 47.6 faded prior to task completion may attenuate detection of Anxious/ Anxious 6 55.6 M D M effects. Subjects were selected for analysis if post-task ratings of neither pleasure nor arousal ratings crossed the center lines in either session. Ratings for 101 subjects remained in the appropriate Affect Grid quadrant in both sessions. A marginal interaction in anxious-calm moods (F(l,17) = 3.44, g < Mood Dependent Memory 42 .09) was the sole term to even approach significance (next largest effect: anxious-sad, F ( l , 18) = 2.37, p<.14). Finally, another feature of induced moods is that they should be genuine. Analyses dropping subjects who reported moderate or worse genuineness revealed precisely the same pattern as did the overall analysis (n = 120). Specifically, though all other combinations did, delighted-anxious and delighted-calm combinations did not support significant or marginal M D M effects. The results duplicate the overall pattern in previous publications. Subjects whose moods changed on both pleasure and arousal showed a larger memory disruption than subjects whose moods changed on pleasure alone. However, more detailed investigations of individual 2 x 2 ANOVAs revealed little or no support for M D M in the recall of delighted-anxious and delighted-calm mood subjects, whereas delighted-sad, calm-sad, calm-anxious, and anxious-sad combinations supported marginal to significant mood dependent effects. 2.4.4 RECOGNITION PERFORMANCE Upon completing free recall, subjects were asked to indicate whether or not they had described a memory for each of 24 words. Items that were produced in free recall were removed for these analyses. Recognition performance was evaluated by calculating hits (percent old items rated old) minus false alarms (new items rated old) and are presented in Table 16a. The interaction between Study and Test moods proved reliable (F(9, 143) = 2.11, p< .03). Planned comparisons revealed that subjects assigned to experience matched moods showed better recognition of items than did subjects assigned to experience arousal change alone or both pleasure and arousal change (both Fs (1, 155) > 7.1, p < Table 16a. % Hits - % False Alarms as a Function of Pleasure/Arousal Change Change in Pleasure/Arousaln % Small/Small Large/Small Small/Large Large/Large 39 40 40 40 87.6 82.3 75.5 75.7 Note: 1 Sad/Sad subject recalled all items Mood Dependent Memory 43 Table 16b. % Hits - % False Alarms as a Function of Induced Mood at Study and Test %Hits Study/Test Mood (SD) -%FA Arousal Change: Delighted/Delighted (19.1) 85.9 Delighted/Calm (25.0) 77.6 Calm/Delighted (28.5) 68.9 Calm/Calm (9.5) 93.3 Anxious/ Anxious (18.2) 87.5 Anxious/Sad (14.7) 77.6 Sad/Anxious (18.7) 77.8 Sad/Sad (13.2) 83.3 Pleasure Change: Delighted/Delighted (19.1) 85.9 Delighted/Anxious (22.1) 74.1 Anxious/Delighted (18.2) 87.4 Anxious/Anxious (18.2) 87.5 Calm/Calm (9.5) 93.3 Calm/Sad (17.3) 88.4 Sad/Calm (23.1) 79.4 Sad/Sad (13.2) 83.3 Pleasure and Arousal Change: Delighted/Delighted (19.1) 85.9 Delighted/Sad (24.6) 68.4 Sad/Delighted (13.5) 85.6 Sad/Sad (13.2) 83.3 Calm/Calm (9.5) 93.3 Calm/Anxious (19.1) 75.2 Anxious/Calm (23.8) 71.5 Anxious/Anxious (18.2) 87.5 Note: Sad/Sad n=9, 1 subject recalled all items .01). However, subjects assigned to pleasure change alone did not show significant memory disruption, and their performance did not differ reliably from that of any other group (all Fs < 2.4, gs > . 12). As with free recall, six separate ANOVAs , considering Study mood, Test mood and their interaction, were performed to compare each mood to each other mood. The data are presented according to individual ANOVAs in Table 16b. Analysis of combinations that vary on arousal revealed a mood dependent interaction between delighted and calm (F(l, 36) = 5.64, g < .02) but not between anxious and sad moods. There were no significant M D M effects in recognition when subject moods varied on pleasure. That is, in neither delighted-anxious, nor calm-sad (Study Mood F ( l , 35) = 3.17, p < .08) mood combinations did the interaction terms approach significance. Delighted and sad moods, that vary on both pleasure and arousal, produced only a marginal test effect (F(l, 35) = 3.41, p < .07), while calm-anxious moods, that also vary on pleasure and arousal, produced a significant interaction (F(l, 36) = 5.64, g < .02). Mood Dependent Memory 44 Table 17. d' as a Function of Induced Mood at Study and Test Study/Test Mood (SD) d' Arousal Change: Delighted/Delighted (1.01) 3.7 Delighted/Calm (1.44) 3.2 Calm/Delighted (1.42) 2.9 Calm/Calm ( .71) 4.1 Anxious/Anxious (1.11) 3.9 Anxious/Sad ( -76) 3.1 Sad/Anxious ( -96) 3.2 Sad/Sad (1.13) 3.2 Pleasure Change: Delighted/Delighted (1.01) 3.7 Delighted/Anxious (1.30) 2.9 Anxious/Delighted (1.08) 3.8 Anxious/ Anxious (1.11) 3.9 Calm/Calm (.IV 4.1 Calm/Sad (1.08) 3.9 Sad/Calm (1.26) 3.4 Sad/Sad (1.13) 3.2 Pleasure and Arousal Change: Delighted/Delighted (1.01) 3.7 Delighted/Sad (1.26) 2.8 Sad/Delighted ( .81) 3.8 Sad/Sad (1.13) 3.2 Calm/Calm ( .71) 4.1 Calm/Anxious ( .84) 3.1 Anxious/Calm (1.31) 2.9 Anxious/Anxious (1.11) 3.9 Note: Sad/Sad n=9, 1 subject recalled all items To reduce possible impact of response bias, recognition was also evaluated via d' scores, presented in Table 17. Analysis of d' scores revealed precisely the same pattern of dependent recognition effects as did the overall analysis. Namely, significant M D M recognition was restricted to delighted-calm and anxious-calm mood combinations (both Fs (1, 36) > 8.5, p_s < .01). Finally, mood manipulations provide the same possibility of attenuating M D M in recognition as in recall. Analyses were performed on subjects selected according to initial induction of more intense Study moods (n=95), moods that remained in the appropriate Affect Grid quadrant post-task in both sessions (n= 100), or moods rated as more than moderately genuine (n=T 19). Only selection of subjects who reported more intense moods made any change to the effects. Specifically, effects in delighted-calm and anxious-calm mood combinations continued to be marginal or significant but application of this selection critierion indicated that the anxious-sad combination supported a marginal interaction (F(l, 21) = 3.86, p < .07). As in free recall, selection of more intense pleasure ratings changed interpretation for a combination that varies on arousal. Comparison of current results to Beck & McBee's (1995) report required recognition to be re-calculated based on all generated items. These scores are presented as a function of Study Mood Dependent Memory 45 Table 18. Recognition of All Generated Items % Hits - % False Alarms as a Function of Induced Mood at Study and Test %Hits Study/Test Mood (SD) -%FAs Arousal Change: Delighted/Delighted ( 8.6) 93.3 Delighted/Calm (19.0) 84.2 Calm/Delighted (18.5) 78.3 Calm/Calm (4.3) 96.7 Anxious/Anxious (10.3) 93.3 Anxious/Sad (8.3) 84.2 Sad/Anxious (10.2) 85.0 Sad/Sad ( 4.0) 94.2 Pleasure Change: Delighted/Delighted (8.6) 93.3 Delighted/ Anxious (14.9) 82.5 Anxious/Delighted (10.7) 92.5 Anxious/Anxious (10.3) 93.3 Calm/Calm (4.3) 96.7 Calm/Sad (12.1) 92.5 Sad/Calm (13.2) 87.5 Sad/Sad (4.0) 94.2 Pleasure and Arousal Change: Delighted/Delighted (8.6) 93.3 Delighted/Sad (14.9) 82.4 Sad/Delighted (5.6) 94.2 Sad/Sad ( 4.0) 94.2 Calm/Calm (4.3) 96.7 Calm/Anxious (10.4) 83.3 Anxious/Calm (20.1) 80.0 Anxious/ Anxious (10.3) 93.3 Note: Sad/Sad n=9, 1 subject recalled all items and Test moods in Table 18. Although, overall, moods induced at Study and Test produced no main effects (both Fs < 1), their interaction was reliable (F(9, 144) = 3.21, g<.01). Considering combinations that vary on arousal, both the delighted-calm (F(l, 36) = 9.49, p < .01), and the anxious-sad combination (F(l, 36) = 11.3, g < .01) produced significant interactions indicating dependent recognition. Of combinations that vary on pleasure, only the calm-sad mood combination supported M D M in a marginal interaction (F(l, 36) = 3.31, g < .08). The delighted-sad mood combination revealed a study mood effect (F(l, 36) = 4.01, g < .04), a marginal main effect of test mood (F(l, 36) = 3.46, g < .07) and a marginal interaction (F(l, 36) = 3.46, g < .07), whereas the anxious-calm combination revealed only a significant interaction (F(l, 36) = 11.1, g < .002). Recognition performance in all but the delighted-anxious mood combination, provided some evidence of mood dependence when the data was drawn on all generated items. Particularly interesting is comparison of results in the delighted-sad combination according to exclusion or inclusion of previously recalled items. This combination fits most closely with previous investigations of M D M in recognition. As predicted, items that had not been recalled did not support M D M effects whereas inclusion of recalled items Mood Dependent Memory 46 showed a marginal M D M interaction, indicating that selection of scoring method played a pivotal role in the apparent contradiction in published reports (Eich & Metcalfe, 1989; Beck & McBee, 1995). Recognition results support the pattern of performance previously described in only free recall, namely, greater evidence of M D M when moods vary on both pleasure and arousal than when they vary on pleasure alone. However, as with free recall, subjects assigned to experience arousal change showed memory disruption equivalent to that demonstrated by subjects whose moods change on both pleasure and arousal, indicating that one-dimensional changes in mood are not all equivalent. 2.4.5 SUMMARY OF RESULTS The majority of subjects who under took the study achieved at least moderately intense pleasure or displeasure simultaneously with moderately high or low arousal on two occasions. However, addition of an arousal criterion to the mood manipulation procedures did significantly lower the proportion of subjects who could successfully experience two moods (to 65% from 88%>), and was the most frequent cause of mood failure. Affect Grid data are consistent across tasks and sessions. The mood criteria ensured large and reliable group differences on pleasure and arousal ratings and indicated comparable initial intensity of pleasure/displeasure and heightened/lowered arousal across inductions. However, differences between moods did become evident post-task. Unpleasant and low arousal moods moods showed a greater tendency to return to neutral ratings than did pleasant or high arousal moods. Yet, the mood manipulations were still significantly influencing post-task ratings. PANAS data confirm mood manipulation success by indicating that delight and sadness, which were expected to show the highest and lowest P A scores respectively, differed significantly from the average P A reported by other subjects. Similarly, anxious and calm groups, which were expected to show the highest and lowest N A respectively, differed significantly from the average N A reported by other subjects. Mood Dependent Memory 47 The induced moods led to mood congruent effects in the Event Generation task. Induction of pleasant moods encouraged subjects to produce a greater number of positive memories than did induction of unpleasant moods. Free recall performance was mood dependent, and replicated the previously reported pattern of greater memory disruption in subjects whose moods change along both pleasure and arousal than in subjects whose moods change on pleasure alone. However, one-dimensional change on arousal appeared to produce as strong an effect as did change in both pleasure and arousal. Analysis of individual 2x2 ANOVAs revealed better evidence for M D M in some combinations (anxious-sad, calm-sad, calm-anxious, delighted-sad), than in others (delighted-calm and delighted-anxious). Considering recognition (for items not previously recalled), M D M appeared to follow the same pattern of greater disruption for two- than one-dimensional mood change, a pattern that in previous reports had applied only to free recall. Detailed investigation via individual 2x2 ANOVAs revealed greatest support for M D M in recognition of subjects assigned to delighted-calm and anxious-calm combinations. Mood Dependent Memory 48 2.5 DISCUSSION Prior research consistently demonstrated greater disturbance in recall performance for subjects who experienced two- as opposed to one-dimensional changes in mood (Eich & Metcalfe, 1989; Eich et al., 1994). Results of the current study supported this pattern in certain aspects but not others. In particular, subjects who experienced change in pleasure alone did not demonstrate as much disruption in recall as did subjects whose moods varied on both pleasure and arousal. However, addition of subjects whose moods varied on arousal alone allowed comparisons that were unavailable in prior research, and their results indicate different memory implications for both types of one-dimensional mood change. Specifically, though recollection performance was only marginally worse after pleasure change, it was equally disrupted by change in arousal alone or both pleasure and arousal change. Figure 8. Relationship among MDM effects in Free Recall. — Anxioi elighted IS u \ • / s v / • Sad " Calm 1 1 1 Interaction p < .01 Interaction p < .10 Interaction p < .12 More detailed analyses revealed that mood combinations which varied on the same dimension or dimensions often differed in the potency of mood effects in recall (Figure 8). Though M D M was generally not supported in delighted-calm moods, it was consistently significant in anxious-sad mood combinations. Whereas M D M was never significant in delighted-anxious moods, it was reliable in calm-sad mood combinations. And though M D M was supported in both conditions of two-dimensional mood change, it appeared larger in anxious-calm than delighted-sad combinations. Mood Dependent Memory 49 Figure 9. Relationship among MDM effects in Recognition. 