Investigating Attentional Bias on Distinct Mechanisms of Attention

Sandersan Onie

A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy

School of Psychology Faculty of Science

August 2019

Thesis Dissertation Sheet

Surname/Family Name : Onie Given Name/s : Sandersan Abbreviation for degree as give in the University calendar : Ph.D. Faculty : Science School : Psychology Thesis Title : Investigating attentional bias on distinct mechanisms of attention

Abstract 350 words maximum:

Attentional biases to negative information are proposed to play an important role in emotional disorders. Moreover, retraining these biases may lead to clinical benefits. However, there is much conflicting evidence, possibly due to the field’s reliance on measures of primarily spatial attention. This thesis investigates the relationship between attentional bias and psychopathology, and its trainability, while considering distinct mechanisms of attention.

Following an introduction (Chapter 1), Chapter 2 investigated whether inhibition training was able to elicit similar patterns as typical attentional bias modification using the Dot Probe (DP). The results provided evidence that none of the training conditions elicited changes in attentional bias, as well as for a lack of relationship between the DP and psychopathology. Subsequent experiments in this thesis shifted focus to emotion-induced blindness (EIB), a non-spatial attention index, and investigated its psychometric properties, underlying mechanisms, links with negative affect, and malleability. Chapter 3 investigated the test-retest reliability of EIB, finding favourable test-retest reliability compared to indices reported in the literature. Chapter 4 examined whether the emotional valence or arousal of stimuli accounted for their impact in EIB and the DP. While there was evidence for an overall effect of emotion in both tasks, valence and arousal uniquely accounted for performance in EIB at lag 2, but neither valence nor arousal accounted for behavioural patterns in the DP. Chapter 5 investigated EIB’s locus of competition by modifying EIB to include a priming task at the end of each stream. The results found that missed targets did not prime responses to subsequent stimuli, suggesting that perceptual impairments for them occurred early in processing. Finally, Chapter 6 investigated the trainability of EIB, finding across two experiments that EIB could not be retrained. Importantly, across chapters 3 to 6, EIB never failed to replicate.

This thesis provides deeper understanding for emotional prioritization measured by EIB, underscoring its robustness, stability across time, and sensitivity to gradations of valence and arousal. Despite this, these studies suggest that the relationship between task-irrelevant, attentional capture by emotional stimuli in EIB and psychopathology is tenuous, as is the ability to retrain EIB.

Declaration relating to disposition of project thesis/dissertation

I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).

…………………………………………………………… ……………………………………..……………… 30/08/2019 Signature Witness Signature Date The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.

FOR OFFICE USE ONLY Date of completion of requirements for Award:

ii ORIGINALITY STATEMENT

ORIGINALITY STATEMENT ‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.’

Signed ……………………………………………......

Date 30/08/2019

‘I hereby grant the University of New South Wales or its agents a non-exclusive licence to archive and to make available (including to members of the public) my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known. I acknowledge that I retain all intellectual property rights which subsist in my thesis or dissertation, such as copyright and patent rights, subject to applicable law. I also retain the right to use all or part of my thesis or dissertation in future works (such as articles or books).’

‘For any substantial portions of copyright material used in this thesis, written permission for use has been obtained, or the copyright material is removed from the final public version of the thesis.’

Signed ……………………………………………......

Date ……………………………………………...... 18/12/19 ......

AUTHENTICITY STATEMENT ‘I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis.’

Signed ……………………………………………......

Date ……………………………………………...... 18/12/19 ...... INCLUSION OF PUBLICATIONS STATEMENT iii

INCLUSION OF PUBLICATIONS STATEMENT

UNSW is supportive of candidates publishing their research results during their candidature as detailed in the UNSW Thesis Examination Procedure.

Publications can be used in their thesis in lieu of a Chapter if: • The student contributed greater than 50% of the content in the publication and is the “primary author”, ie. the student was responsible primarily for the planning, execution and preparation of the work for publication • The student has approval to include the publication in their thesis in lieu of a Chapter from their supervisor and Postgraduate Coordinator. • The publication is not subject to any obligations or contractual agreements with a third party that would constrain its inclusion in the thesis

Please indicate whether this thesis contains published material or not. ☐ This thesis contains no publications, either published or submitted for publication

Some of the work described in this thesis has been published and it has been ☒ documented in the relevant Chapters with acknowledgement

This thesis has publications (either published or submitted for publication) incorporated ☐ into it in lieu of a chapter and the details are presented below

CANDIDATE’S DECLARATION I declare that: • I have complied with the Thesis Examination Procedure • where I have used a publication in lieu of a Chapter, the listed publication(s) below meet(s) the requirements to be included in the thesis. Name Signature Date (dd/mm/yy) Sandersan Onie 30/08/2019

iv ACKNOWLEDGEMENTS

Acknowledgements

First and foremost, I would like to thank God for the amazing opportunity to do what I do, for connecting me with such amazing people. Everything is from you, through you and for you.

I would like to thank my supervisor Steve Most, for his encouragement, support, advice and mentorship. I could not have had a better supervisor. Today, when this is being written, we are sitting across each other working on this thesis. I am intimidated. I am intimidated, in awe and inspired by your brilliance, intellect, unrivaled communication ability, and your patience.

Thank you for guiding me, valuing my opinions, and having my best interests at heart. I aspire to be a mentor like you and I would not be where I am without your support.

I would like to thank old lab mates: Briana Kennedy, Jenna Zhao, Vera Newman and

Poppy Watson. When I first started out my research career, all of you supported me and taught me how to function as a researcher here at UNSW. Thank you for sending me papers during my honours year and being extremely supportive. I cannot express how important that time was for me, and how that helped me get started. I enjoyed the time we spent eating cake, snacks and introducing me to cronuts. I am glad to know you are all well. Thank you to Poppy Watson. your presence in the lab is comforting, as you are always ready to lend a helping hand. Thank you for all the stimulating conversations, for the times we spent (especially at the writing retreat last year), and that you’re always so ready to help me, or anyone else in the lab that may need your expertise. ACKNOWLEDGMENTS v

I would like to thank the academics at UNSW who have guided me and taught me many things along the way. I would like to especially thank Mike Le Pelley for taking the time to teach me new skills, and Chris Donkin for always being around and open to the many questions we’ve had. I enjoyed our chats immensely.

I would like to thank my friends on Level 10 as well as the honours students I’ve had the opportunity of mentoring: David Jin, Alicia Heng, Eliza Rodgers, Wing See Yuen, Clara DeTorres,

Jenny Le, Ananda Vasudevan, Brandon Le and Elizabeth Phung. Meeting such amazing students have been one of the greatest joys during these three and a half years. You are all so much more brilliant than I could have imagined, and I am grateful you are all doing well.

I would like to thank my friends on Level 7, who are still there and who have since left, where I spent many days in laughter and ludicrousness: Manos Konstantinidis, Arthur Kary, Rob

Taylor, Hanbit Cho, Peggy Wei, Jeremy Ngo, Belinda Xie, Aba Szollözi, Yonatan Vanunu and

Annalese Bolton. Thank you for the conversations over lunch, for the dinners we’ve had, and always being there. Level 7 has become a home in a way, thanks to all of you. Thank you for the nights out in Boston, New York and Vancouver. You all have made an otherwise lonely journey into a memorable and fun one. Sitting around in a crowded room, talking about trivial things such as the sheer magnitude of Garston’s lunches, is one of the things I will miss most.

I would like to thank especially Adrian Walker and Garston Liang, who have walked this road with me since we were undergraduates. Thank you for being there in my ups and downs.

Your very presence at the university comforts me, knowing there are people who I can turn to and rely on. I only hope I have been able to be as good of a friend to you both. vi ACKNOWLEDGEMENTS

I would like to the fellow researchers my friends outside of the university, whom I have had the privilege of meeting, chatting, and getting to know: Patrick Clarke, Lies Notebaert, Julie

Ji, Julian Basanovic, Mary Peterson, Gina Grimshaw, Ottmar Lipp, Jolene Cox, Enrique

Mergelsberg, Stefany Christina and most recently: Jemma Todd, Katrina Prior, Henry Austin,

Georgina Wright, Jessie Georgiades and Emily Rose. It has been the greatest honor to be in the presence of amazing researchers and even greater people. I enjoyed the chats and time I spent with many of you. Thank you for showing me that hard and precise science is done by such a warm and friendly community.

A very special thank you to Reinout Wiers, Ron Rapee and Colin MacLeod. Thank you for your warmth and hospitality and for the chats we’ve had. My respect and admiration for you goes beyond your impact on the world and a level of intellectual perspective that if I am lucky, I hope to someday reach – but also your kindness and generosity. You inspire me to think big and challenge myself.

I would like to thank the UNSW Psychology support staff especially Jonathan Solomon,

John Bolzan, Socrates Mantalaba, Danny Chen, Rebecca Sinclair, Gary Mann and Camilla Leung, and so many others I cannot name, for the constant support and friendship you all provide for us. Without you all, everything would grind to a halt. I cannot thank you enough.

I would like to thank my church family, especially Jan Otto Wijaya, Lenny Arief, Nikkiady

Arief, Gwyneth Hidajatno, Stephanie Benita, Ryana Rachmat, and Juventius Kevin, for always being there for me when life outside of my thesis becomes too great a burden to bear.

I would like to thank my partner, Jessica Felisa Nilam for her encouragement and support, for loving me so wholly and wonderfully. I hope I have made you proud. ACKNOWLEDGMENTS vii

I would like to thank my family. Thanks to my mum for always encouraging and believing in me. During this period, both grandma and grandpa entered into eternity, but you always championed me and everything I do. Thanks to my dad who always strove to provide for us, no matter how difficult the circumstances, even sacrificing yourself throughout your life. Thanks to my sister, who would always take good care of her siblings whenever we are together. Thanks to my brother Sammy who always encouraged me through your love language of giving. This family is one of the greatest privileges I have in this world.

Finally, I would like to thank myself for not giving up. You worked hard, Sandy, and you deserve to acknowledge it. You deserve to eat roast pork in Bali and lie down at the beach or soak in a Japanese hot spring after this. Take a break. viii ABSTRACT

Abstract

Following an introduction (Chapter 1), Chapter 2 investigated whether inhibition training was able to elicit similar patterns as typical attentional bias modification using the Dot

Probe (DP). The results provided evidence that none of the training conditions elicited changes in attentional bias, as well as for a lack of relationship between the DP and psychopathology.

Subsequent experiments in this thesis shifted focus to emotion-induced blindness (EIB), a non- spatial attention index, and investigated its psychometric properties, underlying mechanisms, links with negative affect, and malleability.

Chapter 3 investigated the test-retest reliability of EIB, finding favourable test-retest reliability compared to indices reported in the literature. Chapter 4 examined whether the emotional valence or arousal of stimuli accounted for their impact in EIB and the DP. While there was evidence for an overall effect of emotion in both tasks, valence and arousal uniquely accounted for performance in EIB at lag 2, but neither valence nor arousal accounted for behavioural patterns in the DP. Chapter 5 investigated EIB’s locus of competition by modifying

EIB to include a priming task at the end of each stream. The results found that missed targets did not prime responses to subsequent stimuli, suggesting that perceptual impairments for ABSTRACT ix them occurred early in processing. Finally, Chapter 6 investigated the trainability of EIB, finding across two experiments that EIB could not be retrained. Importantly, across chapters 3 to 6, EIB never failed to replicate.

x ABSTRACT

Publications

Chapter 2: Onie, S., Notebaert, L., Clarke, P. & Most, S.B. (2019) Investigating the Effects of Inhibition Training on Attentional Bias Change: A Simple Bayesian Approach. Frontiers in Psychology. Chapter 3: Onie, S. & Most, S.B. (2017) Two roads diverged: Distinct mechanisms of attentional bias differentially predict negative affect and persistent negative thought. Emotion, 15(7), 884.

I certify that these publications were a direct result of my research towards this PhD, and that reproduction in this thesis does not breach copyright regulations.

Sandersan Onie 30th August 2019 TABLE OF CONTENTS xi

Table of Contents

Thesis Dissertation Sheet ………………………………..…………………………………………………………………….... i

Originality Statement ……………………………………….…………………………….………………………………………. ii

Inclusion of Publications Statement ………………………………………………………………………………………..iii

Acknowledgements ……………………………………….…………………………….………………………………………… iv

Abstract ……………………………………….…………………………….………………………………………………………….viii

Publications ……………………………………….…………………………….…………………………………………………..... x

Table of Contents ………………….…….…………………………….…………………….……………………………………. xi

List of Figures ……………………………………….…………………………….…………………………………………………. xv

List of Tables ……………………………………….…………………………….…………………………………………………. xvi

Chapter 1: Introduction and overview ……………………………………….…………………………….…………..….1

The Taxonomy of External and Internal Attention ………….…………………………….……………….3

External Attention ………….…………………………….…………………….……………………………..4

Internal Attention ………….…………………………….…………………….………………………………6

Summary and Interim Discussion ……………….…………………….………………………………..8

Measures of Attentional Bias to Negative Information ………………...…………………….…………9

Spatial Attention Tasks ……………….…………………….………………..……………………………11

Temporal Attention Tasks ……………….…………………….…………………………………………16

Attentional Bias Modification ……………….…………………….……………………………….………………20

Reviewing Past Meta-Analyses ……………….…………………….……………………………….……….…..21

Attentional Bias and Anxiety ……………….…………………….……………………….……………22 xii TABLE OF CONTENTS

Attentional Bias Modification and Subsequent Impact on Anxiety ………..…………24

Possible Publication Bias in the Literature ……………….…………………….……………………………30

Considering other Mechanisms of Attention ….…………………….……………………………………. 33

The Current Study ……………….…………………….………………………………………………………………..34

Chapter 2: Investigation of Inhibition Training on Attentional Bias Change ………………………….36

Methods ……………….…………………….………………………………………….…………………….…….………42

Results ……………….…………………….………………………………………….…………………….……………… 48

Discussion ……………….…………………….………………………………………….…………………….…….…… 54

Chapter 3: Test-Retest Reliability of Emotion-Induced Blindness …….…………………….………..……61

Methods ……………….…………………….………………………………………….…………………….…………… 69

Results ……………….…………………….…………………………………..…….…………………….………….…… 70

Discussion ……………….…………………….……………………………….………….…………………….…………73

Chapter 4: Investigating whether Valence or Arousal drive Attentional Capture in two Distinct

Attentional Paradigms ……….…………………….………………………………………….……………..……….…………76

Experiment 1 …………………………………………………………………………………………….…………………80

Methods ………………………………………………………………………………………….………………80

Results ……………………………………………………………………………………………..………………84

Discussion ………………………………………………………………………………………….…………… 91

Experiment 2 ………………………………………………………………………………………………….……………92

Methods ……………………………………………………………………………………………….…………92

Results ……………………………………………………………………………………………………..………94

Discussion …………………………………………………………………………………………………..……96 TABLE OF CONTENTS xiii

General Discussion ………………………………………………………………………………………….………..…97

Chapter 5: Investigating the Locus of Competition in Emotion-Induced Blindness ………………101

Methods ……………………………………………………………………………………………………………………106

Results ……………………………………………………………………………………………………………….………110

Discussion …………………………………………………………………………………………………………….……115

Chapter 6: Does a Single Session of Training Modify Emotion-Induced Blindness? ……………..119

Experiment 1 …………………………………………………………………………………………………………..…124

Methods ……………………………………………………………..…………………………………………124

Results …………………………………………………………………..………………………………………128

Discussion ………………………………………………………………………………………………………129

Experiment 2 ………………………………………………………………………..……………………………………130

Methods ……………………………………………………………………..…………………………………130

Results ……………………………………………………………………………………………………………132

Discussion ………………………………………………………………………………………………………134

General Discussion ………………………………………………………………….…………………………………135

Chapter 7: General Discussion ………………………………………………………………………………………………139

The lack of Relationship Between Attentional Bias and Anxiety …………………………………144

Possible Moderating Factors ………………………………………….………………………………144

Flawed Measurement Tools and Analyses ………………………..……………………………147

No Relationship between Attention and Psychopathology.……………….……………150

Attentional Bias Modification …………………………………………………………………………………….152

Retraining the Dot Probe ………………………………………………………………….……………152 xiv TABLE OF CONTENTS

Retraining Emotion-Induced blindness ………………………………………………..…………153

Findings from Emotion-Induced Blindness …………………………………………………………………154

Strengths and Limitations …………………………………………………………………….……………………156

Constraints on Generality ………………………………………………………………………………158

Future Directions ………………………………………………………………………………….……………………159

Emotion-Induced Blindness ……………………………………………………………………………159

Formalizing Models ……………………………………………………………………..…………………159

Conclusions and Summary …………………………………………………………………………………………160

References ……………………………………………………………………………………………………………………………161

LIST OF FIGURES xv

List of Figures

Chapter 1

Figure 1. Schematic Representation of the Emotional Stroop Task…………………………………………..10

Figure 2. Schematic Representation of a Dot Probe Trial………………………………………………………….12

Figure 3. Schematic Representation of the Visual Search Task Trial………………………………………….13

Figure 4. Schematic Representation of the Emotional Spatial Cueing Paradigm Trial……………….14

Figure 5. Schematic Representation of the ARDPEI Trial…………………………………………………………..16

Figure 6. Schematic Representation of an Emotional Attentional Blink Trial…………………………….18

Figure 7. Schematic Representation of an Emotion-Induced Blindness Trial…………………………….19

Chapter 2

Figure 8. Schematic of Training Conditions in Experiment 1…………………………………………………….47

Chapter 5

Figure 9. Schematic Representation of the modified Emotion-Induced Blindness Trial ………….109

Figure 10. Example Answer Screen of the modified emotion-induced blindness Task..….………110

Chapter 6

Figure 11. Schematic Representations of the different training conditions in Experiment 1…..127

Figure 12. Schematic Representations of the different training conditions in Experiment 2…..131

xvi LIST OF TABLES

List of Tables

Chapter 2

Table 1. Depression, Anxiety, and Stress Scale 21 Means and Standard Deviations …………………50

Table 2. Attentional Bias Indices for three training conditions …………………………………………………51

Table 3. Reliability Indices of Attentional Bias Indices ……………………………………………………………..51

Chapter 3

Table 4. Accuracy Means and Standard Deviations for Emotion-Induced Blindness …………..……71

Table 5. Test-Retest ICC Coefficients ……………………………………………………………………………………….73

Chapter 4

Table 6. Accuracy Means and Standard Deviations for Emotion-Induced Blindness …………..……90

Table 7. Response Time Means and Standard Deviations for Dot Probe…………………………..………94

Chapter 5

Table 8. Accuracy Means and Standard Deviations for Emotion-Induced Blindness ………………111

Table 9. Aggregated RTs to the arrow, binned according to congruence and correct/incorrect responses in the emotion-induced blindness task………………………………………………………………….112

Chapter 6

Table 10. Experiment 1 Accuracy Means and Standard Deviations for emotion-induced blindness ……………………………………………………………………………………………………………………………….129

Table 11. Experiment 2 Accuracy Means and Standard Deviations for emotion-induced blindness ……………………………………………………………………………………………………………………………….133

Chapter 7 LIST OF FIGURES xvii

Table 12. Median and 95% Credible Intervals of Emotion-Induced Blindness Effect Size

Estimates per Thesis Chapter ………………………………………………………………………………………………..155

18 INTRODUCTION AND OVERVIEW

Chapter 1:

Introduction and overview

CHAPTER 1 19

Introduction and Overview

The work in this thesis aims to establish new understanding of how people spontaneously attend to emotional stimuli, and it tackles three key questions: a) to what degree does spontaneous attention to negative emotional information (i.e., “negative attentional bias”) link to measures of psychopathology; b) to what degree can negative attentional biases be retrained; and c) to what degree is it fruitful to isolate different mechanisms of attention when pursuing links with psychopathology and retrainability?

In a world rich in information, attention mechanisms help us prioritise and condense the environment into useful chunks of information. These selected chunks ultimately form our lived experience. For example, two people at the same place and time may pay attention to different things, thereby experiencing that moment differently, responding with different emotions, and forming different memories. Although people often can purposefully choose how they allocate attention (e.g. searching for a friend in a crowded street or for an outfit in a wardrobe), stimuli can also capture our attention without our intention, either because of their salient physical properties (e.g., brightness) or because of their emotional relevance (e.g., a snake on a wooded path). This can be adaptive, but some conditions – such as heightened anxiety – tend to be linked with disproportionate allocation of attention towards perceived threat, a pattern known as a negative attentional bias (Stöber, 1993; Van Bockstaele et al., 2014). Simply put, if what we pay attention to forms our subjective experience of the world, consistently allocating attention towards perceived threat paints a picture of a world that indeed deserves to be feared.

The Taxonomy of External and Internal Attention. 20 INTRODUCTION AND OVERVIEW

In light of the fact that work on attentional biases is rarely accompanied by consideration of multiple mechanisms subsumed under the umbrella term ‘attention’, the next several pages provide a primer on some mechanistic distinctions delineated within the visual cognition literature. By virtue of this goal, this section describes some mechanisms that are not central to the work in this thesis. However, I try to strike balance by placing particular emphasis on the mechanisms most relevant to the current work and incorporate findings relevant to the attentional bias and psychopathology literature as well.

A helpful guide to understanding attention was developed by Chun, Golomb and Turk-

Brown (2011)in a paper entitled ‘A Taxonomy of External and Internal Attention’. In this review, the authors noted key characteristics of attention and broadly divided attentional processes into two categories: internal and external attention. Even though the authors of that guide discussed a wide range of modalities and topics, in this thesis I limit my discussion to those directly relevant to the literature on anxiety-linked attentional biases.

At a minimum, attention fulfils two key functions. First, attention is selective, a function that stems from limits in the number of items we can process in any moment. At any single point in time, there exist far more sources of internal and external information than can be effectively processed, far more memories than can be accessed and recalled, and far more responses and behaviours to be executed. Much attention research is devoted to understanding what information is selected, how is it selected, and at which stage in processing is it selected (e.g., does unselected information get processed enough to nevertheless affect our behaviour?). CHAPTER 1 21

Secondly, attention is responsible for the modulation of information processing, a process which occurs after selection. For example, once an object is selected, attentional processes determine how quickly it is processed, and/or how well it is remembered. Indeed, just because a stimulus has been selected, it does not necessarily mean it is remembered (Levin

& Simons, 1997; See Yuen et al., in prep).

All of these processes link closely with attentional bias research; however, selection in particular plays a core role in modern cognitive theories in anxiety, in which anxiety is associated with biased competition in favour of emotionally negative information (Bar-Haim et al., 2007; Onie & Most, 2017; Stöber, 1993; Van Bockstaele et al., 2014; but see Kruijt, Parsons,

& Fox, 2019). In the taxonomy, the authors broadly divide attentional processes based on their targets into two categories: external attention and internal attention.

External Attention. External attention refers to the observed attentional processes external to the self. That is, in an existence where objects primarily move in the dimensions of space and time, attentional processes also allow for selection on these two dimensions. They are referred to as spatial and temporal attention respectively. Note that Chun and colleagues

(2011) also discuss other forms of external attention such as object based attention, which is the preferential processing of certain objects in the environment, and feature based attention, which is increased attention towards a certain perceptual feature e.g. the colour red. All the tasks used in the attentional bias literature broadly fall into either spatial attention or temporal attention tasks. Therefore, for the purposes of this thesis, I will only focus on spatial and temporal attention, as they form the basis for the existing attentional bias literature. 22 INTRODUCTION AND OVERVIEW

Spatial Attention. Spatial attention refers to the process of prioritizing an area in the visual field for further processing. Spatial attention has been described as a spotlight, in which information within the spotlight is processed preferentially (Michael I. Posner, Snyder, &

Davidson, 1980). Another metaphor which builds upon the spotlight metaphor is a zoom lens metaphor, in which the size of the spotlight can be widened or narrowed with consequences on processing efficiency (Cave & Bichot, 1999; Eriksen & St. James, 1986). Beyond the widening or narrowing of the zoom lens, previous studies have also shown that spatial can be split into multiple regions, with consequences on processing (Awh & Pashler, 2000). Overall, the greater the area of focus, whether in one or more locations, the less efficient the processing.

Spatial attention is closely linked with foveation, that is, the phenomenon that objects in the area of fixation are higher in resolution. The farther the regions or object from the fovea, the more granular and less focused the image. As a result, eye movements often serve as an index for attention, as objects in fixation are processed further (Deubel & Schneider, 1996).

However, previous evidence has also dissociated spatial attention from eye fixation, in that attention can be deployed in an area that is not foveated (Hunt & Kingstone, 2003). This is called covert attention. Conversely, when attention and eye fixation move in tandem, it is called overt attention. Overt attention is also by far the dominant mode of assessing attention in the attentional bias literature, which we will see in one of the following sections of tasks and paradigms of the attentional bias literature.

Temporal Attention. Distinct from spatial attention, temporal attention refers to the process of deploying attention at specific points in time (Coull & Nobre, 1998). Temporal attention appears to be dissociable from spatial attention as previous evidence shows that they CHAPTER 1 23 do not interfere with each other (Correa & Nobre, 2008) and their effects are additive (Doherty et al., 2005). While spatial attention is often studied by having participants search for an object scattered at multiple locations, temporal attention is typically studied by having participants search for a target in a rapid serial visual presentation stream. That is, stimuli being presented at a very quick rate in the same spatial location. This is to investigate the rate at which these stimuli can be processed.

While participants are easily able to detect a single target in a stream of images in the absence of any task-related distraction or demands (Potter, 1975; Potter, Wyble, Hagmann, &

McCourt, 2014), if the participant must detect two or more stimuli, the limits of temporal attention emerge. This has been studied using a task called the attentional blink (Chun &

Potter, 1995; Raymond, Shapiro, & Arnell, 1992), in which accuracy of the first target is preserved, but if the second target appears within half a second of the first target, reporting accuracy is diminished. These findings suggest that similar to how spatial attentional mechanisms limit the number of items in the environment one can process, temporal attention limits how quickly one can process stimuli in rapid succession.

In sum, external attention investigates how our attentional mechanisms deal with the outside world in both space and time.

Internal Attention. In contrast with external attention, internal attention refers to any attentional process that does not pertain to external stimuli e.g. the number of objects in preserved in working memory. The taxonomy of attention processes laid out by Chun and colleagues distinguishes three different domains of internal attention: response and task selection, long term memory, as well as working memory. As the attentional bias literature 24 INTRODUCTION AND OVERVIEW focuses primarily on external attention, specifically overt spatial attention, the following sections will be brief.

Response and task selection. One form of response and task selection is the inhibition of alternative responses (Aron, Robbins, & Poldrack, 2004; Nee, Wager, & Jonides, 2007). For example, in the Stroop task, participants must call out the ink colour of the word, which may be different from the written colour word (e.g. “RED” in the colour blue; Stroop, 1935). To do the task, there must be an inhibition of the response, “RED”, overriding the automatic process of reading.

Long term memory. Another form of internal attention is long term memory. In particular, attentional mechanisms determine which information is encoded into long-term memory and which ones are retrieved (e.g. Chun & Turk-Brown, 2007; Yi & Chun, 2005). Past evidence has shown that the attentional processes governing external attention and encoding into long-term memory may be distinct (See Yuen et al., in prep; (Durand, Isaac, & Januel,

2019)) and certain forms of anxiety are associated with greater memory encoding for negative information (Durand et al., 2019). Memory retrieval requires selection between specific memories competing for recall (Badre et al., 2005; Ranganath et al., 2000), and biased memory retrieval of negative and self-focused information is considered a core characteristic of depression (J. M. Smith & Alloy, 2010).

Working memory. Working memory lies at the intersection of external and internal memory, because it follows patterns and principles from external attention and is able to function in the absence of external stimuli. That is, purely on the basis of an internal representation of these stimuli. Past studies have shown that maintenance of information in CHAPTER 1 25 working memory biases external memory. For example, maintenance of certain spatial locations in working memory biases spatial attention towards those locations (Awh & Jonides,

2001; Corbetta & Shulman, 2002). In another study, the same effect is found with specific shapes (Downing, 2000; Soto, Heinke, Humphreys, & Blanco, 2005).

Working memory tasks have been shown to interfere with both internal and external attention. For example, manipulating information in working memory has shown to interfere with visual search (Han & Kim, 2004) as well as generate refractory effects in responding

(Jolicoeur, 1998). However, perceptual attention has also shown to affect what is held in working memory (Lepsien, Griffin, Devlin, & Nobre, 2005).

Summary and interim Discussion. Attention is a core process in almost every part of cognition, as it is the natural consequence of there a limited capacity system in the face of abundant stimuli and possible responses. External attention refers to the selective processing of stimuli in the outside world and amongst them are: spatial attention, which is attention towards a particular region in space, as well as temporal attention, which is the deployment of attention in time. Moreover, attention also exists internally, which regulates decisions, responses, stimuli and representations. Critically, attention seems to operate on multiple levels.

In relation to the study of psychopathology, studies have found its effects on a series of different attentional mechanisms e.g. selective retrieval of long term memory (e.g. Nolen-

Hoeksema, 2008; Litz, Prassas, & Weathers, 1994), spatial attention (Bar-Haim et al., 2007;

MacLeod, Mathews, & Tata, 1986), as well as temporal attention (Kennedy & Most, 2015b;

Olatunji, Ciesielski, Armstrong, Zhao, & Zald, 2011; Onie & Most, 2017). However, the attentional bias literature did, and perhaps still, considers attention as a monolithic construct, 26 INTRODUCTION AND OVERVIEW explicitly stated by Van Bockstaele and colleagues (2014, p 1). However, even just in external attention, past findings have shown that spatial and temporal attention both tap into psychopathology through different routes (Onie & Most, 2017). Therefore, it is important that we consider the underlying mechanisms of attention for the paradigms used.

Measures of Attentional Bias to Negative information

Attention in and of itself is a challenge to measure. Therefore, in measuring attentional bias to negative information, the attentional bias field has developed and used a myriad of tasks. The purpose of this following section is to describe previously used attentional bias measures in the field and discuss the findings from these tasks.

Note that the attentional bias field focuses primarily on external attention, which may be due to the fact that while selection of memory traces from long-term memory may be at the heart of emotional disorders (e.g. rumination and worry), the methodology to assess it is lacking. External attention, while seemingly far removed, is simpler to assess, interpret and test.

One common thread amongst all the negative attentional bias tasks is that emotion modulates performance. For example, in one task, a negative image in a stream of images impairs detection of a subsequent target more than does a neutral image in the stream. As elaborated above, findings in the visual cognition literature point towards a multi-component theory of attention (Chun et al., 2011; Lavie, Hirst, De Fockert, & Viding, 2004). Therefore, it is likely that an attentional bias may link to individual differences in different ways depending on what task is being used to index attentional bias (cf. Cisler & Koster, 2010).

One of the earliest techniques of measuring attentional bias is the modified Stroop Task

(Mathews & MacLeod, 1985). In this task, participants are presented with neutral and negative CHAPTER 1 27 words in different colours (e.g. ‘death’ in green, and ‘chair’ in yellow), and asked to call out the colour of the word. Increased response times on trials with a negative word relative to trials with neutral words are believed to signal an attentional bias for negative information. For a schematic of the task, see Figure 1 below.

Neutral Trials Emotional Trials

Table Death Path Kill

Trials Chair Trials Shame Street Murder

Figure 1. Schematic Representation of the Emotional Stroop Task. Words are presented one at a time and participants must call out the ink colour of the word presented. A common version of this task is to have neutral and emotional blocks and compare response times of the two blocks.

Many studies have used the modified Stroop task in anxious or clinical populations (see.

Bar-Haim et al., 2007). For example, McNally and colleagues (1990) found that Vietnam combat veterans had longer response times to PTSD-related words relative to OCD-related words, positive, and neutral words. Notably, a meta-analysis conducted by Phaf and Kan (2007) found that the emotional stroop effect was only present in high anxious individuals and was not robust within a non-clinical population.

In the context of this thesis which is to discuss the benefit of isolating different mechanisms of attention, the emotional Stroop is at a disadvantage, which is that it is 28 INTRODUCTION AND OVERVIEW extremely difficult to isolate what attentional mechanisms are being engaged in this task (Phaf

& Kan, 2007). Furthermore, the Stroop does not allow direct comparison of an emotional and neutral stimulus. Other measures of attentional bias have been made since then that have more clearly tapped into spatial or temporal attention. Below, we discuss some of these tasks, as well as some of the findings that have emerged from using these tasks.

Spatial Attention Tasks.

Dot Probe. To enable direct comparison of emotional vs neutral stimuli, the Dot Probe task was developed (MacLeod et al., 1986). In a typical Dot Probe trial, a negative and neutral stimulus (words, images, or faces) appear, spatially separated, on a computer screen. The stimuli then disappear to reveal a target probe in one of two locations, to which the participant must respond. Faster responses to the probe on congruent trials are believed to reflect preferential allocation of spatial attention towards the negative stimuli. The authors found that individuals who were clinically anxious responded quicker when the probe was behind the negative stimuli, suggesting preferential allocation of attention to the negative information.

This finding has been replicated numerous times and has become a gold standard for the cognitive experimental approach to psychopathology (Bar-Haim et al., 2007).

In response to a concern that participants could complete the task by fixating on the centre of the screen a “probe identification task” was developed, in which participants must not only indicate the location of the probe, but also a given characteristic of the probe (e.g. whether the probe is horizontally (. .) or vertically (:) aligned; see. Imhoff, Lange, & Germar,

2019, for location vs identification spatial tasks). For a schematic of the trial, see Figure 2.

CHAPTER 1 29

100 ms : 500 ms

Until Response

Figure 2. Schematic Representation of a Dot Probe trial. Following a fixation cross, two images appear on opposite sides of the screen, and are quickly replaced with a probe. This schematic is a congruent trial as the probe appears behind the negative stimulus. In the probe identification version of the task, participants must identify the identity of the probe using key press.

Visual Search. Similar to the Dot Probe, the Visual Search task (Öhman, Flykt, & Esteves,

2001; Rinck, Becker, Kellermann, & Roth, 2003) is another experimental paradigm that indexes spatial attention, as it involves the shifting the spotlight of attention in search of a target. As the name implies, the visual search task involves the participant searching for a target stimulus

(words, images or faces) within a grid of similar stimuli. For example, searching for the word

“spider” within a 3 x 3 grid of neutral words. Similar to the Dot Probe, the emotionality of the stimulus seems to facilitate selection, evident by speeded responses when the target is a negative word. For an example of a visual search trial, see Figure 3 below.

Using the visual search task Öhman and colleagues (2001) found that participants were faster to find images of snakes and spiders as opposed to flowers or mushrooms, Eastwood and 30 INTRODUCTION AND OVERVIEW

Smilek (2005) found that participants high in social anxiety or with panic disorder, but not with obsessive-compulsive disorder, were faster in finding negative stimuli amongst neutral stimuli.

Taken together, these findings suggest that emotion is able to facilitate visual search, and that this process may be specific to certain disorders.

Figure 3. Schematic of a Visual Search Task Trial. Participants were instructed to respond with the left arrow key if all the images came from the same category, or the right arrow if there was an odd one out. Responses were faster if the target was either a spider or snake, as opposed to a mushroom or flower. Note, the authors did not provide an illustration, this illustration is derived from descriptions in the paper. This schematic is based off a description in Öhman and colleagues (2001).

Emotional Spatial Cueing Paradigm. The Emotional Spatial Cueing Paradigm (Fox,

Russo, Bowles, & Dutton, 2001) was developed to distinguish between facilitated attention towards, and difficulty disengaging from the negative stimuli. In a typical trial, a fixation cross briefly appears at the centre of the screen, followed by a neutral or negative stimulus (often referred to as a cue in this paradigm) either on the left or right side of the screen. The cue then vanishes, and a probe appears either near the location of the cue, or on the opposite side of the screen. Participants are instructed to respond to the location of the probe, or in a variant, the CHAPTER 1 31 type of probe (Yiend & Mathews, 2001). For an example of a trial, see Figure 4 below. Fox et al.,

(2001) found no difference between anxious and non-anxious participants on valid trials with either negative or neutral stimuli. However, anxious participants were slower to respond to invalid trials (that is when the probe appeared on the opposite side of the screen) when the cue was negative relative to when the cue was neutral. Taken together, this seems to suggest that anxiety is characterized by difficulty disengaging from negative information rather than facilitated engagement to it. However, as discussed in the next task, there is evidence that both of these processes uniquely relate to anxiety.

+ fail + 1000 ms fail + 100 ms Cue + 50 ms

Until Response

Figure 4. Schematic Representation of the Emotional Spatial Cueing Paradigm Trial. Following the presentation of a neutral or emotional cue, participants must respond to the location of the probe. This schematic is directly based off a figure in Fox et al. (2001).

Attentional Response to Distal versus Proximal Emotional Information Paradigm.

