<%BANNER%>

Dissecting Emotion: Towards a Functional Neuroimaging Probe for Affective Disorders

HIDE
 Title Page
 Dedication
 Acknowledgement
 Table of Contents
 List of Tables
 List of Figures
 Abstract
 Introduction
 Methodology
 Preliminary data: Dissecting the...
 Face matching and the amygdale:...
 Dissociating event-related responses...
 Discussion
 Appendix: IAPS pictures codes
 References
 Biographical sketch
 

PAGE 1

1 DISSECTING EMOTION: TOWARDS A FU NCTIONAL NEUROIMAGING PROBE FOR AFFECTIVE DISORDERS By PAUL WRIGHT A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

2 Copyright 2006 by Paul Wright

PAGE 3

3 Dedicated to my parents, who ke pt me curious about how things worked and what things meant.

PAGE 4

4 ACKNOWLEDGMENTS I thank my committee members for their different contributions to my apprenticeship: to Yijun Liu for modeling vision, to Christiana Le onard for modeling rigor, to Dawn Bowers for modeling optimism, and to Russell Bauer for modeling practical insight. Special thanks are due to A ndy James for showing me it can be done, to Jason Craggs for helping me overcome inertia, and to Jessica Couch for keeping me moving. I also thank Emmanuel Mennonite Church for thei r moral and spiritual support.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........8 LIST OF FIGURES................................................................................................................ .........9 ABSTRACT....................................................................................................................... ............10 CHAPTER 1 INTRODUCTION..................................................................................................................12 Specific Aims.................................................................................................................. ........12 Components of Emotion Processing.......................................................................................14 Brain Regions Implicated in Depression................................................................................16 Anatomical Background.........................................................................................................18 The Amygdala.................................................................................................................19 The Insula..................................................................................................................... ...20 The Ventral Prefrontal Cortex.........................................................................................22 The Anterior Cingulate Cortex........................................................................................24 Emotion Processing in Healthy Individuals...........................................................................26 Processing Emotion in Facial Expressions......................................................................27 Processing Emotion in Complex Scenes.........................................................................30 2 METHODOLOGY.................................................................................................................34 How fMRI Works................................................................................................................. ..34 Basics of fMRI Paradigm Design...........................................................................................38 Basics of fMRI Analysis........................................................................................................ .40 Interpreting fMRI results...................................................................................................... ..43 Examining the BOLD Response to Exclude False Activation........................................44 Using Control Conditions to Test Specific Cognitive Components................................45 Using Factorial and Parametric Designs to Overcome Limits in the Subtractive Approach......................................................................................................................46 3 PRELIMINARY DATA: DISSECTING THE NEURAL CORRELATES OF DISGUST........................................................................................................................ .......49 Introduction................................................................................................................... ..........49 Methods........................................................................................................................ ..........51 Subjects....................................................................................................................... .....51 Disgust Picture Paradigm................................................................................................51 Functional Imaging Data Acquisition.............................................................................53 Functional Imaging Data Analysis..................................................................................53

PAGE 6

6 Results........................................................................................................................ .............54 Emotion Ratings..............................................................................................................54 fMRI Data...................................................................................................................... ..54 Discussion..................................................................................................................... ..........57 4 FACE MATCHING AND THE AMYGD ALA: BOTTOM-UP EMOTION PROCESSING OR NOT?......................................................................................................69 Introduction................................................................................................................... ..........69 Methods........................................................................................................................ ..........71 Subjects....................................................................................................................... .....71 Face Matching Task........................................................................................................71 Functional Imaging Data Acquisition.............................................................................72 Functional Imaging Data Analysis..................................................................................73 Results........................................................................................................................ .............74 Behavioral Data...............................................................................................................74 fMRI Data...................................................................................................................... ..74 Discussion..................................................................................................................... ..........76 Relevance Detection Activates the Amygdala................................................................77 Emotion Processing at the Amygdala Habituates...........................................................77 Spatial Processing Bypasses the Amygda la During the Control Condition....................78 Cognitive Processing During Emotion Matching............................................................79 Negative BOLD Responses.............................................................................................80 Complex Contributions to Amygdala Activation............................................................81 5 DISSOCIATING EVENT-RELATED RESPONSES TO TOP-DOWN AND BOTTOM-UP EMOTION PROCESSING............................................................................91 Introduction................................................................................................................... ..........91 Methods........................................................................................................................ ..........95 Subjects....................................................................................................................... .....95 Picture Rating Task Paradigm.........................................................................................95 Functional Imaging Data Acquisition.............................................................................97 Functional Imaging Data Analysis..................................................................................97 Results........................................................................................................................ .............99 Behavioral Data...............................................................................................................99 fMRI Data...................................................................................................................... ..99 Discussion..................................................................................................................... ........100 Top-down Appraisal in the OFC and Insula..................................................................101 Bottom-up Processing in the Amygdala........................................................................103 Mixed Responses at the An terior Cingulate Cortex......................................................104 Response to Frequency Rating in the Parietal Cortex...................................................105 Summary and Conclusions............................................................................................106 6 DISCUSSION..................................................................................................................... ..115 Summary........................................................................................................................ .......115

PAGE 7

7 Proof of Concept: Dissociated Re sponses to Disgust and Arousal...............................116 A Response at the Amygdala, but not Specific to Emotion..........................................117 An Event-related Emotion Rating Task Pa rtially Replicates Responses to Blockrelated Tasks..............................................................................................................119 Technical Considerations......................................................................................................121 Scanning Parameters.....................................................................................................121 Connectivity Analysis...................................................................................................122 Current Trends in Functional Imagi ng of Major Depressive Disorder.................................123 Emotiona l Bias..............................................................................................................123 Emotion Regulation.......................................................................................................126 Cognitive Tasks.............................................................................................................127 Future Direction: the Anterior Cingulate Co rtex and the Interac tion of Emotion and Cognition...................................................................................................................... .....128 APPENDIX IAPS PICTURES CODES...................................................................................134 LIST OF REFERENCES.............................................................................................................140 BIOGRAPHICAL SKETCH.......................................................................................................155

PAGE 8

8 LIST OF TABLES Table page 3-1 Affective ratings (dimensional).........................................................................................60 3-2 Affective ratings (categorical)...........................................................................................61 3-3 Clusters of activation for (threat neutral)........................................................................62 3-4 Clusters of activation for (contamination neutral)...........................................................63 3-5 Clusters of activation for (mutilation neutral).................................................................64 4-1 Behavioral data................................................................................................................ ..82 4-2 Clusters of activation for [(Emotion Identity) (Identity Control)]............................83 4-3 Clusters of activation for [(Emotion Control) (Identity Control)]............................84 4-4 Regions showing significant modulation of BOLD response............................................85 5-1 Response time in milliseconds (standard deviation)........................................................108 5-2 Clusters of activation for inte raction of valence and task................................................109 5-3 Clusters of activation for main effect of valence.............................................................110 5-4 Clusters of activation for main effect of task...................................................................111 A-1 Contamination pictures....................................................................................................134 A-2 Mutilation pictures...........................................................................................................135 A-3 Threat pictures................................................................................................................ .136 A-4 Neutral pictures............................................................................................................... .137 A-5 Emotion rating pictures....................................................................................................138 A-6 Frequency rating pictures.................................................................................................139

PAGE 9

9 LIST OF FIGURES Figure page 3-1 Statistical maps showing contrasts betw een each emotional condition and neutral..........65 3-2 Glass brain view of regions of interest...........................................................................66 3-3 BOLD responses................................................................................................................67 3-4 Correlations with emotion ratings......................................................................................68 4-1 Matching task paradigm.....................................................................................................86 4-2 Selective response to emotion at the left inferior prefrontal sulcus...................................87 4-3 Response to face matching at the left and right amygdala.................................................88 4-4 Regions of deactivation......................................................................................................89 4-5 Habituation.................................................................................................................... .....90 5-1 Main effect of task...........................................................................................................112 5-2 Main effect of valence.....................................................................................................113 5-3 Responses in the anterior cingulate cortex.......................................................................114

PAGE 10

10 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DISSECTING EMOTION: TOWARDS A FU NCTIONAL NEUROIMAGING PROBE FOR AFFECTIVE DISORDERS By Paul Wright December 2006 Chair: Yijun Liu Major Department: Medical Sciences--Neuroscience The goal of this research was to devel op a functional magnetic resonance imaging paradigm for use in investigating major depressi ve disorder. Functional neuroimaging studies of depression have reported altered resting brai n metabolism and altered responses to simple emotion paradigms in the amygdala, ventral prefr ontal cortex, and anterior cingulate cortex. We studied healthy individuals responses to comple x emotion paradigms to attempt to distinguish activity in these regions. We hypothesized that the amygdala me diates bottom-up processing driven by emotional stimulus content, and that the ventral prefrontal cortex and anterior cingulate cortex mediate top-down processing driven by explicit knowledge or in tention. In the first experiment, we tested whether matching faces by emotional expression elicited a bottom-up response in the amygdala. The amygdala responde d during matching of both emotional and nonemotional faces, implying that this response was driven by top-down demands of the task. In the second experiment, we measured responses during rating of emotional pictures. The response in the amygdala was greater to unpleasant than pl easant pictures, regardless of rating task: a bottom-up response. The response in the ventral pr efrontal cortex was greater to emotion rating than non-emotional rating, regardless of picture content: a top-down response. The anterior cingulate cortex showed weak, mixed bottom-up and top-down responses. This emotion rating

PAGE 11

11 paradigm improves existing approaches to imaging the neural bases of major depressive disorder by eliciting dissociated responses in two regi ons implicated in depres sion: the amygdala and orbitofrontal cortex. This paradigm may be used in future studies to investigate in parallel the effects of depression on bottom-up and top-dow n emotion processing. Future studies may attempt to elicit more specific responses in the anterior cingulate cortex by using paradigms in which emotional stimuli interfere with the performance of cognitive tasks.

PAGE 12

12 CHAPTER 1 INTRODUCTION Functional neuroimaging has been used to identify brain regions involved in major depressive disorder (MDD); however the individual contributions of these regions to illness are not known. Specific regional responses may be elic ited by using imaging task paradigms that dissect the components of emotion processing. The goal of the experi ments in this dissertation is to refine the use of functional magnetic resonanc e imaging (fMRI) as a tool for probing emotion processing in healthy individuals, in order to improve its usefulne ss for later investigations of affective disorders. The feasibility of disse cting emotion with fMRI was demonstrated by showing that neural responses to emotional pict ures in different regions of the brain were associated with different ratings of picture content (chapter 3). We then attempted to refine two existing paradigms reported to elic it distinct neural responses to bottom-up (stimulus-driven) and top-down (knowledge-driven) processing of emotional stimuli. A face-matching task was reported to elicit bottom-up responses in the amy gdala that were inhibited during face labeling. We added a novel control condition to the face ma tching task to test whether the amygdala responded to stimulus content and not to task de mands (chapter 4). Rating emotional pictures has also been shown to recruit co rtical regions and to modulate limbic responses. We aimed to reproduce this modulation using an optimized paradigm design, where task components were varied at the event level and re sponses were detected using fact orial analysis (chapter 5). The findings of these studies are discussed in re lation to ongoing functional neuroimaging research on MDD, and future studies are recommended (chapter 6). Specific Aims Studies of MDD have highlighted three br ain regions where resting metabolism or responses to simple emotion tasks were altered: the amygdala, ventral prefrontal cortex (PFC),

PAGE 13

13 and anterior cingulate cortex (ACC) (David son et al., 2003;Drevet s et al., 1992;Mayberg, 2003;Sheline et al., 2001;Siegle et al., 2002). These regions are posit ed to be involved respectively in rapid generation (LeDoux, 2000), contextual modification (Rolls, 1999), and explicit appraisal (Lane et al ., 1997a) of emotional responses. The goal of the experiments presented in this dissertation is to identify an emotion paradigm that can elicit dissociable responses in limbic and cortical regions to bot tom-up and top-down emotion processing, and that can demonstrate modulation of bottom-up res ponses under different top-down conditions. Aim 1. Assess the validity and reliability of the amygdala response to a face matching paradigm. Verbal labeling of emotional stimuli is hypothesized to inhibit emotional feelings. A previous study using emotional faces compared verbal labeling with matching faces by their expression (Hariri et al., 2000). Du ring labeling, responses were de creased at the amygdala and increased in the ventral prefr ontal cortex. The response to fa ce matching was hypothesized to represent associative (or bottom-up) processing of emotion; however, the task design confounded emotional content with task performance. In order to test the hypothe sis that the amygdala response reflected bottom-up processing, we comp ared face matching with a novel control task in which neutral faces were matc hed by identity. In order to asse ss the reliability of the amygdala response, we investigated changes in the res ponse over repeated presentations of the task. Aim 2. Dissociate the neural correlat es of top-down and bottom-up emotion processing using a picture rating task. Responses to emotional stimuli may also be inhibited by tasks requiring that feelings be explicitly attended and appraised. Rating emotional pictures has been shown to recruit the ACC and modulat e the response in the amygdala (Lane et al., 1997a;Taylor et al., 2003). We attempted to repro duce these results using an optimized paradigm design. Stimulus content and rating task instructions were randomized at an event level in order

PAGE 14

14 to eliminate expectancy effects. Factorial analysis of the neural responses was used to identify main effects of bottom-up processing of stimul us content, top-down pro cessing of task demands, and their interaction. Components of Emotion Processing It is not yet known whether spec ific symptoms of depression may be linked with specific components of emotion processing. The symptoms of major depressive disorder may be grouped into attentional or c ognitive symptoms, such as apathy, psychomotor retardation, impaired attention, and executive dysfunc tion, and vegetative symptoms such as disordered sleep, disordered appetite, and endocrine disturbanc es. It has been proposed that cognitive and vegetative symptoms are mediated respectively by hypoactivity in dorsal, cortical brain regions and by hyperactivity in ventral, subcortical br ain regions (Mayberg, 1997) These two circuits may respectively mediate the top-down effects of cognitive behavioral therapy and the bottomup effects of antidepressant me dication (Goldapple et al., 200 4). The chronic negative mood effects of MDD may be probed by measuring th eir influence on acute emotion processing. For example, individuals with a sad mood are more likely to attend to and later to recall sad emotional stimuli (Eysenck, and Keane, 2000). Chr onic alterations in cort ical and subcortical circuits may be probed separately by measur ing acute responses to bottom-up and top-down components of emotion processing. The production and regulation of emotion is complex. The selection of theories below provide the basis for the worki ng hypotheses use in the current re search. Passer and Smith define emotions as positive or negative feeling (affect) states consisting of a pattern of cognitive physiological and behavioral reactions to events that have relevance to important goals or motives (Passer, and Smith, 2001). The physiol ogical reaction to an emotional event was emphasized by the 19th century psychologist William James, who proposed that, the bodily

PAGE 15

15 changes follow directly the per ception of the exciting fact, and that our fe eling of the same changes as they occur IS the emotion (James, 1884, original emphasis). These bodily changes include muscle movement, such as shivering, an d visceral responses or autonomic responses, such as heart palpitations. Despite James em phasis on the physiological reaction, he assumed that when faced with an emotional event, pe rception and appraisal of the event preceded the generation of a bodily response (Ellsworth, 1994). Furthermore, he proposed that bodily signals must recombine in the brain with a representation of the exciting fact in order for an emotion to be felt. Later researchers debated the relativ e influence of immediate perception and deliberate appraisal upon the behavioral and physiological signs of emotion. Zajonc proposed the affective primacy hypothesis, according to which events may be appraised emotionally without conscious awaren ess (Zajonce, 1980, quoted in Eysenck & Keane, 2000). He supported his hypothesis by investigating the behavioral effects of subliminally presented emotional stimuli. Individuals view ing subliminal emotional faces followed by Chinese pictograms rated pictograms preceded by happy faces as more likable. It is not clear, however, whether the change in liking score wa s accompanied by any feelings about the Chinese pictogram. Lazarus emphasized how emotion wa s influenced by the conscious appraisal of events (Lazarus, 1982, quoted in Eysenck & Keane, 2000). He showed that physiological responses to a distressing movie of a surgical procedure altered according to whether the movies narration emphasized or downplayed the emotionality of the events. It has been noted, however, that the movies narration itself may be consid ered an emotional stimulus. Although Zajonc and Lazarus present apparently opposing theories, evid ence from studies of f ear conditioning in rats suggests that emotional behavior may be produc ed by two parallel appraisal systems. LeDoux identified a fast and a slow route by which se nsory stimuli could reach the amygdala and thus

PAGE 16

16 evoke bodily responses. He used this evidence to draw together the theories of Lazarus and Zajonc, stating: The activation of the amygdala by inputs from th e neocortex [slow route] is consistent with the classic notion that emotional proces sing is postcognitive, wher eas the activation of the amygdala by thalamic inputs [fast route] is consistent with the hypothesis, advanced by Zajonc (1980), that emotional processing can be preconscious and precognitive. (Quoted in Eysenck & Keane, 2000, p.493) Thus there is evidence for two anatomic routes for emotion processing. A ppraisal theorists later proposed that the components of emotional processing do not follow one another in linear order, but evolve in parallel (Ellswo rth, 1994). Ellsworth states, Ne ither interpretation, nor bodily feedback, nor subjective experien ce comes first; at the very most, one can talk about which of these complex temporal processes starts first. If these components of emotion processing are indistinguishable temporally, they might be di stinguishable spatially by identifying distinct anatomical correlates of each component. In this dissertation, the neural co rrelates of two levels of emotion processing are operationally defined: Bottom-up processing responses that are determin ed by the emotional content of stimuli, independently of task demands. Top-down processing responses that correlate with the explicit appraisal (naming or evaluation) of perceived or experienced emotion. It is not clear how these two levels of emotion processing ar e linked to specific symptoms of MDD. Such a link may be found by identifying brain regions that shar e a common association with specific components of emotion processi ng and with specific symptoms of depression. Functional imaging studies of MDD therefore pr ovide the anatomic targets for the emotion paradigms used in this dissertation. Brain Regions Implicated in Depression Early studies of resting brain metabolism in patients with MDD measured cerebral glucose metabolism using positron emission tomography (PET), and reported increased metabolism in

PAGE 17

17 the amygdala and ventral PFC (Drevets, 1998). Amygdala metabolism correlated with symptom scores and with plasma cortisol (a hormone asso ciated with response to stress) during depressive episodes. Later fMRI studies extended these fi ndings. The amygdala response was increased in patients viewing pictures of fearful faces, and decreased following successful medication (Sheline et al., 2001). Because the emotional st imuli were presented sufficiently rapidly to prevent their conscious percepti on, the increased response may re flect altered implicit emotion processing in MDD. However, MDD may also alter top-down processing in the amygdala. The amygdala response to unpleasant em otional words lasted around 10 seconds in controls, but in patients lasted around 25 seconds (Siegle et al., 2002). Because respons e duration correlated weakly with self-report measures of rumination, the author s associated thes e changes with prolonged elaborative processing of emotiona l information. Rumination may also underlie increased metabolism in the PFC, as explained by Mayberg: Frontal hyperactivity is now viewed as an exaggerated or maladaptive compensatory process resulting in psychomotor agitation and rumination, servi ng to over-ride a persistent negative mood generated by abnormal chronic activ ity of limbic-subcortical structures. In contrast, frontal hypometabolism seen with incr easing depression severity is the failure to initiate or maintain such a compen satory state. (Mayberg, 2003, p. 197) This view is supported indire ctly by an fMRI study of patie nts viewing pleasant emotional pictures (Mitterschiffthaler et al., 2003). Patients lacked the re sponse seen in controls in the medial PFC, but had increased responses in th e ventrolateral PFC. Because previous studies showed overlapping responses in the ventrolateral PFC to cognitive and emotional tasks (Drevets, and Raichle, 1998a), th e authors suggested that responses to pleasant pictures in this region may reflect an attempt to experience positive emotion. A compensatory process may also involve th e ACC. Studies of resting metabolism in patients with MDD reported different responses in three regions of the ACC, inferior, superior, and anterior to the genu of the corpus callosum. Metabolism in the subgenual ACC is increased

PAGE 18

18 in MDD and during induced sad mood (Mayberg 1997). In patients who respond to medication, metabolism decreases in the subgenual ACC and increases in dorsal regions, including the supragenual ACC. These changes were replicated in two studie s of primary unipolar depression (Kennedy et al., 2001;Mayberg et al ., 2000), one of Parkinsons di sease patients with secondary depression (Stefurak et al., 2001), and one of pati ents responding to place bo treatment (Mayberg et al., 2002). Metabolism in th e pregenual ACC was unaffected by medication, but pretreatment metabolism in this region distinguished patien ts who later responded to medication from those who did not (Mayberg et al., 1997). This finding was replicated in an fMRI study of patients with MDD (Davidson et al., 2003). The response to unpl easant pictures at base line predicted symptom scores following eight weeks of medication. Furt hermore, cognitive-behavioral therapy (CBT) increased metabolism in the pregenual ACC, sugge sting that this region plays an important topdown role in recovery form depression (Goldapp le et al., 2002). The re sponse in the pregenual ACC to mood challenge is greater in remitted de pression patients then in healthy controls or patients with active depression (Liotti et al., 200 2). This altered respon se in the absence of treatment supports a protectiv e role for this region. Anatomical Background The findings reviewed above suggest that the ventral PFC and pregenual ACC mediate top-down compensation in MDD that may modul ate increased bottom-up processing in the amygdala. In general, animal studies, lesion st udies, and functional imaging studies agree, describing the roles of the amygdala in forming si mple emotional associa tions, the orbitofrontal cortex in contextual appraisal of emotion, and the anteri or cingulate cortex in emotionally guided behavior. An additional region, the insula, is associated with awaren ess of visceral and autonomic sensations.

PAGE 19

19 The Amygdala The amygdala appears to be involved in gene rating rapid but rough re sponses to stimuli that have emotional value. Its role in mediating fear behavi ors is often emphasized, although it also responds to rewarding stimuli. Anatomy. The amygdala is an almond-shaped nucl eus in the medial temporal lobe, anterior to the hippocampus. It sh ares reciprocal connections with many cortical and subcortical regions (Aggleton, and Saunders, 2000). Its inputs arrive principally at its lateral and basolateral nuclei and are passed, directly or indirectly, to the central nucle us for output. Of particular interest are cortical afferents fr om insula, temporal and anterior cingulate cortices and from the dorsolateral, medial and orbital pr efrontal cortices. The amygdala se nds efferents to each of these regions, and back-projections to the occipital and temporal cortices to modulate visual processing. Its subcortical outputs include the ta il of caudate, ventral putamen and nucleus accumbens as well as one-way efferents to the me diodorsal thalamus and reciprocal connections with the entire hypothalamus. It receives input s from the midline thalamus and medial pulvinar as well as various brainstem nuclei. Animal studies. LeDoux has described in great detail the involvement of the amygdala in fear conditioning in rats (LeDoux, 2000). In his mode l, fear is operationally defined as defensive posturing in response to an aversive stimulus, su ch as an electric shock (LeDoux, 2000). Rats can be conditioned to display fear responses to a ne utral stimulus (e.g., a tone ) by pairing it with the unconditioned stimulus (shock). These conditioned re sponses are impaired if the amygdala or its connections are damaged. Emotional informa tion may reach the amygdala through a fast, subcortical pathway involving th e collicular visual system, or through a slow, cortical pathway, involving occipital and temporal cortex. Single cell recordings in monkeys have implicated the amygdala in responses to both aversive a nd rewarding conditioning (Rolls, 1999). Some

PAGE 20

20 amygdala neurons responded exclusively to rewa rding food-conditioned stimuli or to aversive fear-conditioned stimuli, and others responded to both. Although amygdala neurons rapidly alter their firing patterns to detect conditioned stimuli, these associations appear to be inflexible. More flexible, context-dependent associations may be learned by the orbitofrontal cortex (see below). Some neurons in the monkey amygdala respond selec tively to faces, demonstrating a role for the amygdala in responding to socially salient visual cues (Leonard et al., 1985). Monkeys with bilateral amygdala lesions are emot ionally unresponsive; they show no fear of snakes or humans and eat items not usually used as food. Human lesion studies. Humans rarely have lesions confin ed to the amygdala. Injuries or damage from encephalitis often affect the hippo campus and temporal lobe as well (Aggleton, and Saunders, 2000). There is no sing le, clear symptom of amygdala damage. It is apparent, however, that such lesions rare ly cause cognitive deficits, bu t generally cause changes in emotionality. Extensive lesions involving th e amygdala are associated with Kluver-Bucy syndrome, which includes symptoms of blunt ed emotions, hypersexuality and hyperorality (placing non-food objects in the mouth). Bilateral amygdala lesions are cons istently associated with impaired recognition of f acial expressions of fear (Adol phs et al., 1994;Broks et al., 1998;Sprengelmeyer et al., 1999;Young et al., 1995) and occasionally associated with impaired recognition of vocal expression s of fear (Scott et al., 1997). The Insula The insula has been described as sensory cort ex for the viscera, and may be involved in processing the bodily responses involved in emotion (Adolphs, 2002). This region is notably associated with the emotion disgust, but is also involved in pain and awareness of visceral sensations (Critchley et al., 2002).

PAGE 21

21 Anatomy. The insula is located inside the Sylvia n fissure, and is so-called because is completely concealed (insulated) by the tempor al and frontal lobes. It is divided from anteroventral to posterodorsal into agranular, dysgranuar, and granular regions (Mesulam, and Mufson, 1982a). The insula receives input from all five sensory modalities, and shares reciprocal connections with the amygdala, lateral orbita l cortex, and ACC (Mufson, and Mesulam, 1982). Limbic connections, as well as olfactory, gustato ry, and autonomic connections, are particularly extensive in the anterovent ral insula (Mesulam, and Mufson, 1982b). Mesulam and Mufson suggest that the insulas connect ions to the amygdala allow viscer al input to the limbic system, and furthermore that its common connectivity patterns with the lateral orbital cortex identify these two regions as part of an integrated paralimbic unit. Human lesion studies. The insula is associated with im paired recognition of disgust. A patient with selective damage to the right in sula and putamen was impaired at recognizing disgust in two sets of face stimuli, and in two sets of vocal stimuli (Calder et al., 2000). Despite being able to recognize disgust conveyed in co mplex scenes, he scored low on questionnaires measuring the experience of disgust. Another pa tient with extensive lesions was impaired at recognition of all static emotional stimuli, but could recognize emotions acted out or described in stories, with the excep tion of disgust (Adolphs et al., 2003). Adolphs et al. suggested that the processing of acted-out emotions bypassed the damaged limbic a nd ventral prefrontal regions, and relied upon parietal and dorsa l prefrontal pathways. They also suggested that recognition of emotion depended upon regions representing somatic states, and that this patients selective impairment at recognizing disgust stemmed fr om his selective lesion of this region of somatosensory cortex, with sparing of the more dorsal postcentral regions. Thus the insula

PAGE 22

22 appears to be particularly important in the se nsory experience of emotion, and in disgust in particular. The Ventral Prefrontal Cortex The ventral PFC is implicated in the flexib le association of st imuli with reward and punishment, social functioning, and regulation of mood. A regi on of the ventral PFC, the orbitofrontal cortex, is heavily connected with the amygdala and may be involved in contextual fine-tuning of primitive emotional signals from th e amygdala. Unreferenced data in this section is taken from The Orbitofrontal Cortex and Reward (Rolls, 2000). Anatomy. The orbitofrontal cortex is located on th e ventral surface of the frontal lobes, adjacent to the orbits of the eyes. The orbitofron tal cortex receives input from primary gustatory, olfactory, auditory, and somatose nsory cortices. The inputs of pr imary relevance to the current research, however, arise from multiple stages alon g the ventral visual pathways in the temporal lobe, which are involved in object recogni tion, and in particular face recognition. The orbitofrontal cortex also rece ives strong inputs from the amyg dala and from the mediodorsal nucleus of the thalamus. Its outpu ts include back-projections to the ventral visual pathways and outputs to the amygdala, ACC, latera l hypothalamus and ventral striatum. Animal studies. The orbitofrontal cortex was first associated with reward with the discovery of neurons with selectiv e responses to taste, a primary re inforcer (that is, certain tastes can be rewarding or punishing, such as sweet or sour). The taste rewa rd neurons activity was enhanced by hunger, and specific to the type of reward, in that a monkey fed to satiety on bananas would maintain an orbito frontal response to the sight of peanuts while the neurons tuned to bananas ceased responding. When a food rewa rd was associated with a visual stimulus, orbitofrontal neurons responded to that stimulus, provided the animal was hungry. Certain orbitofrontal neurons responded to th e reward value of stimuli, even after their associations were

PAGE 23

23 reversed in a visual discriminati on reversal task. In this task, food was paired initially with a square symbol, and after reversal was paired with a triangle. Orbito frontal neurons responding initially to the square rapidly adapted to respond to the triang le after reversal. Some neurons responded to non-reward when reward was expect ed, for example after the switching described above. Other non-reward neurons responded selectiv ely to removal or termination of a reward, possibly enabling a context-specific response. Ce rtain orbitofrontal neurons in macaques carry information about faces. They distinguish both ex pression and identity, probably receiving this information from neurons in the temporal visu al cortex, consistent with the connections described above. Human lesion studies. Human with orbitofrontal lesions tend to be euphoric, and have difficulty in planning and social functioning. The classic prefront al lesion patient is Phineas Gage, a 19th century railway foreman who survived the passage of an explosive-driven railroad spike into his left cheek bone and out of the top of his head, passing through his prefrontal cortex. Previously hard-working and respected, Ga ge began to neglect his work duties and his marriage, engage in drinking and brawling. More recent studies have shown that patients with prefrontal lesions and impaired social functi oning perform badly on certain tasks. When human subjects were asked to perform a visual discri mination reversal task similar to the one above, subjects with ventral prefrontal lesions made more errors than c ontrols, apparently because they were less able to correct their behavior. Test performance correla ted with measures of social impairment. Similar patients were impaired at a gambling task in which two decks of cards were presented, one that gave large rewards but larger penalties and another th at gave small rewards but smaller penalties (Bechara et al., 1994). Patients with ventral prefrontal lesions were more likely than controls to pick the high-reward deck even when net gain was greater with the small-

PAGE 24

24 reward deck. They could apparently discern on ly the short-term positiv e, not the long-term negative consequences of their decision. Some patie nts with orbitofrontal lesions were unable to recognize emotion in facial expres sions and/or speech (Hornak et al., 1996). The latter deficits were distinct from impaired visual discrimination reversal. The Anterior Cingulate Cortex The anterior cingulate cortex is closely related both to medial prefrontal cortex and to motor cortex. It is associated with numerous functions in addition to emotion processing, including detection of pain and control of attention (Kerns et al., 2004;Vogt, 2005). Anatomy. The cingulate gyrus forms a semi-circu lar belt on the medial surface of the cortex, surrounding the corpus callosum. The cingul ate gyrus was first associated with emotion when Papez included it in his famous emotion circ uit (Papez, 1937). Papez postulated that just as the striate cortex was considered to be receptiv e cortex for visual signals from the retina, the cingulate gyrus may be considered to be recep tive cortex for emotional signals from the hypothalamus. He also saw the cingulate gyrus extensive cortical outputs as a means by which emotion could color other experiences, and cortic al inputs to the cingulate, as a means by which emotion could be generated by psychic processes as an alternative to visceral inputs. Papez described the region as the seat of dynamic vi gilance by which environmental experiences are endowed with an emotional consciousness. The ante rior cingulate cortex r eceives afferents from the medial orbitofrontal cortex, the amygdala, the temporal pole cortex and somatosensory cortex, including the insula (Rol ls, 1999). It receives rich dopa minergic innervation from the ventral midbrain (Bannon, and Roth, 1983). Its effere nts extend to the peria queductal gray in the midbrain, the dorsal motor nucleus of the vagus nerve and the ventral striatum and caudate nucleus (Rolls, 1999).

PAGE 25

25 Animal studies. Different behaviors may be associated with afferent signals from different regions of the ACC. The subgenual ACC projects to the medial hypothalamus and ventrolateral periaqueductal gray, whereas the pregenual ACC pr ojects to the dorsal hypothalamus and lateral periaqueductal gray (Ongur, and Price, 2000). Ongur and Price (2000) suggest that these projections allow the subgenual and pregenua l regions to evoke coordinated emotional responses, resulting respec tively in quiescent or confrontatio nal stances. Shima and Tanji trained monkeys to push or to turn a handle, by associat ing a reward with the pr eferred action. Certain neurons in the monkey anterior cingulate cortex responded to decreased reward, but not to constant reward. When the monkeys failed to adjust their behavior in response to reversal, these cells also failed to fire (Shima, and Tanji, 1998). This is similar to the orbitofrontal response to stimulus-reward reversal, but that the association is with action, not a stimulus. Injections of a GABA agonist into the anterior cingulate prevented the monkeys from altering their task behavior in response to changing reward. Thus the error-correc ting activity of the anterior cingulate cortex could be interpreted as part of a reward-seeking mechanism. Human lesion studies. Humans with anterior cingulate strokes appear to lose their initiative, and despite intact cognition and motor function, become quite in active and rarely even talk (Damasio, and Van Hoesen, 1983). Patients treated for chronic pain with bilateral 5mm lesions of the anterior cingulate cortex report that the pain co ntinues but no longer causes them distress. They also showed less spontaneous beha vior compared with c ontrols, producing shorter statements at a written task a nd producing fewer and simpler models when asked to put together Tinker Toys (Cohen et al., 1999). These findings, along with the animal studies above, suggest that the anterior cingulate corte x, particularly its pr egenual region, may medi ate the influence of processed emotional information on the adju stment and initiation of motor behavior.

PAGE 26

26 Emotion Processing in Healthy Individuals In patients with depression, functional im aging studies have pinpointed the amygdala, orbitofrontal cortex, and anterior cingulate cortex as possible s ites for deficits. Anatomical and lesion studies have described th e amygdalas role in associati ng emotional experiences with perceived stimuli, the orbitofrontal cortexs role in assigning context-de pendent emotional value to stimuli, and the anterior cingulate cortex s role in selecting actions based on their consequences anticipated value. These regions ha ve been investigated in healthy humans using PET and fMRI paradigms involving the percep tion and evaluation of emotional stimuli. Although a wide variety of emotional tasks have been used, this dissertation focuses on those that use visual stimuli to elicit emotion proces sing. Standardized sets of visual stimuli are available with rigorous descriptions of their content, allowing creation of precisely-defined emotional and control stimulus sets, and incr easingly the likelihood of reproducible results across experiments. Using visual stimuli to el icit emotion processing allows precise stimulus timing, improving the detection of the resulting neural response s. The Pictures of Facial Emotions consists of faces expressing happiness, sadness, fear, anger, disgust, and surprise (Ekman, and Friesen, 1976). These expressions ar e reliably recognized across cultures, implying that they are not socially learned, but ma y represent basic categories for the social communication of emotion (Ekman, 1982). The Intern ational Affective Pict ure System consists of emotional scenes with a wide variety of c ontent, eliciting a range of emotional responses (Center for the Study of Emotion and Attention [CSEA-NIMH], 2001). These pictures have been rated along three dimensions of emotion: vale nce (from pleasant to unpl easant), arousal (from calm to excited), and dominance (from controlled to in control). Ratings for pictures at either end of the valence scale tend also to have high arousal ratings. These ra ting scales have been validated using physiological measures of emotional responses (L ang, 1995). Eye blink

PAGE 27

27 responses to auditory startl e are increased by unpleasant pict ures and decreased by pleasant pictures. Skin conductance responses to pictures are proportionate to their arousal scores. Functional imaging studies have used both facial expressions and emotional scenes to investigate the neural correlates of bottom-up and top-down processing of emotion. Processing Emotion in Facial Expressions Facial expressions enable the social communication of emoti on. Neural responses to faces cannot be assumed to correlate with the experien ce of emotion, but may correlate with automatic (bottom-up) perception of emotion, or explicit (top-down) recognition of emotion (Davidson, and Irwin, 1999). Facial expressi on recognition is impaired in pa tients with MDD, supporting the use of face stimuli in imaging investigations of MDD (Gur et al., 1992). Furthermore, recent studies have shown that brain re gions that respond to facial e xpressions of disgust or pain overlap with brain regions that respond to the experience of di sgust or pain (Singer et al., 2004;Wicker et al., 2003). This evidence suggest s that the neural correlates of emotional communication and emotional experience may at l east partly overlap, fu rther supporting the use of face stimuli in imaging studi es of affective disorders. Lesions studies have implicated the amygdala in the perception of fear in facial expressions (Adolphs et al., 1994). This role was confirmed in numerous fMRI studies, which reported a bottom-up, stimulus-driven response in the amygdala to faces expressing fear and other salient emotions. The amygdala responded to fearful facial expressions regardless of whether attention was paid to emotion, to the ge nder of the faces (Morris et al., 1996;Winston et al., 2002;Winston et al., 2003), or to the properties of anothe r stimulus (Anderson et al., 2003;Vuilleumier et al., 2001). The amygdala al so responded when emotional faces were displayed very briefly (~30 ms) and rapidly repl aced with a neutral f ace, a technique called backwards masking that prevents conscious awareness of the stimulus (Morris et al.,

PAGE 28

28 1998;Whalen et al., 1998b). Other studies showed that the amygdala response may be modulated by top-down effects, reporting increased res ponses during explicit recognition of happy and disgusted faces (Gorno-Tempini et al., 2001), and decreased respons es during explicit recognition of happy and angry f aces (Critchley et al., 2000), both compared with gender recognition. Top-down modulation of the amygdala may be mediated by the prefrontal cortex. Patients with lesions in the ventral PFC have general social impairments (Rolls, 1999) and specific impairments in recognizi ng facial expressions (Hornak et al., 1996). Comparing emotion recognition with gender recognition elicited spec ific responses to emotion recognition in prefrontal regions (Gorno-Tempini et al., 2001;Winston et al., 2003) and elicited amplified responses to emotional faces in the fusiform gyrus and superior temporal su lcus (Critchley et al., 2000;Winston et al., 2002). These variable result s may reflect multiple strategies for emotion recognition: some participants may hold in mind verbal labels for the candidate emotions in order to guide their response, others may use a non-verbal st rategy, such as holding in mind a visual example of each facial expr ession. Later studies investigated the role of the ventral PFC in emotion recognition by compared verbal a nd non-verbal emotion recognition. One study compared facial expression matching (a perceptu al task) with facial expression labeling (an intellectual task) (Harir i et al., 2000). The amygdala respons e was lower during labeling then matching, and correlated inversely with activity in the right ventral PFC, suggesting that verbal judgments of emotion recruit top-down inhibi tion by the PFC of bottom-up processing in the amygdala. A subsequent study compared verbal and facial cues in a delayed match to sample task involving emotion or gender judgment s (Narumoto et al., 2000). This study found no amygdala response, and reported right prefront al activation to both verbal and non-verbal emotion judgment tasks. The difference in am ygdala responses between these tasks may be

PAGE 29

29 because Hariri et al. (2000) di splayed only fearful or angry faces whereas Narumoto et al. (2000) showed all six of Ekmans basic facial expression s, counterbalancing the valence of the stimuli. The prefrontal response to the non-verbal task in Narumoto et al. (2000) may be driven by the requirement to maintain a representation of the emotional information from the cue in working memory, whereas in the task employed by Hari ri et al. (2000) all stimuli are present simultaneously. Because the Hariri task may elicit top-dow n modulation by the ventral PFC of bottom-up responses at the amygdala, we chose to investigate th is task further. By re quiring participants to read verbal labels in every trial, the labeling condition of the Hariri task is likely to limit participants recognition strategy to verbal pr ocessing, and thereby elic it top-down processing. The matching condition of the Hariri task has b een shown to activate the amygdala reliably in studies looking at the influence on the amygdala of genetics (Hariri et al., 2002b;Pezawas et al., 2005), drugs (Hariri et al., 2002a;Tessitore et al., 2002), and aging (Tes sitore et al., 2005). However, the matching condition is more likely to involve knowledge-based processing than other control tasks, and may not e licit a true bottom-up response. Ha riri et al. (2000) claimed that participants do not match facial expressions using covert verbal labeling, but using perceptual cues such as wide eyes or a furrowed brow. Pe rceptual cue matching may represent intentional, knowledge-based processing in pursuit of task demands, or top-down processing Therefore, in order to distinguish facial feat ure matching from emotion proce ssing, and presumably thereby to dissect an emotional response at the amygdala, we modified the Hariri task to include an intermediate control condition in which neutral faces were matched by identity. This experiment tested the validity of using the face matching task to investigat e bottom-up emotion processing in

PAGE 30

30 the amygdala by examining whether the amygdala response is driven by emotional content or task demands. Processing Emotion in Complex Scenes The emotional scenes in the International A ffective Picture System (IAPS) differ from the Ekman faces in several important respects. First, they are complex, and differences in emotion are conveyed not by small shifts in facial c onfiguration, but by multip le components of the image. Second, whereas faces represent social signals of emotion (eliciting emotional perception), scenes are more like ly to evoke emotion directly, eliciting emotional experience. Whereas top-down processing of faces involves recognition of an emotional category, top-down processing of scenes is more lik ely to involve the appraisal of internal feelings. Third, the subjective ratings categorizing the Ekman faces ar e based on specific categories, but the ratings of IAPS pictures are based on general dimensional scores. As with emotional faces, viewing unpleasant scenes elicits a response in the amygdala (Irwin et al., 1996;Taylor et al., 1998). Studies us ing IAPS pictures also report greater responses to emotional than neutral scenes in the ventra l temporal cortex (Lane et al., 1997b;Lang et al., 1998). This response appears to be specific to unpleasant scenes (Lane et al., 1997b) and correlates with activity in the amygdala (Sabatin elli et al., 2005). Increased ventral temporal responses to unpleasant pictures ar e thought to be driven by back-pro jections to this region from the amygdala. Interestingly, in a st udy of patients with MDD, the contrast between responses to unpleasant and neutral scenes was gr eater in patients than in cont rols in the ventral temporal cortex but not in the amygdala (Davidson et al., 2003). The re sponses in the amygdala and ventral temporal cortex correla te with arousal scores, and t hus appear to reflect bottom-up responses driven by stimulus c ontent (Sabatinel li et al., 2005).

PAGE 31

31 The preliminary data presented in chapter three used IAPS pict ures to dissociate a response in the ventral temporal cortex to arousal from a response in the insula to disgust. By using both dimensional and categorical rati ngs to describe stimulus content, this study dissected two components of bottom-up processing. Previous studi es reported dissociated responses in the the amygdala to fearful faces and in the insula to disgusted faces (Phillips et al., 1997;Sprengelmeyer et al., 1998). A study of patients with obsessive-c ompulsive disorder sup ported these findings by showing that in patients, the insula response was greater to disgust-inducing IAPS pictures than fear-inducing pictures (Shapira et al., 2003). However, two othe r studies using IAPS pictures reported equal activation of the insula to both dis gustand fear-inducing pict ures (Schienle et al., 2002;Stark et al., 2003). The stimuli used to elicit disgust differ ed between studies, the former using only pictures of spoiled food, garbage, a nd other contaminants, and the latter using in addition pictures of injuries, tumors, and other mu tilations. In the study repor ted in chapter three, we examined separately the neural responses to pictures of contamin ation and pictures of mutilation, comparing both with pictures that elic it fear. Because contamination pictures elicit low arousal scores and mutilation pictures elicit high arousal scores, this study allowed us to investigate separately the neural correlates of the emotional cate gory disgust, and the emotional dimension arousal. Top-down processing of IAPS pictures has been investigated both using a labeling task, as described above, and using emotional rating tasks. Hariri et al. (2003) repeated their matching and labeling study using IAPS pictures. In this study, the matching task involved identifying identical threatening photographs (for example, a picture of a gun), while the labeling task involved choosing selecting betw een the verbal descriptors n atural and arti ficial. The amygdala response was larger for matching than la beling, while a larger response to labeling was

PAGE 32

32 found at the ventral prefrontal co rtex (BA 47) (Hariri et al., 20 03). The responses of these two regions were negatively correlated. While these responses echoed those in Hariri et al. (2000), the second task did not examine t op-down processing of emotion. In the original Hariri task, the labels were angry or afrai d, whereas in the second tas k, labeling required semantic processing of non-emotional stimulus content. Replicating the face matching and labeling task using emotional scenes is hindered by scenes co mplexity. While different face pictures within an emotional category share general features, diffe rent scenes eliciting, for example, fear may vary widely in their visual features, making matching more di fficult. Also, while categorical labels for facial expressions of emotion are wi dely recognized, categorical labels for emotional scenes have not been established. Perhaps for th ese reasons, studies of top-down processing of emotional scenes usually require part icipants to rate the scenes simply as pleasant or unpleasant. In the earliest picture rating study, particip ants viewed blocks of mixed unpleasant and neutral pictures during and rate d either whether each pictur e was pleasant or unpleasant or whether it was indoors or outdoors (Lane et al., 1997a). Emotion ra ting elicited larg er responses than location rating in the anterior cingulate cort ex (ACC) and medial pref rontal cortex (PFC). This study could not investigate bottom-up resp onses because unpleasant and neutral pictures were intermingled. In two subsequent studies, pleasant and neutral pi ctures were presented separately to identify regions involved in bo ttom-up responses (Liberzon et al., 2000;Taylor et al., 2003). Liberzon et al. (2000) compared emotion rating with picture recognition, a cognitive task intended to draw attention away from emotion. The right amygdala responded more to unpleasant than neutral pictures, and this contra st was greater during em otion rating than during picture recognition. Taylor et al. (2003) compared emotion rating with passive viewing, in order to test whether top-down pro cessing diminished bottom-up res ponses. The right amygdala and

PAGE 33

33 insula responded to unpleasant pi ctures, but the response was sma ller during emotion rating than during passive viewing. The opposite effect was s een in the ACC and medial PFC: responses to unpleasant pictures were increased during emoti on rating compared with passive viewing. These studies suggest that th e amygdala and insula may mediate bot tom-up responses to the content of emotional scenes, and that these responses may be modulated by top-down processing mediated by the ACC and medial PFC. Because these st udies used block designs, which may confound bottom-up emotion processing with expectation of emotion (a topdown effect), we attempted to replicate these findings using an optimized even t-related paradigm. In our task, both emotional and non-emotional rating tasks requi red ratings on continuous scales : either the pleasantness of the picture, or the frequency of its appear ance on television. Stimulus content and tasks instructions were randomized on a trial-by-trial basis, preven ting expectancy of emotional content, and equalizing the timing of bottom-up and top-down responses. Factorial analysis was used to separate the main effects of bottomup and top-down processing and their interaction.