1 1 - Anxious elighted U dad 1 Calm 1 Interaction p < .05 In prior investigations, (Eich & Metcalfe, 1989), there was no reliable decline in recognition memory following mood change. In the current investigation (Figure 11), pleasure change did not, but both pleasure and arousal change and arousal change alone did reliably impair recognition performance. More detailed analyses suggest that the overall pattern is supported primarily by two particular mood combinations; namely the delighted-calm and anxious-calm mood combinations. Three approaches to emotion were described in the introduction. Each approach considers different characteristics to best describe the fundamental nature of emotion. These models were introduced as a starting point from which to derive predictions about psychological distinctiveness between mood combinations. As the models select different attributes to represent the elementary features of emotion, they can be interpreted to predict greater distinctiveness between certain mood combinations than others; predictions that vary across models. This psychological distinctiveness may also translate into larger mood dependent effects. Predictions drawn from each model received some support. Russell et al's (1989) Affect Grid approach considers pleasure and arousal the fundamental dimensions of mood and its derivation predicted greater mood dependence for moods that vary on both pleasure and arousal over moods that vary on a single dimension. Thus, mood dependence may be greater in contrasts of delighted and sad moods or of anxious and calm moods than in other combinations. Both these contrasts did tend to show dependent effects, supporting the predictions drawn from this model. Mood Dependent Memory 50 However, one-dimensional changes in mood varied considerably in their memory effects. Arousal change and pleasure change sometimes did and other times did not produce dependent effects. Considering arousal change, anxious-sad moods produced strong M D M whereas there was only slim evidence of M D M in contrasts of delighted and calm moods. Considering pleasure change, dependent memory effects were reliable between calm and sad but not between delighted and anxious moods. Watson and Tellegen's (1985) PANAS approach considers the basic dimensions of emotional experience to be P A and NA. Predictions based on this model are for contrasts of moods that vary on both its fundamental features; namely anxious (High NA) - sad (Low PA) and delighted (High PA) - calm (Low NA) mood combinations to encourage large mood dependent effects. While the largest M D M effect was evident in the anxious-sad mood comparison, the delighted-calm mood combination provided only slim evidence of M D M . Moods that varied on either P A (delighted-sad), N A (anxious-calm), or between moods associated with low levels of N A and P A (calm-sad), produced significant dependent effects, but M D M was not evident in combinations which varied across high levels of P A and N A (delighted-anxious). Johnson-Laird and Oatley's (1988) evolutionary approach posits basic cognitive differences among the moods under investigation excluding delighted and calm moods. Extending this to the current realm, mood dependence should be unlikely in delighted-calm mood combination; a prediction borne out by the data. The other moods under investigation were considered to reflect different fundamentally distinct basic emotions which should lead to comparable M D M in their combinations. The delighted-anxious comparison did not support significant dependent effects, a problem which was compounded by the fact that M D M was supported in the calm-anxious combination. If delighted and calm moods do not differ cognitively, as they are both a form of the basic emotion "joy", then combinations that produce mood dependence in one should produce mood dependence for the other. Delight produced dependence only when crossed with sadness, while calm and anxious moods (crossed with each other or sad moods) produced M D M on two occasions. In contrast, Mood Dependent Memory 51 sadness crossed with any other mood produced dependent recall. The data clearly indicate that moods differ fundamentally in their relative susceptibility to M D M , but that predictions drawn from no single theory of emotion reviewed here can easily reconcile all the free recall data. What aspect of a sad mood may be particularly conducive to production of M D M ? According to Lazarus (1991), a sad mood may differ fundamentally from other moods because its action tendency is inaction and withdrawal, and a sad mood may be unique among negative moods in its self-focused nature (Pyszczynski et a l , 1989). An assumption drawn from evolutionary approaches is that moods ready different prototypical action patterns, and therefore, prime certain types of cognitive processing. Studies confirm that subjects under the influence of sad moods tend to learn and remember negatively charged materials better than do subjects under pleasant moods, especially when the material is self-referencing (Blaney, 1986). Comparable studies examining memory for negative material in anxiety have been much more mixed (Macaulay et al., 1993). In contrast, studies of anxious subjects have quite consistently shown an attentional bias to threat on vigilance tasks (Matthews & Eysenck, 1987), whereas sad subjects do not appear to show such a bias (Blaney, 1986). Thus, mood congruent effects tend to show an attentional~but not memorial—bias in anxiety, and a memorial—but not attentional—bias in sadness. This evidence suggests that anxious and sad moods enhance opposing basic cognitive processes. Consistent with this assertion, subjects who were sad at Study showed significantly higher recall than did subjects who were anxious at Study (means of 60.2 and 50.8). Thus, reflecting on and describing past occurrences during the event generation task may be precisely the type of processing that is enhanced by sadness, whereas anxiety—with its biased focus on external threat—may make it more difficult to turn attention inward. These opposing biases in cognition may amplify the psychological distinctiveness between anxious and sad moods leading to the large mood dependent effect between them. Although evidence exists regarding cognitive changes accompanying anxious and sad moods, there is little direct evidence about cognitive changes induced by joy. According to some Mood Dependent Memory 52 theorists, the purpose of joy is to encourage the initiation and maintenance of relationships (Johnson-Laird & Oatley, 1989), which should lead to the action tendencies of "expansiveness and approach to share one's good fortune" (Lazarus, 1991, p 823). Isen and her colleagues (Isen et al., 1978) have concentrated on the cognitive impact of positive affect and base their predictions on the associative network model of emotion and knowledge. However, the particular action tendency associated with joy is less clear. It seems reasonable to agree with Lazarus' (1991) assertion that joy may encourage one to attend to the environment to seek out others with whom to share pleasurable feelings. Joyful moods (delighted and calm) and may share with anxiety this outward orientation and may differ from sadness on this dimension. Initially, the evolutionary model was interpreted to predict differences between moods based on specific differences in their action tendencies. If, instead, orientation of attention is reflective of distinctiveness between moods, then mood dependent effects may be more likely when moods which differ in orientation are crossed with each other. The Affect Grid approach could also be taken to suggest that, among moods sharing a similar orientation, M D M effects may be more likely between moods that vary on both pleasure and arousal rather than a single dimension alone. The integration of these two approaches may be beneficial for the purposes of future cognitive research. According to this more integrated approach, moods that share an internal focus but vary on Affect Grid dimensions, may follow the same pattern as demonstrated here among moods that share an external focus. These suggestions would be untestable under Johnson-Laird & Oatley's (1988) evolutionary scheme given that there are only 5 basic emotions, but other approaches, such as Frijda.(1986) and de Rivera (1977) include a larger group of characteristics which define differences between emotions. These approaches may suggest moods that share an internal orientation but vary on the dimensions of pleasure and arousal. A test of these predictions would then be possible by determining if pleasure and arousal change control the appearance of mood dependent effects. Mood Dependent Memory 53 Other comparisons would also be of interest. The contrast of anger and anxiety, for instance, would compare moods which represent different basic emotions with different evolutionary purposes and action tendencies. These moods would share an external focus as well as displeasure and high arousal. According to the new approach, though anger should not show dependent memory effects when crossed with anxiety, it should produce results similar to those of anxiety. That is, while crossing with delight may not, crossing with calm or sad moods should encourage mood dependence. The development of new mood manipulations would allow evaluation of this approach for predicting the appearance of M D M effects. Although the pattern of free recall results seems understandable, a different pattern emerged in recognition performance. The recognition task was initially included to demonstrate that contradictory reports of dependent recognition (i.e. Eich & Metcalfe, 1989; Beck & McBee, 1995) could be an artifact of scoring method rather than a substantive difference in results. As shown by the delighted-sad mood combination—which is most akin to previous research—scoring method is indeed a crucial consideration. Recognition on items not freely recalled revealed little evidence of M D M but their inclusion revealed, perhaps spurious, M D M recognition. Beyond this demonstration, the recognition data are difficult to reconcile with either this new view of emotional effects in cognition or with previous assertions that recognition may generally be less sensitive to mood dependence. One combination, delighted-calm, showed dependent recognition but not recall; an effect documented only once before using an elated-depressed mood combination (Leight & Ellis, 1981). Interestingly, the anxious-calm combination is the sole comparison that shares an external focus and varies on both pleasure and arousal, and is likewise the sole combination to show dependence on both recall and recognition. Though M D M was supported in all sad mood combinations in free recall, no sad mood combination revealed dependent recognition. These results indicate not only that moods and tasks vary in their sensitivity to dependent effects, but that the factors may be interactive. A more detailed analysis of processing undertaken Mood Dependent Memory 54 in recall and recognition will have to be evaluated while also considering processing changes that accompany different moods. On the whole, pitfalls inherent in both scoring methods render the recognition data difficult to interpret. Recognition analyses based on items that were not recalled has two possible drawbacks. The high rate of recall in the current investigation meant that recognition was evaluated for only 40 to 50% of the total items. Removal of recalled items may also mean that this subset comprises only the most difficult items. Recognition analyses based on all recognized items also has drawbacks. Items having been previously recalled are likely to be recognized and if recall performance was mood dependent then recognition is likely to show the same pattern. Ultimately, neither approach is sufficient to make clear determination on the reality of mood dependent recognition. To answer questions about M D M in recognition, future research should avoid contamination between tasks by ensuring separation between items targeted for recall and those targeted for recognition. Perhaps the most realistic approach is investigation of recognition in its own right, not as a task secondary to recall. Most interesting, would be investigations of the two mood combinations that showed M D M even under conservative scoring methods. 2.5.1 S U M M A R Y The experiment was successful in showing that mood manipulations produced moods that endured task performance, that were rated as moderately genuine, and that lead to mood dependent recall and recognition in some, but not all, mood combinations. The integration of the evolutionary and Affect Grid approach makes greater headway in clarifying which combinations of moods are likely to be profitably explored in M D M research than does any single approach alone. Problems caused by recall and recognition being performed on the same items make interpretation of recognition performance difficult. Future research to address this issue should rely on recognition of items not previously included in a recall task. Mood Dependent Memory 55 C H A P T E R 3. EXPERIMENT 2: INVESTIGATION OF MOOD DEPENDENCE IN BIPOLAR AFFECTIVE DISORDER. 3.1 ISSUES TO BE RESOLVED An important issue remaining to be resolved is validation of mood dependent effects in groups that are experiencing real, yet unmanipulated, changes in mood. Some investigators have shown that mood congruence, considered both robust and reliable (Blaney, 1986), may not be evident in the performance of subjects who are assigned to different mood conditions according to their pre-existing moods (Hasher et al., 1985; Parrott & Sabini, 1990). As M D M is generally a less reliable effect than M C M , it is important to validate the effect among subjects who have not been exposed to any mood induction. This is especially true given that M D M has been proposed to contribute to memory problems in a variety of clinical disorders. If so, mood dependence should be evident in the memory performance of patients, such as manic-depressives, who repeatedly experience strong changes in their mood. Though there have been numerous authors speculating about M D M in clinical groups, there remains only one published clinical evaluation of mood dependence. Weingartner, Miller and Murphy (1977) investigated the impact of manic-depressive illness on basic memory performance. The researchers documented mood patterns in a group of 8 manic-depressives for 8-20 weeks. In this period, subjects experienced both normal and manic mood states. Every four days, mood was evaluated on a number of scales and subjects generated 20 free associations in response to each of 2 standard word probes. These self-generated items served as the target materials in a test of memory four days later. After re-evaluating mood, subjects then were asked to remember words they had generated for both probes. They then generated 20 free associations for 2 new probe words which would serve as the target items in the next session. Mood Dependent Memory 56 The researchers reported that the difference in mood scores for the two sessions correlated -.35 with items remembered. Patients reproduced 93% more associations when they experienced moods during memory testing which matched the moods experienced during generation of the free associations. Specifically, items generated in normal moods were readily reproduced if subjects re-experienced normal moods at test (8.4 items) but were reproduced less often if subjects experienced manic moods at test (4.4 items). Importantly, the reverse was also true. For items generated in manic moods, re-experience of a manic mood at test led to significantly better performance than did the experience of a normal mood at test (5.7 vs 2.9 items). If mania simply disrupted task performance, then memory testing during mania should lead to worse performance than testing during normal moods. Rather, these data suggest that mood change disrupted memory. Conceptual replication of Weingartner Miller & Murphy (1977) is important for two reasons. First, on practical grounds, an attempt to replicate is warranted in light of the previous equivocal history of M D M . Second, on theoretical grounds, the processes which lead to M D M are worth further investigation. The results clearly indicate that matched moods lead to a greater proportion of items being reproduced at test. What remains unclear is why. Subjects were presented with probes that had been previously used in free association and were asked to "recall their associative responses." However, the authors note that subjects may have employed one of two strategies for doing so. They may have attempted to recall, in an explicit and episodic manner, the associates from the previous time. Alternatively, they may have