UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Tracking attention in space and time : the dynamics of human visual attention Jefferies, Lisa N. 2009

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

Item Metadata

Download

Media
24-ubc_2009_fall_jefferies_lisa.pdf [ 1.26MB ]
Metadata
JSON: 24-1.0067342.json
JSON-LD: 24-1.0067342-ld.json
RDF/XML (Pretty): 24-1.0067342-rdf.xml
RDF/JSON: 24-1.0067342-rdf.json
Turtle: 24-1.0067342-turtle.txt
N-Triples: 24-1.0067342-rdf-ntriples.txt
Original Record: 24-1.0067342-source.json
Full Text
24-1.0067342-fulltext.txt
Citation
24-1.0067342.ris

Full Text

TRACKING ATTENTION IN SPACE AND TIME: THE DYNAMICS OF HUMAN VISUAL ATTENTION  by  LISA N. JEFFERIES  B.A., Simon Fraser University, 2002 M.A., The University of British Columbia, 2005  A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE STUDIES  (Psychology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  July, 2009 © Lisa N. Jefferies, 2009  ABSTRACT Attention is essential to everyday life: without some selective function to guide and limit the processing of incoming information, our visual system would be overwhelmed. A description of the spatiotemporal dynamics of attention is critical to our understanding of this basic human cognitive function and is the primary goal of this dissertation. In particular, the research reported here is aimed at examining two aspects of the spatiotemporal dynamics of attention: a) the rate at which the focus of attention is shrunk and expanded along with the factors that influence this rate, and b) the factors governing whether attention is deployed as either a unitary or a divided focus. The present research examines the spatiotemporal dynamics of focal attention by monitoring the pattern of accuracy that occurs when participants attempt to identify two targets embedded in simultaneously presented streams of items. By asking participants to monitor these streams simultaneously, with the spatial and temporal positions of the two targets in the streams being varied incrementally, it is possible to index the extent of focal attention in both space and time. Chapter 2 develops this behavioural procedure and assesses the rate at which the focus of attention is contracted. A qualitative model is put forward and tested. Chapter 3 examines factors that modulate the temporal course of attentional narrowing in young adults who presumably can exercise efficient control of attentional processes. In contrast, Chapter 4 examines the effect of reduced attentional control by examining the same process in older adults. The second goal of this thesis was to examine whether focal attention is deployed as a unitary or a divided focus. These two perspectives are generally viewed as mutually  ii  exclusive. The alternative hypothesis pursued in Chapter 5 is that focal attention can be deployed as either a single, unitary focus or divided into multiple foci, depending on the observers mental set and on the task demands. The final chapter then combines and compares the findings across all experiments and evaluates how they fit in with current theories of visual attention.  iii  TABLE OF CO TE TS  ABSTRACT ....................................................................................................................... ii TABLE OF CO TE TS ................................................................................................. iv LIST OF FIGURES ....................................................................................................... viii LIST OF TABLES ............................................................................................................ x ACK OWLEDGEME TS ............................................................................................ xii DEDICATIO ................................................................................................................ xiii CHAPTER 1: GE ERAL I TRODUCTIO .............................................................. 1 I.I A Zoom-Lens Conception of Focal Attention ........................................................... 2 I.2 Unitary vs. Divided Focus of Attention .................................................................... 5 I.2.1 Evidence for a Unitary Focus of Attention ............................................................ 6 I.2.2 Evidence for a Divided Focus of Attention ....................................................... 11 I.3 Dissertation Objectives ............................................................................................. 18 References ........................................................................................................................ 20 CHAPTER 2: LI EAR CHA GES I THE SPATIAL EXTE T OF THE FOCUS OF ATTE TIO ACROSS TIME ............................................................................... 24 Lag-1 Sparing Across Space ........................................................................................... 26 Control of Attention in the Spatiotemporal Domain ................................................... 27 Experiment 2.1 ................................................................................................................. 31 Method .......................................................................................................................... 31 Observers ....................................................................................................................... 