Similar to the Emotional Spatial Cueing Paradigm, The Attentional Response to Distal versus

Proximal Emotional Information Paradigm (ARDPEI; Grafton & MacLeod, 2014; Rudaizky,

Basanovic, & MacLeod, 2014) was developed to tease apart engagement with and 32 INTRODUCTION AND OVERVIEW disengagement from emotional information. In the ARDPEI, a cue appears on the left or right of centre on the computer screen, anchoring attention at that particular location. Following that, two stimuli (one negative or neutral image, and one abstract image) appear on the screen, with one of the stimuli replacing the cue, thereby anchoring attention to that specific stimulus. The probe then replaces one of the stimuli, similar to the Dot Probe. For an example trial, see Figure

From this task, there are four different trial types derived from using either neutral or negative stimuli, and whether the stimuli appear at the location, or away from the anchoring cue. A negative engagement index is obtained by calculating how much more quickly participants respond when attention is anchored away from the location of the negative stimuli than when attention is anchored on it. A neutral engagement index is calculated in a similar fashion, and a difference score between the negative and neutral engagement index indicates whether there is facilitated engagement for negative relative to neutral information. A negative disengagement bias is obtained by calculating how much more slowly participants respond when attention is anchored at the location of the negative stimuli than if it anchored away from it. A neutral disengagement index is calculated in a similar fashion, and a difference scores between the negative and neutral disengagement indices reveals of how much slower participants are at disengaging from negative stimuli are relative to neutral stimuli. Past findings have found indices of engagement and disengagement to threat, to be uncorrelated – further building the case that these may be unique processes (Rudaizky et al., 2014).

Past studies have found that higher trait anxiety is associated with both facilitated engagement to threat cues, as well as difficulty disengaging from them (Rudaizky et al., 2014). CHAPTER 1 33

Furthermore, individuals high on trait rumination demonstrate difficulty disengaging while showing no engagement effects.

+ 1000 ms + 150 ms Cue + 100 or 500 ms

Until Response

Figure 5. Schematic Representation of an ARDPEI trial Attention is anchored to one of two locations by presenting a red box, followed by a discrimination task within the box to ensure attention was anchored to that location. Following that two images are briefly presented: a neutral/negative image and an abstract image and followed by a probe in one of two locations. This schematic is directly based off a figure in Grafton and MacLeod (2015).

Temporal Attention Tasks.

Emotional Attentional Blink. Distinct from spatial attention tasks, the emotional attentional blink (Anderson, 2005; Ihssen & Keil, 2009; Schwabe et al., 2011; Schwabe & Wolf,

2010) indexes temporal attention.

In a recent iteration of the task by Sigurjónsdóttir, Sigurardóttir, Björnsson, and

Kristjánsson (2015), participants viewed a rapid sequence of photos of faces at a rate of 67 ms with a 40ms black screen in between. Two target images were embedded into the stream. The first target image (T1) was a photo of a face with a dot on either the left or right cheek, and the 34 INTRODUCTION AND OVERVIEW second target image (T2) is overlaid with a green tint. Participants were instructed to indicate using keypress whether the dot appeared on the left or right cheek on the first target and the gender of the second target. Either T1 or T2 can have a threatening expression. Consistent with past studies, Sigurjónsdóttir and colleagues found that emotion modulated the accuracy of the targets of either T1 or T2, in which threatening faces in either T1 or T2 increase accuracy in reporting that particular target. For a schematic of this version of the emotional attentional blink see Figure 6 below. In the same study, the authors also used the Dot Probe, Spatial Cueing

Paradigm, and Singleton Cueing Task, and found that only the emotional attentional blink was sensitive to emotion. Note that all the other tasks indexed spatial attention.

Target 1

Target 2 .

67ms / stimulus 40ms / blank screen

Figure 6. Schematic Representation of the Emotional Attentional Blink. In this example, target 1 Is emotional while target 2 is neutral. Photos of faces were presented at a rate of 67 ms with blank screens between each image presented for 40 ms. After the stream had ended, participants used arrow keys to indicate whether the probe was on the left or right cheek of the target, and identified the photograph of the person tinted green from an array. In this task, CHAPTER 1 35 either or both targets could be emotional. This schematic is adapted off a figure in Sigurjónsdóttir and colleagues (2015).

Emotion-Induced Blindness. The emotion-induced blindness task (Most, Chun, Widders,

& Zald, 2005) allows us to investigate the impact of a task-irrelevant distractor in a temporal setting. In a typical trial, people view a rapid sequence of landscape and architectural images

(100-ms/item) and report the orientation of a single, similar image that is rotated 90-degrees clockwise or counterclockwise; accuracy is robustly impaired when the target is preceded in the stream by an emotional distractor than by a neutral distractor (Most et al., 2005). Similar to the attentional blink, the effect is strongest when the stimuli and target are closest in time. For an example of an emotion-induced blindness trial, see Figure 7 below.

There have been a number of studies demonstrating the relationship between emotion- induced blindness and various forms of psychopathology (see. Mchugo, Olatunji, & Zald, 2013).

In one study, Smith, Most Newsome and Zald (2006) found that images associated with an aversive noise caused greater impairment then similarly valenced images that were not associated with an aversive noise. Furthermore, in another study, Olatunji, Armstrong, McHugo and Zald (2013)found combat-related stimuli elicited a stronger impairment for veterans with

PTSD relative to veterans without PTSD and healthy controls, suggesting that the emotion- induced blindness is sensitive to stimulus valence as well as anxiety. The emotion-induced blindness will play a central role in this thesis, and we will revisit it in chapter three.

36 INTRODUCTION AND OVERVIEW

Distractor

Target

100ms/ image

Figure 7. Schematic Representation of an Emotion-Induced Blindness Trial. In this trial, a negative distractor is used, and appears two images (lag 2) before the target. After the stream, participants indicate the rotation of the target with a button press. Note that in real experiments, the negative images used are much more intense e.g. mutilated bodies and soiled toilets.

Interim Summary. In summary, the attentional bias field has provided evidence for the links between heightened anxiety and an attentional bias towards negative information using an array of tasks. Furthermore, in some tasks the emotional stimuli are task relevant (e.g. Visual

Search paradigm, Attentional Blink) and in other tasks, the emotional stimuli are task irrelevant

(e.g. Dot Probe, emotion-induced blindness). This is an important distinction as delineating between voluntary and involuntary processes has become an important point in clinical research (e.g. rumination; Smith & Alloy, 2010).

The tasks themselves also seem to tap into different mechanisms of attention. For example, the Dot Probe and Visual Search Task seem to tap into spatial attention, and both emotion-induced blindness and the emotional attentional blink tap into temporal attention.

Despite the buffet of attention mechanisms that can be tapped into via tasks available to researchers, a majority of the research done is using spatial attention tasks. CHAPTER 1 37

Attentional Bias Modification

Thus far we have discussed different mechanisms of attention and the various tasks used to measure attentional dysfunction in psychopathology. However, beyond measuring these biases and finding that clinically anxious individuals exhibit a heightened bias for negative information, past studies have also demonstrated that attentional bias towards threat can be modified with a consequent impact on emotional vulnerability, suggesting a potentially causal impact of attentional bias on anxiety (Mathews & MacLeod, 2002; See, MacLeod, & Bridle,

2009). Most commonly, these studies have used the Dot Probe and have incorporated a contingency whereby the probe consistently replaces the neutral stimulus, thus training participants always to attend away from threat. Following the training session, a lab stressor is applied (e.g. spontaneously having to prepare a recorded speech) and post-manipulation stress is recorded. Participants who were trained to attend away from threat have been reported to exhibit less emotional reactivity towards the stressor (Amir, Beard, Burns, & Bomyea, 2009;

Amir, Beard, Taylor, et al., 2009; Schmidt, Richey, Buckner, & Timpano, 2009).

These findings suggest a causal role of attentional bias in the development and maintenance of anxiety disorders, and training individuals to attend away from threat could be beneficial in that it could ameliorate anxiety symptoms. These findings not only demonstrate the potentially causal role of attentional bias in anxiety, but also suggest the possibility of an effective and easily disseminated treatment that can be uploaded to people’s personal devices

(Bar-Haim, 2010; MacLeod & Mathews, 2012). 38 INTRODUCTION AND OVERVIEW

Reviewing Past Meta-Analyses

Since its conception in mid 80s, the attentional bias field has garnered much attention, resulting in a substantial number of studies being done on attentional bias assessment and retraining. In this section, I review several key meta-analyses and their findings to evaluate the current state of the field. Specifically, I assess the evidence for three main claims: a) that anxiety is characterised by an attentional bias towards threat, b) that this attentional bias is malleable, c) that modifying attention away from threat has anxiolytic benefits. There are few areas worth highlighting: first, the majority of these studies employ the Dot Probe task for attentional bias assessment and retraining, meaning that they are largely limited to probing and manipulating spatial attention. Second, the effect size indices entered into the meta-analyses differ from one meta-analysis to another, most commonly taking the form of Hedge’s g and

Cohen’s d. However, with large samples, such as in a meta-analysis, Cohen’s d and Hedge’s g can be considered equivalent (Grissom & Kim, 2014). Furthermore, other effect size measures have been used, but have been converted to Cohen’s d for ease of interpretability. (see

Borenstein, Hedges, Higgins, & Rothstein, 2009). Finally, a number of meta-analyses employ a fail-safe n, which is how many studies with null results required in order to render the overall effect non-significant (Rosenthal, 1979).

Attentional Bias and Anxiety. In this section, I review the meta-analyses in investigating the relationship between emotional disorders and attentional bias.

Bar-Haim and colleagues (2007) conducted one of the earliest meta-analysis and found that across several tasks, an attentional bias to threat was present in clinical individuals, but not healthy controls. The meta-analysis consisted of 172 studies with 4,031 individuals (2,263 CHAPTER 1 39 anxious and 1,768 non-anxious). The analysis included three paradigms: the emotional Stroop task (Mathews & MacLeod, 1985), Emotional Cueing task (Fox et al., 2001), and the Dot Probe.

The overall analysis revealed a significant bias towards threat in anxious individuals (d = 0.45) but not in non-anxious individuals (d = -0.007). Furthermore, attentional bias towards threat was present in both clinically diagnosed (d = 0.45), and non-clinically diagnosed but high self- reported anxious populations (d = 46). When only looking at spatial attentional tasks, both the

Dot Probe and Cueing task indicate an attentional bias in anxious individuals (Dot Probe: d =

0.37; Posner Cueing Task: d = 0.43), but not in non-anxious individuals, (Dot Probe: d = -0.09;

Cueing Task: d = -0.11). Overall, these findings are consistent with the notion that anxiety is characterised by spatial attention towards threat.

Stepping back for a broader view of psychopathology, a meta-analysis by (Peckham,

McHugh, & Otto, 2010) analysed 29 empirical studies investigating attentional bias in depression, finding that an attentional bias to threat was also present in depression. Similar to what studies have observed in anxiety populations, the analysis yielded a significant difference in attentional bias between depressed and non-depressed individuals, with a relatively similar effect size for anxiety as found in Bar-Haim et al. (2007; d = 0.52). This finding is consistent with the notion that a negative attentional bias characterises a wide range of psychopathology

(Amir, Beard, Taylor, et al., 2009; Bar-Haim et al., 2007; Chapman & Martin, 2011; Hertel &

Mathews, 2011).

However, Krujit et al. (2019) conducted a meta-analysis using pre-training AB scores in

ABM RCTs consisting of clinically anxious individuals and found evidence that clinically anxious individuals showed no attentional bias in the Dot Probe. This is in contrast with the findings of 40 INTRODUCTION AND OVERVIEW

Bar-Haim and colleagues (2007) that found a significant relationship between an attentional bias to threat and anxiety. The meta-analysis consisted of 13 randomised controlled trials

(RCTs) and 1005 clinically anxious individuals. Kruijt and colleagues found that there was moderate evidence against a pre-existing bias towards threat in clinically anxious individuals

(BF01 = 4.35), leading to the conclusion that anxiety is not characterised by a negative spatial attentional bias. A critical strength of this meta-analysis is that authors used participants from attentional bias modification RCTs, in which attentional bias towards threat before training was not a key outcome. Therefore, the possibility of publication bias is much lower, due to the outcome of interest of the meta-analysis differing from the outcome of interest in publishing these papers.

The meta-analyses that investigated the relationship between a spatial negative attentional bias and psychopathology (i.e., anxiety and depression) provide mixed evidence for the existence of such a relationship. However, only Kruijt et al. (2019)’s meta-analysis used

Bayesian analyses, which allowed the authors to quantify support for the null hypothesis (i.e., no association between negative attention bias and psychopathology dysfunction). However, the Bayes Factor was only mildly moderate, which does not entirely exclude the possibility that the alternative hypothesis (i.e., association between negative attention bias and psychopathology dysfunction) may be true. In sum, there seems to be mixed evidence for the link between spatial attention biases to emotional stimuli and anxiety, with more recent meta- analysis (A.-W. Kruijt et al., 2019) providing slightly more convincing evidence due to the use of

Bayesian analysis, and a more effective method of overcoming publication bias. CHAPTER 1 41

Attentional Bias Modification and Subsequent Impact on Anxiety. In this section I evaluate some key meta-analyses in the attentional bias modification literature. A vast majority of the studies included in all the attentional bias modification meta-analyses used the Dot

Probe to both assess and retrain attention. Collectively, the meta-analyses provide overwhelming evidence that attentional bias modification was able to successfully modify bias, and that clinical benefits reliably emerged following successful attentional retraining. However, as noted below, more recent attempts at retraining have failed to modify attention.

In an early meta-analysis, Hakamata et al (2010) found that attentional retraining was able to modify attention, and that training attention away from threat had subsequent anxiolytic benefits. The meta-analysis evaluated 12 RCTs consisting of 467 participants, investigating the malleability of negative attentional bias and any subsequent anxiolytic effects.

The analysis revealed that Attention Bias Modification (ABM) had a significant anxiolytic effect on state anxiety (d = .61, fail-safe n = 54), with greater effect when using word rather than face stimuli (d = 1.29). When outliers were removed, the effect size was reduced, but the results still remained statistically significant (d = .36, fail-safe n = 7). In addition, there was a marginal correlation between the magnitude of negative attentional bias change and change in anxiety

(p = 0.52).

In a subsequent meta-analysis, Hallion and Ruscio (2011) investigated the effects of

Cognitive Bias Modification (CBM, an umbrella term that subsumes ABM and treatments that focus on interpretation biases) on anxiety and depression. Consistent with the findings of

Hakamata and colleagues (2010), Hallion and Ruscio found that CBM (including ABM) was able to reduce anxiety. However, this effect only emerged following a stressor, a finding which 42 INTRODUCTION AND OVERVIEW emerges repeatedly in the following meta-analyses. This meta-analysis consisted of 45 studies with 2,591 participants. In an attempt to reduce the effects of publication bias, the authors invited doctoral dissertations and unpublished studies to be considered for inclusion. The analysis revealed a significant effect of ABM on attentional bias change (g = 0.29). CBM also ameliorated anxiety, but only after a stressor was applied (g = 0.23). While the authors did not perform separate analyses for the anxiolytic effects of CBM and ABM, a subsequent analysis found no differential association between type of CBM and change in anxiety.

Beard, Sawyer and Hofmann (2012) conducted a meta-analysis and found that ABM reduced both anxiety and depressive symptoms, but only after a stressor. This meta-analysis consisted of 37 studies and 2,135 participants investigating the effects of ABM on a wide range of outcomes, including anxiety, depression, urge to consume alcohol, or urge to smoke. The study found that averaging across post-training, post-stressor and follow-up, there was significant impact of ABM on both anxiety (g = 0.68) and depression (g = 1.54). Moderator analyses also found that more sessions resulted in larger effect sizes (β = - 0.176). Furthermore, outcomes post-training yielded nonsignificant findings (g = 0.03), while outcomes post-stressor were significant (g = 0.40). For multi-session studies, a significant change in outcome was driven by individuals with high symptomology (g = 0.51) rather than healthy samples (g = - 0.15).

Mogoaşe, David and Koster (2014) investigated the effects of ABM on both anxiety and depression, finding that ABM was able to significantly modify attention, and although there was a reduction in anxiety and depression symptoms right after training, clinical benefits increased following a stressor. Consistent with previous meta-analyses (e.g. Hallion & Ruscio, 2011; Beard et al., 2012), the finding that effects increased post-stressor suggest that any clinical benefits CHAPTER 1 43 may be due to reducing reactivity to a stressor. The meta-analysis included 2,268 participants, the largest ABM meta-analysis to date. They also included a number of negative findings which were not included in previous meta-analysis (e.g., Boettcher, Berger, & Renneberg, 2012;

Carlbring et al., 2012; Julian, Beard, Schmidt, Powers, & Smits, 2012). Note, all effect sizes reported here were computed after removing outliers. The analysis revealed that ABM significantly modified negative attentional bias (g = 0.312), however, this change did not persist at follow up (g = 0.553). Furthermore, ABM significantly reduced anxiety and depression symptomology post-training (g = 0.196), however, this effect increased post-stressor (g = 0.375) but was no longer significant at follow-up (g = 0.553). The change in overall symptomology post-training was primarily driven by anxiety studies (g = 0.260) and studies conducted in healthy participants (g = 0.211). The latter finding is in contrast with Beard et al. (2012) who found that ABM was most effective in individuals with high symptomology. Overall, the results of this meta-analysis suggest that negative attentional bias is malleable, and consistent with

Hallion and Ruscio (2011)’s findings, ABM reduces reactivity to a stressor.

Linetzky, Pergamin-Hight, Pine and Bar-Haim (2015) performed a meta-analysis, finding that following ABM treatment, significantly less patients met formal diagnostic criteria for anxiety. However, ABM was only successful in the lab, suggesting the need for task improvement, focusing on increased engagement. The meta-analysis included 11 randomised control trials and 589 clinically anxious patients. The analysis revealed that clinicians observed greater reductions in patients’ anxiety after ABM training (d = 0.42), despite a nonsignificant change in patients’ self-reported anxiety. Following ABM training, significantly less patients in the treatment group met formal diagnostic criteria for their respective anxiety disorder relative 44 INTRODUCTION AND OVERVIEW to control group. Finally, ABM delivered in the clinic was successful (d = 0.34), whilst ABM delivered at home was not (d = -0.10). The meta-analysis concluded that ABM significantly reduced anxiety, however, better methods were required to retrain attention to improve participant engagement at home.

Price and colleagues (2016) found that ABM not only reliably modified bias, but also reduced diagnostic remission. In this meta-analysis, Price and colleagues investigated the effectiveness of ABM on modifying attentional biases, measures of social anxiety, and diagnostic remission. The meta-analysis consisted of 13 studies and 778 participants and found that ABM significantly modified attentional biases (β* = -.63), and, while there was no significant impact of ABM on self-report measures of social anxiety, ABM significantly reduced diagnostic remission by approximately 11.8% (d = 0.23). Exploratory moderation analyses also found that a higher pre-existing bias predicted better response to ABM for studies done in the lab (β* = -.29) and with patients that were clinician-rated rather than self-report (β* = -.58).

Critically, this meta-analysis found that ABM was able to increase diagnostic remission, evidence that ABM has real-world clinical benefits. Together, the findings of Linetzky et al.

(2015) and Price et al. (2016) suggest that ABM may have real world benefits.

Cristea, Mogoaşe, David and Cuijpers (2015) performed a meta-analysis investigating the efficacy of CBM (including ABM) in children and adolescents, finding that although ABM was able to modify attention, no clinical benefits were observed. However, due to the exclusion criteria set by the authors, no studies with post-training stressors were included. This meta- analysis consists of 23 RCTs. In contrast with past findings, the authors found that while ABM successfully modified negative attention bias (g = 0.53), the intervention did not have any CHAPTER 1 45 subsequent impact on any clinical outcomes (g = 0.02). However, as mentioned, the authors did not include post-stressor or follow up measurements. This limitation is similar to the other meta-analysis conducted by Cristea, Kok and Cuijpers, (2015), which was performed on 49 RCTs that investigated the effects of CBM on anxiety and depressive symptoms in adults. In this meta-analysis, only scores post-training were considered, excluding measures obtained post- stressor or at follow up. Overall, after removing outliers, CBM did not significantly reduce anxiety symptoms (g = 0.16) or depressive symptoms (g = 0.22). Despite not performing separate analyses for ABM and CBM, CBM procedures significantly produced larger effect sizes than ABM for both anxiety (p =.034) and depressive symptoms (p =.014). Consistent with

Linetzky et al., (2015), the meta-analysis of Cristea, Kok and Cuijpers(2015) found that any clinical benefits were larger if the intervention was delivered in the lab (p = -.030). The authors concluded that there may be very little to no clinical benefits of CBM; however, since they did not include studies with a post-training stressor, the outcomes are not directly comparable with other meta-analyses which did find that clinical benefits emerged following a stressor.

Thus far, the collective meta-analyses provide substantial evidence that a) ABM is able to modify attention and b) successful modification of attention has subsequent clinical benefits post-stressor, but these effects are strongest in the lab. However, more recent studies been surprisingly reported more failures to retrain negative attentional biases using traditional training methods (Clarke et al., 2017; Everaert, Mogoaşe, David, & Koster, 2015; Notebaert,

Clarke, Grafton, & Macleod, 2015; Yeung & Sharpe, 2019). To our knowledge, these studies have not been included in any meta-analyses noted above. 46 INTRODUCTION AND OVERVIEW

In response to this, a review by MacLeod and Clarke (2015) noted that it was important to delineate between success in retraining attention using attentional bias modification paradigms, and its consequential effect on anxiety vulnerability. The authors found that in studies where selective attention was successfully modified away from threat, the training group typically showed reduced anxiety vulnerability to a stressor following training.

Furthermore, when negative attentional bias change was not achieved, there was no reliable change in anxiety symptoms. Although their conclusion was not statistically tested, more compelling support for the importance of a successful attention bias change in the ABM paradigm was provided by Grafton et al. (2017), in which the authors controlled for successful attentional bias change in their meta-analysis and found that it significantly moderated the subsequent emotional reactivity after the stressor tasks. Therefore, current understanding seems to suggest that a successful modification of attention leads to reduction in anxiety, but the field’s currently available methods require further improvement in the service of more consistent outcomes (MacLeod & Clarke, 2015; C. MacLeod, Grafton, & Notebaert, 2019).

Altogether, these findings provide somewhat disjointed evidence. How does modifying attention away from threat reliably reduce anxiety symptoms, while the relationship between a negative attentional bias and anxiety is tenuous? Furthermore, although there is abundant evidence to suggest that traditional attentional retraining methods can successfully modify attention (e.g. Mogoaşe et al., 2014; Hakamata et al., 2010), more recent findings have failed to do this using conventional methods (e.g. Clarke et al., 2017; Notebaert et al., 2015). While there is some evidence that ABM is most effective with participants with a pre-existing negative attentional bias (Price et al., 2016) and high symptomology (Beard et al., 2012; Dennison, 2018), CHAPTER 1 47 there is also evidence to suggest ABM is most effective in healthy participants (Mogoaşe et al.,

2014). A common scientific practice is to derive models of the current findings, but for reasons discussed next, this may not be optimal.

Possible Publication Bias in the Literature

Even meta-analyses, which combine results across studies, cannot circumvent all problems with the literature. For example, incentives built into the publication process lead to what is known as a publication bias, which refers to the increased likelihood of studies being published if positive findings are found. As a result of publication bias, the literature and associated meta-analyses are at risk of arriving at inflated conclusions about the size (or even presence) of effects. For example, (Shanks, Vadillo, Riedel, Clymo, Govind, Hickin, Tamman, &

Puhlmann, 2015) conducted a meta-analysis investigating whether mating motives influence consumer choices and risk-taking, yielding a significant effect (d = 0.57). The authors then ran a series of studies (including a fully pre-registered study) and compared their own unbiased effect size estimates with the meta-analysis effect size. The mean effect sizes differed substantially (d = 0.58), and as a result the conclusions significantly changed. While almost all meta-analyses in the attentional bias and ABM field attempt to control for publication bias, the degree to which such attempts have been successful can’t be known without comparing them to data collected in the absence of such publication incentives.

Low powered designs may also contribute to a biased literature. Kruijt et al. (2018) noted that studies investigating the relationship between negative attention bias and anxiety are underpowered when using the effect size reported in Bar-Haim et al. (2007) as reference. A recent methods paper discussing the issue of power in psychology found that average effect 48 INTRODUCTION AND OVERVIEW sizes in psychology may actually be smaller than initially understood (Brysbaert, 2019).

Furthermore, using simulations, the aforementioned study by Brysbaert (2019) suggested that a research field that consists of low powered studies is characterised by significant findings of varying directionality, even if there is no true effect. Therefore, one possible cause for the mixed findings in the meta-analyses above is due to the meta-analyses consisting of low- powered studies.

One solution is to apply bias correction methods as ABM analyses have done (e.g.

Hallion & Ruscio, 2011). Examples of these correction methods are Bayesian Bias Correction

(Guan & Vandekerckhove, 2016), trim-and-fill (Duval & Tweedie, 2000)curve and P- Curve

(Simonsohn, Nelson, & Simmons, 2014). All of the aforementioned methods attempt to estimate the effect of publication bias on effect sizes and/or p-values and fill out the missing data using mathematical models. However, one concern is that these methods often substitute non-existent data with data generated by a model rather than data collected in the lab.

Therefore, the outcomes should be interpreted with particular caution. Furthermore, a recent conference paper presented findings where the authors applied the Bayesian Bias Correction and P-Curve to the effect sizes generated by the meta-analysis in Shanks et al. (2015)and compared the output to the unbiased estimate. The analysis revealed that despite generating effect size estimates closer to the unbiased measure, none of the existing bias correction methods could fully correct for the bias, with a difference of d > 0.40 (Etz & Vandekerckhove,

2017), equivalent to the average effect size found in psychology (Brysbaert, 2019). Therefore, at this present time, it seems that bias correction methods cannot correct for bias in the literature. CHAPTER 1 49

While one might initially assume that a positively skewed literature may mask a true null effect, it can also – counterintuitively – be the case that a positively skewed literature masks real and reliable effects. An initial underestimate of the power needed to detect this effect, combined with publication bias, will result in a literature with inflated effect sizes early on.

However, as publishing null results become more and more important in the field, a number of null results will emerge. These newer findings will cast doubt, despite the fact that these studies are similarly powered to find a large and inflated effect size. Therefore, a positively skewed literature does not preclude the existence of a reliable effect – the studies that make it up may be simply underpowered, and, therefore, yield extremely mixed findings (Brysbaert,

2018).

Due to all the factors discussed above, we may have an inaccurate picture of the attentional bias field. Simply by inspecting past studies, we cannot determine: (a) an unbiased effect size measure of any of the effects; (b) whether these studies have been powered; and (c) whether these findings are reliable or consistent. Therefore, it is necessary to replicate the findings in the literature.

Considering other Mechanisms of Attention

Attentional biases have been measured through a variety of tasks, and although these measures don’t always correlate with each other (Dalgleish et al., 2003; Mogg et al., 2000) researchers sometimes assume that they tap into a common underlying mechanism (e.g., Van

Bockstaele et al., 2014, p. 688). As discussed previously in the introduction, there is evidence against a unitary mechanism of attention (Chun et al., 2011). One study suggests that emotion- 50 INTRODUCTION AND OVERVIEW driven biases that impact these mechanisms link differentially to clinically relevant individual differences (Onie & Most, 2017).

The study by Onie and Most (2017) evaluated the relative degree to which the dot probe and emotion-induced blindness correlated with each other and with individual differences in trait negative affect (i.e., self-reported anxiety and depression) and persistent negative thought

(e.g., worry and rumination). Consistent with the notion that the dot probe and emotion- induced blindness reflect different mechanisms of attentional bias, performances in the two tasks were found not to correlate with each other (Onie & Most, 2017). Furthermore, they accounted for unique variance in negative affect and only emotion-induced blindness predicted persistent negative thought, even though the same emotional images were used in the two tasks.

Such findings contribute to an emerging literature on the potential importance of differentiating between attention mechanisms when investigating attentional bias. Other findings within this literature suggest that different mechanisms of attention may be differentially sensitive to emotional information. For example, Sigurjonsdottir and colleagues

(2015) compared four tasks in their sensitivity to negative faces relative to neutral faces: the dot probe, a spatial cueing task where an emotional or non-emotional face could precede the target at the same or different location (Fox, Russo, Bowles & Dutton, 2001), a search task where a non-target emotional or non-emotional face could appear alongside the search display, and an attentional blink task where an emotional or non-emotional face could appear as the first of two targets within a rapid stream of items. The researchers found that it was the attentional blink, which indexes temporal attention and had no spatial component, that was CHAPTER 1 51 most sensitive to the presence of emotional information (i.e., participants were worse at reporting the second target when it was preceded by an emotional than non-emotional first target). This is consistent with repeated findings that emotion-induced blindness (a temporal- but not spatial- attention effect) is almost invariably robust.

In conclusion, the Dot Probe task has formed the foundation of the attentional bias field, but an evaluation of the literature along with possible publication suggests that replication efforts are needed. Furthermore, depending on the outcome, possible moderating factors or alternative mechanisms for assessing attentional bias may be a potential avenue for investigation.

The Current Study

This thesis investigates the relationship between attentional bias and psychopathology, and its trainability, while considering distinct mechanisms of attention. For this thesis, we employ the Dot Probe due to the large amount of studies conducted using this task, as well as the emotion-induced blindness, to mirror the task-irrelevant nature of the emotional stimuli in the Dot Probe. Using the Dot Probe and emotion-induced blindness, we are able to assess spontaneous attentional capture on both spatial and non-spatial mechanisms of attention respectively. In addition, due to the possible publication bias, replications of key attentional bias findings (e.g. the link between attentional bias and psychopathology) are incorporated into the experiments.

Chapter 2 investigates whether ABM modifies attention through training inhibitory mechanisms, by assessing whether inhibition training is able to elicit similar patterns as the avoid-negative training condition. Furthermore, Chapter 2 also seeks to replicate the Dot Probe 52 INTRODUCTION AND OVERVIEW congruence effect, the link between task performance and psychopathology, as well as traditional attentional retraining conditions (avoid-negative and attend-negative). Chapters 3 to

5 investigate emotion-induced blindness’ reliability and validity as an attentional bias measure, as well as replicate the task’s relationship with measures of psychopathology. Chapter 3 investigates the test-retest reliability of emotion-induced blindness, giving an indication of its stability over time and utility as an individual difference measure. Chapter 4 examines whether stimuli valence or arousal accounts for their impact in emotion-induced blindness and Dot

Probe. Chapter 5 investigates emotion-induced blindness’s locus of competition by modifying the task to include a priming task at the end of the stream. Finally, Chapter 6 investigates the trainability of emotion-induced blindness, an indicator for its utility as an attentional retraining paradigm.

CHAPTER 2 53

Chapter 2:

Investigation of Inhibitory Training on Attentional Bias Change

Adapted from Onie, Sandersan, Notebaert, L., Clarke, P., & Most, S. B. (2019). Investigating the effects of inhibition training on attentional bias change: A simple Bayesian approach. Frontiers in Psychology. https://doi.org/10.3389/fpsyg.2018.02782

The candidate led the design of the study, programming, data collection, data analysis and manuscript preparation. 54 INHIBITORY TRAINING ON ATTENTIONAL BIAS

Investigation of Inhibitory Training on Attentional Bias Change

Due to mixed findings in the literature using the Dot Probe, we first sought to conduct a direct replication of the basic congruence effect of the Dot Probe, in which responses are quicker on congruent trials than incongruent trials, and relationship between the Dot Probe and anxiety, in which individuals with higher levels of anxiety demonstrate a greater negative attentional bias. We also sought to replicate the basic ABM training paradigm, in which attention can be trained towards and away from negative information. In addition, mixed findings in the ABM literature underscore the need to better understand the specific mechanisms contributing to attentional change in order to improve training methods resulting in consistent outcomes.

One mechanism that may contribute to a negative attentional bias is an impairment in inhibitory control. For example, past studies have found that anxious individuals show a deficit in inhibitory control measured using the anti-saccade task (Derakshan, Ansari, Hansard, Shoker,

& Eysenck, 2009; Wieser, Pauli, & Mühlberger, 2009), a task which requires the individual to inhibit reflexive eye-movements towards a neutral stimulus on the screen. Specifically,

Derakshan et al., (2009) showed that high anxious individuals were slower than low anxious individuals in the anti-saccade task, but showed no difference in a pro-saccade task, in which the cue and the target were in the same location. This suggests that anxious individuals did not differ in initiating attention to a neutral stimulus, but rather specifically in inhibiting the allocation of attention to a salient distractor. Models such as the attentional control theory

(Eysenck & Derakshan, 2011; Eysenck, Derakshan, Santos, & Calvo, 2007) have postulated that anxiety-linked deficits in executive control, specifically deficits in the inhibitory function, may CHAPTER 2 55 play a maintaining role in a negative attentional bias (see Heeren, De Raedt, Koster, & Philippot,

2013 for a review).

Therefore, it is possible that ABM procedures achieve their impact on attentional bias by training inhibitory control. Consistent with this notion, previous research has shown that individuals who were trained to attend away from threating stimuli (e.g., words reflecting social threat) showed improved performance in a post-training attentional control task (Chen, Clarke,

Watson, MacLeod, & Guastella, 2015), suggesting that ABM training improved inhibitory attentional control. There was no significant difference between trials using negative, neutral or positive stimuli in the attentional control assessment, suggesting that this increase in inhibitory control was not emotion-specific. Furthermore, an fMRI study showed that ABM increased neural activity in lateral frontal regions, areas found to play a role in inhibitory control

(Browning, Holmes, Murphy, Goodwin, & Harmer, 2010), and another study found that by stimulating the dorsolateral prefrontal cortex (linked to attentional control) attentional bias was more readily modified (Clarke, Browning, Hammond, Notebaert, & MacLeod, 2014). Such findings are again consistent with the notion that one of the mechanisms modified by ABM may be general inhibitory control, and also consistent with recent suggestions that the attentional biases often observed among clinical populations actually reflect a more domain-general inefficiency of attentional control (McNally, 2019).

One potential way to test whether increasing general inhibitory attentional control contributes to reductions in negatively biased attention similar to ABM procedures is to adapt the anti-cueing task into a training task, similar in format to ABM tasks. The anti-cueing task is a behavioural adaptation of the anti-saccade task that uses response time as the dependent 56 INHIBITORY TRAINING ON ATTENTIONAL BIAS variable instead of eye-movements. In the anti-cueing task, a single pre-target cue appears, with the target then appearing in the opposite location (Cain, Prinzmetal, Shimamura, &

Landau, 2014; M. I. Posner, Cohen, & Rafal, 1982). This assessment task can be adapted into a training task to reduce attentional deployment to a non-emotional, but salient, pre-potent stimulus. This might be achieved by presenting multiple trials in which only a single neutral pre- target stimulus appears, with the target then appearing in the opposite location. We hypothesize that by training the individual to consistently orient away from a neutral, pre- potent stimulus, it may be able to train inhibitory control. Because this training condition would not include emotional stimuli, it may better target pure inhibitory control, without a potentially additional role of emotion regulation

Thus, in the current study, we sought to compare the impact of inhibition control training to regular ABM training. Specifically, we assessed whether inhibitory control training produced a similar reduction in attentional bias as a standard avoid-negative condition of the dot probe task. If inhibitory control alone can contribute to attentional bias change in similar fashion as ABM, it is predicted that, relative to an attend-negative ABM condition (where no inhibitory control training is hypothesized to be involved), the inhibitory control training and avoid-negative training conditions should produce similar changes in attentional bias.

An additional aim of the present study was to include analyses to address a limitation of past studies. Specifically, inconsistent outcomes of ABM studies (i.e., observations either of significant change in attentional bias or of no significant change) may stem from lack of power in any given study. A lack of power or precision may lead to either a) not enough power to detect a true change (Type-II error), or b) a lack of power leading to spurious significant findings CHAPTER 2 57

(Type-I error), potentially resulting in inaccurate conclusions in low-powered studies (Button et al., 2013). The low test-retest reliability of typical Dot Probe measures (Schmukle, 2005;

Staugaard, 2009) may also contribute to the issue of power, as simulations demonstrate that low task reliability yields lower power (Kanyongo, Brook, Kyei-Blankson, & Gocmen, 2007).

Therefore, since low power may lead to inaccurate conclusions, better statistical methods are needed to determine whether the study is underpowered.

As part of this study, we compared the use of a Bayesian inference to a frequentist approach (the widely used approach using p-values for inference) in analyzing post-ABM attentional bias. Bayesian approaches are often capable of distinguishing between lack of power/precision and lack of an effect, and they have steadily garnered favor in many corners of the psychological literature (Andrews & Baguley, 2013). This approach starts with a priori beliefs (“priors”) about the direction or size of an effect, which are ultimately combined with observations (evidence) and data to yield updated beliefs (“posteriors”). This can be illustrated with an intuitive example: if you only have superficial familiarity with a colleague who always appears to be carefree, your impression of them might change drastically the first time you encounter them in a depressed state. In contrast, for those who know your colleague better and who have a more well-rounded understanding of them, witnessing the colleague’s depression might not change their impression as much. In short, what we take away from an event (posteriors) depends on our prior expectations (priors), even if the event itself is held consistent (evidence). It is worth noting that if all observers’ perceptions of your colleague were absolutely identical after the depressive episode, the observers would have identical posteriors of the event. However, in most cases, the resulting posteriors will be constrained by the initial 58 INHIBITORY TRAINING ON ATTENTIONAL BIAS priors. Ultimately the heart of Bayesian inference is combining priors with current evidence to yield posteriors.