used the probe for producing associates that came to mind during the test. In the present study, letter association will be used in an attempt to distinguish between these strategies, in order to evaluate if both are equally sensitive to mood dependence. To compare implicit and explicit strategies, it is desirable to provide subjects with equivalent cues, but to vary the instructions regarding how subjects should perform the tasks (Schacter, Bowers, and Booker, 1989). Here, similar cues will be provided for subjects who will be encouraged to deliberately remember previously generated associations (explicit memory task) Mood Dependent Memory 57 or to perform anew the task of listing the first associates that come to mind for a given letter (implicit memory task). M D M in explicit but not implicit memory task performance would indicate that mood was a cue used in subjects' conscious retrieval strategies. M D M in both explicit and implicit tasks would indicate that mood became incorporated into the memory trace, and partial re-presentation in the form of a matching mood allowed broader redintegration of the original event. The question of sensitivity to mood effects remains to be resolved among other memory tasks as well. Past research has often involved 2 to 3 tasks in each experiment. Although this is understandable given the time and effort invested in each subject, it also has some drawbacks. Manipulated moods tend to be strongest when subjects reach the criterion and tend to decline somewhat over the session. Clarification of the relationship among tasks requires examination of a variety of tasks in a group of individuals whose naturally occurring moods remain stable in a session. In studies with undergraduates, consideration of specific processes required by Study and Test tasks, as well as the way these processes overlap, appears more important than the general implicit or explicit nature of the memory demands. Specifically, tasks which rely upon subjects generating materials at study and on subjects generating and selecting among candidates for solution of implicit memory tests have, on some occasions at least, shown mood dependence. Although implicit tasks generally, and visual tasks more specifically, appear less sensitive to mood (see p. 17), it is unclear if the experience of stable and extreme moods could lead to mood dependent effects. To answer these questions, implicit and explicit memory performance was examined among individuals diagnosed with rapid-cycling Bipolar Disorder. 3.2 M E T H O D 3.2.1 SUBJECT RECRUITMENT Through advertisement in the Mood Disorders Association of British Columbia newsletter, and collaboration with Dr. Raymond Lam in U B C Mood Disorders Clinic, clinical Mood Dependent Memory 58 patients with a current DSM-IIIR diagnosis of Bipolar Affective disorder and a history of rapid mood change were identified. By convention, specification of rapid-cycling Bipolar Disorder refers to patients who display 4 or more complete depression/mania cycles per year. 3.2.2 DESIGN Subjects were informed that the study was aimed at understanding how emotional states affect various cognitive processes such as learning, perceiving, and remembering. They were informed that participation was required through periods of depression, periods of mania/hypomania, and periods of mood change and involved a variety of tasks, including evaluation and generation of certain materials, general knowledge, and the ability to remember material from the previous session. By examining mood history and in consultation with the subject, a time was selected which served as the standard interval between study and test sessions for that subject. The interval ranged from 2-7 days. Meetings took place on a regular basis in which the aim was for subjects to complete each of 4 cells in a 2 x 2 within-subject design. In principle, study mood (depression or mania/hypomania) was to be crossed with test mood (depression or mania/hypomania), requiring a minimum of 4 Study/Test combinations. There were 8 unique sets of Study materials limiting each subject to a maximum of 8 Study/Test combinations. In practice, to be covered in detail in the Subjects section, only 5 subjects were able to complete the study in its entirety, with 5 others completing partial designs (viz. 2 or 3 Study/Test combinations). 3.2.3 STUDY SESSION At the beginning of the first session, subjects were provided with a copy of the Affect Grid, presented in Figure 1 (p. 9). The experimenter explained the Grid in detail, as described earlier (pp. 9-10). After marking current levels of pleasure and arousal, the experimenter explained the PANAS scales (p. 24), which subjects then completed. Subjects were then read a list of 10 common unrelated words, such as DESK and SLIPPER. Subjects performed event generation as described earlier (pp. 25-26). Subjects were Mood Dependent Memory 59 instructed to say OK as soon as they generated a suitably specific event. If such an event was not generated within two minutes, then that probe was skipped and the experimenter presented the next word. If generation was successful, the experimenter transcribed generation latency, event description, as well as ratings of event intensity, importance and vividness. These probe words later served as targets for both explicit free recall and implicit category production in the Test session. Next, subjects were shown 4 multi-colour inkblots one at a time for 5 sec each. Their task was to examine each inkblot in order to provide a rating of aesthetic pleasantness on a scale of -2 (extremely unpleasant) to +2 (extremely pleasant). This task provided subjects with the opportunity of studying the inkblots which they were later asked to select from five similar distractors. Subjects were then provided with a letter of the alphabet. Their task was to say aloud 20 associates, or words beginning with that letter, as quickly as possible. The experimenter kept a written record of the 20 associations and the time taken for two different letters. These items later served as targets under both explicit recollection and implicit re-association instructions. Finally, subjects were shown four line drawings of common objects, such as a train or a ball (Snodgrass & Corwin, 1988), one at a time for 5 sec each. Their task was to examine each complete picture (for an example see Figure A in Appendix III) for that time to provide a rating of complexity on a scale of 0 (not at all complex) to 3 (very complex). This task provided subjects with the opportunity of studying the pictures which they were later asked to identify in an implicit picture fragment completion task. 3.2.4 TEST SESSION Subjects returned for testing 2-7 days later. Mood was measured via the Affect Grid and PANAS scale. Testing involved six tasks which assessed either explicit or implicit memory for materials from the immediately preceding Study session. Mood Dependent Memory 60 Two tasks tested for memory of the previously generated autobiographical events. Explicit free recall was performed in the latter part of the Test session. Subjects were reminded that in the prior session they generated personal events in response to certain probe words. Subjects then spent 3-5 minutes trying to remember either the probes or the events they described. The implicit task, category production, required subjects to generate six items belonging to each of 10 categories, such as ARTICLES OF FURNITURE or ARTICLES OF CLOTHING. Event generation probes were candidates for five of the categories. Two tasks tested memory for visual stimuli. Explicit inkblot recognition was tested by presenting subjects with 4 sets of six inkblots. Each set was composed of one inkblot which had previously been rated placed randomly among 5 extremely similar alternatives. Subjects were to select the previously viewed inkblot and rate their confidence in that selection on a scale from 0 (not at all confident) to 3 (very confident). Five extremely similar distractors were employed to ensure that recognition would be of moderate difficulty. The other task tested implicit memory for pictures. Subjects were presented with 8 series of picture fragments (for an example see Appendix III). Each series was composed of eight fragmented versions of the same object. The initially presented fragment (Figure H) contained very little information but each one that followed contained slightly more detail until the complete object was presented in the eighth and final drawing (Figure A). Subjects were asked to name the object as early in the series as possible, a point which was recorded by the experimenter. Four of the objects had been previously viewed and rated by subjects in the Study session. Two tasks tested memory for the letter associations. In the explicit recollection task, subjects were reminded that in the last session they had listed letter associates for certain letters. Subjects then spent 3-5 min trying to recall their associates for one of those letters, and were specifically requested to refrain from guessing. In the implicit memory task, subjects were instructed that letter association was to be performed for more letters. They were reminded that their task was to "go as quickly as possible" naming the first 20 associates that came to mind. They were given two new letters which were followed by a third that had been previously used for Mood Dependent Memory 61 letter association in the Study session. If subjects remarked that the letter had already been performed, the experimenter replied that, because not all letters were being used, some had to be re-cycled so that there would be enough for all sessions. They were reminded that, regardless, the task was to list 20 associates as quickly as possible. The experimenter kept written record of the associates and the time taken for each letter. To reduce the number of trips that subjects were required to make to U B C hospital, a Test session was often followed immediately by a Study session with new materials. Subjects could then return after the appropriate time interval for testing on this new set of items. The decision whether or not to do so was made in consultation with subjects prior to testing by determining which conditions remained for completion and by their educated estimation of how their mood was likely to change in the intervening interval. The order of Test session tasks was as follows: category production, picture fragment completion, implicit testing of letter associations, inkblot recognition, recall of event generation probes, and explicit testing of letter associations. However, if a new Study session was scheduled immediately, implicit testing of letter associations took place during that Study session by asking subjects to produce letter associates for two new letters (the Study letters), then one that they had previously performed (testing implicit memory for those letter associates). 3.3 SUBJECTS Ten subjects (9 women and 1 man) ranging in age from 21 to 47 years (mean of 33.7 years) completed at least 2 of the 4 possible mood combinations while reporting moods that were judged strong enough to reflect depression or mania. Al l four mood combinations (M/M, M/D, D / M , and D/D) were completed by five subjects (4 women and 1 man). Two of these subjects (both women) are noteworthy in that on one occasion they experienced mania associated with unpleasant feelings. Although somewhat atypical, these occasions appear to be true examples of mania, so they are included in the analyses. Five other subjects (all women) completed partial designs before testing was discontinued. Two subjects completed three mood combinations Mood Dependent Memory 62 before the supply of study materials was exhausted. Three subjects completed only two usable mood combinations before their moods stabilized for a considerable length of time (2 subjects) or before the supply of study materials was exhausted (1 subject). These ten subjects contributed 32 conditions corresponding to hypo/mania on two occasions (9 conditions), depression on two occasions (7 conditions), change from hypo/mania to depression (8 conditions) or change from depression to hypo/mania (8 conditions). To accumulate these 32 conditions in which subjects reported strong moods, 53 conditions were collected; that is, on 21 additional occasions, subjects completed both Study and Test sessions. These conditions are not used in most analyses for one of two reasons; either moods duplicated those that had already been collected (7 cases) or only mild moods were reported at Study and/or Test (14 cases). As mood dependence may not be evident unless moods are sufficiently intense (Eich, 1995), occasions when subjects report strong moods are of the greatest current interest, and will constitute the basis of most analyses. Infrequently, events occurred that rendered data from a particular subject incomparable to his or her other data (7 cases: 1 medication change between Study/Test sessions, 1 medically advised sleep deprivation the night before testing, 1 Test appointment that was completed two days late, and 4 conditions in which subjects were undergoing temporary courses of medications noted for causing frequent memory problems). These conditions are considered invalid and as such are never employed in analyses. A total of 5 other subjects (3 women and 2 men) agreed to participate, but either proved to be untestable or completed only one Study/Test pair of sessions before dropping out of the study. 3.3.1 MEDICATION STATUS AND DIAGNOSES A l l but two of these ten subjects received some medication or combination of medications to treat their mood disorder (see Table 19). Two subjects received the same medication or combination for the entire study. The six remaining subjects reported some medication change during their participation. Two subjects began the study unmedicated but later started drug Mood Dependent Memory 63 therapy. The remaining four subjects were on chronic lithium combined with other drugs that changed over the course of the study. Table 19. Medication Status Medications Medication Change 1. Medication Change 2. Medication Change 3. 1. Lith 2. Paro 3. 4. 5. LithDesi 6. Lith Valp Carb 7. Lith Sert 8. Lith Halo Mapr Lith Lith Clon Eltr Lith Sert Lith Luvo Lith Proz Lith Clon Risp Lith Paro Lith Zopi Note: Carb = Carbamazepine, Clon = Clonazepam, Desi = Desipramine, Eltr = Eltroxin, Halo = Haloperidol, Lith = Lithium, Luvo = Fluvoxamine (Luvox), Mapr = Maprotiline, Paro = Paroxetine, Proz = Fluoxetine (Prozac), Risp = Risperidone, Sert = Sertraline Valp = Valproic Acid, Zopi = Zopiclone. Of these fourteen medications, drug monographs (CPS & PDR, 1996) cite frequent memory problems or confusion for only the following three: sertraline (zoloft), fluvoxamine (luvox), and zopiclone (imovane). Three subjects received these medications in combination with lithium. One subject switched from sertraline to fluvoxamine, and would have possible drug-associated memory effects in each session and condition. Given that any such effect would be constant, all conditions completed by this subject remain in the analyses. The other subjects used imovane or sertraline for some temporary period and, as stated earlier, these occasions are not included in the analyses. The 10 subjects were diagnosed according to DSM-IIIR criteria for Bipolar I (n=6) or Bipolar II (n=4) disorder. Bipolar I Disorder refers to patients with a history of both depression and mania. In these individuals the mania is sufficiently severe to require hospitalization on at least one occasion, or to cause marked impairment in occupational and/or social functioning. Mood Dependent Memory 64 Patients diagnosed with Bipolar II Disorder have never experienced a true mania but rather have a history of depression and hypomania. 