31 Apparatus and Stimuli ................................................................................................... 31 Procedure ....................................................................................................................... 32 Results and Discussion .................................................................................................. 34 Second-target Accuracy ................................................................................................ 34 iv  Lag-1 Sparing: Comparing Predicted and Obtained Patterns. ...................................... 36 The Effect of Masking................................................................................................... 38 First-Target Accuracy. .................................................................................................. 40 Temporal Dynamics of the Focus of Attention ............................................................. 42 Experiment 2.2 ................................................................................................................. 44 Method .......................................................................................................................... 47 Observers ....................................................................................................................... 47 Procedure ....................................................................................................................... 47 Results and Discussion .................................................................................................. 47 Experiment 2.3 ................................................................................................................. 50 Method .......................................................................................................................... 50 Observers ....................................................................................................................... 50 Procedure ....................................................................................................................... 51 Results and Discussion .................................................................................................. 51 General Discussion .......................................................................................................... 53 Spatial Dynamics of the Focus of Attention: Analog or quantal?................................. 54 The Focus of Attention: Unitary or Divided? .............................................................. 55 CHAPTER 3. ADJUSTI G THE FOCUS OF ATTE TIO ACROSS TIME: FACTORS WHICH I FLUE CE THE RATE OF ARROWI G ........................ 64 Experiment 3.1 ................................................................................................................. 68 Method .......................................................................................................................... 71 Participants .................................................................................................................... 71 Apparatus and Stimuli ................................................................................................... 71 Procedure ....................................................................................................................... 72 Results and Discussion .................................................................................................. 74 Second-target Accuracy ................................................................................................ 74 Experiment 3.2 ................................................................................................................. 80 Procedure ....................................................................................................................... 82 Results and Discussion .................................................................................................. 83 Experiment 3.3 ................................................................................................................. 85 Participants .................................................................................................................... 86 Stimuli and Procedures.................................................................................................. 86 Results and Discussion .................................................................................................. 87 Summary .......................................................................................................................... 89 References ........................................................................................................................ 92 CHAPTER 4 ASSESSI G THE RATE AT WHICH FOCAL ATTE TIO ARROWS I OLDER ADULTS ................................................................................ 94 v  Experiment 4.1 ................................................................................................................. 99 Method .......................................................................................................................... 99 Observers ....................................................................................................................... 99 Apparatus and Stimuli ................................................................................................. 