Bayesian inference takes the form of model comparison, evaluating the degree to which the data support one model over the other (e.g. null vs. alternative), and asks the question:

‘how many times more likely is one model than the other?’ This is represented with a Bayes

Factor, a numeric value that represents how likely the alternative (H1) is relative to the null

(H0). For example, a Bayes factor of 10 suggests that the alternative is 10 times more likely than the null, a Bayes factor of 0.1 suggests that the null is 10 times more likely than the alternative, and a Bayes factor of 1 suggests the null and the alternative are equally likely, with insufficient evidence to suggest one direction over the other. This contrasts with typical p-values, which allow investigators to infer either that there is evidence to support a claim (p < .05) or that there is not enough evidence to support a claim (p > .05), and it is often unclear whether the latter outcome indicates support for the null or a lack of experimental- or statistical- power. In contrast, Bayes factors allow researchers to distinguish between three outcomes: support for the null, support for the alternative, or ambiguous results. This is a useful way to distinguish between the absence of an effect and absence of power. Note that, here, power is defined not as the probability to detect an effect, but rather the precision of estimates stemming from the amount of evidence collected. This is an important feature, as low powered studies have been shown to lead to spurious findings and a lack of reproducibility (Button et al., 2013). It can be argued that this is a more objective measure of power than a priori power analyses, since power is directly driven by evidence in the present data rather than data collected in another study or meta analyses, which may be prone to publication bias. CHAPTER 2 59

In sum, this study had five main aims: to replicate the basic congruence effect, the relationship between the Dot Probe and anxiety, and bias modification techniques, as well as to test whether inhibitory control training produced a similar reduction in attentional bias as a standard avoid-negative condition of ABM, and to demonstrate the benefits of using Bayesian inference in analyzing ABM data. To test this, we will use a single session ABM training design.

However, we acknowledge that there have been mixed findings with single session ABM training paradigms and therefore, there is a possibility that the training may not yield training effects.

Methods

Participants

120 participants (41 males, 79 females, Mage = 22.37, SDage = 4.23) from the general community were recruited through an online system (SONA) and were compensated AUD $15 for their time.

The sample size was determined a priori, drawing from between-subjects effect sizes from past studies (h2 = 0.06 – 0.15; e.g. Clarke et al., 2014; Grafton et al., 2012; Notebaert et al., 2015).

Power analyses performed in Clarke et al. (2017), using a similar repeated measures design, suggested 105 participants would yield 80% power. 60 INHIBITORY TRAINING ON ATTENTIONAL BIAS

This study was carried out in accordance with the recommendations of 'the Human

Research Ethics Advisory Panel at UNSW' with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Human Research Ethics Advisory Panel.

Design

Participants completed a questionnaire indexing negative affect followed by a pre- training Dot Probe measure of attentional bias. Then, the participants were randomly allocated to one of three between-subjects training conditions based on arrival time to the lab (e.g. participants 1, 4 and 7 would be in the same training condition). One training condition presented negative-neutral stimulus pairs, combined with a contingency that encouraged participants to attend towards the negative stimulus (attend-negative condition). One presented negative-neutral stimulus pairs, combined with a contingency that encouraged participants to attend away from the negative stimulus (avoid-negative condition). A third presented a single non-emotional stimulus (rendering it salient), combined with a contingency that encouraged participants to attend away from it (inhibitory control training condition). All participants then engaged in a post-training Dot Probe measure of attentional bias.

The study took the form of a single session training study to assess the impact of inhibitory suppression on attentional bias.

Materials

Hardware. Stimuli were presented on a 24-in. BenQ XL2420T LED monitor with

1920×1080 resolution and 120-Hz refresh rate. Head position was not fixed. CHAPTER 2 61

Questionnaires. The Depression, Anxiety, and Stress Scale 31 (DASS-21; Lovibond &

Lovibond, 1995) was used to investigate whether there were any pre-existing differences in negative affect amongst the training groups. Participants had to indicate how much statements applied to them over the past week e.g. “I found it hard to wind down”, on a four-point scale ranging from ‘Did not apply to me at all – NEVER’, to ‘Applied to me very much, or most of the time – ALMOST ALWAYS’.

Attentional Task Stimuli. For the negative-neutral stimulus pairs, 75 negative and 75 neutral images consisting of images of people and animals were taken from various sources, including web searches and the International Affective Picture System (Lang, Bradley, &

Cuthbert, 1997), all of which were then collectively rated by independent raters on Amazon’s

Mechanical Turk. The negative images often depicted mutilated bodies and disgust images (e.g. soiled toilets). Neutral images often depicted people in everyday scenes as well as household objects. One-hundred and ninety individuals rated these images on two dimensions: valence and arousal.

To rate the valence, we asked the participants ‘How does this image make you feel?’ to which they answered on a scale ranging from -9 (extremely negative) to 9 (extremely positive).

For arousal, we asked the participants ‘how intense is this image?’ to which they responded on a scale ranging from 0 (not at all intense), to 9 (extremely intense).

Likelihood ratio tests confirmed that negative images were rated more negatively than neutral images χ(1) = 14,459, p < .001, and that negative images were rated as more intense than neutral images χ(1) = 12,736, p < .001 (Valence: Mneg = -5.594, SDneg = 1.1512 , Mneut =

1.186, SDneut = 0.853, Arousal: Mneg = 6.155, SDneg = 1.168, Mneut = 1.887, SDneut = 0.264). 62 INHIBITORY TRAINING ON ATTENTIONAL BIAS

For the inhibitory control training condition, images depicting landscapes – absent of depictions of people or animals – were used. We used photos of landscapes rather than simpler stimuli so that naturalistic scenes were employed across all training conditions.

The same stimulus set was used in the pre-training attentional bias assessment task and the attentional training task, using 50 of the 75 images from the neutral and negative image pool each (except for the inhibitory control training condition, which used only landscapes).

Post-training assessment used a different stimulus set (the other 25 images from each of the neutral and negative image pools) to ensure any training effects were associated with the stimulus valence rather than the specific stimuli themselves.

Attentional bias assessment task. During pre-training and post-training assessment participants were instructed to indicate the left/right direction of an arrow probe using directional keys, and to try ignoring the images that would appear on the screen prior to the target arrow.

On each trial, a fixation cross appeared at the center of the screen for 100ms, followed by two 11.5 cm × 8.5 cm images (a neutral and negative image) placed with their medial edges 7.5 cm above and below central fixation (15cm from one image’s edge to another). We used a

500ms exposure time for the images, consistent with past studies (Clarke et al., 2017). The images disappeared to reveal an arrow (3.5 x 3.5 cm) behind one of the images, which pointed either left or right and remained until participants made a response indicating the arrow’s direction. The two images were placed at the top and bottom of the screen rather than left or right as per recommendations to improve task reliability (Price et al., 2015). CHAPTER 2 63

For each assessment phase, participants completed 120 trials with a short break after 60 trials. Trial type (Congruent, in which the probe appeared behind the negative stimuli, vs

Incongruent, in which the probe appeared behind the non-negative stimuli) and Probe type

(arrow pointing left or right) were equally and randomly allocated throughout. In pre-training,

50 negative and 50 neutral images were randomly distributed amongst 120 trials, (each trial having 1 negative and 1 neutral image), resulting in each image being presented 2 – 3 times. In post-training, 25 negative and 25 neutral images were randomly distributed amongst 120 trials, resulting in each image being presented 4 – 5 times.

Training Task. During the training phase, participants completed 720 trials with a short break after every 80 trials.

Trials and instructions were identical to pre-training assessment except: In the attend- negative condition the probe was always behind the negative stimuli, and in the avoid-negative condition the probe was always behind the neutral stimuli. We adopted this contingency based on past studies (Clarke et al., 2017; Grafton, Mackintosh, Vujic, & MacLeod, 2014; Milkins,

Notebaert, MacLeod, & Clarke, 2016). This was to train attention towards and away from negative stimuli respectively. In the inhibitory control training condition, only one pre-probe stimulus was present, and the probe always appeared in the other location.

For avoid-negative and attend-negative training conditions, participants were trained on the 50 images used in the pre-training assessment randomly distributed amongst 720 trials. In the inhibitory control condition, 50 images depicting landscapes were used instead. A schematic of the different training conditions can be found in Figure 8.

64 INHIBITORY TRAINING ON ATTENTIONAL BIAS

Training Conditions

Avoid Negative Attend Negative Inhibitory Control Training

100ms + + +

500ms

Until Response

Arrow always appears behind Arrow always appears behind Arrow always appears in opposite Neutral Stimulus Negative Stimulus Location to image

Figure 8. Schematic of Training Conditions in Experiment 1. Image location (top and bottom), arrow location (top or bottom), as well as arrow direction (left or right) were randomized.

Procedure

Participants were tested in individual testing rooms. All tasks were completed on computers, starting with a demographics questionnaire and the DASS-21 on Qualtrics (a website based survey tool). Next, participants completed the pre-training attentional bias assessment, before proceeding with the attentional training task. They then finished with the post-training attentional bias assessment. Participants were debriefed at the end of the study.

Data availability

Pre-registration of the aims, methods, and analysis plan, data, and analysis output can be found at: https://osf.io/hfr9s/. CHAPTER 2 65

Results

Data preparation

Accuracy for all participants and across congruency was high for pre- and post-training

(MAccuracy = 97.6%, SDAccuracy = 1.5%). All participants met a pre-defined inclusion criterion of

MAccuracy > 80%.

Probe reaction times were prepared by removing all observations faster than 200ms and slower than 2000ms. On average, 2 – 3 trials were removed for each individual. Following that, reaction times outside three standard deviations from each participant’s own mean, separating pre and post training, congruent and incongruent trials to eliminate outliers. Attentional bias indices were then calculated separately for pre- and post- training: aggregate scores for congruent trials were subtracted from aggregate scores for incongruent trials. A positive attentional bias index indicates a quicker response in congruent trials relative to incongruent trials and suggests an attentional bias towards negative stimuli. Five participants were removed from analyses due to pre-training bias scores that were three standard deviations or more from the mean.

Note that past studies have combined pre and post attentional bias scores when eliminating outliers three standard deviations from an individual’s own mean (e.g. Clarke et al.,

2017). Our pattern of results remains the same between both exclusion methods.

Past studies have criticized the use of aggregate bias scores, favoring other indices such as the bias variability score (Iacoviello et al., 2014). However, simulations have found that these indices can fluctuate without an actual attentional bias, and rather reflect response time 66 INHIBITORY TRAINING ON ATTENTIONAL BIAS variability (Kruijt, Field, & Fox, 2016). Also, despite past issues with reliability, aggregate bias indices have been shown to be reliably associated with anxiety (Bar-Haim et al., 2007).

Seven participants did not have DASS-21 data due to a system error in which their data was not recorded, and these participants were not included in analyses involving DASS-21.

Overall, there were 40 participants in the avoid-negative condition, 37 in the attend-negative condition, and 38 in the inhibitory training condition.

Attentional bias analyses

We first report data analyses using a frequentist method and follow this with a Bayesian analysis. In the interest of making these analyses accessible, both analyses were done in JASP, an open source software which includes both frequentist and Bayesian analyses. This point and click software was developed to introduce people to Bayesian analysis using a familiar interface.

Frequentist analysis. First, we investigated whether there were any pre-existing differences between individuals in the attend-negative, avoid-negative, and inhibitory control conditions prior to training. We performed a one-way ANOVA on the pre-training attentional bias index, which showed no evidence for group differences, F(2, 112) = 0.219, p = .804, h2=

0.004 .

Following that, we investigated whether incongruent trials were slower than congruent trials using a paired samples t-test. The results revealed that there no significant difference between the two trial types, t(114) = 2.326, p = 0.989, � = 0.217, 95% CI [- ∞ , 0.372].

Next, we investigated whether there were any pre-existing differences in negative affect between the different groups via a similar one-way ANOVA on the DASS total score, as well as CHAPTER 2 67 the depression, anxiety and stress subscales. The analysis revealed no significant difference between the groups on any of the scales or subscales, DASS: F(2,106) = 0.180, p = .835, h2=

0.003; DASS-D: F(2,106) = 0.018, p = .982, h2 < 0.001; DASS-A: F(2,106) = 0.219, p = .804, h2=

0.004; DASS-S: F(2,106) = 0.586, p = .558, h2= 0.011. See Table 1 for means and standard deviations, and reliability indices.

Table 1 Depression, Anxiety, and Stress Scale 21 Means and Standard Deviations in Chapter 2 DASS Total DASS-D DASS-A DASS-S

Mean 10.70 3.376 3.138 4.183 SD 8.919 3.310 3.081 3.675 Reliability 0.932 (0.928) 0.864 (0.853) 0.810 (0.794) 0.871(0.869) Note: Means and standard deviations of the Depression, Anxiety and Stress scale. Reliability indices are McDonald’s ω, with Cronbach’s α in the brackets.

To test whether there was a significant difference between the impact of the three training conditions on attentional bias, we performed a repeated measures ANOVA with time

(pre and post training) as a within subjects factor and training condition as a between subjects factor. The analysis yielded no main effect of time, F(1,112) = 0.160, p = 0.690, h2 = 0.001, nor a main effect of training condition, F(2,112) = 0.523, p = 0.594, h2= 0.009, nor an interaction between time and condition F(2,112) = 1.151, p = 0.860, h2= 0.003. Therefore, we failed to reject the hypothesis that there was any impact of training on attentional bias. However, note that it is unclear from this analysis whether this reflects the absence of an effect or a lack of power to find an existing effect, as we cannot give support for the null using the frequentist analysis. 68 INHIBITORY TRAINING ON ATTENTIONAL BIAS

To ensure between-software consistency, the same analysis was performed in SPSS, with the exact same outcome. The output can be found in our osf page noted above. See Table

2 for attentional bias means and Table 3 for reliability values.

Table 2 Attentional Bias Indices for three training conditions Training Conditions Pre-Training AB Index Post-Training AB Index Attentional Bias Change Mean

Avoid Negative 1.256 (11.49) -0.484 (13.62) -1.740 (16.24)

Attend Negative -2.918 (8.915) 0.376 (12.20) 3.295 (16.22)

Inhibitory Control -0.623 (12.40) 0.575 (11.74) 1.204 (13.80)

Note: Attentional Bias Indices for three training conditions. All indices are in ms. Standard deviations are in brackets. AB = Attentional Bias.

Table 3 Reliability Indices of Attentional Bias Indices Training Condition Pre - Training Post - Training

Cong Incong AB Index Cong Incong AB Index

Attend Negative 0.59 (0.96) 0.81 (0.97) 0.53 (0.76)

Avoid Negative 0.97 (0.99) 0.5 (0.99) 0.81 (0.98) 0.78 (0.98) 0.51 (0.78) 0.50 (0.81) Inhibition 0.66 (0.95) 0.65 (0.96) 0.44 (0.69) Note: Reliability indices of Dot Probe scores. Values are McDonald’s ω, whilst values in brackets are Cronbach’s α. Cong refers to the congruent trial type, Incong refers to the incongruent trial type. AB index reliability indices are computed by averaging across reliability indices obtained from differences scores from 100 different combinations of within participant incongruent – congruent scores.

Bayesian analysis. Although Bayes Factors are often reported as BF10, which refers to the probability of the alternative model against the null, here we report findings using BF01, which is the probability of the null relative to the alternative hypothesis (the inverse probability of BF10: BF01 = 1/BF10). This is to help interpretability of the Bayes factors for null findings; for example, ‘we are 15 times more likely to find the null’ is more intuitively interpretable than ‘we are 0.067 times more likely to see the alternative hypothesis’. We used Jeffrey’s scale to CHAPTER 2 69 interpret Bayes factors (Jeffreys, 1998), which places labels on Bayes factors (e.g., BF = 1 – 3 is anecdotal evidence, BF = 3 – 10 is moderate evidence and BF = 10 – 30 is strong evidence).

Whilst unlike the frequentist approach there are no strong cutoffs, a Bayes Factor of 10 is often used to indicate compelling evidence (Aczel et al., 2018; Wagenmakers et al., 2015). The following analyses were done using the Jeffreys-Zellner-Siow (JZS) default priors for ANOVA

(Rouder, Morey, Speckman, & Province, 2012). Note that the default priors selected in JASP has been shown to operate well with a wide range of ANOVA designs and provides a good balance between very strong to uncertain priors.

First, we investigated whether there were any pre-existing attentional bias differences between the groups prior to training. We performed a one-way ANOVA investigating group differences and found moderate evidence to suggest there were no pre-existing differences

BF01 = 7.332.

Following that, we investigated whether responses on incongruent trials were slower than congruent trials using a Bayesian paired samples t-test. The results provided strong evidence that the two groups were equivalent, BF01 = 31.26, � = 0.022, 95% CI [-0.081, -0.001].

Next, we investigated whether there were any pre-existing differences in negative affect between the different groups. We performed four one-way ANOVA to see whether different training groups differed in their DASS subscale or total score. The analysis revealed moderate to strong evidence that there were no pre-existing differences between the groups, DASS: BF01 =

10.063, DASS-D: BF01 = 11.494, DASS-A: BF01 = 9.745; DASS-S: BF01 = 7.238.

To test our main hypothesis, we performed a repeated measures ANOVA with time (pre and post training) as a within subjects factor and training condition as a between subjects 70 INHIBITORY TRAINING ON ATTENTIONAL BIAS

factor. The analysis revealed moderate evidence suggesting there was no effect of time BF01 =

6.135, and strong evidence suggesting there was no effect of training condition BF01 = 10.638.

In line with the model comparison nature of Bayesian statistics, to obtain the evidence for the interaction term we divided the Bayes Factor for the model containing the interaction term by the Bayes Factor for the model with only the main effects. The analysis yielded strong evidence suggesting there was no interaction BF01 = 10.417.

Insufficient power may lead to inaccurate conclusions

To illustrate the relative ability of Bayesian and frequentist approaches to distinguish lack of effect from lack of precision, we repeated both the frequentist and Bayesian analysis, and we intentionally reduced precision and power by reducing sample size to a sample commonly observed in early attentional bias modification research (Mathews & MacLeod,

2002). We repeated the above-described frequentist and Bayesian analysis 1000 times, testing for group differences in attentional bias change while restricting the analyses to 15 randomly selected samples (participants) in each condition. We then obtained a percentage of the results which were considered significant in the frequentist tradition (p < .05) and a percentage of the results that had strong evidence in the Bayesian framework (BF10 > 10). From the 1000 iterations of this analysis, we found that 1.6% of the frequentist analyses yielded significant results (Type I error) and 98.4% would have concluded that there was no effect (a potential

Type II error due to low power), In contrast, using the Bayesian framework, 0.1% of the Bayes

Factors suggested there was a strong effect, 0.1% suggested strong evidence for a lack of an effect, and 99.8% suggested there was only either anecdotal or moderate evidence in either direction. CHAPTER 2 71

Discussion

In this study, we had five overarching aims: to replicate the basic congruence effect, to replicate the relationship between the Dot Probe and anxiety, to replicate attentional training paradigms, to investigate whether training participants to ignore pre-potent, non-emotional stimuli would yield similar results to training individuals to ignore negative stimuli, and to demonstrate the utility of using Bayesian analyses to analyse post-ABM attentional bias.

For the basic Dot Probe effect, both the frequentist and Bayesian analyses suggested that responses on the congruent trials were no slower than on congruent trials. Furthermore, our results provided evidence against a relationship between performance on the Dot Probe and anxiety, or any of the present measures of negative affect. While the frequentist analysis yielded non-significant results, the Bayesian analysis yielded BF01 > 7 for the null. Furthermore, an omnibus ANOVA suggested that a single session of the attend-negative, avoid-negative, or inhibition training (i.e. training participants to ignore pre-potent, non-emotional stimuli) did not shift attentional bias in healthy participants. As there was no significant bias change in the attend-negative or avoid-negative training conditions, our ability to rule out the possibility that inhibition training had no impact similar to those frequently reported in the attentional bias modification literature is limited. Notably, the absence of attentional bias change in the attend- negative and avoid-negative training conditions is consistent with a number of previous studies that have also failed to achieve significant changes in attention bias using tasks based on the traditional dot probe paradigm (Clarke et al., 2014; Notebaert et al., 2015; Clarke et al., 2017,

Everaert et al., 2015). This underscores the need to understand the processes via which bias change is achieved, and also to explore potentially alternative ways of changing attention bias. 72 INHIBITORY TRAINING ON ATTENTIONAL BIAS

There could be several potential reasons as to why there was no significant bias change.

One potential reason there was no reduction of attentional bias in the ‘avoid negative’ condition is that we only used healthy participants without pre-screening for anxiety. It is also possible that multi-session, high dose retraining sessions are required to reliably modify attention.

However, past studies have found success in modifying attentional biases within a single session in healthy participants (e.g. Chen et al., 2015). Therefore, we have reason to believe that rather than simply increasing the dose, we need to further investigate the underlying mechanisms for attentional bias change to achieve consistency in modifying attention. In addition, we sought to investigate whether inhibition training could account for some of the previously observed patterns in the avoid-negative condition, and other studies that have reported success using one session of ABM training have typically included fewer training trials than in the present study. Nevertheless, we acknowledge that past findings with single session training has found mixed findings (e.g. Everaert et al., 2015), and perhaps beyond increasing the dose, the number of sessions also plays an important role.

Other potential reasons may lie with our specific methodology. One possibility is that button press reaction times may be too crude of a measurement to assess the differences present in these biases. Instead, other implicit measures of attention such as eye tracking could be used. Another possibility is that inhibitory control training did not generate change due to using images depicting landscapes. We initially chose to use these images so that naturalistic scenes were used across all training conditions. We did not have these images rated for valence but note that past studies have used these images as neutral controls, demonstrating a CHAPTER 2 73 differential impact to emotional images (e.g. Most et al., 2005; Onie & Most, 2017; Jin, Onie,

Curby & Most, in press). Finally, in our study, we specifically instructed individuals to ignore the images appearing before the probe, but in previous studies participants tended not to be explicitly instructed to ignore the images (e.g. Basanovic, Notebaert, Grafton, Hirsch, & Clarke,

2017; Clarke et al., 2017). Therefore, it is possible that instructions to explicitly ignore these images led participants to exert attentional control in a way that reduced the impact of the emotional images themselves. This may be a more likely possibility among healthy participants

(as in our study) than among highly anxious individuals, who have previously exhibited deficits in attentional control (Eysenck & Derakshan, 2011; Eysenck et al., 2007). It is possible that this contributed to our finding that traditional ABM training conditions did not modify attention.

That is, the images themselves may not have captured attention in the first place, resulting in no consequence of the contingency manipulation.

Furthermore, one issue previously noted is that studies have found the Dot Probe to have low reliability, which may have contributed to previous null findings. However, in this investigation, we found higher reliability estimates than typically reported in the literature (.50

- .53). Considering that the Dot Probe task used is similar to that which is used in other reliability studies, it is unlikely that the underlying reason is theoretical or due to task parameters. Therefore, a likely possibility is that this sample has exceptionally large between- participant variability which is directly correlated with reliability coefficients (Hedge, Powell, &

Sumner, 2018).

Despite the relatively high reliability estimates, they are still lower than the recommended .7 cut-off (Nunnally, 1978), potentially affecting our findings. In this study, we 74 INHIBITORY TRAINING ON ATTENTIONAL BIAS used McDonald’s Omega to compute reliability instead of Cronbach’s Alpha, which is commonly used in the attentional bias literature. We opted to use McDonald’s Omega, as Cronbach’s alpha has shown to be less accurate in the face of violated assumptions, whilst the Omega is less so (Zinbarg, Revelle, Yovel, & Li, 2005). In our case, Cronbach’s Alpha almost always overestimates the reliability value.

Despite the fact that the Dot Probe has been heavily criticised for its lack of reliability, a recent study found that reliability may not be the best way to evaluate the usefulness of a task.

Reliability indices can be heavily affected by between participant-variability, and many cognitive tasks with reliable effects may yield low reliability (Hedge et al., 2018). The authors of that study noted that any calculation designed to reduce noise, e.g. obtaining a bias index by subtracting two values, would almost always reduce between-participant variability, and therefore reliability. Consistent with this, one issue is that not all ABM studies report reliability indices, including those that have found training effects. Therefore, at this time, it is difficult to evaluate the role of reliability in the current context.

In addition to the traditional ABM training conditions, the inhibition control training also did not modify attentional bias. One possibility is it that inhibition training, even if successful, does not causally influence attentional bias. In one study, Heeren et al., (2015) did find that a two session ABM training program improved executive control as measured by the Attention

Network Task (Fan, McCandliss, Fossella, Flombaum, & Posner, 2005), but it may be that this causal relationship is unidirectional. ABM training may improve attentional control, consistent with hypotheses by Chen et al., (2015) and McNally (2019), but attentional training may not modify attentional bias. This is a possibility that should be addressed by future research, which CHAPTER 2 75 would also benefit from delineating between spatial attentional control and executive control, as Heeren et al. (2015) found that ABM modified executive control, but not spatial attentional control. In addition, future studies could also benefit from having pre and post measures of inhibitory control to see whether manipulations of inhibitory control were successful or not.

One potential issue is that traditional attentional retraining conditions have two stimuli on the screen during training whilst the current inhibitory control training condition had only one. Thus, it could be argued that this stimulus-level difference impedes easy comparison across the training conditions. However, attentional training in the ‘inhibitory control’ condition is based on the well-validated anti-cueing task, which involves reorienting away from a pre- potent, attention-grabbing stimulus. Furthermore, the presence of one- versus two- stimuli in the training conditions does not account for our failure to observe a difference between the attend- and avoid- negative groups.

An additional avenue for future research might be to assess whether inhibitory control can improve performance in non-spatial indices of attentional bias, such as emotion-induced blindness (Most et al., 2005), which has been shown to have higher test-retest reliability than typically seen with the dot probe (S. Onie & Most, 2017). It will also be informative to investigate whether inhibitory control mediates the effect of training on attentional bias by assessing attentional control (Basanovic et al., 2017) after training but prior to the assessment of attentional bias.

Importantly, we used Bayesian inference to analyze the robustness of the attentional bias assessment data that was obtained after attentional bias modification. Both the frequentist and Bayesian approach indicated no training effects when using the full sample. 76 INHIBITORY TRAINING ON ATTENTIONAL BIAS

However, the Bayesian approach also distinguished between a lack of an effect and a lack of power, with the resulting Bayes factors suggesting strong evidence to support the null. This was further bolstered when we intentionally reduced power by reducing the sample size to match the sample size used in previous studies, in which we found that over 1000 iterations, the frequentist analysis yielded a 1.6% Type I error, and potentially 98.4% type II error due to low power. The Bayesian approach yielded a 0.1% Type I error, and 0.1% Type-II error, whilst suggesting there was either only anecdotal or moderate evidence 99.8% of the time in the frequentist analysis, a non-significant p value was unable to indicate whether there was true lack of an effect, or a lack of precision. However, the Bayes Factor was sensitive to how much evidence was collected, therefore indicating a measure of precision as well. This demonstrates one of the strengths of the Bayesian approach, which quantifies evidence rather than utilizing a specific cut-off. Note that in this analysis we did use a cut-off of BF = 10 to aid in comparing the two approaches; however, in practice, the Bayes Factor is used to quantify evidence, e.g. we would be able to make stronger conclusions with a Bayes Factor of 9 than a Bayes Factor of 3.5, despite falling into the same bracket in Jeffrey’s scale. An interesting finding is that the frequentist approach to this analysis only yielded 1.6% Type-I error, successfully controlling for error at ⍺ = .05. Therefore, our results seem to suggest that the frequentist approach is not invalid in and of itself in controlling for Type I error. However, this analysis method may prove difficult in controlling for Type II error. Power analyses have been used to reduce these concerns; however, sample size estimates from power analyses rely on effect size estimates from past findings, which may be subject to a number of issues such as publication bias or CHAPTER 2 77 simply random differences in samples. Due to the benefits of Bayesian analyses outlined in the study, subsequent analyses in this thesis will use Bayesian analyses where possible.

In conclusion, we set out to replicate the basic findings in the Dot Probe literature, and compare the outcome of inhibition training with traditional ABM training conditions and found that inhibition training appeared to have no impact on attentional bias. However, the interpretability of this manipulation is limited due to the lack of change in attentional bias in the more traditional attentional bias training conditions as well. Nevertheless, we are able to provide moderate evidence against the relationship between Dot Probe performance and negative affect, as well as strong evidence against any ABM training effects in the current sample of healthy participants.

78 TEST-RETEST RELIABILITY OF EMOTION-INDUCED BLINDNESS

Chapter 3:

Test-Retest Reliability of Emotion-Induced Blindness

Adapted from Onie, S., & Most, S. B. (2017). Two Roads Diverged: Distinct Mechanisms of Attentional Bias Differentially Predict Negative Affect and Persistent Negative Thought. Emotion. https://doi.org/10.1037/emo0000280

The candidate led the design of the study, programming, data collection, data analysis and manuscript preparation. CHAPTER 3 79

Test-Retest reliability of Emotion-Induced Blindness

In previous chapter, I found that Dot Probe indices did not show links with depression, anxiety or stress, and traditional attentional retraining methods did not shift attention. One possibility for these findings (or the lack of) is the Dot Probe task’s reliance on spatial attention.

In this chapter, I seek to investigate the test-retest reliability of emotion-induced blindness, a non-spatial index of attention that has shown links with psychopathology in past studies (See

McHugo et al., 2013).

Spatial attentional bias tasks rely on the assumption that the process of shifting one’s spotlight of attention is synonymous with processing the contents and emotional salience of the target location. Indeed, a majority of the studies using spatial attention tasks investigate differences between high- and low- anxious people do not focus on the subjective emotional salience, but rather attentional shifts towards or away from the said stimulus. (Williams, Watts,

MacLeod, & Mathews, 1988; but see (Mathews & Mackintosh, 1998; Mogg & Bradley, 1998;

Mogg et al., 2000)). The relationship between individual differences and the directional allocation of spatial attention has been found to depend on whether attention is probed very soon (100 ms) or slightly longer (500 ms; Cooper & Langton, 2006) after an emotional stimulus.

It has been suggested that the earlier probes index attentional engagement, whereas later probes index attentional disengagement (Koster, Verschuere, Crombez, & Van Damme, 2005; also see Fox, Russo, & Dutton, 2002, for a similar distinction using a different spatial attention task). However, the distinction may very well be arbitrary. Such reports, together with temporal dynamics of spatial attention itself (e.g., inhibition of return; (Klein & Ivanoff, 2005; Michael I.

Posner, Rafal, Choate, & Vaughan, 1985)), suggest that inferences about attentional biases from 80 TEST-RETEST RELIABILITY OF EMOTION-INDUCED BLINDNESS the dot probe and related spatial attention tasks may be complicated by the possibility that they conflate sensitivity to the emotional salience of stimuli and mechanisms that drive the allocation of spatial attention. Adding to the notion that spatial attention measures are somewhat indirectly related to perceptual prioritization, dissociations have been found between spatial attention to a given location and awareness of stimuli at that location

(Kentridge, Nijboer, & Heywood, 2008; Koch & Tsuchiya, 2007; Lambert, Naikar, McLachlan, &

Aitken, 1999; McCormick, 1997; Most, Simons, Scholl, & Chabris, 2000; Woodman & Luck,

2003), and stimuli that are no longer the focus of spatial attention can continue to be processed and affect subsequent perception (e.g., Fischer & Whitney, 2014).

Therefore, one possible avenue of investigation is to assess attentional bias on a different mechanism of attention, one which does not rely on spatial shifts of attention.

Emotion-Induced Blindness

One task that measures perceptual competition between negative and neutral stimuli, and yet seems to operate on a separate mechanism of attention is emotion-induced blindness

(Most et al., 2005). As mentioned in Chapter 1, on a typical trial, participants view a rapid serial sequence of upright landscape photos and, within each stream, report the clockwise or counter clockwise orientation of a single rotated landscape. When a negative or positive emotional distractor (e.g., a gruesome or erotic picture) appears in the stream just before the target, people spontaneously experience a brief functional “blindness”: For about half a second, they became unable to perceive the target that they were searching for, despite it appearing right in front of their eyes (Most et al., 2005; Most, Smith, Cooter, Levy, & Zald, 2007, but also see

Arnell, Killman, & Fijavz, 2007). The effect is strongest when targets follow distractors by 100 to CHAPTER 3 81

200 ms and diminishes with decreasing temporal proximity (Ciesielski, Armstrong, Zald, &

Olatunji, 2010; Kennedy & Most, 2015); thus, as with the dot probe, later lags are useful for indexing disengagement from the distractors.

Emotion-induced blindness is distinguishable both from spatial attention measures

(Most & Wang, 2011) and from the attentional blink (Wang, Kennedy, & Most, 2012).

Specifically, emotion-induced blindness seems to reflect competitive, spatiotemporally driven interference between targets and emotional distractors (Wang et al., 2012). First, the stimuli are all presented in the same location, so it cannot likely be explained by shifts of attention away from the target location. Second, in an experiment involving two simultaneous streams, in which the distractor could appear in either the same or opposite stream from the target, target perception was worse at than away from the location of the emotional distractor, a pattern opposite that found in typical spatial attention tasks (Most & Wang, 2011).

This suggests that the representations competed for dominance when falling within a shared receptive field (see also (Kennedy, Rawding, Most, & Hoffman, 2014), a mechanism that also appears to distinguish EIB from the attentional blink (Wang et al., 2012; but see Kennedy et al., 2014). Notably, EIB emerges even when the emotional distractor appears immediately after

(instead of before) the target (Most & Jungé, 2008). The spatiotemporal competition account of emotion-induced blindness is consistent with the literature on biased competition (e.g.,

Desimone & Duncan, 1995). Visual neurons that exhibit a given activity level in response to a stimulus falling within their receptive field exhibit a reduced response when a second stimulus also falls within their receptive field, reflecting competition between the two representations

(Chelazzi, Duncan, Miller, & Desimone, 1998; Desimone & Duncan, 1995). Such competitive 82 TEST-RETEST RELIABILITY OF EMOTION-INDUCED BLINDNESS interference can emerge between stimuli separated by a small temporal gap and not just between simultaneous stimuli (Keysers & Perrett, 2002; Wyble & Swan, 2015). Competing items should generally interfere with one another equivalently unless one of them has taken on particular salience (Wyble & Swan, 2015), and emotion-induced blindness appears to be a case in which the emotional power of the distractors imbues them with such salience. Thus, emotion-induced blindness may belong in the company of a small set of behavioral tasks that index prioritization of emotional stimuli in the absence of noise from overlying machinery for the allocation of spatial attention (other such tasks might include binocular rivalry and continuous flash suppression; Alpers & Gerdes, 2007; Yang, Zald, & Blake, 2007).

Emotion-induced blindness has also been found to tap into dysfunctions in selective processing in certain disorders. Olatunji and colleagues (2013) found that combat-exposed veterans with post-traumatic stress disorder showed greater impairment by combat-related stimuli compared to combat-exposed veterans without PTSD and healthy controls.

Furthermore, impairment by other emotional stimuli (e.g. disgust or erotic stimuli) was comparable across groups, suggesting a lack of overall hypervigilance, but rather increased sensitivity disorder specific information. This is consistent with the finding that emotion- induced blindness is sensitive to current stimulus value (Smith et al., 2006). In another study,

Olatunji and colleagues (2011) found that individuals diagnosed with generalized anxiety disorder had overall greater impairment across shorter and longer lags. While this is consistent with an overall elevated threat sensitivity in anxiety, it was unclear whether this emerged due to an impairment in attentional control (see. Derryberry & Reed, 2002). CHAPTER 3 83

Correlational studies have also found that emotion-induced blindness predict clinically- relevant individual differences. Kennedy and Most (2015) found that emotion-induced blindness predicted the brooding subscale of the Ruminative Response Scale (Treynor et al.,

2003). Onie & Most (2017) found that performance on emotion-induced blindness predicted a component score derived from rumination and worry questionnaires, as well as a component score derived from the depression, anxiety and stress subscales of the Depression Anxiety

Stress Scale – 21 (Lovibond & Lovibond, 1995). Furthermore, it was found that persistent negative thought significantly mediated the relationship between emotion-induced blindness at lag 1 and negative affect. However, note that Kennedy and Most, and Onie and Most, used different indices of emotion-induced blindness. While Kennedy and Most used a score obtained subtracting accuracy on lag 2 from lag 1 to index speed of recovery from negative stimuli, Onie and Most used accuracy on negative trials at lag 1 while statistically controlling for accuracy on neutral at lag 1 to index emotional impairment. Further research is needed validate different emotion-induced blindness indices.

Overall, the emotion-induced blindness paradigm seems to be a promising way of assessing attentional bias, albeit on a mechanism removed from spatial attention. From this point forward, the bulk of the thesis will be employing emotion-induced blindness as its primary technique.