3.4 RESULTS Table 20. Pleasure and Arousal ratings as a function of Study/Test Condition and Session. Study/Test Session 1 Session 2 Moods n Pleas Arous Pleas Arous 3.4.1 MOOD MEASURES Pleasure and arousal ratings are reported in Table 20. Although both Affect Grid and PANAS mood measures were collected, comparability to prior studies made it preferable to use pleasure and arousal ratings to determine whether a condition would be selected for analysis (i.e., those designated as manic/hypomanic (M) or depressed (D)). As in mood manipulation studies, groups should show large and reliable differences on pleasure and arousal ratings. A N O V A of pleasure ratings confirms differences between M and D occasions in Study (F(l, 30) = 73.9, p < .001, M and D means = 1.9 and -3.0), and Test sessions (F(l, 30) = 117.6, p < .001, M and D means = 2.2 and -2.1). Arousal ratings also differ across moods in both sessions (Study • , session F ( l , 30) = 90.0, p < .001, M and D means = 2.8 and -2.1; Test session F ( l , 30) = 305.8, p < .001, M and D means = 2.5 and -2.7). P A and N A ratings are reported in Table 21. Although PANAS ratings were not the principle method to classify conditions, they were used as supplemental information. They provided extra information to make decisions whether ambiguous cases should be considered examples of depression, mania, or mood change. A N O V A of PANAS ratings revealed large and reliable M/M M/D D/M D/D 9 2.1 2.6 8 1.8 3.0 8 -3.0 -2.4 7 -3.0 -1.7 2.2 -1.8 2.2 2.4 -2.9 2.5 -2.6 -2.6 Table 21. PA and NA ratings as a function of Study/Test Condition and Session. Study/Test Session 1 Session 2 Moods n PA NA PA NA M/M 9 36.1 16.8 36.4 17.9 M/D 8 34.0 21.5 17.9 19.1 D/M 8 20.8 26.6 38.9 15.4 D/D 7 16.4 27.3 17.9 24.9 Mood Dependent Memory 65 group differences in P A during both Study (F(l, 30) = 47.1, g < .001, M and D means = 35.1 and 18.7) and Test sessions (F(l,30) = 82.7, g < .001, M and D means = 37.6 and 17.9). Differences in N A ratings are modest but significant at Study (M and D means =19.0 and 26.9, F( 1,30) = 5.99, g < .025). At Test, N A differences fail to meet conventional levels of significance (M and D means = 16.7 and 21.8, F ( l , 30) = 2.67, g < .11). As the manias experienced by most participants in this study were associated with euphoric (as opposed to dysphoric) mood, mania and depression should be represented on opposite ends of PA, and it is important to note that the differences reported on the Affect Grid are confirmed on PA. The data suggest that N A is also affected by these moods, albeit to less extent than PA. On the whole, both mood measures verify that subjects experienced different moods on occasions labeled manic and those labeled depressed. 3.4.2 RESULTS OF STUDY SESSION TASKS Autobiographical Event Generation To examine the autobiographical event generation task for M C M effects, events were divided into positive, neutral, and negative events (percentages are reported as a function of Study session mood in Table 22). Interested readers are referred to Appendix II for a detailed analysis of the associated event ratings. Analyses of generated events revealed a main effect of Event Type (F(2, 60) = 46.6, g < .001) such that a greater percentage of events were rated positive than negative and more were rated negative than neutral (means of 55.3, 31.9 and 12.6 respectively). Surprisingly, in this analysis, Study mood did not interact with Event Type. Although there are small differences in percentage of memories rated positive (59.2 and 51.3) and negative (30.1 and 34!0) across M and D occasions, these differences are not significant. Table 22. Percent of Event Types Generated as a Function of Mood. Event Gen Mood Type % Manic Pos 59.2 [17] Neut 10.7 Neg 30.1 Depressed Pos 51.3 [15] Neut 14.7 Neg 34.0 Mood Dependent Memory 66 The ratio of positive to negative events generated is another means of examining mood congruence. Although the ratio for manic Study sessions (3.1) appeared larger than the ratio for depressed Study sessions (2.1), the difference was not significant. Study mood did not produce typical M C M effects that have been demonstrated in undergraduate students' event generation performance (Eich et al, 1994). Unlike studies with undergraduates, however, most of these subjects completed the Study task under both moods. Performance of within-subject analysis seemed a prudent method of reducing error variance. Nine subjects had Study sessions completed in both moods. Subjects who completed more than one condition in the same Study mood had an average calculated for use in the current analysis. Within-subject A N O V A of Study mood on event generation revealed a main effect of Event Type (F(2, 16) = 31.0, g < .001) which, like the earlier analysis, indicated that generated events are more often rated positive than negative, and more often rated negative than neutral (means of 55.2, 32.0, and 12.8, respectively). The interaction between Study mood and Event Type remained non-significant (F(2, 16) = 1.59, g = .25). The difference between ratios of valenced events is smaller (M = 2.5 and D =1.7), but the greater power of within-subject comparison allows this difference to reach conventional significance (F(l, 8) = 5.28, p < .05). This effect is the only indication of mood congruence in event generation and, clearly, it is not the robust effect reported in previous studies. What factors may have diminished M C M effects in this study? Perhaps these subjects, having been diagnosed with Bipolar Disorder and having faced ever-changing moods, have learned to try to minimize mood effects on their judgments. Alternatively, subjects may be engaging in mood regulation. Parrott and Sabini (1990) have investigated mood biases in memory and judgment. They demonstrated that when subjects perform cognitive tasks under the influence of naturally occurring moods, mood incongruent biases can operate on memory tasks. Simply stated, they suggest that naturally occurring moods may lead to incongruent biases as subjects attempt to moderate even pleasant moods. Across a number of studies, subjects, divided into groups by pre-Mood Dependent Memory 67 existing differences in mood or on the basis of subtle mood manipulations in which they were unaware that their moods were being altered, showed mood incongruent biases when generating autobiographical memories. The concept that subjects may adopt mood regulation strategies is intriguing, especially in a population in which marked mood change is inevitable. In the present study, though subjects were grouped according to their own naturally occurring moods, there is no compelling evidence for either mood incongruent or mood congruent biases. Yet, it is conceivable that these biases could operate in opposition across different sessions. For example, one could hypothesize that on certain occasions, say when mood is rising or falling, mood regulation is attempted, leading to mood incongruent memory biases, while on other occasions, say when mood has stabilized to some degree, attempts at mood regulation are abandoned, leading to M C M effects. Of course, it is equally plausible for mood stability to lead to mood regulation attempts and mood change to lead to the abandonment of such attempts. The current investigation does not provide any means of evaluating these speculations, but does suggest that future investigators of autobiographical memory in Bipolar Disorder will need to consider the possibility of both mood congruent and mood incongruent biases. Inkblot and Picture Ratings Inkblots were rated on their aesthetic pleasantness (scale from -2 to +2) whereas pictures were rated for apparent complexity (scale from 0 to 3). Within-subject analysis of inkblot ( M and D means = 0.32 and 0.01), and picture ratings (M and D means = 1.78 and 1.64), did not differ significantly across Study moods. Latency of Letter Associations Time taken to generate 20 letter associates was compared for manic and depressed moods within-subject. The analyses revealed no differences across moods (F(l, 8) < 1, M and D means of 140.4 and 154.0 sec). Mood Dependent Memory 68 Summary of Study Session Results Although subjects clearly experienced strong and different moods across Study sessions, evidence for a mood congruent bias in their performance of Study session tasks is slim at best. One within-subject analysis showed an effect of Study moods on the relative distribution of positive and negative event generated, but ratings of inkblots and pictures, and latency on the letter association task was similar across moods. Strong mood congruent effects were predicted at study onset, particularly due to the intense nature of clinically significant moods. Naturally occurring moods have, on some occasions, produced mood incongruent memory bias, a fact that may play a role in the diminished mood congruent effects in this study. 3.4.3 RESULTS OF TEST SESSION TASKS Although the initial study design called for performance of within-subjects A N O V A to evaluate Study and Test mood effects on memory tasks, the small number (5) of subjects that completed all conditions made reliance upon this approach alone unsound. Rather, performance across Test session tasks will be analyzed in a number of ways. First, parametric (ANOVA) and non-parametric (difference between proportions test) analyses will be performed while treating conditions as though they are entirely between-subjects. Second, within-subject comparison of the effect of matching versus mismatching Test moods given a certain Study mood will be ascertained via parametric (dependent t-test) and non-parametric (Wilcoxon matched pairs test) analyses. Third, as planned, within-subject A N O V A will be calculated on those 5 subjects who completed all Study/Test mood combinations. Fourth, and finally, the general relationship between mood change and memory task performance will be determined by calculation of correlation coefficients. Though the particular pattern of significant effects will vary across these analyses, positive results should predominate for tasks whose performance genuinely depends upon mood. Mood Dependent Memory 69 Free Recall of Autobiographical Event Generation Probes. Subjects were asked to recall probe words or memories described during autobiographical event generation in the preceding Study session. Free recall was scored in two ways: first, as the percentage of the 10 probe words or events recalled; second, because five probe words served as targets in category production, as the percentage of recall among probe words that were untested via category production. Although explicit and implicit task performance is considered by some researchers to rely on independent processes (Dunn & Kirsner, 1988), this scoring ensured that these items were not contaminated by previous implicit testing. Free recall of autobiographical event generation probes is summarized across analyses, mood conditions, and both methods of scoring in Table 23. Visual inspection of means in the between-subjects analysis reveals a mood dependent pattern; however, A N O V A revealed no significant effects either in total free recall or in items untested in category production (Study, Test, and Interaction Fs (1, 28) < 1.28, all ps > .25). The standard deviations indicate a large degree of within-Table 23. Free Recall as a Function of Study/Test Mood % Moods Recall (SD) n Between Subjects Total M/M 34.4 (33.6) 9 Recall M/D 23.8 (20.7) 8 D/M 20.0 (29.8) 8 D/D 25.7 (17.2) 7 Not CP M/M 40.0 (36.1) 9 Tested M/D 27.5 (28.2) 8 D/M 17.5 (32.8) 8 D/D 25.7 (19.0) 7 Within-Subject Matched vs Mismatched Moods Total Mat 33.1 (17.1) 13 Recall Mis 17.7 Not CP Mat 36.9 (30.0) 13 Tested Mis 18.5 ANOVA Total M/M 32.0 (33.6) 5 Recall M/D 18.0 (20.7) D/M 14.0 (29.8) D/D 20.0 (17.2) Not CP M/M 36.0 (36.1) 5 Tested M/D 20.0 (28.2) D/M 12.0 (32.8) D/D 16.0 (19.0) Correlations Absolute Change in Mood n Pleas Arous PA NA • Recall 53 -.33** -.10 -.24* -.06 Not CP 53 _ yg** -.15 -.20 -.06 Note: ** p < .025, * p. < 10 Mood Dependent Memory 70 group variability which would require extensive between groups effects to reach significance. A non-parametric approach to this analysis compared the proportion of recalled items across matched and mismatched mood conditions. The difference between proportions test revealed trends for matched moods (n= 9 M / M , 7 D/D) to produce better recollection than mismatched moods (n= 8 M/D, 8 D/M) in both total free recall (z = 1.83, g < .07, means of 30.6 and 21.9) and items not tested in category production (z = 1.69, g < .10, means of 33.8 and 22.5). Direct comparison of the effect of matched versus mismatched moods within an individual is a more powerful test than the between-subjects approach. On thirteen occasions subjects contributed two mood combinations with a particular Study mood. One Test session was performed in that same mood, and one was performed in the opposing mood. Dependent t-tests revealed a significant advantage of matched over mismatched moods in the recall of event generation probes (total recall t(24) = 3.24, p < .01; items not CP tested t(24) = 2.22, p < .05). Wilcoxon matched pairs tests revealed significant (total recall, z = 2.67, g<.01) and marginal (items not CP tested, z = 1.87, g < .07) advantage to performance given matching moods. Within-subject A N O V A based on those five subjects who completed the study in its entirety revealed a marginal Study x Test mood interaction indicating M D M in total free recall (F(l, 4) = 6.67, g < .07), but not among items not CP tested (all gs > .20). Each previous analysis is based on various sets of selected conditions. When analyses are applied to only part of a dataset, an obvious concern is that selected conditions differ from other conditions on more than the selection variable. It is worth investigating whether the effects exist in the dataset as a whole or if they are restricted to those particular conditions selected for analysis. In this study, conditions were selected according to reported moods because mood dependent effects may require strong moods to exist (Eich, 1995). Nevertheless, the overall relationship between mood change and memory performance warrants evaluation. Reduction of between-subjects differences was achieved by standardizing task performance around each subject's overall mean (based on every condition regardless of mood). Mood Dependent Memory 71 Correlations were calculated between the resulting standardized scores and mood change across sessions (absolute distances between pleasure, arousal, P A and N A scores) for all 53 conditions. Pleasure change was negatively associated to free recall of event generation probes across scoring methods. That is, the greater the distance between pleasure ratings on Study and Test occasions, the fewer event generation probes were remembered. P A change was marginally negatively associated to total free recall of event generation probes. Importantly, this method of analysis is based on every valid condition contributed by every subject regardless of mood intensity or change in mood. It attests that these effects are not arbitrarily restricted to selected conditions but are generally true across the dataset as a whole. Taken together, the results of the six different analyses consistently indicate that M D M is evident in recollection of event generation probes in this population. Category Production Priming In another phase of testing, subjects generated exemplars for 10 different conceptual categories. Probes used for event generation in the immediately preceding Study session were targets in five of these categories. Base rate production of these targets was known through a number of unpublished investigations of implicit memory conducted in our lab. Priming was defined as the difference between percentage production of studied probe words and unstudied production of these same targets in the normative group. Over all valid conditions, priming averaged 6.4% that differs significantly from zero (t(52) =2.15, g < .025), indicating implicit memory for autobiographical event generation probes. Table 24. Category Production Priming as a Function of Study/Test Mood % Moods Priming (SD) n Between Subjects M/M 7.4 (18.2) 9 M/D -4.5 (23.9) 8 D/M 4.9 (26.8) 8 D/D -6.1 (10.9) 7 Within-Subject Matched vs Mismatched Moods Mat -0.1 (21.3) 13 Mis . -0.1 ANOVA M/M 2.1 (20.2) 5 M/D -5.6 (26.1) D/M -1.8 (15.8) D/D -6.4 ( 7.9) Correlations Absolute Change in Mood n Pleas Arous PA NA 53 .11 -.05 -.09 -.06 Mood Dependent Memory 72 Category production priming is presented across analyses and mood conditions in Table 24. In the between-subjects approach to analysis, though visual inspection across mood conditions suggested a Test Mood effect, no significant effects were revealed by A N O V A (Test effect F ( l , 28) = 2.36, p > . 