100 Procedure ..................................................................................................................... 100 Results ......................................................................................................................... 102 Discussion ................................................................................................................... 107 The Rate of Narrowing Focal Attention ...................................................................... 110 Experiment 4.2 ............................................................................................................... 113 Observers ..................................................................................................................... 113 Stimuli and Procedure ................................................................................................. 113 Results and Discussion ................................................................................................ 114 Experiment 4.3 ............................................................................................................... 115 Observers ..................................................................................................................... 116 Stimuli and Procedure ................................................................................................. 116 Results and Discussion ................................................................................................ 117 Conclusions .................................................................................................................... 119 Hemifield Effects ........................................................................................................... 120 CHAPTER 5. IS FOCAL ATTE TIO U ITARY OR DIVIDED? IT DEPE DS O TASK DEMA DS .................................................................................................. 127 Experiment 5.1 ............................................................................................................... 129 Methods ....................................................................................................................... 129 Participants .................................................................................................................. 129 Stimuli and Procedure ................................................................................................. 129 Results and Discussion ................................................................................................ 132 Experiment 5.2 ............................................................................................................... 135 Method ........................................................................................................................ 136 Observers ..................................................................................................................... 136 Apparatus and Stimuli ................................................................................................. 137 Results and Discussion ................................................................................................ 137 References ...................................................................................................................... 141 CHAPTER 6: GE ERAL DISCUSSIO .................................................................. 143 Summary ........................................................................................................................ 143 arrowing the Focus of Attention to the Location of the First Target .................... 146 Comparing Results Across Experiments..................................................................... 148 vi  Incorporating ew Data into Current Models of Visuospatial Attention ............... 151 The Human Attentional etwork Model .................................................................... 153 Orienting and the Posterior Attention Network .......................................................... 154 Incorporating a Focus of Attention that Expands and Contracts ................................ 157 Incorporating a Flexibly Divided Focus of Attention ................................................. 158 The Episodic Theory of Attention ............................................................................... 159 Incorporating a Focus of Attention that Expands and Contracts ................................ 162 Incorporating a Flexibly Divided Focus of Attention ................................................. 162 The Activity Distribution Model .................................................................................. 164 The Deployment of Focal Attention According to the Activity Distribution Model .. 166 Incorporating a Focus of Attention that Expands and Contracts ................................ 173 Incorporating a Flexibly Divided Focus of Attention ................................................. 174 Contributions and Implications ................................................................................... 178 Possible Limitations ...................................................................................................... 179 Future Directions........................................................................................................... 181 References ...................................................................................................................... 