Test-Retest Reliability: The Intra-Class Correlation

However, before further investigation with the emotion-induced blindness, it is important to measure its test-retest reliability. Test-retest reliability refers to a measurement’s stability over time, often quantified as measurement consistency between two distinct time 84 TEST-RETEST RELIABILITY OF EMOTION-INDUCED BLINDNESS points. Test-retest reliability has been used in distinguishing between state and trait characteristics, in that validated measures with higher test-retest reliability may reflect trait characteristics, while low test-retest reliability reflects higher state fluctuation. Typically, tasks are required to have over .7 reliability (Nunnally, 1978); however, further investigations into the source of this reveal that this recommendation is largely arbitrary (Lance, Butts, & Michels,

2006).

A recent paper highlighted the importance of test-retest reliability in individual difference research (Hedge et al., 2018). Test-retest reliability is most commonly measured using the intra-class correlation coefficient. The formula for the intra-class correlation for test- retest reliability purposes is given below:

�� = �� + �� + ��

From the formula, three distinct factors affect intra-class correlation: error variance, between session variance, and most importantly, between subject variance. The greater the error variance and between session variance, the smaller the reliability estimate. However, if two tasks had identical measurement error and between session error, but one had smaller between-participant variability, the task in question would yield lower test-retest reliability. It is extremely important for tasks used in individual difference research to have high between subject variance – and by extension test-retest reliability - as this line of research seeks to rank subjects on two or more measurements. For example, with high between participant variance, even with moderate error, the order or data points remains the same. However, if the data CHAPTER 3 85 points are clustered together, even a small amount of error can change the order of each participant.

Therefore, test-retest reliability is an important indicator of utility in studying clinically relevant individual differences. Several studies have investigated the test–retest reliability of the Dot Probe, often finding it to be low. Test–retest reliability coefficients, for example, have ranged from .243 to .32 (Schmukle, 2005; Staugaard, 2009), suggesting that the task may be suboptimal for indexing stable individual differences. A recent review has acknowledged this fact and calls for more reliable tasks to be used. One possibility is the use of difference scores, as any process which reduces variability will also invariably reduce between-subject variability, resulting in lower test-retest reliability (Hedge et al., 2018). Efforts have been made to use other indices of attentional bias e.g. bias variability scores (Khatibi, Dehghani, Sharpe,

Asmundson, & Pouretemad, 2009; Zvielli, Bernstein, & Koster, 2014); however, simulations have shown that these scores may index random variability than actual attentional bias (Kruijt et al., 2016). Another possibility is that the task’s low reliability stems from its measurement of spatial attention shifts, with the mechanisms driving spatial attention allocation introducing noise that partly obscures individual differences in the representational salience of the emotional content. If so, emotion-induced blindness—which does not rely on spatial attention—might exhibit a more favorable test–retest reliability. Therefore, in this study, we replicated the basic emotion-induced blindness effect and investigated its test-retest reliability.

86 TEST-RETEST RELIABILITY OF EMOTION-INDUCED BLINDNESS

Method

Design

In a repeated-measures design, participants completed the emotion-induced blindness task on Day 1 and came back a week later and completed the task again. Participants were compensated course credits for their time.

Participants

Eighty-four participants (43 women, 41 men; Mage = 22.1 years, SDage = 5.89 years) participated in the study. Fifty-five (29 women, 26 men; Mage = 22.9 years, SDage = 5.93 years) participants returned for the second visit, and thus their scores were used to calculate the test–retest reliability of EIB.

Materials and equipment.

Behavioral tasks were computerized, with stimuli presented on a 24-in. BenQ XL2420T

LED monitor with 1920 x 1080 resolution and 120-Hz refresh rate. Head position was not fixed.

Emotionally negative and neutral images were sourced from the International Affective Picture

System (Lang et al., 1997) and supplemented with similar pictures from publically available sources. Together, these pictures had been rated by a separate group of six men and six women on a scale from 1 to 9 (1 = very negative/low arousal; 9 = very positive/high arousal), with ratings of the negative pictures significantly different from those of the neutral pictures, in both valence (MNegative = 1.727, SDNegative = 0.530; MNeutral = 5.013, SDNeutral = 0.454), t(11) = 35.0, p <

.001) and arousal (MNegative = 6.055, SDNegative = 0.678; MNeutral = 3.196, SDNeutral = 0.549, t(11) =

24.4, p < .001.)

CHAPTER 3 87

Procedure

In the emotion-induced blindness lags one, two and four were used. In each emotion- induced blindness trial, participants were instructed to find one rotated image in a rapid stream of 17 images and, after each stream, to indicate, via keypress, the orientation of the target.

Images measured roughly 11.5 cm x 8.5 cm on the screen. Each stream item appeared for 100 ms before being immediately replaced with the next. Except for the distractors, all items were images depicting landscapes. Trials either contained a negative emotional distractor (96 trials), a neutral distractor (96 trials), or no distractor (baseline; 96 trials). The target could appear at serial 4 to 8. If the participant had correctly indicated the rotation of the target image, a bell sound would play. Incorrect responses had no feedback.

Results

Emotion-Induced Blindness

We first investigated the emotion-induced blindness effect, in which accuracy is lower for targets is following a negative distractor than a neutral distractor. The difference is typically greater at lag 2 than lag 1, but slowly dissipates as the lags increase (Kennedy & Most, 2015b).

We aggregated participant responses in emotion-induced blindness at time 1 and two separately, obtaining percentage accuracies for each lag and distractor type per participant which can be found in Table 4.

88 TEST-RETEST RELIABILITY OF EMOTION-INDUCED BLINDNESS

Table 4 Accuracy Means and Standard Deviations for emotion-induced blindness Time 1 Time 2

Lag Negative Trials Neutral Trials Negative Trials Neutral Trials

1 70.34%(9.775%) 77.08% (10.77%) 75.76% (11.48%) 80.00% (13.03%)

2 72.94% (11.08%) 82.81% (11.89%) 79.81% (13.82%) 87.65% (13.82%)

4 81.57% (10.98%) 85.22% (12.33%) 85.42% (14.11%) 88.75% (14.36%) Note: Standard deviations are provided in parentheses, all values given are percentage scores.

Time 1. We investigated the presence of an emotion-induced blindness effect using a

Repeated Measures ANOVA with valence (negative vs neutral) and lag (1,2 and 4) as fixed within factors and accuracy as a dependent variable. The results revealed that a model containing Valence, Lag and the Valence x Lag interaction accounted for the data best, 333.3 times more likely than a model with just Valence and Lag, and 5.684 x 1050 times more likely

31 24 than the null (Lag: BFinc = 1.293 x 10 , Val: BFinc = 1.327 x 10 , Lag x Val: BFinc = 333.3).

We conducted follow up paired samples t-tests at each lag between negative and neutral trial accuracies to investigate the magnitude of emotion-induced blindness at each temporal distance. We tested the hypothesis that negative trials had lower accuracy than neutral trials at lags one, two and four. The analysis revealed extremely strong evidence that there was greater impairment following negative distractors than neutral distractors at lag one

6 13 (BF10 = 3.220 x 10 , � = -0.687, 95% CI [-0.924, -0.449]), Lag 2 (BF10 = 2.158 x 10 , � = -1.070,

95% CI [-1.341, -0.815]), and Lag 4 (BF10 = 464.5, � = -0.436, 95% CI [-0.658, -0.219]).

Overall, the analyses suggest that we replicated the emotion-induced blindness effect, in which negative distractors cause impaired target detection relative to neutral distractors. CHAPTER 3 89

Time 2. We repeated the analysis above for the data at time 2 and performed a

Repeated Measures ANOVA with valence (negative vs neutral) and lag (1,2 and 4) as fixed within factors and accuracy as a dependent variable. The results revealed that a model containing Valence, Lag and the Valence x Lag interaction accounted for the data best, albeit only 1.954 times more likely than a model with just Valence and Lag, and 3.00 x 1026 times

19 9 more likely than the null (Lag: BFinc = 3.093 x 10 , Val: BFinc = 5.564 x 10 , Lag x Val: BFinc =

1.954).

We conducted follow up paired samples t-tests at each lag between negative and neutral trial accuracies to investigate the magnitude of emotion-induced blindness at each temporal distance. We tested the hypothesis that negative trials had lower accuracy than neutral trials at lags one, two and four. The analysis revealed strong to extremely strong evidence that there was greater impairment following negative distractors than neutral distractors at lag one (BF10 = 69.41, � = -0.457, 95% CI [-0.730, -0.181]), Lag 2 (BF10 = 1.327 x

7 10 , � = -0.943, 95% CI [-1.273, -0.620]), and Lag 4 (BF10 = 103.94, � = -0.475, 95% CI [-0.753, -

0.210]).

Overall, the analyses suggest that we also replicated the emotion-induced blindness effect at time 2, in which negative distractors cause impaired target detection relative to neutral distractors.

Test Re-test Reliability

Test–retest reliability was computed via intra-class coefficient. Note that this analysis was done using a frequentist approach as when this thesis was written, no accessible method for computing Bayesian intra-class correlation coefficients was available. 90 TEST-RETEST RELIABILITY OF EMOTION-INDUCED BLINDNESS

A recent study showed that negative accuracy while controlling for neutral accuracy predicted negative affect and persistent negative thought (Onie & Most, 2017). However, partialling accuracy on neutral trials from that on negative trials was not possible with test– retest reliability analysis methods, the analysis controlling for performance in neutral trials followed precedent (e.g., Most et al., 2005) by employing the difference between accuracy on negative and neutral trials as the attentional bias index. The intra-class coefficients at each lag, for negative trials alone, neutral trials alone, and the difference between negative and neutral trials, are reported in Table 5.

Table 5 Test-Retest ICC Coefficients EIB Score Used Lag ICC p - values Negative 1 .721 <.001 2 .797 <.001 4 .894 <.001 Neutral 1 .864 <.001 2 .907 <.001 4 .911 <.001 Neutral minus Negative 1 .439 <.001 2 .419 4 .333 Note. Values reported were taken from ICC average measures output. ICC = intra-class correlation coefficient; EIB = emotion- induced blindness.

Results revealed that the neutral-minus-negative index had a test–retest reliability coefficient of .439 at Lag 1, .419 at Lag 2, and .333 at Lag 4.

Discussion

In this study, we conducted a replication of the emotion-induced blindness as well measured the test-retest reliability for a series of different emotion-induced blindness indices. CHAPTER 3 91

The results revealed strong to extremely strong evidence of the emotion-induced blindness effect, that is increased impairment on trials with a negative distractor relative to trials with a neutral distractor. This effect was present at both time points, and median effect size estimates suggests this effect is strongest at lag 2. Furthermore, the emotion-induced blindness test- retest reliability indices at early lags (.419 - .907), compare favorably with those reported in the dot probe literature (e.g., .22 to .32 in Schmukle, 2005, and .243 to .262 in Staugaard, 2009), as well with those reported in the broader attentional bias literature (values .33 over a wide range of tasks including the emotional Stroop, visual search, and morph tasks; (H. M. Brown et al.,

2014; Eide, Kemp, Silberstein, Nathan, & Stough, 2002; Kindt, Bierman, & Brosschot, 1996;

Siegrist, 1997), suggesting that emotion-induced blindness may be more suited for individual difference research.

In particular, performance on Lag 1 trials with negative distractors alone (i.e., not controlling for neutral trials) had a relatively robust test–retest reliability of .721, which is higher than the average test–retest reliability found in the attentional bias literature. Taken with the finding that Lag 1 trials with negative distractors alone have the same predictive value as negative distractors while controlling for neutral distractors in predicting negative affect and persistent negative thought (Onie & Most, 2017), suggests that emotion-induced blindness at

Lag 1 may be a suitable index in clinical individual difference research. Of course, using Lag 1 negative alone without controlling for neutral changes the interpretation from the relative attentional capture of emotional vs non-emotional stimuli, to simply how much does emotional stimuli capture attention. Therefore, more research needs to be done in validating negative accuracy without controlling for neutral accuracy. 92 TEST-RETEST RELIABILITY OF EMOTION-INDUCED BLINDNESS

Taken together with previous studies which link emotion-induced blindness with clinically relevant individual differences, as well as the findings from the current study, emotion-induced blindness seems to be a useful measure of stable individual differences in attentional bias. Higher test-retest reliability also aligns with the hypothesis that emotion- induced blindness relatively directly indexes the prioritization of emotional representations, without measurement noise that might result from overlying mechanisms for spatial attention allocation. CHAPTER 4 93

Chapter 4:

Investigating whether Valence or Arousal drive Attentional Capture in two Distinct Attentional Paradigms

Currently in preparation for submission as Sandersan Onie, & Steven B. Most (in prep).

The candidate led the design of the study, programming, data collection, data analysis and manuscript preparation 94 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

Investigating whether Valence or Arousal drive Attentional Capture in two Distinct

Attentional Paradigms

Given that temporal tasks such as emotion-induced blindness are robustly sensitive to the presence of emotional stimuli, it is important to understand whether their sensitivity is driven by any particular emotional qualities. For example, dimensional views of emotion distinguish between continuous dimensions of valence (how positive or negative is this stimulus?) and arousal (how intense is the emotional impact of the stimulus regardless of valence?; (Bradley & Lang, 1994; Russell, 1980; Russell & Barrett, 1999). Therefore, in this chapter, we seek to investigate whether attentional capture in emotion-induced blindness is driven by valence, arousal, or a combination of both. In addition, we also applied the same question for the Dot Probe.

Some models suggest that valence should be a stronger driver of emotional distraction.

For example, the Categorical Negativity Theory posits that a valence evaluation process allocates attention towards emotional, particularly negative stimuli, regardless of intensity

(Pratto & John, 1991). Specifically, the model predicts that negatively valenced information will always attract attention, regardless of the arousal level. This is consistent with the existence of a preattentive threat detection mechanism, which posits that certain specialized feature detectors assess threat prior to conscious awareness (e.g. Mogg & Bradley). Öhman and Mineka

(2001) build upon this and suggest that threat processing have a privileged position in processing and enjoy faster processing through a rudimentary amygdala pathway (but also see,

(Pessoa & Adolphs, 2010)). CHAPTER 4 95

However, other findings suggest that arousal may play a driving role. Using a lexical decision task, where people make speeded responses indicating whether strings of letters form words, Larsen, Mercer, Balota & Strube (2008) found that lexical decision response times were longer for low arousal negative words than high arousal negative words. Using a different task,

Fernandes, Koji, Dixon and Aquino (2011) asked participants to assess whether two numbers on either side of an image (positive or negative, high or low arousal) were both odd, both even, or a mix of both. An interaction emerged: when the arousal ratings for the images were high, response times were slower in the presence of negative than positive images; but when arousal ratings of the images were low, response times were slower for positive images. Using the attentional blink, other studies found that arousal alone modulated attention, regardless of valence (e.g., Anderson, 2005) a conclusion that was similarly reached in studies using the dot probe (Vogt, De Houwer, Koster, Van Damme, & Crombez, 2008).

In the context of attentional bias, the question of what drives attentional capture is an important one, as this directly affects the conclusions drawn between attentional bias and clinically relevant individual differences. For example, if it is found that an attentional bias task is driven by arousal and not valence, and it is found that depression predicts increased impairment by negative distractors (as a distractor label, not mechanistically), then the conclusion would be that depression leads to increased attentional capture by intense or arousing stimuli. Therefore, whether valence or arousal drives attentional capture at each level of attention may directly affect our understanding of how attention is biased in specific disorders. 96 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

Therefore, in this study, we seek to investigate whether valence or arousal drives attentional capture in the Dot Probe and emotion-induced blindness, two attentional bias tasks which both have been shown to link with psychopathology, but which have been suggested to tap into different mechanisms of attention. Furthermore, our investigation pertains to a larger line of work investigating attentional biases in clinical disorders, which focuses on attention towards negative stimuli due to its proposed causal relationship with certain emotional disorders. Therefore, we constrained this study to negative stimuli.

Past studies have sought to investigate the relative impact of valence and arousal using emotion-induced blindness and the Dot Probe. Singh and Sunny (Singh & Sunny, 2017)

investigated the relative impact of valence and arousal in emotion-induced blindness by varying valence (negative vs positive) in one study, and varying arousal (high vs low) in a second. The authors found that while there was no difference in positive vs negative stimuli, there was a difference between high and low arousal, in which high arousal results in more impairment. Sutton and Lutz (Sutton & Lutz, 2018) also investigated this question with the Dot

Probe, in which they included stimuli that varied in valence (low and high) and arousal (low and high). The authors found that for negative stimuli, incongruent trials were always slower than congruent trials; however, for positive trials, only high arousal images resulted in the same pattern of findings.

In our study, instead of varying arousal and valence, we sought to conduct a higher resolution analysis, in which we investigated whether specific ratings of valence and arousal drive attentional capture in spatial attention and temporal attention. In the current literature, arousal (low vs high arousal) is considered to play a stronger role than valence (positive vs. CHAPTER 4 97 negative valence). Two problems with this interpretation are a) it is almost impossible to disentangle the effects of valence and arousal using these designs, as valence could very well be driving attention for both positive and negative distractors, but since there is no neutral comparison, the effects of valence and arousal cannot be separated and b) different mechanisms of attention are not considered, as the typical attention to negative information could be more predominant in some mechanisms of attention but not others. In contrast, the natural variation of valence and arousal in stimuli, rather than manipulations of overall differences of valence and arousal between groups of stimuli, may better enable us to disentangle the two.

In Experiment 1, we investigated whether valence and/or arousal of the images predict how impairing an image is in an emotion-induced blindness task, and in Experiment 2, we investigated whether valence and/or arousal of the images predict attentional capture in the

Dot Probe. we hypothesized that if valence or arousal play a role in competitive interference, then the proportion of incorrect trials for any given image will be predicted by its valence, arousal, or both. However, by lag 8, neither valence nor lag should play a role since typically negative and neutral trials no longer exert influence on accuracy. For the Dot Probe, since two images are used on each trial, if valence or arousal play a role in spatial attentional capture, then the difference of the valence or arousal of the two images should predict RTs on each trial.

Experiment 1

Method

Design. To investigate whether valence and/or arousal ratings predicted distractibility in the emotion-induced blindness task, we performed two separate data collections. First, we 98 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE collected valence and arousal ratings for neutral and negative images using Amazon’s

Mechanical Turk. A second sample of participants completed a session of emotion-induced blindness in the lab, which incorporated the rated neutral and negative images as distractors.

Participants. For the image ratings, we recruited 96 participants (47 males, 49 females;

Age: M = 42.27 years, SD = 13.15 years, Range = 23 - 73 years) on Amazon’s Mechanical Turk without any geographic restriction. Participants were reimbursed USD$ 2.50 for one hour of their time.

For emotion-induced blindness, we recruited 99 participants to the lab (38 males, 61 females; Age: M = 23.19 years, SD = 4.26 years, Range = 18.03 - 42.01 years) from the local community through a paid recruitment system (SONA-P). Participants were reimbursed

AUD$15 for one hour of their time. This study was carried out in accordance with the recommendations of the Human Research Ethics Advisory Panel at UNSW, with written informed consent from all participants.

Materials.

Images. Images for the distractor ratings and emotion-induced blindness tasks were collected from the International Affective Picture System (IAPS; Lang, Bradley, & Cuthbert,

1997), with selection of the images guided by whether they had been rated as neutral or negative in the IAPS norms (Lang et al., 1997), and these were supplemented by images with similar content from publically available sources. A total of 150 images were collected (75 negative and 75 neutral). The negative images depict gore and disgust images (e.g. mutilated corpses and soiled toilets). Neutral images depict similar objects in non-negative settings

(people in everyday scenes as well as household objects). CHAPTER 4 99

Emotion-induced blindness also uses pictures depicting landscapes and buildings in the stream. These images were collected from a variety of sources including Google searches and photographs taken by members of the lab.

Distractor Rating Questionnaire. To collect distractor ratings, we used Qualtrics, an online survey development program.

Emotion-Induced Blindness. Stimuli were presented on a 24-inch BenQ XL2420T LED monitor with 1920 × 1080 resolution and 120-Hz refresh rate, using Matlab via the

Psychophysics Toolbox extensions (Brainard, 1997).

Procedure.

Distractor Ratings. At the start of the session, example images were shown to ensure informed consent regarding the graphic nature of the images. Participants were then asked whether they would like to continue the study. If they agreed, the experiment began. Following that, participants were given brief descriptions of “valence” and “arousal” to help them answer the questions in the task.

The descriptions were as follows for valence and arousal for the first and second question respectively: “The first question will ask how negative the image makes you feel. One extreme of this scale represents finding the pictures to be unhappy, disgusting, bothersome, irritating, or sad. The other end of the extreme represents feeling none of the negative emotions outlined in the previous sentence (note: this other extreme does not represent how positive the image makes you feel).

The second question will ask how intense the picture is. At one extreme, 'very intense' indicates that the picture has high emotional energy and may make viewers feel 'buzzed', jittery, 100 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE wide-awake, stimulated, excited, or frenzied. At the other extreme, 'not at all intense' represents absence of characteristics outlined in the previous sentence. It is useful to note that pictures can sometimes be highly negative or positive without having high intensity (and vice versa).”

On each trial, one image was presented at a time, followed by two questions: for valence ratings, participants were asked ‘How does this image make you feel?’, to which they answered on a scale ranging from -9 (extremely negative) to 9 (extremely positive). For arousal, we asked the participants ‘how intense is this image?’, to which they responded on a scale ranging from 0 (not at all intense), to 9 (extremely intense). For approximately half the participants, the sign on the valence question was flipped to control for biases to respond above or below zero.

To better focus participant responses on discriminating the relative valence and arousal values, rather than impulsively categorizing images as neutral or negative, the 75 neutral and

75 negative images were separately and randomly allocated into three blocks of 25 images each, for a total of 6 blocks. The blocks were then presented in alternating valence with a neutral block presented first. After initial randomization and allocation, the order was set, and each participant did the exact same questionnaire.

Emotion Induced Blindness. This analysis was conducted as part of a larger data collection, which – in addition to testing the relative contributions of valence and arousal to emotion-induced blindness – investigated the relationship between emotion-induced blindness and different depression and anxiety questionnaires and aimed to establish best practices in analyzing emotion-induced blindness (e.g. using random slopes in linear mixed modeling). CHAPTER 4101

After giving consent, participants were given a series of questionnaires: Depression,

Anxiety, Stress Scale - 21 (Lovibond & Lovibond, 1995), State Trait Anxiety Inventory

(Spielberger, Gorsuch, & Lushene, 1970), Participant Health Questionnaire - 9 (Kroenke, Spitzer,

Williams, & Löwe, 2010), Generalised Anxiety Disorder - 7 (RL, Kroenke, JW, & Löwe, 2006),

Penn State Worry Questionnaire (Meyer, Miller, Metzger, & Borkovec, 1990) ,Ruminative

Response Scale (Treynor et al., 2003). These were administered on a computer using Qualtrics.

Participants then completed the emotion-induced blindness task. In each trial, participants were instructed to find one rotated image in a rapid stream of 17 images and, after each stream, to indicate the orientation of the target image via keypress. A bell sound played as feedback for each correct response. Images measured approximately 11.5 cm x 8.5 cm on the screen. Each stream item appeared for 100 ms before being immediately replaced with the next. Except for the distractors, all items were images depicting landscape and architecture.

Trials either contained a negative emotional distractor (297 trials), a neutral distractor

(297 trials), or no distractor (baseline; 297 trials). The target could appear at serial position 4–8 in the stream with equal probability. One third of the negative and neutral distractors preceded the target by one serial position (lag-1; SOA 100 ms), one third preceded the target by two serial positions (lag-2; SOA 200ms), and one third preceded the target by eight serial positions

(lag-8; SOA 800ms).

Results

All analyses were performed in JASP (JASP Team, 2018) and JZS priors unless stated otherwise. The 95% CIs reported here are Credible Intervals, with effect size medians reported. 102 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

A correlation matrix revealed that the valence and arousal ratings of negative images are highly correlated (r = -0.975, BF10 = ∞), which leads to the issue of multi-collinearity. In order to circumvent this issue, we refrained from investigating the contribution of individual predictors within a model and pursued a model comparison approach, as multi-collinearity does not preclude inference from distinct models (Frost, 2017). Therefore, we do not seek to compare the relative predictive values of variables in a single model and thus do not report regression coefficients or their respective Bayes Factors for the predictors within a model. In analyses with potential multi-collinearity, often the variance inflation factor (VIF) is reported, which is how many times the standard deviation of the regression coefficient is multiplied by due to multicollinearity. However, as we do not report individual coefficients, we also cannot report individual VIFs for each predictor. If we did attempt to estimate regression coefficients,

VIF values would be above 10, further justifying a model comparison approach.

Distractor Ratings. To investigate differences in ratings of valence and arousal between neutral and negative images, we aggregated ratings for each image to obtain a mean valence and arousal rating per image. We then applied Bayesian paired samples t-tests to assess the difference in ratings of the stimulus sets.

For valence ratings, we tested the hypothesis that negative images were rated more negatively than neutral images (ValNegative < ValNeutral). The analysis revealed extremely strong evidence that negative images were rated more negatively than neutral images (BF10 = 4.07 x

42 10 , � = -5.326, 95% CI [-5.981, -4.742], MNegative = -5.587, SDNegative = 1.533, MNeutral = 1.766,

SDNeutral = 0.856). CHAPTER 4103

We then investigated whether negative images were rated as more arousing than neutral images (ArNegative > ArNeutral). The analysis revealed extremely strong evidence that

40 negative images had higher arousal than neutral images (BF10 = 4.445 x 10 , � = 4.969, 95% CI

[4.445, 5.472], MNegative = 6.155, SDNegative = 1.168, MNeutral = 1.887, SDNeutral = 0.264).

Emotion-Induced Blindness. We aggregated participant’s emotion-induced blindness responses, obtaining percentages correct for each lag and distractor type per participant. Mean percentages correct for each lag and distractor type can be found in Table 6. Following that, we investigated the presence of an emotion-induced blindness effect using a Repeated Measures

ANOVA with valence (negative vs neutral) and lag (1, 2, and 8) as fixed within factors and percentage correct as a dependent variable. To analyze the output, we used model comparison to infer which model best accounts for the data. For each possible model, a Bayes Factor is generated, quantifying how likely it is compared to the null. This way we can infer which model best accounts for the data, as well as how likely each model is compared to each other.

The analysis revealed that the best model contained main effects of valence and lag, as well as a valence X lag interaction. This model outperformed the second-best model, which contained only lag and valence main effects, by 1.850 x 1038 times, and it outperformed the null

129 49 55 hypothesis by 9.821 x 10 times (Lag: BFinc = 9.528 x 10 , Val: BFinc = 1.410 x 10 , Lag x Val:

38 BFinc = 1.850 x 10 ).

We conducted follow-up paired samples t-tests at each lag between negative and neutral trial accuracies to investigate the magnitude of emotion-induced blindness at each temporal distance between target and distractor. We tested the hypothesis that negative trials yielded lower accuracy than neutral trials at lags 1 and 2. The analysis revealed extremely 104 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE strong evidence for higher accuracy in neutral trials than in negative trials at both lag 1 and lag

2, although median effect size estimates suggest a larger magnitude of emotion-induced blindness at lag 2 than at lag 1 (Lag 1: BF10 = 880,427, � = -0.835, 95% CI [-1.128, -0.543]; Lag 2:

10 BF10 = 1.498 x 10 , � = -1.145, 95% CI [-1.448, -0.837]). Inspection of the means suggests that the increase in magnitude from lag 1 to lag 2 is due to neutral accuracy increasing more than negative accuracy (Negative Lag 2 – Lag 1 = 3.48%; Neutral Lag 2 – Lag 1 = 5.05% ).

Typically, at lag 8, emotion-induced blindness is no longer present (Most et al., 2005;

Wang, Kennedy & Most, 2012; Singh & Sunny, 2017). Therefore, we tested the hypothesis that percentages correct on negative and neutral trials at lag 8 were equivalent. The analysis revealed moderate evidence that this was the case (BF01 = 8.928, � = -0.010, 95% CI [-0.281,

0.263]). Furthermore, a paired-samples t-test show that by lag 8, neutral and negative trials were equivalent in accuracy to baseline trials in which there was no distractor present

(Negative – Baseline: BF10 = 8.808, � = -0.024, 95% CI [-0.244, -0.298]; Neutral-Baseline: BF10 =

8.762, � = -0.025, 95% CI [-0.246, -0.290]).

Together, the analyses suggest that we replicated the emotion-induced blindness effect, in which negative distractors cause impaired target detection relative to neutral distractors at shorter lags, and that distractor effects are no longer present by lag 8.

Do valence and arousal ratings predict distraction in emotion-induced blindness?. We first obtained a distraction score for each individual distractor by calculating the percentage of correct responses across presentations of a given distractor (in this case, the

“distraction score” is inversely associated with the degree of distraction caused). Thus, each image was linked with a valence rating, an arousal rating, and a distraction score. On average, CHAPTER 4105 each distractor was presented 130 times over 99 participants, per trial type, and 390 times in a single experiment.

We then investigated whether valence or arousal best predicted performance at each lag using multiple regression, by entering both valence and arousal in the models. At lag 1, the analysis revealed that a model with valence alone best accounted for the data (R2 = 0.108),

2.024 times over the second-best model which only contains arousal (R2 = 0.099), and 468.238 times greater than the null model, suggesting that at lag 1, there was marginal evidence valence best accounted for the data over and above arousal, but either was extremely more likely than the null.

At lag 2, the analysis revealed that a model with valence alone best accounted for the data (R2 = 0.257), 3.217 times over the second-best model which contains both valence and arousal (R2 = 0.265), and 2.153 x 108 times greater than the null model, suggesting that at lag 2, there was moderate evidence valence best accounted for the data over and above both arousal and valence, but either was extremely more likely than the null.

At lag 8, the analysis revealed that the null model was best, 5.590 times more likely than the second-best model with just valence, 5.688 times more likely than the model with just arousal, and 17.040 times more likely than the model with both valence and arousal. This is expected as distractors typically no longer exert any influence by lag 8 (Most et al., 2005; Singh

& Sunny, 2017; Wang et al., 2012).

Therefore, we have evidence that distraction at early lags is driven by the valence of the images, with more evidence that this is true for lag 2 than lag 1. However, due to preselecting stimuli based on perceived valence, we create two discrete clusters of data points, 106 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE corresponding to neutral and negative images. Therefore, even if there is no relationship, this may force a regression line connecting these two clusters, thereby biasing the results in favour of a valence only model. Therefore, for a more fine-grained analysis, we conducted the same multiple regression analyses with only the negative distractors, investigating whether amongst negative distractors, valence and/or arousal predicts distraction.

Do valence and arousal ratings predict distraction in emotion-induced blindness using only negative distractors? In this section, we repeat the analysis above, using only the negative distractors. This is to circumvent a “forced regression” that would have stemmed from the negative- and neutral- sets of stimuli being extremely different in both valence and arousal.

At lag 1 the analysis revealed that the null model best accounted for the data, which was only 1.226 times more likely than a model with both valence and arousal (R2 = 0.073),

1.969 times more likely than a model with just valence (R2 = 0.024), and 3.052 times more likely than a model with just arousal. Therefore, while there was moderate evidence against a model with just arousal or valence, there was inconclusive evidence to whether the null or a model with both valence and arousal was better.

At lag 2, the analysis revealed that the best model contained both valence and arousal

(R2 = 0.288), which was 68.734 times more likely than the second-best model with just valence

(R2 = 0.155), 759.978 times more likely than the model with just arousal, and 3,907.739 times more likely than the null model. Therefore, the analysis strongly suggests that valence and arousal uniquely contribute to target impairment in emotion-induced blindness at lag 2. This is

2 2 2 2 also supported by observing the R values in which R VAL+AR is almost equivalent to R VAL + R AR, consistent with the notion that both uniquely contribute to distractibility. CHAPTER 4107

At lag 8, the analysis revealed that the null model was preferred, which was 2.878 times more likely than the second-best model with only arousal, 3.466 times more likely than the model with only valence, and 5.128 times more likely than the model with both valence and arousal. Therefore, there was marginal to moderate evidence that neither valence nor arousal predicted distraction. This was to be expected as at lag 8, there was no additional impairment due to distractor presence.

Table 6. Accuracy Means and Standard Deviations for Emotion-Induced Blindness EIB Accuracy (in %) Arousal (range: 1 - 9) Valence (range: -9 - 9)

Negative Neutral Negative Neutral Negative Neutral

Mean 72.57% 77.31% 6.155 1.887 -5.587 1.766

SD 5.84% 4.55% 1.168 0.264 1.533 0.856

Minimum 62.08% 63.08% 3.010 1.542 -8.115 -0.771

Maximum 88.59% 86.30% 8.125 2.958 -1.938 4.615 Note: Under EIB Accuracy, Negative and Neutral refer to trial types while under Arousal and Valence, Negative and Neutral refers to type of image used.

Therefore, while it is unclear whether a model with valence and arousal or the null model best accounted for distraction at lag 1, there was extremely strong evidence that a model with both valence and arousal predicted distraction at lag 2. At lag 8, there was marginal evidence that the null model best accounted for distraction.

Does emotion-induced blindness predict negative affect and/or persistent negative thought. In the study, participants completed the Depression, Anxiety, Stress Scale 21, the

Generalized Anxiety Disorder 7, Participant Health Questionnaire 9, State Trait Anxiety Index, 108 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

Penn State Worry Questionnaire, and the Ruminative Response Scale. We first submitted all the questionnaires to a principal components analysis in a bid to reduce dimensionality in the data and a parallel analysis suggested that one component accounted for all the different questionnaires. We then used the component score to as an overall negative affect and persistent negative thought index (henceforth just a negative affect index) and investigated whether different indices of emotion-induced blindness predicted this index.

In line with Onie & Most (2017), we first investigated whether accuracy in negative trials along or in negative trials while controlling for accuracy in neutral trials predicted the negative affect index. The analysis suggested that the null was 3.925 times more likely than a model with just negative (R2 = 0.004) and 2.256 times more likely than negative while controlling for neutral (R2 = 0.008). Therefore, there is marginal to moderate evidence, that emotion-induced blindness at lag 1 did not predict negative affect. We repeated this analysis for each and every one of the questionnaires, finding moderate to strong support that emotion-induced blindness did not predict any of these measures (BF01 = 2.734 – 14.174).

Discussion

In Experiment 1, we investigated whether valence or arousal predicted distraction in emotion-induced blindness. We first established that negative images were both rated more negative and more arousing than neutral images, followed by the presence of an emotion- induced blindness effect at lags 1 and 2. Following that, analyses investigating whether valence and arousal ratings predicted emotion-induced blindness revealed marginal and moderate evidence that valence best accounted for distraction at lags 1 and 2 respectively. At lag 8, the null model best accounted for distraction, but this was expected as we observed no emotion- CHAPTER 4109 induced blindness at lag 8. However, large differences in valence and arousal in the stimuli may cause a ‘forced regression’, in which we observe a regression due to there being two distinct clusters of data points. Therefore, to circumvent this, we repeated the analysis using negative distractors only. The analyses using negative distractors revealed extremely strong evidence that valence and arousal uniquely contributed to an image’s distractibility at lag 2, whereas at lag 1 there was ambiguous evidence for either the null model or for a model with both valence and arousal. As predicted, at lag 8 the null model was preferred.

Thus far, the data suggest that emotion-induced blindness at lag 2 is most sensitive to the emotionality of the images, with both its valence and arousal, suggesting the impact of other factors in other trial types.

Interestingly, we failed to replicate Onie and Most (2017)’s finding that negative accuracy or negative accuracy while controlling for neutral accuracy predicted negative affect.

Experiment 2

Following Experiment 1’s finding that emotion-induced blindness is sensitive to stimulus valence and arousal, in Experiment 2 we investigated whether the Dot Probe was similarly sensitive to the valence or arousal of the same stimuli.

Method

Participants. We recruited as many participants as possible prior to the end of the calendar year (i.e., prior to the unavailability of participants due to the holidays), with an ambition to roughly match the sample size of participants in Experiment 1. However, this period ended prior to our reaching this goal. We did not deem this to be a major concern, as our use of Bayesian statistics should provide an indication of whether we had achieved enough 110 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE precision in the specific analyses we conducted (e.g., see Onie, Notebaert, Clarke, & Most,

2019).

We recruited 46 participants (21 males, 25 females; Age: M = 23.38 years, SD = 3.31 years, Range = 19 - 37 years) from the general community through a paid recruitment system.

Participants were reimbursed $10 for 45 minutes of their time. This study was carried out in accordance with the recommendations of ‘the Human Research Ethics Advisory Panel at UNSW’ with written informed consent from all subjects.

Design. To investigate whether valence or arousal of the images accounted for attentional capture in the Dot Probe, participants completed one session of the Dot Probe task using the images rated in Experiment 1.