10). The non-parametric test was not based on priming, rather, it compared the proportion of target items produced across matched and mismatched moods (% targets produced = 35.0 and 31.2 respectively). It was also calculated across manic and depressed moods at Test (% targets produced = 38.8 and 26.7). The test revealed no advantage in matched over mismatched moods (z = 0.54), but a trend indicating more items were produced in manic than depressed Test moods (z=1.71,p<.10). A striking feature of the category production data is that, though overall there appears to be significant priming (6.4%), no priming is evident across the 32 conditions selected on the basis of strong mood or mood change across sessions. Contrasting performance between those 32 selected (0.1%) and the 21 remaining conditions (14.9%), revealed that significantly greater priming was supported by the unselected conditions (F(l, 51) = 5.82, p_ < .025). The conditions selected due to strong subject mood or mood change appear to foster less category production priming than do other conditions; an effect that appears particularly true given a depressed mood during memory testing. Within-subject dependent t-test (t(24) = .01) and Wilcoxon matched pairs test (z = .04) both compare, within a subject, the effect of Test moods that do versus do not match a given Study mood, and revealed no significant or marginal effects in category production priming. Similarly, within-subject A N O V A based on those five subjects that completed all conditions revealed no mood effects in category production priming (all Fs < 1.1). Priming was standardized in precisely the same manner as was free recall performance. Specifically, each subject's performance was standardized around his or her own mean drawn from all conditions regardless of mood. The standardized scores were not significantly correlated with absolute mood change across sessions. Mood Dependent Memory 73 Though category production priming was the most promising implicit memory task, across analyses, the current data clearly do not support M D M . The only effect to approach significance was revealed via non-parametric analysis of targets produced indicating that testing during depression lead to production of fewer category targets than did testing during mania. Inkblot Recognition During testing for recognition of inkblots, subjects were to select previously rated inkblots from 5 extremely similar distractors and rate their confidence in those choices. Greater confidence ratings reflected a greater percentage of identified targets (means for increasing confidence ratings: 26%, 43%, 48%, and 57%). Inkblots were correctly recognized above chance levels (t(52) = 7.10, p < .001; means of 43.4% and 16.7% respectively). Mean inkblot recognition across analyses and conditions is presented in Table 25. Inspection of mean inkblot recognition performance across mood conditions appeared, like recall of event generation probes, to support a mood dependent pattern, but neither between subject A N O V A (all Fs (1, 28) < 1:5, all ps > .20) nor non-parametric difference between proportions test (z = 1.27, p > .15) revealed significant mood dependent effects. Only when subjects reported feeling "Not at all confident" in their recognition choice did matching versus mismatching mood means differ by more than 10% (matched = 36.1% and mismatched = 20.0%) but t-tests revealed no significant differences at any confidence level. Table 25. Inkblot Recognition as a Function of Study/Test Mood % Moods Recog (SD) n Between Subjects M/M 41.7 (21.7) 9 M/D 34.4 (26.5) 8 D/M 34.4 (37.6) 8 D/D 50.0 (14.4) 7 Within-Subject Matched vs Mismatched Moods Mat 48.1 (29.1) 13 Mis 28.8 ANOVA M/M 40.0 (22.4) 5 M/D 40.0 (28.5) D/M 30.0 (32.6) D/D 50.0 (17.7) Correlations Absolute Change in Mood n Pleas Arous PA NA 53 -.16 -.08 -.35* -.05 Note: * p_ < .025 Mood Dependent Memory 74 When the effects of Study and Test moods were evaluated within-subject, both dependent t (t(24) = 2.38, g < .05) and Wilcoxon tests (z = 2.00, p < .05) revealed a significant advantage in performance when Study and Test moods matched. Yet, the overall within-subject A N O V A revealed no significant effect of mood on inkblot recognition (all Fs < 2.7, all gs > . 15). Correlations between mood change scores and standardized inkblot recognition performance across all valid conditions revealed a significant negative relationship between P A change and inkblot recognition. These results are at least moderate indication that inkblot recognition performance is influenced in a mood dependent manner. Picture Fragment Priming During Test sessions, subjects saw as many as 8 picture fragments for each of 8 objects. Their task was to name each object as quickly as possible. Four objects had been rated in the prior Study session; the remaining four were new. Priming was calculated by dividing the advantage in number of cards needed to identify old items by the cards required to identify new items (% Priming = (# cards for new items - # cards for old items)/ # cards for new items). Average priming on picture fragment completion was 12.3% which differs significantly from zero (t(50) = 3.70, g < .01), indicating imglicit memory for gictures. Picture fragment comgletion griming is presented across analyses and mood conditions in Table 26. Like category production performance, by visual inspection the largest differences were between conditions of manic and depressed Test moods, but no effects were significant by between-subjects A N O V A (all Fs (1, 26) < 1). As the difference between Table 26. Picture Fragment Priming as a Function of Study/Test Mood % Moods Priming (SD) n Between Subjects M/M 11.9 M/D 6.1 D/M 10.2 D/D 3.1 (18.0) 9 (35.1) 8 (33.7) 7 (25.0) 6 Within-Subject Matched vs Mismatched Moods Mat 7.2 (43.3) 12 Mis 0.9 ANOVA M/M 4.1 (17.3) 4 M/D -9.7 (34.6) D/M -5.2 (34.1) D/D -1.6 (30.8) Correlations Absolute Change in Mood n Pleas Arous PA NA 51 -.10 -.15 -.15 -.01 Mood Dependent Memory 75 proportions test could not be applied to picture fragment completion (each object was eventually identified by each subject), the Mann-Whitney U test was performed in its place, revealing no mood effects in picture fragment completion (z = 0.71). Evaluation of the effect of matched versus mismatched moods via within-subjects comparison (dependent t(22) = 0.26, and Wilcoxon z = 0.16), via within-subject A N O V A (all Fs < 1) and via correlations between standardized performance and mood change were all consistent in indicating that mood plays little role in performance of picture fragment completion. Explicit Memory for Letter Associates During one phase of testing, subjects were given a letter and were asked to spend 3-5 min. recalling their prior associations from the preceding Study session. Recall of letter associates is presented across analyses and mood conditions in Table 27. Unlike the other explicit memory tasks in this study, visual inspection of means in the between-subjects analyses indicates little effect of mood in recollection of letter associates. Both A N O V A (all Fs < 1) and difference between proportions test (z = 0.25) confirm this impression. Similarly, within-subject comparison of matched versus mismatched moods (dependent t(22) = 0.08, Wilcoxon z = 0.06) and overall correlations between mood change and standardized performance indicated no mood effects in recall of letter associates. However, within-subject A N O V A did reveal a marginal interaction that indicates better recall of letter associates when subjects Test moods did not match their mood at Study (F(l, 3) = 8.44, p < .07). Table 27. Recall of Letter Associates as a Function of Study/Test Mood % Moods Recall (SD) n Between Subjects M/M 28.3 (15.2) 9 M/D 27.5 (11.9) 8 D/M 29.3 (18.4) 7 D/D 24.3 ( 9.3) 7 Within-Subject Matched vs Mismatched Moods Mat 25.0 (17.2) 12 Mis 24.6 ANOVA M/M 20.2 (14.1) 4 M/D 27.5 ( 9.6) D/M 31.2 (15.5) D/D 25.0 (12.2) Correlations • Absolute Change in Mood n Pleas Arous PA NA 52 .01 -.07 -.03 -.09 Mood Dependent Memory 76 Considering the analyses overall, a preponderance of the results suggest, somewhat surprisingly, that mood condition had little effect on explicit recollection of letter associates. Implicit Memory for Letter Associates In another phase of testing, subjects produced 20 associates, as quickly as possible, for each of three letters, two new, and one that had been performed in the previous Study session. Overall, duplication of letter associations averaged 26.2% that differs significantly from zero (t(51) = 15.0, p < .001), indicating implicit memory for those previous associates. Re-performance of letter association is represented across analyses and mood conditions in Table 28. Between-subjects analysis revealed no main effects (both Study and Test mood Fs (1, 27) < 1.25, p > .25) but a significant interaction between Study and Test moods (F(l, 27) = 5.31, p < .03). Surprisingly, significantly more words were duplicated when subjects reported dissimilar Study and Test moods. The difference between proportions test also supports this pattern (z = 3.06, p < .01). Visual examination of means suggests that D/D may differ from the other conditions. However, Tukey tests revealed no significant pairwise comparisons between means. Remaining analyses also reveal no significant mood effects in the performance of letter re-association (dependent t(22) = 1.5, g > .15; Wilcoxon test, z = 1.5, g > .10; within-subject A N O V A , all Fs (1, 3) < 4.6, all gs > . 10; correlations all gs > . 10), suggesting that, on the whole, M D M is not evident in implicit letter re-association. Table 28. Re-Association to Letter as a Function of Study/Test Mood % Moods Re-Assoc (SD) n Between Subjects M / M 26.7 (15.4) 9 M/D 31.9 ( 7.5) 8 D/M 32.1 (12.9) 7 D/D 17.1 (10.7) 7 Within-Subject Matched vs Mismatched Moods Mat 22.9 (18.3) 12 Mis 30.8 ANOVA M / M 25.0 (15.8) 4 M/D 30.0 ( 7.1) D/M 35.0 (10.0) D/D 11.2 (10.3) Correlations Absolute Change in Mood n Pleas Arous PA NA 52 .12 .17 .16 .01 Mood Dependent Memory 77 Notably, subjects produced similar proportions of previous associates in both implicit and explicit methods of testing. Though instructions differed between tasks, they may not have had the intended effect. Asked to re-associate, subjects generated 5.4 old items in a list of 20; asked to explicitly recall the prior associates, subjects listed 5.9 old items in a list of 13.5, producing a very high number of false alarms (7.6) for a recall task. Although the instructions to subjects clearly differed between tasks, these numbers suggest that subjects may have approached both tasks similarly. During the explicit recall task, subjects would often indicate that they could not remember any more items. If 3 min. had not yet passed they were asked to keep trying, without guessing, until time was up. Inevitably, subjects would produce more items but it is not clear if they simply began to use the letter as a cue for re-associating rather than spending that time trying to remember items. In retrospect, the strict adherence to a 3 min. limit may have pressured subjects to begin using re-association as a strategy during recall. Summary of Mood Dependent Memory Effects Table 29. Summary of MDM Effects across Analyses and Tasks Method of Analysis Event Generation Free Cat Recall Prod Visual Stimuli Inkblot Picture Recog Priming Letter Associates Re-Recall Assoc Between Subjects n= 30 to 32 ANOVA Diff in Proportions ns SIG ns ns ns ns ns ns ns ns SIG SIG Within-Subject n = 12 or 13 Dependent t Mat vs Mis Wilcoxon Matched Pairs SIG SIG ns ns SIG SIG ns ns ns ns ns ns Within-Subject n = 4 or 5 ANOVA MAR ns ns ns MAR ns Correlations n=51 to 53 Mood Change (1/4 measures) SIG ns SIG ns ns ns Note: SIG: p. < .05, MAR: g < 10, ns: not significant, xxx: reversed effect Mood Dependent Memory 78 The results of all analyses are summarized in Table 29. Mood dependent effects are evident when memory of event generation probes is tested via free recall, but not when it is tested via category production priming. Memory for visual information may be dependent upon mood when tested by inkblot recognition but is clearly not when it is tested by picture fragment completion priming. Regarding memory for letter associations, though both recall of associates and re-association showed at least one apparent reversed M D M effect, most analyses revealed no mood effects in their performance. 