PAGE 34

34 CHAPTER 2 METHODOLOGY Functional magnetic resonance imaging (fMRI) measures neur onal activity indirectly by detecting changes in blood oxygenation. By localiz ing these changes during a stimulus or task, fMRI may indicate the neural correlates of that task. FMRI is non-invasive, and has comparatively good spatial resolu tion. However, the temporal re solution of fMRI is limited by the sluggishness of the blood oxygenation leveldependent (BOLD) response. Furthermore, responses must be detected by comparing signal during an experimental task and a control condition. Therefore, in order to elicit reliable, valid responses fMRI paradigms must present stimuli with optimal timing, and must compare task conditions that are matched for every cognitive factor but the one bei ng studied. This chapter describes the physical basis for fMRI, the basics of paradigm design, and some advanced techniques for confirming the validity of statistical maps and for investigating responses to interacting cognitive functions. Unreferenced material is taken Functional magnetic resonance imaging (Huettel et al., 2004) or from class notes from BCH 6741, Magnetic Resonance Imag ing and Spectroscopy, taught by Dr. Thomas Mareci. How fMRI Works Magnetic resonance imaging is based on a radio signal produced by excited hydrogen nuclei (spins) in a strong magnetic field. The im ages produced by MRI are divided into slices, and each slice is composed of units called voxels A voxel is comparable to a pixel (picture element) in a computer image, but because MRI slices have thickness, their cubic constituents are called volume elements. Th e intensity of every voxel in a slice is calculated by decoding a single, complex radio signal that combines the individual radio signals from each voxel. The individual signals are encoded by varying thei r frequency and phase according to their spatial

PAGE 35

35 locations. Frequency and phase encoding, and the complex calculations used to reconstruct MR images, will not be discussed here. We will assume that MR signal intensity at a given voxel is proportional to the radio signal produ ced at that voxel. Functional MRI measures signal that is sensitive to changes in blood oxygenation. This si gnal decays over time, as described in Equation 1. (1) S(t) = S0e-t/T2* Where S = signal, t = time, S0 = signal at t=0, and T2* = decay constant. Equation 1 shows that signal decay rate is exponential, and varies according to the T2* decay constant. T2* describes the combined infl uence of a number of factors upon signal decay. One factor, T2, describes interac tions between the spins, which varies between different tissue types but is essentially constant. In an ideal situation T2* = T2, but in reality T2* is shorter than T2, and real signal decay rates are more rapid than ideal signal decay rates. This is due to the influence of inhomogeneity effects, or local im perfections in the magnetic field. Inhomogeneity may be caused by the presence of paramagnetic ma terial, such as deoxyhemoglobin, or interfaces between air and water. It is th e inhomogeneity component that causes T2* to vary over time, which in turn causes the cha nges in signal seen in fMRI. The variations in T2* that form the basi s of fMRI are caused inhomogeneity produced by deoxygenated hemoglobin. Because the hemoglobin mo lecule contains an iron atom at its core, its magnetic properties depend on whether the ir on is exposed. Oxygenated hemoglobin has a concealed iron molecule and is diamagnetic, causing no magne tic effects. In deoxygenated hemoglobin, the iron is exposed, making the mol ecule paramagnetic and capable of producing inhomogeneity. The net result is a decrease in MR signal in the region of blood vessels containing deoxyhemoglobin. This effect was descri bed in detail by systematically varying blood

PAGE 36

36 oxygenation in rats (Ogawa et al., 1990). When perfused with oxygena ted blood, vessels are indistinguishable on MRI from the surrounding brain tissue, but when perfused with deoxygenated blood, vessels become dark. Ogawa et al. named this effect the blood oxygenation level-dependent, or BOLD, contrast. BOLD co ntrast was altered by changes in inhaled carbon dioxide, blood glucose level, a nd level of anesthesia, indicating that it is determined by both cerebral blood flow (supply) and cerebral metabolism (demand). From these data, Ogawa et al. predicted that BOLD contrast could be detectable within phy siological parameters for blood oxygenation, and that BOLD MRI could be used as a complement to PET for imaging functional brain activity. Shortly after the discovery by Ogawa et al., the first fMRI studies appeared. Changes in the BOLD signal were induced in the occipital lobe us ing an alternating visual stimulus, and in the central sulcus using alterna ting hand movements (Kwong et al ., 1992). While these experiments measured responses to extended periods of stim ulation, a later experiment demonstrated that changes in BOLD signal were detectable to visu al stimuli lasting only two seconds (Blamire et al., 1992). The change in BOLD signal was dela yed about 2-3 seconds after the stimulus, a delayed response known as the BOLD hemodynam ic response (HDR). The HDR is thought to reflect an increase in blood oxyge n that occurs, after a delay, in response to neural activity. The BOLD HDR is now known to have reasonably predictable characteris tics. It begins to rise about 2-3 seconds after the start of the stimulus, and falls about 10-15 seconds after the end of the stimulus. Thus the temporal characterist ics of neural responses are delayed and smoothed over time in the BOLD HRD. However, the char acteristics of the BOLD HDR are sufficiently predictable to allow detection of neural events that are less th an 10-15 seconds apart. The BOLD responses to two visual stimuli presented two sec onds apart appear to add roughly linearly (Dale,

PAGE 37

37 and Buckner, 1997). Subtracting the response to a single stimulus from the response to two stimuli revealed that the remaining response to the second stimulus was comparable to the response to the first stimulus. For most fMRI an alyses, it is assumed that BOLD responses are invariant over time, and between brain regions, and that they add roughly linearly. This allows the use of General Linear Modeling, and even t-related analyses, as discussed below. Several studies have sought to explain in de tail the relationship between neural activity and the BOLD HDR. This question was addressed directly in a study using monkeys in which BOLD signal and electrical responses we re recorded simultaneously (Logot hetis et al., 2001). In this study, a novel MRI-compatible elect rical recording system was used to demonstrate that increases in the BOLD signal were indeed relate d to increased electrical activity. Specifically, the main driver of the BOLD HDR was the local fi eld potential. This electr ical signal represents the sum of postsynaptic events at the recording site, or the net input to the neuron. A number of theories propose to explain the link between neural activity and increased oxyhemoglobin. If increased cerebral blood flow ma tched increased neuronal metabolis m, then the concentration of blood oxygen would remain constant, and no BO LD response would be detectable. This mismatch between supply and demand that produces the BOLD response may reflect an overcompensation that anticipates future increa ses in demand, or may be characteristic of anaerobic metabolism. The latter view was propos ed by Shulman et al. (2001) in their astrocyteneuron lactate shuttle model. In this model, anaerobic metabolism is required for the rapid clearance of glutamate from the synaptic cleft following a burst of firing (Shulman et al., 2001). This model supports the findings of Logothetis et al. (2001) because it links the BOLD HDR to postsynaptic activity.

PAGE 38

38 In summary, fMRI generates images of the brain in which the intensity of signal is modulated by changes in blood oxygenation. The BO LD responses to primary visual and motor stimuli occur in well-established visual and mo tor regions of the brain (Blamire et al., 1992;Kwong et al., 1992). However, the location of BOLD responses probably does not reflect the location of firing neurons, but more likely reflects post-synaptic act ivity (Logothetis et al., 2001). Despite the sluggish nature of the BOLD response, individual responses appear to add linearly, and thus may be modeled even in response to brief stimuli. A number of experimental designs and analysis techniques ha ve been developed in order to increase certainty that the detected BOLD response reflect s task-related brain activity. Basics of fMRI Paradigm Design The primary goal when designing an fMRI paradi gm is to evoke a response that will be distinguishable from noise. Task-related response s are detected as differences in BOLD signal during task and control conditions and these differences are typi cally small, around one percent. Task-related responses must be distinguished from changes in signal due to non-task-related factors (noise). Thermal noise or intrinsic noise is the unavoidable, random variation in the signal due atomic vibrations within system components. Scanner drift is a gradual, monotonic decrease in signal, which may occur due to sl ow changes in temperature in the scanners magnetic coils. Physiological noise may arise from visceral motilit y, breathing, or the heart beat. Motion artifact occurs due to the movement of the head during scanning. Finally, fMRI signal changes may be due to non-task-related neural activity Thermal noise is unavoidable, and is overcome by choosing a stimulus that will evoke as large a response as possible. Monotonic sources of noise, such as scanner drift, are ov ercome by repeating the stimulus on/off conditions several times. Thus, task-related signal changes w ill be distinguishable by their periodic nature. Physiological noise, however, is also periodic To avoid this noise, the frequency of task

PAGE 39

39 presentation must be chosen such that is dist inct from the frequencies at which physiological noise occurs. Practically, this means that peri ods of stimulation should last between 2 30 seconds. Motion artifact is difficult to avoid, but maybe overcome by immobilizing the subjects head. Any surviving motion may be corrected du ring later analysis, and if the motion is too large, the subjects data are discarded. Since an y movement of the jaw in creases motion artifact, fMRI tasks involving speech are difficult to anal yze, and where possible subjects should make task responses by manual button presses. Non-task -related neural activity is the most difficult variable to control. Analyzing data from a larg er group of subjects theoretically maximizes taskrelated activity, and minimizes nontask-related activity, which is assumed to be different for each individual. It is important to choose the co ntrol conditions carefully, so that the comparison used to search from brain activity truly re flects the cognitive functions of interest. Two types of fMRI paradigms are commonl y employed, and both are used in this dissertation. These are the block design and the event-related design In the block design, a series of trials of the same type are presented consecu tively, followed by a series of trials of another type, or a rest period. Blocks of each stimulus ty pe are usually alternated several times using the so-called boxcar design This repetition avoids false signal de tection due to scanner drift. In the event-related design, trials are pres ented in random order, either mixed together or separated by a period of rest. The advantage of the block design is that it produces a large, easily detectable signal. The disadvantage of the bl ock design is that the subject is able to anticipate upcoming trials, since the series of trials are all alike. The event-related design overcomes this expectancy effect by randomizing trial types. Furthermore, because trials are analyzed individually, eventrelated designs allow analysis based on the subject s responses to individua l trials, such as error rate or response time. The disadvantages of the event-related design are that the BOLD responses

PAGE 40

40 to individual trials are small, re quiring many repetitions of each tr ial type to produce a detectable response, and that the BOLD responses to ev ent-related tria ls may overlap. One way of overcoming this overlap is to space events > 15 seconds apart. Alternatively, both these hemodynamic issues may be overcome by jittering th e timing of event pr esentation. By jittering the delay between events, the variance in th e resulting BOLD HDR beco mes larger the more rapidly events are presented (Bur ock et al., 1998). This increases the detectability of the BOLD response, and allows a greater number of repeti tions of each event type. Therefore the block design is appropriate if the neural response to the task should not va ry when the trial is expected. The event-related design is pref erred when expectancy effects must be avoided, or when the study is investigating differences in neural activity with different responses to the task, such as error trials vs. correct trials. Basics of fMRI Analysis What most fMRI analyses ha ve in common is that they create maps of the brain representing differences between the task and control conditions. These differences are usually represented as statistic al values resulting from a comparis on of the BOLD response to each condition. Plotted as a colored overlay on an anatom ic image of the brain, these statistical maps create the well-known colored clusters of activa tion that illustrate f unctional imaging studies. This section explores how these clusters are generated, in order to show that they do not represent a direct photograph of br ain activity, but instead are inform ed by a series of statistical decisions made by the experimenter. Before statistical maps are created, fMRI data are preprocessed. In studies of groups of subjects, the images from each subject are orie nted within a standard space defined by the midline of the brain, the line between the anteri or commissure and the posterior commissure, and the outermost surfaces of the cortex (Talair ach, and Tournoux, 1988). The data from each subject

PAGE 41

41 may be corrected for motion. This correction ope rates by comparing the whole-brain image at each time point with the initial time point, and then minimizing differences using iterated small rigid-body transformations of the later time point. The correction process re sults in a series of rotation and translation values th at allows head movement to be approximated and thus allows the exclusion of data with large motion. Although early fMRI experiments created statis tical maps by applying t-tests to the raw signal during each task condition, the experiments in this dissertation were analyzed by modeling an estimated BOLD HDR. This approach uses General Linear Model (GLM) statistics. The GLM may be applied using the fixed or random effects approaches. Both are used in this dissertation. The fixed effects approa ch is applicable to small data sets, but is susceptible to one highly-responsive individual dominating the re sult. The random-effects approach is more conservative, detecting consistent changes across the group, but requ ires larger sample sizes. In both approaches, the analysis begins by m odeling the BOLD HDR to each task condition. Although this modeling approach may approximate a square wave in slow block designs, and thus represents only a slight improvement over th e t-test, modeling is particularly important in event-related designs, where there are greater di fferences between the timing of stimuli and the timing of the modeled BOLD HDR. Based on stimulus timing, the BOLD HDR may be estimated using a standard response function (Boynton et al., 1996). The modeled response, or reference time course may then be fit to the fMRI si gnal at each voxel. After estimating a baseline value, the magnitudes of the reference time courses for each task condition are varied until the difference between the model and the data is minimized, using the partial least squares approach. The estimate of the response magnitude is called the beta weig ht. The solution to the general linear model for a given voxel is given in Equation 2.

PAGE 42

42 S(t) = x Mx(t) + y My(t) + e(t) Where S = MR signal, t = time, task = beta weight for task x or y, M = modeled hemodynamic response for task x or y, a nd e = error, or baseline. The modeled response for each condition is identical for each voxel, being determined by the timing of the paradigm. The error term is the same at each voxel, being based upon the signal during a baseline or resting task condition. Statisti cal maps are derived fr om the beta weights. In the fixed effects approach, a single beta weig ht (for each task condition at each voxel) is calculated for the whole group. The statistical si gnificance of the response to each task condition (or the contrast between selected task conditions) is calculated using the magnitude of the beta weight (or contrast) and its standard error. Th e standard error is based upon the differences between model value and actual si gnal value, or residuals. Th e mixed effects approach is vulnerable to bias resulting from a strong res ponse from a single individual. This bias is overcome by using the random effects approach, in which individual beta weights are calculated for each task condition at each voxel, one for each subject. The statistical significance of the response to each task condition (or the contrast between selected task conditions) is then calculated by performing a t-test on the sample of beta weights. Although this approach is less susceptible to contamination from one highly-responsive individual, it requires a larger sample size. In the fixed effects analysis, the number of degrees of freedom is determined by the number of time points in the group data. This is the number of subjects multiplied by the number of data points in one fMRI run, which usually is in th e thousands. In the random effects analysis, the number of degrees of freedom is determined by the number of subjects only. Thus to reach a given level of significance, a hi gher statistical score must be obtained in the random effects analysis compared with fixed effects.

PAGE 43

43 In both approaches, statistical scores are ove rlaid upon an anatomical image of the brain. Typically, the statistical values are represented by a range of colors. To make the map more readable, a threshold is usually applied to eliminate areas where the response was less significant. The choice of statis tical threshold is the topic of ongoing debate in the fMRI literature. The main issue involves correction for multiple comparisons. Since a statistical test is applied at every voxel, then for every 100 voxels 5 will have a significa nt response at p < 0.05. In fact, there are over 10,000 voxels, so we woul d expect at least 500 false positive voxels. The choice of statistical threshold, and method fo r correction for multiple comparisons, will be discussed in each studys chapter. In summary, fMRI analysis produces brain maps of statistical values. In this dissertation, these values are calculated by f itting a model of the hemodynamic response to the data at each voxel, and then performing statistics upon the mode l. This may be done using a fixed effects approach or a random effects approach. The former is better suited to small sample sizes, but is vulnerable to contamination by a single, highly -responsive individual. Th e latter is a more conservative approach, requiring larger sample si zes. Because statistical maps represent highly processed information, and because there are multip le opinions about how best to decide which clusters of activation are significant, an impor tant skill in fMRI analysis is interpretation. Interpreting fMRI results As noted above, there are various sources of noise in fMRI da ta. Although a statistical map may reveal clusters of significant signal change, these may or not reflect th e neural correlates of the task condition being tested. This is immediately obvious when activation is seen outside the brain. Significant signal change s outside the brain may occur due to task-related head or eye movement. The former is particularly prevalent at the borders of light and dark regions, such as the skull, the edges of the brain, or large sulci. It may be possi ble to identify motion and other

PAGE 44

44 physical noise effects (like arterial pulsation) by examining the BOLD response at each cluster of activation. Unlike physical noise effects, however, non-task-relat ed neural activity have not have a pattern that can distinguish it from task-relat ed neural activity. It ca n only be eliminated by careful task design and by having an a priori hypothesis about the brain regions involved in the task. The approaches given below are each applied in at least one of the experiments in this dissertation in an attempt to increase the certai nty that the apparent fMRI activations indeed reflect use of the hypothe sized cognitive function. Examining the BOLD Response to Exclude False Activation As explained above, the clusters of activati on illustrated on fMRI statistical maps are indirect indicators of brain act ivity. In order to confirm whet her the underlying MR signal changes support the conclusion of the statistical tests, a retrospective examination of the BOLD HDR at each cluster of statistica l activation may be performed. This is achieved by averaging the responses to each task condition within subjec ts, and then averaging the responses across subjects. The MR signal for each response is norma lized to a percentage change from baseline before averaging. In block design paradigms, th e response should rise and fall within the time window for averaging th e BOLD HDR. In event-related paradigms, the time window for averaging the BOLD HDR to a given event may incl ude one or more subsequent events. In order to minimize the influence of subsequent events the delay between events is randomly jittered. Once the group average BOLD HDR has been ca lculated for each task condition, certain judgments may be made about the re gion where the response occurred. The expected BOLD response to a stimulus is a smooth curve that ri ses approximately 2-3 seconds after stimulus onset, and then falls a pproximately 10-15 seconds after the stimulus ends. In reality, there is a large amount of variability in the BOLD responses between individuals, and within individuals, both in time (between repeat ed blocks, or different scanning sessions) and

PAGE 45

45 between brain regions (Aguirre et al., 1998;Me nz et al., 2006;Miez in et al., 2000). One consequence of this is that the BOLD HDR may reve al patterns in the responses to different task conditions that are missed by the GLM, because of discrepancies between the modeled and actual response. Post-hoc analysis of the av erage BOLD response may confirm whether these patterns are statistically significant. Several unexpected patterns of BOLD response may be seen. First, the signal may be inverted. An apparent activation on a statistic al map may be due to a decrease in BOLD signal during the control condition, while signal during the task does not cha nge. Deactivation during task performance may implicate a brain region in cognitive processes that occur during the control condition. This eff ect has been reported consistently in a set of brain regions including the anterior and posterior cingula te cortices, and the bilateral angular gyri (Shulman et al., 1997). Because these deactivations are deeper with increasingly demandi ng tasks, they are taken to represent mental functions performed at rest th at require attention (M cKiernan et al., 2003). A second unexpected pattern of BOLD response is a steadily rising curve. This may represent changes in tonic activation of a brain region, or may represent noise effects, such as task-related head motion. One indication that a st atistical activation is due to a noise effect is a high degree of variability in the BOLD response, as measured by the standard error of each time point. Further examination of individual subjects responses ma y reveal that a large jump in signal in one individual led to a statis tically significant result. Using Control Conditions to Test Specific Cognitive Components Having ruled out physical noise effects, the i nvestigator must ask whether a particular cluster of activation corresponds to a partic ular cognitive function. The earliest fMRI experiments compared simple visual stimuli with a resting condition: a blank screen. In a more complex task, for example a face recognition tas k, a more detailed control condition may be

PAGE 46

46 required. For example, to look for specific neural correlate of face recognition, an experiment may include a condition in which subjects must recognize common household objects. In addition to determining the response to face r ecognition and object recognition versus rest, the two conditions may be compared with each other. This is achieved in the GLM approach by looking for significant differences between the beta weights for each task condition. This subtractive method assumes that subtracting object rec ognition from face recognition eliminates any brain regions that respond equa lly to both tasks, and are ther efore not specifically activated by face recognition. Although the subtractive method has weaknesses, as discussed below, this approach can be used successfully provided the cognitive factors involved in the compared tasks are considered carefully. Using Factorial and Parametric Designs to Overcome Limits in the Subtractive Approach The subtractive method may be used to finely dissect a particular c ognitive function, but it may be inadequate due because it falsely assu mes that the neural co rrelates of cognitive functions add by pure insertion In a serial subtractive design, task conditions are designed in which cognitive components are added one by one. Under the assumption pure insertion, differences between each task c ondition will reflect processing of the new cognitive component. However, if pure insertion is not true, the a ddition of a new cognitive component may alter the way the brain handles the existing components. Because there may be interaction between cognitive components, factorial analysis may be used to account for this interaction (Friston et al., 1996). This approach was illustrated by Fristo n et al. (1996) by investigating the neural correlates of object recognition and phonological retrieval in an object naming task. The authors tested the hypothesis that the inferior tem poral activation during object naming reflected phonological retrieval. They did so using both a se rial subtractive method and using a factorial method. The serial subtraction experiment invol ved three conditions: task A, saying yes in

PAGE 47

47 response to a colored shape, task B, saying yes in response to a colo red object, and task C, naming a colored object. Task A involved visual analysis and speech, task B involved visual analysis, speech, and object r ecognition, and task C involved vi sual analysis, speech, object recognition, and phonological retrieva l. The responses to these task conditions were examined using PET. The inferior temporal region was activated when subtra cting A from B, but not when subtracting B from C. This implied that this region was involved in object recognition, but not phonological retrieval. In the factorial design, a ta sk D was added: naming the color of a colored shape. Task D involved visual analysis speech, and phonologi cal retrieval, but not object recognition. The PET responses to these four ta sk conditions were anal yzed using a two-way ANOVA, allowing the detection of a main effect of object recognition (B & C vs. A & D), a main effect of phonological retrieval (C & D vs A & B), and an interaction between the two. The results showed that at the inferior tem poral lobe, there was both a main effect of phonological retrieval and an interaction betw een phonological retrieval and object recognition. That is, the response to object recognition varied with the addition of pho nological retrieval, and vice versa. The interaction demonstrated that for this experiment, the assumption of pure insertion was not true by revea ling a response in the inferior te mporal lobe to object recognition that was missed using the subtractive approach. T hus factorial designs can be used to examine separately the neural correlates of cognitive components that may interact. Like factorial designs, parametric designs attempt to overcome the problem with cognitive subtraction. They do so by comparing the MR signa l to a task parameter with multiple values. This allows the identification of brain regions whose activity covaries with the task-related parameter, rather than regions that are merely more active during the task. For example, one experiment regions whose respons es covaried with word presen tation rate (Buchel et al., 1998).

PAGE 48

48 Whereas all words activated bila teral frontal regions irrespecti ve of rate, presentation rate predicted responses in bilatera l occipitotemporal regions. Th is shows that investigating covariations of fMRI responses with experime ntal parameters may finely dissect neural correlates of one task parameter. In this case, the parametric approach distinguished set-related activity during word reading from individual stimu lus perception. Parametric analysis may also be used in experiments involving complex stimuli, such as photographs, to separate the effects of interest from other confounds pres ent in the complex stimulus. This approach was applied in this dissertation using emotional rati ngs of photographs as a parameter in analyzing fMRI data.

PAGE 49

49 CHAPTER 3 PRELIMINARY DATA: DISSECTING THE NEURAL CORRELATES OF DISGUST This experiment investigated the ability of fMRI to distinguish nuances in the negative emotional content of picture stimuli. Previous studies showed that th e amygdala and occipitotemporal cortex (OTC) respond to pictures ra ted as emotionally arousing. Responses in the amygdala to fearful faces have been dissociated fr om responses in the insula to disgusted faces. It is not clear whether the amygdala/OTC and insula respond selectively to arousing and disgusting pictures. In th is fMRI study, healthy volunteers viewed pictures of contamination, human mutilation, threat, and neutral scenes durin g scanning, and then rated pictures for the basic emotions, including dis gust. The anterior insula re sponded to contamination and mutilation but not threat, while the OTC responded to threat and mutilations more strongly than contamination. The above activatio ns were predicted by disgust a nd arousal ratings respectively. Additionally, mutilations uniquely activated the right superior parietal cortex. No response was detected at the amygdala. These results support se lective disgust processi ng at the insula, and suggest distinct neural responses to contamination and mutilation. The use of fMRI in investigating components of emotion processing is feasible, but subs equent studies should improve imaging of the amygdala. Introduction The role of the insula in affective processi ng has been the topic of recent debate. Case reports from patients with insula lesions and Huntingtons diseas e describe impaired recognition of facial expressions of disgust, and in some cases impaired ability to feel the emotion of disgust itself (Adolphs et al., 2003;Calder et al., 2000;Sprengelmeyer et al ., 1996). In healthy volunteers, functional brain imaging experiments using faci al expression recognition tasks have indicated selective activation of the insula to facial expressions of disg ust (Phillips et al.,

PAGE 50

50 1997;Sprengelmeyer et al., 1998). However, results from experiments using pictures such as body products to induce disgust, que stion the notion of selective di sgust processing at the insula. Although an insular response to di sgust-inducing pictures was re ported in a study of obsessivecompulsive disorder (OCD) patients (Phillips et al., 2000), two subsequent studies of healthy controls found equal activation of the insula to both disgusta nd fear-inducing pictures (Schienle et al., 2002;Stark et al., 2003). Su rprisingly, disgust-inducing pi ctures activated the amygdala more than fear-inducing pictures in these studies; the authors e xplain that their disgust-inducing pictures elicited high disgust ratings, while thei r fear-inducing pictures elicited only moderate fear ratings. A recent study in our lab extended the previous OCD study (Phillips et al., 2000) by adding fear-inducing pictur es (Shapira et al., 2003). In th is case, we found greater insula activation for disgust-inducing pictures than f ear-inducing pictures in both healthy volunteers and patients with OCD. Considering the discrepa ncy between our results a nd those of the studies above (Schienle et al., 2002;Star k et al., 2003), we decided to ex amine potential differences in study methodologies. The former studies used pict ures of contamination (e.g. spoiled food and body products) and mutilations (e.g. injuries and corp ses) to induce disgust, whereas ours used only pictures of contamination. We therefore desi gned a study to test separately the effects of these two types of pictures. The insula has been associated with a range of functions, including vi sceral and gustatory processing (Wicker et al., 2003), autonomic regu lation (Critchley et al., 2003), and selfgenerated affective experiences (P han et al., 2002); thus a general affective or disgust-specific role for the insula are both plausible. Schienle et al. (2002) suggest th at a shared affective pathway is sufficient to explain the insular res ponse to affective pictur es. The current study re-

PAGE 51

51 examines this conclusion by presenting pictures of contamination and mutilation separately to test whether the insular response is specific to food-related disgust. We hypothesize first that the in sula responds selectively to disgust-inducing pictures, and second that there is a distinct neural response to pi ctures of contamina tion and pictures of mutilation. The first hypothesis predicts that pictur es of contamination or mutilation or both will cause greater activation at the in sula than fear-inducing pictures The present study therefore uses functional magnetic resonance imaging (fMRI) to compare these two types of disgust-inducing pictures with fear-inducing and ne utral pictures, in order to a ssess the validity of combining pictures of mutilation and contamination in a single category. In addition to performing standard exploratory analysis using stat istical activation maps, we exam ine the neural response to the affective pictures in detail usi ng signal time-courses from select ed regions of interest (ROIs). Methods Subjects Eight healthy volunteers (4 male) aged 20-26 ga ve written informed consent in accordance with a protocol approved by the Institutional Re view Board at the University of Florida. According to self-report, 7 were right-handed, and 1 was left-handed. The volunteers denied taking any psychiatric medication at the time of the scan and gave no history of psychiatric or neurological disorders. Disgust Picture Paradigm Pictures were selected from the Internationa l Affective Picture System (IAPS) (Center for the Study of Emotion and Attention [CSEA-NI MH], 2001) and were divided into four categories: contamination, mutilation, threat and neutral which were defined as follows: contamination pictures depicted of scenes associ ated with poor hygiene or poisons (e.g. spoiled food, bodily waste, garbage, pollution); mutilation pictures showed human injuries or disease

PAGE 52

52 (e.g. murder victims, traffic accidents, tumors, bi rth defects); threat pictures showed imminent attacks (e.g. humans with guns or knives, dogs, sn akes); and neutral pict ures depicted various scenes with low arousal and medium pleasur e ratings (Lang et al., 2001) (e.g. landscapes, household tools, non-threatening animals). The stimuli were presented using an Integr ated Functional Imaging System (IFIS, MRI Devices, Inc., Waukesha, WI) w ith a 7 LCD screen at 640 X 480 pixel resolution, mounted over the subjects head and viewed using a fixed mirror. The screen subtended approximately 14 x 11 of the visual field. A PC running E-Prim e (Psychology Software Tools, Pittsburgh, PA) began presenting each task in synchronization with the first RF pulse of each scan. Each emotion category was presented during a separate MRI scan in order to avoid the fa tigue or boredom that may result from viewing a single, long sequence. The order of the runs was randomized for each participant. Each run consisted of six alternating emotional and neutral picture blocks (21 sec long), interspersed with 9-sec ond fixation blocks. Each block c ontained 14 pictures selected randomly (without replacement) from the list for that category, and each picture was presented for one second, followed by 0.5 seconds of fixation. Participants were instructed to view the pictures passively, keep their eyes open, and to avoid repressing or exaggerating their emotional response. After scanning, each participant rated 15 randomly chosen pictures on a scale from 1 to 5 (5 being the most intense emotion) for each of the following basic emotions: happiness, sadness, fear, anger, disgust a nd surprise (Ekman, and Friesen, 1976). Dimensional ratings were taken from the normative set provided with th e IAPS. These were ratings from 1 to 9 for pleasure, arousal, and dominance, with 9 indica ting the viewer felt most pleasant, most aroused,

PAGE 53

53 and most dominant respectively. Finally, each participant completed a 32-item questionnaire designed to indicate their sensitivit y to disgust (Haidt et al., 1994). Functional Imaging Data Acquisition MR images were acquired using a 3 Tesla Al legra system (Siemens, Munich, Germany). Anatomical imaging used a standard MPRAGE se quence with a 240mm square field of view at 256 256 pixel resolution in the axial plane, and 160 slices of 1.0-1.4 mm thickness. Functional data were gathered using echo-planar imaging (EPI) sensitive to blood oxygen level-dependant (BOLD) signal (TR = 3000ms, TE = 30ms, flip angle = 90 FOV = 240mm, matrix = 64 64). Twenty-four slices were collect ed in the axial plane with a 6 mm thickness and 0 mm gap. Each functional run lasted 3 min 9 sec and consiste d of 63 volumes, the firs t two of which were discarded before analysis due to their T1 saturation. Functional Imaging Data Analysis Data were analyzed using BrainVoyager 2000 v. 4.9.6 (Brain Innovations, Maastricht, Holland). Each anatomic scan was normalized to Talairach space, and the transformation parameters saved. The in-plane functional images from each participant were then co-registered with the pre-transformation anatomic scan, a nd converted into a 3D volume time-course in Talairach space using the saved transformation parameters. Finally, the 3D functional data underwent 3D motion correction and linear trend removal. Voxel-wise statistical activation maps were generated using a gene ral linear model (GLM) in which the predictors were an estimated hemodynamic response to each emotional condition. Contrasts between predictors were used to calcul ate the relative contributi on of each predictor to the variance in the BOLD signal. Unless otherwise stated, the statistical threshold was set to p <

PAGE 54

54 0.05 with Bonferroni correction for multiple comparisons, and th e minimum cluster size was 100 mm3. Region of interest (ROI) analyses were performed within select ed clusters of significantly activated voxels. Within each ROI, the BOLD responses for each condition were visualized using time-locked averaging of the percentage signal change relative to fixation. A GLM was calculated for the mean signal from the ROI, and the modeled amplitude of each predictor (the beta weight) was used to describe the size of the hemodynamic response. Unlike the statistical activation value, which reflects how well the model fits the data, the beta weight describes the BOLD response, which is assumed to be proporti onal to neural activation (Ogawa et al., 1992). Results Emotion Ratings All three emotional conditions were rated as evoking significantly less pleasure, more arousal, and more dominance than neutral according to the mean IAPS scores for each picture set (Lang et al., 2001) (Tables 3-1 & 3-2 ). Mutilation was less pleasant than contamination and threat, and contamination was less arousing than threat and mutilation. According to our own subjects ratings of basic emotions, the threat condition elicited more fear than the other three conditions and the contamination a nd mutilation conditions each elicit ed more disgust than threat and neutral. The mutilation condition also elicited more sadness than neutral. (For all the above comparisons p < 0.001, corrected for multiple comparisons.) The mean standard deviation disgust sensitivity score was 13.4 4.0 (males: 13.3 3.4, females: 13.6 5.2). The mean for American adults is approximately 16 (mal es: 14, females: 18) (Haidt et al., 1994). fMRI Data Exploratory statistical activ ation maps were generated by contrasting each emotional condition with neutral using the GL M (see Methods: data analysis). Figure 3-1 illustrates clusters

PAGE 55

55 of activation seen in the anterior insula a nd occipito-temporal cortex (OTC). See Tables 3-3 3-4 & 3-5 for a full list of activated regions. The anterior insula was activated bilaterally in both the contamination and mutilation conditions. No significant activat ion was found in the insula for the threat condition at the threshold p < 0.05 corrected. The extent of activation in the OTC increased in from contamination to threat to mutilation respectiv ely. Activation for mutilation extended into the midline occipital cortex and posterior cingulate and was additionally s een in the thalamus, ventral striatum, superior parietal co rtex and several prefrontal regions (Table 3-5 ). No significant signal changes were found at the amygdala, but detailed examination of the EPI (functional) images revealed lo ss of signal at the amygdala due to susceptibility artifact. Comparisons between emotions revealed no uni que activation for contamination or threat, but mutilation condition activated the right superior parietal cortex. The contrasts (contamination threat) and (mutilation threat) eac h showed activation in the left anterior insula at a threshold of p < 0.0001 uncorrected, but this did not achieve the stricter threshold of p < 0.05 corrected. Clusters of activation for ROI analysis were se lected from those contra sts showing significant differences between emotional c onditions; thus the insula ROI was derived from the contrast [(contamination + mutilation) neutral] and the OTC ROI from the contrast [(mutilation + threat) neutral]. The left and right ROIs were co mbined for analysis. The right superior parietal ROI was selected from the contrast [mutilation (contamination + threat)]. All three ROIs are illustrated in Figure 3-2 Time-locked averaging of the BO LD signal across conditions (see Figure 3-3 A-C ) showed a phasic response to all picture conditions (inclu ding neutral) in the OTC. This response was enhanced in the emotional conditions: the enhan cement was smallest for contamination, greater

PAGE 56

56 for threat, then greatest for mutilation. At the in sula, viewing neutral pictures evoked no change in signal, but contamination and mutilation ag ain caused a phasic increase. Viewing threat pictures caused a small response, although this failed to reach the thre shold for statistical significance during the e xploratory analysis (Figure 3-1 ). Signal in the righ t superior parietal cortex increased in response to mutilation pictures but was indistinguishable from neutral during the other conditions. The widespread activation for mutilation pictures (see Figure 3-1 ) may reflect the high affective arousal ratings for these pictures, particularly the amplit ude of the signal increases in the OTC; also, activity in the anterior insula suggested a re lationship with the disgust rating. We therefore tested the correlations between pictur e ratings and BOLD signal change, represented by beta weight (see Methods: Data analysis). Since the experi mental design did not include comprehensive picture ratings for each subject, the ratings were pooled across subjects. Arousal rating predicted signal change in the OTC [r2 = 0.98, p < 0.05], and disgust rating marginally predicted signal change in the anterior insula [r2 = 0.85, p = 0.08] (see Figure 3-3 D+E ). The complementary correlations were not significan t: disgust rating with OTC signal change [r2 = 0.61, p = 0.22] and arousal rating with insular signal change [r2 = 0.28, p = 0.47]. OTC signal change was also predicted by rati ngs for happiness, pleasure and do minance, but these were each correlated with the arousal rating [respectively, r2 = 0.97, 0.85 & 0.999, p < 0.05, p = 0.08 & p = 0.0005], suggesting that, in this case, these rati ngs are confounded with a common factor. The disgust rating did not correlat e significantly with any other ra tings. Each subjects disgust sensitivity score was compared w ith that individuals signal cha nge in the OTC and insula for each emotional condition, but no significant correlations were found.

PAGE 57

57 Discussion The aim of this study was to compare the ne ural responses to tw o potentially different types of disgust. Contrary to previous studies comparing disgusta nd fear-inducing pictures, (Schienle et al., 2002;Stark et al ., 2003) we found that disgust si gnificantly activat ed the insula while fear did not. Furthermore, we showed that the insular response correlated with feelings of disgust, but not with feelings of arousal. Secondly, we showed distinct neural responses to viewing pictures of contamination and mutilati on. Specifically, viewing pictures of mutilation caused greater activation of the OT C, and unique activation of the ri ght superior parietal cortex. The data presented here are insufficient to explain the failure of two previous studies (Schienle et al., 2002;Stark et al., 2003) to find a specific insular re sponse to disgust in terms of the effect of combining pictures of contamina tion and mutilation, since the insula responded to both conditions. It is possible that our small (and statistically nonsignificant) insular response to threat was because our pictures evoked less fear than those of the other two studies. A comprehensive comparison of picture ratings betw een studies is not possible here but the fear ratings for our threat picture set (2.7 out of 5, equivalent to 4.8 out of 9) are close to those of Schienle et al. and Star k et al. (5.5 and 4.8 out of 9 respectively). Furthe rmore, if we are to accept the interpretation that activity in the insula reflects a shar ed affective system, then our study should have shown greater acti vity in the insula to threat than to contamination, since the threat pictures were rated as more arousing a nd less pleasant than contamination. One possible explanation is that 1.5 Tesla MRI (used in the pr evious studies) is not sufficiently sensitive to BOLD effects to detect the relatively small differe nces between the fear a nd disgust responses at the insula that are dete ctable at 3 Tesla (used in the current study). Previous studies have also s uggested that activation of the OTC is influenced by emotional intensity (Lang et al., 1998;Schi enle et al., 2002;Shapira et al ., 2003;Stark et al., 2003). These

PAGE 58

58 visual areas do not encode emotion, but receive feedback from emotion-processing regions such as the amygdala (Rolls, 1999). Thus, although we failed to image the amygdala in this study, enhancement of ventral visual processing may be thought of as a proxy for amygdala activity (Sabatinelli et al., 2005). The most compelli ng evidence we found for a specific response to disgust in the insula is found in the correlations between disgust rating an d insular response, and arousal rating and occipi to-temporal response (Figure 3-3 D+E ). These suggest a double dissociation between the insula, processing info rmation related to disgust, and the OTC, processing general affective arousal. These findi ngs are compatible with the existence of a common affective pathway, but sugges t that this simple model is insufficient to explain activity at the insula. Activation of the insula by both mu tilation and contamination pictures suggests that the insular response to disgust is more related to the emotional feeling of disgust rather than the gustatory content of th e eliciting stimulus. An electrical recording study in humans provides support for a late (300 ms) response to emoti onal stimuli at the insu la, likely reflecting a conscious feeling rather than earlier processi ng of gustatory content (Krolak-Salmon et al., 2003). We recognize several shortcomings of this study. Since we were unable to image the amygdala, we had to use occipito-temporal activ ation as a proxy for the amygdala response. Although in this study, activity in the insula was not correlated with affec tive arousal, the insula influences autonomic arousal (Critchley et al ., 2003), and we cannot rule out the insulas influence on occipito-temporal activity. The affective ratings of our picture sets may be confounded with other features unique to each se t, such as the lack of human faces in the contamination set, or the abundance of the color red in the mutilation set. Future studies should use imaging parameters able to image the amygdala, take physiological measures of arousal

PAGE 59

59 (such as heart rate and skin conductance) a nd specifically account for possible confounds during selection of picture sets. (It s hould be noted that not all stud ies report confirmation of proper amygdala imaging, and that artifact is common at higher magnetic fields, i.e. 3 Tesla (Merboldt et al., 2001).) The unique activation of the right superior pa rietal cortex by mutilation pictures is an interesting, new finding that should be further explored by future studies. A previous case study proposed a parietal pathway for processing acte d-out emotions (Adolphs et al., 2003). This pathway may be more responsive to mutilation pictures if the viewer processes them by mentally re-enacting the bodily condition of the victim in the picture. This view is further supported by studies locating mirror neurons for bodily actions in the parietal cortex (Buccino et al., 2004). Whether mutilation pictures evoke a distinct emotion, for exampl e horror, is an interesting question for further study. It has been suggested that horror is a blend of disgust and fear, and it is interesting to note that mutilation may be viewed for pleasure in art or entertainment (McNally, 2002). In conclusion, our findings sugge st that the OTC and the insula process different affective information, as reflected by arousal and disgust ratings respectively. M odulation of occipitotemporal activity by feelings of arousal is we ll modeled by the concept of a shared affective network processing basic affective dimensions. Howe ver, the apparently disgust-specific activity in the insula supports the idea th at emotional categories may have distinct neural representations. We also suggest that future studies consider contamination and mutilation pictures separately. Whether mutilation pictures evoke a distinct emotion (perhaps horror) is a question best answered by future research.

PAGE 60

60 Table 3-1 Affective ratings (dimensional) Picture set Pleasure Arousal Dominance Contamination 3.2 0.8 4.9 0.84.7 0.6 Mutilation 1.9 0.6 6.3 0.73.3 0.6 Threat 3.0 0.8 6.3 0.93.3 0.7 Neutral 5.6 0.9 3.4 1.06.0 0.6 Pleasure, arousal and dominance ratings are out of nine, and were taken from the IAPS data.

PAGE 61

61 Table 3-2 Affective ratings (categorical) Picture set Happiness Sadness Fear Anger Disgust Surprise Contamination 1.2 0.7 1.3 0.61.4 0.81.1 0.32.6 1.0 1.3 0.5 Mutilation 1.0 0.0 2.4 1.31.6 0.91.4 0.63.2 1.3 1.6 0.9 Threat 1.0 0.0 1.3 0.52.7 1.01.6 1.01.5 1.0 2.0 1.3 Neutral 1.6 0.9 1.0 0.01.1 0.21.0 0.01.1 0.2 1.2 0.4 Happiness, sadness, fear, anger, disgust, and surp rise ratings are out of five, and were obtained from subjects in the current study.

PAGE 62

62 Table 3-3 Clusters of activa tion for (threat neutral) Region Side BA X Y Z Size t(478) OTC R 3746-62-443148.7 OTC L 37-42-64-328558.0 Parahippocampal gyrus L 36-26-39-3219-6.6 Only clusters >100 voxels shown. L: left, R: ri ght. BA: Brodmanns Area. X, Y and Z refer to Talairach co-ordinates (mm). Si ze: number of 1mm3 voxels. t: mi xed effects statistical score (degrees of freedom). Negative t score denotes decrease relative to neutral. OTC: occipitotemporal cortex.

PAGE 63

63 Table 3-4 Clusters of activati on for (contamination neutral) Region Side BA X Y Z Size t(478) Insula R 1331202571 6.6 Insula / frontal operculum L 13, 47-382701650 7.7 Middle frontal gyrus R 46441723119 6.2 OTC L 37-45-55-81909 7.4 OTC R 3745-49-10554 6.7 Only clusters >100 voxels shown. L: left, R: ri ght. BA: Brodmanns Area. X, Y and Z refer to Talairach co-ordinates (mm). Size: number of 1mm3 voxels. t: mixed effects statistical score (degrees of freedom). OTC: occipito-temporal cortex.

PAGE 64

64 Table 3-5 Clusters of activation for (mutilation neutral) Region Side BA X Y Z Size t(478) Cerebellum B 1-72-24136 6.2 Insula L 13-33221526 6.5 Insula / frontal operculum R 13, 473922-11614 7.4 Medial frontal gyrus R 854036248 6.6 Middle frontal gyrus R 10, 46334613134 -6.1 Midline occipital 17-19, 29-31 0-6796818 8.2 OTC L 37-38-61-108419 10.0 OTC R 3742-60-714555 11.3 Parahippocampal gyrus R 3619-520290 6.1 Parahippocampal gyrus L 36-19-52-3844 7.4 Precentral gyrus R 645-335481 6.3 Superior frontal gyrus B 9-457302165 8.7 Superior frontal gyrus L 8-112457144 6.6 Cuneus L 19-8-8933279 6.4 Cuneus & precuneus R 1925-69398924 9.6 Cuneus & precuneus L 19-23-76321132 6.4 Intra-parietal sulcus L 7-30-5538151 6.1 Thalamus L -6-1513109 6.0 Ventral striatum R 210-4108 6.6 Only clusters >100 voxels shown. L: left, R: righ t, B: bilateral. BA: Brodmanns Area. X, Y and Z refer to Talairach co-ordinat es (mm). Size: number of 1mm3 voxels. t: mixed effects statistical score (degrees of freedom). Nega tive t score denotes decrease relativ e to neutral. OTC: occipitotemporal cortex. : this cluster may not reflect neural activity because it is partly outside the brain and its shape corresponds to the anterior sagittal sinus.

PAGE 65

65 Figure 3-1. Statistical maps showing contrast s between each emotional condition and neutral. Contamination and mutilation activated the anterior insula. The occipito-temporal cortex (OTC) responded to al l three emotional conditions, but comparatively weakly to contamination. Red / yellow: emotion > neutral, blue: neutral > emotion. Green line in inset shows slice a ngle (8 from ACPC). Threshold: p < 0.05, corrected for multiple comparisons, minimum cluster size: 100 mm3.

PAGE 66

66 Figure 3-2. Glass brain view of regions of in terest. Green: insula, or ange: occipito-temporal cortex, red: right superior pa rietal cortex. A: view from front, B: view from left, C: view from top. The contrasts from which th ese ROIs were selected are described in Methods: Functional imaging data analysis. Internal axes denote the anterior and posterior commissures. L: left, R: right, A: anterior, P: posteri or, S: superior, I: inferior.

PAGE 67

67 Figure 3-3. BOLD responses. A) Ther e is an occipitotemporal respon se to all conditions, but this is enhanced in the emotional conditions. B) The insula did not respond to neutral pictures, but showed the gr eatest response to contamination pictures. The small response to threat pictures did not reach significance in the exploratory analysis (Figure 1). C) The superior parietal ROI responded only to mutilation pictures. : contamination, : mutilation, : threat, : neutral. Solid vertical lines indicate start of picture block, dotted vertical lines indicate start of fixation block.

PAGE 68

68 Figure 3-4. Correlations with emotion ratings. For each picture set, BOLD response amplitude (Beta weight) is plotted against A) arous al rating & B) disgust rating. Arousal predicted ventral visual activity, whereas disgust predicted activity in the anterior insula. : occipito-temporal cortex, : anterior insula.