Materials. Images and hardware used in the Dot Probe were the same as in Experiment

Procedure. At the beginning of the task, participants were instructed to indicate the left/right direction of an arrow probe using directional keys, and to try ignoring the images that would appear on the screen prior to the target arrow. They were also told that if they took too long or responded incorrectly, there would be a 3s penalty before the next trial.

On each trial, a fixation cross appeared at the center of the screen for 100ms, followed by two 11.5 cm × 8.5 cm images (a neutral and negative image) placed with their medial edges

7.5 cm above and below central fixation (15cm from one image’s edge to another). We used a

500ms exposure time for the images, consistent with past studies (Clarke et al., 2017; Koster et al., 2005; C. MacLeod et al., 1986; Sandersan Onie et al., 2019). The images disappeared to reveal an arrow (3.5 x 3.5 cm) behind one of the images, which pointed either left or right and CHAPTER 4111 remained until participants made a response indicating the arrow’s direction. The two images were placed at the top and bottom of the screen rather than left or right as per recommendations to improve task reliability (Price et al., 2015).

Participants completed 384 trials with a short break after every 96 trials, forming four blocks. Trial type (Congruent, in which the probe appeared behind the negative stimuli, vs

Incongruent, in which the probe appeared behind the non-negative stimuli) was randomly allocated within each block. 75 negative and 75 neutral distractors were used, resulting in each distractor being presented on average approximately 5 times during the experiment.

Results

Data Preparation. Trials in which RTs were longer than 3 seconds and shorter than 0.3 seconds were eliminated.

Dot Probe. We first investigated whether negative images captured attention by performing paired samples t-tests, testing the hypothesis that RTs on congruent trials would be slower than RTs on incongruent trials (RTCong < RTIncong). The analysis revealed moderate evidence that this was the case (BF10 = 3.084, � = -0.312, 95% CI [-0.603, -0.054]), providing moderate evidence that participants responded more slowly on incongruent trials than on congruent trials. See Table 7 for descriptive statistics of the response times.

Table 7. Response Time Means and Standard Deviations for Dot Probe Incongruent Congruent

Mean 483.6 480.2

SD 69.72 69.33

Minimum 402 399.7 112 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

Incongruent Congruent

Maximum 785 784 Note: Units are in ms.

Do valence and arousal ratings predict attentional performance in the dot probe?

We investigated whether valence or arousal ratings for each negative image accounted for attentional capture, hypothesizing that: if attentional capture was driven by distractor valence or arousal, then more negative images should elicit quicker RTs on congruent trials and slower RTs on incongruent trials. Therefore, we obtained an average RT for when each negative distractor appeared in a congruent trial, and for when each negative distractor appeared in an incongruent trial. Note that a key difference between congruent and incongruent trial types was that the emotional stimulus could elicit either a shortened or a prolonged response latency respectively (assuming the predicted dot probe main effect). Therefore, for incongruent trials, we flipped the sign of the valence and arousal ratings to create an attentional bias index that incorporated both types of trial. Across all participants, on average, each negative image appeared 115 times in a congruent trial, and 115 in an incongruent trial.

We then regressed valence and arousal ratings on RT. The analysis suggested that the null model best accounted for the data, outperforming a model with both valence and arousal by

2.966 times, a model with just valence by 3.067 times, and a model with just arousal by 3.630 times. Therefore, the analysis gives marginal to moderate support that neither valence, arousal, nor a combination of the two accounts for performance in the Dot Probe.

In the analysis above, we assumed that any attentional effects were driven purely by the valence and arousal of the emotional stimulus. However, the Dot Probe presents both negative CHAPTER 4113 and neutral stimuli on each trial, and variability in valence and arousal of the neutral stimulus may add noise to the analysis above. Therefore, below, we incorporated the ratings of the neutral stimuli in the analysis to increase the resolution of our analysis. That is, for every combination of neutral-negative distractors present in the experiment, a difference score for valence and arousal value, as well as average RT, was calculated.

We regressed valence and arousal on RTs for each combination of images. The analysis once again suggested that the null model best accounted for the data, outperforming a model with just valence by 12.513 times, a model with just arousal by 20.414 times, and a model with both valence and arousal by 20.488 times. Therefore, this analysis bolters the finding that Dot

Probe performance – at least in this experiment – was not sensitive to stimulus valence or arousal.

Discussion

In Experiment 2, we investigated whether valence or arousal predicted attentional capture indexed by RTs in the Dot Probe. A paired samples t-test revealed moderate evidence that participants were responding faster on congruent trials than on incongruent trials, suggesting that the Dot Probe may be sensitive to overarching categorical differences in emotion. However, an analysis using the valence and arousal of the negative images revealed marginal to moderate evidence that neither valence nor arousal accounted for this pattern.

Further analysis considering ratings of the neutral images yielded stronger evidence that this was the case. Therefore, our evidence suggests that while the Dot Probe appeared to reflect sensitivity to categorical differences in valence and arousal, it was not sensitive to finer gradations of these dimensions. 114 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

Additionally, an exploratory analysis suggests that the more negative the distractor image, the slower the responding. This suggests the negative image was processed to some degree.

General Discussion

In this study, we investigate whether ratings of valence and arousal predict attentional capture in two paradigms: emotion induced blindness and the Dot Probe. In the first study, we collected distractor ratings on MTurk, and recruited a separate pool of participants to the lab to complete a session of emotion-induced blindness. This resulted in a replication of the emotion- induced blindness effect, whereby negative distractors caused more impairment than neutral distractors at early lags. Multiple regression analyses revealed marginal and moderate evidence, at lag 1 and lag 2 respectively, that valence predicted distractor impairment; however, the R2 value and Bayes Factor were larger at lag 2. At lag 8, there was strong evidence that neither valence nor arousal predicted impairment. This is consistent with the preceding analysis in which there was no longer a distractor effect at lag 8. In a follow-up analysis, in order to circumvent a “forced regression” that would have stemmed from the negative- and neutral- sets of stimuli being extremely different in both valence and arousal, we repeated this analysis using only the negative stimuli and found that, at lags 1 and 8, the null model best accounted for impact; however, at lag 2 there was extremely strong evidence that valence and arousal uniquely contributed to attentional capture. One possibility as to why we observed an emotion- induced blindness effect at lag 1 without observing a predictive relationship with valence or arousal ratings of the distractors is due to a ceiling effect, adding another source of variability. CHAPTER 4115

This is consistent with our finding that emotion-induced blindness appears to be smaller at lag 1 than at lag 2.

In this experiment, we did not find a relationship between emotion-induced blindness performance on negative trials with or without controlling performance on neutral trials, in contrast to Onie & Most (2017). However, unlike Onie and Most’s findings, our principal components analysis did not separate questionnaires indexing negative affect and persistent negative thought into separate components. Therefore, we are unable to make a direct comparison between the two studies.

In the second experiment, we investigated whether ratings of valence and arousal predicted response times in the Dot Probe. A paired samples t-test revealed moderate evidence that participants responded more slowly on incongruent trials (where the target appeared opposite the emotional picture) than congruent trials (where the target appeared at the same location as the emotional picture). However, the Bayes Factor for this analysis revealed that the alternative hypothesis was only slightly more than three times the probability of the null hypothesis and must therefore be interpreted with caution. Although this pattern appeared consistent with a predicted dot probe effect, multiple regression analyses provided moderate to strong evidence that the effect did not seem to be linked to valence or arousal ratings.

Therefore, there are at least two possible explanations for why there was an overall congruence effect in the dot probe when neither valence nor arousal predicted performance: one possibility is that negative images indeed did attract spatial attention in the dot probe, resulting in slower responding on incongruent trials, but that this was driven by non-valence-, non-arousal- related features such as colour differences. In principle, this would be consistent 116 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE with the wider literature in which evidence is mixed regarding the degree to which the dot probe is a reliable or sensitive measure of emotion processing (see Bar-Haim, Lamy, Pergamin,

Bakermans-Kranenburg, & van IJzendoorn, 2007; Kruijt, Parsons, & Fox, 2019; Schmukle, 2005).

If the dot probe is indeed a relatively insensitive or unreliable index of preferential emotion processing, one underlying reason might be because the allocation of spatial attention itself is driven by multiple concurrent processes, such as decisional mechanisms about whether to direct attention towards or away (e.g., Onie & Most, 2017). The fact that emotion-induced blindness does not involve the allocation of spatial attention may be one reason why it seems to provide a cleaner index of emotion prioritization, as reflected in its higher test-retest reliability. Nevertheless, the findings that neither the valence nor arousal predicted attentional capture in the dot probe underscore the need to improve spatial attention measures, a sentiment echoed by a recent review of the literature (MacLeod et al., 2019). Note that despite the use of Bayes Factors, experiment 2 had a small sample size, and thus its findings must be taken with caution. We recommend a replication with a larger sample size.

A key strength of this study is that ratings were not obtained from the individuals who performed the attentional tasks, but rather a separate population. One risk in having the same sample perform the task and rate the images is the possibility of carry over effects from one task to another. On the other hand, since a separate group of participants rated the stimuli in the current study, this precludes our ability to make inferences about whether stimuli the individual finds subjectively emotional captures their attention. This is especially true with the

Dot Probe, in which past research has repeatedly found that heightened spatial vigilance to threat is observed when it relates closely to participants’ predominant worry (Amir, Beard, CHAPTER 4117

Burns, et al., 2009; Mathews & Macleod, 1985). Therefore, while overall negativity may capture attention in the Dot Probe, each distractor may carry different weight for each individual – resulting in unaccounted variance that could not be capture in our analysis.

One future study that might provide further insight into the contribution of valence and arousal to attentional biases would be to incorporate positive stimuli as well. The scope of this study was limited to negative stimuli due to the focused interest of the attentional bias field; however, a study incorporating positive stimuli would give more insight to the processes underlying attentional capture by emotion. This way, we could see whether attentional capture in spatial and temporal mechanisms of attention are driven by the same or different combinations of valence and arousal for positive and negative stimuli, relative to neutral stimuli. Also, even if it is driven by the same factors, the degree to which they impact attentional capture may differ.

In conclusion, the present study suggests that while the Dot Probe may not be sensitive to fine gradations of valence and arousal, emotion-induced blindness is relatively sensitive to such gradations. In particular, the results of this study show that emotion-induced blindness at lag 2, using only negative distractors, is sensitive to both valence and arousal. Moving forward, we recommend that future research drawing links with constructs such as attentional bias in anxiety consider whether the tasks employed are sensitive specifically to valence, to arousal, or to a combination of both.

118 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

Chapter 5:

Investigating the Locus of Competition in Emotion-Induced Blindness

Currently in preparation for submission as Sandersan Onie, Colin MacLeod & Steven B. Most (in prep). CHAPTER 5 119

The candidate designed the study with CM and SBM, and the candidate was responsible for programming, data collection, data analysis and manuscript preparation

Investigating the Locus of Competition in Emotion-Induced Blindness

In the previous chapters, we found that the perceptual impairments reflected in emotion-induced blindness seemed to be driven by the valence and arousal of the distractors, and that indices of emotion-induced blindness had higher test-retest reliability than other attentional bias tasks. In this chapter, we seek to further understand the underlying processes in emotion-induced blindness. Specifically, whether the competition occurs early or late in visual processing.

In addition to investigating links between attentional effects and individual differences linked with psychopathology, it is of critical importance that we also investigate the underlying mechanisms of the tasks used. One reason is that by understanding the attentional processes engaged in the task, we have a deeper understanding about the attentional dysfunctions in disorders. For example, in a recent study, Price, Brown and Siegle (2019) attempted to break down the response process in the Dot Probe using computational modelling, in a bid to investigate the mixed findings in the attentional bias literature. Consistent with theories of the

Dot Probe, the authors found that congruent and incongruent trials did not differ in terms of how well participants were able to respond and process the probe, or how cautious participants 120 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE were in responding, but differed on a parameter believed to include the process of shifting attention in space. This is consistent with the idea that RTs measured in the Dot Probe reflect the shifting of spatial attention and not other processes – bolstering our confidence on conclusions based on findings with the Dot Probe. This is an example of how a study peeling away at surface level effects and into deeper mechanistic processes are able to bolster our understanding of emotional dysfunction.

A core question in the underlying mechanisms of emotion-induced blindness is whether the observed effects reflect an early or late selection mechanism, as this distinction may be an indicator of the task’s malleability and by extension, its utility in attentional bias modification.

Early-selection, as proposed by Broadbent (1958) suggests that there is a limited perceptual capacity, and therefore the attentional ‘filter’ occurs after basic features, but before semantic meaning has been processed. On the other hand, late-selection, as proposed by Deutsch and

Deutsch (1963) suggests that attention is an unlimited resource and that selection occurs after all processing is completed. The perceptual load theory was introduced to bridge these two theories suggesting that early selection was observed under high perceptual load due to its limited capacity, but given low perceptual load, attentional mechanisms could operate in a way similar to what is observed in late selection (Lavie, 1995; Lavie & Tsal, 1994).

There are two lines of work which suggest that late selection mechanisms may be more malleable. The first comes from findings in the literature that suggest that early perception is impenetrable by later, higher-order cognitions (Pylyshyn, 1999). The author provides evidence that perception should not be viewed as a continuous early stage of cognition, thereby allowing cognition to actively influence early perception. However, perception should stand alone and CHAPTER 5 121 only links with cognition through well-defined interactions. This sentiment is echoed in a more recent paper by Firestone and Scholl (2019) where the authors highlight a myriad of pitfalls when generating the conclusion that cognition is able to influence early perception. For example, the authors argue that many studies confound perception and judgement in their design, so that even if participants judge a situation differently due to outside factors, it does not necessarily give support that their perception changes. The second comes from the malleability of the attentional blink. The attentional blink is the failure to report the second of two targets in a stream of stimuli when the two targets are close in temporal proximity

(Raymond et al., 1992). A key point is that the attentional blink is proposed to occur later in visual processing due to central processes such as bottlenecks gating access to working memory

(e.g., Chun & Potter, 1995; Jolicœur & Dell’Acqua, 1998; M. C. Potter, Staub, & O’Connor,

2002), or errors in target retrieval from memory (K. L. Shapiro, Raymond, & Arnell, 1994).

(Slagter et al., 2007) conducted a study and found that mental training over a three-month period reduced the magnitude of the attentional blink. Furthermore, past studies have also found that repeated attentional blink training was also able to reduce the effect (Choi, Chang,

Shibata, Sasaki, & Watanabe, 2012; Maki & Padmanabhan, 1994; Tang, Badcock, & Visser, 2014; but also see Enns, Kealong, Tichon, & Visser, 2017 for evidence for the contrary). Therefore, it appears that early perceptual processes seem more immune to modification.

The issue of locus of competition in emotion-induced blindness is inextricably linked with the core question of whether the emotion-induced blindness shares the same underlying mechanism as the attentional blink (Kennedy et al., 2014; J. MacLeod, Stewart, Newman, &

Arnell, 2017), due to the fact that a proposed core distinction between the two tasks is that the 122 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE attentional blink occurs at a later stage in the visual process, the emotion-induced blindness is postulated to occur at an earlier stage (Most & Wang, 2011). Evidence from this emerges in particular from investigating the spatial localization of each task, as the greater the spread of a temporal effect, the later and more central the attentional filter (Wyble & Swan, 2015). Indeed, a previous study using the attentional blink found that in a dual-stream version of the task, target impairment occurred even when the targets were in different streams (Holländer,

Corballis, & Hamm, 2005) suggesting a later, more central process. In contrast, in emotion- induced blindness, the impairment only occurred when the distractor occurred in the same stream (Most & Wang, 2011), with spatial localization suggesting a relatively early locus of competition. However, this method of investigation does not directly index how much processing had occurred before competition occurs.

Another way previous studies have investigated the locus of selection in the attentional blink more directly is modifying the task to see whether missed targets are able to prime a subsequent stimuli or probe. In the study, the participants were asked to report three letters in a stream of numeric characters. Traditionally, the attentional blink task only contains two targets, but inclusion of a third target would allow investigation of whether the second target would prime detection of the third target. The third target would either match the second target, in which the third target would be a lowercase version of the second target, or a mismatch, in which the third target would be a lowercase version of another letter. The authors found that even if participants had not been able to correctly report the second target, if it matched the third target, participants were more likely to correctly report the third target, suggesting that the second target had primed detection of the first target, whether participants CHAPTER 5 123 could accurately report it. This suggests that missed targets in the attentional blink had been processed to a substantial degree (Shapiro, Driver, Ward & Sorensen, 1997). These findings have been replicated several times in the literature (e.g. Harris & Little, 2010; Rolke, Heil, Streb,

& Hennighausen, 2001; Visser, Merikle, & Di Lollo, 2005).

In this study, we investigated whether missed targets in emotion-induced blindness are processed sufficiently to prime subsequent responses, as missed targets have been found to do in previous attentional blink studies (e.g. Shapiro, Driver, Ward, & Sorensen, 1997). If, instead, emotional dictators interfere with target perception at a relatively early point in processing, we should observe a lack of priming caused by missed targets. In this study, we held the distractor- target temporal separation constant at lag 1, as previous studies have found a relationship between lag 1 of emotion-induced blindness and clinically relevant individual differences (Onie

& Most, 2017).

Methods

Participants

We recruited 111 participants (Mage = 22.18, SDage = 4.19, Range = 18 – 48, Men = 41,

Women = 70) from the general community and compensated them AUD$15 for their time. The experimental protocol was approved by the UNSW Human Research Ethics Advisory Panel.

The pre-registered aims, methods and analyses of this study can be found at https://osf.io/a8c5u

Design

We assessed whether missed targets during emotion-induced bilndness caused priming by measuring the degree to which the targets’ rotation to the right or left facilitated or 124 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE interfered with a speeded judgment about whether a subsequently presented arrow pointed right or left. The left- or right-pointing arrow appeared immediately after each emotion- induced blindness trial and remained visible until the participant’s response, and response time served as the primary measure of interest.

Materials and Hardware

Images for the distractor ratings and emotion-induced blindness tasks were collected from the International Affective Picture System (IAPS; Lang, Bradley, & Cuthbert, 1997), with selection of the images guided by whether they had been rated as neutral or negative in the

IAPS norms (Lang et al., 1997), and these were supplemented by images with similar content from publically available sources. A total of 150 images were collected (75 negative and 75 neutral). The negative images depict gore and disgust images (e.g. mutilated corpses and soiled toilets). Neutral images depict similar objects in non-negative settings (people in everyday scenes as well as household objects). These images were validated in Chapter four of this thesis, in the Distractor Ratings subsection of the Results section.

Individual differences in anxiety were indexed using the Generalized Anxiety Disorder 7 scale (GAD7; Spitzer, Kroenke, Williams, & Löwe, 2006) CHAPTER 5 125

Stimuli were presented on a 24-in. BenQ XL2420T LED monitor with 1920×1080 resolution and 120-Hz refresh rate. Head position was not fixed. The GAD 7 questionnaire was administered using Qualtrics (Qualtrics, Provo, UT), an online survey program. The emotion- induced blindness task was administered, and responses were collected, using the

Psychophysics Toolbox extensions for MATLAB (Brainard, 1997; Keleiner et al., 2007; Pelli,

1997).

Procedure

Participants first completed the GAD 7, followed by two practice phases of the task and one experimental phase.

In the first practice phase, participants responded to 50 trials in which they searched for rotated target in each stream of images and then selected it from a line-up on a subsequent display (the ‘answer screen’). Each emotion-induced blindness stream ended with three images following the target. The answer screen contained 4 possible choices (one ‘foil’, which had not been in the stream, and the image that had appeared in the stream, both presented in left- and right- rotated orientations). Each answer screen also allowed participants to instead indicate ‘I did not see a rotated target image in that trial’. Following the selection, either ‘CORRECT’ or

‘INCORRECT’ flashed up on the screen depending on the response. Answers were only correct if participants selected the correct image in the correct orientation. The purpose of this phase – in addition to giving participants opportunity to acclimate to the task – was to render the target’s orientation salient, a necessary pre-requisite for the priming manipulation. 126 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

In a second practice phase containing 12 trials, we introduced the post-stream arrow judgment task at the end of each trial. After each stream’s final image, but prior to the answer screen, an arrow pointing left or right appeared for one second or until participants’ keypress response indicating the direction in which the arrow was pointing (D for left and F for right).

Participants were instructed to respond to the arrow as accurately as possible and before the arrow disappeared.

In the third, non-practice phase, distractor images could appear one image before the target (lag 1). Each trial contained either a negative or neutral distractor or, in a baseline condition, no distractor (where the distractor’s serial position was occupied simply by a landscape image drawn from the same image set as the other landscapes in the stream).

Participants completed 80 trials of each trial type, for a total of 240 trials, with a short break after every 60 trials. For a schematic of a trial, see Figure 9 below.

Distractor Target

Arrow

100ms/ image Until Response

Figure 9. Schematic Representation of the modified emotion-induced blindness trial. Participants would view the stream and respond to the arrow at the end of the stream before selecting the target in the answer screen. Note that the target and arrow are pointed in the same direction, thus making this a congruent trial.

CHAPTER 5 127

Note that participants in most emotion-induced blindness studies indicate target orientation on each trial by pressing the left or right arrow keys (Most et al., 2005; Onie &

Most, 2017; Jin et al., 2018). However, in the present study participants reported the rotated target via the answer screen instead, to avoid response competition between responses in the distractor and the arrow judgment tasks. For an example of the answer screen, see Figure 10 below.

I did not see a rotated target image in that trial

Figure 10. Example Answer Screen of the modified emotion-induced blindness task. Participants indicate their answer by clicking on one of five options.

Results

We analyzed the data using JASP with JZS priors for ANOVA (Rouder & Morey, 2012), and interpreted the Bayes Factors according to Jeffrey’s scale (Jeffrey, 1961). Bayes factor for each effect was derived by dividing the Bayes factor of a model with the effect with the model without the effect, similar to a chi-squared ratio test (e.g. for effect of A: Bayes Factor of model containing A and B / Bayes Factor of model containing B; (Mathôt, 2015). Effect sizes are reported as Cohen’s � and interpreted according to Cohen’s scale (Cohen, 1988): 128 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

� = 0.2: small, � = 0.5: medium, � = 0.8: large.

Data Cleaning

We first eliminated trials in which the participant took longer than one second to respond to the arrow, as they were instructed to respond within one second.

Emotion-Induced Blindness

We first tested for the presence of an emotion-induced blindness effect using a

Bayesian mixed effects ANOVA, entering valence as a within-subjects factor. The analysis

47 revealed extremely strong evidence for a main effect of distractor valence, BF10 = 3.359 x 10 , with follow up analyses confirming that participants performed substantially worse when there

21 was a neutral distractor relative to no distractor, BF10 = 1.302 x 10 , δ Base-Neut = 1.735, 95%CI

[1.438, 2.036], and when there was a negative distractor relative to a neutral distractor BF10 =

15 9.031 x 10 , δ Neut-Neg = 1.425, 95%CI [1.115, 1.731]. Consistent with this, participants performed substantially worse when there was a negative distractor relative to no distractor,

27 BF10 = 5.151 X 10 , δ Base-Neg = 2.155, 95%CI [1.824, 2.498]. Emotion-induced blindness percentage accuracies are given in Table 8.

Table 8 Accuracy Means and Standard Deviations for Emotion-Induced Blindness Negative Neutral Baseline

Accuracy 63.97% (20.05%) 75.17% (16.24%) 88.75% (9.37%) Note: Means are given, with standard deviations given in parentheses. All values are in percentages correct. Negative, Neutral, and Baseline refer to distractor type.

Priming CHAPTER 5 129

Response Times. In these analyses, we tested for the presence of a congruence effect,

whether this effect changed as a function of distractor valence, and whether persisted when

targets were missed. Response time means and standard deviations are reported in Table 8.

Table 9 Aggregated RTs to the arrow, binned according to congruence and correct/incorrect responses in the emotion-induced blindness task.

Distractor Correct Incorrect

Congruent Incongruent Incong - Cong Congruent Incongruent Incong – Cong

Negative 0.481 (0.084) 0.500 (0.091) 0.019 (0.049) 0.539 (0.106) 0.545 (0.098) 0.007 (0.073)

Neutral 0.461 (0.080) 0.484 (0.082) 0.024 (0.038) 0.526 (0.098) 0.541 (0.100) 0.013 (0.101)

Baseline 0.443 (0.073) 0.479 (0.08) 0.036 (0.037) 0.529 (0.124) 0.524 (0.111) -0.015 (0.126)

Note: Mean response times with standard deviations in parentheses. All values are in seconds. Negative, Neutral, and Baseline refer to the distractor type in the trial. Congruent and Incongruent refers to whether the orientation of the target matched the orientation of the arrow in the arrow task. Correct and incorrect refers to whether the emotion-induced blindness target was correctly identified.

We first performed a manipulation check by investigating whether there was a

congruence effect in the baseline condition using a Bayesian paired samples t-test using

baseline trials where participants had correctly reported the target image, testing whether

response times to the arrow were slower when the directions of the arrow and target were

incongruent. The analysis revealed extremely strong evidence that individuals were slower to

respond to the arrow on incongruent trials than on congruent trials, with a very large effect size

estimate, thus confirming that the experimental task was sensitive to congruence in the

14 absence of a distractor, BF10 = 2.547 x 10 , δ Incong-Cong = 0.929, 95%CI [0.700, 1.158]. 130 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

We then investigated whether the congruence effect was present in trials where distractors were present but where targets were perceived. We employed a repeated measures

ANOVA, entering in valence and congruence as fixed factors, for all trials in which the target landscape image was correctly identified. The term of interest is the valence x congruence interaction, indicating whether the congruence effect differs as a function of valence. The analysis revealed extremely strong evidence that reaction times were slower on incongruent

23 trials than on congruent trials, BF10 = 7.753 x 10 . There was also a main effect of valence BF10

= 5.027 x 1020, in which post hoc analyses revealed evidence that reaction time to the arrow slowed from baseline to neutral trials, BF10 = 12,000, and from neutral to negative trials, BF10 =

2.087 x 107. Finally, there was marginal evidence to suggest that there was a valence x congruence interaction, BF10 = 2.094. Follow up paired-samples t-test corroborated this account, suggesting that the effect size diminished when there was a distractor, and further

14 when there was an emotional distractor, Base: BF10 = 2.547 x 10 , δ Base = 0.932, 95%CI [10.710,

7 1.146], Neut: BF10 = 1.087 x 10 , δ Neut = 0.611, 95%CI [0.419, 0.815], Neg: BF10 = 418.491, δ Neg =

0.379, 95% CI [0.190, 0.573]. Therefore, there is marginal evidence that the congruence effect diminished as the distractor potency increased.

Finally, we investigated whether the congruence effects differed as a function of correctly identifying the target image using a repeated measures ANOVA, entering in valence, congruence, and target accuracy as factors. Note that in this analysis we did not include the baseline condition, as failures to see targets in the baseline condition would stem from other, non-perceptual-competition reasons (e.g., “zoning out”). The analysis revealed extremely strong evidence for a main effect of valence, BF10 = 223.755, in which RTs to the arrow CHAPTER 5 131 increased following negative (compared to neutral) distractors; a main effect of congruence,

BF10 = 199.380, in which RTs on incongruent trials were much slower; and a main effect of

41 whether participants had correctly identified the distractor, BF10 = 1.734 x 10 , such that RTs were much longer on trials where participants was not able to correctly identify the distractor.

Critically, there was moderate evidence suggesting that there was no valence x congruence x target identification interaction, BF01 = 6.494. That is, the marginal valence x congruence interaction that had been observed when participants were able to identify the target was not substantially changed when participants did correctly identify the target. Consistent with this, another repeated measures ANOVA entering in valence and congruence, but only using trials in which the participant could not report the target, provided moderate evidence against a congruence effect, BF01 = 8.547. Furthermore, there was strong evidence to suggest that this did not differ between different distractor types, BF01 = 16.67.

Overall, the results suggest that there was a congruence effect when the rotated target image was seen, and this may differ between distractor valences. However, there is some evidence suggesting that this congruence effect did not change when participants missed the target image in conditions where perceptual competition occurred (consistent with the current study’s finding of worse accuracy in distractor-present than distractor-absent trials, which in turn revealed that participants were not simply missing targets due to “zoning out”).

In this study, we did not eliminate trials where the arrow was reported incorrectly, opening up the possibility of a speed-accuracy trade-off. If this was the case, we would see an opposite pattern of findings for accuracy. Analyses confirmed that this was not the case, and patterns of finding for accuracy mirrored those of response times. 132 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

Attentional Bias and Anxiety

Emotion-Induced Blindness. We investigated whether indices of emotion-induced blindness predicted anxiety by regressing total GAD7 scores on negative emotion-induced blindness accuracy while controlling for accuracy on neutral trials. There was moderate evidence to suggest that accuracy following negative distractors when controlling for accuracy following neutral distractors did not predict anxiety, BF01 = 4.405, β = -0.105, 95 CI [-4.689

1.610].

Priming indices. We also investigated whether anxiety predicted the magnitude of the congruence effect when participants were unable to correctly identify the rotated target image.

The rationale for this analysis was to test whether the competition in emotion-induced blindness occurs at different stages as a function of anxiety, hypothesizing that there may be less priming among high anxious individuals due to the distractor image competing at an earlier stage of perceptual processing. We subtracted congruent RTs from incongruent RTs for both negative and neutral trials on trials where the target was incorrectly reported. This gives us a single index of how much priming occurs with each distractor type. We then performed a multiple regression, entering both negative and neutral priming indices in a model predicting

GAD 7 scores to investigate whether depth of processing predicted self-reported anxiety.

There was moderate evidence to suggest no relationship between anxiety and priming when targets were missed, BF01 = 5.494, β = 1.920, 95 CI [-2.282 12.103], suggesting that the locus of competition did not differ between those scoring high and low in anxiety.

Discussion CHAPTER 5 133

In this study, we investigated whether targets missed during emotion-induced blindness are nevertheless processed sufficiently to prime a response to subsequent stimuli. Although targets that were seen in this task did cause priming (indexed as slower response times when the left/right rotation of a subsequently presented arrow was incongruent with the left/right rotation of the target within the stream), the data revealed that priming did not occur when the targets were missed. Analyses taking into account self-reported anxiety provided moderate evidence against a relationship between emotion-induced blindness and anxiety in this sample, as well as against a relationship between anxiety and priming caused by missed targets.

Assessing whether priming can be caused by missed targets provided a means for assessing the locus of competition during emotion-induced blindness. Phenomenally, emotion- induced blindness is similar to the well-studied attentional blink, but some evidence suggests that emotion-induced blindness stems from different, earlier-acting mechanisms than the attentional blink (e.g., (Most & Jungé, 2008; Most & Wang, 2011; Wang et al., 2012). There isn’t universal agreement on this point, as evidence also exists that the two phenomena involve the same mechanisms (Kennedy et al., 2014; J. MacLeod et al., 2017). Evidence that the attentional blink stems from relatively late mechanisms limiting what can be consolidated into visual working memory (e.g., Chun & Potter, 1995) includes findings that missed targets during the attentional blink prime responses to subsequent stimuli (Harris & Little, 2010; Rolke et al.,

2001; Shapiro et al., 1997; Visser et al., 2005). In contrast, the absence of priming in the present study is consistent with a distinction between emotion-induced blindness and attentional blink mechanisms. 134 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

The present study yielded evidence against the relationship between emotion-induced blindness and anxiety. This is consistent with the findings in the preceding chapter, in which indices of emotion-induced blindness failed to predict negative affect. There may be several explanations for this finding, which stands in contrast to patterns reported elsewhere in the literature. For example, Onie and Most (2017) found a relationship between emotion-induced blindness and negative affect but used different measures than the GAD 7 used here. Instead, they used a component score from a principal components analysis to reduce dimensionality in the data. Therefore, the possibility remains that emotion-induced blindness (or one of many underlying mechanisms) may link to negative affect as a whole, rather than a single individual differences measure. However, note that the aforementioned study has not yet been replicated

(although other links to individual differences have been found; e.g. (Kennedy & Most, 2015b;

Olatunji et al., 2013, 2011) , and if there was a shared mechanism amongst different manifestations of negative affect in general, then this underlying mechanism would also need to be inconsistently present amongst these different manifestations. Another possible explanation is that, in line with Bar-Haim et al.’s (Bar-Haim et al., 2007) meta-analysis, attentional bias reliably occur only in clinical samples (e.g., Olatunji et al., 2011), whereas the studies in this thesis only used non-clinical samples.

However, note that more recent conceptualizations psychopathology also incorporate a dimensional approach of symptomology, suggesting that a continuous approach is also warranted (S. L. Brown, Svrakic, Przybeck, & Cloninger, 1992; Den Hollander-Gijsman et al.,

2012; Lebeau et al., 2012), which will also yield more power and precision (Altman & Royston,

2006; Dawson & Weiss, 2012). Indeed, a recent study also suggested that a dimensional CHAPTER 5 135 approach to anxiety and depression would yield greater power in research (Bjelland et al.,

2009). Therefore, the previous finding that emotion-induced blindness differs amongst clinically anxious individuals and healthy controls must be replicated.

It is important to note that the current assessment of priming in emotion-induced blindness differs from the attentional blink priming tasks. Specifically, the attentional blink priming tasks had three targets in the stream that all involved the same task demands (i.e., letter identification). Here, however, participants responded to the arrow using arrow key press, and to the target using mouse click. Nevertheless, despite this difference, analyses of the present data verified that a congruence priming effect emerged when targets were seen. This suggests that despite surface level task differences, the manipulation was valid.

The present study is also somewhat limited by the fact that only tested lag-1 was tested, leaving open the possibility that at lags that occur later in the temporal window corresponding to emotion-induced blindness, the locus of competition may shift and enable fuller processing of even the missed targets. This possibility remains an important one for future investigation.

In conclusion, emotion-induced blindness seems to occur early on in visual processing, in which target stimuli knocked out due to emotional prioritization were not processed substantially. Emotion-induced blindness is the first demonstration of early, non-spatial perceptual selection with task-irrelevant emotional stimuli, expanding the range of attentional mechanisms we can access, as well as bolstering the conclusions from past and future studies using emotion-induced blindness.

136 INVESTIGATING WHETHER VALENCE OR AROUSAL DRIVE ATTENTIONAL CAPTURE

Chapter 6:

Does a Single Session of Training Modify Emotion-Induced Blindness?

CHAPTER 6 137

Currently in preparation for submission as Sandersan Onie, Chris Donkin & Steven B. Most (in prep).

The candidate led the design of the study, programming, data collection, data analysis and manuscript preparation

Does a single session of Training Modify Emotion-Induced Blindness?

In the previous chapters, we found that emotion-induced blindness had favourable test- retest reliability compared to other attentional bias tasks, was sensitive to the gradations of valence and arousal of individual stimuli and reflected early perceptual competition. However, we have also found evidence against the relationship between emotion-induced blindness and negative affect measured by various self-report questionnaires. Therefore, while emotion- induced blindness seems to be a relatively reliable task, evidence from this thesis suggests its link with negative affect appears tenuous. However, past published studies have found attentional affects in emotion-induced blindness in a range of psychopathologies using clinical patients, including obsessive compulsive disorder, generalised anxiety disorder and posttraumatic stress disorder (for review. See McHugo et al., 2013). Moreover, emotion-induced blindness performance has been shown to link with subclinical measures of negative affect (e.g.

Most et al., 2005; Kennedy & Most, 2015; Onie & Most, 2017), suggesting a relationship between emotion-induced blindness and clinically relevant individual differences. Therefore, in 138 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS this chapter, we investigate the trainability of emotion-induced blindness to assess its utility for

ABM.

If attentional biases play a role in psychopathology, then one potential avenue for intervention could be the development of attentional retraining tasks that modify these processes with consequent clinical benefits. Attempts to retrain attentional biases are known within the literature as attentional bias modification (ABM). While retraining efforts are primarily dominated by the Dot Probe task (e.g. MacLeod, Rutherford, Campbell, Ebsworthy, &

Holker, 2002; as featured in Chapter 2), other forms of retraining task have emerged. For example, Notebaert, Clarke, Grafton & MacLeod (2015) developed a gamified approach called the person matching task, in which two faces appeared on the screen. One had a happy expression whilst the other had an angry expression. On each trial, each pair would be replaced with a new pair of faces, and participants were instructed to either indicate whether the person with the happy expression remained the same on this trial, or whether the angry person remained the same. This manipulated whether participants attended the positive or negative stimuli. The authors found greater attentional bias change following this training than the standard Dot Probe, using a standard Dot Probe assessment.

However, like the assessment of attentional bias, much of the attentional retraining literature has focused on modifying spatial attention, despite there being multiple distinct mechanisms of attention (Chun et al., 2011; Wyble & Swan, 2015). Emotion-induced blindness in particular, may be a promising candidate for attentional retraining due its demonstrated links with clinically relevant individual differences. That is, although data collected in the context of this thesis have not pointed to a relationship between emotion-induced blindness and anxiety, CHAPTER 6 139 several published reports have found that emotion-induced blindness is sensitive to a multitude of different disorders. For example, Olatunji, Armstrong, McHugo and Zald (2013) found that emotion-induced blindness differentiated veterans with and without PTSD when using combat- related distractors. Olatunji and colleagues (2011) found that emotion-induced blindness differed between individuals diagnosed with generalized anxiety disorder and healthy controls, with clinically anxious individuals demonstrating overall greater impairment. Furthermore, Onie and Most (2017) found strong evidence that emotion-induced blindness had links with both negative affect and persistent negative thought, which is a core characteristic of anxiety and depression. Therefore, if emotion-induced blindness has links with mechanisms more closely related to certain disorders, retraining emotion-induced blindness may yield greater clinical benefits. To that end, a key question is whether emotion-induced blindness is malleable at all.