3.4.4 SUMMARY OF RESULTS Ten subjects contributed 32 conditions which represented occasions of depression, mania/hypomania or change between these states. These selected conditions differed substantially on average Affect Grid and PANAS measures when compared by mood. Surprisingly, however, these relatively strong moods did not produce strong mood congruent effects in the event generation task. In particular, only within-subject differences revealed M C M effects in the relative proportion of positive and negative events generated. No mood congruent effects were noted in the ratings of inkblots or pictures nor in time required to produce letter associates. Mood dependent effects in memory task performance were evaluated by way of six different analyses. These approaches almost unanimously agree in supporting mood dependent effects in recollection of event generation probes, but no effects in implicit category production memory for a subset of the same probes. Memory for visual information was unaffected by mood condition when tested by picture fragment completion, but there was some evidence of M D M in inkblot recognition performance. Interestingly, exploration of significant effects in letter association tasks revealed an effect opposite to mood dependence. On the whole, though, analyses of associate recall and re-association support no significant mood effects. Overall, mood dependent effects appeared restricted to explicit tasks, specifically to free recall of event generation probes and inkblot recognition. Mood Dependent Memory 79 3.5 DISCUSSION The primary purpose of the current investigation was to determine the reality of M D M effects among subjects who had experienced no mood manipulation but, rather, reported real, rapid, and involuntary changes in their endogenous mood; namely, persons diagnosed with Bipolar Disorder. The study confirmed a reliable decline in subjects' later ability to recall when their mood at Test had changed, replicating Weingartner et al. (1977). Mood dependence appears reliable in this population, validating it as a real world phenomenon. During debriefing, most participants discussed memory problems that they considered to be a consequence of depression. Many were surprised to find that the current study focussed on times of mood change as possibly even more disrupting to memory than periods of depression. The study demonstrated that MDM was strongest among recall of personal events, unfortunately, precisely the aspect of the memory disruption that participants found most distressing. If patients diagnosed with Bipolar Disorder are unaware of the prospect of M D M effects, then they are certainly unaware that the M D M aspect of their memory problems can potentially be countered. The evidence suggests that, though information may vary in sensitivity to mood change, provision of cues during memory retrieval may reduce MDM effects. Further, there is some evidence that re-instatement of mental context, by contemplation of the original learning circumstances, may overcome place-dependent memory effects (Smith, 1984) and this tactic may be of some practical use as a simple strategy to overcome mood dependent effects in Bipolar Disorder. Although this study has provided a wealth of information, clinical investigations do pose certain problems. The fundamental problem entailed here was identification and recruitment of sufficient numbers of subjects. Even once subjects were identified and agreed to participate, the amount of time and number of sessions made it difficult for subjects to complete the study. Outpatients occasionally did not appear for appointments when their moods became severe; some were hospitalized, others were unable to leave their beds. Restricting the study to in-patients Mood Dependent Memory 80 would have eliminated that problem but would have further limited the number of eligible subjects. Even among those subjects who kept all their appointments, this study required proficient prediction of when the next mood change would occur. Though other groups are unlikely to experience the severity of mood change associated with Bipolar Disorder, future research may nevertheless benefit by focusing on groups who experience repeated and predictable mood change. Thus, patients with Seasonal Affective Disorder or Unipolar Affective Disorder with diurnal variation in mood may be populations for future investigation. The secondary purpose of the study was to evaluate sensitivity to M D M effects across various implicit and explicit memory tasks in a group whose moods are not diminished by task performance. Explicit free recall of items that had been studied via autobiographical event generation showed the most consistent evidence of M D M revealing significant effects in 5 of 6 analyses. Clearly, subjects remembered more items when Study and Test moods matched. Importantly, these results validate three prior studies involving mood manipulation in undergraduate students (Eich et al. 1994). Investigation of category production, the most promising implicit memory task, was undertaken to replicate a M D M effect that had been shown on one prior occasion. Though event generation paired with later free recall showed M D M , no such effects were evident in their category production priming. Although the task appeared to conform to the requirements of the "do-it-yourself hypothesis, that is" subject reliance upon internally guided processing at both Study and Test, mood had little effect on category production priming. Mood dependent effects were evaluated via both implicit and explicit memory tasks for visual stimuli. While implicit picture fragment completion was uninfluenced by moods, explicit inkblot recognition was sensitive to mood change, and three of six analyses supported M D M effects. Although both tasks relied upon study of experimenter-presented materials, the amorphous and ambiguous nature of the inkblots may have been more amenable to establishing a connection between mood and those materials. A different means of forming such a connection Mood Dependent Memory 81 with visual stimuli in the future would be instructing subjects to generate complex visual mental images in the study session for later recall at test1. In the inkblot recognition task, five extremely similar distractors at test may have meant that target items would not readily stand out, diminishing the inherent potential of having a direct copy of the original item as a retrieval cue. Rather, subjects may have been forced to search for other relevant information to make their choice, allowing mood to influence retrieval processing and resulting in the M D M effects. Letter association with subsequent implicit and explicit memory testing was included to ascertain how mood exerts its influence on memory. The task was based upon free association to cue words that had demonstrated M D M (Weingartner et al., 1977). The letter cues were used to ensure that associates would occur for only one cue and that they could be specifically identified for implicit or explicit memory testing. M D M in explicit but not implicit tasks would indicate that mood was a cue used in subjects' conscious retrieval strategies. M D M in both explicit and implicit tasks would indicate that mood became incorporated into the memory trace, and partial re-presentation in the form of a matching mood allowed broader redintegration of the original event. Unfortunately, neither outcome occurred. Overall analyses indicate that mood effects are not reliable in letter association, leaving unanswered the questions posed regarding the connection between mood and memory. What factors may have decreased M D M effects in letter association tasks? Mood congruent effects can be diminished by presentation of to-be-remembered materials that have their own inherent organization (Fiedler, Pampe & Scherfe, 1986). Could the organization imposed by the letter cue be sufficient to negate any M D M effects in this task? Specification of the initial letter did restrict subjects' responses to a set of items, thereby imposing some organization on the materials. However, the task imposed little or no conceptual organization on item selection. Free 1 Note: As suggested by Dr. P. Suedfeld (Apr. 1997). Mood Dependent Memory 82 association to probe words (Weingartner et al. 1977) is likely to have resulted in greater conceptual similarity across items than the current task, yet it revealed significant M D M . Although the letter provided little conceptual organization, it did provide a production rule that would apply to every item on the list. As noted in the results section, careful application of the 3 min. requirement for letter associate recall may have encouraged subjects to perform at least part of the explicit recall task as implicit re-association, that is, by using the letter as a cue for producing items regardless of their occurrence on the original list. Given that the implicit task showed no significant effects, use of this strategy may have diminished M D M effects in recall. Considered individually, neither of these possibilities appear adequate to account for the lack of M D M in this explicit task. Yet, in combination, these features may be sufficient to curtail M D M . Although letter association, like free association, required subjects to rely upon internal decisions and resources to select some items but not others during task performance, M D M was not evident in letter association performance. The impetus to investigate mood dependence in implicit tasks was, in part, as a counterpoint to arguments that demand characteristics may be responsible for M D M effects in subjects who undertake a mood manipulation. If M D M was evident when subjects were unaware that memory was being tested, demand could not be offered as a reasonable alternative. Although M D M was not apparent on implicit tasks, demand remains an unlikely explanation of the effects in this study. As became apparent during debriefing, subjects expected to see an effect of depressed mood on performance, an effect that appeared on some tasks but was significant only once. Further, most subjects were unaware of the possibility of M D M , yet it was the most consistent effect of mood on memory performance, playing a role in both free recall and inkblot recognition. The "do-it-yourself hypothesis (Eich, 1995) has allowed substantial progress in our understanding of M D M effects. Prior to its introduction, reports of positive instances of M D M and failures to replicate seemed to fall in no clear pattern. As a framework, it has provided an organization to those prior results and has allowed predictions about paths worth investigating. Previously, evidence of M D M seemed clear and consistent given that subjects played an active Mood Dependent Memory 83 role in both generating the targets and producing the cues required during retrieval of the events (Eich et al. 1994; Eich & Metcalfe, 1989; Beck & McBee, 1995) However, this study raises questions about the necessity and sufficiency of these pre-conditions. Category production and letter re-association both appear to conform to the pre-conditions as specified by the "do-it-yourself hypothesis, yet results of neither showed significant M D M effects. It is becoming clear that, although M D M may eventually be replicated in implicit memory, the effect is more consistently obtained via explicit memory tasks. Mood dependence was demonstrated on two explicit tasks. Free recall of subject generated personal events met both criteria of the "do-it-yourself hypothesis, while recognition of inkblots met none. Upon first reflection, the tasks appear to have little in common other than they both rendered M D M results. But speculation about the attributes that these two tasks share is essential to make further progress in our understanding of the conditions that lead to M D M . Personal events and inkblots seem ripe for description, interpretation and evaluation in emotional terms. Prior tasks, that have proven insensitive to M D M effects, tended to require subjects to learn a list of unrelated neutral nouns while co-incidentally experiencing an unrelated mood. Bower (1987) hypothesized that if subjects did not experience a meaningful connection between the mood and materials, say via causal belongingness, they were unlikely to be connected in memory. If mood was not connected to the materials, later demonstration of M D M would be impossible. Both personal events and inkblots seem amenable to a connection with emotions, and differ from other materials in their inherent "emotionality". Unfortunately, identification of particular details that create this emotionality seems doubtful. Another aspect that these tasks share is that their performance requires more than the simple contemplation of previously acquired concepts. In tasks that do not demonstrate M D M , subjects typically read, hear aloud, or see pictures referring to unrelated nouns and rate some attribute like pleasantness or complexity. Thus task performance requires subjects to simply refer to pre-existing conceptual or semantic knowledge. Generation of personal events and rating of inkblots represent more than simple instantiation of a concept. Mood Dependent Memory 84 But, what more do they represent? I believe they create mental representations that more clearly refer to distinct episodes. Tulving (1972) introduced the distinction between episodic and semantic memory. Episodic memory is comprised of events that occured at a particular time, in a particular context. Semantic memory consists of representations of concepts and their associations. Though Tulving considered the terms to refer to different underlying memory systems, others have argued that they may represent ends of a continuum (Kinstch, 1974). Hintzman (1986; 1988) has developed a mathematical model of memory in which a single process can produce results that mirror the episodic-semantic distinction. In Hintzman's multiple-trace memory model, known as MTNERVA 2, each consciously experienced event gives rise to its own memory trace. Memory for prior episodes is represented in individual events, whereas semantic memory is derived from the pool of individual instances. In his model, primitive components, ranging from emotional tone to abstract relations are the features of each event. Phenomena that are repeated will have these features represented in memory over and over again, in various configurations according to the particular context of each instantiation. Instances representing the same concept will share associations to, say abstract relations, but will not share a consistent connection to any particular context or mood. Information will become decontextualized as it is experienced in different emotional states, different physical modalities or contexts. Rather than representing another instantiation of a concept, the event generation task requires detailed description and ratings of personal experiences in response to a word like DESK. In Hinztman's model, performance of the task will lay down a new instance in reference to the word DESK. However, as the task requires more than reference to the simple connotation of the word, it will have features that distinguish it from those types of instances. Though the events may have been described before, memory for these instances should not be decontextualized to the same extent as simple reference to the meaning of the word DESK. Thus, contextual features of such an event may prove useful as retrieval cues. Mood Dependent Memory 85 The inkblots presented in the current investigation are unique and were designed not to have a clear conceptual referent. Though subjects may attempt to relate inkblots to existing concepts, doing so surely differs from, say, seeing a picture or reading a word. This new material may be especially sensitive to contextual change, as few other similar instances would have been experienced before, making it unlikely that the context cues could be overloaded during retrieval by numerous similar instances. Ultimately only new research will decide if either emotionality of materials or emphasis on episodic elements adds to our understanding of M D M effects. "Emotionality" of materials