183  vii  LIST OF FIGURES  Figure 2.1. Illustration of Theoretical Model ...…...……………………….....…….29 Figure 2.2. Illustration of Sequence of Events in Experiment 2.1 ......………..........34 Figure 2.3. Second-Target Accuracy Data from Experiment 2.1 …..……...….…..35 Figure 2.4. Second-Target Accuracy Data from Experiment 2.1 …..…………..….37 Figure 2.5. Second-Target Accuracy Data from Experiment 2.1 ..……….….……39 Figure 2.6. First-Target Accuracy Data from Experiment 2.1 ...………...….....…..41 Figure 2.7. Magnitude of Lag-1 Sparing in Experiment 2.1 …………...…….….….44 Figure 2.8. Illustration of Theoretical Model ..…………………..………….....……46 Figure 2.9. Second-Target Accuracy Data from Experiment 2.3 ...………..……….52 Figure 3.1. Illustration of Theoretical Model ……….………………………..…..….66 Figure 3.2. Schematic illustration of the Sequence of Events in Experiment 3.1 .…70 Figure 3.3. Second-Target Accuracy Data from Experiment 3.1 ...………......…….75 Figure 3.4. Magnitude of Lag-1 Sparing in Experiment 3.1 …………….…....…….76 Figure 3.5. Schematic Illustration of the Sequence of Events in Experiment 3.2 …82 Figure 3.6. Second-Target Accuracy Data from Experiment 3.2 ...…………….…..84 Figure 3.7. Magnitude of Lag-1 Sparing in Experiment 3.2 ……………….…....….85 Figure 3.8. Second-Target Accuracy Data from Experiment 3.3 ………………..…87 Figure 3.9. Magnitude of Lag-1 Sparing in Experiment 3.3 ...……………………...89 Figure 4.1. Illustration of Theoretical Model………………………………………...98 Figure 4.2. Schematic Illustration of Sequence of Events in Experiment 4.1…..…101 Figure 4.3. Second-Target Accuracy Data from Experiment 4.1 ...………......…..103  viii  Figure 4.4. Lag-1 Sparing Results from Experiment 4.2 ...……..………...……......112 Figure 4.5. Second-Target Accuracy Data from Experiment 4.2 ..……….……....114 Figure 4.6. Second-Target Accuracy Data from Experiment 4.3…………....…….117 Figure 5.1. Illustration of Sequence of Events and Data from Experiment 5.1..…132 Figure 5.2. Second-Target Accuracy Data from Experiment 5.2 ...…………...…..139 Figure 6.1. Graph Comparing Results of Experiments in Chapters 2 – 5………...149 Figure 6.2. Components of the Activity Distribution Model ..……………...……..165  ix  LIST OF TABLES Table 4.1. Table Listing First-Target Identification Accuracy for Young and Older Adults …………………………………………………………………....110  x  PREFACE The following is a manuscript-based rather than traditional dissertation. It begins with an introduction to the question addressed by this dissertation, both in terms of situating this question in the previous literature and in introducing the methodology used throughout the dissertation (Chapter 1). The following four self-contained chapters (Chapters 2-5) are manuscripts that are either submitted to or are in press in Journal of Experimental Psychology: Human Perception and Performance or are prepared for submission to similar journals. As a result of the choice to use a manuscript-based dissertation and to provide a general context for each manuscript, several sections of this document will appear similar and perhaps redundant. For instance, there is a great deal of similarity and overlap in the introduction to Chapter 1 and the introduction of various manuscripts presented throughout the dissertation, as well as between the method sections of the manuscripts presented in Chapters 2-5. Chapter 6 is a general discussion and contains several comparisons and analyses between the results presented in the previous chapters, as well as discussions and conclusions that relate the manuscripts together.  xi  ACK OWLEDGEME TS I wish to first thank my supervisor, Dr. Jim Enns, for his untiring dedication throughout my graduate training. Without his willingness to so generously share his time and knowledge, I would not have been able to accomplish what I have so far. I would also like to thank all of the dedicated members of my Ph.D. committee for their time and invaluable comments. It has been my great fortune to work with many enthusiastic professors and researchers and to be inspired by them: Dr. Chris Davis who turned my sights to Psychology, Dr. Vito Modigliani whose passion for cognitive psychology and dedication to his students kindled my own passion for cognitive psychology, and Dr. Richard Wright who funneled that passion towards the areas of attention and perception and who taught me a love for theoretical models. Thank you! The research presented in this dissertation was supported by an Alexander Graham Bell Canada Graduate Scholarship to LNJ, by grants from the Natural Sciences and Engineering Research Council of Canada to JTE and VDL, and by a Senior Graduate Trainee Scholarship to LNJ from the Michael Smith Foundation for Health Research. Chapter 2: We thank Tram Neill, Mark Nieuwenstein, and Brad Wyble for helpful comments on an earlier version of this in-press manuscript. Chapter 3: We thank Rachel Fouladi and Bruno Zumbo for assistance with statistical analysis of the data.  xii  DEDICATIO  Many people teach you a few things in life, A few people teach you many things. But one or two truly teach you everything you know and make you who you are. It is those who teach your heart to beat, your soul to sing … those who are with you always.  To my parents: To my Mom who walked every step of the way with me, not just as mother but as bestfriend. My deepest love and thanks. And to my Dad who would have been so proud.  To Vince Di Lollo, Who makes research a fascinating voyage of discovery, who has inspired me with his passion for science and the brain, and who has truly been my mentor. Grazie!  