Past studies have shown that under certain conditions, emotion-induced blindness can be either attenuated or amplified. Kennedy, Newman and Most (2018) found that warning participants of the upcoming distractor type (aversive, erotic or neural) could significantly improve target perception relative to trials in which no warning was present, although this did not eliminate emotion-induced blindness entirely. In another study, Most and colleagues (2010) found that people exhibited greater emotion-induced blindness while experiencing romantic insecurity, when their partners were rating the attractiveness of potential romantic rivals.

However, past studies have also shown that emotion-induced blindness may be immune to change. In a recent study, (Haddara et al., 2018) investigated the effect of shock-induced anticipatory anxiety on emotion-induced blindness. While they found that anticipatory anxiety prolonged emotion-induced blindness to later lags, there was virtually no difference at the early 140 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS lags whatsoever. Zhao and Most (2019) investigated whether increasing the frequency with which emotional distractors appeared would mitigate emotion-induced blindness, as it does in the other attentional paradigms such as the Stroop task (e.g. Cheesman & Merikle, 1986;

Lindsay & Jacoby, 1994). The authors found no evidence that increasing distractor frequency of the negative or neutral images modified emotion-induced blindness. Indeed, in the previous chapter, we found that emotion-induced blindness reflected relatively early perceptual competition, which has been argued to be immune to cognitive interference (Firestone &

Scholl, 2019; Pylyshyn, 1999). Therefore, while some studies suggest that emotion-induced blindness may be malleable (if not strongly so), other studies have provided evidence to the contrary. However, none of these studies have attempted to retrain emotion-induced blindness through repeated training akin to attentional bias modification.

In this study, across two experiments, we investigated the malleability of emotion- induced blindness in a training paradigm akin to attentional bias modification tasks. To our knowledge this is the first attempt to retrain attentional biases to emotional stimuli using a non-spatial task. We investigated whether requiring participants to consistently ignore negative distractors within a rapid serial visual presentation would decrease the degree to which negative distractors subsequently capture attention and thus the degree to which they elicited emotion-induced blindness. Therefore, in one training condition, all distractors were negative at lag 1. In a second training condition, to control for the possibility that training to ignore any distractor at all (i.e., non-emotional ones) would have the same effect, distractors at lag 1 were always neutral. Finally, a third condition addressed the possibility that any training-linked changes in emotion-induced blindness might arise simply due to exposure to negative CHAPTER 6 141 distractors. In this condition, negative distractors appeared but did not need to be ignored, as the trial was fashioned in such a way that the distractor would not interfere with detection of the target: with targets appearing outside the typical emotion-induced blindness window (lag 8) and with no masking items before or after the target.

There are several possible outcomes from this study. One is that training participants to ignore negative distractors might yield the greatest change in bias, followed by training to ignore neutral distractors, followed by mere exposure to negative distractors. This finding would suggest that actively suppressing attention to emotionally negative stimuli would yield the greatest training effects. Another possible outcome is that ignoring either negative or neutral distractors would attenuate emotion-induced blindness, whereas mere exposure to negative would not, and this would suggest that practice ignoring distractors could lead to attentional changes independent of the emotional quality of the ignored distractors. A third possible outcome is that emotion-induced blindness could be attenuated when people practiced ignoring negative distractors and when they were merely exposed to negative distractors, but not when they practiced ignoring neutral distractors. This would suggest simple desensitization to negative images through increased exposure. This last possibility may be unlikely, given that Zhao and Most (2019) found that the frequency of negative distractors did not mitigate emotion-induced blindness (though in the current study, the ratio between negative and neutral exposure is much greater). Finally, another possible outcome is that emotion-induced blindness is resistant to training, and that the relative impairment caused by negative versus neutral distractors remains the same regardless of training condition. This 142 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS would be consistent with the idea that emotion-induced blindness is an early perceptual process that cannot be modified.

Experiment 1

Method

Design. In this study, participants first completed an anxiety questionnaire, followed by a pre-training emotion-induced blindness assessment. They then underwent one of three training conditions, in which they: a) practiced ignoring a negative distractor on each trial

(Ignore Negative), b) practiced ignoring a neutral distractor on each trial (ignore neutral), or c) were simply exposed to a negative distractor on each trial without a need to ignore (negative exposure). They were then assessed for levels of emotion-induced blindness in a post-training assessment.

Participants. A power analysis indicated it would require approximately 400 participants to detect a 5% difference in accuracy for the training condition x pre-post interaction. This was a conservative approach, and we sought to collect as many participants as possible during the semester. Furthermore, we set out to employ Bayesian statistics to allow us to quantify strength of evidence.

In the end, we recruited 243 first year university students (Mage = 19.73, SDage = 3.91,

Range = 17 – 56 years, Males = 76, Females = 167) in exchange for course credit for 1.5 hours of their time. Data collection was run in accordance with UNSW Human Research Ethics Advisory

Panel guidelines and recommendations.

Materials. CHAPTER 6 143

Images. Images for the distractor ratings and emotion-induced blindness tasks were collected from the International Affective Picture System (IAPS; Lang et al., 1997), with selection of the images guided by whether they had been rated as neutral or negative according to the IAPS norms (Lang et al., 1997), and these were supplemented by images with similar content from publically available sources. A total of 150 images were collected (75 negative and

75 neutral). The negative images included gory and disgust images (e.g. medical trauma and soiled toilets). Neutral images were non-negative depictions of objects and people (people in everyday scenes as well as household objects). The emotional nature of these images was validated in Chapter four of this thesis, in the Distractor Ratings subsection of the Results section.

Emotion-induced blindness tasks also typically use pictures depicting landscapes and buildings in the stream. These images were collected from a variety of sources including Google searches and photographs taken by members of the lab.

Hardware. Stimuli were presented on a 24-inch BenQ XL2420T LED monitor with 1920 ×

1080 resolution and 120-Hz refresh rate, using Matlab via the Psychophysics Toolbox extensions (Brainard, 1997; Pelli, 1997; Keleiner, Brainard & Pelli, 2007).

Procedure.

To assess the effects of different types of training, the experiment was comprised of three distinct parts: a pre-training assessment phase, a training phase and a post-training assessment phase.

Throughout the experiment, participants were given only one set of instructions: to find the one rotated target image amongst each stream of images and indicate the orientation of 144 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS the target image using arrow key press. If they were correct, they were given auditory feedback in the form of a bell sound.

Pre-training Assessment. In this phase participants completed 120 trials with a short break after 60 trials. In total, there were 60 negative trials and 60 neutral trials, evenly distributed between the two blocks. The target appeared as the item immediately following the distractor on each trial (lag 1). This followed from Onie and Most (2017), in which the authors found that emotion-induced blindness at lag 1 predicted both negative affect and persistent negative thought.

Training. Following the pre-training assessment phase, participants completed 720 trials with a short break after every 60 trials. Participants were randomly allocated into one of the three training conditions based on their order of arrival at the lab (e.g., participants one and four and seven would have the same condition).

In the Ignore Negative training condition, all trials contained a negative distractor at lag

1. In the ignore neutral training condition, all trials contained a neutral distractor at lag 1. In the negative exposure condition, all trials contained a negative distractor, which appeared at lag 8.

Further, to ensure that participants did not need to ignore the distractor in this condition, blank screens were placed 400ms before and between 400 – 600ms after the target CHAPTER 6 145

a) Ignore Negative b) Ignore Neutral c) Exposure Negative

Negative Distractor Negative Distractor Neutral Distractor

Target Target 500ms blank screen

100ms/ image Target last item in 100ms/ image 100ms/ image stream

Figure 11. Schematic Representations of the different training conditions in Experiment 1. In the Ignore Negative condition, the target appeared one image after a negative distractor. In the ignore neutral condition, the target appeared one image after the neutral distractor. In the Negative Exposure condition, a negative distractor appeared, followed by two landscape images, followed by a 500 ms blank screen, and followed by the target. Having blank screens before and after the target facilitates detection and removes the impact of any distractors.

Pre-training Assessment. The post training assessment was identical to the pre-training assessment with the exception that a different set of distractor images were used.

Results

To prepare the data for analyses, we aggregated responses to obtain percentage accuracies pre- and post-training for each participant. All analyses were performed in JASP

(JASP Team, 2018) with default priors provided. Note that null effects will be reported as BF01, that is, the amount of evidence in favour of the null.

Emotion-induced blindness. We first investigated the presence of an emotion-induced blindness effect. We submitted pre-test emotion-induced blindness accuracies for negative and 146 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS neutral trials to a paired samples t-test with a default prior, testing the hypothesis that negative distractors caused more impairment (MNeg < MNeut). The analysis yielded very strong evidence that accuracies were lower on trials with negative distractors than trials with neutral distractors

16 (BF10 = 9.7417 x 10 , � = -0.849, 95% CI [-1.021, -0.677], MNeg = 69.79%, SDNeg = 9.384%, MNeut =

74.68%, SDNeut = 9.860%). For emotion-induced blindness means and standard deviations, see

Table 10.

Training. To investigate the effects of training on emotion-induced blindness, we submitted the data to a 3 (condition: Ignore Negative, ignore neutral, and negative exposure) x

2 (distractor valence: negative and neutral) x 2 (time: pre- and post-training). The output was interpreted using analysis of effects in which models with the effect are compared with models without the effect (Mathôt, 2017).

The analysis provided extremely strong evidence of a main effect of valence, in which averaging across two time points, negative distractors caused more impairment (BF10 = 8.159 x

1019), and extremely strong evidence for an effect of time, in which overall accuracy increased

32 from pre- to post-training (BF10 = 1.197 x 10 ). However, there was strong evidence against a valence, time and condition interaction (BF01 = 22.72), and moderate evidence against a time and valence interaction (BF01 = 7.936), suggesting that there was no change in the effect of valence as a function of time or training condition. Therefore, the analysis gives strong evidence that we were unable to change emotion-induced blindness through training.

Table 10 Experiment 1 Accuracy Means and Standard Deviations for emotion-induced blindness CHAPTER 6 147

Phase Training Condition Negative Neutral Pre -Training - 69.79% (9.384%) 74.68% (9.860%) Post-Training Ignore Negative 76.61% (13.57%) 79.49% (13.68%) Ignore Neutral 75.80% (11.59%) 80.69% (9.222%) Negative Exposure 75.28% (9.583%) 79.26% (9.529%) Note: Table values are emotion-induced blindness accuracies in percentages, of pre- and post- assessment. Ignore Negative refers to the training condition in which participants were only given negative distractors at lag 1. Ignore Neutral is the same but with neutral images. Negative Exposure refers to the training condition in which a negative distractor was used, at lag 8, with no masks before and after the target. Note that accuracies only reflect pre and post training EIB assessments and thus are all measured at lag 1.

Discussion

In Experiment 1, we investigated the malleability of emotion-induced blindness by giving participants practice ignoring negative distractors, practice ignoring neutral distractors, or exposure to negative distractors. While an emotion-induced blindness effect was established, analyses revealed that there was evidence against differential effects of the different training conditions on emotion-induced blindness. One pattern of findings was that there was an increase in accuracy from pre- to post- training regardless of training type. Inspection of means in Table 9 show that this increase was relatively uniform from one training condition to another.

Experiment 2

In Experiment 2, we investigated the possibility that training effects do occur for emotion-induced blindness, but that they are only observable at slightly later lags and therefore 148 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS had been missed in Experiment 1. Therefore, Experiment 2 employed lags 2 and 4 rather than lag 1. Furthermore, this experiment followed up on Experiment 1’s finding of an overall increase in performance from pre- to post- training by testing whether practice simply searching for a rotated target in the stream could reduce emotion-induced blindness. Towards this end, Experiment 2 included a fourth training condition in which there were no distractors

(baseline trials) and also assessed baseline at pre- and post-training assessments.

Method

For this experiment, we aimed to recruit as many people as possible before the end of semester. In the end, we recruited 113 first year university students (Mage = 22.692, SDage =

3.327, Range = 18 – 37, Males = 44, Females = 69) in exchange for course credit for 1.5 hours of their time. Data collection was run in accordance with UNSW Human Research Ethics Advisory

Panel guidelines and recommendations.

The design and materials were identical with Experiment 1 with the exception that the attentional bias assessments incorporated lag 2 and lag 4 instead of lag 1, and also included baseline trials – where no distractor was present in the stream. Each trial type had 40 trials, for a total of 200 trials. A short break was given after every 50 trials. Furthermore, in the training phase, a baseline (no-distractor) condition was included. For a schematic of the different training conditions, see Figure 12. CHAPTER 6 149

a) Ignore Negative b) Ignore Neutral

Negative Distractor Neutral Distractor

Target Target

c) Exposure Negative b) Baseline

Negative Distractor

4 x Blank Screen

Target is last item in stream

Figure 12. Schematics of different training conditions in Experiment 2. In the Ignore Negative condition, the target appears one image after a negative distractor. In the Ignore Neutral condition, the target appears one image after the neutral distractor. In the Negative Exposure condition, a negative distractor appears, followed by two landscape images, followed by a 500 ms blank screen, followed by the target. Having blank screens before the target is to facilitate target processing. In the baseline condition, there was no distractor present.

Results 150 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS

Emotion-induced blindness. To investigate the presence of an emotion-induced blindness effect prior to training, we performed a repeated measures Bayesian Analysis, with lag (lags 2 and 4) and valence (negative and neutral), only using pre-training data. The analysis revealed strong evidence for a critical valence x lag interaction (BF10 = 536.2), characteristic of emotion-induced blindness.

Post hoc t-tests which tested the hypothesis that neutral accuracy is higher than negative accuracy, revealed that at lag 2 there was extremely strong evidence to suggest that there was more impairment when there was a negative distractor relative to a neutral

12 distractor, BF10 = 2.274 x 10 , � = -0.833, 95% CI [-1.047, -0.627]. At lag 4, there was extremely strong evidence to suggest there was more impairment with negative distractors relative to neutral distractors, BF10 = 2241.711, � = -0.419, 95% CI [-0.607, -0.234]. Therefore, consistent with the valence x lag interaction found above, there was lower accuracy on negative trials, and this effect seemed to decrease from lag 2 to lag 4. Mean accuracies can be found in Table 11.

CHAPTER 6 151

Table 11 Experiment 2 Accuracy Means and Standard Deviations for emotion-induced blindness

Phase Training Condition Negative Neutral Baseline

Lag 2 Lag 4 Lag 2 Lag 4

77.33% 85.61% 84.76% 88.60% 89.36% Pre-Training - (9.88%) (7.50%) (7.56%) (6.58%) (6.45%) 83.00% 85.25% 84.19% 87.38% 87.50% Post-Training Ignore Negative (8.47%) (7.90%) (10%) (8.66%) (7.77%) 78.29% 85.25% 85.04% 87.04% 89.50% Ignore Neutral (8.99%) (8.58%) (7.12%) (9.36%) (7.69%) 81.67% 86.30% 81.93% 86.78% 86.74% Negative Exposure (7.40%) (8.05%) (9.07%) (7.49%) (7.75%) 72.54% 84.23% 78.85% 78.85% 85.54% Baseline (7.28%) (9.67%) (9.28%) (9.28%) (8.81%) Note: Table values are emotion-induced blindness accuracies in percentages, of pre- and post- assessment. Ignore Negative refers to the training condition in which participants were only given negative distractors at lag 1. Ignore Neutral is the same but with neutral images. Negative Exposure refers to the training condition in which a negative distractor was used, at lag 8, with no masks before and after the target. Baseline refers to the training conditions in which there were no distractors.

Training. We then investigated whether different training conditions had differential impact on emotion-induced blindness by using a repeated measures Bayesian ANOVA, with time (pre- and post- training), lag (lags 2 and 4), and valence (negative, neutral, baseline) as within subject factors, and training condition as a between subject factor. As baseline trials have no distractor – and therefore have no lag – the experiment was programmed in such a way that each baseline trial had a lag identifier attached to the trial to enable the use of a repeated measures ANOVA.

The analysis revealed moderate evidence for an effect of time (BF10 = 7.906), in which accuracy decreased from pre- to post - training. Furthermore, while there was strong evidence 152 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS

against a time x lag interaction (BF01 = 8), there was strong evidence for a time x valence interaction (BF10 = 23.826) in which accuracy on baseline and neutral accuracies decreased from pre- to post- training, but accuracy on negative trials remained the same. Critically, there was moderate evidence against the presence of a time, lag, training condition, and lag interaction

(BF01 = 9.174), a time x training condition x valence effect (BF01 = 27.78) and a time x training condition x lag effect (BF01 = 12.05). Therefore, there is moderate to strong evidence to suggest that there was no effect of the different training conditions on emotion-induced blindness.

Discussion

In Experiment 2, we replicated emotion-induced blindness prior to training and investigated the possibility that training effects were present, but that they had been masked by ceiling effects at lag 1 in Experiment 1. Therefore, we used lags 2 and 4 in the assessment tasks. Furthermore, due to the overall increase in accuracy in Experiment 1, we investigated the possibility that training to identify a target in a stream could account for this increase and included a fourth training condition in which participants only performed baseline trials.

Once again, the analysis indicated that we replicated the emotion-induced blindness effect, but there was moderate evidence against any differential effects of the training conditions. Therefore, it appears that the relative prioritization of negative and neutral distractors could not be shifted even when assessment occurred at lags 2 and 4.

General Discussion CHAPTER 6 153

In this study, we investigated the trainability of emotion-induced blindness. In the first experiment, we only used lag 1 while randomly assigning individuals to one of three training conditions: a) training to ignore negative distractors, b) training to ignore neutral distractors, and c) exposure to negative distractors, in which the negative distractor was shown, but the target was always easy to identify. The analysis yielded very strong evidence for the existence of an emotion-induced blindness effect prior to training but also yielded strong evidence that there were no differential effects of training whatsoever.

In the second study, we considered the possibility that training effects had been present but were not observable as early as lag 1, and additionally tested the possibility that simply searching through a stream of images reduced susceptibility to emotion-induced blindness. The analysis again yielded strong evidence for a pre-existing emotion-induced blindness effect but yielded strong evidence against any differential effects of training even when tested at lags 2 and 4. Together, the analyses suggest that these methods of single-session training failed to reduce emotion-induced blindness at any of the early lags.

There are several possibilities for why we observed no differential training effects. For example, there may have been a lack of motivation. The differential training conditions were designed based on the assumption that participants would exert effort ignoring negative stimuli and would exert no such effort in the no-distractor training or negative exposure conditions.

However, if participants had simply not exerted any effort whatsoever, then we would not see any differential effects of training, consistent with the findings that emerged. However, in a previous study, Most and colleagues (2007 offered a large monetary incentive (up to USD$90), and found that even with large incentives, individuals still exhibited emotion-induced blindness. 154 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS

Therefore, there is reason to suspect that motivation may not have played a major role.

Another possible reason for the absence of training effects is that the “dosage” of training was insufficient. While the number of training trials in this study (720) is higher than typical ABM studies (e.g. 384 training trials in Notebaert et al, 2015; 578 training trials in Chen, et al., 2015), the robustness of emotion-induced blindness may suggest that it is more difficult to modify.

Therefore, emotion-induced blindness may require a much greater dose. Future studies testing whether training can modify emotion-induced blindness should consider increasing the amount or duration of training involved.

Finally, another possibility for the absence of training effects is in any training condition, the temporal distance between the distractor and target is held constant. Therefore, the participant may be able to use the distractor image as a temporal marker for the target as has been shown in the attentional blink (Nieuwenstein, Chun, Van Der Lubbe, & Hooge, 2005).

Furthermore, since greater attention is already allocated towards the negative distractors, it may actually increase attention towards threat, washing out any potential benefits of training.

Future studies should consider this possibility and check whether the participant was aware of the contingency between distractor and target.

At this point, it seems unclear what the common thread is between studies that were able to modify emotion-induced blindness, and studies that were not. Indeed, it seems that some manipulations are able to modify emotion-induced blindness (e.g. Most et al., 2010;

Kennedy et al., 2018) while other are not (e.g. Zhao & Most, 2017; Haddara et al., 2018).

Further investigation is required to elucidate under what specific conditions emotion-induced blindness is modulated. CHAPTER 6 155

An interesting finding is that in the first experiment, there was extremely strong evidence that overall accuracy improved from pre- to post- training; however, in the second experiment, there was moderate evidence that accuracy decreased from pre- to post-training.

The key difference is that in the first experiment, lag 1 was used for assessment, while in experiment 2, lags 2 and 4 were used. One possibility is that while practice effects were present in all lags, lags 2 and 4 were much more susceptible to fatigue. Emotion-induced blindness is typically smaller at lag 1 than at lag 2, suggesting that it some form of ceiling may be encountered, and that lags 2 onward may be more open to effects of cognitive control.

However, more research is required to understand the relative impact of external factors on each lag e.g. training and cognitive control.

The findings in this study are consistent with findings from the attentional blink literature. While some studies show that the attentional blink can be reduced through training

(Braun, 1998; Choi et al., 2012; Taatgen et al., 2009), other studies have shown otherwise

(Martens et al., 2006). However, one weakness of these studies is that there were possible ceiling effects, in which the attentional blink effect was eliminated due to the fact that accuracy for both the first and second target approached a ceiling (Enns et al., 2017). Enns and colleagues (2017) corrected for this by making the task more difficult for individuals who were approaching ceiling, and as a result, still found an attentional blink effect after training. Given we found that emotion-induced blindness is also immune to training, it is possible that temporal tasks are more stable and robust, but less malleable, than spatial attention tasks.

While efforts to reduce emotion-induced blindness through training were unsuccessful in the present experiments (which are the first to have tried this), the robustness and relatively 156 SINGLE SESSION TRAINING EMOTION-INDUCED BLINDNESS high test-retest reliability of emotion-induced blindness nevertheless suggests that it may be a useful tool for individual differences research. For example, Jin, Onie, Curby and Most (2018) found that emotion-induced blindness was smaller among violent video-gamers that among non-gamers even when controlling for other violent media consumption. While further investigation is needed to elucidate the relationship between emotion-induced blindness and individual differences in anxiety, emotion-induced blindness may be a useful tool in assessing other individual differences.

In conclusion, these preliminary, single-session efforts to retrain emotion-induced blindness were unsuccessful, as we were unable to modify the relative impairment of negative and neutral distractors from pre- to post- training. However, this is consistent with the idea that emotion-induced blindness reflects stable early perceptual processes, which may be resistant to retraining.

CHAPTER 7 157

Chapter 7:

General Discussion

158 GENERAL DISCUSSION

General Discussion

Over the course of five studies, we investigated the links between attentional bias towards threat as well as the attempted modification of said biases in both spatial and temporal attentional tasks. In addition, we investigated the psychometric properties of the emotion-induced blindness. Note that, in line with the attentional bias literature, I use the term

‘attentional bias’ to refer to the capture of externally directed attention by task-irrelevant negative stimuli. The data yielded some very interesting findings which I discuss later in the discussion, but in the first section of the discussion I address the surprising failures to replicate previously observed links with individual differences and impacts of attentional retraining.

In the first study, we investigated the possibility that training participants to inhibit a response to a pre-potent stimulus could achieve similar attentional outcomes to typical Dot

Probe retraining. Participants were randomly allocated into one of three training conditions: to ignore negative stimuli, attend to negative stimuli, or inhibit attention to non-emotional-but- physically-salient stimuli (i.e., inhibitory control training). The analyses revealed strong evidence that both traditional retraining methods and inhibitory control training failed to shift measures of attentional bias. Furthermore, the analyses provided strong evidence against a relationship between anxiety, depression, or stress and the attentional bias index in the Dot Probe, thereby contrasting with a foundational finding in the attentional bias literature. Thus, in the subsequent studies, we turned our focus from spatial attention to temporal attention, using the emotion-induced blindness task. Studies two to four focused on establishing the psychometric properties of emotion-induced blindness, while study five investigated its malleability for attentional bias modification purposes. CHAPTER 7 159

In study two, we investigated the test-retest reliability of emotion-induced blindness by having participants complete two emotion-induced blindness sessions seven days apart. The analysis revealed that while using difference scores (i.e., accuracy following neutral distractors minus accuracy following negative distractors) yielded higher test-retest reliability than other measures of attentional bias, accuracies following negative distractors alone yielded much higher test-retest reliability indices. In addition, we replicated the emotion-induced blindness effect in which there was greater impairment by negative distractors than neutral distractors at early lags. Therefore, these findings suggest that emotion-induced blindness may be a more stable phenomenon than widely used spatial attention tasks and may be better suited for individual differences research.

In study three, we investigated whether valence or arousal drives attentional capture in emotion-induced blindness and the Dot Probe task. This was important for understanding whether attentional dysfunction in psychopathology may be characterised by senstivitiy to emotional valence and/or arousal of stimuli. Participants either complete the emotion-induced blindness or the Dot Probe task, while an additional group of participants rated the images used in the tasks. In addition, participants completed a series of questionnaires indexing negative affect and persistent negative thought. In the emotion-induced blindness task, we replicated the basic effect, in which greater target impairment was observed following negative relative to neutral distractors. Consistent with past findings, there was extremely strong evidence that the effect was present at lags 1 and 2, but also extremely strong evidence that this effect was no longer present at lag 8. Effect size estimates suggest that this effect is larger at lag 2 than at lag

1, with inspection of the means suggesting this may be due to neutral accuracy improving more 160 GENERAL DISCUSSION than negative accuracy from lag 1 to lag 2. Multiple regression analyses yielded moderate evidence that valence best accounted for attentional capture at lags 1 and 2, while neither valence nor arousal predicted attentional capture at lag 8. However, due to pre-selection of stimuli based on valence, the results of this regression may have been caused by two distinct clusters of data points, without any real relationship. Therefore, we repeated the analysis using only the negative stimuli for a more fine-grained analysis. When only looking at the negative distractors, the analysis yielded extremely strong evidence that at lag 2, the valence and arousal ratings of distractors uniquely contributed to emotion-induced blindness. The R2 of this model, was almost exactly the summed R2 values of valence and arousal models alone. At lag 1, there was marginal evidence for the null, possibly due to ceiling effects. At lag 8, there was moderate evidence that neither valence nor arousal predicted attentional capture – consistent with the finding that at lag 8, the effect is no longer present. To analyze whether emotion-induced blindness predicted measures of negative affect and/or persistent negative thought, we submitted the questionnaires to a principal components analysis, yielding a single component.

However, there was evidence that emotion-induced blindness at lag 1 did not predict this index of overall negative affect. Indeed, this outcome held true when we used emotion-induced blindness performance to predict the questionnaires individually. In the Dot Probe, we found moderate evidence that RTs were faster when the probe appeared behind the negative stimuli than when the probe appeared behind the neutral stimulus. However, there was strong evidence that neither ratings of valence nor arousal accounted for attentional capture. These findings suggest that emotion-induced blindness was more sensitive to the finer gradations of CHAPTER 7 161 valence and arousal than the Dot Probe. However, our results suggest that relationship between emotion-induced blindness and psychopathology require further investigation.

In study four, we investigated whether emotion-induced blindness represented early or late selection in visual processing. We did this by using a priming paradigm to see whether targets missed due to emotion-induced blindness were still processed and thus interfered with a subsequent arrow judgement task. In addition, we also included the GAD 7 as a measure of anxiety. We tested this using a modified emotion-induced blindness task to investigate whether target direction primed responses to a post-stream arrow judgment. The analysis found that if the participant had accurately reported the target in emotion-induced blindness, trials in which the target and arrow were in different directions yielded slower RTs than if they were in the same direction. However, for the trials in which the target was not reported accurately, there was strong evidence against this effect, i.e. no priming effect. Therefore, emotion-induced blindness seems to reflect early selection. Consistent with the results of study 3, emotion- induced blindness did not demonstrate any relationship with anxiety.

Finally, we investigated whether emotion-induced blindness was malleable. In the first experiment, we tested whether a) training participants to ignore negative distractors, b) training them to ignore neutral distractors, or c) merely exposing them to negative images would reduce the emotion-induced blindness effect. The analysis provided strong evidence that none of these training conditions reduced the magnitude of emotion-induced blindness. In contrast, we found strong evidence that over time, there was an overall increase in accuracy regardless of training condition. In the second experiment, we tested whether the absence of any training effects was due to ceiling effects at lag 1. Therefore, instead of using lag 1 in the 162 GENERAL DISCUSSION final test phase, we used lags 2 and 4. Furthermore, to assess whether simply searching through a stream of images could increase accuracy, we also added an additional baseline (no- distractor) training condition and baseline trials to the assessment. Once again, there was moderate evidence to suggest no difference in training conditions whatsoever. However, in contrast with Experiment 1, in Experiment 2 there was moderate evidence to suggest that performance decreased over time. Taken together, the study suggested that although general practice effects can lead to increased accuracy in the task, over a single session of training and/or exposure to emotional stimuli, emotion-induced blindness itself is not malleable.

In summary, our findings suggest that neither spatial nor temporal task-irrelevant attentional capture has easily reliable links with negative affect or persistent negative thought.

Our evidence also suggests that single session training may not reliable impact attentional biases in either task. That said, emotion-induced blindness seems to be a relatively stable task, demonstrating sensitivity to the gradations of valence and arousal and potentially reflecting early perceptual competition.

The lack of Relationship between Attentional Bias and Anxiety

In relation to the finding that neither the Dot Probe nor emotion-induced blindness showed any links with negative affect, there are three key possibilities: a) there may be certain moderating factors linking attentional bias and negative affect, b) our current measurement tools were flawed, or c) subclinical variations in negative affect and persistent negative thought are not linked with differences in attentional bias as operationalized in the literature.

Possible Moderating Factors. The first possibility is that the link between attention and emotional dysfunction does exist, but only under certain conditions. Indeed, this is supported CHAPTER 7 163 by the meta-analysis by Bar-Haim and colleagues (2007), in which the authors found that while an attentional bias was present in individuals with clinical anxiety, it was absent in healthy controls.

One possible explanation that we did not observe a relationship between measures of attentional bias and clinically related individual difference is that the emotional state of the individual play a role, not just trait differences. One piece of evidence to suggest this comes from the attentional bias modification literature, in that successful modification of attention reduces attentional vulnerability towards a stressor. Past meta analyses have shown that the differential benefits of attentional bias modification most strongly emerge following a lab stressor (Hallion & Ruscio, 2011; Mogoaşe et al., 2014). This suggests that the state of the individual is an important link with attentional bias. Therefore, state anxiety may also play a large role in attentional biases (Frewen, Dozois, Joanisse, & Neufeld, 2008). If this was the case, then it would likely explain the mixed findings in the literature, where the relationship between trait anxiety and attention would only sometimes emerge, due to individuals with high trait anxiety being more likely to have elevated state anxiety. This may also explain the null findings between attentional bias tasks and anxiety in this thesis.

One way this can be investigated in the future is by using stress or anxiety induction methods such as the Trier Social Stress Test (Kirschbaum, Pirke, & Hellhammer, 1993) or shock induced anticipatory anxiety (e.g. Shechner, Pelc, Pine, Fox, & Bar-Haim, 2012). This way, rather than seeing whether a certain index (often chosen for arbitrary reasons) shows links with anxiety, we might be able to conduct an exploratory analysis to assess what indices or values are sensitive to fluctuations in state anxiety. 164 GENERAL DISCUSSION

Past studies have examined effects of state vs trait anxiety/depression on spatial attention bias (e.g. Beevers & Carver, 2003; MacLeod & Mathews, 1988; Mogg, Bradley, &

Hallowell, 1994). In some of these studies, the interaction between trait and state anxiety was investigated using lab induced stressors or externally occurring stressors (e.g. upcoming examinations for students). However, many of these studies were too underpowered to find an attentional bias X anxiety effect according to the meta-analysis by Bar-Haim et al. (2007), undermining the interpretability of the findings. Using emotion-induced blindness as a non- spatial measure of attentional bias, a past study showed that threat of shock did not modulate the effect at lag 2 but did prolong emotional interference at lags 4 and 7 (Haddara et al., 2018).

This is at odds with an earlier finding in which the researchers induced relationship anxiety in heterosexual females by having their partners rate other females on campus for attractiveness

(Most et al., 2010). The authors found that the degree to which the participants reported discomfort correlated significantly with emotion-induced blindness. In sum, more studies using high powered designs are required to disentangle the influence of state and trait anxiety on attentional biases.

Echoing the findings from a previous chapter in this thesis, another possible moderating factor is the subjective value of the stimuli used. In attentional bias studies, a wide range of stimuli are used that are often rated by other people. However, it may be that each individual only demonstrates a reliable attentional effect when stimuli are used that the individual finds subjectively emotional. For example, Olatunji and colleagues (2013) found that veterans with

PTSD showed increased emotion-induced blindness to combat related stimuli compared to veterans without PTSD and healthy controls. Jin, Onie, Curby & Most (2018) found that CHAPTER 7 165 subjective ratings of negative stimuli predicted the magnitude of emotion-induced blindness. In the Dot Probe, past studies have observed hypervigilance towards stimuli relating to the individual’s predominant worry (Amir et al., 2009; Mathews and MacLeod, 1985; McNally et al.,

1994). Consistent with a personalized approach, in the treatment of anxiety and depression, cognitive behavioural therapy seeks to challenge the particular and specific cognitions relevant to each individual. If indeed only stimuli that are subjectively aversive capture attention, then much noise would be introduced when a) the proportion of stimuli that each individual finds emotional differs from one person to another, b) the degree to which each individual finds each stimulus aversive differs and c) aggregating responses washes out any fine-grained details, combining unaccounted noise in an imprecise index. Therefore, future research may benefit from incorporating subjective ratings of each stimulus as a possible moderator.

Flawed Measurement Tools and Analyses. The second possibility is that our current measurement tools are unable to properly investigate attention towards negative information in anxiety. This was highlighted in a recent publication by MacLeod and colleagues (2019), which addressed the claims that the Dot Probe had low reliability. Previous publications have noted that keypress response times may be too crude a measurement tool (Onie et al., 2019).

New versions of the Dot Probe task have reported relationships between the indices and anxiety (e.g. Attentional Response to Distal versus Proximal Emotional Information Paradigm;

Grafton & MacLeod, 2014; Rudaizky, Basanovic & MacLeod, 2014). However, they need to be replicated.

While our tasks indeed need to be improved, improvements in task reliability alone may not be optimal. A recent study demonstrated that the reliability of a task may not be tied to 166 GENERAL DISCUSSION how robustly it measures an effect (Hedge et al., 2018). The authors found that low test-retest and internal reliability can be attributed to having low between participant variability.

Therefore, if two tasks have the same measurement noise, the task with greater between- participant variability will have greater reliability indices. In fact, in the same paper, the authors argued that in experimental designs that do not target individual differences, low between- participant variability, and thus low test-retest reliability, yields more statistical power.

Furthermore, the authors noted that any manipulations that reduce variability (e.g., deriving difference scores) will almost always yield lower reliability indices because they reduce between-participant variability.

For example, consider the possibility that in fact the Dot Probe is an almost perfect index of spatial attention. Assume in this scenario that there is a strong relationship between attentional bias and anxiety. If we tested a predominantly low anxious sample we would have minimal between participant variability, and low test-retest reliability despite the strong relationship between the Dot Probe and anxiety. It may also be difficult to find individual differences in attentional bias if publication bias leads to underestimations of the range of anxiety scores necessary for individual differences to emerge; in such a case, we may believe that we are testing a heterogeneous sample despite it being homogenous relative to the fuller spectrum of scores. Coupled with the addition of even minimal measurement noise, this would result in sub-par reliability. In an alternative scenario, an attentional bias may be a perfect measure of spatial attention allocation, but there may be no true link between attentional bias and anxiety, resulting in a completely normal distribution of attentional bias indices in the population. This too would yield sub-par reliability due to sampling from a homogenous CHAPTER 7 167 population. In sum, reliability indices by themselves are unable to distinguish whether the Dot

Probe is an extremely good or an extremely poor measure of spatial attention.

One solution may be to devote more time and effort to investigating the psychometric properties of these tasks rather than immediately applying them to a clinical setting. In this spirit, the approach I have taken throughout this thesis is to consistently attempt to replicate the effect (e.g. the relationship between emotion-induced blindness and anxiety measures) while investigating the specific underlying mechanisms (e.g. whether it is early vs late visual processing). By understanding further how the perceptual system works, we may be able to generate stronger theories about attentional dysfunction in psychopathology.