PAGE 69

69 CHAPTER 4 FACE MATCHING AND THE AMYGDALA: BOTTOM-UP EMOTION PROCESSING OR NOT? Previous studies have inves tigated top-down modulation of bottom-up emotion processing in the amygdala using a face matching and labeli ng task. The face matching component of this task has been shown to activate the amygdala reliably, and is thought to elicit bottom-up processing of the facial features that communi cate emotion. However, facial feature matching may also elicit intentional, knowledge-based pro cessing in pursuit of task demands, or top-down processing. Therefore, in order to distinguish facial feature matching from emotion processing, and presumably thereby to dissect an emotiona l response at the amygdala, we modified the face matching task to include an intermediate contro l condition in which neutral faces were matched by identity. The left and right amygdalae responded to both th e emotion and the identity matching conditions, and the only selective response to emotion matching was at the left inferior prefrontal sulcus. The left amygdala response hab ituated to emotion matching but not to identity matching. Although the amygdala has been descri bed as a "fear module", a growing body of work suggests that it is a more general "relev ance detector". We concl uded that the amygdalar response to face matching is driven at least in part by relevance detection, independent of emotion processing, although there appears to be additional emotion-specific processing at the left amygdala. These results suggest that the f ace matching task is not a valid paradigm to investigate bottom-up processing of facial emotion in the amygdala. Introduction The amygdala has been described as an emo tion processing module specialized for fear, enabling both the experience of fear, and the rec ognition of fear in others. While this notion is supported by lesion studies (Adolphs et al., 1994 ) and functional neuroi maging (Morris et al., 1996), other data suggest that the amygdala has a more general role, de scribed as relevance

PAGE 70

70 detection (Sander et al., 2003). For example, th e amygdala also responds to disgusting scenes (Schienle et al., 2002), abstract figures associated with food re ward (Gottfried et al., 2003), neutral faces of a different race (Hart et al ., 2000), and novel neutral faces (compared with previously-viewed neutral faces) (Schwartz et al., 2003). The amygdalar response to facial expressions of emotion is gaining clinical relevance as a paradigm for studying anxiety and affective disorders. An increas ed response to fearful or sad faces has been shown in patients with depres sion (Surguladze et al., 2005) as well as posttraumatic stress disorder (Rauch et al., 2000;Shin et al., 2005). Furthermore, in depression, this hyperresponsivity has been reversed by drug tr eatment (Fu et al., 2004;Sheline et al., 2001). Several of the above investigators have sugge sted a correlation between altered amygdala activity and previous behavioral data showing impaired facial expr ession recognition in patients with depression (Gur et al., 1992). These studies used a variety of tasks, and often did not control for non-emotional face processing. A more integrated picture of the neurob ehavioral alterations surrounding amygdalar dysfunction could be obtained using a well-establishe d task to dissociate emotional and non-emotional face processing. Several groups have reported a robust am ygdala response to a task that requires participants to match faces by emotional e xpression (Hariri et al., 2000;Hariri et al., 2002c;Paulus et al., 2005;Piggot et al., 2004;Wang et al., 2004). Matching facial expressions requires explicit evaluation of em otion, but without the explicit use of verbal labels, which may inhibit the amygdala response by act ivating the ventral prefrontal cortex (Hariri et al., 2000). In order to clarify whether the amygdalar response to matching emotional faces is specifically due to the perception of emotion, we designed a variant of the matching task from Hariri et al. (2000). We added an intermedia te control condition matching neutral faces by

PAGE 71

71 identity to dissect more precisely the emo tional component of the matching task. If the amygdalar response specifically reflects emotion pr ocessing, then it should appear only in the emotion matching condition. Other regions involved in the perception of faces, such as the fusiform face area (Kanwisher et al., 1997) and the superior tem poral sulcus (Chao et al., 1999), should respond to both emotion matching and identity matching. Methods Subjects Twelve healthy participants (six female), aged 18 53 (mean 29) years, were recruited as approved by the University of Floridas Institut ional Review Board. All participants were righthanded and had normal or correct ed-to-normal vision. None re ported any neurological or psychiatric history, nor use of psychoactive medications for the previous six months. Face Matching Task The matching task consisted of three conditi ons: emotion, identity, and control. In each condition, participants were shown a target face above two probe faces, and then had to choose which probe matched the target (Figure 4-1 ). In the emotion condition, pa rticipants were asked to match the faces by their expressed emotion (happine ss, fear, or anger). In the identity condition, participants were asked to matc h neutral faces by identity. In the control condition, participants were asked to match the pixilated patterns deri ved from neutral face pictures; thus all three conditions presented objects with the same dimensi ons and shades of gray. The task was ordered in blocks of six 3-second trials of the same condition, preceded by a 3-second instruction screen. The block condition was varied in a fixed se quence that repeated four times and was counterbalanced across participan ts (emotion > identity > cont rol or control > identity > emotion). The entire run consisted of twelve 21-s econd tasks blocks interspe rsed with thirteen 9second rest blocks and lasted 3 mi n 9 sec. During rest, a fixation cross was displayed. A total of

PAGE 72

72 48 grayscale face portraits were presented from th e series Pictures of Facial Affect (Ekman, and Friesen, 1976), with six actors of each ge nder posing happy, fearful, angry, and neutral expressions. Twelve control patter ns were created by shrinking ne utral face pictures to 8 x 12 resolution, randomizing the pixels, a nd enlarging to original size. W ithin trials, probe and target faces were the same gender, and an equal number of trials of each gender were presented in each block. Each actors face appeared an equal num ber of times during the experiment. In the emotion condition, one actor was selected for the pr obe face, and a second actor for both of the target faces. The pictures subtended approximate ly 3.6 x 5.4 (target) and 2.9 x 4.3 (probes) of the visual field (the target was larger to help distinguish it from the probes). Participants selected the left or right target by pressing a button under their index or middle finger respectively, causing the selected ta rget to be outlined in yellow. The participants practiced each condition inside the scanner before the experimental run until th ey felt confident performing the task. The stimuli were presented using an Integr ated Functional Imaging System (IFIS, MRI Devices, Inc., Waukesha, WI) w ith a 7 LCD screen at 640 X 480 pixel resolution, mounted over the subjects head and viewed using a fixed mirror. The screen subtended approximately 14 x 11 of the visual field. A PC running E-Prim e (Psychology Software Tools, Pittsburgh, PA) began presenting each task in synchronization wi th the first RF pulse of each scan. Responses were collected with a MRI-compatible button gl ove attached to the pa rticipants right hand. Functional Imaging Data Acquisition Brain images were acquired using a Siemens Allegra 3 Tesla scanner (Siemens, Munich, Germany) with a standard head coil. Anatom ic images were acquired using an MPRAGE sequence in the sagittal plane at 1.0 mm3 resolution, TR = 1780ms, TE = 4.38ms, flip angle = 8. Functional images were acquired using a gradient echo planar imaging (EPI) sequence sensitive

PAGE 73

73 to blood oxygen level-dependant (BOLD) contrast in the axial orientation (parallel to the AC-PC line), covering the whole brain with 36 slices, 3.8mm thick (0mm gap) with a 240mm field of view and a matrix size of 64 x 64 voxels (i n-plane resolution = 3.75mm), TR = 3000ms, TE = 30ms, flip angle = 90. A total of 125 brain vo lumes were acquired (3min 15sec scan time) and the first two volumes were discarded before analysis to allow for T1 equilibration. Functional Imaging Data Analysis MR data were analyzed using BrainVoyage r 2000 (v. 4.9.6, Brain Innovations, Maastricht, Holland). The functional images were coregister ed with anatomic images, and normalized to Talairach space for each participant. Functiona l data underwent 3D mo tion correction, linear trend removal and slice scan time correction (the slice data in each volume were time-shifted to the start of the TR by interpolation). Higha nd lowfrequency noise was removed using lowand highpass filters with cut-off frequencies of 10/123 Hz and 1/123 Hz respectively. Spatial smoothing was applied using a Gaussian fi lter of 5.7mm full-width half maximum. Regions of task-related brai n activity were estimated us ing general linear modeling. A reference function, or predictor, was crea ted for each condition by convolving the block presentation time course with an estimated hemodynamic response function (Boynton et al., 1996). The signal at each voxel was modeled with a we ighted combination of the three predictors using least squares fitting. Sta tistical maps were created usin g random effects analysis. This conservative approach looks for consistent di fferences between predictors' weighting across participants, preventing data from one or two participants from dominating the analysis. Clusters of voxels with significant differences between pred ictors were selected by setting a statistical threshold of t(11) > 4.0 (p < 0.002 uncorrect ed) and a minimum cluster size of 100 mm3 (except at the amygdala our a priori region of interest where sm aller clusters were allowed).

PAGE 74

74 For each cluster of significant voxels, we validated the GLM results by plotting the mean BOLD response to each condition, as previously described (Wright et al., 2004). The percentage signal change was calculated for each block relati ve to the preceding resting signal, and then averaged across blocks and participants for each condition. In order to investigate habituation, the BOLD response to each condition was calculate d separately for each of the four repetitions of that block, averaged across participants. Th e mean peak BOLD response was calculated from the period of peak activity at 9 18 seconds afte r block onset. The magnitude and significance of any modulation of mean amplitude over time were calculated using linear regression. To confirm that the MRI parameters used were able to detect signal at the amygdala, we visually inspected the functional images usi ng an outline of the amygdala drawn from the average anatomic image according to the guidelines of Brierley et al. (Brierley et al., 2002). In two out of the twelve participan ts, susceptibility artifact from the nasal sinuses obscured the amygdala. Because amygdala activation could be de tected with or without these participants, they were included in the analysis to maximize statistical power in th e rest of the brain. Results Behavioral Data Participants' responses were significantly faster and more accur ate in the identity condition, compared with emotion and control (Table 4-1 ). On debriefing most participants reported that the identity conditi on was the easiest of the three. fMRI Data Significant differences in modeled signal amplitude (activations) are summarized in Tables 4-2 & 4-3 Emotion-specific activations found us ing the conjunction of the contrasts (emotion identity) and (emotion control) o ccurred at the left infe rior frontal sulcus (Figure 4-2 ) and right precentral gyrus (not shown). The BOLD response at the inferior frontal

PAGE 75

75 sulcus appeared to be specif ic to the emotion condition (Figure 4-2A ), while at the precentral gyrus, the magnitude of the response to identity a ppeared to be intermed iate between that of control and emotion. Emotion-specif ic deactivations were found at the left transverse temporal sulcus and the pregenual an terior cingulate gyrus (Figure 4-4B ). Specifically, the BOLD response at the temporal region was negative during emotion matching but remained near baseline levels during the identity and c ontrol conditions. The pr egenual cingulate BOLD response decreased from baselin e during all three conditions, w ith the largest decrease during emotion matching (Figure 4-4A ). The emotion condition did not selectively activate the amygdala. Activations at the left and right amygdalae were found when contrasting either face matching condition (emotion or identity) with th e control condition. The activation was more statistically significant at the ri ght amygdala than the left (pea k t(11) = 7.25 vs. 5.23 for emotion, 8.63 vs. 5.11 for identity), but the mean peak BO LD response was larger at the left amygdala than the right (0.5% vs. 0.2% for emotion, 0.7% vs. 0.3% for identity). The BOLD response appeared larger in magnitude and duration to id entity than to emotion in both hemispheres. Figure 4-3 and Table 4-3 describe amygdala activation obta ined using the conjunction of the contrasts (emotion control) and (identity co ntrol). The amygdalar clus ters for the individual contrasts (emotion control) and (identity co ntrol) were overlappi ng, and are therefore not depicted separately. Activation for both face matching conditions was also seen in the right fusiform gyrus, occipitotemporal cortex, and anterior and posterior cingulate cortices (Figure 4-4B shows cingulate activations). While the fusiform and o ccipitotemporal activations reflected positive BOLD responses that were larger during face matching than during the control condition, the

PAGE 76

76 anterior and posterior cingulat e BOLD responses were negative during the control condition and remained near baseline levels during the emotion and identity conditions. The control condition activated ex tensive, bilateral regions of the occipital and parietal cortex and smaller clusters within the right anteri or insula, left collateral sulcus, left fusiform gyrus, and left superior frontal sulcus. These ac tivations reflected positive BOLD responses that were larger during pattern matching than during identity and emotion matching. Habituation was also investigated by looking for changes over time in the behavioral and BOLD responses. Neither accuracy nor response time showed significant changes over time, indicating that fatigue did not occur. The response to emotion at the left amygdala decreased markedly over repeated blocks and showed a significant, negative correlation with time (Table 4-4 Figure 4-5 ). Other regions showing significant m odulation of peak amplitude of the BOLD response with time are described in Table 4-4 with the slope of the linear regression representing BOLD modulation. Note that in some regions the re sponse increased over time, and that compared with the change in response at the amygdala, the next largest modulation was only about half the size. Discussion Contrary to the "emotion processor" hypothesis, this study found equal amygdala activation for emotional and neutral face matchi ng. Hariri et al. (2000) proposed that the amygdala encoded emotion at an associative le vel during matching of em otional faces, but the current findings require further explanation. Sander et al. (200 3) describe the amygdala as a system for relevance detection stating, An event is relevant for an organism if it can significantly influence (positively or negatively) the attainment of his or her goals. This definition may be a more useful starting poi nt for interpretation of the present study.

PAGE 77

77 Relevance Detection Activates the Amygdala While Hariri and colleagues acknowledge that the blocked design of their task limits its ability to investigate proce ssing of specific emotions (Har iri et al., 2002c), they have demonstrated that matching facial expression s robustly activates th e amygdala, and have successfully employed the task to probe the eff ects of genetics (Hariri et al., 2002b;Pezawas et al., 2005), drugs (Hariri et al., 2002 a;Tessitore et al., 2002), and agi ng (Tessitore et al., 2005) on the amygdala and an associated affective network. Previous studies have shown amygdala activa tion to neutral faces based on gaze direction, novelty, and race (Sander et al., 2003). To our knowledge the only face matching study with neutral faces investigated the e ffect of matching neutral faces by race (Lieberman et al., 2005). In the current study, however, matchi ng neutral faces activated the amygdalae as much as matching emotional faces without the addition of relevance fr om gaze, race, etc. It seems that the matching task itself adds relevance to neutral face stimuli. Viewed alone, neutral faces would be expected to have less inherent relevance than emotiona l faces, but during a matching task, they must acquire task-related, goal-oriented relevance. That is, one face becomes the right face, while the other face becomes the wrong face. If we accept that the matching task itself evokes amygdalar relevance detection, we must still acc ount for the absence of additional activation of the amygdala by emotional content, and for the a pparent lack of releva nce detection during the control condition. Emotion Processing at the Amygdala Habituates At the left amygdala, the mean amplitude of the BOLD response decreased over time for emotion but not identity matching. This suggest s that while both face matching conditions activate the amygdala, there is still a distinct pattern of emotion pr ocessing at the left amygdala. Figure 4-5 shows that the initia l BOLD response at the left amygdala is larger to emotion than to

PAGE 78

78 identity, but then rapidly diminishes. Although th e emotion condition initially evokes greater left amygdalar activation than the identity condition, ha bituation apparently prevented this difference from being detected with the statis tical approach used in this study. A previous study found greater hab ituation to repeated, passivel y viewed faces in the right amygdala compared with the left, and that activa tion in the left amygdala distinguished fearful and happy faces (Wright et al., 2001). The author s described the right amygdala as a rapid but general relevance detector, and the left amygdala as slower but capable of distinguishing emotional valence. The current findings fit this mode l but do not test it e xplicitly. It is possible that habituation of the right am ygdalar response in the current study occurs rapidly within the first block, making it difficult to detect. An event -related study of habitu ation during explicit and incidental emotion processing may shed more light on the lateralization of the speed and specificity of emotion processing at the amygdala. An earlier study divided the left amygdalar resp onse to unconsciously perceived emotional faces into a ventral valence pr ocessing domain and a dorsal salience processing domain (Whalen et al., 1998b). Since the face matching task does not investigate differences in valence, this may explain the dorsal location of the amygdalar activation in the current study. Spatial Processing Bypasses the Amygdala During the Control Condition Although the control condition involved patter n matching it elicited no response from the amygdala compared with rest. While face matchi ng activated the ventral visual pathway (in particular the right fusiform gyrus), pattern ma tching activated mostly the dorsal visual pathway (the interior parietal lobule and intraparietal sulcus bilaterally, see Table 4-3 ). Several participants described a pattern-matching strategy of aligning the white sq uares in the probe and target patterns. It is possible th at this spatial alignment operati on may be performed by the dorsal visual pathway, avoiding the need for relevance detection by the amygdala.

PAGE 79

79 However, it is not clear whethe r the amygdala detects relevance in visual information using the categorical, ventral pathway exclusively. A dolphs and colleagues reported initial evidence for emotion processing via the dorsal visual path way in one patient with extensive encephalitis lesions of the ventral surface. This patient wa s impaired at recognizing emotions in facial expressions, but could recognize em otions when they were acted out (Adolphs et al., 2003). We are unaware of any functional imaging study show ing amygdala activation vi a the dorsal visual pathway. Cognitive Processing During Emotion Matching The emotion condition selectivel y activated a region of the le ft inferior frontal sulcus (Figure 4-2 ). A previous study found activation of the corresponding region in the right hemisphere to explicit, but not incidental, evalua tion of facial expressions (Gorno-Tempini et al., 2001). Activity in this region of the left hemisphere has been associated with word reading (Matsuo et al., 2003) and with act ion recognition (Hamzei et al., 2003). Since several subjects in the current study reported mentally naming the targ et emotion, the observed left-sided activation may reflect covertly naming or "reading" facial expressions, especially since unfamiliar neutral faces and pixilated patterns are not easily na med. While hemispheric differences in amygdala activity have been hypothesized to result from is psilateral prefrontal efferents (Irwin et al., 2004), and Hariri and colleagues demonstrated inhibition of th e amygdala by the prefrontal cortex during emotion labeling (Hariri et al ., 2000), it seems unlikely that the prefrontal activation to emotion matching in the current study is responsible for the habituation of the left amygdala. The active region in the cu rrent study is in the left dorso lateral prefrontal cortex, and corresponds neither with the ventro medial prefrontal region associ ated with fear extinction in rats (Quirk, and Gehlert, 2003), nor with the right vent rolateral region described by Hariri et al. (2000). The correlation between the signal time cour se at the left amygdala and left inferior

PAGE 80

80 frontal sulcus was small and not significant (r = 0.16), implying that these regions are not functionally connected during this task. While identity matching requires a simple, pe rceptual match, emotion matching requires additional categorical processing (reflected in increased reaction time and decreased accuracy [Table 4-1 ]). This cognitive component may indirec tly link prefrontal ac tivation and amygdalar habituation during the emotion condition. The left amygdalas BO LD response to the emotion condition decreases to below the level of its response to the identity condition (Figure 4-5 A+B ). Thus increased cognitive processing during th e emotion condition may inhibit the amygdala, resulting in lower baseline (tas k-related) activation once the eff ect of emotion has habituated. We investigated the effect of task difficu lty with a multivariate ANOVA using task condition (emotion, identity, or control) and block order (1st, 2nd, 3rd, or 4th exposure to block) to predict mean peak BOLD response in the prefrontal co rtex and amygdala. Introducing reaction time as a covariate measure of task difficulty did not modu late the effect of task condition on mean peak BOLD response. We conclude that the observed brain activity was not qua ntitatively linked with task difficulty in the present study. Negative BOLD Responses Of additional interest are the negative BOLD responses found at the an terior and posterior cingulate cortices (Figure 4-4 ). These regions have been describe d as part of a network that is active during rest and deactivated by a variety of tasks (McKiernan et al., 2003). Because this network deactivated to both visual and auditory tasks (proportionate to difficulty in the latter case), McKiernan et al. (2003) sp eculated that it mediates atten tion-dependant processing during the conscious resting state, in cluding monitoring emotional stat e. The posterior cingulate and subgenual anterior cingulate deactivated duri ng the control condition alone, whereas the pregenual anterior cingulate deactivated during a ll three conditions, with the greatest decrease

PAGE 81

81 for emotion. It is possible, ther efore, that the posterior cingula te and subgenual anterior cingulate are involved in similar functions at rest and during the emotion and identity conditions. Conversely, matching emotions decreases activity in the pregenual cingulate cortex. A recent meta-analysis found that the peri-genual anterior cingulate cortex was ac tivated in emotional studies and deactivated in cogni tive studies (Bush et al., 2000). Furthermore, the emotional cingulate interacts strongly w ith the amygdala (Pezawas et al., 2005). Because the emotion condition involves a cognitive operation on emotional stimuli, the responses we observed at the pregenual cingulate cortex have several possibl e interpretations. The la rger deactivation may reflect a coincidental need for increased atte ntional resources elsewh ere or alternatively, deactivation of the pregenual cingulate may be necessary for emotional processing in the matching task. Complex Contributions to Amygdala Activation Activation of the amygdala by emotional face ma tching appears to reflect a combination of processes. Both the identity a nd the emotion task involve rele vance detection at the amygdala simply because a choice between two faces must be made, while the control condition appears to utilize a separate spatial processing network. It appears that additional left amygdalar activation due to emotional content was not detected because of rapid habituation. Selective activation of the left inferior prefrontal sulcus and habituation of the left amygdala during the emotion condition may be indirectly li nked via the common influence of increased top-down processing during emotion matching. We conclude that activ ation of the amygdala to the emotional facematching task cannot be interpreted as bottom-up emotion processing alone, but likely involves more general relevance detection involved in perceptual matching.

PAGE 82

82 Table 4-1 Behavioral data Task condition Response time (ms) Accuracy (%) Control 1625 86.1 Identity 1155 99.7 Emotion 1704 90.6 Significantly different to both control a nd emotion, p < 0.001, unpaired Students t-test.

PAGE 83

83 Table 4-2 Clusters of activation for [(Emotion Identity) (Identity Control)] Region Side BA x y z Size t(11) Inferior frontal sulcus L 9, 44-451530183 4.8 Precentral gyrus R 449-452144 6.0 Pregenual cingulate cortex L 24, 32-43710169 -5.2 Transverse temporal gyrus L 41-52-2915210 -6.2 Only clusters >100 voxels shown except for a priori region (amygdala). L: left, R: right. BA: Brodmanns Area. X, Y and Z refer to Tala irach co-ordinates. Size: number of 1mm3 voxels. t: random effects statistical score (degrees of freedom). Negative t values indi cate deactivation. denotes conjunction of two contrasts.

PAGE 84

84 Table 4-3 Clusters of activation for [(Emotion Control) (Identity Control)] Region Side BA x y z Size t(11) Amygdala R 22-7-10487 6.0 Amygdala L -9-3-1035 6.0 Subgenual cingulate cortex R 32433-10103 6.6 Fusiform gyrus R 3739-42-19198 5.1 Inferior temporal sulcus R 3749-70-3200 7.6 Middle temporal gyrus R 3953-599883 8.6 Posterior cingulate gyrus R 231-55231491 6.0 Insula R 32203122 -6.5 Collateral sulcus L 35, 36-26-40-9114 -6.8 Fusiform gyrus L 19, 37-24-58-12126 -6.7 Inferior parietal lobule R 4040-3736350 -6.6 Inferior parietal lobule L 40-33-4640404 -6.2 Intraparietal sulcus R 7, 1922-68423344 -5.9 Intraparietal sulcus L 7, 19-17-68421081 -6.6 Middle occipital gyrus R 1934-7913870 -5.3 Middle occipital gyrus L 19-25-74221015 -5.6 Superior frontal sulcus L 6-19-355176 -6.1 Only clusters >100 voxels shown except for a priori region (amygdala). L: left, R: right. BA: Brodmanns Area. X, Y and Z refer to Tala irach co-ordinates. Size: number of 1mm3 voxels. t: random effects statistical score (degrees of freedom). Negative t values indi cate deactivation. denotes conjunction of two contrasts.

PAGE 85

85 Table 4-4 Regions showing signifi cant modulation of BOLD response Region BOLD modulation (% / run) Emotion Identity Control L amygdala -1.39-0.14-0.53 L transverse temporal gyrus -0.520.33-0.58 R amygdala 0.14* 0.40-0.22 R middle temporal gyrus 0.02* 0.31-0.12 B posterior cingulate co rtex -0.32* 0.76-0.45 L pregenual cingulate cortex -0.430.31* -0.50 Values represent difference in peak BOLD ac tivation (% signal change ) over one run (four blocks of each condition). significant correlation, p < 0.05.

PAGE 86

86 Figure 4-1 Matching task paradigm. Participants had to select which of the lower two probe images matched the upper target image. Th e selected probe was outlined in yellow. Faces were matched by emotion (A) or iden tity (B) and in the control condition (C), participants matched pixilated patterns.

PAGE 87

87 Figure 4-2 Selective response to emotion at the left inferior prefront al sulcus. A: BOLD response; vertical dotted lines indicate beginning and end of block; error bars denote standard error of mean. B: Cluster of activation for [(Emotion Identity) (Identity Control)] with a threshold of t(11) > 4.0; slice locati on given in Talairach coordinates; slice in radiologi cal convention (Table 4-2). C: Left inferior prefrontal sulcus activation for the group illustrated on a rendered 3D brain from a single participant.

PAGE 88

88 Figure 4-3 Response to face matching at the le ft and right amygdala. A+C: BOLD responses; vertical dotted lines indicate beginning and e nd of block; error bars denote standard error of mean. B: Clusters of activation for [(Emotion Control) (Identity Control)] with a threshold of t(11) > 4.0; slice locati on given in Talairach coordinates; slice in radiologi cal convention (Table 4-3).

PAGE 89

89 Figure 4-4 Regions of deactivation. B: "Glass br ain" showing clusters of activation with a threshold of t(11) > 4.0 in the cingulate cortex (Tables 4-2 & 4-3). Gray borders denote the anterior and posterior commisure s and the borders of the cerebrum. A, C+D: BOLD responses; vertical dotted line s indicate beginning and end of block; error bars denote standard error of mea n. A: Pregenual cingulat e cortex (-4, 37, 10). The BOLD response is negative for all thr ee conditions, with a significantly greater decrease for emotion. C+D: Subgenual cingula te cortex (4, 33, -10) and posterior cingulate cortex (1, 23). The BOLD response is negative in the control condition, but remains at baseline for emotion and identity.

PAGE 90

90 Figure 4-5 Habituation. A+B: The BOLD response at the amygdala decreases over time in the emotion condition but not in the identity condition. BOLD responses are derived from left amygdala activation clusters for the cont rasts (emotion control) (A) and (identity control) (B). C+D: The peak BOLD respons e habituates only at the left amygdala in the emotion condition. Peak response = mean % signal change values for 9 sec 18 sec time points. Responses derived from le ft and right amygdala clusters for the contrast [(Emotion Control) (Identity Control)] (Table 4-3).

PAGE 91

91 CHAPTER 5 DISSOCIATING EVENT-RELATED RESPON SES TO TOP-DOWN AND BOTTOM-UP EMOTION PROCESSING Conscious rationalization of emotional stimuli may modulate automatic physiological responses to stimulus content. Functional ne uroimaging studies have investigated top-down modulation of emotion using tasks that require participan ts to rate their re sponses to emotional pictures. These studies suggest that the amygdala and insula ma y mediate bottom-up responses to the content of emotional scenes, and that th ese responses may be modulated by top-down processing mediated by the anteri or cingulate cortex (A CC) and medial pref rontal cortex. We attempted to replicate these findings using an optimized pa radigm design. Pleasant and unpleasant pictures were rated using emotional and non-emotiona l rating scales : either the pleasantness of the picture, or the frequency of its appearance on television. Stimulus content and tasks instructions were randomized on a trial-by-trial basis, prev enting expectancy of emotional content, and equalizing the timing of bottom-up and top-down responses. Factorial analysis was used to separate the main effects of bottomup and top-down processing and their interaction. The amygdala responded to unpleasant pictures du ring both tasks. The orb itofrontal cortex and amygdala responded during emotional rating of bot h pleasant and unpleasant pictures. The ACC responded to both stimulus content and task dema nds. No significant interaction effects were found. These results show that an event-relate d picture rating task may dissociate bottom-up processing in the amygdala from top-down proce ssing in the orbitofrontal cortex and insula. Introduction Psychological and neuroanatomic evidence sugg ests that responses to emotional stimuli may be mediated both by automatic, bottom-up pr ocesses and by intentiona l, top-down processes (Eysenck, and Keane, 2000). Zajonc proposed the affective primacy hypothesis, which states the appraisal of emotional stimuli is rapid a nd automatic (Zajonce, 1980, quoted in Eysenck &

PAGE 92

92 Keane, 2000). He supported this hypothesis by showing that sub liminally presented emotional faces may influence pleasantness ratings of a subsequently presented Chinese pictogram. Lazarus, on the other hand, argued that emoti onal responses are influenced by conscious, contextual appraisal (Lazar us, 1982, quoted in Eysenck & Ke ane, 2000). He supported this hypothesis by demonstrating that physiological responses to a di sturbing film varied depending on the narrative that accompanied the film. Alt hough these hypotheses appear to disagree, later theorists proposed that emotional experience may result from the appraisal of a stimulus by both bottom-up and top-down processes (Ellsworth, 1 994). Anatomical evidence supports multiple, interacting pathways for emotional appraisal. The idea of two parallel emotion systems is supported by anatomical evidence. In rats, cond itioned fear responses are dependent upon the amygdala. LeDoux showed that fear-inducing vi sual signals may reach the amygdala by two pathways: a fast subcortical pathway, and a sl ow cortical pathway (LeDoux, 2000). Subsequent functional neuroimaging studies in humans have sought to describe the roles of the amygdala and associated cortical regions in emotional appraisal. Several investigators have investigated botto m-up and top-down components of emotional appraisal by employing emotion rating tasks. Bottom-up processes may be mediated by brain regions that respond to differences in the emoti onal content of stimuli. Top-down processes may be mediated by brain regions that respond only duri ng explicit rating of emotion. By altering task demands, these studies investigated the m odulation of bottom-up responses by top-down appraisal. In the earliest rating study, participants view ed blocks of mixed unpleasant and neutral pictures during positron emission tomography (PET) and rated either whether each picture was pleasant or unpleasant or whet her it was indoors or outdoors (L ane et al., 1997a). Comparing emotion rating with location rating, activation wa s seen in a cluster spanning the anterior

PAGE 93

93 cingulate cortex (ACC) and medial prefrontal cortex (PFC) (bot h Brodmann Area 32) and in the insula. The authors suggested that the ACC mediated internal attention specifically associated with emotion processing, and that increased activ ity at the insula represented amplification of interoceptive processing. This study could not investigate bottom-up responses because unpleasant and neutral pictures were intermingled. In two s ubsequent studies, pleasant and neutral pictures were presente d separately to identify region s involved in bottom-up responses (Liberzon et al., 2000;Taylor et al., 2003). Liberz on et al. (2000) compared emotion rating with picture recognition, a cognitive task intended to draw attention away from emotion. The right amygdala was activated by unpleasant compared with neutral pictures, and activation was greater during emotion rating than duri ng picture recognition. Taylor et al. (2003) compared emotion rating with passive viewing, in order to test whether top-down processing diminished bottom-up responses. The right amygdala a nd insula responded to unpleasant pictures, but the response was smaller during emotion rating than during passiv e viewing. The opposite effect was seen in the ACC and medial prefrontal co rtex: activation by unpleasant pi ctures was increased during emotion rating compared with passive viewing. A later study compared emotion rating with ratings of personal relevance (Phan et al., 2004). The left amygdala responded more strongly to emotional pictures during emotion rating; howev er the ACC and medial PFC were deactivated during emotion rating. This finding suggests a specifi c role for these region s in the appraisal of self-relevance, which Zajonc and Smith (1993) de fined as a component of emotional appraisal (quoted in Eysenck and Keane, 2000). This vi ew is supported by the recent finding that the medial PFC responds more strongly when rating ones own emotions than when rating others emotions (Ochsner et al., 2004). In summary, thes e studies suggest that the amygdala and insula

PAGE 94

94 may mediate bottom-up responses to emotional stim ulus content, and that these responses may be modulated by top-down pro cessing mediated by the ACC. The studies above employ several task designs. In all but one study, emotional pictures are presented in blocks, so neur al responses to emotional c ontent may be confounded with expectancy of emotion, a top-down effect. Similarly, most studies te sted for interactions between bottom-up and top-down factors using single statis tical contrasts, which do not account for the main effects of each factor (Friston et al., 1996). Phan et al. (2004) presented emotional pictures in a random, event-related design and tested inte ractions using a factor ial analysis. However, their rating tasks were presented in blocks. B ecause the bottom-up and top-down factors varied at different time intervals, the neural responses to each factor may have unequal power. Finally, these studies used a variety of control tasks. Pa ssive viewing fails to control for the attentional and motor components of emotion rating. Recogniti on and self-relevance rating include internal appraisal and thus may be confounded with emo tion rating. The indoor / outdoor judgment task avoids these problems, but elicits only two cat egorical responses, while emotion rating tasks typically elicit responses on a continuous scale. In the current study, we investigate bottom-up and top-down emotional appraisal using an optimized emotion rating paradigm. We employ an event-related picture rating task in which both emotional content and task instructions are randomized. By randomizing both bottom-up and top-down factors at the same temporal fr equency we avoided contaminating our findings with differential signal-to-noi se characteristics of eventand block-level hemodynamic responses. As a control task, participants were asked how frequently images similar to the one displayed are shown on television. We chose freq uency rating as a control condition because it controls for attentional and motor components of the rating task, requires attention to external

PAGE 95

95 stimulus features, and elicits a response on a co ntinuous scale. We used factorial analysis to identify the neural responses to the task. Regions showing a main ef fect of stimulus content were defined as mediated bottom-up responses. Region s showing a main effect of rating task were defined as mediating top-down appraisal. Regions showing an interaction effect were defined as mediating bottom-up responses that were modul ated by top-down appraisal. We tested the hypotheses that the amygdala mediates a botto m-up response that is modulated by top-down appraisal, and that the ACC mediates top-down appraisal. Methods Subjects Sixteen healthy male participants gave inform ed consent as approved by the University of Floridas Institutional Review Board. The part icipants had no history of psychiatric or neurological illness, and were taking no psychot ropic medication at the time of the study. One participant was excluded due to discrete head movements gr eater than 1mm during scanning. Picture Rating Task Paradigm Participants were presented w ith pictures from the Internat ional Affective Picture System (IAPS) (Center for the Study of Emotion and Attention [CSEA-NIMH], 2001). Below each picture, a cue instructed particip ants to make either an emotion or frequency rating. The emotion rating cue read, How pleasant do you find the cont ent of this image? The frequency rating cue read, How frequent do images with similar cont ent appear on television? The participants selected one of four responses: for emotion ra tings, very unpleasant, moderately unpleasant, moderately pleasant, or very pleasant, and for frequency ratin gs, weekly, daily, hourly, or continuously. We selected emotionally arousing IAPS pictures and assigned them, based on the valence ratings, to two groups: plea sant or unpleasant (Lang et al ., 2001). The mean ratings were, for the pleasant set, pleasure = 6.7 +/0.9, ar ousal = 4.7 +/1.0, and for the unpleasant set,

PAGE 96

96 pleasure = 3.7 +/1.1, arousal = 4.8 +/1.3 (mean +/standard deviation). IAPS picture codes are listed in appendix A. The trials were categori zed by rating task and valence, giving four trial types: emotion rating pleasant (EP), emotion rating unpleasant (EU), frequency rating pleasant (FP), and frequency rating unpleasant (FU). Prior to performing the task, each participant was familiarized with the task and the scanner environment by completing a training run consisting only of emot ion ratings. Different sets of pictures were used for the training a nd task runs. The results of the training run are reported elsewhere, in a study of the effect s of training upon emotional and non-emotional ratings (Li et al., 2006). During the task run, all four trial types were presented in a random order using an eventrelated design. Fifteen trials of each type were presented for 3 sec each, along with 30 null trials, during which a fixation cross was displayed for 3 sec. The test run lasted 4 min 30 sec. Null trials were included in the random sequence in orde r to jitter the stimulus onset asynchrony (SOA) between trials. This increases the variance in the resulting fM RI response, making the response to rapid stimuli (SOA < 15 sec) detectable (B urock et al., 1998). Jittering the SOA with randomly interspersed null trials creates a ge ometric distribution of SOAs, which is more efficient than uniform randomization (Serence s, 2004). The resulting mean SOA was 4.5 sec, with a minimum of 3 sec. This timing was chosen to minimize response attenuation when repeating emotional stimuli, but to maximize the number of trials in th e run (Soon et al., 2003). The stimuli were presented using an Integr ated Functional Imaging System (IFIS, MRI Devices, Inc., Waukesha, WI) w ith a 7 LCD screen at 640 X 480 pixel resolution, mounted over the subjects head and viewed using a fixed mirror. The screen subtended approximately 14 x 11 of the visual field. A PC running E-Prim e (Psychology Software Tools, Pittsburgh, PA)

PAGE 97

97 began presenting each task in synchronization wi th the first RF pulse of each scan. Responses were collected with a MRI-compatible button gl ove attached to the pa rticipants right hand. Functional Imaging Data Acquisition Participants were scanned usi ng a 3 Tesla Siemens Allegra sc anner with a standard head coil (Siemens, Munich, Germany). Anatomic imag es were acquired using an MPRAGE sequence with TR = 1500 ms, TE = 4.38 ms, and flip angle = 8. In the axial plane, 160 slices were acquired (thickness 1.0 1.2 mm, according to the he ight of the brain) with in-plane field of view 240 mm X 180 mm and matrix size 256 X 192. F unctional images covering the whole brain were acquired using echo-planar imaging se nsitive to blood-oxygena tion level dependent (BOLD) effects, with TR = 3000 ms, TE = 30 ms, f lip angle = 90. In the axial plane, 38 slices with a thickness of 3.8 mm were al igned with the plane of the in tercommissural line and had an in-plane field of view 240 X 240 mm and matrix size 64 X 64. The first two volumes of each functional run were discarded to allow for T1 eq uilibration. These settings have previously been shown to provide reasonable c overage of the amygdala while allowing coverage of the whole brain, and without sacrificing BOLD sensitivity (Wright, and Liu, 2005). Visual inspection of functional images showed coverage in the am ygdala was adequate in ten out of fifteen participants. Because our a priori hypotheses predicted responses in the ACC, coverage of this region was inspected. Signal in the subgenual A CC was lost to susceptibility artifact, and responses within the affect ed region were discarded. Functional Imaging Data Analysis Data were analyzed using BrainVoyager QX version 1.7.6 (Brain Innovations, Maastricht, Holland). The functional images were coregister ed with anatomic images, and normalized to Talairach space for each participant. Functiona l data underwent 3D mo tion correction, linear trend removal and slice scan time correction. The test runs underwent Gaussian spatial

PAGE 98

98 smoothing using a kernel of 5.7 mm (1.5 voxels ) full-width half-maximum (FWHM). The training run underwent spatial smoothing as abov e and temporal Gaussian smoothing using a kernel of 4 data points (12 sec) FWHM. Task-related activity was mapped using a voxelwise general linear modeling analysis. For both event-related and block analyses, the BOLD response to each task condition was estimated using a standard hemodynamic model (Friston et al., 1998). The estimated responses were fit to the MR signal for each individua l to generate a beta weight, reflecting the magnitude of the contribution of each task type to the overall model. Using the conservative random-effects approach, statistical maps were generated by appl ying second-order statisti cs to the groups beta weights at each voxel. For the test runs, a tw o-way ANOVA was used to estimate separately the main effects of judgment type and emotional vale nce, and their interactio n. This approach avoids the assumption of pure insertion, allowing the locali zation of neural responses that correlate with cognitive task components in a nonlinear fashion (Friston et al., 1996). In th e training run a t-test was used to compare early vs. late training response s (the first two blocks of ten vs. the last two). For whole-brain analysis, thre sholds were set to exclude clusters smaller than 100 mm3 (after functional data were resampled to 1 mm resoluti on), and statistical scores below F (1,14) = 12, p < 0.005. For regions for which we had an a priori hypothesis (amygdala, OFC, and ACC) the statistical threshold was lowered to F(1,14) = 5, p < 0.05. At each region, t-score were calculated post-hoc using single statistical contrasts to indicate the dire ction of the main effect or interaction. Mean BOLD responses were plotted usi ng BrainVoyagers event-related averaging function. In the test runs, BOLD re sponses were calculated for each of the four trial types. In the training run, BOLD responses were plotted for early, middle, and late blocks. For each test run

PAGE 99

99 event, percentage signal change was calculated relative to the si gnal during the two seconds prior to stimulus onset. These values were averaged by event type in a time window from to 13 seconds relative to stimulus onset Contamination from subsequent stimuli occurr ing within the 13-second window was eliminated in the overall av erage due to the jittered SOA (Dale, and Buckner, 1997). Events were not time-locked to the fMRI sampling period (3 seconds) to allow finer sampling of the BOLD response (Serences 2004). Data therefore were resampled to 1 second resolution by interpolation. BOLD responses to the training paradigm were calculated in a similar way, but using the original sampling pe riod. Signal change was cal culated relative to the two periods preceding the start of each block (6 seconds). A time window from to 36 sec was used to cover the BOLD response to the entire block. Results Behavioral Data Participants rated pleasant and unpleasant stim uli appropriately. Pleasure ratings, adjusted to the standard scale used in the IAPS of 1 9 were significantly higher for pleasant pictures compared with unpleasant pictures (6.6 +/0.9 vs. 3.6 +/0.8, p < 0.001). Response times were slower during frequency rating, re gardless of emotion, implying that the control task was more difficult (Table 5-1 ). fMRI Data Responses were observed in the regions for which an a priori hypothesis was proposed. A main effect of task was observed in the left OFC and right insula (Figure 5-1 ). These regions showed positive BOLD responses to emotion rati ng and negative responses to frequency rating. Supporting the statistical main e ffect, the BOLD responses appeared equal for pleasant and unpleasant pictures (Figure 5-1 B & D ). The left amygdala responded selectively to unpleasant pictures (Figure 5-2 ). Activation was detected at reduced threshold (F(1,14) = 5, p < 0.05) due to

PAGE 100

100 the a priori selection of this region. A positive BOLD response was seen for negative pictures during both tasks, supporting the main effect of valence. Examination of the ACC, also at reduced threshold, revealed three activations, two exhibiting a main effect of valence, and one a main effect of task (Figure 5-3 ). These clusters were distinct, but overlapped slightly. The BOLD responses in these regions suggested that in both clusters exhibiting a main effects of valence, the result was driven by positive responses to pleas ant pictures. In the cl uster exhibiting a main effect of task, the result appear ed to be driven by a positiv e response to emotion rating (Figure 53 B-D ). However, the BOLD response curves were not as clearly separated in these regions as they were in the OFC, insula, and amygdala. Regions for which we proposed no a priori hypothesis also showed interaction effects (Table 5-2 ) and main effects of valence (Table 5-3 ) and task (Table 5-4 ). An interaction between task and valence was located between the left in sula and temporal operculum, a region anterior and posterior to the main effect of task at th e right insula. The BOLD responses at this region (not shown) were more variable than at other regions, and did not have a typical curve shape, suggesting they may have arisen from the nearby middle cerebral artery. Several regions showed a main effect of valence, name ly the bilateral posterior cingulat e cortex, postcentral gyrus, and posterior fusiform cortex. A main effect of task was seen in the bilateral parieto-occipital sulcus, showing greater BOLD responses to the frequenc y task, regardless of valence. Several other regions showed greater responses to frequency ratings, predominantly in the left PFC, including premotor cortex and supplementary motor area (SMA). Discussion The aim of the current study was to identify re gions of the brain involved in bottom-up and top-down emotional appraisal. We pursued this aim using an ev ent-related, factorial design to identify regional brain responses that varied with stimulus valence, task instructions, or an

PAGE 101

101 interaction of both factors. Prev ious studies implicated the am ygdala, insula, anterior cingulate cortex (ACC), and orbitofrontal cortex (OFC) in task involving emoti onal ratings. The current study found responses to the emotion rating task at the OFC and insula, an d to stimulus content at the amygdala. Both effects, and an interaction between stimulus content and task instructions, were seen in different su bdivisions of the ACC. Top-down Appraisal in the OFC and Insula Rating emotion activated the left OFC and bila teral insulae. These re gions showed a main effect of task, and as the BOLD responses c onfirm, responded equally to both pleasant and unpleasant stimuli. These findings suggest that both regions are implicated in top-down emotional appraisal, independently of bottom-up responses. Previous studies associated the OFC with em otional rating of pictures, odors and words (Cunningham et al., 2004;Liberzon et al., 2000;Roye t et al., 2003). In th e picture rating study, the OFC showed an increased response to unpleasant relative to pleasant pi ctures during emotion rating, but not during picture rec ognition (Liberzon et al., 2000). Ho wever, direct comparison of the rating and recognition tasks re vealed no interaction effect. Furthermore the region of the OFC that was activated (1, 23, 18) was medial to the cluster found in the current study (-26, 35, -7), therefore a direct comparison of the two responses may not be meaningful. Studies in monkeys have shown that OFC neurons respond to the reward value of abstract objects, even when the objects reward associations are change d (Rolls, 1999). In humans, lesions of the OFC impair social functioning, decision making, and long-term planning (Damasio, and Van Hoesen, 1983). The OFC is commonly activated in reward studi es, in which participants learn to associate abstract stimuli with monetary va lue. It is possible that in the current study, the response in the OFC reflects an intentional, explicit j udgment of stimulus value (O'Doherty, 2004).

PAGE 102

102 The insula receives sensory input from the visc era, and has been associated with awareness of internal sensations and with the emoti on of disgust (Critchle y, 2005;Phillips et al., 1997;Wright et al., 2004). Although the insula appears closely involve d in the visceral component of emotion, and therefore would be a ssociated with bottom-up emotional appraisal, Adolphs (2002) proposed a sequence of regi onal brain responses dur ing facial emotion processing in which the insula responds last, possibly corresponding with conscious awareness of emotion (Adolphs, 2002). A previous study interpre ted activation of the insula during emotion rating as an amplification of inte roceptive cues, but this studys design did not assess the effect of the stimulus content (Lane et al., 1997a). A la ter study showed an increased response at the insula to unpleasant versus neutral pictures ; the response decreased during emotion rating compared with passive viewing (Taylor et al., 2003). In a third study, th e insular response was proportional to emotional intensity1, and was increased during bot h emotion rating and selfrelatedness rating (Phan et al., 2004). In Phan et al. (2004), all insu la responses were anterior to the response reported here, whereas in Taylor et al. (2003), the insula response corresponded with the current study. However, Taylor et al. compared unpleasant and ne utral pictures, whereas the current study compared unpleasant and pleasant pictures. The two picture sets used in the current study thus differed in pleasure ratings, bu t not in arousal ratings. Therefore it is possible that our equal insula responses re flected our two picture sets equa l arousal scores. Thus the main effect of task in the current study may in fact reflect top-down modulation of bottom-up responses that are equal to bot h pleasant and unpleasant pictures. Taylor et al. (2003) found that the insular response was decreased during emoti on rating, whereas in the current study it was increased. This may be explained by the use of different control conditions in the two studies. 1 In Phan et al. (2004), intensity is distinct from arousal and refers to a measure deri ved from the pleasure scale, where high intensity means high or low pleasure, and low intensity means moderate pleasure.