xiii  STATEME T OF CO-AUTHORSHIP The manuscripts presented in this dissertation were co-authored by my supervisor, Dr. James T. Enns and my co-supervisor, Dr. Vincent Di Lollo, who supervised me in each stage of the research process. Chapter 3 was also co-authored by Dr. Shahab Ghorashi, and Chapter 4 was co-authored by Dr. Alexa Roggeveen, Dr. Patrick Bennett, and Dr. Alison Sekuler. Under my supervisors’ guidance, I developed the theory that is examined in the dissertation, designed each of the studies, tested most of the participants, and performed all of the statistical analysis. I also prepared the first draft of the manuscripts and then worked closely with my supervisors to finalize them for publication. To recognize these collaborative efforts, plural pronouns (e.g., we, our, us) are used throughout the manuscripts.  xiv  CHAPTER 1: General Introduction The visual world contains such a rich array of information that there is need for a selective mechanism to limit the information entering the visual system at any given moment. Visual attention serves this selective function by guiding visual processing to the most relevant information in a scene. There is little doubt that focused attention has been understood at an intuitive level since the emergence of consciousness; but the scientific study of human visual attention is a relatively recent endeavour. Substantial progress has been made in probing the functions of visual attention, but still many questions remain unanswered, including questions about how visual attention is deployed across space and time and what rules govern such deployment. A description of the spatiotemporal dynamics of attention is critical to our understanding of this most basic human cognitive function and is the primary goal of this dissertation. In particular, the research reported here is aimed at two aspects of the spatiotemporal dynamics of attention: a) the rate at which the focus of attention is shrunk and expanded and the factors that influence this rate, and b) the factors governing whether attention is deployed as either a unitary or a divided focus. In the experiments reported here, we explored the limits of human visual attention as it operates in healthy adults. While not taking an explicitly developmental perspective, the present work explored how the spatiotemporal dynamics of attention change as a function of ageing (Chapter 4). Chapter 2 develops a behavioural procedure to assess the rate at which the focus of attention is contracted. This procedure is based on the accuracy of perceptual report for targets that are separated by varying degrees of space and time. Small differences in the spatial or temporal separation of the stimuli or subtle changes to their characteristics result in predictable 1  changes in the spatiotemporal dynamics of focal attention, as is demonstrated in each of the chapters in this dissertation. The remainder of this chapter provides an overview of the existing literature on two aspects of the spatiotemporal dynamics of attention. The first section outlines the research showing that the focus of attention can be adjusted in spatial extent; the second section considers the research showing that the focus of attention is unitary and the research showing that it can be divided. The final section describes the specific hypotheses investigated in each chapter of this dissertation. I.I A Zoom-Lens Conception of Focal Attention There are many ways of conceptualizing focused attention. William James (1890) provided an early, intuitive perspective when he said that: “My experience is what I agree to attend to. Only those items which I notice, shape my mind – without selective interest, experience is an utter chaos.” Since then, many metaphors for focused attention have been employed, some of the more enduring being a spotlight of attention (Hernandez-Peon, 1964), a zoom-lens, (Eriksen & Yeh, 1985), and a Mind’s Eye (Jonides, 1980). Research has clearly demonstrated that the focus of attention is highly flexible. Not only can it be shifted from one location to another in order to process the most relevant information in a scene, but it can also be adjusted in spatial extent from broad to narrow and vice versa. The fact that focal attention can be shrunk or expanded to match the size of the area to be attended has been supported by a variety of studies (e.g., Egeth, 1977; Eriksen & Hoffman, 1974; Eriksen, Pan, & Botella, 1993; Eriksen & Rohrbaugh, 1970; Eriksen & St. James; Jonides, 1980, 1983; LaBerge, 1983; Murphy & Eriksen, 1987).  2  In general, it was found that there is an inverse relationship between the size of the focus of attention and the reaction time (RT) to stimuli appearing in the attended area (Egeth, 1977). If the focus of attention was broad processing speed was slower than if the focus of attention was narrow. This is consistent with the idea attentional resources are fixed in amount – if they must be distributed more diffusely over a larger region of space processing will be impaired while if they can be focused in a more concentrated manner on a smaller region, then processing will be enhanced. LaBerge (1983, 1995) probed in detail what the precise spatial limits of the focus of attention are, and found that the focus of attention could be reduced to as little as 0.3° of visual angle or expanded to an area of approximately 2° (see also LaBerge et al, 1991). Jonides (1980, 1983) built