However, a recent paper by Rouder, Kumar and Haaf (2019) suggests that using inhibition tasks to study individual differences may be a practice in futility. Even though the authors specifically use inhibition tasks e.g. Stroop or Flanker tasks, past findings have shown that in the Dot Probe, attentional control and inhibition mechanisms may be engaged (Heeren et al., 2013) and in emotion-induced blindness, participants must inhibit the representation of the emotional stimulus to detect the target. The authors show that correlations between common inhibition tasks do not emerge in typical correlation analyses due to the high level of measurement error or trial noise. Furthermore, using a hierarchical model which substantially accounts for more variability on a group, individual and trial level, the correlations between inhibition tasks still do not emerge due to the high level of measurement noise. Many attentional bias tasks such as the Dot Probe and Spatial Cueing task have been shown to link with inhibition or executive control (Heeren et al., 2013). Furthermore, attentional bias tasks may fall prey to this issue to an even greater extent due to the use of emotional stimuli – 168 GENERAL DISCUSSION increasing the variation and between-trial noise. The authors propose several solutions for improving measurement sensitivity, including increasing trial number rather than just increasing sample size, and using hierarchical models that account for trial level variation (e.g. random effects linear models).

In sum, it possible that attentional biases to threat are linked with individual differences in anxiety, but that our tasks are too insensitive to robustly detect this relationship. In improving these tasks, reliability cannot be the only benchmark, but we need a strong understanding of what participants are doing and what mechanisms these tasks are engaging.

There is also the possibility that many experimental tasks (including widely used ones) do not constitute fruitful tools for investigating individual differences.

No relationship between Attention and Psychopathology. Finally, given our findings, one possibility we must consider is that clinically relevant individual differences in healthy participants are not characterized by enhanced attentional capture by task-irrelevant emotional stimuli. At the time of this writing, many key theories across science have failed to find support in high powered studies despite being tied to large literatures of findings that would predict otherwise. Some notable examples are the ego depletion effect from social psychology (Hagger et al., 2016), genetic markers of depression (Border et al., 2019), and mating motives affecting consumer choices (Shanks, Vadillo, Riedel, Clymo, Govind, Hickin, Tamman, Puhlmann, et al.,

2015). A similarity between all of these findings is that there was a wealth of evidence for the existence of such effect. Therefore, we must cautiously assess our own field.

With the Dot Probe, Kruijt et al. (2019) noted that – using the effect size estimate given by a notable meta-analysis showing that attentional bias is linked with anxiety (Bar-Haim et al., CHAPTER 7 169

2007) – existing studies are not sufficiently powered. This is highly problematic, as a large collection of underpowered studies is not a substitute for a smaller set of sufficiently precise studies (Brysbaert, 2019). This casts healthy doubt upon past findings with the Dot Probe.

With regard to a relationship between emotion-induced blindness and negative affect, investigation into this link is relatively new. While the use of Bayesian analyses in this thesis does help circumvent the issue of power, the literature has typically used a frequentist approach, which requires accurate power analyses. The only other published employing

Bayesian analyses with emotion-induced blindness is by Onie and Most (2017), which investigated the relative degree to which emotion-induced blindness and the Dot Probe predicted negative affect and persistent negative thought. While the study reported moderate evidence for a relationship between emotion-induced blindness performance and negative affect, the authors used a component score rather than the individual subscales of the DASS-21.

This is problematic due to the fact that the component score is not a validated measure of negative affect or persistent negative thought. Combined with the findings of this thesis, the relationship between emotion-induced blindness and negative affect may be tenuous.

However, as outlined by Chun et al. (2011), there are still many more mechanisms of attention. Indeed, external attention is more easily measured than attention to internal representations; however, current models of psychopathology place maladaptive cognitions (in other words, biased selection towards certain memory traces), at the core of almost all disorders (McLaughlin & Nolen-Hoeksema, 2011). Therefore, following on from more recent theories of anxiety, perhaps research should focus more on negative cognitions – which in itself, can be construed as an internally directed attentional bias. 170 GENERAL DISCUSSION

In sum, we may not have observed a relationship between task-irrelevant attentional capture by negative stimuli and psychopathology due to a) moderating factors, b) the inadequacy of our measurement tools, or c) the true absence of this relationship. Nevertheless, the inconsistent relationship between attentional bias measures and clinically relevant individual differences is an important issue to address as the inconsistent findings between attentional bias tasks (especially the Dot Probe), compromise the validity of attention bias modification treatments stemming from these findings. Therefore, it is important to first establish under what conditions this link reliably emerges to properly assess the effects of training.

Attentional Bias Modification

In addition to the finding that neither the Dot Probe nor emotion-induced blindness showed links with negative affect, we failed to modify biases in both these tasks. In the following sections, we discuss possible reasons for the failure to train these attentional mechanisms

Retraining the Dot Probe. In the first study, we attempted to retrain spatial attention bias using a typical Dot Probe attentional retraining paradigm, in which participants are trained to attend to or away from threat. In addition, we investigated whether inhibitory control training would result in similar patterns as the avoid-negative patterns of training. None of these training conditions yielded any change in attention.

This finding is surprising, as over 3 meta-analyses that were noted in the introduction suggest a robust effect of attentional retraining using the traditional retraining methods. While there have been some studies showing failures to retrain (e.g. Onie et al., 2019; Clarke et al., CHAPTER 7 171

2014; Notebaert et al., 2015; Clarke et al., 2017, Everaert et al., 2015), the evidence from meta- analyses showing traditional retraining methods are able to modify attention vastly outnumber these studies (e.g. Hakamata et al., 2010; Hallion & Ruscio, 2011; Mogoase et al., 2014; Cristea et al., 2015; Heeren et al., 2015; Price et al., 2016). Therefore, there are three main possibilities.

The first is that we found a rare sample of participants in which the training manipulation did not work. The second is that this training method is able to modify attention, but in higher doses. However, as discussed in Chapter 2, the study had higher number of training trials than other studies which could modify attention in a single session (e.g. Chen et al., 2015). The third and final possibility is that, as recent studies have suggested, this training method is not optimal in retraining biases. Note that if future studies conclude the second possibility is more likely, then this would be evidence for a large a pervasive publication bias issue in this field.

Retraining Emotion-induced blindness. With emotion-induced blindness, we attempted to retrain this bias by exposing participants to a large number of repetitive trials. This is the first documented attempt at retraining emotion-induced blindness.

One possibility is that emotion-induced blindness cannot be fully eliminated through attentional retraining. While past studies have showed that emotion-induced blindness can be modulated, to date there is only one study which reported the absence of emotion-induced blindness. Specifically in the second experiment of Most and colleagues (2005), the authors found that when participants were told what type of target would be present, individuals with low harm-avoidance demonstrated no emotional impairment. Enns et al. (2017) found that after controlling for ceiling effects, the attentional blink was robust against retraining, concluding that the attentional mechanisms engaged in the attentional blink represent the 172 GENERAL DISCUSSION upper limits of human perception. Perhaps this holds true for emotion-induced blindness as well.

However, another possibility is that we are able to retrain emotion-induced blindness, with the conditions enabling such retraining yet to be discovered. This claim has support from findings in the Dot Probe literature, as a current line of investigation tests what conditions might support a reliable training effect in the Dot Probe emerge. Notebaert et al. (2015) found that a novel card task was able to elicit a greater training effect than the traditional training condition. In the novel card task, participants were continuously presented with a pair of faces

(one angry one happy). On each trial, participants had to indicate whether the happy or angry face (depending on which condition the participant as assigned to) on the current trial was the same person as the previous trial. In particular, during the card task stimuli are no longer task irrelevant, but are task-relevant targets. Therefore, it may be that our current method of exposing participants to a large number of trials in which participants repeatedly perform the task may not be optimal in retraining the mechanisms underlying emotion-induced blindness.

Findings from Emotion-Induced Blindness

Despite the surprising failures to replicate (e.g. no relationship between attentional bias tasks and clinically relevant individual differences, and our inability to retrain the Dot Probe), the experiments in this thesis provided important insight into the processes underlying emotion-induced blindness task. Emotion-induced blindness is a spatially localized form of early perceptual competition, distinguishing it from the attentional blink which seems to operate with late selection mechanisms, which do not appear to be spatially localized (Lunau & Olivers,

2010). In addition, at no point in chapters three to six did we fail to replicate the emotion- CHAPTER 7 173

induced blindness effect. As seen in Table 11, the effect size of emotion-induced blindness at

the early lags are substantial. These values compare favorably to other tasks indexing

emotional prioritization in the literature, including the Emotional Stroop (Phaf & Khan, 2007)

and Dot Probe (Bar-Haim et al., 2007; Kruijt et al., 2018). Therefore, emotion-induced blindness

may be amongst the most replicable and robust tasks used to assess emotional prioritization.

Table 12

Median and 95% Credible Intervals of Emotion-Induced Blindness Effect Size Estimates per Thesis Chapter

Chapter 3 Chapter 4 Chapter 5 Chapter 6

Lag 1 0.687 [0.924, 0.449] 0.835, [1.128, 0.543] 1.425, [1.115, 1.731] 0.849, [1.021, 0.677]

Lag 2 1.070, [1.341, 0.815] 1.145, [1.448, 0.837] Not Tested 0.833, [1.047, 0.627] Note: Median effect sizes are reported while credible intervals are reported in parentheses. Effect sizes are in Cohen’s � and were obtained using Bayesian t-test in JASP testing the hypothesis that neutral accuracy is greater than negative accuracy using default priors.

One interesting pattern of findings from this thesis and the wider literature is that while

emotion-induced blindness can be modulated, e.g. due to relational discomfort (Most et al.,

2010) and proactive control (Kennedy et al. 2018), there has only been one reported incidence

where emotion-induced blindness was completely eliminated (Most et al., 2005). As previously

mentioned, the authors found that individuals with low measures of harm-avoidance were able

to tune their attentional set towards the target. However, note that this manipulation didn’t

reduce the impact of emotional distractors, but rather enhanced detection of the target by

allowing participants to form an attentional set towards the type of target. Therefore, it is

unclear whether the emotional distractors had any less impact. Together, past findings suggests

that while emotional prioritization indexed by emotion-induced blindness can be attenuated or 174 GENERAL DISCUSSION exacerbated, it is extremely difficult to eliminate. This is consistent with the idea that early perceptual processes are more immune to cognitive influence (Firestone & Scholl, 2016;

Plyshyn, 1999). This in turn, may be the reason why emotion-induced blindness demonstrates favourable test-retest reliability compared to spatial attention tasks, due to its stability stemming from its early perceptual nature. Furthermore, the stability of emotion-induced blindness, especially at lag 2 where it has been shown to be most robust, may suggest that its sensitivity to the valence and arousal may be due to the lack of interference from cognitive processes as well.

However, one trade-off may be that in exchange for the stability it offers, emotion- induced blindness is inflexible to modification. In this thesis, we show evidence for and against a basic Dot Probe congruence effect, speaking to its instability. However, reported successes in modifying attentional patterns in the Dot Probe are numerous. One area of investigation may be to investigate the possibility of a third variable that may moderate effects of training, such as the ability to exert executive control over the attentional mechanism.

Overall, emotion-induced blindness seems to be a robust task that occurs at a relatively early stage in visual processing, which may contribute to its stability over time, as well as its sensitivity to the finer gradations of distractors’ valence and arousal values. The effect sizes reported in this thesis are medium to very large; however, there is a possibility that stability and inflexibility come hand in hand.

Strengths and Limitations

One strength of this thesis is the use of Bayesian analyses. Studies relying on frequentist analyses are often unable to determine whether a non-significant finding is due to a lack of an CHAPTER 7 175 effect or a lack of precision. Indeed, even significant findings from low-powered findings are found to be imprecise (Onie et al., 2019). Therefore, the use of Bayesian analyses where available, helps to quantify the evidence for or against a given effect. In the spatial attentional bias domain, to my knowledge, there have only been two studies that use Bayesian analyses.

One is a recent meta-analysis by Kruijt and colleagues (2018), in which they provide moderate evidence against the existence of congruence effect in clinically anxious individuals. The second is Chapter 2 of this thesis, published as Onie and colleagues (2019), which is one of the first fully pre-registered attentional bias modification study. This study provided strong evidence against a relationship between attentional bias and anxiety.

One limitation of this thesis is its reliance on recruitment of non-clinical populations.

Bar-Haim et al. (2007) found no presence of attentional bias in healthy participants. Therefore, one potential explanation for our failure to find links between attentional bias tasks and anxiety is our use of non-clinical samples. However, whereas past research has indeed distinguished between clinical and non-clinical samples, dichotomizing a variable that can be continuous is not recommended for individual difference research (Goodhew & Edwards, 2019).

Furthermore, dichotomizing a continuous variable introduces a large array of problems, including the reduction of power (Altman & Royston, 2006; Dawson & Weiss, 2012). Therefore, there is the possibility that a continuous approach may in fact yield more accurate findings.

Studies including clinical patients often have a tendency to categorically separate them.

However, even in clinical populations, the degree of impairment and symptomology varies.

Furthermore a past study suggests that approaching dimensional approach is just as valid and yields more power in studying depression and anxiety (Bjelland et al., 2009). Therefore, future 176 GENERAL DISCUSSION studies might include both healthy and clinical populations while adopting use a continuous approach in clinical symptoms and investigate the interaction between the attentional effect and level of symptomology.

One solution to the uncertainty between attention on any mechanistic level and negative affect is to use a high powered registered report (Nosek & Lakens, 2014), featuring moderator variables drawn from published meta-analyses and publications. This particular style of reporting involves the peer review process prior to submission and guaranteed publication prior to data collection. This process eliminates the possibility of publication bias by guaranteeing publication and allows input from reviewers during the design phase, to maximize the quality of the study conducted.

Constraints on Generality. As suggested by Simons, Shoda & Lindsay (2017), I discuss the limits of generality of this thesis to facilitate accumulative science and help future studies build upon or replicate these findings. All the studies from this thesis come from a UNSW sample, with many testing a first-year student sample. This limits our ability to generalize to a larger population due to a predominantly young age-group, as past studies have found that attentional bias changes as a function of age (Mather & Carstensen, 2003; Talbot, Ksander, &

Gutchess, 2018). Furthermore, we only sample from UNSW participant pools, limiting our generalizability to different populations and cultures, as past studies have found that different cultures show differences in visual processing (Amer, Ngo, & Hasher, 2017; Boduroglu, Shah, &

Nisbett, 2009) and on the Stroop task (Ishii, Reyes, & Kitayama, 2003). While the student population at UNSW is known to be highly diverse (UNSW Media and Communications office, n.d.), we did not collect information on ethnicity or cultural background, lacking the evidence to CHAPTER 7 177 make this claim. Therefore, future studies seeking to build upon this work should consider the age group as well as cultural backgrounds when interpreting their findings.

Future Directions

Emotion-Induced Blindness. As previously discussed, given the finding that emotion- induced blindness is robust but does not lend itself to retraining (in this thesis) or a variety of other manipulations in the literature, it may be that emotion-induced blindness represents a critical and universal mechanism necessary for emotional prioritization. Two studies can be done to test this. The first would seek to establish whether the emotion-induced blindness is present across different populations. Almost all studies on emotion-induced blindness have been conducted in North America or Australia, strictly confining its current generalizability to populations in these areas. While it can be argued that many of the samples contain multi- cultural populations, this does not represent a global population. Secondly, a study can be conducted using methodology developed by Haaf & Rouder (2019), which determines the proportion of individuals in a sample that a) show the effect, b) do not show any effect, or c) show the reverse effect using Bayesian model fitting. If this is indeed a mechanism present in all humans, then this study should show results consistent with this.

Formalizing Models. One common approach in cognitive psychology is to use computational models to formalize theories. While statistics also employ models (e.g., linear modeling), cognitive models are also known as process models, which simulate the cognitive processes underlying a given response. Using cognitive models allows us to speak more towards the underlying processes and mechanisms rather than just surface level observations, and more strongly test these theories. This practice is beginning to appear in the attentional bias 178 GENERAL DISCUSSION literature; for example, Price, Brown, and Siegel (2019) applied drift diffusion modelling to Dot

Probe data. Drift diffusion modelling posits that a response (e.g. a key press in response to a probe) is the result of multiple cognitive processes and maps on these processes to parameters within the model. For example, the drift rate is how quickly the individual processes information, allowing them to make a correct response. Non-decision time is a parameter which includes orienting to the stimulus as well as motor responses. With the Dot Probe, we would expect congruent and incongruent trials to differ in RTs due to the time taken to reorient their attention towards the probe, but we would not expect the differ in terms of how quickly participants respond to the probe once their attention is on it. Hence, we would expect there to be a difference in non-decision time, but not drift rate. Price and colleagues found exactly that.

Therefore, using cognitive models allow us to speak to the underlying mechanisms in responding above and beyond surface level observations.

In addition, cognitive models may allow us to make inferences about the underlying mechanisms engaged in the task, allowing us to tie the attentional bias field more closely to the visual cognition field. For example, using certain types of cognitive models, we might be able to see whether certain tasks are driven by common mechanisms and develop a taxonomy of attentional bias tasks based on their latent mechanisms.

Conclusion and Summary

The findings from this thesis cast reasonable doubt on the relationship between attentional capture by task-irrelevant emotional stimuli and individual differences related to psychopathology, as well as on the trainability of such attentional biases. Despite this, evidence from this thesis and the wider literature supports suggestions that emotion-induced blindness CHAPTER 7 179 is a non-spatial, early selection mechanism that is stable across time points and sensitive to gradations of the valence and arousal of stimuli. In the ongoing mission to understanding links between what people spontaneously pay attention to and the implications of such biases for emotional health, future research should continue to integrate the attentional bias literature with the visual cognition literature, while attempting to replicate existing findings in high- powered samples.

180 REFERENCES

References

Aczel, B., Palfi, B., Szollosi, A., Kovacs, M., Szaszi, B., Szecsi, P., … Wagenmakers, E.-J. (2018).

Quantifying Support for the Null Hypothesis in Psychology: An Empirical Investigation.

Advances in Methods and Practices in Psychological Science, 251524591877374.

https://doi.org/10.1177/2515245918773742

Alpers, G. W., & Gerdes, A. B. M. (2007). Here Is Looking at You: Emotional Faces Predominate

in Binocular Rivalry. Emotion. https://doi.org/10.1037/1528-3542.7.3.495

Altman, D. G., & Royston, P. (2006). The cost of dichotomising continuous variables. British

Medical Journal.

Amer, T., Ngo, K. W. J., & Hasher, L. (2017). Cultural differences in visual attention: Implications

for distraction processing. British Journal of Psychology.

https://doi.org/10.1111/bjop.12194

Amir, N., Beard, C., Burns, M., & Bomyea, J. (2009). Attention Modification Program in

Individuals With Generalized Anxiety Disorder. Journal of Abnormal Psychology.

https://doi.org/10.1037/a0012589

Amir, N., Beard, C., Taylor, C. T., Klumpp, H., Elias, J., Burns, M., & Chen, X. (2009). Attention

Training in Individuals With Generalized Social Phobia: A Randomized Controlled Trial.

Journal of Consulting and Clinical Psychology. https://doi.org/10.1037/a0016685

Anderson, A. K. (2005). Affective influences on the attentional dynamics supporting awareness.

Journal of Experimental Psychology: General. https://doi.org/10.1037/0096-

3445.134.2.258

Andrews, M., & Baguley, T. (2013). Prior approval: The growth of Bayesian methods in REFERENCES 181

psychology. British Journal of Mathematical and Statistical Psychology.

https://doi.org/10.1111/bmsp.12004

Arnell, K. M., Killman, K. V., & Fijavz, D. (2007). Blinded by Emotion: Target Misses Follow

Attention Capture by Arousing Distractors in RSVP. Emotion.

https://doi.org/10.1037/1528-3542.7.3.465

Aron, A. R., Robbins, T. W., & Poldrack, R. A. (2004). Inhibition and the right inferior frontal

cortex. Trends in Cognitive Sciences. https://doi.org/10.1016/j.tics.2004.02.010

Awh, E., & Jonides, J. (2001). Overlapping mechanisms of attention and spatial working

memory. Trends in Cognitive Sciences. https://doi.org/10.1016/S1364-6613(00)01593-X

Awh, E., & Pashler, H. (2000). Evidence for split attentional foci. Journal of Experimental

Psychology: Human Perception and Performance. https://doi.org/10.1037/0096-

1523.26.2.834

Bar-Haim, Y. (2010). Research Review: Attention bias modification (ABM): A novel treatment for

anxiety disorders. Journal of Child Psychology and Psychiatry and Allied Disciplines.

https://doi.org/10.1111/j.1469-7610.2010.02251.x

Bar-Haim, Y., Lamy, D., Pergamin, L., Bakermans-Kranenburg, M. J., & van IJzendoorn, M. H.

(2007). Threat-related attentional bias in anxious and nonanxious individuals: A meta-

analytic study. Psychological Bulletin, 133(1), 1–24. https://doi.org/10.1037/0033-

2909.133.1.1

Basanovic, J., Notebaert, L., Grafton, B., Hirsch, C. R., & Clarke, P. J. F. (2017). Attentional

control predicts change in bias in response to attentional bias modification. Behaviour

Research and Therapy. https://doi.org/10.1016/j.brat.2017.09.002 182 REFERENCES

Beevers, C. G., & Carver, C. S. (2003). Attentional bias and mood persistence as prospective

predictors of dysphoria. Cognitive Therapy and Research.

https://doi.org/10.1023/A:1026347610928

Bjelland, I., Lie, S. A., Dahl, A. A., Mykletun, A., Stordal, E., & Kraemer, H. C. (2009). A

dimensional versus a categorical approach to diagnosis: Anxiety and depression in the

HUNT 2 study. International Journal of Methods in Psychiatric Research, 18(2), 128–137.

https://doi.org/10.1002/mpr.284

Boduroglu, A., Shah, P., & Nisbett, R. E. (2009). Cultural differences in allocation of attention in

visual information processing. Journal of Cross-Cultural Psychology.

https://doi.org/10.1177/0022022108331005

Boettcher, J., Berger, T., & Renneberg, B. (2012). Internet-based attention training for social

anxiety: A randomized controlled trial. Cognitive Therapy and Research.

https://doi.org/10.1007/s10608-011-9374-y

Border, R., Johnson, E. C., Evans, L. M., Smolen, A., Berley, N., Sullivan, P. F., & Keller, M. C.

(2019). No support for historical candidate gene or candidate gene-by-interaction

hypotheses for major depression across multiple large samples. American Journal of

Psychiatry. https://doi.org/10.1176/appi.ajp.2018.18070881

Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Converting Among

Effect Sizes. In Introduction to Meta-Analysis. https://doi.org/10.1002/9780470743386.ch7

Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: the self-assessment manikin and the

semantic differential. Journal of Behavior Therapy and Experimental Psychiatry, 25(1), 49–

59. https://doi.org/10.1016/0005-7916(94)90063-9 REFERENCES 183

Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision.

https://doi.org/10.1163/156856897X00357

Brown, H. M., Eley, T. C., Broeren, S., MacLeod, C., Rinck, M., Hadwin, J. A., & Lester, K. J.

(2014). Psychometric properties of reaction time based experimental paradigms

measuring anxiety-related information-processing biases in children. Journal of Anxiety

Disorders. https://doi.org/10.1016/j.janxdis.2013.11.004

Brown, S. L., Svrakic, D. M., Przybeck, T. R., & Cloninger, R. C. (1992). The relationship of

personality to mood and anxiety states: A dimensional approach. Journal of Psychiatric

Research, 26(3), 197–211. https://doi.org/10.1016/0022-3956(92)90023-H

Browning, M., Holmes, E. A., Murphy, S. E., Goodwin, G. M., & Harmer, C. J. (2010). Lateral

Prefrontal Cortex Mediates the Cognitive Modification of Attentional Bias. Biological

Psychiatry. https://doi.org/10.1016/j.biopsych.2009.10.031

Brysbaert, M. (2019). How Many Participants Do We Have to Include in Properly Powered

Experiments? A Tutorial of Power Analysis with Reference Tables. Journal of Cognition.

https://doi.org/10.5334/joc.72

Button, K. S., Ioannidis, J. P. A., Mokrysz, C., Nosek, B. A., Flint, J., Robinson, E. S. J., & Munafò,

M. R. (2013). Power failure: Why small sample size undermines the reliability of

neuroscience. Nature Reviews Neuroscience. https://doi.org/10.1038/nrn3475

Cain, M. S., Prinzmetal, W., Shimamura, A. P., & Landau, A. N. (2014). Improved control of

exogenous attention in action video game players. Frontiers in Psychology.

https://doi.org/10.3389/fpsyg.2014.00069

Carlbring, P., Apelstrand, M., Sehlin, H., Amir, N., Rousseau, A., Hofmann, S. G., & Andersson, G. 184 REFERENCES

(2012). Internet-delivered attention bias modification training in individuals with social

anxiety disorder - a double blind randomized controlled trial. BMC Psychiatry.

https://doi.org/10.1186/1471-244X-12-66

Cave, K. R., & Bichot, N. P. (1999). Visuospatial attention: Beyond a spotlight model.

Psychonomic Bulletin and Review. https://doi.org/10.3758/BF03212327

Chapman, S., & Martin, M. (2011). Attention to pain words in irritable bowel syndrome:

Increased orienting and speeded engagement. British Journal of Health Psychology.

https://doi.org/10.1348/135910710X505887

Cheesman, J., & Merikle, P. M. (1986). Distinguishing conscious from unconscious perceptual

processes. Canadian Journal of Psychology. https://doi.org/10.1037/h0080103

Chelazzi, L., Duncan, J., Miller, E. K., & Desimone, R. (1998). Responses of neurons in inferior

temporal cortex during memory-guided visual search. Journal of Neurophysiology.

https://doi.org/10.1152/jn.1998.80.6.2918

Chen, N. T. M., Clarke, P. J. F., Watson, T. L., MacLeod, C., & Guastella, A. J. (2015). Attentional

bias modification facilitates attentional control mechanisms: Evidence from eye tracking.

Biological Psychology, 104, 139–146. https://doi.org/10.1016/j.biopsycho.2014.12.002

Choi, H., Chang, L. H., Shibata, K., Sasaki, Y., & Watanabe, T. (2012). Resetting capacity

limitations revealed by long-lasting elimination of attentional blink through training.

Proceedings of the National Academy of Sciences of the United States of America.

https://doi.org/10.1073/pnas.1203972109

Chun, M. M., Golomb, J. D., & Turk-Browne, N. B. (2011). A Taxonomy of External and Internal

Attention. Annual Review of Psychology, 62(1), 73–101. REFERENCES 185

https://doi.org/10.1146/annurev.psych.093008.100427

Chun, M. M., & Potter, M. C. (1995). A Two-Stage Model for Multiple Target Detection in Rapid

Serial Visual Presentation. Journal of Experimental Psychology: Human Perception and

Performance. https://doi.org/10.1037/0096-1523.21.1.109

Ciesielski, B. G., Armstrong, T., Zald, D. H., & Olatunji, B. O. (2010). Emotion modulation of

visual attention: Categorical and temporal characteristics. PLoS ONE.

https://doi.org/10.1371/journal.pone.0013860

Cisler, J. M., & Koster, E. H. W. (2010). Mechanisms of attentional biases towards threat in

anxiety disorders: An integrative review. Clinical Psychology Review.

https://doi.org/10.1016/j.cpr.2009.11.003

Clarke, P. J. F., Branson, S., Chen, N. T. M., Van Bockstaele, B., Salemink, E., MacLeod, C., &

Notebaert, L. (2017). Attention bias modification training under working memory load

increases the magnitude of change in attentional bias. Journal of Behavior Therapy and

Experimental Psychiatry. https://doi.org/10.1016/j.jbtep.2017.02.003

Clarke, P. J. F., Browning, M., Hammond, G., Notebaert, L., & MacLeod, C. (2014). The causal

role of the dorsolateral prefrontal cortex in the modification of attentional bias: Evidence

from transcranial direct current stimulation. Biological Psychiatry.

https://doi.org/10.1016/j.biopsych.2014.03.003

Cohen J. (1988). Statistical Power Analysis for the Behavioural Science (2nd Edition). In

Statistical Power Anaylsis for the Behavioural Science (2nd Edition).

Cooper, R. M., & Langton, S. R. H. (2006). Attentional bias to angry faces using the dot-probe

task? It depends when you look for it. Behaviour Research and Therapy. 186 REFERENCES

https://doi.org/10.1016/j.brat.2005.10.004

Corbetta, M., & Shulman, G. L. (2002). Control of goal-directed and stimulus-driven attention in

the brain. Nature Reviews Neuroscience. https://doi.org/10.1038/nrn755

Correa, Á., & Nobre, A. C. (2008). Spatial and temporal acuity of visual perception can be

enhanced selectively by attentional set. Experimental Brain Research.

https://doi.org/10.1007/s00221-008-1429-2

Coull, J. T., & Nobre, A. C. (1998). Where and when to pay attention: The neural systems for

directing attention to spatial locations and to time intervals as revealed by both PET and

fMRI. Journal of Neuroscience.

Cristea, I. A., David, D., & Cuijpers, P. (2015). Practitioner Review : Cognitive bias modification

for mental health problems in children and adolescents : a meta-analysis. Journal of Child

Psychology and Psychiatry, 56(7), 723–734. https://doi.org/10.1111/jcpp.12383

Cristea, I. A., Kok, R. N., & Cuijpers, P. (2015). Efficacy of cognitive bias modification

interventions in anxiety and depression: Meta-analysis. British Journal of Psychiatry.

https://doi.org/10.1192/bjp.bp.114.146761

Dalgleish, T., Taghavi, R., Neshat-Doost, H., Moradi, A., Canterbury, R., & Yule, W. (2003).

Patterns of processing bias for emotional information across clinical disorders: A

comparison of attention, memory, and prospective cognition in children and adolescents

with depression, generalized anxiety, and posttraumatic stress disorder. Journal of Clinical

Child and Adolescent Psychology. https://doi.org/10.1207/S15374424JCCP3201_02

Dawson, N. V., & Weiss, R. (2012). Dichotomizing continuous variables in statistical analysis: A

practice to avoid. Medical Decision Making. https://doi.org/10.1177/0272989X12437605 REFERENCES 187

Den Hollander-Gijsman, M. E., Wardenaar, K. J., De Beurs, E., Van Der Wee, N. J. A., Mooijaart,

A., Van Buuren, S., & Zitman, F. G. (2012). Distinguishing symptom dimensions of

depression and anxiety: An integrative approach. Journal of Affective Disorders, 136(3),

693–701. https://doi.org/10.1016/j.jad.2011.10.005

Derakshan, N., Ansari, T. L., Hansard, M., Shoker, L., & Eysenck, M. W. (2009). Anxiety,

inhibition, efficiency, and effectiveness: An investigation using the Antisaccade task.

Experimental Psychology. https://doi.org/10.1027/1618-3169.56.1.48

Derryberry, D., & Reed, M. A. (2002). Anxiety-related attentional biases and their regulation by

attentional control. Journal of Abnormal Psychology. https://doi.org/10.1037/0021-

843X.111.2.225

Desimone, R. (1995). Neural Mechanisms of Selective Visual Attention. Annual Review of

Neuroscience. https://doi.org/10.1146/annurev.neuro.18.1.193

Deubel, H., & Schneider, W. X. (1996). Saccade target selection and object recognition: Evidence

for a common attentional mechanism. Vision Research. https://doi.org/10.1016/0042-

6989(95)00294-4

Deutsch, J. A., & Deutsch, D. (1963). Attention: Some theoretical considerations. Psychological

Review. https://doi.org/10.1037/h0042712

Downing, P. E. (2000). Interactions between visual working memory and selective attention.

Psychological Science. https://doi.org/10.1111/1467-9280.00290

Durand, F., Isaac, C., & Januel, D. (2019). Emotional memory in post-traumatic stress disorder: A

systematic PRISMA review of controlled studies. Frontiers in Psychology. Frontiers Media

S.A. https://doi.org/10.3389/fpsyg.2019.00303 188 REFERENCES

Duval, S., & Tweedie, R. (2000). Trim and fill: A simple funnel-plot-based method of testing and

adjusting for publication bias in meta-analysis. Biometrics. https://doi.org/10.1111/j.0006-

341X.2000.00455.x

Eide, P., Kemp, A., Silberstein, R. B., Nathan, P. J., & Stough, C. (2002). Test-retest reliability of

the emotional stroop task: Examining the paradox of measurement change. Journal of

Psychology: Interdisciplinary and Applied. https://doi.org/10.1080/00223980209605547

Enns, J. T., Kealong, P., Tichon, J. G., & Visser, T. A. W. (2017). Training and the attentional blink:

Raising the ceiling does not remove the limits. Attention, Perception, and Psychophysics,

79(8), 2257–2274. https://doi.org/10.3758/s13414-017-1391-9

Eriksen, C. W., & St. James, J. D. (1986). Visual attention within and around the field of focal

attention: A zoom lens model. Perception & Psychophysics.

https://doi.org/10.3758/BF03211502

Everaert, J., Mogoaşe, C., David, D., & Koster, E. H. W. (2015). Attention bias modification via

single-session dot-probe training: Failures to replicate. Journal of Behavior Therapy and

Experimental Psychiatry. https://doi.org/10.1016/j.jbtep.2014.10.011

Eysenck, M. W., & Derakshan, N. (2011). New perspectives in attentional control theory.

Personality and Individual Differences. https://doi.org/10.1016/j.paid.2010.08.019

Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). Anxiety and cognitive

performance: Attentional control theory. Emotion. https://doi.org/10.1037/1528-

3542.7.2.336

Fan, J., McCandliss, B. D., Fossella, J., Flombaum, J. I., & Posner, M. I. (2005). The activation of

attentional networks. NeuroImage. https://doi.org/10.1016/j.neuroimage.2005.02.004 REFERENCES 189

Fernandes, M. A., Koji, S., Dixon, M. J., & Aquino, J. M. (2011). Changing the focus of attention:

The interacting effect of valence and arousal. Visual Cognition.

https://doi.org/10.1080/13506285.2011.618151

Firestone, C., & Scholl, B. J. (2019). Cognition does not affect perception : Evaluating the

evidence for “ top-down ” effects, (2016). https://doi.org/10.1017/S0140525X15000965

Fischer, J., & Whitney, D. (2014). Serial dependence in visual perception. Nature Neuroscience.

https://doi.org/10.1038/nn.3689

Fox, E., Russo, R., Bowles, R., & Dutton, K. (2001). Do threatening stimuli draw or hold visual

attention in subclinical anxiety? Journal of Experimental Psychology: General.

https://doi.org/10.1037/0096-3445.130.4.681

Fox, E., Russo, R., & Dutton, K. (2002). Attentional bias for threat: Evidence for delayed

disengagement from emotional faces. Cognition and Emotion.

https://doi.org/10.1080/02699930143000527

Frewen, P. A., Dozois, D. J. A., Joanisse, M. F., & Neufeld, R. W. J. (2008). Selective attention to

threat versus reward : Meta-analysis and neural-network modeling of the dot-probe task,

28, 307–337. https://doi.org/10.1016/j.cpr.2007.05.006

Goodhew, S. C., & Edwards, M. (2019). Translating experimental paradigms into individual-

differences research: Contributions, challenges, and practical recommendations.

Consciousness and Cognition. https://doi.org/10.1016/j.concog.2019.01.008

Grafton, B., Mackintosh, B., Vujic, T., & MacLeod, C. (2014). When ignorance is bliss: Explicit

instruction and the efficacy of CBM-A for anxiety. Cognitive Therapy and Research.

https://doi.org/10.1007/s10608-013-9579-3 190 REFERENCES

Grafton, B., & MacLeod, C. (2014). Enhanced probing of attentional bias: The independence of

anxiety-linked selectivity in attentional engagement with and disengagement from

negative information. Cognition and Emotion.

https://doi.org/10.1080/02699931.2014.881326

Grissom, R. J., & Kim, J. J. (2014). Effect sizes for research: A broad practical approach. Effect

Sizes for Research: A Broad Practical Approach. https://doi.org/10.4324/9781410612915

Guan, M., & Vandekerckhove, J. (2016). A Bayesian approach to mitigation of publication bias.

Psychonomic Bulletin and Review. https://doi.org/10.3758/s13423-015-0868-6

Haaf, J. M., & Rouder, J. N. (2019). Some do and some don ’ t ? Accounting for variability of

individual difference structures, 772–789.

Haddara, N., Ravid, J., Miller, E. L., O’Hagan, M., Caracciolo, C., & Miskovic, V. (2018). Anxious

Anticipation Prolongs Emotional Interference for Rapid Visual Detection. Emotion.

https://doi.org/10.1037/emo0000521

Hagger, M. S., Chatzisarantis, N. L. D., Alberts, H., Anggono, C. O., Batailler, C., Birt, A. R., …

Zwienenberg, M. (2016). A Multilab Preregistered Replication of the Ego-Depletion Effect.