seems somewhat less satisfactory for making research decisions. It does not easily translate into a set of specific attributes of materials or tasks, that should reflect an emotional nature, making a priori identification of greater emotionality difficult. Creating an emphasis on episodic elements during a study task, on the other hand, may be more useful. Smith (1994) suggests that mood and other contextual features may be more important when new information is being acquired. In future M D M research, subjects could be asked to learn the meaning of very low frequency words, or of words in a different language, events that do not have prior decontextualized semantic memory representations. These tasks should encourage contextual features like mood to be a useful part of the memory trace during retrieval. The "do-it-yourself hypothesis has suggested generation of materials and of retrieval cues as one method of enhancing the connection in memory between mood and materials. Though these suggestions have often led to M D M , their use does not guarantee its occurrence. Both recollection of personal events and recognition of inkblots appeared mood dependent and may share that they are complex, emotional, and open to interpretation. They also differ from tasks, like letter association and picture rating, whose performance relies upon instantiation of previously acquired simple concepts. This study demonstrates that exploration of mood dependent effects may be enhanced by use of materials that are unique, open to interpretation and mood. Mood Dependent Memory 86 3.5.1 SUMMARY The experiment successfully demonstrated that strong mood change can lead to memory deficits and validates MDM effects as real world phenomena. Mood dependent effects were evident in the recollection of both personal events and inkblots in subjects diagnosed with Bipolar disorder. The pattern of results across memory tasks indicates that dependent effects are more likely in explicit than implicit tasks, and though generation of materials at Study and reliance of internal cues for Test performance has led to significant results in the past, their use does not guarantee dependent recollection. Mood dependent effects may also be encouraged by use of emotional materials, or by using memory tasks that create mental representations that are better characterized by features of episodic than semantic memory. Mood Dependent Memory 87 CHAPTER 4. GENERAL DISCUSSION This dissertation consists of two alternative approaches to investigating the puzzle of mood dependent memory. The first study examined four manipulated moods that varied on the fundamental dimensions of emotion. Results revealed that there are indeed differences in the size of mood dependent effects according to the relationship between the two moods, bearing out Bower's (1981) prediction. Change from one mood to another was not always equivalent, that is, dependent effects were more striking under certain mood combinations than others. Although the pattern of dependent effects was not entirely predicted from any single theory of emotion, integration of two approaches did provide a reasonable explanation of current results, and indicates that ideas drawn from emotion theory provided a means of formulating testable predictions about the psychological distinctiveness between moods. This integrated model would point to future exploration of dependent effects in moods that vary in orientation of attention as determined by action tendencies. Specific comparisons between certain moods, like anger and anxiety, would allow comprehensive testing of the model. Investigations of this type may not only provide insight into the nature of the relationship between mood and memory, but the cross-fertilization of ideas may also benefit research into emotion. The second study examined mood dependent effects in Bipolar Disorder. Its results validate M D M effects as phenomena that can significantly impair explicit memory performance in subjects who experience unmanipulated changes in mood. The research indicates that in the real world, persons experiencing changing moods may be facing a disruption in memory on the basis of mood change; a memory disruption that has the potential of being reduced. Future attention to the concept of mental reinstatement may be of practical use to this and other clinical groups experiencing substantial changes in mood. A secondary reason for undertaking this study was to evaluate dependent memory effects across numerous tasks in subjects whose moods would be unaffected by task performance. Although most tasks had shown some promise in previous results with undergraduates, only free Mood Dependent Memory 88 recall and inkblot recognition appeared to support dependent memory effects. Thus, this study demonstrates that investigation of cognitive effects in relevant clinical groups can provide a means of validating tasks developed in the laboratory. Prior investigations of M D M have relied almost exclusively on the induction of happy and sad moods. Decisions about the methods and materials that should encourage M D M have, therefore, been based on research with these two moods. Investigation of M D M in Bipolar Disorder patients, a group that experiences natural mood changes, was fairly consistent with prior studies, replicating M D M in free recall and pointing to inkblot recognition as a task for future research. However, the study of M D M across four different moods showed that the pattern of results seen in delighted and sad moods may not hold in other combinations. Results obtained in delighted-sad, anxious-sad, and calm-sad mood combinations revealed M D M in recall but not recognition whereas delighted-calm moods showed stronger M D M in recognition than recall. The anxious-calm mood combination revealed M D M in both recall and recognition while the delighted-anxious mood combination revealed no M D M effects. Of course, this pattern of performance remains to be replicated. But, the current study suggests that sensitivity to M D M effects varies across both mood combination and tasks. One interpretation of this complex pattern of results could be that M D M is a fragile and unreliable effect, appearing in certain conditions but not others without notable reason (Bower & Mayer, 1989). Although the point is well-taken given the past history of M D M research, its recent replicability in happy and sad subjects (Eich & Metcalfe, 1989; Eich et al., 1994; Beck & McBee, 1995) indicates that current understanding of the effect has allowed it to be controlled. Further, the current clinical study shows considerable overlap with prior laboratory investigations of manipulated mood in undergraduates. If M D M is simply unreliable, this consistency would be unlikely. The current confirmation of M D M in Bipolar disorder, makes it compelling to continue to pursue understanding of the effect. Replication of M D M in Bipolar disorder suggests that, while Mood Dependent Memory 89 researchers may find it difficult to examine in the lab and while progress has been irregular, the effect is nonetheless an authentic phenomemon in the lives of clinical patients. On this basis alone, the continued investigation of M D M is warranted. Though mood changes may be inevitable in the lives of clinical patients, the M D M disruption has reasonable potential for being reduced. In fact, the complex pattern reported in the current investigation could be seen as reason for optimism in this regard. As M D M effects are not pervasive across all types of materials and memory requirements, the possibility of finding strategies that clinical patients could use to counter the M D M effects remains open. On these practical grounds, researchers should continue to strive for greater understanding of M D M rather than reject the area as unyielding. A more positive view is that a complex pattern of results is to be expected given the nature of emotion. Emotions are complex phenomena subtending visceral and cognitive responses. They have provided the basis of volumes describing the qualities that distinguish one from the other and have inspired hot debate among theorists devoted to elucidating their fundamental features. Given this complexity it seems clear that previous cognitive approaches to the relationship between emotion and memory have been based on over-simplified views of emotions. Typically, in cognitive research, emotional states have been viewed as functionally equivalent to provision of a simple cue during the study of target material. Given this assumed simplicity, it is understandable that cognitive researchers viewed the pattern of M D M results as excessively variable. Yet, recognition that emotions act not only as cues but can fundamentally alter the way that subjects deal with information may be necessary for further progress to be made. Taken together, the results of the current studies suggests that more comprehensive understanding of the complex relationships between moods, methods, and materials, may require laboratory examination of a much wider range of tasks and mood combinations. Furthermore, once a pattern of performance can be established in the lab, validation studies in other clinical groups may be profitable. Given significant mood dependent recall and recognition performance in subjects assigned to anxious-calm moods, a particularly interesting future possibility would be Mood Dependent Memory 90 examination of M D M in patients with Anxiety Disorder. Investigation of a range of both implicit and explicit tasks may reveal a different pattern of M D M performance in this group compared to the current findings in Bipolar Affective Disorder. Both investigations provide new information about the relationship between mood and memory. Alternative approaches—emotion theory on the one hand and clinical groups on the other—granted greater insight than could be gained from within cognitive psychology alone. Future complementary use of cognitive psychology with appropriate outside ideas and populations may continue to enhance our understanding of the relations between emotion and cognition, by recommending possibilities that no solitary approach would allow. Mood Dependent Memory 91 References Bartlett, J. C , Burleson, G., & Santrock, J. W. (1982). Emotional mood and memory in young children. Journal of Experimental Child Psychology, 34, 59-76. Bartlett, J. C , & Santrock, J. W. (1979). Affect-dependent episodic memory in young children. Child Development. 50, 513-518. Beck, A., T. (1967). Depression: Clinical, experimental, and theoretical aspects. New York: Harper & Row. Beck, R. C , & McBee, W. (1995). Mood-dependent memory for generated and repeated words: Replication and extension. Cognition and Emotion. 9, 289-307. Blaney, P. H. (1986). Affect and memory: A review. Psychological Bulletin. 99, 229-246. Bower, G. H . (1981). Mood and memory. American Psychologist. 36, 129-149. Bower, G. H . (1987). Commentary on mood and memory. Behaviour Research and Therapy. 25, 443-455. Bower, G. H. (1992). How might emotions affect learning? In. S.A. Christianson (Ed.), The handbook of emotion and memory: Research and Theory (pp. 3-31). Hillsdale, NJ: Erlbaum. Bower, G. H. , Gilligan, S. G., & Monteiro, K. P. (1981). Selectivity of learning caused by affective states. Journal of Experimental Psychology: General, 110. 451-473. Bower, G. H. , & Mayer, J. D. (1985). Failure to replicate mood-dependent retrieval. Bulletin of the Psychonomic Society. 23, 39-42. Bower, G. H. , & Mayer, J. D. (1989). In search of mood-dependent retrieval. Journal of Social Behavior and Personality. 4, 121-156. Bower, G. H. , Monteiro, K. P., & Gilligan, S. G. (1978). Emotional mood as a context for learning and recall. Journal of Verbal Learning and Verbal Behavior, 17, 573-585. Brown, W. P., & Ure, D. M . J. (1969). Five rated characteristics of 650 word association stimuli. British Journal of Psychology. 60, 233-249. Buchwald, A. M . , Strack, S., & Coyne, J. C. (1981). Demand characteristics and the Velten mood induction procedure. Journal of Consulting and Clinical Psychology. 49, 478-479. Chartier, G. M . , & Ranieri, D. J. (1989). Comparison of two mood induction procedures. Cognitive Therapy and Research, 13, 275-282. Mood Dependent Memory 92 Clark, D. M . (1983). On the induction of depressed mood in the laboratory: Evaluation and comparison of the Velten and musical procedures. Advances in Behavior Research and Therapy. 5, 27-49. Compendium of Pharmaceuticals and Specialities (Canada). (1996). Toronto, Ontario: Canadian Pharmaceutical Association. De Rivera, J. (1977). A structural theory of the emotions. New York: International Universities Press. Dunn, J. C , & Kirsner, K. (1988). Discovering functionally independent mental processes: The principle of reversed association. Psychological Review, 95, 91-101. Eich, E. (1995). Searching for mood dependent memory. Psychological Science, 6, 67-75. Eich, E., Macaulay, D., & Ryan, L. (1994). Mood dependent memory for events of the personal past. Journal of Experimental Psychology: General. 123, 201-215. Eich, E., & Metcalfe, J. (1989). Mood dependent memory for internal versus external events. Journal of Experiment Psychology: Learning. Memory, and Cognition, 15, 443-455. Ellis, H . C , & Ashbrook, P. W. (1989). The "state" of mood and memory research: A selective review. Journal of Social Behavior and Personality. 4, 1-21. Evans, S. H. , & Anastasio, E. J. (1968). Misuse of analysis of covariance when treatement effect and covariate are confounded. Psychological Bulletin. 69, 225-234. Fiedler, K., Pampe, H. , & Scherfe, U . (1986). Mood and memory for tightly organized social information. European Journal of Social Psychology. 16, 149-164. Forgas, J. P. & Bower, G. H. (1987). Mood effects on person-perception judgments, Journal of Personality and Social Psychology. 53, 53-60. Forgas, J. P., Bower, G. H. , & Moylan, S. J. (1990). Praise or blame? Affective influences on attributions for achievement. Journal of Personality and Social Psychology. 59, 809-819. Frijda, N . H . (1986). The emotions. Cambridge, England: Cambridge University Press. Graf, P., Mandler, G., & Hayden, M . (1982). Simulating amnesic symptoms in normal subjects. Science. 218. 1243-1244. Graf, P., & Ryan, L. (1990). Transfer appropriate processing for implicit and explicit memory. Journal of Experimental Psychology: Learning. Memory, and Cognition. 16, 978-992. Mood Dependent Memory 93 Graf, P., & Schacter, D. (1985). Implicit and explicit memory for new associations in normal and amnesic subjects. Journal of Experimental Psychology: Learning. Memory, and Cognition. 11, 501-518. Haaga, D. A. (1989). Mood state-dependent retention using identical or non-identical inductions at learning and recall. British Journal of Clinical Psychology. 28, 75-83. Hasher, L. , Rose, K. C , Zacks, R., Sanft, H. , & Doren, B. (1985). Mood, recall and selectivity effects in normal college students. Journal of Experimental Psychology: General. 114. 104-118. Hintzman, D. L. (1986). "Schema abstraction" in a multiple-memory trace model. Psychological Review. 93, 411-428. Hintzman, D. L. (1988). Judgements of frequency and recognition memory in a multiple-trace memory model. Psychological Review. 95, 528-551. Horowitz, L. M . , & Prytulak, L. S. (1969). Redintegrative memory. Psychological Review. 