on findings such as these and conceptualized attention as having two discrete functional modes. One mode involved attention being distributed diffusely over a large portion of the visual field. The alternative mode was a highly-focused operation with resources being deployed to a small region of space. The two-stage model allows for the visual system to switch rapidly back and forth between these two modes of operation, but maintains that the two modes are distinct from one another. Perhaps the first formal statement of a focus of attention that could shrink and expand continuously was put forward by Eriksen and Yeh (1985) and developed further by Eriksen and colleagues (Eriksen, Pan, & Botella, 1993; Eriksen & St. James; Murphy & Eriksen, 1987). In the zoom lens model of attention, it is proposed that the focus of attention can shrink to a narrower, more “concentrated” beam if the task at-hand required detailed analysis of a small region of space. If a larger area needed analysis, on the other hand, then the focus of attention would be expanded, although this caused attention to be more diffuse within the attended region. The critical difference between this model and that put forward by Jonides (1983) is that the  3  zoom lens model suggests that the process of expanding and contracting the focus of attention is dynamic and the breadth of focal attention fluctuates continuously. The difference between a broad focus and a narrow one is no longer seen as being two discrete forms of processing but rather two poles on opposite ends of a continuum. More recently, Barriopedro and Botella (1998) used the attentional blink paradigm to test the proposal that a broad attentional focus results in a more diffuse dispersal of attentional resources and poorer performance than a narrower focus of attention. In a typical attentional blink paradigm, if two targets are presented in close temporal proximity to one another, the first target is identified accurately, the identification accuracy of the second target is very poor. Barriopedro and Botella presented two rapid serial visual presentation (RSVP) streams of distractors, presented 1.15°, 2.34°, or 3.53° of visual angle apart from one another. The two targets appeared one in each RSVP stream, and Barriopedro and Botella found that the second target was identified more accurately (i.e., the magnitude of the attentional blink was reduced) if the two RSVP streams were at a close spatial separation. This supports the idea of an inverse relationship between the size of the focus of attention and the processing resources available to items within that focus – a smaller focus leads to more concentrated attentional resources. As outlined above, there is clear evidence that the focus of attention can be varied in spatial extent – that is, it can be contracted or expanded as required by the task at hand. Unlike shifts of spatial attention, which have been probed in great detail with regards to their time course and the factors which influence their initiation and rate, relatively little is known about readjustments of the spatial extent of focal attention. To date, only a single study has examined the rate at which the focus of attention can be expanded from narrow to broad.  4  An estimate of the time required to expand focal attention has been provided by Benso, Turatto, Mascetti, and Umiltà (1998), who used a pre-cue to draw attention from the center of the screen to a randomly-chosen location, at which a cue was presented. The cue consisted of a ring with a diameter of either 2.5˚ or 7.5˚. After a variable delay, a target was presented within the cued area. The critical assumption was that, upon presentation of the cue, the attentional focus expanded to cover the cued area. The results suggested that the process of expansion was completed within about 33 to 66 ms. One limitation of Benso et al.’s (1998) study was that it was confined to the case in which the attentional spotlight was expanded. It is clear from the research outlined above that not only can the focus of attention be shifted from one location to another, it can also be expanded and contracted to encompass a larger or smaller region of space. Although it is apparent that this expanding/contracting process takes place, very little research has examined this particular aspect of the spatiotemporal dynamics of attention. In fact, basic aspects such as the rate at which the focus of attention narrows have not been examined. Chapter 2 of this thesis will provide such an estimate, and will be complemented by Chapters 3 and 4 which will examine some of the factors which influence the rate at which the focus of attention contracts and which factors initiate or delay the initiation of the narrowing process. 1.2 Unitary vs. Divided Focus of Attention This dissertation has a two-fold goal with regards to clarifying the spatiotemporal dynamics of attention. The first is to examine the process by which the focus of attention expands and contracts – to measure the time course of the narrowing process and to identify and map out factors which influence the rate at which the narrowing process occurs. Research relevant to this goal was described in the section above. The second goal of this dissertation is to  5  examine another aspect of the spatiotemporal dynamics of attention —namely, to examine whether the focus of attention is unitary or divided. Considerable evidence exists to show that attention can be unitary and at the same time there is also considerable evidence showing that the focus of attention is deployed as multiple independent foci. Although there is support for both perspectives, the two modes have, in general, been viewed as mutually exclusive. In this dissertation, I test whether the two modes can be flexibly deployed based on task demands and show one set of conditions which determine whether attention is deployed as either a unitary or a divided focus. Before outlining that hypothesis in detail, however, the existing evidence for a unitary and a divided focus will be briefly reviewed below. 