Perspectives on Psychological Science. https://doi.org/10.1177/1745691616652873

Hallion, L. S., & Ruscio, A. M. (2011). A Meta-Analysis of the Effect of Cognitive Bias

Modification on Anxiety and Depression. Psychological Bulletin, 137(6), 940–958.

https://doi.org/10.1037/a0024355

Harris, I. M., & Little, M. J. J. (2010). Priming the semantic neighbourhood during the attentional

blink. PLoS ONE. https://doi.org/10.1371/journal.pone.0012645

Hedge, C., Powell, G., & Sumner, P. (2018). The reliability paradox: Why robust cognitive tasks REFERENCES 191

do not produce reliable individual differences. Behavior Research Methods.

https://doi.org/10.3758/s13428-017-0935-1

Heeren, A., De Raedt, R., Koster, E. H. W., & Philippot, P. (2013). The (neuro)cognitive

mechanisms behind attention bias modification in anxiety: Proposals based on theoretical

accounts of attentional bias. Frontiers in Human Neuroscience.

https://doi.org/10.3389/fnhum.2013.00119

Heeren, A., Mogoaşe, C., McNally, R. J., Schmitz, A., & Philippot, P. (2015). Does attention bias

modification improve attentional control? A double-blind randomized experiment with

individuals with social anxiety disorder. Journal of Anxiety Disorders.

https://doi.org/10.1016/j.janxdis.2014.10.007

Heeren, A., Mogoase, C., Philippot, P., & McNally, R. J. (2015). Attention bias modification for

social anxiety: A systematic review and meta-analysis. Clinical Psychology Review, 40, 76–

90. https://doi.org/10.1016/j.cpr.2015.06.001

Hertel, P. T., & Mathews, A. (2011). Cognitive bias modification: Past perspectives, current

findings, and future applications. Perspectives on Psychological Science.

https://doi.org/10.1177/1745691611421205

Holländer, A., Corballis, M. C., & Hamm, J. P. (2005). Visual-field asymmetry in dual-stream

RSVP. Neuropsychologia. https://doi.org/10.1016/j.neuropsychologia.2004.06.006

Hunt, A. R., & Kingstone, A. (2003). Inhibition of Return: Dissociating Attentional and

Oculomotor Components. Journal of Experimental Psychology: Human Perception and

Performance. https://doi.org/10.1037/0096-1523.29.5.1068

Iacoviello, B. M., Wu, G., Abend, R., Murrough, J. W., Feder, A., Fruchter, E., … Charney, D. S. 192 REFERENCES

(2014). Attention Bias Variability and Symptoms of Posttraumatic Stress Disorder. Journal

of Traumatic Stress. https://doi.org/10.1002/jts.21899

Ihssen, N., & Keil, A. (2009). The costs and benefits of processing emotional stimuli during rapid

serial visual presentation. Cognition and Emotion.

https://doi.org/10.1080/02699930801987504

Imhoff, R., Lange, J., & Germar, M. (2019). Identification and location tasks rely on different

mental processes: a diffusion model account of validity effects in spatial cueing paradigms

with emotional stimuli. Cognition and Emotion.

https://doi.org/10.1080/02699931.2018.1443433

Ishii, K., Reyes, J. A., & Kitayama, S. (2003). Spontaneous Attention to Word Content Versus

Emotional Tone: Differences Among Three Cultures. Psychological Science.

https://doi.org/10.1111/1467-9280.01416

JASP Team. (2018). JASP (Version 0.9.0.1). [Computer Software].

https://doi.org/10.1017/S095457941700030X

Jeffreys, H. (1998). Theory of probability. Clarendon Press.

Jin, M., Onie, S., Curby, K. M., & Most, S. B. (2018). Aversive images cause less perceptual

interference among violent video game players: evidence from emotion-induced

blindness. Visual Cognition. https://doi.org/10.1080/13506285.2018.1553223

Jolicoeur, P. (1998). Modulation of the attentional blink by on-line response selection: Evidence

from speeded and unspeeded task1 decisions. Memory and Cognition.

https://doi.org/10.3758/BF03201180

Jolicœur, P., & Dell’Acqua, R. (1998). The Demonstration of Short-Term Consolidation. Cognitive REFERENCES 193

Psychology. https://doi.org/10.1006/cogp.1998.0684

Julian, K., Beard, C., Schmidt, N. B., Powers, M. B., & Smits, J. A. J. (2012). Attention training to

reduce attention bias and social stressor reactivity: An attempt to replicate and extend

previous findings. Behaviour Research and Therapy.

https://doi.org/10.1016/j.brat.2012.02.015

Kanyongo, G. Y., Brook, G. P., Kyei-Blankson, L., & Gocmen, G. (2007). Reliability and statistical

power: How measurement fallibility affects power and required sample sizes for several

parametric and nonparametric statistics. Journal of Modern Applied Statistical Methods.

https://doi.org/10.22237/jmasm/1177992480

Kennedy, B. L., & Most, S. B. (2015a). Affective stimuli capture attention regardless of

categorical distinctiveness: An emotion-induced blindness study. Visual Cognition, 23(1–2),

105–117. https://doi.org/10.1080/13506285.2015.1024300

Kennedy, B. L., & Most, S. B. (2015b). The rapid perceptual impact of emotional distractors.

PLoS ONE. https://doi.org/10.1371/journal.pone.0129320

Kennedy, B. L., Newman, V. E., & Most, S. B. (2018). Proactive Deprioritization of Emotional

Distractors Enhances Target Perception. Emotion. https://doi.org/10.1037/emo0000362

Kennedy, B. L., Rawding, J., Most, S. B., & Hoffman, J. E. (2014). Emotion-induced blindness

reflects competition at early and late processing stages: An ERP study. Cognitive, Affective

and Behavioral Neuroscience. https://doi.org/10.3758/s13415-014-0303-x

Kentridge, R. W., Nijboer, T. C. W., & Heywood, C. A. (2008). Attended but unseen: Visual

attention is not sufficient for visual awareness. Neuropsychologia.

https://doi.org/10.1016/j.neuropsychologia.2007.11.036 194 REFERENCES

Keysers, C., & Perrett, D. I. (2002). Visual masking and RSVP reveal neural competition. Trends

in Cognitive Sciences. https://doi.org/10.1016/S1364-6613(00)01852-0

Kindt, M., Bierman, D., & Brosschot, J. F. (1996). Stroop versus Stroop: Comparison of a card

format and a single-trial format of the standard color-word Stroop task and the emotional

Stroop task. Personality and Individual Differences. https://doi.org/10.1016/0191-

8869(96)00133-x

Kirschbaum, C., Pirke, K. M., & Hellhammer, D. H. (1993). The “Trier social stress test” - A tool

for investigating psychobiological stress responses in a laboratory setting. In

Neuropsychobiology. https://doi.org/10.1159/000119004

Klein, R. M., & Ivanoff, J. (2005). Inhibition of return. In Neurobiology of Attention.

https://doi.org/10.1016/B978-012375731-9/50020-3

Koch, C., & Tsuchiya, N. (2007). Attention and consciousness: two distinct brain processes.

Trends in Cognitive Sciences. https://doi.org/10.1016/j.tics.2006.10.012

Koster, E. H. W., Verschuere, B., Crombez, G., & Van Damme, S. (2005). Time-course of

attention for threatening pictures in high and low trait anxiety. Behaviour Research and

Therapy. https://doi.org/10.1016/j.brat.2004.08.004

Kroenke, K., Spitzer, R. L., Williams, J. B. W., & Löwe, B. (2010). The Patient Health

Questionnaire Somatic, Anxiety, and Depressive Symptom Scales: A systematic review.

General Hospital Psychiatry. https://doi.org/10.1016/j.genhosppsych.2010.03.006

Kruijt, A.-W., Parsons, S., & Fox, E. (2019). A meta-analysis of bias at baseline in RCTs of

attention bias modification: No evidence for dot-probe bias towards threat in clinical

anxiety and PTSD. Journal of Abnormal Psychology. https://doi.org/10.1037/abn0000406 REFERENCES 195

Kruijt, A. W., Field, A. P., & Fox, E. (2016). Capturing dynamics of biased attention: Are new

attention variability measures the way forward. PLoS ONE.

https://doi.org/10.1371/journal.pone.0166600

Lambert, A., Naikar, N., McLachlan, K., & Aitken, V. (1999). A new component of visual

orienting: Implicit effects of peripheral information and subthreshold cues on covert

attention. Journal of Experimental Psychology: Human Perception and Performance.

https://doi.org/10.1037/0096-1523.25.2.321

Lance, C. E., Butts, M. M., & Michels, L. C. (2006). What Did They Really Say ? Organizational

Research Methods, 9(2), 202–220. https://doi.org/10.1177/1094428105284919

Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1997). International affective picture system

(IAPS): Technical manual and affective ratings. NIMH Center for the Study of Emotion and

Attention.

Larsen, R. J., Mercer, K. A., Balota, D. A., & Strube, M. J. (2008). Not All Negative Words Slow

Down Lexical Decision and Naming Speed: Importance of Word Arousal. Emotion.

https://doi.org/10.1037/1528-3542.8.4.445

Lavie, N. (1995). Perceptual Load as a Necessary Condition for Selective Attention. Journal of

Experimental Psychology: Human Perception and Performance.

https://doi.org/10.1037/0096-1523.21.3.451

Lavie, N., Hirst, A., De Fockert, J. W., & Viding, E. (2004). Load theory of selective attention and

cognitive control. Journal of Experimental Psychology: General.

https://doi.org/10.1037/0096-3445.133.3.339

Lavie, N., & Tsal, Y. (1994). Perceptual load as a major determinant of the locus of selection in 196 REFERENCES

visual attention. Perception & Psychophysics. https://doi.org/10.3758/BF03213897

Lebeau, R. T., Glenn, D. E., Hanover, L. N., Beesdo-Baum, K., Wittchen, H. U., & Craske, M. G.

(2012). A dimensional approach to measuring anxiety for DSM-5. International Journal of

Methods in Psychiatric Research, 21(4), 258–272. https://doi.org/10.1002/mpr.1369

Lepsien, J., Griffin, I. C., Devlin, J. T., & Nobre, A. C. (2005). Directing spatial attention in mental

representations: Interactions between attentional orienting and working-memory load.

NeuroImage. https://doi.org/10.1016/j.neuroimage.2005.02.026

Levin, D. T., & Simons, D. J. (1997). Failure to detect changes to attended objects in motion

pictures. Psychonomic Bulletin and Review. https://doi.org/10.3758/BF03214339

Lindsay, D. S., & Jacoby, L. L. (1994). Stroop Process Dissociations: The Relationship Between

Facilitation and Interference. Journal of Experimental Psychology: Human Perception and

Performance. https://doi.org/10.1037/0096-1523.20.2.219

Linetzky, M., Pergamin-Hight, L., Pine, D. S., & Bar-Haim, Y. (2015). Quantitative evaluation of

the clinical efficacy of attention bias modification treatment for anxiety disorders.

Depression and Anxiety. https://doi.org/10.1002/da.22344

Litz, B. T., Prassas, A., & Weathers, F. W. (1994). Emotional Priming of Autobiographical

Memory in Post-Traumatic Stress Disorder. Cognition and Emotion.

https://doi.org/10.1080/02699939408408946

Lovibond, P. F., & Lovibond, S. H. (1995). The structure of negative emotional states:

Comparison of the Depression Anxiety Stress Scales (DASS) with the Beck Depression and

Anxiety Inventories. Behaviour Research and Therapy. https://doi.org/10.1016/0005-

7967(94)00075-U REFERENCES 197

MacLeod, C., & Clarke, P. J. F. (2015). The attentional bias modification approach to anxiety

intervention. Clinical Psychological Science. https://doi.org/10.1177/2167702614560749

MacLeod, C., Grafton, B., & Notebaert, L. (2019). Anxiety-Linked Attentional Bias: Is It Reliable?

Annual Review of Clinical Psychology. https://doi.org/10.1146/annurev-clinpsy-050718-

095505

MacLeod, C., & Mathews, A. (1988). Anxiety and the Allocation of Attention to Threat. The

Quarterly Journal of Experimental Psychology Section A.

https://doi.org/10.1080/14640748808402292

MacLeod, C., & Mathews, A. (2012). Cognitive Bias Modification Approaches to Anxiety. Annual

Review of Clinical Psychology. https://doi.org/10.1146/annurev-clinpsy-032511-143052

MacLeod, C., Mathews, A., & Tata, P. (1986). Attentional Bias in Emotional Disorders. Journal of

Abnormal Psychology. https://doi.org/10.1037/0021-843X.95.1.15

MacLeod, C., Rutherford, E., Campbell, L., Ebsworthy, G., & Holker, L. (2002). Selective attention

and emotional vulnerability: Assessing the causal basis of their association through the

experimental manipulation of attentional bias. Journal of Abnormal Psychology.

https://doi.org/10.1037/0021-843X.111.1.107

MacLeod, J., Stewart, B. M., Newman, A. J., & Arnell, K. M. (2017). Do emotion-induced

blindness and the attentional blink share underlying mechanisms? An event-related

potential study of emotionally-arousing words. Cognitive, Affective and Behavioral

Neuroscience. https://doi.org/10.3758/s13415-017-0499-7

Maki, W. S., & Padmanabhan, G. (1994). Transient suppression of processing during rapid serial

visual presentation: Acquired distinctiveness of probes modulates the attentional blink. 198 REFERENCES

Psychonomic Bulletin & Review. https://doi.org/10.3758/BF03210954

Mather, M., & Carstensen, L. L. (2003). Aging and attentional biases for emotional faces.

Psychological Science. https://doi.org/10.1111/1467-9280.01455

Mathews, A., & Mackintosh, B. (1998). A cognitive model of selective processing in anxiety.

Cognitive Therapy and Research. https://doi.org/10.1023/A:1018738019346

Mathews, A., & Macleod, C. (1985). Selective Cues Processing in Anxiety of Threat. Behavior

Research Ther. https://doi.org/10.1016/0005-7967(90)90160-K

Mathews, A., & MacLeod, C. (1985). Selective processing of threat cues in anxiety states.

Behaviour Research and Therapy, 23(5), 563–569. https://doi.org/10.1016/0005-

7967(85)90104-4

Mathews, A., & MacLeod, C. (2002). Induced processing biases have causal effects on anxiety.

Cognition and Emotion. https://doi.org/10.1080/02699930143000518

McCormick, P. A. (1997). Orienting Attention Without Awareness. Journal of Experimental

Psychology: Human Perception and Performance. https://doi.org/10.1037/0096-

1523.23.1.168

McLaughlin, K. A., & Nolen-Hoeksema, S. (2011). Rumination as a transdiagnostic factor in

depression and anxiety. Behaviour Research and Therapy.

https://doi.org/10.1016/j.brat.2010.12.006

McNally, R. J., Kaspi, S. P., Riemann, B. C., & Zeitlin, S. B. (1990). Selective processing of threat

cues in posttraumatic stress disorder. Journal of Abnormal Psychology.

https://doi.org/10.1037/0021-843X.99.4.398

McNally, Richard J. (2019). Attentional bias for threat: Crisis or opportunity? Clinical Psychology REFERENCES 199

Review. https://doi.org/10.1016/j.cpr.2018.05.005

Meyer, T. J., Miller, M. L., Metzger, R. L., & Borkovec, T. D. (1990). Development and validation

of the penn state worry questionnaire. Behaviour Research and Therapy.

https://doi.org/10.1016/0005-7967(90)90135-6

Milkins, B., Notebaert, L., MacLeod, C., & Clarke, P. J. F. (2016). The Potential Benefits of

Targeted Attentional Bias Modification on Cognitive Arousal and Sleep Quality in Worry-

Related Sleep Disturbance. Clinical Psychological Science.

https://doi.org/10.1177/2167702615626898

Mogg, K., & Bradley, B. P. (1998). A cognitive-motivational analysis of anxiety. Behaviour

Research and Therapy. https://doi.org/10.1016/S0005-7967(98)00063-1

Mogg, K., Bradley, B. P., & Hallowell, N. (1994). Attentional Bias to Threat: Roles of Trait

Anxiety, Stressful Events, and Awareness. The Quarterly Journal of Experimental

Psychology Section A. https://doi.org/10.1080/14640749408401099

Mogg, K., McNamara, J., Powys, M., Rawlinson, H., Seiffer, A., & Bradley, B. P. (2000). Selective

attention to threat: A test of two cognitive models of anxiety. Cognition and Emotion.

https://doi.org/10.1080/026999300378888

Mogoaşe, C., David, D., & Koster, E. H. W. (2014). Clinical efficacy of attentional bias

modification procedures: An updated meta-analysis. Journal of Clinical Psychology, 70(12),

1133–1157. https://doi.org/10.1002/jclp.22081

Most, S. B., Chun, M. M., Widders, D. M., & Zald, D. H. (2005). Attentional rubbernecking:

Cognitive control and personality in emotion-induced blindness. Psychonomic Bulletin &

Review, 12(4), 654–661. https://doi.org/10.3758/BF03196754 200 REFERENCES

Most, S. B., & Jungé, J. (2008). Don’t look back: Retroactive, dynamic costs and benefits of

emotional capture. Visual Cognition. https://doi.org/10.1080/13506280701490062

Most, S. B., Laurenceau, J. P., Graber, E., Belcher, A., & Smith, C. V. (2010). Blind Jealousy?

Romantic Insecurity Increases Emotion-Induced Failures of Visual Perception. Emotion.

https://doi.org/10.1037/a0019007

Most, S. B., Simons, D. J., Scholl, B. J., & Chabris, C. F. (2000). Sustained Inattentional Blindness.

Psyche.

Most, S. B., Smith, S. D., Cooter, A. B., Levy, B. N., & Zald, D. H. (2007). The naked truth: Positive,

arousing distractors impair rapid target perception. Cognition and Emotion, 21(5), 964–

981. https://doi.org/10.1080/02699930600959340

Most, S. B., & Wang, L. (2011). Dissociating spatial attention and awareness in emotion-induced

blindness. Psychological Science. https://doi.org/10.1177/0956797610397665

Nee, D. E., Wager, T. D., & Jonides, J. (2007). Interference resolution: Insights from a meta-

analysis of neuroimaging tasks. Cognitive, Affective and Behavioral Neuroscience.

https://doi.org/10.3758/CABN.7.1.1

Neuroscience, H., Mchugo, M., Olatunji, B. O., & Zald, D. H. (2013). The emotional attentional

blink : what we know so far, 7(April), 1–9. https://doi.org/10.3389/fnhum.2013.00151

Nieuwenstein, M. R., Chun, M. M., Van Der Lubbe, R. H. J., & Hooge, I. T. C. (2005). Delayed

attentional engagement in the attentional blink. Journal of Experimental Psychology:

Human Perception and Performance, 31(6), 1463–1475. https://doi.org/10.1037/0096-

1523.31.6.1463

Nosek, B. A., & Lakens, D. (2014). Registered reports: A method to increase the credibility of REFERENCES 201

published results. Social Psychology. https://doi.org/10.1027/1864-9335/a000192

Notebaert, L., Clarke, P. J. F., Grafton, B., & Macleod, C. (2015). Behaviour Research and

Therapy Validation of a novel attentional bias modi fi cation task : The future may be in the

cards. Behaviour Research and Therapy, 65, 93–100.

https://doi.org/10.1016/j.brat.2014.12.007

Notebaert, L., Clarke, P. J. F., Grafton, B., & MacLeod, C. (2015). Validation of a novel

attentional bias modification task: The future may be in the cards. Behaviour Research and

Therapy. https://doi.org/10.1016/j.brat.2014.12.007

Nunnally, J. . (1978). Psychometric Theory(2nd ed). Journal of Chemical Information and

Modeling. https://doi.org/10.1017/CBO9781107415324.004

Öhman, A., Flykt, A., & Esteves, F. (2001). Emotion drives attention: Detecting the snake in the

grass. Journal of Experimental Psychology: General. https://doi.org/10.1037/0096-

3445.130.3.466

Ohman, A., & Mineka, S. (2001). Fears, phobias, and preparedness: toward an evolved module

of fear and fear learning. Psychological Review, 108(3), 483–522. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/11488376

Olatunji, B. O., Armstrong, T., McHugo, M., & Zald, D. H. (2013). Heightened attentional capture

by threat in veterans with PTSD. Journal of Abnormal Psychology.

https://doi.org/10.1037/a0030440

Olatunji, B. O., Ciesielski, B. G., Armstrong, T., Zhao, M., & Zald, D. H. (2011). Making something

out of nothing: Neutral content modulates attention in generalized anxiety disorder.

Depression and Anxiety. https://doi.org/10.1002/da.20806 202 REFERENCES

Onie, S., & Most, S. B. (2017). Two Roads Diverged: Distinct Mechanisms of Attentional Bias

Differentially Predict Negative Affect and Persistent Negative Thought. Emotion.

https://doi.org/10.1037/emo0000280

Onie, Sandersan, Notebaert, L., Clarke, P., & Most, S. B. (2019). Investigating the effects of

inhibition training on attentional bias change: A simple Bayesian approach. Frontiers in

Psychology. https://doi.org/10.3389/fpsyg.2018.02782

Peckham, A. D., McHugh, R. K., & Otto, M. W. (2010). A meta-analysis of the magnitude of

biased attention in depression. Depression and Anxiety, 27(12), 1135–1142.

https://doi.org/10.1002/da.20755

Pessoa, L., & Adolphs, R. (2010, November). Emotion processing and the amygdala: From a “low

road” to “many roads” of evaluating biological significance. Nature Reviews Neuroscience.

https://doi.org/10.1038/nrn2920

Phaf, R. H., & Kan, K. J. (2007). The automaticity of emotional Stroop: A meta-analysis. Journal

of Behavior Therapy and Experimental Psychiatry.

https://doi.org/10.1016/j.jbtep.2006.10.008

Posner, M. I., Cohen, Y., & Rafal, R. D. (1982). Neural systems control of spatial orienting.

Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences.

https://doi.org/10.1098/rstb.1982.0081

Posner, Michael I., Rafal, R. D., Choate, L. S., & Vaughan, J. (1985). Inhibition of Return: Neural

Basis and Function. Cognitive Neuropsychology.

https://doi.org/10.1080/02643298508252866

Posner, Michael I., Snyder, C. R., & Davidson, B. J. (1980). Attention and the detection of signals. REFERENCES 203

Journal of Experimental Psychology: General. https://doi.org/10.1037/0096-

3445.109.2.160

Potter, M. (1975). Meaning in visual search. Science. https://doi.org/10.1126/science.1145183

Potter, M. C., Staub, A., & O’Connor, D. H. (2002). The Time Course of Competition for

Attention: Attention Is Initially Labile. Journal of Experimental Psychology: Human

Perception and Performance. https://doi.org/10.1037/0096-1523.28.5.1149

Potter, M. C., Wyble, B., Hagmann, C. E., & McCourt, E. S. (2014). Detecting meaning in RSVP at

13 ms per picture. Attention, Perception, and Psychophysics.

https://doi.org/10.3758/s13414-013-0605-z

Pratto, F., & John, O. P. Automatic vigilance: The attention-grabbing power of negative social

information., 61 Journal of Personality and Social Psychology § (1991). US: American

Psychological Association. https://doi.org/10.1037/0022-3514.61.3.380

Price, R. B., Brown, V., & Siegle, G. J. (2019). Computational Modeling Applied to the Dot-Probe

Task Yields Improved Reliability and Mechanistic Insights. Biological Psychiatry, 85(7), 606–

612. https://doi.org/10.1016/j.biopsych.2018.09.022

Price, R. B., Kuckertz, J. M., Siegle, G. J., Ladouceur, C. D., Silk, J. S., Ryan, N. D., … Amir, N.

(2015). Empirical recommendations for improving the stability of the dot-probe task in

clinical research. Psychological Assessment. https://doi.org/10.1037/pas0000036

Price, R. B., Wallace, M., Kuckertz, J. M., Amir, N., Graur, S., Cummings, L., … Bar-Haim, Y.

(2016). Pooled patient-level meta-analysis of children and adults completing a computer-

based anxiety intervention targeting attentional bias. Clinical Psychology Review.

https://doi.org/10.1016/j.cpr.2016.09.009 204 REFERENCES

Pylyshyn, Z. W. (1999). Is vision continuous with cognition? The case for cognitive

impenetrability of visual perception. Behavioral and Brain Sciences, 341–423.

https://doi.org/10.1017/S0140525X99002022

Raymond, J. E., Shapiro, K. L., & Arnell, K. M. (1992). Temporary Suppression of Visual

Processing in an RSVP Task: An Attentional Blink? Journal of Experimental Psychology:

Human Perception and Performance. https://doi.org/10.1037/0096-1523.18.3.849

Rinck, M., Becker, E. S., Kellermann, J., & Roth, W. T. (2003). Selective attention in anxiety:

Distraction and enhancement in visual search. Depression and Anxiety.

https://doi.org/10.1002/da.10105

RL, S., Kroenke, K., JW, W., & Löwe, B. (2006). A brief measure for assessing generalized anxiety

disorder: The gad-7. Archives of Internal Medicine, 166(10), 1092–1097. Retrieved from

http://dx.doi.org/10.1001/archinte.166.10.1092

Rolke, B., Heil, M., Streb, J., & Hennighausen, E. (2001). Missed prime words within the

attentional blink evoke an N400 semantic priming effect. Psychophysiology.

https://doi.org/10.1017/S0048577201991504

Rosenthal, R. (1979). The file drawer problem and tolerance for null results. Psychological

Bulletin. https://doi.org/10.1037/0033-2909.86.3.638

Rouder, J. N., Kumar, A., & Haaf, J. M. (2019). Why Most Studies of Individual Differences with

Inhibition Tasks Are Bound to Fail. PsyArxiv.

Rouder, J. N., & Morey, R. D. (n.d.). Multivariate Behavioral Default Bayes Factors for Model

Selection in Regression, (February 2013), 37–41.

Rouder, J. N., Morey, R. D., Speckman, P. L., & Province, J. M. (2012). Default Bayes factors for REFERENCES 205

ANOVA designs. Journal of Mathematical Psychology.

https://doi.org/10.1016/j.jmp.2012.08.001

Rudaizky, D., Basanovic, J., & MacLeod, C. (2014). Biased attentional engagement with, and

disengagement from, negative information: Independent cognitive pathways to anxiety

vulnerability? Cognition and Emotion. https://doi.org/10.1080/02699931.2013.815154

Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology,

39(6), 1161–1178. https://doi.org/10.1037/h0077714

Russell, J. A., & Barrett, L. F. Core affect, prototypical emotional episodes, and other things

called emotion: Dissecting the elephant., 76 Journal of Personality and Social Psychology §

(1999). US: American Psychological Association. https://doi.org/10.1037/0022-

3514.76.5.805

Schmidt, N. B., Richey, J. A., Buckner, J. D., & Timpano, K. R. (2009). Attention Training for

Generalized Social Anxiety Disorder. Journal of Abnormal Psychology.

https://doi.org/10.1037/a0013643

Schmukle, S. C. (2005). Unreliability of the dot probe task. European Journal of Personality.

https://doi.org/10.1002/per.554

Schwabe, L., Merz, C. J., Walter, B., Vaitl, D., Wolf, O. T., & Stark, R. (2011). Emotional

modulation of the attentional blink: The neural structures involved in capturing and

holding attention. Neuropsychologia.

https://doi.org/10.1016/j.neuropsychologia.2010.12.037

Schwabe, L., & Wolf, O. T. (2010). Emotional Modulation of the Attentional Blink: Is There an

Effect of Stress? Emotion. https://doi.org/10.1037/a0017751 206 REFERENCES

See, J., MacLeod, C., & Bridle, R. (2009). The Reduction of Anxiety Vulnerability Through the

Modification of Attentional Bias: A Real-World Study Using a Home-Based Cognitive Bias

Modification Procedure. Journal of Abnormal Psychology.

https://doi.org/10.1037/a0014377

Shanks, D. R., Vadillo, M. A., Riedel, B., Clymo, A., Govind, S., Hickin, N., … Puhlmann, L. M. C.

(2015). Journal of Experimental Psychology : General Romance , Risk , and Replication :

Can Consumer Choices and Risk-Taking Be Primed by Mating Motives ? Romance , Risk ,

and Replication : Can Consumer Choices and Risk-Taking Be Primed by Mating Motives ?

Journal of Experimental Psychology: General, 144(6), 142–158.

Shanks, D. R., Vadillo, M. A., Riedel, B., Clymo, A., Govind, S., Hickin, N., … Puhlmann, L. M. C.

(2015). Romance, risk, and replication: Can consumer choices and risk-taking be primed by

mating motives? Journal of Experimental Psychology: General.

https://doi.org/10.1037/xge0000116

Shapiro, K., Driver, J., Ward, R., & Sorensen, R. E. (1997). Priming from the attentional blink: A

Failure to Extract Visual Tokens but Not Visual Types. Psychological Science.

https://doi.org/10.1111/j.1467-9280.1997.tb00689.x

Shapiro, K. L., Raymond, J. E., & Arnell, K. M. (1994). Attention to Visual Pattern Information

Produces the Attentional Blink in Rapid Serial Visual Presentation. Journal of Experimental

Psychology: Human Perception and Performance. https://doi.org/10.1037/0096-

1523.20.2.357

Shechner, T., Pelc, T., Pine, D. S., Fox, N. A., & Bar-Haim, Y. (2012). Flexible attention

deployment in threatening contexts: An instructed fear conditioning study. Emotion. REFERENCES 207

https://doi.org/10.1037/a0027072

Siegrist, M. (1997). Test-retest reliability of different versions of the stroop test. Journal of

Psychology: Interdisciplinary and Applied. https://doi.org/10.1080/00223989709603516

Sigurjónsdóttir, Ó., Sigurardóttir, S., Björnsson, A. S., & Kristjánsson, Á. (2015). Barking up the

wrong tree in attentional bias modification? Comparing the sensitivity of four tasks to

attentional biases. Journal of Behavior Therapy and Experimental Psychiatry, 48, 9–16.

https://doi.org/10.1016/j.jbtep.2015.01.005

Simons, D. J., Shoda, Y., & Lindsay, D. S. (2017). Constraints on Generality (COG): A Proposed

Addition to All Empirical Papers. Perspectives on Psychological Science.

https://doi.org/10.1177/1745691617708630

Simonsohn, U., Nelson, L. D., & Simmons, J. P. (2014). p-Curve and Effect Size: Correcting for

Publication Bias Using Only Significant Results. Perspectives on Psychological Science.

https://doi.org/10.1177/1745691614553988

Singh, D., & Sunny, M. M. (2017). Emotion induced blindness is more sensitive to changes in

arousal as compared to valence of the emotional distractor. Frontiers in Psychology,

8(AUG), 1–8. https://doi.org/10.3389/fpsyg.2017.01381

Slagter, H. A., Lutz, A., Greischar, L. L., Francis, A. D., Nieuwenhuis, S., Davis, J. M., & Davidson,

R. J. (2007). of Limited Brain Resources, 5(6).

https://doi.org/10.1371/journal.pbio.0050138

Smith, J. M., & Alloy, L. B. (2010). Rumination theories. Psychology, 29(2), 116–128.

https://doi.org/10.1016/j.cpr.2008.10.003.A

Smith, S. D., Most, S. B., Newsome, L. A., & Zald, D. H. (2006). An emotion-induced attentional 208 REFERENCES

blink elicited by aversively conditioned stimuli. Emotion, 6(3), 523–527.

https://doi.org/10.1037/1528-3542.6.3.523

Soto, D., Heinke, D., Humphreys, G. W., & Blanco, M. J. (2005). Early, involuntary top-down

guidance of attention from working memory. Journal of Experimental Psychology: Human

Perception and Performance. https://doi.org/10.1037/0096-1523.31.2.248

Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970). STAI manual for the state-trait anxiety

inventory. Self-Evaluation Questionnaire. MANUAL. https://doi.org/10.1037/t06496-000

Spitzer, R. L., Kroenke, K., Williams, J. B. W., & Löwe, B. (2006). A brief measure for assessing

generalized anxiety disorder: The GAD-7. Archives of Internal Medicine.

https://doi.org/10.1001/archinte.166.10.1092

Staugaard, S. R. (2009). Reliability of two versions of the dot-probe task using photographic

faces. Psychology Science Quarterly.

Stöber, J. (1993). Anxiety: The cognitive perspective, by Michael W. Eysenck Hove, UK:

Lawrence erlbaum; 1992; 195 pp. . Anxiety, Stress & Coping.

https://doi.org/10.1080/10615809308248376

Sutton, T. M., & Lutz, C. (2018). Attentional Capture for Emotional Words and Images: The

Importance of Valence and Arousal. Canadian Journal of Experimental Psychology, 73(1),

47–54. https://doi.org/10.1037/cep0000154

Talbot, C. E., Ksander, J. C., & Gutchess, A. (2018). Aging impairs disengagement from negative

words in a dot probe task. Frontiers in Psychology.

https://doi.org/10.3389/fpsyg.2018.02361

Tang, M. F., Badcock, D. R., & Visser, T. A. W. (2014). Training and the attentional blink: Limits REFERENCES 209

overcome or expectations raised? Psychonomic Bulletin and Review.

https://doi.org/10.3758/s13423-013-0491-3

Treynor, W., Gonzalez, R., & Nolen-Hoeksema, S. (2003). Rumination reconsidered: A

psychometric analysis. Cognitive Therapy and Research.

https://doi.org/10.1023/A:1023910315561

Van Bockstaele, B., Verschuere, B., Tibboel, H., De Houwer, J., Crombez, G., & Koster, E. H. W.

(2014). A review of current evidence for the causal impact of attentional bias on fear and

anxiety. Psychological Bulletin, 140(3), 682–721. https://doi.org/10.1037/a0034834

Visser, T. A. W., Merikle, P. M., & Di Lollo, V. (2005). Priming in the attentional blink: Perception

without awareness? Visual Cognition, 12(7), 1362–1372.

https://doi.org/10.1080/13506280444000733

Vogt, J., De Houwer, J., Koster, E. H. W., Van Damme, S., & Crombez, G. (2008). Allocation of

Spatial Attention to Emotional Stimuli Depends Upon Arousal and Not Valence. Emotion.

https://doi.org/10.1037/a0013981

Wagenmakers, E. J., Beek, T. F., Rotteveel, M., Gierholz, A., Matzke, D., Steingroever, H., …

Pinto, Y. (2015). Turning the hands of time again: A purely confirmatory replication study

and a Bayesian analysis. Frontiers in Psychology, 6(APR), 1–6.

https://doi.org/10.3389/fpsyg.2015.00494

Wang, L., Kennedy, B. L., & Most, S. B. (2012). When emotion blinds: A spatiotemporal

competition account of emotion-induced blindness. Frontiers in Psychology.

https://doi.org/10.3389/fpsyg.2012.00438

Wieser, M. J., Pauli, P., & Mühlberger, A. (2009). Probing the attentional control theory in social 210 REFERENCES

anxiety: An emotional saccade task. Cognitive, Affective and Behavioral Neuroscience.

https://doi.org/10.3758/CABN.9.3.314

Woodman, G. F., & Luck, S. J. (2003). Serial Deployment of Attention During Visual Search.

Journal of Experimental Psychology: Human Perception and Performance.

https://doi.org/10.1037/0096-1523.29.1.121

Wyble, B., & Swan, G. (2015). Mapping the spatiotemporal dynamics of interference between

two visual targets. Attention, Perception, and Psychophysics, 2331–2343.

https://doi.org/10.3758/s13414-015-0938-x

Yang, E., Zald, D. H., & Blake, R. (2007). Fearful Expressions Gain Preferential Access to

Awareness During Continuous Flash Suppression. Emotion. https://doi.org/10.1037/1528-

3542.7.4.882

Yeung, E. S., & Sharpe, L. (2019). Cognitive Bias Modification for Social Anxiety: The Differential

Impact of Modifying Attentional and/or Interpretation Bias. Cognitive Therapy and

Research. https://doi.org/10.1007/s10608-019-10012-3

Yiend, J., & Mathews, A. (2001). Anxiety and attention to threatening pictures. Quarterly

Journal of Experimental Psychology Section A: Human Experimental Psychology.

https://doi.org/10.1080/713755991

Yuko Hakamata, PhD, Shmuel Lissek, PhD, Yair Bar-Haim, PhD, Jennifer C. Britton, PhD, Nathan

Fox, PhD, Ellen Leibenluft, MD, Monique Ernst, MD, PhD, and Daniel S. Pine, M., Hakamata,

Y., Lissek, S., Bar-Haim, Y., Britton, J. C., Fox, N. A., … Pine, D. S. (2010). Attention Bias

Modification Treatment: A meta-analysis towards the establishment of novel treatment

for anxiety. Biological Psychiatry, 68(11), 982–990. REFERENCES 211

https://doi.org/10.1016/j.biopsych.2010.07.021

Zhao, J. L., & Most, S. B. (2019). Manipulations of distractor frequency do not mitigate emotion-

induced blindness. Cognition and Emotion.

https://doi.org/10.1080/02699931.2018.1459490

Zinbarg, R. E., Revelle, W., Yovel, I., & Li, W. (2005). Cronbach’s, α Revelle’s β and McDonald’s ω

H: Their relations with each other and two alternative conceptualizations of reliability.

Psychometrika. https://doi.org/10.1007/s11336-003-0974-7

212 REFERENCES

Thank you for reading this thesis.