84, 519-531. Hull, C. L. (1943). Principles of Behavior. New York: Appleton-Century-Crofts. Ingram, R. E. (1984). Information processing and feedback: Effects of mood and information favorability on the cognitive processing of personally relevant information. Cognitive Therapy and Research. 8, 371-386. Isen, A. M . , Shalker, T., Clark, M . , & Karp, L. (1978). Positive affect, accessibility of material in memory, and behavior: A cognitive loop? Journal of Personality and Social Psychology, 36, 1-12. Jacoby, L. L. , & Dallas, M . (1981). On the relationship between autobiographical memory and perceptual learning. Journal of Experimental Psychology: General. 110. 306-340. Johnson, T. L. , & Klinger, E. (1988). A nonhypnotic failure to replicate mood dependent recall. Bulletin of the Psychonomic Society. 26, 191-194. Johnson-Laird, P. N . , & Oatley, K. (1988). Are there only two primitive emotions? A reply to Frijda. Cognition and Emotion. 2, 89-93. Kihlstrom, J. F. (1989). On what does mood dependent memory depend? Journal of Social Behavior and Personality, 4, 23-32. Kinstch, W. (1974). The representation of meaning in memory. Hillsdale, NJ: Erlbaum. Mood Dependent Memory 94 Lazarus, R. S. (1991). Progress on a cognitive-motivational-relational theory of emotion. American Psychologist 46, 819-834. Leight, K. A. , & Ellis, H . C. (1981). Emotional mood states, strategies, and state-dependency in memory. Journal of Verbal Learning and Verbal Behavior. 20, 251-275. Macaulay, D., Ryan, L. , & Eich, E. (1993). Mood dependence in implicit and explicit memory. In P. Graf & M . E. J. Masson (Eds.), Implicit memory: New directions in cognition, development, and neuropsychology (pp. 75-94). Hillsdale, NJ: Erlbaum. Mandler, G. (1988). Memory: Conscious and unconscious. In P. R. Solomon, G. R. Goethals, C. M . Kelley, & B. R. Stephens (Eds.), Memory: Interdisciplinary approaches, (pp. 84-106). New York: Springer-Verlag. Martin, M . (1990). On the induction of mood. Clinical Psychology Review, 10, 669-697. Mayer, J. D., Gaschke, Y . N . , Braverman, D. L. , & Evans, T. W. (1992). Mood-congruent judgment is a general effect. Journal of Personality and Social Psychology. 63, 119-132. Mathews, A. , & Eysenck, M . (1987). Clinical anxiety and cognition. In M.J. Eysenck & I. Martin, (Eds.), Theoretical foundations of behavior therapy (pp. 217-234). New York: Plenum Press. Morris, C D . , Bransford, J. D., & Franks, J. J. (1977). Levels of processing versus transfer approapriate processing, Journal of Verbal Learning and Verbal Behavior. 16, 519-533. Nissen, M . J., Ross, J. L. , Willingham, D. B., MacKenzie, T. B., & Schacter, D. L. (1988). Memory and awareness in a patient with multiple personality disorder. Brain and Cognition, 8, 21-38. Parrott, W. G., & Sabini, J. (1990). Mood and memory under natural conditions: Evidence for mood incongruent recall. Journal of Personality and Social Psychology. 59, 321-336. Perrig, W. J., & Perrig, P. (1988). Mood and memory: Mood-congruity effects in absence of mood. Memory & Cognition, 16, 102-109. Physician's Desk Reference. (United States). (1996). Oradell, N . J.: Medical Economics Co. Plutchik, R. (1980). Emotion: A psychoevolutionary synthesis. New York: Harper & Row. Mood Dependent Memory 95 Polivy, J., & Doyle, C. (1980). Laboratory induction of mood states through the reading of self-referent mood statements: Affective changes or demand characteristics? Journal of Abnormal Psychology. 89, 286-290. Pyszczynski, T., Hamilton, J., C , Herring, F. J., & Greenberg, J. (1989). Depression, self-focused attention, and the negative memory bias. Journal of Personality and Social Psychology. 39. 1161-1178. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology. 39. 1161-1178. Russell, J. A. , Weiss, A., & Mendelsohn, G. A. (1989). Affect Grid: A single item scale of pleasure and arousal., Journal of Personality and Social Psychology. 57, 493-502. Schacter, D. L. Bowers, J., Booker, J. (1989). Intention, awareness and implicit memory: The retrieval intentionality criterion. In S. Lewandowsky, J. C. Dunn, & K. Kirsner (Eds.), Implicit memory: Theoretical issues (pp. 47-65). Hillsdale, NJ: Erlbaum. Schacter, D. L. , & Kihlstrom, J. F. (1989). Functional Amnesia. In F. Boiler and J. Grafman, (Eds). Handbook of Neuropsychology, (pp. 209-231). Amsterdam: Elsevier Publications. Schare, M . L. , Lisman, S. A., & Spear, N . E. (1984). The effects of mood variation of state-dependent retention. Cognitive Therapy and Research, 8, 387-408. Seibert, P. S., & Ellis, H . C. (1991). A convenient self-referencing mood induction procedure. Bulletin of the Psychonomic Society. 29, 121-123. Slamecka, N . J., & Graf, P. (1978). The generation effect: Delineation of a phenomenon. Journal of Experimental Psychology: Human Learning and Memory. 4, 592-604. Smith, S. M . (1984). A comparison of two techniques for reducing contest-dependent forgetting. Memory and Cognition. 12, 477-482. Smith, S. M . (1994). Theoretical principles of context-dependent memory. In D. Morris & M . Gruneberg (Eds.), Theoretical aspects of memory (pp. 168-195). New York: Rutledge. Snodgrass, J. G., & Corwin, J. (1988). Perceptual identification thresholds for 150 fragmented pictures from the Snodgrass and Venderwart picture set. Perceptual and Motor Skills. 67, 3-36. Sutherland, G., Newman, B., & Rachman, S. (1982). Experimental investigations of the relations between mood and intrusive unwanted cognitions. British Journal of Medical Psychology. 55, 127-138. Mood Dependent Memory 96 Teasdale, J. D. (1983). Negative thinking in depression: Cause, effect, or reciprocal relationship? Advances in Behaviour Research and Therapy. 5, 3-25. Tulving, E. (1983). Elements of episodic memory. Oxford: Oxford University Press. Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory (pp. 590-600). New York: Academic Press. Ucros, C. G. (1989). Mood state-dependent memory: A meta-analysis. Cognition and Emotion. 3, 139-167. Velten, E. (1968). A laboratory task for induction of mood states. Behavior Research & Therapy. 6, 473-482. Watson, D., Clark, L. A. , & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology. 54, 1063-1070. Watson, D., & Tellegen, A. (1985). Toward a consensual structure of mood. Psychological Bulletin. 98, 219-235. Weingartner, H. , Miller, H. , & Murphy, D. L. (1977). Mood-state-dependent retrieval of verbal associations. Journal of Abnormal Psychology, 86, 276-284. Wetzler, S. (1985). Mood state-dependent retrieval: A failure to replicate. Psychological Reports. 56, 759-765. Mood Dependent Memory 97 Appendix I: Musical Selections Pleasant and High arousal (delighted) Side a 1. Mozart Eine Kleine Nachtmusik: Allegro 5'38 2. Mozart Eine Kleine Nachtmusik: Rondo 3'01 3: Mozart Serenade #9: Finale 4'26 4. Bach Brandenburg Concerto #3: Allegro 4'57 5. Tchaikovsky The Nutcracker: Waltz of the Flowers 6'36 6. Tchaikovsky The Nutcracker: Trepak l'OO repeat Sideb 1. Beethoven Symphony #9: Presto 2'36 2. Vivaldi Four Seasons: Spring I Allegro 3'01 3. Vivaldi Four Seasons: Spring III Allegro 3'56 4. Vivaldi Four Seasons: Autumn I Allegro 4'24 5. Vivaldi Four Seasons: Autumn III Allegro 3'10 6. Tchaikovsky The Nutcracker: Dance of the Flutes 2'04 7. Tchaikovsky 1812 Overture 2'04 repeat Pleasant and Low arousal music (calm) Side a 1. Hoist The Planets. 8'20" Venus: The Bringer of Peace 2. Bach Ave Maria 5'00" 3. Grieg Peer Gynt. Morning: Prelude to Act IV 2'00" 4. Copland Appalachian Spring: Pastoral Interlude 6'00" repeat Sideb 1. Copland Rodeo: Corral Nocturne 3'34" 2. Copland Rodeo: Saturday Night Waltz 4'22" 3. Saint-Saens Le Cygne 3T7" 4. Brahms Symphony #2: 2nd Movement 3'30" repeat X 2 Mood Dependent Memory 98 Unpleasant and High arousal (anxious) Side a 1. Bruckner Symphony #8 (opening of last movement) 1*30" 2. Hoist The Planets. Mars:The Bringer of War 6'41" 3: Beethoven Grosse Fugue (op. 133 of string quartet) 4'30" 4. Shostakovich Symphony #5 (last movement) 10'59" repeat Sideb 1. Beethoven Symphony #6: Thunderstorm 3T0" 2. Grieg Peer GyntAbduction of the Bride 3'22" 3. Stravinsky Rite of Spring (part 1) 4'45" 4. Hoist The Planets. Uranus: The Magician 3'53" 5. Vivaldi Four Seasons: Summer 7'48" repeat #2 Presto #4 I Allegro Non Molto III Allegro Unpleasant and Low arousal music (sad) Side a 1. Albinoni Adagio in G minor 4' 52" 2. Sibelius Violin Concerto: Adagio di molto 3T4" 3. Lehar Viha-lied 2'23" 4. Schumman Traummeri 3'31" 5. Dvorak Symphony #9: Largo 5'50" repeat X 2 Sideb 1. Grieg Peer Gynt: The death of Ase 5'49" 2. Tchaikovsky Swan Lake: Dances des cygnes 3'00" 3. Chopin Prelude #4 in E minor 2'00" 4. Vivaldi Four Seasons: Autumn Adagio 2'20" 5. Stravinsky Firebird: lullaby (5'14" to 8'30") 3'15" repeat X 2 Mood Dependent Memory 99 Appendix II: Analysis of M C M in Autobiographical Event Generation Experiment #1: To examine the Event Generation task for M C M effects, events and their associated ratings were divided into positive, neutral, and negative groups, and are reported as a function of Study session mood in Table 30. In this section, unless otherwise specified, only subjects who generated and provided the associated ratings for at least one memory of each Event Type (positive, neutral, and negative) were included in the analyses. This was necessary in order to evaluate Event Type as a within-subject factor. Generation Latency Analysis of event generation latencies revealed only a main effect of Event Type (F(2, 230) = 3.58, p < .03, both other Fs < 1). Tukey tests indicated that, on average, neutral events required more time for generation than either positive or negative events (means of 14.7, 10.2 and 10.1 respectively). There was no indication of congruent effects in generation latency, replicating previously published results (Eich et al. 1994). Intensity Ratings Analysis of intensity ratings compared positive to negative events given that neutral events, by definition, do not vary in intensity. The analyses unveiled a main effect of Event Type Table 30 Percent of Events Generated, Generation Latency, Rated Intensity, Importance and Vividness as a Function of Induced Mood and Event Type (positive , neutral, and negative). Induced Event Gen Lat Rating Mood Type % n (sec) Int Imp Viv Delighted pos 70.3 40 7.9 2.7 2.8 3.4 [40] neut 11.7 29 14.6 1.6 2.5 neg 17.9 34 9.9 2.1 2.6 3.3 Calm pos 66.4 40 10.1 2.4 2.5 3.2 [40] neut 12.9 33 14.1 1.6 2.4 neg 20.6 36 9.7 2.0 2.5 3.2 Anxious pos 40.2 37 12.9 2.3 2.6 3.2 [40] neut 15.4 33 14.6 1.5 2.5 neg 44.3 40 9.4 2.3 2.8 3.4 Sad pos 51.8 39 11.9 2.4 2.6 3.3 [40] neut 12.7 37 15.5 1.6 2.5 neg 35.4 38 11.9 2.1 2.7 3.2 Mood Dependent Memory 100 (F(l , 140) = 23.0, p < .004) which was qualified by an Event Type by Study Mood interaction (F(3, 140) = 3.97, p < .01). Typically, positive events were attributed greater intensity ratings than were negative events (means of 2.4 vs 2.1 respectively). However, investigation of the interaction via Tukey tests shows that P mood groups reported higher intensity ratings for positive than for negative events (means of 2.6 and 2.0) whereas U mood groups recorded equal intensity ratings for both event types (means of 2.3 and 2.2). Importance Ratings Analyses of importance ratings show a main effect of Event Type (F(2, 230) = 84.7, p < .001) and a marginal Event Type by Study Mood interaction (F(6, 230) = 1.82, p < .10). Typically, both positive and negative events are rated as more important than neutral events, (means of 2.7, 2.5, and 1.6 respectively). However, the interaction reveals a mild mood congruent effect. Comparison of P to U mood subjects (Tukey tests) shows that while P mood group subjects assigned greater importance to positive than to negative events (means of 2.6 and 2.3), U mood group subjects assigned comparable importance to both positive and negative events (means of 2.7 and 2.8). Vividness Ratings Analyses of vividness ratings unveiled a main effect of Event Type (F(2, 230) = 43.7, p < .001), but no interaction (F < 1). Positive and negative events were rated as more vivid than neutral events (means of 3.3, 3.2, and 2.5, respectively) but these ratings were not influenced by mood. Summary The results suggest that mood manipulations produced effects that were sufficiently potent to influence the way that subjects perceived the intensity and importance of memories they produced. It may be that moods impact on the memories that are selected or the way they are remembered. In either case, the extent of the mood congruent effects indicates that mood manipulations had substantial impact on subjects' event generation performance. Mood Dependent Memory 101 Experiment #2: To examine the autobiographical event generation task for M C M effects, events were divided into positive, neutral, and negative events (percentages are reported as a function of Study session mood in Table 31). As M C M effects in the proportion of Event Types generated was evident only via within-subject comparion, the remaining analyses also entail within-subject comparisons and, unless otherwise specified, use seven subjects who had one memory from each Event Type. Ratings based on these 23 occasions (12 M and 11 D) are also presented in Table 31. Analysis of generation latency revealed neither a main effect of Event Type, nor an interaction between Event Type and Study mood (both Fs < 1). However, a marginal main effect of Study mood (F(l , 6) = 4.84, p < .075) indicated that, regardless of Event Type, subjects tended to generate events more quickly when manic than when depressed (means of 7.7 and 11.9 sec). Examination of event intensity draws on only positive and negative events and is therefore based on 9 subjects who generated both positive and negative events in each mood condition. Analyses revealed neither main effects nor an interaction between Event Type and Study mood (all Fs (1, 8) < 2.4, all p > .15) indicating that subjects characterized positive and negative events as comparably intense (2.7 and 2.4) regardless of Study mood. Analysis of event importance and event vividness ratings revealed reliable main effects of Event Type (both Fs(2, 12) > 8.3, g < .01). Tukey tests indicate that neutral events were rated as significantly less important and less vivid than were either positive or negative events which did Table 31. Percent of Events Generated, Generation Latency, Rated Intensity, Importance and Vividness as a Function of Mood and Event Type. Mood Event Type Gen % #Cnds All Evnt Types Lat (sec) Int Rating Imp Viv Manic [17] pos neut neg 59.2 10.7 30.1 12 7.3 9.7 9.2 2.7 2.5 2.6 3.6 2.0 2.5 2.9 3.4 Depressed [15] pos neut neg 51.3 14.7 34.0 11 11.0 13.4 11.3 2.5 2.6 2.7 3.2 1.9 2.2 2.9 3.6 Mood Dependent Memory 102 not differ from each other (neut, pos and neg importance ratings of 2.0, 2.7, and 2.8 respectively; neut pos and neg vividness ratings of 2.2, 3.3, and 3.6 respectively). Given that there were no interactions with Study mood these relationships were true of ratings performed in both manic and depressed moods. Event ratings agree with the analysis of Event Types generated and reveal little evidence of M C M effects in Bipolar patients' performance in the Event Generation Task. Mood Dependent Memory 103 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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


Related Items