1.2.1 Evidence for a Unitary Focus of Attention Research testing whether focal attention can be divided between two or more locations has generally been designed to probe the following fundamental prediction: if focal attention can be divided between two separate spatial locations, intervening locations should be unattended. Items appearing at these locations should, therefore, neither benefit from attentional processing nor interfere with the processing of items at the attended locations. Posner, Snyder, and Davidson (1980) provided one of the earliest tests of whether focal attention could be divided or not, using a probabilistic methodology. In their experiments, observers were presented with four horizontally aligned potential target locations. These locations varied with respect to the probability that the target, when it appeared, would appear at that location. The target would appear at one of the locations 65% of the time (which of the four locations that would be was specified on each trial). The target appeared at a second location 25% of the time and at the remaining two locations 5% of the time each (improbable locations). The target was a dot, and simple reaction times (RTs) were recorded to its onset. Posner, 6  Snyder, and Davidson found that RTs were fastest to the 65%-location, somewhat slower to the 25%-location, and slowest to the improbable locations, indicating that attention was allocated to both the 65%- and the 25%-probable locations. This was true, however, only when the two most probable locations were adjacent to one another. When the two most probable locations were non-adjacent, RTs to the second-probable location were practically identical to RTs to the improbable location. Given this, Posner, Snyder, and Davidson concluded that focal attention could not be divided to two separate spatial locations simultaneously. There were, however, two problems with Posner, Snyder, and Davidson’s (1980) study that made their conclusions somewhat questionable. First, the unequal probabilistic split of 6525 meant that observers might not be sufficiently motivated to split attention to both of the “probable” locations. Second, in a quirk of their experimental design, although the first-probable location varied from trial to trial, the second-probable location remained constant for an entire block, which might have resulted in observers employing an unusual strategy. These issues were addressed in a study by Kiefer and Stiple (1987), who replicated Posner, Snyder, and Davidson (1980) with some modifications to their experimental technique. Their first modification was to change the target probabilities to better motivate observers to divide focal attention, providing that it is possible for them to do so. Instead of using a 65-25-55 split, Kiefer and Stiple used a 40-40-10-10 split, thereby discouraging observers from strategically allocating focal attention only to the single most likely location. The second modification was that each expected location changed on a trial-to-trial basis so that inter-trial expectancies were not a confound. In order to indicate the locations at which the target was most likely to occur, Kiefer and Stiple presented two digits from the set 1-4 displayed simultaneously both above and below the fixation cross. These two digits referred to the two  7  probable target locations, numbered 1-4 from left to right. These cues indicated whether the probable target locations would be adjacent to one another, near non-adjacent (locations 1 and 3, or 2 and 4), or far non-adjacent (1 and 4). Kiefer and Stiple (1987) found that RTs were significantly faster to both the cued locations and to any locations intervening between the two cued locations, but not to the location which fell outside of the cued locations. Interestingly, they also found that the attention effect was reduced in the far non-adjacent (locations 1 and 4) condition relative to the near nonadjacent condition (locations 1 and 3 or 2 and 4), consistent with the finding that the larger the extent of focal attention, the “thinner” it is spread (Castiello & Umiltà, 1992; Castiello & Umiltà, 1990; Barriopedro & Botella, 1998; Egeth, 1977; Eriksen & St. James, 1986). This set of results confirmed Posner, Snyder, and Davidson’s (1980) findings and supported their conclusion that focal attention is deployed as a unitary focus. Heinze, Luck, Munte, Gös, Mangun, and Hillyard (1994) used electrophysiological techniques to probe whether focal attention is deployed as a single focus or as multiple foci when spatially separated areas must be attended simultaneously. Heinze et al. asked two related questions: first, whether when attention is deployed to two separate locations the intervening locations are attended or unattended, and second, whether attention is more efficient when allocated to items that are adjacent or separated. In addressing these questions, Heinze et