PAGE 103

103 Relative to emotion rating, frequency rating may decrease emotional resp onses, whereas passive viewing may increase them. The above hypotheses could be tested in a future study combining three picture conditions (pleasant, neutral, a nd unpleasant) and three task conditions (emotion rating, passive viewing, and a non-emotional rating task). Bottom-up Processing in the Amygdala The left amygdala responded selectively to unpleasant pictures, regardless of task instructions (Figure 5-2 ). Bottom-up processing in the amygdal a is consistent with previous studies, although these tended to report responses in the right amygdala that were modulated by different rating tasks (Liberzon et al., 2000;Taylor et al., 2003), or constant across tasks (Phan et al., 2004). The difference in laterality between pr evious studies and the current study, and the lack of modulation of the amygdala response, ma y be due to the training period that preceded testing. A previous study of facial expressi on recognition reported that amygdala responses shifted over time from right to left, which the au thors suggested represented a shift in processing style (Gur et al., 2002). Other stud ies supported differences in proces sing styles in the right and left amygdala. Lesions of the right amygdala impaired general, autonomic responses to emotional faces, but lesions of the left amygda la impaired recognition of specific facial expressions (Glascher, and Adolphs, 2003). While the right amygdala responded to subliminallypresented facial expressions of fear, the left amygdala responded to seen faces (Morris et al., 1999). The right amygdala responded preferentially to emotional pictures, whereas the left amygdala responded preferentially to words (Mar kowitsch, 1998;Phelps et al., 2001). The left amygdala response in the current study may ther efore represent comparatively more conscious and specific processing of emotion than the ri ght amygdala response reported in previous emotion rating studies. A behavior al study of the effects of trai ning in emotional rating showed that after training, individuals are more likely to make emotional ratings spontaneously when

PAGE 104

104 presented with an image and allowed to respond freely (Li et al., 2006). The authors suggested that explicit emotion rating may become proce duralized, or implicitly learned. Thus, in the current study, training in emotion rating may ha ve produced an automatic early appraisal of emotional valence regardless of whether an em otion rating was required. The left amygdala response may therefore mediate implicitly learne d emotion rating that is immune to modulation by task instructions. Future studies may specifica lly investigate how the neural correlates of emotion rating change over time, and may also inve stigate the differential sensitivity of the left and right amygdalae to modulati on by different rating tasks. Mixed Responses at the An terior Cingulate Cortex Adjacent regions of the ACC responded to emo tion ratings and to pleasant pictures (Figure 5-3 ). While previous studies associated this region with top-down processing, the current findings suggest that this region mediates multiple levels of emotion processing. Both types of responses reported here have been reported in previous studies. The ACC responded during emotion ratings (compared with a control task) in seve ral previous studies, although the location of the responses in these studies was more dorsa l than those in the cu rrent study (Cunningham et al., 2004;Lane et al., 1997a;Ochsner et al., 2004). Several other studies have reported responses to happy mood induction in the pregenual ACC, the same region that in the current study responded to pleasant pictures (s ee Vogt, 2005). Researchers are cu rrently investigating the roles of different subdivisions of the ACC in emo tion processing (Bush et al., 2000;Vogt, 2005). The current findings implicate different subdivision s in bottom-up and top-down emotion processing. However, because the BOLD responses in the ACC were less clearly separated than the responses in the OFC, insula, and amygdala (Figure 5-3 B-D ), the statistical findings are supported with less confidence in this region. Previous studies reported a strong top-down response in the ACC (Cunningham et al., 2004;Lane et al., 1997a;Ochsner et al., 2004). The

PAGE 105

105 apparently weak top-down res ponse in the current study may be due to the event-related design. Previous studies varied rating ta sks in a block-related fashion, and it is possible that clearly differentiated responses in the ACC to top-down emotion processing requir e the establishment of a cognitive set over several stimuli, as is though t to occur during block presentation. The current study varied both emotional content and task instruc tions at the event level so that both factors would elicit responses of the same temporal magn itude, and therefore have comparable statistical power. The stimulus onset asynchrony (mean 4.5 seconds) was previously shown to maximize the efficiency of detecting responses to emo tional faces, a bottom-up effect (Serences, 2004). Future studies may investigate most efficient timing of stimuli for th e detection of top-down effects. Response to Frequency Rating in the Parietal Cortex Frequency judgments activated th e bilateral parieto-occipital sulcus. The precise region of activation in the current study was at the border of the pari eto-occipital sulcus and the intraparietal sulcus. Bilateral parietal activations at these co-ordinates were seen in a previous study during the spatial control task (indoor / outdoor judgment) (Lane et al., 1997a). Although the parietal lobe is generally associated with the dorsal visual pathway, which processes spatial location, this region is not associated with a sing le, specific function. The parietal cortex is an associative region, receiving multimodal inputs, and being activated by a wide range of cognitive tasks. The region activated in the current st udy may correspond with th e caudal intraparietal sulcus (CIP), which studies in monkeys and humans have implicated in grasping, and in processing object orientation (C ulham, and Kanwisher, 2001). Th e parietal cortex is also implicated in arithmetic tasks. A recent metana lysis posits a mental num ber line represented at the horizontal section of the in traparietal sulcus (Dehaene et al., 2004). While this section is anterior to the region found in the current st udy, the above authors propose that additional

PAGE 106

106 attentional orientation on the mental number lin e may occur more posteriorly. Thus it is possible that the activation in the current study refl ects the numerical approximation involved in the frequency judgment task. Summary and Conclusions The aim of this study was to use an emotion rati ng task to dissociate the neural correlates of automatic, bottom-up responses to pictures emotional content a nd of voluntary, top-down responses during explicit emotion rating. This wa s the first study of its kind to employ an eventrelated, factorial design. Event-rela ted stimulus presentation elimin ated the effect of anticipation of either task or emotion. Factorial design and analysis controlled for non-linear differences in the neural responses to each task condition by accounting for the main effect of emotional valence (pleasant or unpleasant), the main eff ect of task (emotion or frequency rating), and interactions between the two (representing top-down modulati on of responses to stimulus content). In agreement with previous studies, we demonstrated a top-down response in the orbitofrontal cortex, and a bottom-up response in the left amygdala. Unlike previous studies, which found top-down responses in the ACC, an d top-down modulation of bottom-up responses in the insula, we found both top-down and bo ttom-up responses in the ACC, and a top-down response in the insula. These results demonstrat e the potential of fMRI to distinguish bottom-up and top-down emotion processing at the event leve l. The pleasant and unpleasant stimuli used in the current study varied in emotional valence but not in arousal. Future studies may test whether including neutral pictures may reveal top-down modulation of bottom-up responses in the insula that are driven by differences in arousal. The ti ming of stimulus presentation used in the current study was previously shown to elicit reliabl e bottom-up responses. Future studies may investigate whether a slower event-related desi gn or a blocked design is necessary to elicit reliable top-down responses in th e ACC. With these improvements, this emotion rating paradigm

PAGE 107

107 may be used to study differential alterations in top-down and bottom-up pathways that may be seen in patients with affective disorders.

PAGE 108

108 Table 5-1 Response time in milliseconds (standard deviation) Stimulus valence Rating task Emotion Frequency Pleasant 1902 (295) 2088 (263) Unpleasant 1899 (265) 2087 (295) Responses < 500 ms were excluded as they most likely reflected carried-o ver late responses to previous trials. ANOVA of task a nd valence revealed a significant ma in effect of task (F (1, 14) = 36.5, p < 0.001).

PAGE 109

109 Table 5-2 Clusters of activation fo r interaction of valence and task. Region Side BA x y z Size F(1,14) t(14) Subgenual cingulate cortex L 25 -422-925410.9 2.4 Insula / MCA L 3611-1510921.8 5.1 Only clusters > 100 mm3 shown. BA: Brodmann Area. X, Y and Z refer to Talairach coordinates (mm). Size: number of 1mm3 voxels. F: result of whole-brain ANOVA (degrees of freedom); score is taken from the peak voxel of th e cluster. t: random effect s statistical score for post-hoc cluster-based contrast [(EU+FP) (EP+FU)] (degrees of freedom). EU: emotion rating on unpleasant pictures, EP: emotion rating on pleasant pictures, FU: frequency rating on unpleasant pictures, FP: freque ncy rating on pleasant pictures BA: Brodmann area. MCA: middle cerebral artery. Italic text: a priori region of interest, tested at lo wer statistical threshold.

PAGE 110

110 Table 5-3 Clusters of activation for main effect of valence. Region Side BA x y z Size F(1,14) t(14) Pregenual cingulate cortex R 32 839181092 15.0 2.9 Postcentral gyrus L 2-39-3156397 16.4 4.2 Posterior cingulate R 308-5212583 23.1 4.5 Posterior cingulate L 30-10-5312288 20.5 4.4 Amygdala L -21-7-10203 13.0 -2.2 Inferior occipital gyrus L 18-39-75-8362 24.9 -5.3 Only clusters > 100 mm3 shown. BA: Brodmann Area. X, Y and Z refer to Talairach coordinates (mm). Size: number of 1mm3 voxels. F: result of whole-brain ANOVA (degrees of freedom); score is taken from the peak voxel of th e cluster. t: random effect s statistical score for post-hoc cluster-based cont rast [(EP + FP) (EU + FU)] (deg rees of freedom); positive values: greater response to positive pict ures, negative values: greater re sponse to negative pictures. EU: emotion rating on unpleasant pictures, EP: emo tion rating on pleasant pictures, FU: frequency rating on unpleasant pictures, FP: frequency ra ting on pleasant pictures. BA: Brodmann area. Italic text: a priori region of interest tested at lower statistical threshold.

PAGE 111

111 Table 5-4 Clusters of activati on for main effect of task. Region Side BA x y z Size F(1,14) t(14) Orbitofrontal cortex L 11-2635-7315 19.04.8 Pregenual cingulate cortex B 24 -13410588 10.43.4 Insula R 42-15879 26.14.6 Insula L -40-55151 16.74.3 Superior temporal gyrus L 42-61-89104 17.94.6 Parahippocampal gyrus R 2816-14-21207 22.05.7 Callosomarginal gyrus L 4, 31-7-3346103 19.54.3 Precuneus R 714-6246172 25.04.9 Inferior frontal gyrus L 45-48348245 20.3-4.7 Middle frontal gyrus R 9472933278 21.6-5.1 Medial frontal gyrus L 6, 8-31949651 21.4-4.6 Inferior frontal gyrus L 9-4516281130 37.0-5.3 Middle frontal gyrus L 6-343521060 28.4-5.1 Superior temporal sulcus L 21, 22-61-4001013 30.6-5.0 Parieto-occipital sulcus L 7, 19-37-67411680 22.9-5.2 Parieto-occipital sulcus R 7, 1940-70401478 21.6-5.4 Only clusters > 100 mm3 shown. BA: Brodmann Area. X, Y and Z refer to Talairach coordinates (mm). Size: number of 1mm3 voxels. F: result of whole-brain ANOVA (degrees of freedom); score is taken from the peak voxel of the cluster. t: post-hoc cluster-based contrast [(EP + EU) (FP + FU)] (degrees of freedom ); positive scores: greater response to emotion rating task, negative scores: greater response to frequency rating task. EU: emotion rating on unpleasant pictures, EP: emotion rating on pleasan t pictures, FU: freque ncy rating on unpleasant pictures, FP: frequency rating on pleasant pictures. BA: Brodma nn area. MCA: middle cerebral artery. Italic text: a priori region of interest tested at lower statistical threshold.

PAGE 112

112 Figure 5-1 Main effect of task. A & C) statistical maps showing a ma in effect of task in the right insula (A) and the left OFC (C). L: left. R: right. Z: position of axial slice in Talairach space. Color scale indicates F score on two-way ANOVA (see methods). B & D) BOLD responses in the right insula (B) and left OFC (D). Error bars show standard error. Vertical lines show beginning and end of trial. EP: em otion rating on pleasant pictures, EU: emotion rating on unpleasant pi ctures, FP: frequency rating on pleasant pictures, FU: frequency ra ting on unpleasant pictures.

PAGE 113

113 Figure 5-2 Main effect of valen ce. A) reduced-threshold statistical map showing a main effect of valence at the left amygdala. Res ponses are only shown inside the a priori regions of interest (yellow circles). L: left. R: right. Y: position of coronal slice in Talairach space. Color scale indicates F score on two-way ANOVA. B) BOLD response in the left amygdala. Error bars show standard er ror. Vertical lines show beginning and end of trial. EP: emotion rating on pleasant pictures, EU: emotion rating on unpleasant pictures, FP: frequency rating on pleasa nt pictures, FU: frequency rating on unpleasant pictures.

PAGE 114

114 Figure 5-3 Responses in the anteri or cingulate cortex. A) Sagittal slice from a single participant indicating regions showing main effects of valence and task at the pACC. Responses are only shown within the a priori region of interest (yello w line). A: anterior. P: posterior. X: position of slice in Talair ach space. B-D) BOLD responses in the regions illustrated in A. Error bars s how standard error. EP: emotion rating on pleasant pictures, EU: emotion rating on unpleasant pictures, FP: frequency rating on pleasant pictures, FU: frequenc y rating on unpleasant pictures.

PAGE 115

115 CHAPTER 6 DISCUSSION Summary The experiments in this dissertation represent an effort to develop a functional magenetic resonance imaging (fMRI) probe for affective diso rders. Previous studies of major depressive disorder (MDD) highlighted three brain regions where resting meta bolism or responses to simple emotion tasks were altered: the amygdala, orb itofrontal cortex (OFC), and anterior cingulate cortex (ACC) (Davidson et al., 2003;Drev ets et al., 1992;Mayber g, 2003;Sheline et al., 2001;Siegle et al., 2002). These regions are posite d to be involved respectively in rapid generation (LeDoux, 2000), contextual modificati on (Rolls, 1999), and exp licit appraisal (Lane et al., 1997a) of emotional responses. The worki ng hypothesis of this work is that the amygdala mediates bottom-up (stimulus-driven) emotion processing and that the OFC and ACC mediate top-down (knowledge-based) emoti on processing. This hypothesis was investigat ed under two specific aims: Aim 1. Assess the validity and reliability of the amygdala response to a face matching paradigm. Face matching elicits activat ion the amygdala reliably, due, it is thought, to bottomup processing of the facial features that commun icate emotion. In order to separate responses to facial emotion from responses to non-emotiona l components of face matching, we modified the face matching task to include an intermediate control condition in whic h neutral faces were matched by identity. The left and right amygdalae responded to both the emotion and the identity matching conditions, and the left amygdala response habituated to emotion matching but not to identity matching. These results suggest that in the face matching task, the amygdala response to emotion is confounded with non-emotional releva nce detection involved in matching facial

PAGE 116

116 features. We concluded that amygdala activati on in the face matching task is not a specific indicator of bottom-up processing of facial emotion in the amygdala. Aim 2. Dissociate the neural correlat es of top-down and bottom-up emotion processing using a picture rating task. We designed an emotion rating paradigm in which the emotional content of pictures and the type of rating task were varied independently. Bottom up responses were identified mapping the main effect of picture c ontent (pleasant or unpleasant), top-down responses were identified mapping the main effect rating task (emotion rating or frequency rating), and top-down modulation of bottom-up responses was identified by mapping the interaction of picture content and rating ta sk. Bottom-up activation in the amygdala occurred to unpleasant pictures during bot h emotion rating and frequency ra ting tasks. The orbitofrontal cortex and insula responses were selective for top-down processing regardless of emotional content. No significant interaction effects were found. The ACC responded to both stimulus content and task demands, and the responses we re less specific than those in other regions. Future studies should investigate whether including neut ral pictures may elic it interaction effects by varying emotional arousal, and should investigat e whether a slower stimulus presentation rate may elicit stronger top-down responses in the ACC. These results showed that an event-related picture rating task can dissociate bottom-up processing in the amygdala from top-down processing in the orbitofrontal cortex and insu la. This paradigm allo ws two brain regions implicated in major depressive disorder to be st udied separately and in pa rallel, and may be used to investigate the effect of illness of bottom-up and top-down emotion processing. Proof of Concept: Dissociated Responses to Disgust and Arousal In study one, the disgust study (Chapter 3) we demonstrated dissociated neural correlates of the emotional category disgust and the emoti onal dimension arousal. We separated disgusting pictures into low arousal (contam ination) and high arousal (mutil ation) sets, and compared them

PAGE 117

117 with low disgust, high arousal pictures (threat) We showed that disgusting pictures elicited responses in the anterior insula, while emoti onally arousing pictures evoked a response in the occipito-temporal cortex. In this study we illustra ted the importance of visualizing the changes in BOLD signal that underlie regional responses on the statistical ac tivation map. Comparing signals in this way suggested a relationship betw een regional responses and emotion ratings that was not obvious from the statistical maps. Inves tigating this relationshi p further revealed a quantitative relationship between rating of disgust and the respons e at the insula, and between ratings of arousal and the response at the occipito-temporal cortex. A Response at the Amygdala, but not Specific to Emotion In study one, technical factors prevented det ection of responses in the amygdala, a region that has been consistently associated with affective disorders and perception of negative emotional stimuli (Drevets et al., 1992;Le Doux, 2000;Sheline et al., 2001).In study two, we employed a face matching task that had been repo rted to activate the amygdala reliably, with improved scanning parameters that reduce susceptib ility artifact (Chapter 4). We attempted to dissect the emotional component of the face matchi ng task by comparing that task with a similar one in which all faces were neutral. In this way, we hoped to test whether face matching elicited a bottom-up response in the amygdala. Contrary to expectations, activation in the amygdala did not differ when matching emotional faces by ex pression and when matching neutral faces by identity. A search of the liter ature suggested that the amygdala responds not only to aversive emotional stimuli, but during any task involving relevance detec tion, which is th e identification of stimuli that may alter the individuals percei ved chances of reward (S ander et al., 2003). This notion suggests two interpretations of our result s in light of the hypothesized bottom-up role of the amygdala. First, pairs of faces that match ar e intrinsically relevant, regardless of their emotional content. The amygdala response is ther efore bottom-up, being driv en by the perceptual

PAGE 118

118 cues presented in facial features, but is not specific to perception of facial emotion. Second, neutral faces usually less relevant than emotional faces, sin ce they convey no social signal of emotion (as suggested by greater amygdala responses to emotional faces than neutral faces (Morris et al., 1996;Whalen et al., 1998b)) and in the identity matching task, the desire to perform a correct match may confer relevance on the matching face 2. Thus neutral faces acquire temporary relevance, and the associated amygdala response is driven by th e demands of the task: a top-down influence. The conclusion drawn from both these interpretations is that the amygdala response during the emotional face matching is co ntaminated by factors other than bottom-up processing of emotion. However, some features of the responses were selective for emotion. By examining the BOLD responses during separate bl ocks of each task, we found that the response in the amygdala to emotion matching habituated, a characteristic of responses to emotional stimuli (Breiter et al., 1996;Wright et al., 2001). During identity matc hing, by contrast, the amygdala response did not habituate. This anal ysis shows how subtle differences in task paradigm can affect neural re sponses and blur interpretation. We chose to investigate the face matching task because it had previously been used in conjunction with a labeling task to demonstr ate top-down modulation of bottom-up responses (Hariri et al., 2000). In the previous study, th e amygdala response to face matching decreased during face labeling, and was negatively correlated with activity in th e right ventrolateral prefrontal cortex. We did not repeat the face labe ling experiment because the results of the face matching experiment suggest that the amygdala response to emotion is confounded by task 2 The control condition pixel-pattern matching also required a correct match, but did not activate the amygdala. This may be because of specialization of the amygdala fo r naturalistic stimuli, especially faces (Leonard et al., 1985); furthermore, parietal activation seen only in control condition suggested pattern matching was performed using spatial processing in the dorsal visual pathway, rath er than object processing in the ventral visual pathway (Culham, and Kanwisher, 2001)

PAGE 119

119 demands. We therefore turned to a paradigm in which emotional content and task demands could be controlled separately. An Event-related Emotion Rating Task Partia lly Replicates Responses to Block-related Tasks Previous studies have investig ated top-down processing of em otion with rating paradigms. In study three, the emotion rating study (Chapter 5), we designed a task in which the emotional content of presented pictures and the type of rating task were i ndependently controlled. The aims of this task were to dissociate bottom-up respons es to picture content from top-down responses during explicit emotion rating, and to demonstrat e modulation of bottom-up responses different rating tasks. Participants were shown pleasant and unpleasant pictur es, and were either asked to rate how pleasant they found the picture, or how frequently it appeared on television. We varied stimulus content at the event level in order to ensure that bottom-up responses were not confounded with expectation, a top-down factor that may be elicited by blocked stimulus presentation. We also varied task instructions at the event-level to ensure that responses to the two factors were not biased by temporal effects. We employed a factorial analysis in order to identify brain regions responding exclusively to bottom-up or to top-down factors (main effects), and to identify regions in which the two factor s interact. Replicating the findings of previous studies, we found a main effect of rating task in the OFC and ACC, and a main effect of stimulus content in the left amygdala. Examination of the BOLD responses in these regions confirmed that the OFC responded to emotion rating regardle ss of stimulus valence, and that the amygdala responded to unpleasant pictures regardless of th e task. Different regions of the ACC exhibited main effects of both top-down and bottom-up fa ctors, but BOLD responses at the ACC only weakly supported the statistica l findings. The current study thus succeeded in separating bottomup and top-down processing in the amygdala and OF C, but failed to replicate previous findings

PAGE 120

120 associating the ACC with topdown processing and showing top-down modulation of responses in the amygdala and insula. Although previous studies reporte d that rating pictures modul ated the responses in the amygdala and insula (Liberzon et al., 2000;Phan et al., 2004;Taylor et al., 2003), our study failed to replicate these findings. Seve ral features of the design may ha ve contributed to this. First, previous studies compared unpleasant and neutral pictures, which differ in emotional ratings of both valence and arousal, but th e current study compar ed unpleasant and pleasant pictures, which differed in valence but not in arousal. Comparing stimuli with different arousal ratings may be necessary to distinguish botto m-up responses in the insula because this region has been associated with autonomic arousal responses (Critchley et al., 2004). Second, in the current study, the rating task was preceded by a training period. Training may have caused a shift from automatic, general arousal res ponses mediated by the right am ygdala to habitual, conscious responses to specific emotional content mediated by the left amygdala (Gur et al., 2002). Future studies may include neutral pictures in order to compare different levels of emotional arousal, and may investigate how neural responses during training in emotion rating evolve over time. The current study only found a weak top-down response in the ACC. While the current study used an event-related design, previous stud ies using block-related designs reported strong top-down responses in the ACC (Lane et al., 19 97a;Taylor et al., 2003). The response in the ACC may depend on the establishment of a c ognitive set over several consecutive emotion ratings, an effect that is eliminated in even t-related designs. The curre nt study also failed to detect top-down modulation of bottom-up responses The modulating effect may only be elicited by a block-related design or by an event-rela ted design with a slower presentation rate. A previous study investigated the most efficient presentation rate for eliciting bottom-up responses

PAGE 121

121 to emotional stimuli (Serences, 2004). Future stud ies may investigate the effect of stimulus blocking and presentation rate on top-dow n responses to emotion rating tasks. Technical Considerations Scanning Parameters In study two, improved scanning parameters f acilitated detection of activation in the amygdala. This region is prone to susceptibility ar tifact, a loss of signa l due to the nearby interface between air and tissue at the nasal sinuses. Because scanning parameters that reduce susceptibility artifact may also reduce sensi tivity to BOLD signal changes, defining optimal parameters is an ongoing challenge. Several authors have proposed optimized scanning parameters for minimizing suscepti bility artifact at the amygdala; however none of these studies were able to validate their appr oach using an emotion task be cause neural responses in the amygdala to emotional stimuli habituate, pr eventing the comparison of BOLD responses obtained during two successive s cans (Chen et al., 2003;Merboldt et al., 2001;Robinson et al., 2004). We nonetheless tested two sets of para meters using emotional stimuli from study one, comparing cubic voxels with side of 3.8 mm or 3.0 mm. Scanning using 3.0 mm voxels produced better images than with 3.8 mm voxels, but BOLD responses in the ventral temporal cortex appeared to lose their sensitivity to different le vels of emotional arousal when scanning using 3.0 mm voxels (results not shown). We therefore co ncluded that scanning with 3.8 mm voxels minimizes susceptibility artifact at the amygdala while maintaini ng sensitivity to BOLD effects. These parameters achieved coverage of ten out of twelve participants in study two, but in study three coverage of the amygdala was achieved in only ten out of fi fteen participants. An additional challenge arose in the emoti on rating study, in which both the amygdala and ACC were regions of interest. Responses in th e subgenual ACC (sACC) we re discarded in this study due to susceptibility artifact The interface between air and tissue that causes artifact at the

PAGE 122

122 sACC is nearly perpendicular to the interface near the amygda la. Because the severity of susceptibility artifact depends upon the relationshi p between the orientati ons of the air-tissue interface and the plane of sca nning, a scanning plane that provi des good images of the amygdala is likely to provide poor images of the sACC. Future work should investigate the optimal scanning parameters for imaging both the amygda la and sACC using advanced pulse sequences (such as spiral echo-planar imaging), oral magnetic shim devices (Wilson, and Jezzard, 2003), or smart phantoms designed to mimic BOLD respons es in regions of susceptibility artifact. Connectivity Analysis The original proposal for this work con ceived of the amygdala, OFC, and ACC as a network. For this reason we planned to perform functional connectivity analyses to investigate the interactions between th ese regions. Functional conne ctivity analysis estimates communication between brain regi ons by calculating the correlation between the signals at each region (Buchel et al., 1999). Within-condition interregional covariance analysis (WICA) investigates how this correlati on changes as a function of task conditions (He et al., 2003). Our original goal was to repeat the face matching and labeling experiment of Hariri et al. (2000) in order to replicate interactions between the amygdala and prefront al cortex. Since, however, the results of this study cast doubt upon the emotional si gnificance of the amygdala response, we did not continue to develop this paradigm. In th e next experiment, the emotion rating task, we employed an event-related design. Both WICA and the psychophysiological interactions approach used originally in Ha riri et al. (2000) inve stigate the influence of task on functional connectivity by calculating BOLD si gnal correlations between brain regions during different task blocks (Friston et al., 1997). In our rapid, event-related design, th e responses to different event types were commingled, and could not be analyzed in this way. Further development of novel analysis techniques may allow functional connect ivity analysis within event-related designs.

PAGE 123

123 Current Trends in Functional Imaging of Major Depressive Disorder Recent functional neuroimaging studies of M DD have focused on three features of the illness: emotional bias, impaired regulation of emotion, and impaired concentration (Leppanen, 2006). Patients with depression pay more attent ion to negative stimuli and less attention to positive stimuli, whereas healthy controls tend to have the opposite bias. This shift in bias may be revealed by behavioral measur es of attention, or by identifying brain regions that reverse their responses to positive and negative stimuli in patien ts with MDD. Impaired emotional regulation has so far been investigated only indirectly, by looking at altered patterns of cortico-limbic connectivity in patients compared with controls Impaired concentration has been investigated with memory and attention paradigms. Emotional Bias Several imaging studies have shown that in M DD the subcortical and cortical responses to positive and negative stimuli revers e their sign. The putamen and fusiform gyrus showed greater activation to happy faces than sad faces in contro ls, but showed greater activation to sad faces than happy faces in patients (Surguladze et al., 2005). Similarly, responses to sad faces at the ventral striatum and amygdala were elevated in patients with depression (Fu et al., 2004). Treatment with antidepressant medication reduced these limbic responses, and increased the response in the lateral prefr ontal cortex. A corresponding reve rsal was observed in the ventromedial prefrontal corte x, which responded to happy mood induc tion in controls, and to sad mood induction in patients (Keedwell et al., 2005). Th ese findings show that biased responses to emotional stimuli may be elicited in both cortical and subcortical regions. However, the passive viewing paradigms used to elicit subcortical responses and the m ood induction paradigm used to elicit cortical responses are not directly compar able. Future studies may investigate subcortical and cortical biases side by side using tasks designed to separate top-down and bottom-up

PAGE 124

124 emotion processing, such as the emotion rati ng task presented in Chapter 5. As already discussed, this task must be improved in order to elicit reliable res ponses in ventromedial prefrontal regions. Following such an improvement, studies of this kind w ith a sufficiently large patient group may reveal heterogeneity in neur al response bias, with some patients exhibiting bias in cortical regions, and othe rs exhibiting bias in subcortical regions. This approach may be used to investigate the neural basis for heterogeneity in symptom profiles. The effect of emotional bias may be reveal ed behaviorally by measuring its effect on memory. Patients with depression recall a larg er proportion of negative words and a smaller proportion of positive words than healthy controls. Electroencephalographic evidence suggests that this bias in recall is mediated by attenti onal bias during the encodi ng of stimuli (Leppanen, 2006). The neural correlates of encoding were investigated usi ng event-related fMRI, comparing responses to faces that were later recalled wi th those that were not. Successfully recalled negative faces evoked greater responses in the left amygdala in adolescents with MDD compared with controls (Roberson-Nay et al., 2006). A sim ilar effect was seen in adults with remitted depression, but only after mood challenge (Ramel et al., 2006). This implies that bias towards negative stimuli is enhanced following an insult to emotional regulation. By using a behavioral measure of emotional bias to gui de the analysis of functional imaging data, these experiments provide evidence linking increase d amygdala activity in MDD with increased attention to and encoding of negative emotional stimuli. Behavioral measures of emotional bias may al so be obtained using a dot probe task or by measuring facial expression recogn ition thresholds. In dot-probe tasks, two faces are displayed, followed by a dot in the location of one of the tw o faces. The participant must respond as quickly as possible to the location of the dot. Patients wi th MDD respond faster when the dot is displayed

PAGE 125

125 in the location previously occupied by a sad face, and several studies have confirmed that this effect is specific to MDD, and is not seen in patients with general anxi ety disorder or social phobia (Leppanen, 2006). A dot probe task with f earful faces was used in an fMRI study of healthy individuals (Pourtois, and Vuilleumier, 2006) This study compared valid trials, in which the fearful face was shown in the same location as the subsequent dot, with invalid trials, in which the fearful face was shown on the opposite si de as the dot. During valid trials, response increased in the occipital cortex. During invalid trials, responses d ecreased in the parietal cortex and increased in the orbitofrontal cortex. These results were proposed to reflect the attentional draw of fearful faces during vali d trials, and the increased cost of disengaging attention from fearful faces during invalid trials. Because the do t probe task provides a behavioral measure of the effect of emotion on atten tion, it represents a more powerfu l paradigm than passive face viewing for the investigati on of affective disorders. Facial expression recognition thresholds are measured by having participants judge pictures of faces that have been morphed to di splay varying degrees of emotion. Patients with MDD and healthy individuals who report anhedoni a fail to recognize happy expressions that are recognized by healthy controls (Leppanen, 2006). Wh ile morphed faces have been used in fMRI studies of MDD (Fu et al., 2004;Surguladze et al ., 2005), recognition thresholds have not yet been used as a behavioral covariate in fMRI analysis. Patients with MDD may experience a greater implicit attentional draw toward sad f aces in the dot-probe task, and fail to explicitly recognize happy faces in the recogniti on threshold tasks. These results imply that bias toward sad faces and bias against happy faces may be me diated respectively by bottom-up and top-down systems. Future imaging studies of patient s performing these tasks should address this hypothesis.

PAGE 126

126 Emotion Regulation Recent studies of patients with MDD have inve stigated emotional regulation indirectly, by measuring correlation between cortical and li mbic responses. These interactions were investigated using a new method that measures low frequency fluctuations in the blood oxygenation-level dependent signal (LFBF) (Ana nd et al., 2005a). In patients with MDD compared with healthy controls, negative emoti onal pictures evoked increased responses in the ACC, amygdala, ventral striatum and medial thalamus. However, correlations in LFBF between ACC and limbic regions were reduced. These corre lations were increased in patients following antidepressant treatment, suggesting that medi cation restored regulat ion (Anand et al., 2005b). Impairment of regulation may al so predict patients responses to cognitive behavioral therapy. Successful therapy was predicted by low responses at the subgenual ACC and high responses at the amygdala during a task in whic h participants briefly saw emo tional words and then attempted to sustain their emotional responses (Siegle et al ., 2006). This pattern of activity was proposed to represent decreased regulation by the sACC and in creased emotional reacti vity at the amygdala, suggesting that patients with impaired regulati on benefit most from cognitive behavioral therapy. The latter experiment may be improved by meas uring correlations between the subgenual ACC and amygdala. The former experiments may be im proved using a paradigm with several task condition that are designed to va ry the degree of emotional regul ation, and to thus potentially alter cortico-limbic connectivity. As already discussed, this effect was demonstrated by comparing emotional face matching with face labeling (Hariri et al., 2000). The correlation between BOLD signal in the amygdala and in the right ventral PFC reversed its sign between the two conditions. Increased prefr ontal and decreased amygdala re sponses during labeling were interpreted as representing emotion regulation. However, our investigation of the face matching task suggested that the amygdala response to emotion was confounded with non-emotional

PAGE 127

127 recognition of matching perceptual features. Other studies have investigated emotion regulation in healthy individuals, but none has repeated the connectivity approach of Hariri et al. (2000). A review of studies of the cogni tive control of emotion propos ed two systems for emotional regulation in the prefrontal corte x, a dorsal indirect system, and a ventral direct system (Ochsner, and Gross, 2005). The former was associated wi th tasks in which emotions were altered by reappraisal, and the second with tasks in which emotions were altered by extinction or reversal. The same regions associated with indirect regu lation also showed greater activation in patients than controls performing cognitive tasks, as discussed below. Cognitive Tasks MDD is characterized by impaired concentratio n; therefore patients ma y also have altered neural responses to non-emotional, cognitivelydemanding tasks. Several authors have studied the neural correlates of atten tion and working memory in patients with MDD. The Stroop task tests attention by requiring participants to name the ink color of a word that names a different color. This requires them to pay attention to on e aspect of a stimulus and ignore another. The nback task exercises working memory by pres enting a stream of characters, and asking participants to respond to stimuli that match a pr eviously presented stimulus. This requires them to hold in memory a sequence of characters, a nd to shift and update that sequence with every new character. Patients with MDD exhibit incr eased responses to both the Stroop and n-back tasks in the ACC and the lateral and dorsola teral PFC (Harvey et al., 2005;Matsuo et al., 2006;Wagner et al., 2006). Conversely, load-depende nt responses to an n-back task were decreased in patients at the medial orbitofront al cortex and rostral ACC (Rose et al., 2006). Despite altered neural responses, patients performe d as well as controls on all tasks. Increased responses in dorsolateral PFC were interpreted as representing great er effort required to achieve equal performance. The results may reflect an effect of mood state on the neural responses

PAGE 128

128 during cognitive task performance. To date, no f unctional imaging studies of patients with MDD have investigated whether responses during cognitive tasks may be modulated by the addition of emotional stimuli. However, such studies in h ealthy volunteers have supported the idea that cognitive and emotional circuits compete, particul arly in the anterior cingulate cortex. Because this region is being seen as increasingly important in affective disorders, combining cognitive tasks with emotional stimuli may prove a promising approach for future studies. Future Direction: the Anterior Cingulate Co rtex and the Interaction of Emotion and Cognition Maybergs model of major depressive diso rder emphasizes the competition between a dorsal, cognitive or sensorimotor circuit and a vent ral, emotional or visceral / autonomic circuit (Mayberg, 2003). This competition between cognitiv e and emotional circuits was also noted in an early review of PET studies. Tasks involvi ng visual discrimination and working memory activated the dorsolateral PFC and dorsal ACC, but deactivat ed the ventromedial PFC and ventral ACC (Drevets, and Raichle, 1998b). The opposite pattern was seen for tasks involving mood induction, emotional stimulus viewing, and symptom provocati on in patients with affective disorders: these deactivated the dorso lateral PFC and dorsal ACC, but activated the ventromedial PFC, ventral ACC, and in tasks involving visual stimuli the amygdala. A later review of fMRI studies confir med opposing patterns of responses in the dorsal and ventral ACC to cognitive and emotional tasks, suggesting that different regions of the ACC perform similar functions in cognitive and emotional contexts (Bush et al., 2000). The ventral ACC may be further divided into pregenual and subgenual subd ivisions. Resting metabolism in the pregenual ACC may predict recovery from depression, a nd resting metabolism in the subgenual ACC is elevated in patients. The importance of the subgenual ACC in depression was highlighted by a recent pilot study of deep brain stimulation (M ayberg et al., 2005). Six patients with multiple

PAGE 129

129 drug-resistant depression had electrodes impl anted near the subgenual ACC. Electrical stimulation was accompanied by improved symptoms in four out of six patients. Functional imaging paradigms designed to activate the su bdivisions of the ACC selectively may prove informative in the study of major depressive disorder. The dorsal and ventral divisions of the ACC may mediate conf lict monitoring in cognitive and emotional domains (Bush et al., 2000). This was suggested by studies dissociating dorsal and ventral ACC responses to cogniti ve and emotional Stroop tasks (B ush et al., 1998;Whalen et al., 1998a). In Stroop tasks, conflicting stimuli are used to interfere with respons es to a task; in this case, participants were instructed to count the number of words on the sc reen. Cognitive conflict was elicited by displaying distr acting number words, for example, the number two displayed four times, which prompted two co nflicting responses within the ta sk set. Emotional conflict was elicited by displaying emotional words, with the assumption that the resulting emotional response would interfere with counting. Cognitive c onflict activated the dorsal ACC (Bush et al., 1998), whereas emotional conflict activated the ventral ACC (Whalen et al., 1998a). Cognitive theorists debate the extent to which cognitive and emotional conflict are comparable, but agree that both involve the diversion of attentional re sources from correct resp onse selection (Algom et al., 2004;Chajut et al., 2005;Dalgle ish, 2005). In cognitive conflict th is occurs within the taskrelated response set, whereas in emotional confli ct attention is drawn out side the task-related response set. A mechanism for regulation of atte ntion during cognitive conflict was proposed in the conflict hypothesis, which stat es that detection of conflict results in increased cognitive control (Cohen et al., 1990;Cohen et al., 2000). Conflict monitoring and cognitive control were associated respectively with the dorsal A CC and dorsolateral PFC. The dorsal ACC was activated selectively by conflicting Stroop trials, whereas th e dorsolateral PFC was activated

PAGE 130

130 selectively by instructions announcing difficult trials (which may elicit anticipatory cognitive control) (MacDonald, III et al., 2000). Furthermor e, trial-by-trial analysis revealed that dorsal ACC activation to conflicting Stroop trials wa s followed in subsequent trials by increased dorsolateral PFC activation and sl ower response times (indicating increased contro l) (Kerns et al., 2004). Extending this model of cognitive conf lict to the emotional domain may provide a useful framework for working hypotheses about the role of the ventral ACC in emotion processing, and its functional inte ractions with other regions. The ventral ACC consists of two subdivisions, which may each play a distinct role in conflict monitoring, and which may each have dis tinct functional relations hips with other brain regions. The pregenual ACC and su bgenual ACC lie anterior and inferior to the genu of the corpus callosum. The subgenual A CC may detect emotional conflict in the visceral / autonomic circuit and enhance processing in this circuit while decreasing pr ocessing in the sensorimotor circuit. This region was associat ed with symptoms of depression and in healthy individuals with sad mood induction (Mayberg et al., 1997;Vogt, 2005). The role of the subgenual ACC in the withdrawal component of sad m ood is supported by evidence in monkeys that the subgenual ACC sends projections to the ventrolateral region of the periaqueductal gr ay area, allowing it to coordinate a quiescent stance (Ongur, and Price, 2000), and by evidence that neurons in the subgenual ACC fire as monkeys prepare to sleep (Rolls et al., 2003). The pregenual ACC may detect conflict at a higher level, between the se nsorimotor and visceral / autonomic circuits, and recruit emotional control, allowing engagement in sensorimotor activity despite concurrent emotional conflict. This region was shown to pred ict recovery in patients with depression and in healthy individuals responded to happy mood induction (Mayberg et al., 1997;Vogt, 2005). The pregenual ACC responded during successful perf ormance of a Stroop task with emotional

PAGE 131

131 distractors (Whalen et al., 1998a) and responded selectively to tasks involving symbolic representation of emotion (such as placebo and e xpectancy tasks) (Eisenberger, and Lieberman, 2004). The role of the pregenual ACC in engagement is supported by evidence in monkeys that the pregenual ACC sends projections to the late ral region of the peria queductal gray, allowing it to coordinate a confrontational or engaged stan ce (Ongur, and Price, 2000). Studies of cognitive conflict described a functional relationship between the dorsal ACC and the dorsolateral PFC (Kerns et al., 2004;MacDonald, III et al., 2000). The pregenual and subgenual ACC may have parallel functional relationships. A recent study in humans used diffusion-weighted imaging to trace fibers from the ACC, showing specifi c connections between the subgenual ACC and amygdala, and between the pregenual ACC and ventromedial PFC. The amygdala has been associated with both the generati on of visceral / autonomic res ponses (Critchley et al., 2002), which may signal emotional conflict, and with the direction of attention towards emotionally salient stimuli (Adolphs et al ., 2005;Sander et al., 2003), wh ich may be a component of withdrawal from sensorimotor activity follo wing detection of emotional conflict. The ventromedial PFC has been associated with em otional control (Ochsner, and Gross, 2005), and may be recruited via connections from the pr egenual ACC after this region detects conflict between cognitive and emotional circuits. Taken t ogether, these findings suggest that activation of the subgenual ACC may occur when emotiona l stimuli elicit withdrawal from a cognitive task, and that activation of the pregenual ACC may occur when emotional stimuli are successfully ignored, allowing continued engageme nt with a cognitive task. Functional imaging paradigms involving cognitive tasks with emotional distractors may be used to test these predictions, and to investigate functional relationships between the subgenual ACC and amygdala and between the pregenua l ACC and ventromedial PFC.

PAGE 132

132 Designing a task paradigm in which emotional stimuli impair the par ticipants involvement in a cognitive task may be challenging, but may advance our understanding of the role of the ACC in depression and in normal emotion pr ocessing. The emotional Stroop task above activated the pregenual ACC, but not the subgenual ACC (Whalen et al., 1998a). In this task, emotional words did not affect task performance. Therefore, one challenge for future studies of this type will be to elicit emotional responses th at are sufficiently strong to impair cognitive task performance. Furthermore, emotional conflict sh ould be elicited at several levels: increasing conflict should result in decreasi ng task performance. Emotional ra tings of stimuli at each level and measures of task performance could be used to define individual engagement thresholds, enabling parametric analysis of the neural corr elates of disengagement. A second challenge for these studies will be to investigate functional relationships between the subgenual ACC and amygdala, and between the pregenual ACC and ve ntromedial PFC. This may be pursued first using functional connectivity analysis, and second using trial-by-trial an alysis to dissociate conflict detection from subsequent recruitment of control, as was done for cognitive conflict (Kerns et al., 2004). These approaches may elucidate normal emotion processing by testing whether the subdivisions of the ACC play specific roles in me diating emotional influences on cognitive task performance. Testing functional re lationships of the vent ral subdivisions of the ACC will test whether the functional interactions seen in cognitive conflict can be extended to emotional conflict. These approaches may also elucidate the neural bases for depression by providing a specific probe for activity in the subg enual ACC, which is implicated in depressed mood, and in the pregenual ACC, which is prop osed to mediate recovery from depression. Dissociating emotional conflict monitoring from emotional control may reveal whether depressed mood is mediated by hypersensitivity to conflict, or by impaired control. Future

PAGE 133

133 studies may continue to build more refined models of emotion processing, and thus may pave the way for improved understanding of the neural ba ses of affective diso rders. The functional neuroimaging paradigms developed in these studies may eventually be used to guide the choice of conventional treatments, and to guide the deve lopment of novel treatments such as deep brain stimulation.

PAGE 134

134 APPENDIX IAPS PICTURES CODES Table A-1 Contamination pictures. Code Description 1270 Roach 1274 Roaches 1275 Roaches 1280 Rat 1945 Turtle 2720 Urinating 2730 NativeBoy 7360 FliesOnPie 7380 RoachOnPizza 9005 HIV Tattoo 9006 HIV Tattoo 9008 Needle 9090 Exhaust 9280 Smoke 9290 Garbage 9300 Dirty 9320 Vomit 9330 Garbage 9340 Garbage 9373 Garbage 9390 Dishes 9560 Duck in oil 9561 Sick kitty 9700 Workers-trash 9830 Cigarettes

PAGE 135

135 Table A-2 Mutilation pictures. Code Description 3000 Mutilation 3010 Mutilation 3010 Mutilation 3015 Accident 3030 Mutilation 3051 Mutilation 3053 Burn victim 3060 Mutilation 3061 Mutilation 3062 Mutilation 3063 Mutilation 3064 Mutilation 3071 Mutilation 3080 Mutilation 3100 Burn victim 3102 Burn victim 3110 Burn victim 3120 Dead body 3130 Mutilation 3140 Dead body 3150 Mutilation 3168 Mutilation 3170 Baby tumor 3181 BatteredFem 3250 OpenChest 3261 Tumor 3266 Injury 3400 Severed hand 9253 Mutilation 9265 Hung man 9405 Sliced hand 9430 Burial 9433 Dead man 9490 Corpse

PAGE 136

136 Table A-3 Threat pictures Code Description 1050 Snake 1051 Snake 1052 Snake 1120 Snake 1230 Spider 1300 Pit Bull 1931 Shark 2682 Police 3500 Attack 3530 Attack 6230 Aimed Gun 6242 Gang 6243 AimedGun 6244 AimedGun 6250 AimedGun 6260 AimedGun 6300 Knife 6314 Attack 6350 Attack 6510 Attack

PAGE 137

137 Table A-4 Neutral pictures Code Description Code Desc ription Code Description 1450 Gannet 5740 Plant 7190 Clock 1510 Dog 5750 Nature 7205 Scarves 1601 Giraffes 5800 Leaves 7207 Beads 1603 Butterfly 5831 Seagulls 7211 Clock 1670 Cow 5875 Bicyclist 7217 Clothes rack 1670 Cow 5900 Desert 7224 File cabinets 1812 Elephants 6150 Outlet 7233 Plate 1999 Mickey 7000 Rolling Pin 7234 Ironing board 2000 Adult 7002 Towel 7235 Chair 2058 Baby 7004 Spoon 7285 Tomatoes 2312 Mother 7006 Bowl 7325 Watermelon 2331 Chef 7009 Mug 7490 Window 2383 Secretary 7010 Basket 7491 Building 2389 Teens 7020 Fan 7495 Store 2487 Musician 7025 Stool 7496 Street 2514 Woman 7030 Iron 7500 Building 2575 Propeller 7031 Shoes 7501 City 2580 Chess 7034 Hammer 7545 Ocean 2655 Child 7035 Mug 7550 Office 2791 Balloons 7040 Dust pan 7560 Freeway 2840 Chess 7050 Hair dryer 7595 Traffic 2870 Teenager 7060 Trash Can 7600 Dragon 2880 Shadow 7080 Fork 7620 Jet 5020 Flower 7090 Book 7705 Cabinet 5120 Pine needles 7095 Headlight 7710 Bed 5200 Flowers 7096 Car 7900 Violin 5220 Nature 7100 Fire hydrant 7950 Tissue 5300 Galaxy 7110 Hammer 8160 RockClimber 5390 Boat 7130 Truck 8161 Hang glider 5410 Violinist 7140 Bus 8162 HotAirBalloon 5510 Mushroom 7150 Umbrella 5534 Mushrooms 7160 Fabric 5600 Mountains 7170 Light Bulb 5720 Farmland 7175 Lamp

PAGE 138

138 Table A-5 Emotion rating pictures Pleasant Unpleasant Code Description Code Description 1670 Cow 1050 Snake 1726 Jaguars 1052 Snake 1812 Elephants 1101 Snake 2391 Boy 1120 Snake 4533 AttractiveMan 1280 Rat 5260 Waterfall 1300 Pit Bull 5480 Fireworks 2440 NeutGirl 5750 Nature 2691 Riot 5780 Nature 2870 Teenager 5910 Fireworks 4621 Harassment 7080 Fork 5531 Mushroom 7270 IceCream 6000 Prison 8021 Skier 7006 Bowl 8186 Skydivers 9102 Heroin 8340 Plane 9600 Ship

PAGE 139

139 Table A-6 Frequenc y rating pictures Pleasant Unpleasant Code Description Code Description 1313 Frog 2141 GrievingFem 1340 Women 2312 Mother 1920 Porpoise 2690 Terrorist 1999 Mickey 2752 Alcoholic 2160 Father 2880 Shadow 2170 Mother 6200 AimedGun 2579 Propeller 6243 AimedGun 4274 AttractiveFem 6315 Attack 4689 EroticCouple 7037 Mug 5270 Nature 7090 Book 5623 Windsurfers 7110 Hammer 5991 Sky 9110 Puddle 6250.2 IceCream 9182 Horses 8178 Sailboat 9190 Woman 8370 Rafting 9415 Handicapped

PAGE 140

140 LIST OF REFERENCES Adolphs,R. (2002). Neural systems for recogniz ing emotion. Current Opinion in Neurobiology 12 169-177. Adolphs,R., Gosselin,F., Buchanan,T.W., Tran el,D., Schyns,P., and Damasio,A.R. (2005). A mechanism for impaired fear recogn ition after amygdala damage. Nature 433 68-72. Adolphs,R., Tranel,D., and Damasio,A.R. (2003). Dissociable neural sy stems for recognizing emotions. Brain and Cognition 52 61-69. Adolphs,R., Tranel,D., Damasio,H., and Damasio,A. (1994). Impaired recognition of emotion in facial expressions following bilateral damage to the human amygdala. Nature 372 669-672. Aggleton,J.P., and Saunders,R.C. (2000). The amygdala what's hapenned in the last decade? In The Amygdala, J. P. Aggleton, ed. (Oxfor d: Oxford University Press), pp. 1-30. Aguirre,G.K., Zarahn,E., and D'Esposito,M. (1998). The variabil ity of human, BOLD hemodynamic responses. Neuroimage 8 360-369. Algom,D., Chajut,E., and Lev,S. (2004). A rational look at the em otional Stroop phenomenon: a generic slowdown, not a stroop effect. Journa l of Experimental Psychology: General 133 323338. Anand,A., Li,Y., Wang,Y., Wu,J., Gao,S., Bukhari,L ., Mathews,V.P., Kalnin,A., and Lowe,M.J. (2005a). Activity and connectivity of brain mood regulating circuit in de pression: a functional magnetic resonance study. Biological Psychiatry 57 1079-1088. Anand,A., Li,Y., Wang,Y., Wu,J., Gao,S., Bukhari,L ., Mathews,V.P., Kalnin,A., and Lowe,M.J. (2005b). Antidepressant effect on connectivity of the mood-regulating circuit: an FMRI study. Neuropsychopharmacology 30 1334-1344. Anderson,A.K., Christoff,K., Panitz,D., De Rosa,E ., and Gabrieli,J.D. (2003). Neural correlates of the automatic processing of threat facial signals. Journal of Neuroscience 23 5627-5633. Bannon,M.J., and Roth,R.H. (1983). Pharmacol ogy of mesocortical dopamine neurons. Pharmacol.Rev. 35 53-68. Bechara,A., Damasio,A.R., Damasio,H., and An derson,S.W. (1994). Insensitivity to future consequences following damage to human prefrontal cortex. Cognition 50 7-15. Blamire,A.M., Ogawa,S., Ugurbil,K., Roth man,D., McCarthy,G., Ellermann,J.M., Hyder,F., Rattner,Z., and Shulman,R.G. (1992). Dynamic ma pping of the human visual cortex by highspeed magnetic resonance imaging. Proceedings of the National Academy of Sciences of the United States of America 89 11069-11073.

PAGE 141

141 Boynton,G.M., Engel,S.A., Glover,G.H., and Heeger ,D.J. (1996). Linear systems analysis of functional magnetic resonance imaging in human V1. Journal of Neuroscience 16 4207-4221. Breiter,H.C., Etcoff,N.L., Whalen,P.J., Kennedy,W .A., Rauch,S.L., Buckner,R.L., Strauss,M.M., Hyman,S.E., and Rosen,B.R. (1996). Response a nd habituation of the human amygdala during visual processing of f acial expression. Neuron 17 875-887. Brierley,B., Shaw,P., and David,A.S. (2002). Th e human amygdala: a systematic review and meta-analysis of volumetric magnetic resonanc e imaging. Brain Research.Brain Research Reviews 39 84-105. Broks,P., Young,A.W., Maratos,E.J., Coffey,P.J ., Calder,A.J., Isaac,C.L., Mayes,A.R., Hodges,J.R., Montaldi,D., Cezayirli,E., Robert s,N., and Hadley,D. (1998). Face processing impairments after encephalitis: amygdala dama ge and recognition of fear. Neuropsychologia 36 59-70. Buccino,G., Binkofski,F., and Riggio,L. (2004) The mirror neuron system and action recognition. Brain and Language 89 370-376. Buchel,C., Coull,J.T., and Friston,K.J. (1999). Th e predictive value of changes in effective connectivity for human learning. Science 283 1538-1541. Buchel,C., Holmes,A.P., Rees,G., and Friston,K .J. (1998). Characterizing stimulus-response functions using nonlinear regr essors in parametric fMRI experiments. Neuroimage 8 140-148. Burock,M.A., Buckner,R.L., Woldorff,M.G., Rosen,B.R., and Dale,A.M. (1998). Randomized event-related experimental designs allow for extr emely rapid presentation rates using functional MRI. Neuroreport 9 3735-3739. Bush,G., Luu,P., and Posner,M.I. (2000). Cognitive and emotiona l influences in anterior cingulate cortex. Trends in Cognitive Science 4 215-222. Bush,G., Whalen,P.J., Rosen,B.R., Jenike,M.A., McInerney,S.C., and Rauch,S.L. (1998). The counting Stroop: an interference task specialized for functiona l neuroimaging--validation study with functional MRI. Human Brain Mapping 6 270-282. Calder,A.J., Keane,J., Manes,F., Antoun,N., a nd Young,A.W. (2000). Impaired recognition and experience of disgust following br ain injury. Nature Neuroscience 3 1077-1078. Center for the Study of Emotion and Attent ion [CSEA-NIMH] (2001). The international affective picture system: Digitized photographs. (Gainesville, FL: The Center for Research in Psychophysiology, University of Florida). Chajut,E., Lev,S., and Algom,D. (2005). Vicissitude s of a misnomer: reply to Dalgleish (2005). Journal of Experimental Psychology: General 134 592-595. Chao,L.L., Martin,A., and Haxby,J.V. (1999). Are face-responsive regions selective only for faces? Neuroreport 10 2945-2950.

PAGE 142

142 Chen,N.K., Dickey,C.C., Yoo,S.S., Guttmann,C.R ., and Panych,L.P. (2003). Selection of voxel size and slice orientation for fMRI in the presence of susceptibility field gradients: application to imaging of the amygdala. Neuroimage 19 817-825. Cohen,J.D., Botvinick,M., and Carter,C.S. (2000). Anterior cingulate and prefrontal cortex: who's in control? Nature Neuroscience 3 421-423. Cohen,J.D., Dunbar,K., and McClelland,J.L. (1990). On the control of au tomatic processes: a parallel distributed processing account of the Stroop effect. Psychological Review 97 332-361. Cohen,R.A., Kaplan,R.F., Zuffant e,P., Moser,D.J., Jenkins,M.A., Salloway,S., and Wilkinson,H. (1999). Alteration of intention a nd self-initiated action associat ed with bilate ral anterior cingulotomy. Journal of Neuropsychi atry and Clinical Neurosciences 11 444-453. Critchley,H., Daly,E., Phillips,M., Brammer,M ., Bullmore,E., Williams,S., Van Amelsvoort,T., Robertson,D., David,A., and Murphy,D. (2000). Exp licit and implicit neural mechanisms for processing of social information from facial expressions: a functiona l magnetic resonance imaging study. Human Brain Mapping 9 93-105. Critchley,H.D. (2005). Neural mechanisms of au tonomic, affective, and cognitive integration. Journal of Comparative Neurology 493 154-166. Critchley,H.D., Mathias,C.J., and Dolan,R.J. (200 2). Fear conditioning in humans: the influence of awareness and autonomic arousal on functional neuroanatomy. Neuron 33 653-663. Critchley,H.D., Mathias,C.J., Josephs,O., O'Doherty,J., Zanini,S., Dewar,B.K., Cipolotti,L., Shallice,T., and Dolan,R.J. (2003). Human cingulat e cortex and autonomic control: converging neuroimaging and clinical evidence. Brain 126 2139-2152. Critchley,H.D., Wiens,S., Rots htein,P., Ohman,A., and Dola n,R.J. (2004). Neural systems supporting interoceptive awaren ess. Nature Neuroscience 7 189-195. Culham,J.C., and Kanwisher,N.G. (2001). Ne uroimaging of cognitive functions in human parietal cortex. Current Opinion in Neurobiology 11 157-163. Cunningham,W.A., Raye,C.L., and Johnson,M.K. (2004) Implicit and explic it evaluation: FMRI correlates of valence, emotional intensity, and control in the pro cessing of attitudes. Journal of Cognitive Neuroscience 16 1717-1729. Dale,A.M., and Buckner,R.L. (1997). Selective Aver aging of Rapidly Presen ted Individual Trials Using fMRI. Human Brain Mapping 5 329-340. Dalgleish,T. (2005). Putting some feeling into it --the conceptual and empirical relationships between the classic and emotional Stroop task s: comment on Algom, Chajut, and Lev (2004). Journal of Experimental Psychology: General 134 585-591.

PAGE 143

143 Damasio,A., and Van Hoesen,G.W. (1983). Emotional disturbances associated with focal lesions of the limbic frontal lobe. In Neuropsychology in Human Emotion, K. M. Heilman and P. Satz, eds. (New York: Guilford Press), pp. 85-110. Davidson,R.J., and Irwin,W. (1999). The functiona l neuroanatomy of emotion and affective style. Trends in Cognitive Science 3 11-21. Davidson,R.J., Irwin,W., Anderle,M.J., and Kalin,N.H (2003). The neural substrates of affective processing in depressed patients treated with ve nlafaxine. American Journal of Psychiatry 160 64-75. Dehaene,S., Molko,N., Cohen,L., and Wilson,A.J (2004). Arithmetic and the brain. Current Opinion in Neurobiology 14 218-224. Drevets,W.C. (1998). Functional neuroimaging studies of depression: the anatomy of melancholia. Annual Review of Medicine 49 341-361. Drevets,W.C., and Raichle,M.E. (1998a). Reci procal Suppression of Regional Cerebral Blood Flow during Emotional versus High er Cognitive Processes: Implicat ions for Interactions between Emotion and Cognition. Cognition and Emotion 12 353-385. Drevets,W.C., and Raichle,M.E. (1998b). Recipro cal suppression of region al cerebral blood flow during emotional versus higher cognitive proces ses: implications for interactions between emotion and cognition. Cognition and Emotion 12 353-385. Drevets,W.C., Videen,T.O., Price,J.L., Pres korn,S.H., Carmichael,S.T., and Raichle,M.E. (1992). A functional anatomical study of unipol ar depression. Journa l of Neuroscience 12 36283641. Eisenberger,N.I., and Lieberman,M.D. (2004). Wh y rejection hurts: a common neural alarm system for physical and social pain. Trends in Cognitive Sciences 8 294-300. Ekman,P. (1982). Emotion in the human face. (Cambridge [Cambridgeshire]: Cambridge University Press). Ekman,P., and Friesen,W.V. (1976). Pictures of facial affect. (Palo Alto, Ca.: Consulting Psychologists Press). Ellsworth,P.C. (1994). James,William and emotion is a century of fame worth a century of misunderstanding? Ps ychological Review 101 222-229. Eysenck,W., and Keane,M.T. (2000). Cognitive ps ychology. (Philadelphia: Taylor and Francis). Friston,K.J., Buechel,C., Fink,G.R., Morri s,J., Rolls,E., and Dolan,R.J. (1997). Psychophysiological and modulatory inter actions in neuroimaging. Neuroimage 6 218-229. Friston,K.J., Josephs,O., Rees,G., and Turner,R. (1998). Nonlinear event-related responses in fMRI. Magnetic Resonance in Medicine 39 41-52.

PAGE 144

144 Friston,K.J., Price,C.J., Fletcher,P., Moore, C., Frackowiak,R.S., and Dolan,R.J. (1996). The trouble with cognitive su btraction. Neuroimage 4 97-104. Fu,C.H., Williams,S.C., Cleare,A.J., Bramme r,M.J., Walsh,N.D., Kim,J., Andrew,C.M., Pich,E.M., Williams,P.M., Reed,L.J., Mitterschiffthaler,M.T., Suckling,J., and Bullmore,E.T. (2004). Attenuation of the neural response to sa d faces in major depression by antidepressant treatment: a prospective, eventrelated functional magnetic resona nce imaging study. Archives of General Psychiatry 61 877-889. Glascher,J., and Adolphs,R. (2003). Processing of the arousal of subliminal and supraliminal emotional stimuli by the human amygdala. Journal of Neuroscience 23 10274-10282. Goldapple,K., Segal,Z., and Garson,C. (2002). E ffects of cognitive behavioral therapy on brain glucose metabolism in patients with ma jor depression. Biological Psychiatry 51 66S. Goldapple,K., Segal,Z., Garson,C., Lau,M., Bieling,P., Kennedy,S., and Mayberg,H. (2004). Modulation of cortical-limbic pathways in majo r depression: treatmentspecific effects of cognitive behavior therapy. Arch ives of General Psychiatry 61 34-41. Gorno-Tempini,M.L., Pradelli,S., Serafini,M., Pagnoni,G., Baraldi,P., Po rro,C., Nicoletti,R., Umita,C., and Nichelli,P. (2001). Ex plicit and incidental facial e xpression processing: an fMRI study. Neuroimage 14 465-473. Gottfried,J.A., O'Doherty,J., and Dolan,R.J. (2003) Encoding predictive reward value in human amygdala and orbitofrontal cortex. Science 301 1104-1107. Gur,R.C., Erwin,R.J., Gur,R.E., Zwil,A.S., He imberg,C., and Kraemer,H.C. (1992). Facial emotion discrimination: II. Behavioral findi ngs in depression. Ps ychiatry Research 42 241-251. Gur,R.C., Schroeder,L., Turner,T., McGrath,C., Chan,R.M., Turetsky,B.I., Alsop,D., Maldjian,J., and Gur,R.E. (2002). Brain activation duri ng facial emotion processing. Neuroimage 16 651662. Haidt,J., McCauley,C., and Rozin,P. (1994). Indivi dual-differences in sensitivity to disgust a scale sampling 7 domains of disgust elicitors Personality and Individual Differences 16 701713. Hamzei,F., Rijntjes,M., Dettmers,C., Glauche,V ., Weiller,C., and Buchel,C. (2003). The human action recognition system and its relationship to Broca's area: an fMRI study. Neuroimage 19 637-644. Hariri,A.R., Bookheimer,S.Y., and Mazziotta,J .C. (2000). Modulating emotional responses: effects of a neocortical network on the limbic system. Neuroreport 11 43-48. Hariri,A.R., Mattay,V.S., Tess itore,A., Fera,F., Smith,W.G., a nd Weinberger,D.R. (2002a). Dextroamphetamine modulates the response of the human amygdala. Neuropsychopharmacology 27 1036-1040.

PAGE 145

145 Hariri,A.R., Mattay,V.S., Tess itore,A., Fera,F., and Weinberg er,D.R. (2003). Neocortical modulation of the amygdala response to f earful stimuli. Biological Psychiatry 53 494-501. Hariri,A.R., Mattay,V.S., Tess itore,A., Kolachana,B., Fera,F., Goldman,D., Egan,M.F., and Weinberger,D.R. (2002b). Serotonin transporter genetic variation and the response of the human amygdala. Science 297 400-403. Hariri,A.R., Tessitore,A., Matt ay,V.S., Fera,F., and Weinberg er,D.R. (2002c). The amygdala response to emotional stimuli: a compar ison of faces and scenes. Neuroimage. 17 317-323. Hart,A.J., Whalen,P.J., Shin,L.M., McInerne y,S.C., Fischer,H., and Rauch,S.L. (2000). Differential response in the human amygdala to racial outgroup vs ingroup face stimuli. Neuroreport 11 2351-2355. Harvey,P.O., Fossati,P., Pochon,J.B., Levy,R., Leba stard,G., Lehericy,S., Allilaire,J.F., and Dubois,B. (2005). Cognitive control and brain re sources in major depression: an fMRI study using the n-back task. Neuroimage 26 860-869. He,A.G., Tan,L.H., Tang,Y., James,G.A., Wright ,P., Eckert,M.A., Fox,P.T., and Liu,Y. (2003). Modulation of neural connectiv ity during tongue movement and reading. Human Brain Mapping 18 222-232. Hornak,J., Rolls,E.T., and Wade,D. (1996). Face a nd voice expression identification in patients with emotional and behavioural changes following ventral frontal lobe damage. Neuropsychologia 34 247-261. Huettel,S.A., Song,A.W., and McCarthy,G. ( 2004). Functional magnetic resonance imaging. (Sunderland, MA.: Sinauer). Irwin,W., Anderle,M.J., Abercrombie,H.C., Schaefer,S.M., Kalin,N.H., and Davidson,R.J. (2004). Amygdalar interhemispheric functional c onnectivity differs between the non-depressed and depressed human brain. Neuroimage 21 674-686. Irwin,W., Davidson,R.J., Lowe,M.J., Mock,B.J., Sorenson,J.A., and Turski,P.A. (1996). Human amygdala activation detected with echo-planar functional magnetic resonance imaging. Neuroreport 7 1765-1769. James,W. (1884). What is emotion? Mind 19 188-205. Kanwisher,N., McDermott,J., and Chun,M.M. (1 997). The fusiform face area: a module in human extrastriate cortex specialized fo r face perception. Journal of Neuroscience 17 43024311. Keedwell,P.A., Andrew,C., Williams,S.C., Bram mer,M.J., and Phillips,M.L. (2005). A double dissociation of ventromedial pref rontal cortical responses to sa d and happy stimuli in depressed and healthy individuals. Biological Psychiatry 58 495-503.

PAGE 146

146 Kennedy,S.H., Evans,K.R., Kruger,S., Mayberg,H.S ., Meyer,J.H., McCann,S., Arifuzzman,A.I., Houle,S., and Vaccarino,F.J. (2001). Changes in regional brain glucose metabolism measured with positron emission tomography after paroxeti ne treatment of major depression. American Journal of Psychiatry 158 899-905. Kerns,J.G., Cohen,J.D., MacDonald,A.W., III, C ho,R.Y., Stenger,V.A., and Carter,C.S. (2004). Anterior cingulate conflict monitoring and adjustments in control. Science 303 1023-1026. Krolak-Salmon,P., Henaff,M.A., Isnard,J., Tallon-Baudry,C., Guenot,M., Vighetto,A., Bertrand,O., and Mauguiere,F. (2003). An attent ion modulated response to disgust in human ventral anterior insula. Annals of Neurology 53 446-453. Kwong,K.K., Belliveau,J.W., Chesler,D.A., Gol dberg,I.E., Weisskoff,R.M., Poncelet,B.P., Kennedy,D.N., Hoppel,B.E., Cohen,M.S., Turner,R ., and (1992). Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulati on. Proceedings of the National Academy of Sciences of the United States of America 89 5675-5679. Lane,R.D., Fink,G.R., Chau,P.M., and Dolan,R.J. (1997a). Neural activa tion during selective attention to subjective emo tional responses. Neuroreport 8 3969-3972. Lane,R.D., Reiman,E.M., Bradley,M.M., Lang,P.J., Ahern,G.L., Davidson,R.J., and Schwartz,G.E. (1997b). Neuroanatomical corre lates of pleasant and unpleasant emotion. Neuropsychologia 35 1437-1444. Lang,P.J. (1995). The emotion probe studie s of motivation and attention. American Psychologist 50 372-385. Lang,P.J., Bradley,M.M., and Cuthbert,B.N. (2001) International affective picture system (IAPS): Instruction manual and affective ratings Technical report A-5. (Gainesville, FL: The Center for Research in Psychophysiology, University of Florida). Lang,P.J., Bradley,M.M., Fitzsimmons,J.R., Cu thbert,B.N., Scott,J.D., Moulder,B., and Nangia,V. (1998). Emotional arousal and activati on of the visual cortex: an fMRI analysis. Psychophysiology 35 199-210. Lazarus,R.S. (1982). Thought on the relations between emotion and cognition. American Psychologist 37 1024. LeDoux,J.E. (2000). Emotion circuits in th e brain. Annual Reviews in Neuroscience 23 155184. Leonard,C.M., Rolls,E.T., Wilson,F.A., and Baylis,G.C (1985). Neurons in the amygdala of the monkey with responses selective for faces. Behavioral Brain Research 15 159-176. Leppanen,J.M. (2006). Emotional information pr ocessing in mood disorders: a review of behavioral and neuroimaging findings. Current Opinion in Psychiatry 19 34-39.

PAGE 147

147 Li,H., Albarracin,D., Wright,P., Brown,R.D., and Liu,Y., (2006). Ev aluation proceduralization and its neural correlates. Unpublished manuscrip t, University of Flor ida, Gainesville, FL. Liberzon,I., Taylor,S.F., Fig,L.M., Decker,L.R., Koeppe,R.A., and Minoshima,S. (2000). Limbic activation and psychophysiologi c responses to aversive visual s timuli. Interaction with cognitive task. Neuropsychopharmacology 23 508-516. Lieberman,M.D., Hariri,A., Jarcho,J.M., Eisenbe rger,N.I., and Bookheimer,S.Y. (2005). An fMRI investigation of race-related amygdala activity in African-Ame rican and CaucasianAmerican individuals. Nature Neuroscience 8 720-722. Liotti,M., Mayberg,H.S., McGinnis,S., Brannan,S .L., and Jerabek,P. (2002). Unmasking diseasespecific cerebral blood flow abnormalities: mood challenge in patients with remitted unipolar depression. American Journal of Psychiatry 159 1830-1840. Logothetis,N.K., Pauls,J., Augath,M., Trinath, T., and Oeltermann,A. (2001). Neurophysiological investigation of the basis of the fMRI signal. Nature 412 150-157. MacDonald,A.W., III, Cohen,J.D., Stenger,V.A., and Carter,C.S. (2000). Dissociating the role of the dorsolateral prefrontal and anterior cingu late cortex in cognitive control. Science 288 18351838. Markowitsch,H.J. (1998). Different ial contribution of right and left amygdala to affective information processing. Behavioural Neurology 11 233-244. Matsuo,K., Glahn,D.C., Peluso,M.A., Hatch,J .P., Monkul,E.S., Najt,P., Sanches,M., Zamarripa,F., Li,J., Lancaster,J.L., Fox,P.T., Ga o,J.H., and Soares,J.C. (2006). Prefrontal hyperactivation during working memory task in untreated individuals wi th major depressive disorder. Molecular Psychiatry. Matsuo,K., Kato,C., Sumiyoshi,C., Toma,K., Duy Thuy,D.H., Moriya,T., Fukuyama,H., and Nakai,T. (2003). Discrimination of Exner's area and the frontal eye fiel d in humans--functional magnetic resonance imaging during language and saccade tasks. Neuroscience Letters 340 1316. Mayberg,H.S. (1997). Limbic-cortical dysregulation: a proposed model of depression. Journal of Neuropsychiatry & Clinical Neurosciences 9 471-481. Mayberg,H.S. (2003). Modulating dy sfunctional limbic-cortical circ uits in depression: towards development of brain-based algorithms for dia gnosis and optimised treatment. British Medical Bulletin 65 193-207. Mayberg,H.S., Brannan,S.K., Mahurin,R.K., Jerabek,P .A., Brickman,J.S., Tekell,J.L., Silva,J.A., McGinnis,S., Glass,T.G., Martin,C.C., and Fox,P .T. (1997). Cingulate func tion in depression: a potential predictor of treatment response. Neuroreport 8 1057-1061.

PAGE 148

148 Mayberg,H.S., Brannan,S.K., Tekell,J.L., S ilva,J.A., Mahurin,R.K., McGinnis,S., and Jerabek,P.A. (2000). Regional metabo lic effects of fluoxetine in majo r depression: serial changes and relationship to clinical re sponse. Biological Psychiatry 48 830-843. Mayberg,H.S., Lozano,A.M., Voon,V., McN eely,H.E., Seminowicz,D., Hamani,C., Schwalb,J.M., and Kennedy,S.H. (2005). Deep br ain stimulation for treatment-resistant depression. Neuron 45 651-660. Mayberg,H.S., Silva,J.A., Brannan,S.K., Te kell,J.L., Mahurin,R.K., McGinnis,S., and Jerabek,P.A. (2002). The functional neuroanatomy of the placebo effect. American Journal of Psychiatry 159 728-737. McKiernan,K.A., Kaufman,J.N., Kucera-Thomps on,J., and Binder,J.R. (2003). A parametric manipulation of factors affecting task-induced deactivation in functional neuroimaging. Journal of Cognitive Neuroscience 15 394-408. McNally,R.J. (2002). Disgust has arri ved. Journal of Anxiety Disorders 16 561-566. Menz,M.M., Neumann,J., Muller,K., and Zysset,S. (2006). Variability of the BOLD response over time: an examination of within-session differences. Neuroimage 32 1185-1194. Merboldt,K.D., Fransson,P., Bruhn,H., and Frahm, J. (2001). Functional MRI of the human amygdala? Neuroimage 14 253-257. Mesulam,M.M., and Mufson,E.J. (1982a). Insula of the old world monkey. I. Architectonics in the insulo-orbito-temporal component of th e paralimbic brain. Journal of Comparative Neurology 212 1-22. Mesulam,M.M., and Mufson,E.J. (1982b). Insula of the old world monkey. III: Efferent cortical output and comments on function. Jo urnal of Comparative Neurology 212 38-52. Miezin,F.M., Maccotta,L., Ollinger,J.M., Petersen ,S.E., and Buckner,R.L. (2000). Characterizing the hemodynamic response: effects of presentati on rate, sampling procedur e, and the possibility of ordering brain activity based on relative timing. Neuroimage 11 735-759. Mitterschiffthaler,M.T., Kumari,V., Malhi,G.S ., Brown,R.G., Giampietro,V.P., Brammer,M.J., Suckling,J., Poon,L., Simmons,A., Andrew,C., and Sharma,T. (2003). Neural response to pleasant stimuli in anhedonia: an fMRI study. Neuroreport 14 177-182. Morris,J.S., Frith,C.D., Perrett,D.I., Rowla nd,D., Young,A.W., Calder,A.J., and Dolan,R.J. (1996). A differential neural response in the human amygdala to fearful and happy facial expressions. Nature 383 812-815. Morris,J.S., Ohman,A., and Dolan,R.J. (1998). Cons cious and unconscious emotional learning in the human amygdala. Nature 393 467-470.

PAGE 149

149 Morris,J.S., Ohman,A., and Dolan,R.J. (1999). A s ubcortical pathway to the right amygdala mediating "unseen" fear. Proceedings of the Nati onal Academy of Sciences of the United States of America 96 1680-1685. Mufson,E.J., and Mesulam,M.M. (1982). Insula of the old world monkey. II: Afferent cortical input and comments on the claustrum. Journal of Comparative Neurology 212 23-37. Narumoto,J., Yamada,H., Iidaka,T., Sadato,N., Fukui,K., Itoh,H., and Yonekura,Y. (2000). Brain regions involved in verbal or non-verbal asp ects of facial emotion recognition. Neuroreport 11 2571-2576. O'Doherty,J. (2004). Reward representations and re ward-related learning in the human brain: insights from neuroimaging. Curre nt Opinion in Neurobiology 14 769-776. Ochsner,K.N., and Gross,J.J. (2005). The cognitive control of emotion. Trends in Cognitive Sciences 9 242-249. Ochsner,K.N., Knierim,K., Ludlow,D.H., Ha nelin,J., Ramachandran,T., Glover,G., and Mackey,S.C. (2004). Reflecting upon feelings: an fMRI study of neural systems supporting the attribution of emotion to self and other. Journal of Cognitive Neuroscience 16 1746-1772. Ogawa,S., Lee,T.M., Kay,A.R., and Tank,D.W. ( 1990). Brain magnetic resonance imaging with contrast dependent on blood oxygena tion. Proceedings of the Nationa l Academy of Sciences of the United States of America 87 9868-9872. Ogawa,S., Tank,D.W., Menon,R., Ellermann,J.M., Kim, S.G., Merkle,H., and Ugurbil,K. (1992). Intrinsic signal changes accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging. Proceedings of the National Academy of Sciences of the United States of America 89 5951-5955. Ongur,D., and Price,J.L. (2000). The organization of networks within the orbital and medial prefrontal cortex of rats, monke ys and humans. Cerebral Cortex 10 206-219. Papez,J.W. (1937). A proposed mechanism of emotion. Archives of Neurology & Psychiatry 38 725-743. Passer,M.W., and Smith,R.E. (2001). Psychology: frontiers and applic ations. (New York: McGraw-Hill). Paulus,M.P., Feinstein,J.S., Castillo,G., Simmon s,A.N., and Stein,M.B. (2005). Dose-dependent decrease of activation in bila teral amygdala and insula by lorazepam during emotion processing. Archives of General Psychiatry 62 282-288. Pezawas,L., Meyer-Lindenberg,A., Drabant,E.M., Verchinski,B.A., Munoz,K.E., Kolachana,B.S., Egan,M.F., Mattay,V.S., Hariri ,A.R., and Weinberger,D.R. (2005). 5-HTTLPR polymorphism impacts human cingulate-amygdala interactions: a ge netic susceptibility mechanism for depression. Nature Neuroscience 8 828-834.

PAGE 150

150 Phan,K.L., Taylor,S.F., Welsh,R.C., Ho,S.H., Britton,J.C., and Liberzon,I. (2004). Neural correlates of individual ratings of emotional salience: a tria l-related fMRI study. Neuroimage 21 768-780. Phan,K.L., Wager,T., Taylor,S.F., and Liberzon,I. (2002). Functional neuroanatomy of emotion: a meta-analysis of emotion activation studies in PET and fMRI. Neuroimage. 16 331-348. Phelps,E.A., O'Connor,K.J., Gatenby,J.C., Gore,J.C., Grillon,C., and Davis,M. (2001). Activation of the left amygdala to a cognitive representation of fear. Nature Neuroscience 4 437441. Phillips,M.L., Marks,I.M., Senior,C., Lyt hgoe,D., O'Dwyer,A.M., Meehan,O., Williams,S.C., Brammer,M.J., Bullmore,E.T., and McGuire,P.K. (2000). A differential neural response in obsessive-compulsive disorder patients with wa shing compared with checking symptoms to disgust. Psychological Medicine 30 1037-1050. Phillips,M.L., Young,A.W., Senior,C., Brammer, M., Andrew,C., Calder,A.J., Bullmore,E.T., Perrett,D.I., Rowland,D., Williams,S.C., Gray,J .A., and David,A.S. (1997) A specific neural substrate for perceiving facial e xpressions of disgust. Nature 389 495-498. Piggot,J., Kwon,H., Mobbs,D., Blasey,C., Lots peich,L., Menon,V., Bookheimer,S., and Reiss,A.L. (2004). Emotional attribution in high -functioning individuals w ith autistic spectrum disorder: a functional imaging study. Journal of the American Academy of Child and Adolescent Psychiatry 43 473-480. Pourtois,G., and Vuilleumier,P. (2006). Chapter 4 Dynamics of emotional effects on spatial attention in the human visual co rtex. Progress in Brain Research 156 67-91. Quirk,G.J., and Gehlert,D.R. (2003). Inhibition of the amygdala: key to pathological states? Annals of the New York Academy of Sciences 985 263-272. Ramel,W., Goldin,P.R., Eyler,L.T., Brown,G.G ., Gotlib,I.H., and McQuaid,J.R. (2006). Amygdala reactivity and mood-congruent memory in individuals at risk for depressive relapse. Biological Psychiatry. Rauch,S.L., Whalen,P.J., Shin,L.M., McInerney, S.C., Macklin,M.L., Lasko,N.B., Orr,S.P., and Pitman,R.K. (2000). Exaggerated amygdala response to masked facial stimuli in posttraumatic stress disorder: a functional MRI study. Biological Psychiatry 47 769-776. Roberson-Nay,R., McClure,E.B., Monk,C.S., Nelson,E.E., Guyer,A.E., Fromm,S.J., Charney,D.S., Leibenluft,E., Blair,J., Ernst, M., and Pine,D.S. (2006). Increased amygdala activity during successful memory encoding in ad olescent major depressive disorder: an fMRI study. Biological Psychiatry. Robinson,S., Windischberger,C., Ra uscher,A., and Moser,E. (2004). Optimized 3 T EPI of the amygdalae. Neuroimage 22 203-210. Rolls,E.T. (1999). The brain and emotion. (Oxford: Oxford University Press).

PAGE 151

151 Rolls,E.T. (2000). The orbitofrontal cortex and reward. Cereb.Cortex 10 284-294. Rolls,E.T., Inoue,K., and Browning,A. (2003). Activ ity of primate subgenual cingulate cortex neurons is related to slee p. Journal of Neurophysiology 90 134-142. Rose,E.J., Simonotto,E., and Ebmeier,K.P. (2006) Limbic over-activity in depression during preserved performance on the n-back task. Neuroimage 29 203-215. Royet,J.P., Plailly,J., Delon-Martin,C., Kare ken,D.A., and Segebarth,C. (2003). fMRI of emotional responses to odors: influence of hedonic valence and judgment, handedness, and gender. Neuroimage 20 713-728. Sabatinelli,D., Bradley,M.M., Fitzsimmons,J.R., and Lang,P.J. (2005). Parallel amygdala and inferotemporal activation reflect emotional intensity and fear relevance. Neuroimage 24 12651270. Sander,D., Grafman,J., and Zalla,T. (2003). The human amygdala: an evolved system for relevance detection. Review s in the Neurosciences 14 303-316. Schienle,A., Stark,R., Walter,B., Blecker,C., O tt,U., Kirsch,P., Sammer,G., and Vaitl,D. (2002). The insula is not spec ifically involved in di sgust processing: an fMRI study. Neuroreport 13 2023-2026. Schwartz,C.E., Wright,C.I., Shin,L.M., Kaga n,J., and Rauch,S.L. (2003). Inhibited and uninhibited infants "grown up": adult amygdalar response to novelty. Science 300 1952-1953. Scott,S.K., Young,A.W., Calder,A.J., Hellawe ll,D.J., Aggleton,J.P., and Johnson,M. (1997). Impaired auditory recognition of fear and a nger following bilateral amygdala lesions. Nature 385 254-257. Serences,J.T. (2004). A comparison of methods for characterizing the event-related BOLD timeseries in rapid fMRI. Neuroimage 21 1690-1700. Shapira,N.A., Liu,Y., He,A.G., Bradley,M.M., Lessig,M.C., James,G.A., Stein,D.J., Lang,P.J., and Goodman,W.K. (2003). Br ain activation by disgust-indu cing pictures in obsessivecompulsive disorder. Biological Psychiatry 54 751-756. Sheline,Y.I., Barch,D.M., Donnelly,J.M., Ollinger,J.M., Snyder,A.Z., and Mintun,M.A. (2001). Increased amygdala response to masked emotiona l faces in depressed subjects resolves with antidepressant treatment: an fMRI study. Biological Psychiatry 50 651-658. Shima,K., and Tanji,J. (1998). Role for cingul ate motor area cells in voluntary movement selection based on reward. Science 282 1335-1338. Shin,L.M., Wright,C.I., Cannistraro,P.A., Wedi g,M.M., McMullin,K., Martis,B., Macklin,M.L., Lasko,N.B., Cavanagh,S.R., Krangel,T.S., Orr,S. P., Pitman,R.K., Whalen,P.J., and Rauch,S.L. (2005). A functional magnetic resonance imagi ng study of amygdala and medial prefrontal

PAGE 152

152 cortex responses to overtly pres ented fearful faces in posttrauma tic stress disorder. Archives of General Psychiatry 62 273-281. Shulman,G.L., Fiez,J.A., Corbetta,M., Buc kner,R.L., Miezin,F.M., Raichle,M.E., and Petersen,S.E. (1997). Common blood flow changes acr oss visual tasks: II. Decreases in cerebral cortex. Journal of Cognitive Neuroscience 9 648-663. Shulman,R.G., Hyder,F., and Rothman,D.L. (2001). Cerebral energetics and the glycogen shunt: neurochemical basis of functional imaging. Procee dings of the National A cademy of Sciences of the United States of America 98 6417-6422. Siegle,G.J., Carter,C.S., and Thase,M.E. (2006). Us e of FMRI to predict recovery from unipolar depression with cognitive behavior ther apy. American Journal of Psychiatry 163 735-738. Siegle,G.J., Steinhauer,S.R., Thase,M.E., Stenger, V.A., and Carter,C.S. (2002). Can't shake that feeling: event-related fMRI asse ssment of sustained amygdala ac tivity in response to emotional information in depressed indivi duals. Biological Psychiatry 51 693-707. Singer,T., Seymour,B., O'Doherty,J., Kaube,H., Do lan,R.J., and Frith,C.D. (2004). Empathy for pain involves the affective but not sensory components of pain. Science 303 1157-1162. Soon,C.S., Venkatraman,V., and Chee,M.W. (20 03). Stimulus repetition and hemodynamic response refractoriness in eventrelated fMRI. Human Brain Mapping 20 1-12. Sprengelmeyer,R., Rausch,M., Eysel,U.T., and Prz untek,H. (1998). Neural st ructures associated with recognition of facial expr essions of basic emotions. Proceed ings of the Royal Society of London.Series B: Biological Sciences 265 1927-1931. Sprengelmeyer,R., Young,A.W., Calder,A.J., Karnat ,A., Lange,H., Homberg,V., Perrett,D.I., and Rowland,D. (1996). Loss of disgust Perception of faces and emotions in Huntington's disease. Brain 119 1647-1665. Sprengelmeyer,R., Young,A.W., Schroeder,U., Gro ssenbacher,P.G., Federlein,J., Buttner,T., and Przuntek,H. (1999). Knowing no fear. Proceedings of the Royal Society of London.Series B: Biological Sciences 266 2451-2456. Stark,R., Schienle,A., Walter,B., Kirsch,P., Samm er,G., Ott,U., Blecker,C., and Vaitl,D. (2003). Hemodynamic responses to fear and disgust-in ducing pictures: an fMRI study. International Journal of Psychophysiology 50 225-234. Stefurak,T., Manhurin,R., and Soloman,D. ( 2001). Response specific regional metabolism changes with fluoxetine treatment in depresse d Parkinson's patients. Movement Disorders 16 S39. Surguladze,S., Brammer,M.J., Keedwell,P ., Giampietro,V., Young,A.W., Travis,M.J., Williams,S.C., and Phillips,M.L. (2005). A differe ntial pattern of neural response toward sad versus happy facial expressions in major depressive disorder. Biological Psychiatry 57 201-209.

PAGE 153

153 Talairach,J., and Tournoux,P. (1988). Co-Planar St ereotaxic Atlas of the Human Brain. Thieme Medical Publishers, Inc.). Taylor,S.F., Liberzon,I., Fig,L.M., Decker,L.R ., Minoshima,S., and Koeppe,R.A. (1998). The effect of emotional content on visual recognition memory: a PET activation study. Neuroimage 8 188-197. Taylor,S.F., Phan,K.L., Decker,L.R., and Liberz on,I. (2003). Subjective rating of emotionally salient stimuli modulates ne ural activity. Neuroimage 18 650-659. Tessitore,A., Hariri,A.R., Fera,F., Smith,W.G., Chase,T.N., Hyde,T.M., Weinberger,D.R., and Mattay,V.S. (2002). Dopamine m odulates the response of the human amygdala: a study in Parkinson's disease. Journal of Neuroscience 22 9099-9103. Tessitore,A., Hariri,A.R., Fera,F., Smith,W.G ., Das,S., Weinberger,D.R., and Mattay,V.S. (2005). Functional changes in the activity of brain regions underl ying emotion processing in the elderly. Psychiatry Research 139 9-18. Vogt,B.A. (2005). Pain and emotion interactions in subr egions of the cingul ate gyrus. Nature Reviews.Neuroscience 6 533-544. Vuilleumier,P., Armony,J.L., Driver,J., and Dolan,R .J. (2001). Effects of attention and emotion on face processing in the human brain: an event-related fMRI study. Neuron 30 829-841. Wagner,G., Sinsel,E., Sobanski,T., Kohler,S ., Marinou,V., Mentzel,H.J., Sauer,H., and Schlosser,R.G. (2006). Cortical in efficiency in patients with unipolar depression: an eventrelated FMRI study with the Str oop task. Biological Psychiatry 59 958-965. Wang,A.T., Dapretto,M., Hariri ,A.R., Sigman,M., and Bookheimer,S.Y. (2004). Neural correlates of facial affect processing in childre n and adolescents with autism spectrum disorder. Journal of the American Academy of Child and Adolescent Psychiatry 43 481-490. Whalen,P.J., Bush,G., McNally,R.J., Wilhelm,S., McInerney,S.C., Jenike,M.A., and Rauch,S.L. (1998a). The emotional counting Stroop paradi gm: a functional magnetic resonance imaging probe of the anterior ci ngulate affective divisi on. Biological Psychiatry 44 1219-1228. Whalen,P.J., Rauch,S.L., Etcoff,N.L., McInerne y,S.C., Lee,M.B., and Jenike,M.A. (1998b). Masked presentations of emotional facial expressions modulate am ygdala activity without explicit knowledge. Journal of Neuroscience 18 411-418. Wicker,B., Keysers,C., Plailly,J., Royet,J.P., Ga llese,V., and Rizzolatti,G. (2003). Both of us disgusted in my insula. The common neural ba sis of seeing and feeling disgust. Neuron 40 655664. Wilson,J.L., and Jezzard,P. (2003). Utilization of an intra-oral diamagnetic passive shim in functional MRI of the inferior frontal cortex. Magnetic Resonance in Medicine 50 1089-1094.

PAGE 154

154 Winston,J.S., O'Doherty,J., and Dolan,R.J. (2003). Common and distinct ne ural responses during direct and incidental processing of multiple facial emotions. Neuroimage 20 84-97. Winston,J.S., Strange,B.A., O'Doherty,J., and Dola n,R.J. (2002). Automatic and intentional brain responses during evaluation of trustworthiness of faces. Nature Neuroscience 5 277-283. Wright,C.I., Fischer,H., Whalen,P.J., McIner ney,S.C., Shin,L.M., and Rauch,S.L. (2001). Differential prefrontal cortex and amygdala habituation to re peatedly presented emotional stimuli. Neuroreport 12 379-383. Wright,P., He,G., Shapira,N.A., Goodman,W.K., and Liu,Y. (2004). Disgust and the insula: fMRI responses to pictures of mu tilation and contamination. Neuroreport 15 2347-2351. Wright,P., and Liu,Y. (2005). Neutral faces activ ate the amygdala during identity matching. Neuroimage 29 628-636. Young,A.W., Aggleton,J.P., Hellawell,D.J., Johns on,M., Broks,P., and Hanley,J.R. (1995). Face processing impairments after amygdalotomy. Brain 118 ( Pt 1) 15-24. Zajonce,R.B. (1980). Feeling and thinking: Pr eferences need no inferences. American Psychologist 35 151-175.

PAGE 155

155 BIOGRAPHICAL SKETCH Paul Wright was born in Dover, England, in 1974. He received his secondary education at the Ashcombe School, Dorking, where he obta ined A-levels in chemistry, physics, and mathematics. After spending a year as a volunte er teaching secondary school science in rural Kenya, Paul enrolled at the University of Southampton, England, to study medicine. He completed a study on the rat hippocampus in his f ourth year, but on resuming clinical training decided against life on th e wards. He was awarded a Bachelor of Medical Science in 1997, and worked on studies of lipid metabolism as a techni cian in the Institute for Human Nutrition at his alma mater. He joined the University of Floridas Interdisciplinary Program in Medical Science in 2001, where his interest in the brain reasserted itself. He has since published two first author articles on the functional neuroimaging of em otion. After graduating in December 2006, Paul plans to complete some articles and analysis in the lab of Yijun Liu, and will then pursue postdoctoral employment in England.


Permanent Link: http://ufdc.ufl.edu/UFE0017288/00001

Material Information

Title: Dissecting Emotion: Towards a Functional Neuroimaging Probe for Affective Disorders
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0017288:00001

Permanent Link: http://ufdc.ufl.edu/UFE0017288/00001

Material Information

Title: Dissecting Emotion: Towards a Functional Neuroimaging Probe for Affective Disorders
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0017288:00001


This item has the following downloads:


Table of Contents
    Title Page
        Page 1
        Page 2
    Dedication
        Page 3
    Acknowledgement
        Page 4
    Table of Contents
        Page 5
        Page 6
        Page 7
    List of Tables
        Page 8
    List of Figures
        Page 9
    Abstract
        Page 10
        Page 11
    Introduction
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
    Methodology
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
    Preliminary data: Dissecting the neural correlates of disgust
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
    Face matching and the amygdale: Bottom-up emotion processing or not?
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
    Dissociating event-related responses to top-down and bottom-up emotion processing
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
    Discussion
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
        Page 130
        Page 131
        Page 132
        Page 133
    Appendix: IAPS pictures codes
        Page 134
        Page 135
        Page 136
        Page 137
        Page 138
        Page 139
    References
        Page 140
        Page 141
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
        Page 149
        Page 150
        Page 151
        Page 152
        Page 153
        Page 154
    Biographical sketch
        Page 155
Full Text





DISSECTING EMOTION: TOWARDS A FUNCTIONAL NEUROIMAGING PROBE FOR
AFFECTIVE DISORDERS














By

PAUL WRIGHT


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2006

































Copyright 2006

by

Paul Wright
































Dedicated to my parents, who kept me curious about how things worked and what things meant.









ACKNOWLEDGMENTS

I thank my committee members for their different contributions to my apprenticeship: to

Yijun Liu for modeling vision, to Christiana Leonard for modeling rigor, to Dawn Bowers for

modeling optimism, and to Russell Bauer for modeling practical insight.

Special thanks are due to Andy James for showing me it can be done, to Jason Craggs for

helping me overcome inertia, and to Jessica Couch for keeping me moving. I also thank

Emmanuel Mennonite Church for their moral and spiritual support.









TABLE OF CONTENTS



A C K N O W L E D G M E N T S ..............................................................................................................4

L IST O F T A B L E S ......................................................................................................... ........ .. 8

LIST OF FIGURES ............................................. ............ ...........................9

A B S T R A C T .......................................................................................................... ..................... 10

CHAPTER

1 INTRODUCTION .................................. .. ........... ..................................... 12

S p ecific A im s......................................................................................................... ........ .. 12
C om ponents of Em otion Processing........................................ ....................... ............... 14
B rain R egions Im plicated in D expression ........................................................... ............... 16
A natom ical B background ......................................................................................................... 18
T h e A m y g d ala ............................................................................................................... .. 19
T he Insula ...................................................................................................... . 20
T he V central Prefrontal C ortex ......................................... ........................ ................ 22
The A interior Cingulate C ortex ..................................... ........................ ................ 24
Em otion Processing in H healthy Individuals ...................................................... ................ 26
Processing Emotion in Facial Expressions.................................................... 27
Processing Em otion in Com plex Scenes .................................................... ................ 30

2 METHODOLOGY .................................... ............ .............................34

How fMRI Works ........................ ....... ..................................... 34
B asics of fM R I Paradigm D esign .......................................... ......................... ................ 38
Basics of fMRI Analysis........................ ......... ...............40
Interpreting fM RI results ........................................................... ........................ 43
Examining the BOLD Response to Exclude False Activation...................................44
Using Control Conditions to Test Specific Cognitive Components................................45
Using Factorial and Parametric Designs to Overcome Limits in the Subtractive
A p p ro ach ...................................................................................................... ........ .. 4 6

3 PRELIMINARY DATA: DISSECTING THE NEURAL CORRELATES OF
D IS G U S T ............................................................................................................. ....... .. 4 9

In tro du ctio n ............................................................................................................ ........ .. 4 9
M e th o d s ....................................................................................................... .................... 5 1
S u bjects......................................................................................................... ....... .. 5 1
D isgust P picture P aradigm .. ...................................................................... ................ 51
Functional Im aging D ata A acquisition ........................................................ ................ 53
Functional Im aging D ata A analysis ..................................... ..................... ................ 53









Results ............................................................ .................. 54
Emotion Ratings ................................. .. ........... ............................. 54
fM RI Data ................................................... .................. 54
Discussion ...................................................... .................. 57

4 FACE MATCHING AND THE AMYGDALA: BOTTOM-UP EMOTION
PR O C E SSIN G O R N O T ? ................................................... ............................................. 69

Introduction ................................................. ............................. 69
M e th o d s ........................................................................................................ ..................... 7 1
Subjects ................................. ................ ................ ...................71
Face Matching Task .............................................................71
Functional Im aging D ata A acquisition ..........................................................................72
Functional Im aging D ata A analysis ...............................................................................73
R results ...................................................... ................. ...................74
B e h av io ra l D ata ...............................................................................................................7 4
fM RI Data ................................................... .................. 74
Discussion............................... ... ... ... ............. ................................. 76
Relevance Detection Activates the Amygdala .............................................................77
Emotion Processing at the Amygdala Habituates ........................................................77
Spatial Processing Bypasses the Amygdala During the Control Condition....................78
Cognitive Processing During Emotion M atching.........................................................79
Negative BOLD Responses ............................................80
Complex Contributions to Amygdala Activation.........................................................81

5 DISSOCIATING EVENT-RELATED RESPONSES TO TOP-DOWN AND
BOTTOM -UP EM OTION PROCESSING .........................................................................91

Introduction ...................................................... .................. 91
M e th o d s ........................................................................................................ ..................... 9 5
Subjects ......................... ........................... ......... ..................... 95
Picture R ating Task Paradigm ......................................................................................95
Functional Im aging D ata A acquisition ..........................................................................97
Functional Im aging D ata A analysis ...............................................................................97
R results ...................................................... ................. ...................99
B e h av io ra l D ata ...............................................................................................................9 9
fM RI Data ................................................... .................. 99
D iscussion....................... ......... ....... ............. ...........................100
Top-down Appraisal in the OFC and Insula................................101
Bottom-up Processing in the Amygdala..................................103
Mixed Responses at the Anterior Cingulate Cortex ...........................................104
Response to Frequency Rating in the Parietal Cortex.........................................105
Sum m ary and C conclusions ....................................................................... ...............106

6 D IS C U S S IO N ....................................................................................................................... 1 1 5

S u m m a ry ...................................................................................................... ..................... 1 1 5



6









Proof of Concept: Dissociated Responses to Disgust and Arousal .............................116
A Response at the Amygdala, but not Specific to Emotion .....................................117
An Event-related Emotion Rating Task Partially Replicates Responses to Block-
related T asks .............. .......................................................................................119
Technical Considerations.................................. ............................... 121
Scanning Parameters .................. .................... ........ ...................121
C onnectivity A analysis ......................................................... ............... .... ............ 122
Current Trends in Functional Imaging of Major Depressive Disorder.............................. 123
E m o tio n a l B ia s ..............................................................................................................12 3
E m otion R regulation ............. .. .................. ................ ........................... ............... 126
Cognitive Tasks .........................................................................................................127
Future Direction: the Anterior Cingulate Cortex and the Interaction of Emotion and
Cognition ............................................... .............................. 128

APPENDIX IAPS PICTURES CODES ........................................................134

L IST O F R E F E R E N C E S ....................................................... ................................................ 140

B IO G R A PH IC A L SK E T C H .................................................... ............................................. 155

































7









LIST OF TABLES

Table page

3-1 A effective ratings (dim ensional) ......................................... ........................ ................ 60

3-2 A effective ratings (categorical) .......................................... ......................... ................ 6 1

3-3 Clusters of activation for (threat neutral) ...................................................................62

3-4 Clusters of activation for (contamination neutral)......................................................63

3-5 Clusters of activation for (mutilation neutral) ............... ....................................64

4 -1 B eh av io ral d ata ................................................................................................................. 8 2

4-2 Clusters of activation for [(Emotion Identity) rn (Identity Control)]......................... 83

4-3 Clusters of activation for [(Emotion Control) rn (Identity Control)] ......................... 84

4-4 Regions showing significant modulation of BOLD response.......................................85

5-1 Response time in milliseconds (standard deviation)...... ........................................108

5-2 Clusters of activation for interaction of valence and task..................... ................... 109

5-3 Clusters of activation for main effect of valence. ...... ... ...................................... 110

5-4 Clusters of activation for main effect of task.................................................................111

A -i C ontam nation pictures. ................................................. ............................................ 134

A -2 M utilation pictures. .... ........................................................ ............................... ............. 135

A -3 T h re at p ictu re s .................................................................................................................13 6

A-4 Neutral pictures ............................... .. ........... ..................................... 137

A -5 E m otion rating pictures.................................................. ............................................ 138

A -6 Frequency rating pictures.. .................................................................... ............... 139











8









LIST OF FIGURES

Figure page

3-1 Statistical maps showing contrasts between each emotional condition and neutral..........65

3-2 "G lass brain" view of regions of interest ...................................................... ................ 66

3-3 BOLD responses ............................................. ............................. 67

3-4 Correlations w ith em option ratings.................................... ....................... ................ 68

4-1 M watching task paradigm ................................................... ............................................ 86

4-2 Selective response to emotion at the left inferior prefrontal sulcus...............................87

4-3 Response to face matching at the left and right amygdala............................................88

4-4 R regions of deactivation ................................................... ............................................. 89

4-5 H abituation ................................................................................................ 90

5-1 Main effect of task ......................................... .............................112

5-2 M ain effect of valence. ................................................. ............................................. 113

5-3 Responses in the anterior cingulate cortex........................................... 114


























9









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

DISSECTING EMOTION: TOWARDS A FUNCTIONAL NEUROIMAGING PROBE FOR
AFFECTIVE DISORDERS

By

Paul Wright

December 2006

Chair: Yijun Liu
Major Department: Medical Sciences--Neuroscience

The goal of this research was to develop a functional magnetic resonance imaging

paradigm for use in investigating major depressive disorder. Functional neuroimaging studies of

depression have reported altered resting brain metabolism and altered responses to simple

emotion paradigms in the amygdala, ventral prefrontal cortex, and anterior cingulate cortex. We

studied healthy individuals' responses to complex emotion paradigms to attempt to distinguish

activity in these regions. We hypothesized that the amygdala mediates "bottom-up" processing

driven by emotional stimulus content, and that the ventral prefrontal cortex and anterior cingulate

cortex mediate "top-down" processing driven by explicit knowledge or intention. In the first

experiment, we tested whether matching faces by emotional expression elicited a bottom-up

response in the amygdala. The amygdala responded during matching of both emotional and non-

emotional faces, implying that this response was driven by top-down demands of the task. In the

second experiment, we measured responses during rating of emotional pictures. The response in

the amygdala was greater to unpleasant than pleasant pictures, regardless of rating task: a

bottom-up response. The response in the ventral prefrontal cortex was greater to emotion rating

than non-emotional rating, regardless of picture content: a top-down response. The anterior

cingulate cortex showed weak, mixed bottom-up and top-down responses. This emotion rating









paradigm improves existing approaches to imaging the neural bases of major depressive disorder

by eliciting dissociated responses in two regions implicated in depression: the amygdala and

orbitofrontal cortex. This paradigm may be used in future studies to investigate in parallel the

effects of depression on bottom-up and top-down emotion processing. Future studies may

attempt to elicit more specific responses in the anterior cingulate cortex by using paradigms in

which emotional stimuli interfere with the performance of cognitive tasks.









CHAPTER 1
INTRODUCTION

Functional neuroimaging has been used to identify brain regions involved in major

depressive disorder (MDD); however the individual contributions of these regions to illness are

not known. Specific regional responses may be elicited by using imaging task paradigms that

dissect the components of emotion processing. The goal of the experiments in this dissertation is

to refine the use of functional magnetic resonance imaging (fMRI) as a tool for probing emotion

processing in healthy individuals, in order to improve its usefulness for later investigations of

affective disorders. The feasibility of dissecting emotion with fMRI was demonstrated by

showing that neural responses to emotional pictures in different regions of the brain were

associated with different ratings of picture content (chapter 3). We then attempted to refine two

existing paradigms reported to elicit distinct neural responses to bottom-up (stimulus-driven) and

top-down (knowledge-driven) processing of emotional stimuli. A face-matching task was

reported to elicit bottom-up responses in the amygdala that were inhibited during face labeling.

We added a novel control condition to the face matching task to test whether the amygdala

responded to stimulus content and not to task demands (chapter 4). Rating emotional pictures has

also been shown to recruit cortical regions and to modulate limbic responses. We aimed to

reproduce this modulation using an optimized paradigm design, where task components were

varied at the event level and responses were detected using factorial analysis (chapter 5). The

findings of these studies are discussed in relation to ongoing functional neuroimaging research

on MDD, and future studies are recommended (chapter 6).

Specific Aims

Studies of MDD have highlighted three brain regions where resting metabolism or

responses to simple emotion tasks were altered: the amygdala, ventral prefrontal cortex (PFC),









and anterior cingulate cortex (ACC) (Davidson et al., 2003;Drevets et al., 1992;Mayberg,

2003;Sheline et al., 2001;Siegle et al., 2002). These regions are posited to be involved

respectively in rapid generation (LeDoux, 2000), contextual modification (Rolls, 1999), and

explicit appraisal (Lane et al., 1997a) of emotional responses. The goal of the experiments

presented in this dissertation is to identify an emotion paradigm that can elicit dissociable

responses in limbic and cortical regions to bottom-up and top-down emotion processing, and that

can demonstrate modulation of bottom-up responses under different top-down conditions.

Aim 1. Assess the validity and reliability of the amygdala response to a face matching

paradigm. Verbal labeling of emotional stimuli is hypothesized to inhibit emotional feelings. A

previous study using emotional faces compared verbal labeling with matching faces by their

expression (Hariri et al., 2000). During labeling, responses were decreased at the amygdala and

increased in the ventral prefrontal cortex. The response to face matching was hypothesized to

represent associative (or bottom-up) processing of emotion; however, the task design confounded

emotional content with task performance. In order to test the hypothesis that the amygdala

response reflected bottom-up processing, we compared face matching with a novel control task

in which neutral faces were matched by identity. In order to assess the reliability of the amygdala

response, we investigated changes in the response over repeated presentations of the task.

Aim 2. Dissociate the neural correlates of top-down and bottom-up emotion

processing using a picture rating task. Responses to emotional stimuli may also be inhibited

by tasks requiring that feelings be explicitly attended and appraised. Rating emotional pictures

has been shown to recruit the ACC and modulate the response in the amygdala (Lane et al.,

1997a;Taylor et al., 2003). We attempted to reproduce these results using an optimized paradigm

design. Stimulus content and rating task instructions were randomized at an event level in order









to eliminate expectancy effects. Factorial analysis of the neural responses was used to identify

main effects of bottom-up processing of stimulus content, top-down processing of task demands,

and their interaction.

Components of Emotion Processing

It is not yet known whether specific symptoms of depression may be linked with specific

components of emotion processing. The symptoms of major depressive disorder may be grouped

into attentional or cognitive symptoms, such as apathy, psychomotor retardation, impaired

attention, and executive dysfunction, and vegetative symptoms, such as disordered sleep,

disordered appetite, and endocrine disturbances. It has been proposed that cognitive and

vegetative symptoms are mediated respectively by hypoactivity in dorsal, cortical brain regions

and by hyperactivity in ventral, subcortical brain regions (Mayberg, 1997). These two circuits

may respectively mediate the top-down effects of cognitive behavioral therapy and the bottom-

up effects of antidepressant medication (Goldapple et al., 2004). The chronic negative mood

effects of MDD may be probed by measuring their influence on acute emotion processing. For

example, individuals with a sad mood are more likely to attend to and later to recall sad

emotional stimuli (Eysenck, and Keane, 2000). Chronic alterations in cortical and subcortical

circuits may be probed separately by measuring acute responses to bottom-up and top-down

components of emotion processing.

The production and regulation of emotion is complex. The selection of theories below

provide the basis for the working hypotheses use in the current research. Passer and Smith define

emotions as "positive or negativefeeling (affect) states consisting of a pattern of cognitive,

physiological, and behavioral reactions to events that have relevance to important goals or

motives" (Passer, and Smith, 2001). The physiological reaction to an emotional event was

emphasized by the 19th century psychologist William James, who proposed that, "the bodily









changes follow directly the perception of the exciting fact, and that our feeling of the same

changes as they occur IS the emotion" (James, 1884, original emphasis). These bodily changes

include muscle movement, such as shivering, and visceral responses or autonomic responses,

such as heart palpitations. Despite James' emphasis on the physiological reaction, he assumed

that when faced with an emotional event, perception and appraisal of the event preceded the

generation of a bodily response (Ellsworth, 1994). Furthermore, he proposed that bodily signals

must recombine in the brain with a representation of the "exciting fact" in order for an emotion

to be felt. Later researchers debated the relative influence of immediate perception and deliberate

appraisal upon the behavioral and physiological signs of emotion.

Zajonc proposed the affective primacy hypothesis, according to which events may be

appraised emotionally without conscious awareness (Zaj once, 1980, quoted in Eysenck & Keane,

2000). He supported his hypothesis by investigating the behavioral effects of subliminally

presented emotional stimuli. Individuals viewing subliminal emotional faces followed by

Chinese pictograms rated pictograms preceded by happy faces as more likable. It is not clear,

however, whether the change in liking score was accompanied by any feelings about the Chinese

pictogram. Lazarus emphasized how emotion was influenced by the conscious appraisal of

events (Lazarus, 1982, quoted in Eysenck & Keane, 2000). He showed that physiological

responses to a distressing movie of a surgical procedure altered according to whether the movie's

narration emphasized or downplayed the emotionality of the events. It has been noted, however,

that the movie's narration itself may be considered an emotional stimulus. Although Zajonc and

Lazarus present apparently opposing theories, evidence from studies of fear conditioning in rats

suggests that emotional behavior may be produced by two parallel appraisal systems. LeDoux

identified a fast and a slow route by which sensory stimuli could reach the amygdala and thus









evoke bodily responses. He used this evidence to draw together the theories of Lazarus and

Zajonc, stating:

The activation of the amygdala by inputs from the neocortex [slow route] is ... consistent
with the classic notion that emotional processing is postcognitive, whereas the activation of
the amygdala by thalamic inputs [fast route] is consistent with the hypothesis, advanced by
Zajonc (1980), that emotional processing can be preconscious and precognitive. (Quoted in
Eysenck & Keane, 2000, p.493)

Thus there is evidence for two anatomic routes for emotion processing. Appraisal theorists later

proposed that the components of emotional processing do not follow one another in linear order,

but evolve in parallel (Ellsworth, 1994). Ellsworth states, "Neither interpretation, nor bodily

feedback, nor subjective experience comes first; at the very most, one can talk about which of

these complex temporal processes starts first." If these components of emotion processing are

indistinguishable temporally, they might be distinguishable spatially by identifying distinct

anatomical correlates of each component. In this dissertation, the neural correlates of two levels

of emotion processing are operationally defined:

* Bottom-up processing responses that are determined by the emotional content of
stimuli, independently of task demands.

* Top-down processing responses that correlate with the explicit appraisal (naming or
evaluation) of perceived or experienced emotion.

It is not clear how these two levels of emotion processing are linked to specific symptoms

of MDD. Such a link may be found by identifying brain regions that share a common association

with specific components of emotion processing and with specific symptoms of depression.

Functional imaging studies of MDD therefore provide the anatomic targets for the emotion

paradigms used in this dissertation.

Brain Regions Implicated in Depression

Early studies of resting brain metabolism in patients with MDD measured cerebral glucose

metabolism using positron emission tomography (PET), and reported increased metabolism in









the amygdala and ventral PFC (Drevets, 1998). Amygdala metabolism correlated with symptom

scores and with plasma cortisol (a hormone associated with response to stress) during depressive

episodes. Later fMRI studies extended these findings. The amygdala response was increased in

patients viewing pictures of fearful faces, and decreased following successful medication

(Sheline et al., 2001). Because the emotional stimuli were presented sufficiently rapidly to

prevent their conscious perception, the increased response may reflect altered implicit emotion

processing in MDD. However, MDD may also alter top-down processing in the amygdala. The

amygdala response to unpleasant emotional words lasted around 10 seconds in controls, but in

patients lasted around 25 seconds (Siegle et al., 2002). Because response duration correlated

weakly with self-report measures of rumination, the authors associated these changes with

prolonged elaborative processing of emotional information. Rumination may also underlie

increased metabolism in the PFC, as explained by Mayberg:

Frontal hyperactivity is now viewed as an exaggerated or maladaptive compensatory
process resulting in psychomotor agitation and rumination, serving to over-ride a persistent
negative mood generated by abnormal chronic activity of limbic-subcortical structures. In
contrast, frontal hypometabolism seen with increasing depression severity is the failure to
initiate or maintain such a compensatory state. (Mayberg, 2003, p. 197)

This view is supported indirectly by an fMRI study of patients viewing pleasant emotional

pictures (Mitterschiffthaler et al., 2003). Patients lacked the response seen in controls in the

medial PFC, but had increased responses in the ventrolateral PFC. Because previous studies

showed overlapping responses in the ventrolateral PFC to cognitive and emotional tasks

(Drevets, and Raichle, 1998a), the authors suggested that responses to pleasant pictures in this

region may reflect an attempt to experience positive emotion.

A compensatory process may also involve the ACC. Studies of resting metabolism in

patients with MDD reported different responses in three regions of the ACC, inferior, superior,

and anterior to the genu of the corpus callosum. Metabolism in the subgenual ACC is increased









in MDD and during induced sad mood (Mayberg, 1997). In patients who respond to medication,

metabolism decreases in the subgenual ACC and increases in dorsal regions, including the

supragenual ACC. These changes were replicated in two studies of primary unipolar depression

(Kennedy et al., 2001;Mayberg et al., 2000), one of Parkinson's disease patients with secondary

depression (Stefurak et al., 2001), and one of patients responding to placebo treatment (Mayberg

et al., 2002). Metabolism in the pregenual ACC was unaffected by medication, but pretreatment

metabolism in this region distinguished patients who later responded to medication from those

who did not (Mayberg et al., 1997). This finding was replicated in an fMRI study of patients with

MDD (Davidson et al., 2003). The response to unpleasant pictures at baseline predicted symptom

scores following eight weeks of medication. Furthermore, cognitive-behavioral therapy (CBT)

increased metabolism in the pregenual ACC, suggesting that this region plays an important top-

down role in recovery form depression (Goldapple et al., 2002). The response in the pregenual

ACC to mood challenge is greater in remitted depression patients then in healthy controls or

patients with active depression (Liotti et al., 2002). This altered response in the absence of

treatment supports a protective role for this region.

Anatomical Background

The findings reviewed above suggest that the ventral PFC and pregenual ACC mediate

top-down compensation in MDD that may modulate increased bottom-up processing in the

amygdala. In general, animal studies, lesion studies, and functional imaging studies agree,

describing the roles of the amygdala in forming simple emotional associations, the orbitofrontal

cortex in contextual appraisal of emotion, and the anterior cingulate cortex in emotionally guided

behavior. An additional region, the insula, is associated with awareness of visceral and

autonomic sensations.









The Amygdala

The amygdala appears to be involved in generating rapid but rough responses to stimuli

that have emotional value. Its role in mediating fear behaviors is often emphasized, although it

also responds to rewarding stimuli.

Anatomy. The amygdala is an almond-shaped nucleus in the medial temporal lobe,

anterior to the hippocampus. It shares reciprocal connections with many cortical and subcortical

regions (Aggleton, and Saunders, 2000). Its inputs arrive principally at its lateral and basolateral

nuclei and are passed, directly or indirectly, to the central nucleus for output. Of particular

interest are cortical afferents from insula, temporal and anterior cingulate cortices and from the

dorsolateral, medial and orbital prefrontal cortices. The amygdala sends efferents to each of these

regions, and back-projections to the occipital and temporal cortices to modulate visual

processing. Its subcortical outputs include the tail of caudate, ventral putamen and nucleus

accumbens as well as one-way efferents to the mediodorsal thalamus and reciprocal connections

with the entire hypothalamus. It receives inputs from the midline thalamus and medial pulvinar

as well as various brainstem nuclei.

Animal studies. LeDoux has described in great detail the involvement of the amygdala in

fear conditioning in rats (LeDoux, 2000). In his model, fear is operationally defined as defensive

posturing in response to an aversive stimulus, such as an electric shock (LeDoux, 2000). Rats can

be conditioned to display fear responses to a neutral stimulus (e.g., a tone) by pairing it with the

unconditioned stimulus (shock). These conditioned responses are impaired if the amygdala or its

connections are damaged. Emotional information may reach the amygdala through a fast,

subcortical pathway involving the collicular visual system, or through a slow, cortical pathway,

involving occipital and temporal cortex. Single cell recordings in monkeys have implicated the

amygdala in responses to both aversive and rewarding conditioning (Rolls, 1999). Some









amygdala neurons responded exclusively to rewarding food-conditioned stimuli or to aversive

fear-conditioned stimuli, and others responded to both. Although amygdala neurons rapidly alter

their firing patterns to detect conditioned stimuli, these associations appear to be inflexible. More

flexible, context-dependent associations may be learned by the orbitofrontal cortex (see below).

Some neurons in the monkey amygdala respond selectively to faces, demonstrating a role for the

amygdala in responding to socially salient visual cues (Leonard et al., 1985). Monkeys with

bilateral amygdala lesions are emotionally unresponsive; they show no fear of snakes or humans

and eat items not usually used as food.

Human lesion studies. Humans rarely have lesions confined to the amygdala. Injuries or

damage from encephalitis often affect the hippocampus and temporal lobe as well (Aggleton, and

Saunders, 2000). There is no single, clear symptom of amygdala damage. It is apparent,

however, that such lesions rarely cause cognitive deficits, but generally cause changes in

emotionality. Extensive lesions involving the amygdala are associated with Kluver-Bucy

syndrome, which includes symptoms of blunted emotions, hypersexuality and hyperorality

(placing non-food objects in the mouth). Bilateral amygdala lesions are consistently associated

with impaired recognition of facial expressions of fear (Adolphs et al., 1994;Broks et al.,

1998;Sprengelmeyer et al., 1999;Young et al., 1995) and occasionally associated with impaired

recognition of vocal expressions of fear (Scott et al., 1997).

The Insula

The insula has been described as sensory cortex for the viscera, and may be involved in

processing the bodily responses involved in emotion (Adolphs, 2002). This region is notably

associated with the emotion disgust, but is also involved in pain and awareness of visceral

sensations (Critchley et al., 2002).









Anatomy. The insula is located inside the Sylvian fissure, and is so-called because is

completely concealed (insulated) by the temporal and frontal lobes. It is divided from

anteroventral to posterodorsal into agranular, dysgranuar, and granular regions (Mesulam, and

Mufson, 1982a). The insula receives input from all five sensory modalities, and shares reciprocal

connections with the amygdala, lateral orbital cortex, and ACC (Mufson, and Mesulam, 1982).

Limbic connections, as well as olfactory, gustatory, and autonomic connections, are particularly

extensive in the anteroventral insula (Mesulam, and Mufson, 1982b). Mesulam and Mufson

suggest that the insula's connections to the amygdala allow visceral input to the limbic system,

and furthermore that its common connectivity patterns with the lateral orbital cortex identify

these two regions as part of an integrated paralimbic unit.

Human lesion studies. The insula is associated with impaired recognition of disgust. A

patient with selective damage to the right insula and putamen was impaired at recognizing

disgust in two sets of face stimuli, and in two sets of vocal stimuli (Calder et al., 2000). Despite

being able to recognize disgust conveyed in complex scenes, he scored low on questionnaires

measuring the experience of disgust. Another patient with extensive lesions was impaired at

recognition of all static emotional stimuli, but could recognize emotions acted out or described in

stories, with the exception of disgust (Adolphs et al., 2003). Adolphs et al. suggested that the

processing of acted-out emotions bypassed the damaged limbic and ventral prefrontal regions,

and relied upon parietal and dorsal prefrontal pathways. They also suggested that recognition of

emotion depended upon regions representing somatic states, and that this patient's selective

impairment at recognizing disgust stemmed from his selective lesion of this region of

somatosensory cortex, with sparing of the more dorsal postcentral regions. Thus the insula









appears to be particularly important in the sensory experience of emotion, and in disgust in

particular.

The Ventral Prefrontal Cortex

The ventral PFC is implicated in the flexible association of stimuli with reward and

punishment, social functioning, and regulation of mood. A region of the ventral PFC, the

orbitofrontal cortex, is heavily connected with the amygdala and may be involved in contextual

fine-tuning of primitive emotional signals from the amygdala. Unreferenced data in this section

is taken from The Orbitofrontal Cortex and Reward (Rolls, 2000).

Anatomy. The orbitofrontal cortex is located on the ventral surface of the frontal lobes,

adjacent to the orbits of the eyes. The orbitofrontal cortex receives input from primary gustatory,

olfactory, auditory, and somatosensory cortices. The inputs of primary relevance to the current

research, however, arise from multiple stages along the ventral visual pathways in the temporal

lobe, which are involved in object recognition, and in particular face recognition. The

orbitofrontal cortex also receives strong inputs from the amygdala and from the mediodorsal

nucleus of the thalamus. Its outputs include back-projections to the ventral visual pathways and

outputs to the amygdala, ACC, lateral hypothalamus and ventral striatum.

Animal studies. The orbitofrontal cortex was first associated with reward with the

discovery of neurons with selective responses to taste, a primary reinforcer (that is, certain tastes

can be rewarding or punishing, such as sweet or sour). The taste reward neurons' activity was

enhanced by hunger, and specific to the type of reward, in that a monkey fed to satiety on

bananas would maintain an orbitofrontal response to the sight of peanuts while the neurons tuned

to bananas ceased responding. When a food reward was associated with a visual stimulus,

orbitofrontal neurons responded to that stimulus, provided the animal was hungry. Certain

orbitofrontal neurons responded to the reward value of stimuli, even after their associations were









reversed in a visual discrimination reversal task. In this task, food was paired initially with a

square symbol, and after reversal was paired with a triangle. Orbitofrontal neurons responding

initially to the square rapidly adapted to respond to the triangle after reversal. Some neurons

responded to non-reward when reward was expected, for example after the switching described

above. Other non-reward neurons responded selectively to removal or termination of a reward,

possibly enabling a context-specific response. Certain orbitofrontal neurons in macaques carry

information about faces. They distinguish both expression and identity, probably receiving this

information from neurons in the temporal visual cortex, consistent with the connections

described above.

Human lesion studies. Human with orbitofrontal lesions tend to be euphoric, and have

difficulty in planning and social functioning. The classic prefrontal lesion patient is Phineas

Gage, a 19th century railway foreman who survived the passage of an explosive-driven railroad

spike into his left cheek bone and out of the top of his head, passing through his prefrontal

cortex. Previously hard-working and respected, Gage began to neglect his work duties and his

marriage, engage in drinking and brawling. More recent studies have shown that patients with

prefrontal lesions and impaired social functioning perform badly on certain tasks. When human

subjects were asked to perform a visual discrimination reversal task similar to the one above,

subjects with ventral prefrontal lesions made more errors than controls, apparently because they

were less able to correct their behavior. Test performance correlated with measures of social

impairment. Similar patients were impaired at a gambling task in which two decks of cards were

presented, one that gave large rewards but larger penalties and another that gave small rewards

but smaller penalties (Bechara et al., 1994). Patients with ventral prefrontal lesions were more

likely than controls to pick the high-reward deck even when net gain was greater with the small-









reward deck. They could apparently discern only the short-term positive, not the long-term

negative consequences of their decision. Some patients with orbitofrontal lesions were unable to

recognize emotion in facial expressions and/or speech (Hornak et al., 1996). The latter deficits

were distinct from impaired visual discrimination reversal.

The Anterior Cingulate Cortex

The anterior cingulate cortex is closely related both to medial prefrontal cortex and to

motor cortex. It is associated with numerous functions in addition to emotion processing,

including detection of pain and control of attention (Kerns et al., 2004;Vogt, 2005).

Anatomy. The cingulate gyrus forms a semi-circular belt on the medial surface of the

cortex, surrounding the corpus callosum. The cingulate gyrus was first associated with emotion

when Papez included it in his famous emotion circuit (Papez, 1937). Papez postulated that just as

the striate cortex was considered to be receptive cortex for visual signals from the retina, the

cingulate gyrus may be considered to be receptive cortex for emotional signals from the

hypothalamus. He also saw the cingulate gyrus' extensive cortical outputs as a means by which

emotion could color other experiences, and cortical inputs to the cingulate, as a means by which

emotion could be generated by "psychic processes", as an alternative to visceral inputs. Papez

described the region as "the seat of dynamic vigilance by which environmental experiences are

endowed with an emotional consciousness." The anterior cingulate cortex receives afferents from

the medial orbitofrontal cortex, the amygdala, the temporal pole cortex and somatosensory

cortex, including the insula (Rolls, 1999). It receives rich dopaminergic innervation from the

ventral midbrain (Bannon, and Roth, 1983). Its efferents extend to the periaqueductal gray in the

midbrain, the dorsal motor nucleus of the vagus nerve and the ventral striatum and caudate

nucleus (Rolls, 1999).









Animal studies. Different behaviors may be associated with afferent signals from different

regions of the ACC. The subgenual ACC projects to the medial hypothalamus and ventrolateral

periaqueductal gray, whereas the pregenual ACC projects to the dorsal hypothalamus and lateral

periaqueductal gray (Ongur, and Price, 2000). Ongur and Price (2000) suggest that these

projections allow the subgenual and pregenual regions to evoke coordinated emotional

responses, resulting respectively in quiescent or confrontational stances. Shima and Tanji trained

monkeys to push or to turn a handle, by associating a reward with the preferred action. Certain

neurons in the monkey anterior cingulate cortex responded to decreased reward, but not to

constant reward. When the monkeys failed to adjust their behavior in response to reversal, these

cells also failed to fire (Shima, and Tanji, 1998). This is similar to the orbitofrontal response to

stimulus-reward reversal, but that the association is with action, not a stimulus. Injections of a

GABA agonist into the anterior cingulate prevented the monkeys from altering their task

behavior in response to changing reward. Thus the error-correcting activity of the anterior

cingulate cortex could be interpreted as part of a reward-seeking mechanism.

Human lesion studies. Humans with anterior cingulate strokes appear to lose their

initiative, and despite intact cognition and motor function, become quite inactive and rarely even

talk (Damasio, and Van Hoesen, 1983). Patients treated for chronic pain with bilateral 5mm

lesions of the anterior cingulate cortex report that the pain continues but no longer causes them

distress. They also showed less spontaneous behavior compared with controls, producing shorter

statements at a written task and producing fewer and simpler models when asked to put together

Tinker Toys (Cohen et al., 1999). These findings, along with the animal studies above, suggest

that the anterior cingulate cortex, particularly its pregenual region, may mediate the influence of

processed emotional information on the adjustment and initiation of motor behavior.









Emotion Processing in Healthy Individuals

In patients with depression, functional imaging studies have pinpointed the amygdala,

orbitofrontal cortex, and anterior cingulate cortex as possible sites for deficits. Anatomical and

lesion studies have described the amygdala's role in associating emotional experiences with

perceived stimuli, the orbitofrontal cortex's role in assigning context-dependent emotional value

to stimuli, and the anterior cingulate cortex's role in selecting actions based on their

consequences' anticipated value. These regions have been investigated in healthy humans using

PET and fMRI paradigms involving the perception and evaluation of emotional stimuli.

Although a wide variety of emotional tasks have been used, this dissertation focuses on those

that use visual stimuli to elicit emotion processing. Standardized sets of visual stimuli are

available with rigorous descriptions of their content, allowing creation of precisely-defined

emotional and control stimulus sets, and increasingly the likelihood of reproducible results

across experiments. Using visual stimuli to elicit emotion processing allows precise stimulus

timing, improving the detection of the resulting neural responses. The Pictures of Facial

Emotions consists of faces expressing happiness, sadness, fear, anger, disgust, and surprise

(Ekman, and Friesen, 1976). These expressions are reliably recognized across cultures, implying

that they are not socially learned, but may represent basic categories for the social

communication of emotion (Ekman, 1982). The International Affective Picture System consists

of emotional scenes with a wide variety of content, eliciting a range of emotional responses

(Center for the Study of Emotion and Attention [CSEA-NIMH], 2001). These pictures have been

rated along three dimensions of emotion: valence (from pleasant to unpleasant), arousal (from

calm to excited), and dominance (from controlled to in control). Ratings for pictures at either end

of the valence scale tend also to have high arousal ratings. These rating scales have been

validated using physiological measures of emotional responses (Lang, 1995). Eye blink









responses to auditory startle are increased by unpleasant pictures and decreased by pleasant

pictures. Skin conductance responses to pictures are proportionate to their arousal scores.

Functional imaging studies have used both facial expressions and emotional scenes to investigate

the neural correlates of bottom-up and top-down processing of emotion.

Processing Emotion in Facial Expressions

Facial expressions enable the social communication of emotion. Neural responses to faces

cannot be assumed to correlate with the experience of emotion, but may correlate with automatic

(bottom-up) perception of emotion, or explicit (top-down) recognition of emotion (Davidson,

and Irwin, 1999). Facial expression recognition is impaired in patients with MDD, supporting the

use of face stimuli in imaging investigations of MDD (Gur et al., 1992). Furthermore, recent

studies have shown that brain regions that respond to facial expressions of disgust or pain

overlap with brain regions that respond to the experience of disgust or pain (Singer et al.,

2004;Wicker et al., 2003). This evidence suggests that the neural correlates of emotional

communication and emotional experience may at least partly overlap, further supporting the use

of face stimuli in imaging studies of affective disorders.

Lesions studies have implicated the amygdala in the perception of fear in facial

expressions (Adolphs et al., 1994). This role was confirmed in numerous fMRI studies, which

reported a bottom-up, stimulus-driven response in the amygdala to faces expressing fear and

other salient emotions. The amygdala responded to fearful facial expressions regardless of

whether attention was paid to emotion, to the gender of the faces (Morris et al., 1996;Winston et

al., 2002;Winston et al., 2003), or to the properties of another stimulus (Anderson et al.,

2003;Vuilleumier et al., 2001). The amygdala also responded when emotional faces were

displayed very briefly (-30 ms) and rapidly replaced with a neutral face, a technique called

backwards masking that prevents conscious awareness of the stimulus (Morris et al.,









1998;Whalen et al., 1998b). Other studies showed that the amygdala response may be modulated

by top-down effects, reporting increased responses during explicit recognition of happy and

disgusted faces (Gorno-Tempini et al., 2001), and decreased responses during explicit

recognition of happy and angry faces (Critchley et al., 2000), both compared with gender

recognition. Top-down modulation of the amygdala may be mediated by the prefrontal cortex.

Patients with lesions in the ventral PFC have general social impairments (Rolls, 1999) and

specific impairments in recognizing facial expressions (Hornak et al., 1996). Comparing emotion

recognition with gender recognition elicited specific responses to emotion recognition in

prefrontal regions (Gorno-Tempini et al., 2001;Winston et al., 2003) and elicited amplified

responses to emotional faces in the fusiform gyrus and superior temporal sulcus (Critchley et al.,

2000;Winston et al., 2002). These variable results may reflect multiple strategies for emotion

recognition: some participants may hold in mind verbal labels for the candidate emotions in

order to guide their response, others may use a non-verbal strategy, such as holding in mind a

visual example of each facial expression. Later studies investigated the role of the ventral PFC in

emotion recognition by compared verbal and non-verbal emotion recognition. One study

compared facial expression matching (a perceptual task) with facial expression labeling (an

intellectual task) (Hariri et al., 2000). The amygdala response was lower during labeling then

matching, and correlated inversely with activity in the right ventral PFC, suggesting that verbal

judgments of emotion recruit top-down inhibition by the PFC of bottom-up processing in the

amygdala. A subsequent study compared verbal and facial cues in a delayed match to sample

task involving emotion or gender judgments (Narumoto et al., 2000). This study found no

amygdala response, and reported right prefrontal activation to both verbal and non-verbal

emotion judgment tasks. The difference in amygdala responses between these tasks may be









because Hariri et al. (2000) displayed only fearful or angry faces whereas Narumoto et al. (2000)

showed all six of Ekman's basic facial expressions, counterbalancing the valence of the stimuli.

The prefrontal response to the non-verbal task in Narumoto et al. (2000) may be driven by the

requirement to maintain a representation of the emotional information from the cue in working

memory, whereas in the task employed by Hariri et al. (2000) all stimuli are present

simultaneously.

Because the Hariri task may elicit top-down modulation by the ventral PFC of bottom-up

responses at the amygdala, we chose to investigate this task further. By requiring participants to

read verbal labels in every trial, the labeling condition of the Hariri task is likely to limit

participants' recognition strategy to verbal processing, and thereby elicit top-down processing.

The matching condition of the Hariri task has been shown to activate the amygdala reliably in

studies looking at the influence on the amygdala of genetics (Hariri et al., 2002b;Pezawas et al.,

2005), drugs (Hariri et al., 2002a;Tessitore et al., 2002), and aging (Tessitore et al., 2005).

However, the matching condition is more likely to involve knowledge-based processing than

other control tasks, and may not elicit a true bottom-up response. Hariri et al. (2000) claimed that

participants do not match facial expressions using covert verbal labeling, but using perceptual

cues such as wide eyes or a furrowed brow. Perceptual cue matching may represent intentional,

knowledge-based processing in pursuit of task demands, or top-down processing Therefore, in

order to distinguish facial feature matching from emotion processing, and presumably thereby to

dissect an emotional response at the amygdala, we modified the Hariri task to include an

intermediate control condition in which neutral faces were matched by identity. This experiment

tested the validity of using the face matching task to investigate bottom-up emotion processing in









the amygdala by examining whether the amygdala response is driven by emotional content or

task demands.

Processing Emotion in Complex Scenes

The emotional scenes in the International Affective Picture System (IAPS) differ from the

Ekman faces in several important respects. First, they are complex, and differences in emotion

are conveyed not by small shifts in facial configuration, but by multiple components of the

image. Second, whereas faces represent social signals of emotion (eliciting emotional

perception), scenes are more likely to evoke emotion directly, eliciting emotional experience.

Whereas top-down processing of faces involves recognition of an emotional category, top-down

processing of scenes is more likely to involve the appraisal of internal feelings. Third, the

subjective ratings categorizing the Ekman faces are based on specific categories, but the ratings

of IAPS pictures are based on general dimensional scores.

As with emotional faces, viewing unpleasant scenes elicits a response in the amygdala

(Irwin et al., 1996;Taylor et al., 1998). Studies using IAPS pictures also report greater responses

to emotional than neutral scenes in the ventral temporal cortex (Lane et al., 1997b;Lang et al.,

1998). This response appears to be specific to unpleasant scenes (Lane et al., 1997b) and

correlates with activity in the amygdala (Sabatinelli et al., 2005). Increased ventral temporal

responses to unpleasant pictures are thought to be driven by back-projections to this region from

the amygdala. Interestingly, in a study of patients with MDD, the contrast between responses to

unpleasant and neutral scenes was greater in patients than in controls in the ventral temporal

cortex but not in the amygdala (Davidson et al., 2003). The responses in the amygdala and

ventral temporal cortex correlate with arousal scores, and thus appear to reflect bottom-up

responses driven by stimulus content (Sabatinelli et al., 2005).









The preliminary data presented in chapter three used IAPS pictures to dissociate a response

in the ventral temporal cortex to arousal from a response in the insula to disgust. By using both

dimensional and categorical ratings to describe stimulus content, this study dissected two

components of bottom-up processing. Previous studies reported dissociated responses in the the

amygdala to fearful faces and in the insula to disgusted faces (Phillips et al., 1997;Sprengelmeyer

et al., 1998). A study of patients with obsessive-compulsive disorder supported these findings by

showing that in patients, the insula response was greater to disgust-inducing IAPS pictures than

fear-inducing pictures (Shapira et al., 2003). However, two other studies using IAPS pictures

reported equal activation of the insula to both disgust- and fear-inducing pictures (Schienle et al.,

2002;Stark et al., 2003). The stimuli used to elicit disgust differed between studies, the former

using only pictures of spoiled food, garbage, and other contaminants, and the latter using in

addition pictures of injuries, tumors, and other mutilations. In the study reported in chapter three,

we examined separately the neural responses to pictures of contamination and pictures of

mutilation, comparing both with pictures that elicit fear. Because contamination pictures elicit

low arousal scores and mutilation pictures elicit high arousal scores, this study allowed us to

investigate separately the neural correlates of the emotional category disgust, and the emotional

dimension arousal.

Top-down processing of IAPS pictures has been investigated both using a labeling task, as

described above, and using emotional rating tasks. Hariri et al. (2003) repeated their matching

and labeling study using IAPS pictures. In this study, the matching task involved identifying

identical threatening photographs (for example, a picture of a gun), while the labeling task

involved choosing selecting between the verbal descriptors "natural" and "artificial". The

amygdala response was larger for matching than labeling, while a larger response to labeling was









found at the ventral prefrontal cortex (BA 47) (Hariri et al., 2003). The responses of these two

regions were negatively correlated. While these responses echoed those in Hariri et al. (2000),

the second task did not examine top-down processing of emotion. In the original Hariri task, the

labels were "angry" or "afraid", whereas in the second task, labeling required semantic

processing of non-emotional stimulus content. Replicating the face matching and labeling task

using emotional scenes is hindered by scenes' complexity. While different face pictures within

an emotional category share general features, different scenes eliciting, for example, fear may

vary widely in their visual features, making matching more difficult. Also, while categorical

labels for facial expressions of emotion are widely recognized, categorical labels for emotional

scenes have not been established. Perhaps for these reasons, studies of top-down processing of

emotional scenes usually require participants to rate the scenes simply as pleasant or unpleasant.

In the earliest picture rating study, participants viewed blocks of mixed unpleasant and

neutral pictures during and rated either whether each picture was pleasant or unpleasant or

whether it was indoors or outdoors (Lane et al., 1997a). Emotion rating elicited larger responses

than location rating in the anterior cingulate cortex (ACC) and medial prefrontal cortex (PFC).

This study could not investigate bottom-up responses because unpleasant and neutral pictures

were intermingled. In two subsequent studies, pleasant and neutral pictures were presented

separately to identify regions involved in bottom-up responses (Liberzon et al., 2000;Taylor et

al., 2003). Liberzon et al. (2000) compared emotion rating with picture recognition, a cognitive

task intended to draw attention away from emotion. The right amygdala responded more to

unpleasant than neutral pictures, and this contrast was greater during emotion rating than during

picture recognition. Taylor et al. (2003) compared emotion rating with passive viewing, in order

to test whether top-down processing diminished bottom-up responses. The right amygdala and









insula responded to unpleasant pictures, but the response was smaller during emotion rating than

during passive viewing. The opposite effect was seen in the ACC and medial PFC: responses to

unpleasant pictures were increased during emotion rating compared with passive viewing. These

studies suggest that the amygdala and insula may mediate bottom-up responses to the content of

emotional scenes, and that these responses may be modulated by top-down processing mediated

by the ACC and medial PFC. Because these studies used block designs, which may confound

bottom-up emotion processing with expectation of emotion (a top-down effect), we attempted to

replicate these findings using an optimized event-related paradigm. In our task, both emotional

and non-emotional rating tasks required ratings on continuous scales: either the pleasantness of

the picture, or the frequency of its appearance on television. Stimulus content and tasks

instructions were randomized on a trial-by-trial basis, preventing expectancy of emotional

content, and equalizing the timing of bottom-up and top-down responses. Factorial analysis was

used to separate the main effects of bottom-up and top-down processing and their interaction.









CHAPTER 2
METHODOLOGY

Functional magnetic resonance imaging (fMRI) measures neuronal activity indirectly by

detecting changes in blood oxygenation. By localizing these changes during a stimulus or task,

fMRI may indicate the neural correlates of that task. FMRI is non-invasive, and has

comparatively good spatial resolution. However, the temporal resolution of fMRI is limited by

the sluggishness of the blood oxygenation level-dependent (BOLD) response. Furthermore,

responses must be detected by comparing signal during an experimental task and a control

condition. Therefore, in order to elicit reliable, valid responses, fMRI paradigms must present

stimuli with optimal timing, and must compare task conditions that are matched for every

cognitive factor but the one being studied. This chapter describes the physical basis for fMRI, the

basics of paradigm design, and some advanced techniques for confirming the validity of

statistical maps and for investigating responses to interacting cognitive functions. Unreferenced

material is taken Functional magnetic resonance imaging (Huettel et al., 2004) or from class

notes from BCH 6741, Magnetic Resonance Imaging and Spectroscopy, taught by Dr. Thomas

Mareci.

How fMRI Works

Magnetic resonance imaging is based on a radio signal produced by excited hydrogen

nuclei (spins) in a strong magnetic field. The images produced by MRI are divided into slices,

and each slice is composed of units called voxels. A voxel is comparable to a pixel (picture

element) in a computer image, but because MRI slices have thickness, their cubic constituents

are called 'volume elements'. The intensity of every voxel in a slice is calculated by decoding a

single, complex radio signal that combines the individual radio signals from each voxel. The

individual signals are encoded by varying their frequency and phase according to their spatial









locations. Frequency and phase encoding, and the complex calculations used to reconstruct MR

images, will not be discussed here. We will assume that MR signal intensity at a given voxel is

proportional to the radio signal produced at that voxel. Functional MRI measures signal that is

sensitive to changes in blood oxygenation. This signal decays over time, as described in Equation

1.

(1) S(t) = Soe-t/T2*

Where S = signal, t = time, So = signal at t=0, and T2* = decay constant.

Equation 1 shows that signal decay rate is exponential, and varies according to the T2*

decay constant. T2* describes the combined influence of a number of factors upon signal decay.

One factor, T2, describes interactions between the spins, which varies between different tissue

types but is essentially constant. In an ideal situation T2* = T2, but in reality T2* is shorter than

T2, and real signal decay rates are more rapid than ideal signal decay rates. This is due to the

influence of inhomogeneity effects, or local imperfections in the magnetic field. Inhomogeneity

may be caused by the presence of paramagnetic material, such as deoxyhemoglobin, or interfaces

between air and water. It is the inhomogeneity component that causes T2* to vary over time,

which in turn causes the changes in signal seen in fMRI.

The variations in T2* that form the basis of fMRI are caused inhomogeneity produced by

deoxygenated hemoglobin. Because the hemoglobin molecule contains an iron atom at its core,

its magnetic properties depend on whether the iron is exposed. Oxygenated hemoglobin has a

concealed iron molecule and is diamagnetic, causing no magnetic effects. In deoxygenated

hemoglobin, the iron is exposed, making the molecule paramagnetic and capable of producing

inhomogeneity. The net result is a decrease in MR signal in the region of blood vessels

containing deoxyhemoglobin. This effect was described in detail by systematically varying blood









oxygenation in rats (Ogawa et al., 1990). When perfused with oxygenated blood, vessels are

indistinguishable on MRI from the surrounding brain tissue, but when perfused with

deoxygenated blood, vessels become dark. Ogawa et al. named this effect the blood oxygenation

level-dependent, or BOLD, contrast. BOLD contrast was altered by changes in inhaled carbon

dioxide, blood glucose level, and level of anesthesia, indicating that it is determined by both

cerebral blood flow (supply) and cerebral metabolism (demand). From these data, Ogawa et al.

predicted that BOLD contrast could be detectable within physiological parameters for blood

oxygenation, and that BOLD MRI could be used as a complement to PET for imaging functional

brain activity.

Shortly after the discovery by Ogawa et al., the first fMRI studies appeared. Changes in the

BOLD signal were induced in the occipital lobe using an alternating visual stimulus, and in the

central sulcus using alternating hand movements (Kwong et al., 1992). While these experiments

measured responses to extended periods of stimulation, a later experiment demonstrated that

changes in BOLD signal were detectable to visual stimuli lasting only two seconds (Blamire et

al., 1992). The change in BOLD signal was delayed about 2-3 seconds after the stimulus, a

delayed response known as the BOLD hemodynamic response (HDR). The HDR is thought to

reflect an increase in blood oxygen that occurs, after a delay, in response to neural activity.

The BOLD HDR is now known to have reasonably predictable characteristics. It begins to

rise about 2-3 seconds after the start of the stimulus, and falls about 10-15 seconds after the end

of the stimulus. Thus the temporal characteristics of neural responses are delayed and smoothed

over time in the BOLD HRD. However, the characteristics of the BOLD HDR are sufficiently

predictable to allow detection of neural events that are less than 10-15 seconds apart. The BOLD

responses to two visual stimuli presented two seconds apart appear to add roughly linearly (Dale,









and Buckner, 1997). Subtracting the response to a single stimulus from the response to two

stimuli revealed that the remaining response to the second stimulus was comparable to the

response to the first stimulus. For most fMRI analyses, it is assumed that BOLD responses are

invariant over time, and between brain regions, and that they add roughly linearly. This allows

the use of General Linear Modeling, and event-related analyses, as discussed below.

Several studies have sought to explain in detail the relationship between neural activity and

the BOLD HDR. This question was addressed directly in a study using monkeys in which BOLD

signal and electrical responses were recorded simultaneously (Logothetis et al., 2001). In this

study, a novel MRI-compatible electrical recording system was used to demonstrate that

increases in the BOLD signal were indeed related to increased electrical activity. Specifically,

the main driver of the BOLD HDR was the local field potential. This electrical signal represents

the sum of postsynaptic events at the recording site, or the net input to the neuron. A number of

theories propose to explain the link between neural activity and increased oxyhemoglobin. If

increased cerebral blood flow matched increased neuronal metabolism, then the concentration of

blood oxygen would remain constant, and no BOLD response would be detectable. This

mismatch between supply and demand that produces the BOLD response may reflect an

overcompensation that anticipates future increases in demand, or may be characteristic of

anaerobic metabolism. The latter view was proposed by Shulman et al. (2001) in their astrocyte-

neuron lactate shuttle model. In this model, anaerobic metabolism is required for the rapid

clearance of glutamate from the synaptic cleft following a burst of firing (Shulman et al., 2001).

This model supports the findings of Logothetis et al. (2001) because it links the BOLD HDR to

postsynaptic activity.









In summary, fMRI generates images of the brain in which the intensity of signal is

modulated by changes in blood oxygenation. The BOLD responses to primary visual and motor

stimuli occur in well-established visual and motor regions of the brain (Blamire et al.,

1992;Kwong et al., 1992). However, the location of BOLD responses probably does not reflect

the location of firing neurons, but more likely reflects post-synaptic activity (Logothetis et al.,

2001). Despite the sluggish nature of the BOLD response, individual responses appear to add

linearly, and thus may be modeled even in response to brief stimuli. A number of experimental

designs and analysis techniques have been developed in order to increase certainty that the

detected BOLD response reflects task-related brain activity.

Basics of fMRI Paradigm Design

The primary goal when designing an fMRI paradigm is to evoke a response that will be

distinguishable from noise. Task-related responses are detected as differences in BOLD signal

during task and control conditions, and these differences are typically small, around one percent.

Task-related responses must be distinguished from changes in signal due to non-task-related

factors (noise). Thermal noise or intrinsic noise is the unavoidable, random variation in the

signal due atomic vibrations within system components. Scanner drift is a gradual, monotonic

decrease in signal, which may occur due to slow changes in temperature in the scanner's

magnetic coils. Physiological noise may arise from visceral motility, breathing, or the heart beat.

Motion artifact occurs due to the movement of the head during scanning. Finally, fMRI signal

changes may be due to non-task-related neural activity. Thermal noise is unavoidable, and is

overcome by choosing a stimulus that will evoke as large a response as possible. Monotonic

sources of noise, such as scanner drift, are overcome by repeating the stimulus on/off conditions

several times. Thus, task-related signal changes will be distinguishable by their periodic nature.

Physiological noise, however, is also periodic. To avoid this noise, the frequency of task









presentation must be chosen such that is distinct from the frequencies at which physiological

noise occurs. Practically, this means that periods of stimulation should last between 2 30

seconds. Motion artifact is difficult to avoid, but maybe overcome by immobilizing the subject's

head. Any surviving motion may be corrected during later analysis, and if the motion is too

large, the subject's data are discarded. Since any movement of the jaw increases motion artifact,

fMRI tasks involving speech are difficult to analyze, and where possible subjects should make

task responses by manual button presses. Non-task-related neural activity is the most difficult

variable to control. Analyzing data from a larger group of subjects theoretically maximizes task-

related activity, and minimizes non-task-related activity, which is assumed to be different for

each individual. It is important to choose the control conditions carefully, so that the comparison

used to search from brain activity truly reflects the cognitive functions of interest.

Two types of fMRI paradigms are commonly employed, and both are used in this

dissertation. These are the block design and the event-related design. In the block design, a series

of trials of the same type are presented consecutively, followed by a series of trials of another

type, or a rest period. Blocks of each stimulus type are usually alternated several times using the

so-called boxcar design. This repetition avoids false signal detection due to scanner drift. In the

event-related design, trials are presented in random order, either mixed together or separated by a

period of rest. The advantage of the block design is that it produces a large, easily detectable

signal. The disadvantage of the block design is that the subject is able to anticipate upcoming

trials, since the series of trials are all alike. The event-related design overcomes this expectancy

effect by randomizing trial types. Furthermore, because trials are analyzed individually, event-

related designs allow analysis based on the subject's responses to individual trials, such as error

rate or response time. The disadvantages of the event-related design are that the BOLD responses









to individual trials are small, requiring many repetitions of each trial type to produce a detectable

response, and that the BOLD responses to event-related trials may overlap. One way of

overcoming this overlap is to space events > 15 seconds apart. Alternatively, both these

hemodynamic issues may be overcome by jittering the timing of event presentation. By jittering

the delay between events, the variance in the resulting BOLD HDR becomes larger the more

rapidly events are presented (Burock et al., 1998). This increases the detectability of the BOLD

response, and allows a greater number of repetitions of each event type. Therefore the block

design is appropriate if the neural response to the task should not vary when the trial is expected.

The event-related design is preferred when expectancy effects must be avoided, or when the

study is investigating differences in neural activity with different responses to the task, such as

error trials vs. correct trials.

Basics of fMRI Analysis

What most fMRI analyses have in common is that they create maps of the brain

representing differences between the task and control conditions. These differences are usually

represented as statistical values resulting from a comparison of the BOLD response to each

condition. Plotted as a colored overlay on an anatomic image of the brain, these statistical maps

create the well-known colored clusters of "activation" that illustrate functional imaging studies.

This section explores how these clusters are generated, in order to show that they do not

represent a direct photograph of brain activity, but instead are informed by a series of statistical

decisions made by the experimenter.

Before statistical maps are created, fMRI data are preprocessed. In studies of groups of

subjects, the images from each subject are oriented within a standard space defined by the

midline of the brain, the line between the anterior commissure and the posterior commissure, and

the outermost surfaces of the cortex (Talairach, and Tournoux, 1988). The data from each subject









may be corrected for motion. This correction operates by comparing the whole-brain image at

each time point with the initial time point, and then minimizing differences using iterated small

rigid-body transformations of the later time point. The correction process results in a series of

rotation and translation values that allows head movement to be approximated and thus allows

the exclusion of data with large motion.

Although early fMRI experiments created statistical maps by applying t-tests to the raw

signal during each task condition, the experiments in this dissertation were analyzed by modeling

an estimated BOLD HDR. This approach uses General Linear Model (GLM) statistics. The

GLM may be applied using the fixed or random effects approaches. Both are used in this

dissertation. The fixed effects approach is applicable to small data sets, but is susceptible to one

highly-responsive individual dominating the result. The random-effects approach is more

conservative, detecting consistent changes across the group, but requires larger sample sizes. In

both approaches, the analysis begins by modeling the BOLD HDR to each task condition.

Although this modeling approach may approximate a square wave in slow block designs, and

thus represents only a slight improvement over the t-test, modeling is particularly important in

event-related designs, where there are greater differences between the timing of stimuli and the

timing of the modeled BOLD HDR. Based on stimulus timing, the BOLD HDR may be

estimated using a standard response function (Boynton et al., 1996). The modeled response, or

reference time course, may then be fit to the fMRI signal at each voxel. After estimating a

baseline value, the magnitudes of the reference time courses for each task condition are varied

until the difference between the model and the data is minimized, using the partial least squares

approach. The estimate of the response magnitude is called the beta weight. The solution to the

general linear model for a given voxel is given in Equation 2.









S(t) = Px Mx(t) + Py My(t) + e(t)

Where S = MR signal, t = time, Ptask = beta weight for task x or y, M = modeled hemodynamic

response for task x or y, and e = error, or baseline.

The modeled response for each condition is identical for each voxel, being determined by

the timing of the paradigm. The error term is the same at each voxel, being based upon the signal

during a baseline or resting task condition. Statistical maps are derived from the beta weights.

In the fixed effects approach, a single beta weight (for each task condition at each voxel) is

calculated for the whole group. The statistical significance of the response to each task condition

(or the contrast between selected task conditions) is calculated using the magnitude of the beta

weight (or contrast) and its standard error. The standard error is based upon the differences

between model value and actual signal value, or residuals. The mixed effects approach is

vulnerable to bias resulting from a strong response from a single individual. This bias is

overcome by using the random effects approach, in which individual beta weights are calculated

for each task condition at each voxel, one for each subject. The statistical significance of the

response to each task condition (or the contrast between selected task conditions) is then

calculated by performing a t-test on the sample of beta weights. Although this approach is less

susceptible to contamination from one highly-responsive individual, it requires a larger sample

size. In the fixed effects analysis, the number of degrees of freedom is determined by the number

of time points in the group data. This is the number of subjects multiplied by the number of data

points in one fMRI run, which usually is in the thousands. In the random effects analysis, the

number of degrees of freedom is determined by the number of subjects only. Thus to reach a

given level of significance, a higher statistical score must be obtained in the random effects

analysis compared with fixed effects.









In both approaches, statistical scores are overlaid upon an anatomical image of the brain.

Typically, the statistical values are represented by a range of colors. To make the map more

readable, a threshold is usually applied to eliminate areas where the response was less

significant. The choice of statistical threshold is the topic of ongoing debate in the fMRI

literature. The main issue involves correction for multiple comparisons. Since a statistical test is

applied at every voxel, then for every 100 voxels, 5 will have a significant response at p < 0.05.

In fact, there are over 10,000 voxels, so we would expect at least 500 false positive voxels. The

choice of statistical threshold, and method for correction for multiple comparisons, will be

discussed in each study's chapter.

In summary, fMRI analysis produces brain maps of statistical values. In this dissertation,

these values are calculated by fitting a model of the hemodynamic response to the data at each

voxel, and then performing statistics upon the model. This may be done using a fixed effects

approach or a random effects approach. The former is better suited to small sample sizes, but is

vulnerable to contamination by a single, highly-responsive individual. The latter is a more

conservative approach, requiring larger sample sizes. Because statistical maps represent highly

processed information, and because there are multiple opinions about how best to decide which

clusters of activation are significant, an important skill in fMRI analysis is interpretation.

Interpreting fMRI results

As noted above, there are various sources of noise in fMRI data. Although a statistical map

may reveal clusters of significant signal change, these may or not reflect the neural correlates of

the task condition being tested. This is immediately obvious when "activation" is seen outside

the brain. Significant signal changes outside the brain may occur due to task-related head or eye

movement. The former is particularly prevalent at the borders of light and dark regions, such as

the skull, the edges of the brain, or large sulci. It may be possible to identify motion and other









physical noise effects (like arterial pulsation) by examining the BOLD response at each cluster of

activation. Unlike physical noise effects, however, non-task-related neural activity have not have

a pattern that can distinguish it from task-related neural activity. It can only be eliminated by

careful task design and by having an a priori hypothesis about the brain regions involved in the

task. The approaches given below are each applied in at least one of the experiments in this

dissertation in an attempt to increase the certainty that the apparent fMRI activations indeed

reflect use of the hypothesized cognitive function.

Examining the BOLD Response to Exclude False Activation

As explained above, the clusters of activation illustrated on fMRI statistical maps are

indirect indicators of brain activity. In order to confirm whether the underlying MR signal

changes support the conclusion of the statistical tests, a retrospective examination of the BOLD

HDR at each cluster of statistical activation may be performed. This is achieved by averaging the

responses to each task condition within subjects, and then averaging the responses across

subjects. The MR signal for each response is normalized to a percentage change from baseline

before averaging. In block design paradigms, the response should rise and fall within the time

window for averaging the BOLD HDR. In event-related paradigms, the time window for

averaging the BOLD HDR to a given event may include one or more subsequent events. In order

to minimize the influence of subsequent events, the delay between events is randomly jittered.

Once the group average BOLD HDR has been calculated for each task condition, certain

judgments may be made about the region where the response occurred.

The expected BOLD response to a stimulus is a smooth curve that rises approximately 2-3

seconds after stimulus onset, and then falls approximately 10-15 seconds after the stimulus ends.

In reality, there is a large amount of variability in the BOLD responses between individuals, and

within individuals, both in time (between repeated blocks, or different scanning sessions) and









between brain regions (Aguirre et al., 1998;Menz et al., 2006;Miezin et al., 2000). One

consequence of this is that the BOLD HDR may reveal patterns in the responses to different task

conditions that are missed by the GLM, because of discrepancies between the modeled and

actual response. Post-hoc analysis of the average BOLD response may confirm whether these

patterns are statistically significant.

Several unexpected patterns of BOLD response may be seen. First, the signal may be

inverted. An apparent "activation" on a statistical map may be due to a decrease in BOLD signal

during the control condition, while signal during the task does not change. Deactivation during

task performance may implicate a brain region in cognitive processes that occur during the

control condition. This effect has been reported consistently in a set of brain regions including

the anterior and posterior cingulate cortices, and the bilateral angular gyri (Shulman et al., 1997).

Because these deactivations are deeper with increasingly demanding tasks, they are taken to

represent mental functions performed at rest that require attention (McKiernan et al., 2003). A

second unexpected pattern of BOLD response is a steadily rising curve. This may represent

changes in tonic activation of a brain region, or may represent noise effects, such as task-related

head motion. One indication that a statistical activation is due to a noise effect is a high degree of

variability in the BOLD response, as measured by the standard error of each time point. Further

examination of individual subjects' responses may reveal that a large jump in signal in one

individual led to a statistically significant result.

Using Control Conditions to Test Specific Cognitive Components

Having ruled out physical noise effects, the investigator must ask whether a particular

cluster of activation corresponds to a particular cognitive function. The earliest fMRI

experiments compared simple visual stimuli with a resting condition: a blank screen. In a more

complex task, for example a face recognition task, a more detailed control condition may be









required. For example, to look for specific neural correlate of face recognition, an experiment

may include a condition in which subjects must recognize common household objects. In

addition to determining the response to face recognition and object recognition versus rest, the

two conditions may be compared with each other. This is achieved in the GLM approach by

looking for significant differences between the beta weights for each task condition. This

subtractive method assumes that subtracting object recognition from face recognition eliminates

any brain regions that respond equally to both tasks, and are therefore not specifically activated

by face recognition. Although the subtractive method has weaknesses, as discussed below, this

approach can be used successfully provided the cognitive factors involved in the compared tasks

are considered carefully.

Using Factorial and Parametric Designs to Overcome Limits in the Subtractive Approach

The subtractive method may be used to finely dissect a particular cognitive function, but it

may be inadequate due because it falsely assumes that the neural correlates of cognitive

functions add by pure insertion. In a serial subtractive design, task conditions are designed in

which cognitive components are added one by one. Under the assumption pure insertion,

differences between each task condition will reflect processing of the new cognitive component.

However, if pure insertion is not true, the addition of a new cognitive component may alter the

way the brain handles the existing components. Because there may be interaction between

cognitive components, factorial analysis may be used to account for this interaction (Friston et

al., 1996). This approach was illustrated by Friston et al. (1996) by investigating the neural

correlates of object recognition and phonological retrieval in an object naming task. The authors

tested the hypothesis that the inferior temporal activation during object naming reflected

phonological retrieval. They did so using both a serial subtractive method and using a factorial

method. The serial subtraction experiment involved three conditions: task A, saying "yes" in









response to a colored shape, task B, saying "yes" in response to a colored object, and task C,

naming a colored object. Task A involved visual analysis and speech, task B involved visual

analysis, speech, and object recognition, and task C involved visual analysis, speech, object

recognition, and phonological retrieval. The responses to these task conditions were examined

using PET. The inferior temporal region was activated when subtracting A from B, but not when

subtracting B from C. This implied that this region was involved in object recognition, but not

phonological retrieval. In the factorial design, a task D was added: naming the color of a colored

shape. Task D involved visual analysis, speech, and phonological retrieval, but not object

recognition. The PET responses to these four task conditions were analyzed using a two-way

ANOVA, allowing the detection of a main effect of object recognition (B & C vs. A & D), a

main effect of phonological retrieval (C & D vs. A & B), and an interaction between the two.

The results showed that at the inferior temporal lobe, there was both a main effect of

phonological retrieval and an interaction between phonological retrieval and object recognition.

That is, the response to object recognition varied with the addition of phonological retrieval, and

vice versa. The interaction demonstrated that for this experiment, the assumption of pure

insertion was not true by revealing a response in the inferior temporal lobe to object recognition

that was missed using the subtractive approach. Thus factorial designs can be used to examine

separately the neural correlates of cognitive components that may interact.

Like factorial designs, parametric designs attempt to overcome the problem with cognitive

subtraction. They do so by comparing the MR signal to a task parameter with multiple values.

This allows the identification of brain regions whose activity covaries with the task-related

parameter, rather than regions that are merely more active during the task. For example, one

experiment regions whose responses covaried with word presentation rate (Buchel et al., 1998).









Whereas all words activated bilateral frontal regions irrespective of rate, presentation rate

predicted responses in bilateral occipitotemporal regions. This shows that investigating

covariations of fMRI responses with experimental parameters may finely dissect neural

correlates of one task parameter. In this case, the parametric approach distinguished set-related

activity during word reading from individual stimulus perception. Parametric analysis may also

be used in experiments involving complex stimuli, such as photographs, to separate the effects of

interest from other confounds present in the complex stimulus. This approach was applied in this

dissertation using emotional ratings of photographs as a parameter in analyzing fMRI data.









CHAPTER 3
PRELIMINARY DATA: DISSECTING THE NEURAL CORRELATES OF DISGUST

This experiment investigated the ability of fMRI to distinguish nuances in the negative

emotional content of picture stimuli. Previous studies showed that the amygdala and occipito-

temporal cortex (OTC) respond to pictures rated as emotionally arousing. Responses in the

amygdala to fearful faces have been dissociated from responses in the insula to disgusted faces.

It is not clear whether the amygdala/OTC and insula respond selectively to arousing and

disgusting pictures. In this fMRI study, healthy volunteers viewed pictures of contamination,

human mutilation, threat, and neutral scenes during scanning, and then rated pictures for the

"basic" emotions, including disgust. The anterior insula responded to contamination and

mutilation but not threat, while the OTC responded to threat and mutilations more strongly than

contamination. The above activations were predicted by disgust and arousal ratings respectively.

Additionally, mutilations uniquely activated the right superior parietal cortex. No response was

detected at the amygdala. These results support selective disgust processing at the insula, and

suggest distinct neural responses to contamination and mutilation. The use of fMRI in

investigating components of emotion processing is feasible, but subsequent studies should

improve imaging of the amygdala.

Introduction

The role of the insula in affective processing has been the topic of recent debate. Case

reports from patients with insula lesions and Huntington's disease describe impaired recognition

of facial expressions of disgust, and in some cases, impaired ability to feel the emotion of disgust

itself (Adolphs et al., 2003;Calder et al., 2000;Sprengelmeyer et al., 1996). In healthy volunteers,

functional brain imaging experiments using facial expression recognition tasks have indicated

selective activation of the insula to facial expressions of disgust (Phillips et al.,









1997;Sprengelmeyer et al., 1998). However, results from experiments using pictures such as

body products to induce disgust, question the notion of selective disgust processing at the insula.

Although an insular response to disgust-inducing pictures was reported in a study of obsessive-

compulsive disorder (OCD) patients (Phillips et al., 2000), two subsequent studies of healthy

controls found equal activation of the insula to both disgust- and fear-inducing pictures (Schienle

et al., 2002;Stark et al., 2003). Surprisingly, disgust-inducing pictures activated the amygdala

more than fear-inducing pictures in these studies; the authors explain that their disgust-inducing

pictures elicited high disgust ratings, while their fear-inducing pictures elicited only moderate

fear ratings. A recent study in our lab extended the previous OCD study (Phillips et al., 2000) by

adding fear-inducing pictures (Shapira et al., 2003). In this case, we found greater insula

activation for disgust-inducing pictures than fear-inducing pictures in both healthy volunteers

and patients with OCD. Considering the discrepancy between our results and those of the studies

above (Schienle et al., 2002;Stark et al., 2003), we decided to examine potential differences in

study methodologies. The former studies used pictures of contamination (e.g. spoiled food and

body products) and mutilations (e.g. injuries and corpses) to induce disgust, whereas ours used

only pictures of contamination. We therefore designed a study to test separately the effects of

these two types of pictures.

The insula has been associated with a range of functions, including visceral and gustatory

processing (Wicker et al., 2003), autonomic regulation (Critchley et al., 2003), and self-

generated affective experiences (Phan et al., 2002); thus a general affective or disgust-specific

role for the insula are both plausible. Schienle et al. (2002) suggest that a shared affective

pathway is sufficient to explain the insular response to affective pictures. The current study re-









examines this conclusion by presenting pictures of contamination and mutilation separately to

test whether the insular response is specific to food-related disgust.

We hypothesize first that the insula responds selectively to disgust-inducing pictures, and

second that there is a distinct neural response to pictures of contamination and pictures of

mutilation. The first hypothesis predicts that pictures of contamination or mutilation or both will

cause greater activation at the insula than fear-inducing pictures. The present study therefore uses

functional magnetic resonance imaging (fMRI) to compare these two types of disgust-inducing

pictures with fear-inducing and neutral pictures, in order to assess the validity of combining

pictures of mutilation and contamination in a single category. In addition to performing standard

exploratory analysis using statistical activation maps, we examine the neural response to the

affective pictures in detail using signal time-courses from selected regions of interest (ROIs).

Methods

Subjects

Eight healthy volunteers (4 male) aged 20-26 gave written informed consent in accordance

with a protocol approved by the Institutional Review Board at the University of Florida.

According to self-report, 7 were right-handed, and 1 was left-handed. The volunteers denied

taking any psychiatric medication at the time of the scan and gave no history of psychiatric or

neurological disorders.

Disgust Picture Paradigm

Pictures were selected from the International Affective Picture System (IAPS) (Center for

the Study of Emotion and Attention [CSEA-NIMH], 2001) and were divided into four

categories: "contamination", "mutilation", "threat" and "neutral" which were defined as follows:

contamination pictures depicted of scenes associated with poor hygiene or poisons (e.g. spoiled

food, bodily waste, garbage, pollution); mutilation pictures showed human injuries or disease









(e.g. murder victims, traffic accidents, tumors, birth defects); threat pictures showed imminent

attacks (e.g. humans with guns or knives, dogs, snakes); and neutral pictures depicted various

scenes with low arousal and medium pleasure ratings (Lang et al., 2001) (e.g. landscapes,

household tools, non-threatening animals).

The stimuli were presented using an Integrated Functional Imaging System (IFIS, MRI

Devices, Inc., Waukesha, WI) with a 7" LCD screen at 640 X 480 pixel resolution, mounted over

the subject's head and viewed using a fixed mirror. The screen subtended approximately 140 x

110 of the visual field. A PC running E-Prime (Psychology Software Tools, Pittsburgh, PA)

began presenting each task in synchronization with the first RF pulse of each scan. Each emotion

category was presented during a separate MRI scan in order to avoid the fatigue or boredom that

may result from viewing a single, long sequence. The order of the runs was randomized for each

participant. Each run consisted of six alternating emotional and neutral picture blocks (21 sec

long), interspersed with 9-second fixation blocks. Each block contained 14 pictures selected

randomly (without replacement) from the list for that category, and each picture was presented

for one second, followed by 0.5 seconds of fixation. Participants were instructed to view the

pictures passively, keep their eyes open, and to avoid repressing or exaggerating their emotional

response.

After scanning, each participant rated 15 randomly chosen pictures on a scale from 1 to 5

(5 being the most intense emotion) for each of the following "basic" emotions: happiness,

sadness, fear, anger, disgust and surprise (Ekman, and Friesen, 1976). Dimensional ratings were

taken from the normative set provided with the IAPS. These were ratings from 1 to 9 for

pleasure, arousal, and dominance, with 9 indicating the viewer felt most pleasant, most aroused,









and most dominant respectively. Finally, each participant completed a 32-item questionnaire

designed to indicate their sensitivity to disgust (Haidt et al., 1994).

Functional Imaging Data Acquisition

MR images were acquired using a 3 Tesla Allegra system (Siemens, Munich, Germany).

Anatomical imaging used a standard MPRAGE sequence with a 240mm square field of view at

256 x 256 pixel resolution in the axial plane, and 160 slices of 1.0-1.4mm thickness. Functional

data were gathered using echo-planar imaging (EPI) sensitive to blood oxygen level-dependant

(BOLD) signal (TR = 3000ms, TE = 30ms, flip angle = 900, FOV = 240mm, matrix = 64 x 64).

Twenty-four slices were collected in the axial plane with a 6 mm thickness and 0 mm gap. Each

functional run lasted 3 min 9 sec and consisted of 63 volumes, the first two of which were

discarded before analysis due to their TI saturation.

Functional Imaging Data Analysis

Data were analyzed using BrainVoyager 2000 v. 4.9.6 (Brain Innovations, Maastricht,

Holland). Each anatomic scan was normalized to Talairach space, and the transformation

parameters saved. The in-plane functional images from each participant were then co-registered

with the pre-transformation anatomic scan, and converted into a 3D volume time-course in

Talairach space using the saved transformation parameters. Finally, the 3D functional data

underwent 3D motion correction and linear trend removal.

Voxel-wise statistical activation maps were generated using a general linear model (GLM)

in which the predictors were an estimated hemodynamic response to each emotional condition.

Contrasts between predictors were used to calculate the relative contribution of each predictor to

the variance in the BOLD signal. Unless otherwise stated, the statistical threshold was set to p <









0.05 with Bonferroni correction for multiple comparisons, and the minimum cluster size was 100

3
mm3.

Region of interest (ROI) analyses were performed within selected clusters of significantly

activated voxels. Within each ROI, the BOLD responses for each condition were visualized

using time-locked averaging of the percentage signal change relative to fixation. A GLM was

calculated for the mean signal from the ROI, and the modeled amplitude of each predictor (the

beta weight) was used to describe the size of the hemodynamic response. Unlike the statistical

activation value, which reflects how well the model fits the data, the beta weight describes the

BOLD response, which is assumed to be proportional to neural activation (Ogawa et al., 1992).

Results

Emotion Ratings

All three emotional conditions were rated as evoking significantly less pleasure, more

arousal, and more dominance than neutral according to the mean IAPS scores for each picture set

(Lang et al., 2001) (Tables 3-1 & 3-2). Mutilation was less pleasant than contamination and

threat, and contamination was less arousing than threat and mutilation. According to our own

subjects' ratings of "basic" emotions, the threat condition elicited more fear than the other three

conditions and the contamination and mutilation conditions each elicited more disgust than threat

and neutral. The mutilation condition also elicited more sadness than neutral. (For all the above

comparisons p < 0.001, corrected for multiple comparisons.) The mean standard deviation

disgust sensitivity score was 13.4 4.0 (males: 13.3 3.4, females: 13.6 5.2). The mean for

American adults is approximately 16 (males: 14, females: 18) (Haidt et al., 1994).

fMRI Data

Exploratory statistical activation maps were generated by contrasting each emotional

condition with neutral using the GLM (see Methods: data analysis). Figure 3-1 illustrates clusters









of activation seen in the anterior insula and occipito-temporal cortex (OTC). See Tables 3-3, 3-4,

& 3-5 for a full list of activated regions.

The anterior insula was activated bilaterally in both the contamination and mutilation

conditions. No significant activation was found in the insula for the threat condition at the

threshold p < 0.05 corrected. The extent of activation in the OTC increased in from

contamination to threat to mutilation respectively. Activation for mutilation extended into the

midline occipital cortex and posterior cingulate and was additionally seen in the thalamus,

ventral striatum, superior parietal cortex and several prefrontal regions (Table 3-5). No

significant signal changes were found at the amygdala, but detailed examination of the EPI

(functional) images revealed loss of signal at the amygdala due to susceptibility artifact.

Comparisons between emotions revealed no unique activation for contamination or threat,

but mutilation condition activated the right superior parietal cortex. The contrasts (contamination

- threat) and (mutilation threat) each showed activation in the left anterior insula at a threshold

of p < 0.0001 uncorrected, but this did not achieve the stricter threshold of p < 0.05 corrected.

Clusters of activation for ROI analysis were selected from those contrasts showing significant

differences between emotional conditions; thus the insula ROI was derived from the contrast

[(contamination + mutilation) neutral] and the OTC ROI from the contrast [(mutilation +

threat) neutral]. The left and right ROIs were combined for analysis. The right superior parietal

ROI was selected from the contrast [mutilation (contamination + threat)]. All three ROIs are

illustrated in Figure 3-2.

Time-locked averaging of the BOLD signal across conditions (see Figure 3-3 A-C) showed

a phasic response to all picture conditions (including neutral) in the OTC. This response was

enhanced in the emotional conditions: the enhancement was smallest for contamination, greater









for threat, then greatest for mutilation. At the insula, viewing neutral pictures evoked no change

in signal, but contamination and mutilation again caused a phasic increase. Viewing threat

pictures caused a small response, although this failed to reach the threshold for statistical

significance during the exploratory analysis (Figure 3-1). Signal in the right superior parietal

cortex increased in response to mutilation pictures, but was indistinguishable from neutral during

the other conditions.

The widespread activation for mutilation pictures (see Figure 3-1) may reflect the high

affective arousal ratings for these pictures, particularly the amplitude of the signal increases in

the OTC; also, activity in the anterior insula suggested a relationship with the disgust rating. We

therefore tested the correlations between picture ratings and BOLD signal change, represented by

beta weight (see Methods: Data analysis). Since the experimental design did not include

comprehensive picture ratings for each subject, the ratings were pooled across subjects. Arousal

rating predicted signal change in the OTC [r2 = 0.98, p < 0.05], and disgust rating marginally

predicted signal change in the anterior insula [r2 = 0.85, p = 0.08] (see Figure 3-3 D+E). The

complementary correlations were not significant: disgust rating with OTC signal change [r2 =

0.61, p = 0.22] and arousal rating with insular signal change [r2 = 0.28, p = 0.47]. OTC signal

change was also predicted by ratings for happiness, pleasure and dominance, but these were each

correlated with the arousal rating [respectively, r2 = 0.97, 0.85 & 0.999, p < 0.05, p = 0.08 & p =

0.0005], suggesting that, in this case, these ratings are confounded with a common factor. The

disgust rating did not correlate significantly with any other ratings. Each subject's disgust

sensitivity score was compared with that individual's signal change in the OTC and insula for

each emotional condition, but no significant correlations were found.









Discussion

The aim of this study was to compare the neural responses to two potentially different

types of disgust. Contrary to previous studies comparing disgust- and fear-inducing pictures,

(Schienle et al., 2002;Stark et al., 2003) we found that disgust significantly activated the insula

while fear did not. Furthermore, we showed that the insular response correlated with feelings of

disgust, but not with feelings of arousal. Secondly, we showed distinct neural responses to

viewing pictures of contamination and mutilation. Specifically, viewing pictures of mutilation

caused greater activation of the OTC, and unique activation of the right superior parietal cortex.

The data presented here are insufficient to explain the failure of two previous studies

(Schienle et al., 2002;Stark et al., 2003) to find a specific insular response to disgust in terms of

the effect of combining pictures of contamination and mutilation, since the insula responded to

both conditions. It is possible that our small (and statistically non-significant) insular response to

threat was because our pictures evoked less fear than those of the other two studies. A

comprehensive comparison of picture ratings between studies is not possible here but the fear

ratings for our threat picture set (2.7 out of 5, equivalent to 4.8 out of 9) are close to those of

Schienle et al. and Stark et al. (5.5 and 4.8 out of 9 respectively). Furthermore, if we are to

accept the interpretation that activity in the insula reflects a shared affective system, then our

study should have shown greater activity in the insula to threat than to contamination, since the

threat pictures were rated as more arousing and less pleasant than contamination. One possible

explanation is that 1.5 Tesla MRI (used in the previous studies) is not sufficiently sensitive to

BOLD effects to detect the relatively small differences between the fear and disgust responses at

the insula that are detectable at 3 Tesla (used in the current study).

Previous studies have also suggested that activation of the OTC is influenced by emotional

intensity (Lang et al., 1998;Schienle et al., 2002;Shapira et al., 2003;Stark et al., 2003). These









visual areas do not encode emotion, but receive feedback from emotion-processing regions such

as the amygdala (Rolls, 1999). Thus, although we failed to image the amygdala in this study,

enhancement of ventral visual processing may be thought of as a proxy for amygdala activity

(Sabatinelli et al., 2005). The most compelling evidence we found for a specific response to

disgust in the insula is found in the correlations between disgust rating and insular response, and

arousal rating and occipito-temporal response (Figure 3-3 D+E). These suggest a double

dissociation between the insula, processing information related to disgust, and the OTC,

processing general affective arousal. These findings are compatible with the existence of a

common affective pathway, but suggest that this simple model is insufficient to explain activity

at the insula. Activation of the insula by both mutilation and contamination pictures suggests that

the insular response to disgust is more related to the emotional feeling of disgust rather than the

gustatory content of the eliciting stimulus. An electrical recording study in humans provides

support for a late (300 ms) response to emotional stimuli at the insula, likely reflecting a

conscious feeling rather than earlier processing of gustatory content (Krolak-Salmon et al.,

2003).

We recognize several shortcomings of this study. Since we were unable to image the

amygdala, we had to use occipito-temporal activation as a proxy for the amygdala response.

Although in this study, activity in the insula was not correlated with affective arousal, the insula

influences autonomic arousal (Critchley et al., 2003), and we cannot rule out the insula's

influence on occipito-temporal activity. The affective ratings of our picture sets may be

confounded with other features unique to each set, such as the lack of human faces in the

contamination set, or the abundance of the color red in the mutilation set. Future studies should

use imaging parameters able to image the amygdala, take physiological measures of arousal









(such as heart rate and skin conductance) and specifically account for possible confounds during

selection of picture sets. (It should be noted that not all studies report confirmation of proper

amygdala imaging, and that artifact is common at higher magnetic fields, i.e. 3 Tesla (Merboldt

et al., 2001).)

The unique activation of the right superior parietal cortex by mutilation pictures is an

interesting, new finding that should be further explored by future studies. A previous case study

proposed a parietal pathway for processing acted-out emotions (Adolphs et al., 2003). This

pathway may be more responsive to mutilation pictures if the viewer processes them by mentally

re-enacting the bodily condition of the victim in the picture. This view is further supported by

studies locating mirror neurons for bodily actions in the parietal cortex (Buccino et al., 2004).

Whether mutilation pictures evoke a distinct emotion, for example "horror", is an interesting

question for further study. It has been suggested that horror is a blend of disgust and fear, and it

is interesting to note that mutilation may be viewed for pleasure in art or entertainment

(McNally, 2002).

In conclusion, our findings suggest that the OTC and the insula process different affective

information, as reflected by arousal and disgust ratings respectively. Modulation of occipito-

temporal activity by feelings of arousal is well modeled by the concept of a shared affective

network processing basic affective dimensions. However, the apparently disgust-specific activity

in the insula supports the idea that emotional categories may have distinct neural representations.

We also suggest that future studies consider contamination and mutilation pictures separately.

Whether mutilation pictures evoke a distinct emotion (perhaps "horror") is a question best

answered by future research.









Table 3-1 Affective ratings (dimensional)
Picture set Pleasure Arousal Dominance
Contamination 3.2 0.8 4.9 0.8 4.7 0.6
Mutilation 1.9 + 0.6 6.3 0.7 3.3 + 0.6
Threat 3.0 + 0.8 6.3 0.9 3.3 + 0.7
Neutral 5.6 0.9 3.4 + 1.0 6.0 + 0.6
Pleasure, arousal and dominance ratings are out of nine, and were taken from the IAPS data.









Table 3-2 Affective ratings (categorical)
Picture set Happiness Sadness Fear Anger Disgust Surprise
Contamination 1.2 + 0.7 1.3 + 0.6 1.4 + 0.8 1.1 + 0.3 2.6 + 1.0 1.3 + 0.5
Mutilation 1.0 + 0.0 2.4 + 1.3 1.6 0.9 1.4 0.6 3.2 + 1.3 1.6 0.9
Threat 1.0 + 0.0 1.3 0.5 2.7 1.0 1.6 1.0 1.5 1.0 2.0 1.3
Neutral 1.6+ 0.9 1.0 + 0.0 1.1 0.2 1.0 + 0.0 1.1 0.2 1.2 0.4
Happiness, sadness, fear, anger, disgust, and surprise ratings are out of five, and were obtained
from subjects in the current study.









Table 3-3 Clusters of activation for (threat neutral)
Region Side BA X Y Z Size t(478)
OTC R 37 46 -62 -4 4314 8.7
OTC L 37 -42 -64 -3 2855 8.0
Parahippocampal gyrus L 36 -26 -39 -3 219 -6.6
Only clusters >100 voxels shown. L: left, R: right. BA: Brodmann's Area. X, Y and Z refer to
Talairach co-ordinates (mm). Size: number of lmm3 voxels. t: mixed effects statistical score
(degrees of freedom). Negative t score denotes decrease relative to neutral. OTC: occipito-
temporal cortex.









Table 3-4 Clusters of activation for (contamination neutral)
Region Side BA X Y Z Size t(478)
Insula R 13 31 20 2 571 6.6
Insula / frontal operculum L 13,47 -38 27 0 1650 7.7
Middle frontal gyrus R 46 44 17 23 119 6.2
OTC L 37 -45 -55 -8 1909 7.4
OTC R 37 45 -49 -10 554 6.7
Only clusters >100 voxels shown. L: left, R: right. BA: Brodmann's Area. X, Y and Z refer to
Talairach co-ordinates (mm). Size: number of 1mm3 voxels. t: mixed effects statistical score
(degrees of freedom). OTC: occipito-temporal cortex.









Table 3-5 Clusters of activation for (mutilation neutral)
Region Side BA X Y Z Size t(478)
Cerebellum B 1 -72 -24 136 6.2
Insula L 13 -33 22 1 526 6.5
Insula / frontal operculum R 13, 47 39 22 -1 1614 7.4
Medial frontal gyrus R 8 5 40 36 248 6.6
Middle frontal gyrus R 10,46 33 46 13 134 -6.1
Midline occipital 17-19, 0 -67 9 6818 8.2
29-31
OTC L 37 -38 -61 -10 8419 10.0
OTC R 37 42 -60 -7 14555 11.3
Parahippocampal gyrus R 36 19 -52 0 290 6.1
Parahippocampal gyrus L 36 -19 -52 -3 844 7.4
Precentral gyrus R 6 45 -3 35 481 6.3
Superior frontal gyrus t B 9 -4 57 30 2165 8.7
Superior frontal gyrus L 8 -11 24 57 144 6.6
Cuneus L 19 -8 -89 33 279 6.4
Cuneus & precuneus R 19 25 -69 39 8924 9.6
Cuneus & precuneus L 19 -23 -76 32 1132 6.4
Intra-parietal sulcus L 7 -30 -55 38 151 6.1
Thalamus L -6 -15 13 109 6.0
Ventral striatum R 21 0 -4 108 6.6
Only clusters >100 voxels shown. L: left, R: right, B: bilateral. BA: Brodmann's Area. X, Y and
Z refer to Talairach co-ordinates (mm). Size: number of 1mm3 voxels. t: mixed effects statistical
score (degrees of freedom). Negative t score denotes decrease relative to neutral. OTC: occipito-
temporal cortex. t: this cluster may not reflect neural activity because it is partly outside the
brain and its shape corresponds to the anterior sagittal sinus.
















R L








Contamination Mutilation Threat

Figure 3-1. Statistical maps showing contrasts between each emotional condition and neutral.
Contamination and mutilation activated the anterior insula. The occipito-temporal
cortex (OTC) responded to all three emotional conditions, but comparatively weakly
to contamination. Red / yellow: emotion > neutral, blue: neutral > emotion. Green
line in inset shows slice angle (8 from ACPC). Threshold: p < 0.05, corrected for
multiple comparisons, minimum cluster size: 100 mm3.




































Figure 3-2."Glass brain" view of regions of interest. Green: insula, orange: occipito-temporal
cortex, red: right superior parietal cortex. A: view from front, B: view from left, C:
view from top. The contrasts from which these ROIs were selected are described in
Methods: Functional imaging data analysis. Internal axes denote the anterior and
posterior commissures. L: left, R: right, A: anterior, P: posterior, S: superior, I:
inferior.












1.2 1.2 1.2
S 1.0 1.0 1.0
c 0.8 0.8 0.8
0.6 0.6 0.6
c 0.4 0.4 0.4
S0.2 0.2 0.2
0.0 0.0 0.0
0
-0.2 -0.2 -0.2
-0.4 -0.4 -0.4
-0.6 -0.6 -0.6
-9 0 21 -9 0 21 -9 0 21


Figure 3-3.BOLD responses. A) There is an occipitotemporal response to all conditions, but this
is enhanced in the emotional conditions. B) The insula did not respond to neutral
pictures, but showed the greatest response to contamination pictures. The small
response to threat pictures did not reach significance in the exploratory analysis
(Figure 1). C) The superior parietal ROI responded only to mutilation pictures.
*: contamination, 0: mutilation, V: threat, *: neutral. Solid vertical lines indicate
start of picture block, dotted vertical lines indicate start of fixation block.


A OTC


B Insula


C Parietal













* OTC: r2 = 0.98
V Insula: r2 = 0.28





v V



T


* OTC: r2 = 0.61
V Insula: r2 0.85

0


V
^-- V


3 4 5 6 7 0.5 1.0 1.5 2.0 2.5 3.0 3.5


Arousal


Disgust rating


Figure 3-4. Correlations with emotion ratings. For each picture set, BOLD response amplitude
(Beta weight) is plotted against A) arousal rating & B) disgust rating. Arousal
predicted ventral visual activity, whereas disgust predicted activity in the anterior
insula.@: occipito-temporal cortex, V: anterior insula.









CHAPTER 4
FACE MATCHING AND THE AMYGDALA: BOTTOM-UP EMOTION PROCESSING OR
NOT?

Previous studies have investigated top-down modulation of bottom-up emotion processing

in the amygdala using a face matching and labeling task. The face matching component of this

task has been shown to activate the amygdala reliably, and is thought to elicit bottom-up

processing of the facial features that communicate emotion. However, facial feature matching

may also elicit intentional, knowledge-based processing in pursuit of task demands, or top-down

processing. Therefore, in order to distinguish facial feature matching from emotion processing,

and presumably thereby to dissect an emotional response at the amygdala, we modified the face

matching task to include an intermediate control condition in which neutral faces were matched

by identity. The left and right amygdalae responded to both the emotion and the identity

matching conditions, and the only selective response to emotion matching was at the left inferior

prefrontal sulcus. The left amygdala response habituated to emotion matching but not to identity

matching. Although the amygdala has been described as a "fear module", a growing body of

work suggests that it is a more general "relevance detector". We concluded that the amygdalar

response to face matching is driven at least in part by relevance detection, independent of

emotion processing, although there appears to be additional emotion-specific processing at the

left amygdala. These results suggest that the face matching task is not a valid paradigm to

investigate bottom-up processing of facial emotion in the amygdala.

Introduction

The amygdala has been described as an emotion processing module specialized for fear,

enabling both the experience of fear, and the recognition of fear in others. While this notion is

supported by lesion studies (Adolphs et al., 1994) and functional neuroimaging (Morris et al.,

1996), other data suggest that the amygdala has a more general role, described as relevance









detection (Sander et al., 2003). For example, the amygdala also responds to disgusting scenes

(Schienle et al., 2002), abstract figures associated with food reward (Gottfried et al., 2003),

neutral faces of a different race (Hart et al., 2000), and novel neutral faces (compared with

previously-viewed neutral faces) (Schwartz et al., 2003).

The amygdalar response to facial expressions of emotion is gaining clinical relevance as a

paradigm for studying anxiety and affective disorders. An increased response to fearful or sad

faces has been shown in patients with depression (Surguladze et al., 2005) as well as post-

traumatic stress disorder (Rauch et al., 2000;Shin et al., 2005). Furthermore, in depression, this

hyperresponsivity has been reversed by drug treatment (Fu et al., 2004;Sheline et al., 2001).

Several of the above investigators have suggested a correlation between altered amygdala

activity and previous behavioral data showing impaired facial expression recognition in patients

with depression (Gur et al., 1992). These studies used a variety of tasks, and often did not control

for non-emotional face processing. A more integrated picture of the neurobehavioral alterations

surrounding amygdalar dysfunction could be obtained using a well-established task to dissociate

emotional and non-emotional face processing.

Several groups have reported a robust amygdala response to a task that requires

participants to match faces by emotional expression (Hariri et al., 2000;Hariri et al.,

2002c;Paulus et al., 2005;Piggot et al., 2004;Wang et al., 2004). Matching facial expressions

requires explicit evaluation of emotion, but without the explicit use of verbal labels, which may

inhibit the amygdala response by activating the ventral prefrontal cortex (Hariri et al., 2000).

In order to clarify whether the amygdalar response to matching emotional faces is

specifically due to the perception of emotion, we designed a variant of the matching task from

Hariri et al. (2000). We added an intermediate control condition matching neutral faces by









identity to dissect more precisely the emotional component of the matching task. If the

amygdalar response specifically reflects emotion processing, then it should appear only in the

emotion matching condition. Other regions involved in the perception of faces, such as the

fusiform face area (Kanwisher et al., 1997) and the superior temporal sulcus (Chao et al., 1999),

should respond to both emotion matching and identity matching.

Methods

Subjects

Twelve healthy participants (six female), aged 18 53 (mean 29) years, were recruited as

approved by the University of Florida's Institutional Review Board. All participants were right-

handed and had normal or corrected-to-normal vision. None reported any neurological or

psychiatric history, nor use of psychoactive medications for the previous six months.

Face Matching Task

The matching task consisted of three conditions: emotion, identity, and control. In each

condition, participants were shown a target face above two probe faces, and then had to choose

which probe matched the target (Figure 4-1). In the emotion condition, participants were asked to

match the faces by their expressed emotion (happiness, fear, or anger). In the identity condition,

participants were asked to match neutral faces by identity. In the control condition, participants

were asked to match the pixilated patterns derived from neutral face pictures; thus all three

conditions presented objects with the same dimensions and shades of gray. The task was ordered

in blocks of six 3-second trials of the same condition, preceded by a 3-second instruction screen.

The block condition was varied in a fixed sequence that repeated four times and was

counterbalanced across participants (emotion > identity > control or control > identity >

emotion). The entire run consisted of twelve 21-second tasks blocks interspersed with thirteen 9-

second rest blocks and lasted 3 min 9 sec. During rest, a fixation cross was displayed. A total of









48 grayscale face portraits were presented from the series "Pictures of Facial Affect" (Ekman,

and Friesen, 1976), with six actors of each gender posing happy, fearful, angry, and neutral

expressions. Twelve control patterns were created by shrinking neutral face pictures to 8 x 12

resolution, randomizing the pixels, and enlarging to original size. Within trials, probe and target

faces were the same gender, and an equal number of trials of each gender were presented in each

block. Each actor's face appeared an equal number of times during the experiment. In the

emotion condition, one actor was selected for the probe face, and a second actor for both of the

target faces. The pictures subtended approximately 3.60 x 5.40 (target) and 2.90 x 4.30 (probes)

of the visual field (the target was larger to help distinguish it from the probes). Participants

selected the left or right target by pressing a button under their index or middle finger

respectively, causing the selected target to be outlined in yellow. The participants practiced each

condition inside the scanner before the experimental run until they felt confident performing the

task.

The stimuli were presented using an Integrated Functional Imaging System (IFIS, MRI

Devices, Inc., Waukesha, WI) with a 7" LCD screen at 640 X 480 pixel resolution, mounted over

the subject's head and viewed using a fixed mirror. The screen subtended approximately 14 x

110 of the visual field. A PC running E-Prime (Psychology Software Tools, Pittsburgh, PA)

began presenting each task in synchronization with the first RF pulse of each scan. Responses

were collected with a MRI-compatible button glove attached to the participant's right hand.

Functional Imaging Data Acquisition

Brain images were acquired using a Siemens Allegra 3 Tesla scanner (Siemens, Munich,

Germany) with a standard head coil. Anatomic images were acquired using an MPRAGE

sequence in the sagittal plane at 1.0 mm3 resolution, TR = 1780ms, TE = 4.38ms, flip angle = 8.

Functional images were acquired using a gradient echo planar imaging (EPI) sequence sensitive









to blood oxygen level-dependant (BOLD) contrast in the axial orientation (parallel to the AC-PC

line), covering the whole brain with 36 slices, 3.8mm thick (0mm gap) with a 240mm field of

view and a matrix size of 64 x 64 voxels (in-plane resolution = 3.75mm), TR = 3000ms, TE =

30ms, flip angle = 900. A total of 125 brain volumes were acquired (3min 15sec scan time) and

the first two volumes were discarded before analysis to allow for TI equilibration.



Functional Imaging Data Analysis

MR data were analyzed using BrainVoyager 2000 (v. 4.9.6, Brain Innovations, Maastricht,

Holland). The functional images were coregistered with anatomic images, and normalized to

Talairach space for each participant. Functional data underwent 3D motion correction, linear

trend removal and slice scan time correction (the slice data in each volume were time-shifted to

the start of the TR by interpolation). High- and low- frequency noise was removed using low-

and high- pass filters with cut-off frequencies of 10/123 Hz and 1/123 Hz respectively. Spatial

smoothing was applied using a Gaussian filter of 5.7mm full-width half maximum.

Regions of task-related brain activity were estimated using general linear modeling. A

reference function, or predictor, was created for each condition by convolving the block

presentation time course with an estimated hemodynamic response function (Boynton et al.,

1996). The signal at each voxel was modeled with a weighted combination of the three predictors

using least squares fitting. Statistical maps were created using random effects analysis. This

conservative approach looks for consistent differences between predictors' weighting across

participants, preventing data from one or two participants from dominating the analysis. Clusters

of voxels with significant differences between predictors were selected by setting a statistical

threshold oft(1 1) > 4.0 (p < 0.002 uncorrected) and a minimum cluster size of 100 mm3 (except

at the amygdala our a priori region of interest where smaller clusters were allowed).









For each cluster of significant voxels, we validated the GLM results by plotting the mean

BOLD response to each condition, as previously described (Wright et al., 2004). The percentage

signal change was calculated for each block relative to the preceding resting signal, and then

averaged across blocks and participants for each condition. In order to investigate habituation,

the BOLD response to each condition was calculated separately for each of the four repetitions

of that block, averaged across participants. The mean peak BOLD response was calculated from

the period of peak activity at 9 18 seconds after block onset. The magnitude and significance of

any modulation of mean amplitude over time were calculated using linear regression.

To confirm that the MRI parameters used were able to detect signal at the amygdala, we

visually inspected the functional images using an outline of the amygdala drawn from the

average anatomic image according to the guidelines of Brierley et al. (Brierley et al., 2002). In

two out of the twelve participants, susceptibility artifact from the nasal sinuses obscured the

amygdala. Because amygdala activation could be detected with or without these participants,

they were included in the analysis to maximize statistical power in the rest of the brain.

Results

Behavioral Data

Participants' responses were significantly faster and more accurate in the identity

condition, compared with emotion and control (Table 4-1). On debriefing most participants

reported that the identity condition was the easiest of the three.

fMRI Data

Significant differences in modeled signal amplitude ("activations") are summarized in

Tables 4-2 & 4-3. Emotion-specific activations found using the conjunction of the contrasts

(emotion identity) and (emotion control) occurred at the left inferior frontal sulcus

(Figure 4-2) and right precentral gyms (not shown). The BOLD response at the inferior frontal









sulcus appeared to be specific to the emotion condition (Figure 4-2A), while at the precentral

gyrus, the magnitude of the response to identity appeared to be intermediate between that of

control and emotion. Emotion-specific deactivations were found at the left transverse temporal

sulcus and the pregenual anterior cingulate gyrus (Figure 4-4B). Specifically, the BOLD

response at the temporal region was negative during emotion matching but remained near

baseline levels during the identity and control conditions. The pregenual cingulate BOLD

response decreased from baseline during all three conditions, with the largest decrease during

emotion matching (Figure 4-4A). The emotion condition did not selectively activate the

amygdala.

Activations at the left and right amygdalae were found when contrasting either face

matching condition (emotion or identity) with the control condition. The activation was more

statistically significant at the right amygdala than the left (peak t(1 1) = 7.25 vs. 5.23 for emotion,

8.63 vs. 5.11 for identity), but the mean peak BOLD response was larger at the left amygdala

than the right (0.5% vs. 0.2% for emotion, 0.7% vs. 0.3% for identity). The BOLD response

appeared larger in magnitude and duration to identity than to emotion in both hemispheres.

Figure 4-3 and Table 4-3 describe amygdala activation obtained using the conjunction of the

contrasts (emotion control) and (identity control). The amygdalar clusters for the individual

contrasts (emotion control) and (identity control) were overlapping, and are therefore not

depicted separately.

Activation for both face matching conditions was also seen in the right fusiform gyrus,

occipitotemporal cortex, and anterior and posterior cingulate cortices (Figure 4-4B shows

cingulate activations). While the fusiform and occipitotemporal activations reflected positive

BOLD responses that were larger during face matching than during the control condition, the









anterior and posterior cingulate BOLD responses were negative during the control condition and

remained near baseline levels during the emotion and identity conditions.

The control condition activated extensive, bilateral regions of the occipital and parietal

cortex and smaller clusters within the right anterior insula, left collateral sulcus, left fusiform

gyrus, and left superior frontal sulcus. These activations reflected positive BOLD responses that

were larger during pattern matching than during identity and emotion matching.

Habituation was also investigated by looking for changes over time in the behavioral and

BOLD responses. Neither accuracy nor response time showed significant changes over time,

indicating that fatigue did not occur. The response to emotion at the left amygdala decreased

markedly over repeated blocks and showed a significant, negative correlation with time

(Table 4-4, Figure 4-5). Other regions showing significant modulation of peak amplitude of the

BOLD response with time are described in Table 4-4, with the slope of the linear regression

representing BOLD modulation. Note that in some regions the response increased over time, and

that compared with the change in response at the amygdala, the next largest modulation was only

about half the size.

Discussion

Contrary to the "emotion processor" hypothesis, this study found equal amygdala

activation for emotional and neutral face matching. Hariri et al. (2000) proposed that the

amygdala encoded emotion at an associative level during matching of emotional faces, but the

current findings require further explanation. Sander et al. (2003) describe the amygdala as a

system for relevance detection stating, "An event is relevant for an organism if it can

significantly influence (positively or negatively) the attainment of his or her goals." This

definition may be a more useful starting point for interpretation of the present study.









Relevance Detection Activates the Amygdala

While Hariri and colleagues acknowledge that the blocked design of their task limits its

ability to investigate processing of specific emotions (Hariri et al., 2002c), they have

demonstrated that matching facial expressions robustly activates the amygdala, and have

successfully employed the task to probe the effects of genetics (Hariri et al., 2002b;Pezawas et

al., 2005), drugs (Hariri et al., 2002a;Tessitore et al., 2002), and aging (Tessitore et al., 2005) on

the amygdala and an associated affective network.

Previous studies have shown amygdala activation to neutral faces based on gaze direction,

novelty, and race (Sander et al., 2003). To our knowledge the only face matching study with

neutral faces investigated the effect of matching neutral faces by race (Lieberman et al., 2005). In

the current study, however, matching neutral faces activated the amygdalae as much as matching

emotional faces without the addition of relevance from gaze, race, etc. It seems that the matching

task itself adds relevance to neutral face stimuli. Viewed alone, neutral faces would be expected

to have less inherent relevance than emotional faces, but during a matching task, they must

acquire task-related, goal-oriented relevance. That is, one face becomes the "right" face, while

the other face becomes the "wrong" face. If we accept that the matching task itself evokes

amygdalar relevance detection, we must still account for the absence of additional activation of

the amygdala by emotional content, and for the apparent lack of relevance detection during the

control condition.

Emotion Processing at the Amygdala Habituates

At the left amygdala, the mean amplitude of the BOLD response decreased over time for

emotion but not identity matching. This suggests that while both face matching conditions

activate the amygdala, there is still a distinct pattern of emotion processing at the left amygdala.

Figure 4-5 shows that the initial BOLD response at the left amygdala is larger to emotion than to









identity, but then rapidly diminishes. Although the emotion condition initially evokes greater left

amygdalar activation than the identity condition, habituation apparently prevented this difference

from being detected with the statistical approach used in this study.

A previous study found greater habituation to repeated, passively viewed faces in the right

amygdala compared with the left, and that activation in the left amygdala distinguished fearful

and happy faces (Wright et al., 2001). The authors described the right amygdala as a rapid but

general relevance detector, and the left amygdala as slower but capable of distinguishing

emotional valence. The current findings fit this model but do not test it explicitly. It is possible

that habituation of the right amygdalar response in the current study occurs rapidly within the

first block, making it difficult to detect. An event-related study of habituation during explicit and

incidental emotion processing may shed more light on the lateralization of the speed and

specificity of emotion processing at the amygdala.

An earlier study divided the left amygdalar response to unconsciously perceived emotional

faces into a ventral valence processing domain and a dorsal salience processing domain (Whalen

et al., 1998b). Since the face matching task does not investigate differences in valence, this may

explain the dorsal location of the amygdalar activation in the current study.

Spatial Processing Bypasses the Amygdala During the Control Condition

Although the control condition involved pattern matching it elicited no response from the

amygdala compared with rest. While face matching activated the ventral visual pathway (in

particular the right fusiform gyms), pattern matching activated mostly the dorsal visual pathway

(the interior parietal lobule and intraparietal sulcus bilaterally, see Table 4-3). Several

participants described a pattern-matching strategy of aligning the white squares in the probe and

target patterns. It is possible that this spatial alignment operation may be performed by the dorsal

visual pathway, avoiding the need for relevance detection by the amygdala.









However, it is not clear whether the amygdala detects relevance in visual information using

the categorical, ventral pathway exclusively. Adolphs and colleagues reported initial evidence

for emotion processing via the dorsal visual pathway in one patient with extensive encephalitis

lesions of the ventral surface. This patient was impaired at recognizing emotions in facial

expressions, but could recognize emotions when they were acted out (Adolphs et al., 2003). We

are unaware of any functional imaging study showing amygdala activation via the dorsal visual

pathway.

Cognitive Processing During Emotion Matching

The emotion condition selectively activated a region of the left inferior frontal sulcus

(Figure 4-2). A previous study found activation of the corresponding region in the right

hemisphere to explicit, but not incidental, evaluation of facial expressions (Gorno-Tempini et al.,

2001). Activity in this region of the left hemisphere has been associated with word reading

(Matsuo et al., 2003) and with action recognition (Hamzei et al., 2003). Since several subjects in

the current study reported mentally naming the target emotion, the observed left-sided activation

may reflect covertly naming or "reading" facial expressions, especially since unfamiliar neutral

faces and pixilated patterns are not easily named. While hemispheric differences in amygdala

activity have been hypothesized to result from ispsilateral prefrontal efferents (Irwin et al.,

2004), and Hariri and colleagues demonstrated inhibition of the amygdala by the prefrontal

cortex during emotion labeling (Hariri et al., 2000), it seems unlikely that the prefrontal

activation to emotion matching in the current study is responsible for the habituation of the left

amygdala. The active region in the current study is in the left dorsolateral prefrontal cortex, and

corresponds neither with the ventromedial prefrontal region associated with fear extinction in

rats (Quirk, and Gehlert, 2003), nor with the right ventrolateral region described by Hariri et al.

(2000). The correlation between the signal time course at the left amygdala and left inferior









frontal sulcus was small and not significant (r = 0.16), implying that these regions are not

functionally connected during this task.

While identity matching requires a simple, perceptual match, emotion matching requires

additional categorical processing (reflected in increased reaction time and decreased accuracy

[Table 4-1]). This cognitive component may indirectly link prefrontal activation and amygdalar

habituation during the emotion condition. The left amygdala's BOLD response to the emotion

condition decreases to below the level of its response to the identity condition (Figure 4-5 A+B).

Thus increased cognitive processing during the emotion condition may inhibit the amygdala,

resulting in lower "baseline" (task-related) activation once the effect of emotion has habituated.

We investigated the effect of task difficulty with a multivariate ANOVA using task condition

(emotion, identity, or control) and block order (1st, 2nd, 3rd, or 4th exposure to block) to predict

mean peak BOLD response in the prefrontal cortex and amygdala. Introducing reaction time as a

covariate measure of task difficulty did not modulate the effect of task condition on mean peak

BOLD response. We conclude that the observed brain activity was not quantitatively linked with

task difficulty in the present study.

Negative BOLD Responses

Of additional interest are the negative BOLD responses found at the anterior and posterior

cingulate cortices (Figure 4-4). These regions have been described as part of a network that is

active during rest and deactivated by a variety of tasks (McKiernan et al., 2003). Because this

network deactivated to both visual and auditory tasks (proportionate to difficulty in the latter

case), McKiernan et al. (2003) speculated that it mediates attention-dependant processing during

the conscious resting state, including monitoring emotional state. The posterior cingulate and

subgenual anterior cingulate deactivated during the control condition alone, whereas the

pregenual anterior cingulate deactivated during all three conditions, with the greatest decrease









for emotion. It is possible, therefore, that the posterior cingulate and subgenual anterior cingulate

are involved in similar functions at rest and during the emotion and identity conditions.

Conversely, matching emotions decreases activity in the pregenual cingulate cortex. A recent

meta-analysis found that the peri-genual anterior cingulate cortex was activated in emotional

studies and deactivated in cognitive studies (Bush et al., 2000). Furthermore, the "emotional"

cingulate interacts strongly with the amygdala (Pezawas et al., 2005). Because the emotion

condition involves a cognitive operation on emotional stimuli, the responses we observed at the

pregenual cingulate cortex have several possible interpretations. The larger deactivation may

reflect a coincidental need for increased attentional resources elsewhere or alternatively,

deactivation of the pregenual cingulate may be necessary for emotional processing in the

matching task.

Complex Contributions to Amygdala Activation

Activation of the amygdala by emotional face matching appears to reflect a combination of

processes. Both the identity and the emotion task involve relevance detection at the amygdala

simply because a choice between two faces must be made, while the control condition appears to

utilize a separate spatial processing network. It appears that additional left amygdalar activation

due to emotional content was not detected because of rapid habituation. Selective activation of

the left inferior prefrontal sulcus and habituation of the left amygdala during the emotion

condition may be indirectly linked via the common influence of increased top-down processing

during emotion matching. We conclude that activation of the amygdala to the emotional face-

matching task cannot be interpreted as bottom-up emotion processing alone, but likely involves

more general relevance detection involved in perceptual matching.









Table 4-1 Behavioral data
Task condition Response Accuracy (%)
time (ms)
Control 1625 86.1
Identity 1155 99.7
Emotion 1704 90.6
* Significantly different to both control and emotion, p < 0.001, unpaired Student's t-test.









Table 4-2 Clusters of activation for [(Emotion Identity) n (Identity Control)]
Region Side BA x y z Size t(11)
Inferior frontal sulcus L 9,44 -45 15 30 183 4.8
Precentral gyrus R 4 49 -4 52 144 6.0
Pregenual cingulate cortex L 24, 32 -4 37 10 169 -5.2
Transverse temporal gyrus L 41 -52 -29 15 210 -6.2
Only clusters >100 voxels shown except for a priori region (amygdala). L: left, R: right. BA:
Brodmann's Area. X, Y and Z refer to Talairach co-ordinates. Size: number of 1mm3 voxels. t:
random effects statistical score (degrees of freedom). Negative t values indicate deactivation. n
denotes conjunction of two contrasts.










Table 4-3 Clusters of activation for [(Emotion -


Region
Amygdala
Amygdala
Subgenual cingulate cortex
Fusiform gyms
Inferior temporal sulcus
Middle temporal gyrus
Posterior cingulate gyrus
Insula
Collateral sulcus
Fusiform gyrus
Inferior parietal lobule
Inferior parietal lobule
Intraparietal sulcus
Intraparietal sulcus
Middle occipital gyrus
Middle occipital gyrus
Superior frontal sulcus
Only clusters >100 voxels sh


Side
R
L
R
R
R
R
R
R
L
L
R
L
R
L
R
L
L


32
37
37
39
23

35, 36
19, 37
40
40
7, 19
7, 19
19
19
6


y z
-7 -10
-3 -10
33 -10
-42 -19
-70 -3
-59 9
-55 23
20 3
-40 -9
-58 -12
-37 36
-46 40
-68 42
-68 42
-79 13
-74 22
-3 55


Size
487
35
103
198
200
883
1491
122
114
126
350
404
3344
1081
870
1015
176


t(11)
6.0
6.0
6.6
5.1
7.6
8.6
6.0
-6.5
-6.8
-6.7
-6.6
-6.2
-5.9
-6.6
-5.3
-5.6
-6.1


own except for a priori region (amygdala). L: left, R: right. BA:


Brodmann's Area. X, Y and Z refer to Talairach co-ordinates. Size: number of 1mm3 voxels. t:
random effects statistical score (degrees of freedom). Negative t values indicate deactivation. rn
denotes conjunction of two contrasts.


Control) r) (Identity Control)]









Table 4-4 Regions showing significant modulation of BOLD response
Region BOLD modulation (% / run)
Emotion Identity Control
L amygdala -1.39 -0.14 -0.53
L transverse temporal gyms -0.52 0.33 -0.58
R amygdala 0.14 0.40 -0.22
R middle temporal gyrus 0.02 0.31 -0.12
B posterior cingulate cortex -0.32 0.76 -0.45
L pregenual cingulate cortex -0.43 0.31 -0.50
Values represent difference in peak BOLD activation (% signal change) over one run (four
blocks of each condition). significant correlation, p < 0.05.








A Emotion


B Identity


C Control


Figure 4-1 Matching task paradigm. Participants had to select which of the lower two probe
images matched the upper target image. The selected probe was outlined in yellow.
Faces were matched by emotion (A) or identity (B) and in the control condition (C),
participants matched pixilated patterns.











A 1.0
0.8
0.6
0.4
0.2
S0.0
-0.2
-*- Control
-0 4 4 rer ,i.v
-0.6 .
-9 -6 -3 0 3 6 9 12 15 18 21 24 27 30 33
Time (s)


y= 16 mm


4.0 t score 8.0


Figure 4-2 Selective response to emotion at the left inferior prefrontal sulcus. A: BOLD
response; vertical dotted lines indicate beginning and end of block; error bars denote
standard error of mean. B: Cluster of activation for [(Emotion Identity) rn (Identity -
Control)] with a threshold of t(1 1) > 4.0; slice location given in Talairach co-
ordinates; slice in radiological convention (Table 4-2). C: Left inferior prefrontal
sulcus activation for the group illustrated on a rendered 3D brain from a single
participant.












A Right amygdala
1.0
o 8 -- Control
0.8 -r Identity
-* Emotion
0.6



0.0


-0.2
-0.4
-0.6
-9 -6 -3 0 3 6 9 12 15 18 21 24 27 30 33
Time (s)


y = -5 mm


4.0 t score 8.0


C Left amygdala
1.0
0,8

0.6
c 0.4
S0.2

S0.0


-9 -6 -3 0 3 6 9 12 15 18 21 24 27 30 33
Time (s)


Figure 4-3 Response to face matching at the left and right amygdala. A+C: BOLD responses;
vertical dotted lines indicate beginning and end of block; error bars denote standard
error of mean. B: Clusters of activation for [(Emotion Control) rn (Identity -
Control)] with a threshold of t(1 1) > 4.0; slice location given in Talairach co-
ordinates; slice in radiological convention (Table 4-3).












A Pregenual cingulate
0.4 1


-0.8 .
-9 -6-3 0 3 6 9 12 15 1821 24273033
Time (s)
C Subgenual cingulate D Posterior cingulate
1.0 0.4

0.5 0.2




S-0.4
-1.5

-0.6
-2.0
-0.8 .
-9 -6 -3 0 3 6 9 121518 21 24273033 -9 -6 -3 0 3 6 9 12 15 18 21 24 27 30 33
Time (s) Time (s)


Figure 4-4 Regions of deactivation. B: "Glass brain" showing clusters of activation with a
threshold of t(1 1) > 4.0 in the cingulate cortex (Tables 4-2 & 4-3). Gray borders
denote the anterior and posterior commisures and the borders of the cerebrum. A,
C+D: BOLD responses; vertical dotted lines indicate beginning and end of block;
error bars denote standard error of mean. A: Pregenual cingulate cortex (-4, 37, 10).
The BOLD response is negative for all three conditions, with a significantly greater
decrease for emotion. C+D: Subgenual cingulate cortex (4, 33, -10) and posterior
cingulate cortex (1, -55, 23). The BOLD response is negative in the control
condition, but remains at baseline for emotion and identity.











A Emolion (left amygdala)
1.5

1.0

.0.5





--- Block 3
10 .-Bl. ock 4

-9 -6 -3 0 3 6 9 12 15 18 21 24 27 30 33
Time (s)
C Left amygdala


B Identity (left amygdala)
1.5


-9-6-3 0 3 6 9 12 15 18 21 24 2730 33
Time (s)
D Right amygdala


1.0

0 0.8

0.6

S0.4

S0.2

r 0.0
E
-0.2

-0.4


1 2 3 4


Figure 4-5 Habituation. A+B: The BOLD response at the amygdala decreases over time in the
emotion condition but not in the identity condition. BOLD responses are derived from
left amygdala activation clusters for the contrasts (emotion control) (A) and (identity
control) (B). C+D: The peak BOLD response habituates only at the left amygdala in
the emotion condition. Peak response = mean % signal change values for 9 sec 18
sec time points. Responses derived from left and right amygdala clusters for the
contrast [(Emotion Control) rn (Identity Control)] (Table 4-3).


2 3 4
Block









CHAPTER 5
DISSOCIATING EVENT-RELATED RESPONSES TO TOP-DOWN AND BOTTOM-UP
EMOTION PROCESSING

Conscious rationalization of emotional stimuli may modulate automatic physiological

responses to stimulus content. Functional neuroimaging studies have investigated top-down

modulation of emotion using tasks that require participants to rate their responses to emotional

pictures. These studies suggest that the amygdala and insula may mediate bottom-up responses to

the content of emotional scenes, and that these responses may be modulated by top-down

processing mediated by the anterior cingulate cortex (ACC) and medial prefrontal cortex. We

attempted to replicate these findings using an optimized paradigm design. Pleasant and

unpleasant pictures were rated using emotional and non-emotional rating scales: either the

pleasantness of the picture, or the frequency of its appearance on television. Stimulus content and

tasks instructions were randomized on a trial-by-trial basis, preventing expectancy of emotional

content, and equalizing the timing of bottom-up and top-down responses. Factorial analysis was

used to separate the main effects of bottom-up and top-down processing and their interaction.

The amygdala responded to unpleasant pictures during both tasks. The orbitofrontal cortex and

amygdala responded during emotional rating of both pleasant and unpleasant pictures. The ACC

responded to both stimulus content and task demands. No significant interaction effects were

found. These results show that an event-related picture rating task may dissociate bottom-up

processing in the amygdala from top-down processing in the orbitofrontal cortex and insula.

Introduction

Psychological and neuroanatomic evidence suggests that responses to emotional stimuli

may be mediated both by automatic, bottom-up processes and by intentional, top-down processes

(Eysenck, and Keane, 2000). Zajonc proposed the affective primacy hypothesis, which states the

appraisal of emotional stimuli is rapid and automatic (Zajonce, 1980, quoted in Eysenck &









Keane, 2000). He supported this hypothesis by showing that subliminally presented emotional

faces may influence pleasantness ratings of a subsequently presented Chinese pictogram.

Lazarus, on the other hand, argued that emotional responses are influenced by conscious,

contextual appraisal (Lazarus, 1982, quoted in Eysenck & Keane, 2000). He supported this

hypothesis by demonstrating that physiological responses to a disturbing film varied depending

on the narrative that accompanied the film. Although these hypotheses appear to disagree, later

theorists proposed that emotional experience may result from the appraisal of a stimulus by both

bottom-up and top-down processes (Ellsworth, 1994). Anatomical evidence supports multiple,

interacting pathways for emotional appraisal. The idea of two parallel emotion systems is

supported by anatomical evidence. In rats, conditioned fear responses are dependent upon the

amygdala. LeDoux showed that fear-inducing visual signals may reach the amygdala by two

pathways: a fast subcortical pathway, and a slow cortical pathway (LeDoux, 2000). Subsequent

functional neuroimaging studies in humans have sought to describe the roles of the amygdala and

associated cortical regions in emotional appraisal.

Several investigators have investigated bottom-up and top-down components of emotional

appraisal by employing emotion rating tasks. Bottom-up processes may be mediated by brain

regions that respond to differences in the emotional content of stimuli. Top-down processes may

be mediated by brain regions that respond only during explicit rating of emotion. By altering task

demands, these studies investigated the modulation of bottom-up responses by top-down

appraisal. In the earliest rating study, participants viewed blocks of mixed unpleasant and neutral

pictures during positron emission tomography (PET), and rated either whether each picture was

pleasant or unpleasant or whether it was indoors or outdoors (Lane et al., 1997a). Comparing

emotion rating with location rating, activation was seen in a cluster spanning the anterior









cingulate cortex (ACC) and medial prefrontal cortex (PFC) (both Brodmann Area 32) and in the

insula. The authors suggested that the ACC mediated internal attention specifically associated

with emotion processing, and that increased activity at the insula represented amplification of

interoceptive processing. This study could not investigate bottom-up responses because

unpleasant and neutral pictures were intermingled. In two subsequent studies, pleasant and

neutral pictures were presented separately to identify regions involved in bottom-up responses

(Liberzon et al., 2000;Taylor et al., 2003). Liberzon et al. (2000) compared emotion rating with

picture recognition, a cognitive task intended to draw attention away from emotion. The right

amygdala was activated by unpleasant compared with neutral pictures, and activation was greater

during emotion rating than during picture recognition. Taylor et al. (2003) compared emotion

rating with passive viewing, in order to test whether top-down processing diminished bottom-up

responses. The right amygdala and insula responded to unpleasant pictures, but the response was

smaller during emotion rating than during passive viewing. The opposite effect was seen in the

ACC and medial prefrontal cortex: activation by unpleasant pictures was increased during

emotion rating compared with passive viewing. A later study compared emotion rating with

ratings of personal relevance (Phan et al., 2004). The left amygdala responded more strongly to

emotional pictures during emotion rating; however the ACC and medial PFC were deactivated

during emotion rating. This finding suggests a specific role for these regions in the appraisal of

self-relevance, which Zajonc and Smith (1993) defined as a component of emotional appraisal

(quoted in Eysenck and Keane, 2000). This view is supported by the recent finding that the

medial PFC responds more strongly when rating one's own emotions than when rating other's

emotions (Ochsner et al., 2004). In summary, these studies suggest that the amygdala and insula









may mediate bottom-up responses to emotional stimulus content, and that these responses may

be modulated by top-down processing mediated by the ACC.

The studies above employ several task designs. In all but one study, emotional pictures are

presented in blocks, so neural responses to emotional content may be confounded with

expectancy of emotion, a top-down effect. Similarly, most studies tested for interactions between

bottom-up and top-down factors using single statistical contrasts, which do not account for the

main effects of each factor (Friston et al., 1996). Phan et al. (2004) presented emotional pictures

in a random, event-related design and tested interactions using a factorial analysis. However,

their rating tasks were presented in blocks. Because the bottom-up and top-down factors varied

at different time intervals, the neural responses to each factor may have unequal power. Finally,

these studies used a variety of control tasks. Passive viewing fails to control for the attentional

and motor components of emotion rating. Recognition and self-relevance rating include internal

appraisal and thus may be confounded with emotion rating. The indoor / outdoor judgment task

avoids these problems, but elicits only two categorical responses, while emotion rating tasks

typically elicit responses on a continuous scale.

In the current study, we investigate bottom-up and top-down emotional appraisal using an

optimized emotion rating paradigm. We employ an event-related picture rating task in which

both emotional content and task instructions are randomized. By randomizing both bottom-up

and top-down factors at the same temporal frequency we avoided contaminating our findings

with differential signal-to-noise characteristics of event- and block-level hemodynamic

responses. As a control task, participants were asked how frequently images similar to the one

displayed are shown on television. We chose frequency rating as a control condition because it

controls for attentional and motor components of the rating task, requires attention to external









stimulus features, and elicits a response on a continuous scale. We used factorial analysis to

identify the neural responses to the task. Regions showing a main effect of stimulus content were

defined as mediated bottom-up responses. Regions showing a main effect of rating task were

defined as mediating top-down appraisal. Regions showing an interaction effect were defined as

mediating bottom-up responses that were modulated by top-down appraisal. We tested the

hypotheses that the amygdala mediates a bottom-up response that is modulated by top-down

appraisal, and that the ACC mediates top-down appraisal.

Methods

Subjects

Sixteen healthy male participants gave informed consent as approved by the University of

Florida's Institutional Review Board. The participants had no history of psychiatric or

neurological illness, and were taking no psychotropic medication at the time of the study. One

participant was excluded due to discrete head movements greater than 1mm during scanning.

Picture Rating Task Paradigm

Participants were presented with pictures from the International Affective Picture System

(IAPS) (Center for the Study of Emotion and Attention [CSEA-NIMH], 2001). Below each

picture, a cue instructed participants to make either an emotion or frequency rating. The emotion

rating cue read, "How pleasant do you find the content of this image?" The frequency rating cue

read, "How frequent do images with similar content appear on television?" The participants

selected one of four responses: for emotion ratings, very unpleasant, moderately unpleasant,

moderately pleasant, or very pleasant, and for frequency ratings, weekly, daily, hourly, or

continuously. We selected emotionally arousing IAPS pictures and assigned them, based on the

valence ratings, to two groups: pleasant or unpleasant (Lang et al., 2001). The mean ratings were,

for the pleasant set, pleasure = 6.7 +/- 0.9, arousal = 4.7 +/- 1.0, and for the unpleasant set,









pleasure = 3.7 +/- 1.1, arousal = 4.8 +/- 1.3 (mean +/- standard deviation). IAPS picture codes

are listed in appendix A. The trials were categorized by rating task and valence, giving four trial

types: emotion rating pleasant (EP), emotion rating unpleasant (EU), frequency rating pleasant

(FP), and frequency rating unpleasant (FU).

Prior to performing the task, each participant was familiarized with the task and the

scanner environment by completing a training run consisting only of emotion ratings. Different

sets of pictures were used for the training and task runs. The results of the training run are

reported elsewhere, in a study of the effects of training upon emotional and non-emotional

ratings (Li et al., 2006).

During the task run, all four trial types were presented in a random order using an event-

related design. Fifteen trials of each type were presented for 3 sec each, along with 30 null trials,

during which a fixation cross was displayed for 3 sec. The test run lasted 4 min 30 sec. Null trials

were included in the random sequence in order to jitter the stimulus onset asynchrony (SOA)

between trials. This increases the variance in the resulting fMRI response, making the response

to rapid stimuli (SOA < 15 sec) detectable (Burock et al., 1998). Jittering the SOA with

randomly interspersed null trials creates a geometric distribution of SOAs, which is more

efficient than uniform randomization (Serences, 2004). The resulting mean SOA was 4.5 sec,

with a minimum of 3 sec. This timing was chosen to minimize response attenuation when

repeating emotional stimuli, but to maximize the number of trials in the run (Soon et al., 2003).

The stimuli were presented using an Integrated Functional Imaging System (IFIS, MRI

Devices, Inc., Waukesha, WI) with a 7" LCD screen at 640 X 480 pixel resolution, mounted over

the subject's head and viewed using a fixed mirror. The screen subtended approximately 140 x

110 of the visual field. A PC running E-Prime (Psychology Software Tools, Pittsburgh, PA)









began presenting each task in synchronization with the first RF pulse of each scan. Responses

were collected with a MRI-compatible button glove attached to the participant's right hand.

Functional Imaging Data Acquisition

Participants were scanned using a 3 Tesla Siemens Allegra scanner with a standard head

coil (Siemens, Munich, Germany). Anatomic images were acquired using an MPRAGE sequence

with TR = 1500 ms, TE = 4.38 ms, and flip angle = 80. In the axial plane, 160 slices were

acquired (thickness 1.0 1.2 mm, according to the height of the brain) with in-plane field of

view 240 mm X 180 mm and matrix size 256 X 192. Functional images covering the whole brain

were acquired using echo-planar imaging sensitive to blood-oxygenation level dependent

(BOLD) effects, with TR = 3000 ms, TE = 30 ms, flip angle = 900. In the axial plane, 38 slices

with a thickness of 3.8 mm were aligned with the plane of the intercommissural line and had an

in-plane field of view 240 X 240 mm and matrix size 64 X 64. The first two volumes of each

functional run were discarded to allow for TI equilibration. These settings have previously been

shown to provide reasonable coverage of the amygdala while allowing coverage of the whole

brain, and without sacrificing BOLD sensitivity (Wright, and Liu, 2005). Visual inspection of

functional images showed coverage in the amygdala was adequate in ten out of fifteen

participants. Because our a priori hypotheses predicted responses in the ACC, coverage of this

region was inspected. Signal in the subgenual ACC was lost to susceptibility artifact, and

responses within the affected region were discarded.

Functional Imaging Data Analysis

Data were analyzed using BrainVoyager QX version 1.7.6 (Brain Innovations, Maastricht,

Holland). The functional images were coregistered with anatomic images, and normalized to

Talairach space for each participant. Functional data underwent 3D motion correction, linear

trend removal and slice scan time correction. The test runs underwent Gaussian spatial









smoothing using a kernel of 5.7 mm (1.5 voxels) full-width half-maximum (FWHM). The

training run underwent spatial smoothing as above and temporal Gaussian smoothing using a

kernel of 4 data points (12 sec) FWHM.

Task-related activity was mapped using a voxel-wise general linear modeling analysis. For

both event-related and block analyses, the BOLD response to each task condition was estimated

using a standard hemodynamic model (Friston et al., 1998). The estimated responses were fit to

the MR signal for each individual to generate a beta weight, reflecting the magnitude of the

contribution of each task type to the overall model. Using the conservative random-effects

approach, statistical maps were generated by applying second-order statistics to the group's beta

weights at each voxel. For the test runs, a two-way ANOVA was used to estimate separately the

main effects of judgment type and emotional valence, and their interaction. This approach avoids

the assumption of pure insertion, allowing the localization of neural responses that correlate with

cognitive task components in a nonlinear fashion (Friston et al., 1996). In the training run a t-test

was used to compare early vs. late training responses (the first two blocks often vs. the last two).

For whole-brain analysis, thresholds were set to exclude clusters smaller than 100 mm3 (after

functional data were resampled to 1 mm resolution), and statistical scores below F (1,14) = 12, p

< 0.005. For regions for which we had an a priori hypothesis (amygdala, OFC, and ACC) the

statistical threshold was lowered to F(1,14) = 5, p < 0.05. At each region, t-score were calculated

post-hoc using single statistical contrasts to indicate the direction of the main effect or

interaction.

Mean BOLD responses were plotted using BrainVoyager's event-related averaging

function. In the test runs, BOLD responses were calculated for each of the four trial types. In the

training run, BOLD responses were plotted for early, middle, and late blocks. For each test run









event, percentage signal change was calculated relative to the signal during the two seconds prior

to stimulus onset. These values were averaged by event type in a time window from -2 to 13

seconds relative to stimulus onset. Contamination from subsequent stimuli occurring within the

13-second window was eliminated in the overall average due to the jittered SOA (Dale, and

Buckner, 1997). Events were not time-locked to the fMRI sampling period (3 seconds) to allow

finer sampling of the BOLD response (Serences, 2004). Data therefore were resampled to 1

second resolution by interpolation. BOLD responses to the training paradigm were calculated in

a similar way, but using the original sampling period. Signal change was calculated relative to

the two periods preceding the start of each block (6 seconds). A time window from -10 to 36 sec

was used to cover the BOLD response to the entire block.

Results

Behavioral Data

Participants rated pleasant and unpleasant stimuli appropriately. Pleasure ratings, adjusted

to the standard scale used in the IAPS of 1 9 were significantly higher for pleasant pictures

compared with unpleasant pictures (6.6 +/- 0.9 vs. 3.6 +/- 0.8, p < 0.001). Response times were

slower during frequency rating, regardless of emotion, implying that the control task was more

difficult (Table 5-1).

fMRI Data

Responses were observed in the regions for which an a priori hypothesis was proposed. A

main effect of task was observed in the left OFC and right insula (Figure 5-1). These regions

showed positive BOLD responses to emotion rating and negative responses to frequency rating.

Supporting the statistical main effect, the BOLD responses appeared equal for pleasant and

unpleasant pictures (Figure 5-1 B & D). The left amygdala responded selectively to unpleasant

pictures (Figure 5-2). Activation was detected at reduced threshold (F(1,14) = 5, p < 0.05) due to









the a priori selection of this region. A positive BOLD response was seen for negative pictures

during both tasks, supporting the main effect of valence. Examination of the ACC, also at

reduced threshold, revealed three activations, two exhibiting a main effect of valence, and one a

main effect of task (Figure 5-3). These clusters were distinct, but overlapped slightly. The BOLD

responses in these regions suggested that in both clusters exhibiting a main effects of valence, the

result was driven by positive responses to pleasant pictures. In the cluster exhibiting a main

effect of task, the result appeared to be driven by a positive response to emotion rating (Figure 5-

3 B-D). However, the BOLD response curves were not as clearly separated in these regions as

they were in the OFC, insula, and amygdala.

Regions for which we proposed no a priori hypothesis also showed interaction effects

(Table 5-2) and main effects of valence (Table 5-3) and task (Table 5-4). An interaction between

task and valence was located between the left insula and temporal operculum, a region anterior

and posterior to the main effect of task at the right insula. The BOLD responses at this region

(not shown) were more variable than at other regions, and did not have a typical curve shape,

suggesting they may have arisen from the nearby middle cerebral artery. Several regions showed

a main effect of valence, namely the bilateral posterior cingulate cortex, postcentral gyrus, and

posterior fusiform cortex. A main effect of task was seen in the bilateral parieto-occipital sulcus,

showing greater BOLD responses to the frequency task, regardless of valence. Several other

regions showed greater responses to frequency ratings, predominantly in the left PFC, including

premotor cortex and supplementary motor area (SMA).

Discussion

The aim of the current study was to identify regions of the brain involved in bottom-up and

top-down emotional appraisal. We pursued this aim using an event-related, factorial design to

identify regional brain responses that varied with stimulus valence, task instructions, or an