al. (1994) employed an event-related electroencephalographic technique. In recording event-related potentials (ERPs), the voltage fluctuations arising from cortical processing are measured at the scalp. The ERP is recorded relative to the onset of a specific event (e.g., a target onset), and averaged across multiple presentations. Heinze et al. focused on a specific component of the ERP waveform, known as the P1 (a positive deflection occurring  8  approximately 100 milliseconds post-stimulus). The P1 has been shown to index the lateralized focusing of attention to the contralateral hemifield (that is, when attention is focused to the left, for example, an enhanced P1 is measured at right-hemisphere electrodes; Heinze, Luck, Mangun, & Hillyard, 1990) and can therefore be used to assess whether the spatial location at which a probe stimulus appears is attended or not. Heinze et al. (1994) presented observers with four horizontal stimuli, arranged two on either side of fixation (and referred to as positions 1, 2, 3, and 4 from left to right). The observer’s task was to monitor two specific locations (either adjacent or separated) and to indicate with a button press if two identical stimuli appeared simultaneously at the two monitored locations. Occasionally, a small probe would appear either at an intervening location (position 2 if positions 1 and 3 were monitored; position 3 if positions 2 and 4 were monitored) or at an outside location (position 4 if positions 1 and 3 were monitored; position 1 if positions 2 and 4 were monitored). The probe was task-irrelevant and required no response from the observer – it was simply an anchor stimulus to which the ERP response could be recorded in order to determine whether the probe location was attended or not. If when instructed to attend to separated locations observers split focal attention into two separate foci, the amplitude of the P1 component should be indistinguishable when the probe stimulus appeared between the two attended locations, or outside the locations. That is if, for example, the observer is instructed to attend to positions 1 and 3, a divided focus of attention should result in the amplitude of the P1 being equal for probes at locations 2 and 4. If, on the other hand, focal attention is deployed as a single unitary focus, then when attending to positions 1 and 3, the P1 component to probe stimuli appearing as position 2 should be indistinguishable from the P1 response to targets appearing at  9  positions 1 and 3, and considerably larger in amplitude than the P1 to probe stimuli appearing at position 4. Heinze et al.’s (1994) finding precisely matched the predictions that would be made if a unitary focus of attention is deployed: when the two monitored locations were separated, a large P1 was found to probe stimuli appearing at the intervening location while a small P1 occurred to probe stimuli appearing at the outside location. Heinze et al. also found that in the condition in which observers were instructed to monitor separated locations and a probe appeared at the intermediate location, accuracy on target responses decreased, responses were slower, and the amplitude of the P1 to the target stimuli was reduced. This indicates that the probe stimuli interfered with target processing, again supporting the argument that both monitored locations and the intervening location were all encompassed within a single unitary focus of attention. Given the reduction in accuracy as well as the reduction in P1 magnitude to target stimuli in this case, Heinze et al. argued that if observers could have divided attention into two separate foci and “ignored” the intervening locations they would have done so. As such, they concluded that observers are unable to divide attention into discontinuous zones. One concern which has plagued the debate about whether focal attention is unitary or divided was neatly addressed in this study. Specifically, it has been argued that data patterns which seem to be consistent with a divided focus of attention could also stem from a single focus of attention which is shifted rapidly and strategically from one location to another (see, e.g., McCormick, Klein, & Johnston, 1998). In Heinze et al.’s (1994) study, a correct response to the target required that both target locations be monitored simultaneously, rendering the observers unable to employ a rapid-switching strategy in this experiment. Since when a rapid-switching strategy is prevented, as it was in this experiment, there is evidence that focal attention is unitary,  10  Heinze et al. suggest that research finding evidence in support of a divided focus of attention has in fact been monitoring rapid-switching patterns. In summary, there is strong evidence that focal attention is necessarily deployed in a unitary fashion and cannot be divided, even if a divided focus would be more efficient for the task to be completed. Evidence for this position comes from both behavioural and electrophysiological sources and shows that when spatially separated items must be attended simultaneously: a) intervening items appear to necessarily receive the benefits of attention and b) interfere with the processing of attended items. 1.2.2  Evidence for a Divided Focus of Attention The previous section outlined evidence showing that focal attention is exclusively unitary  in nature and cannot be divided. There is also, however, research showing that the intervening region is unattended, suggesting that focal attention can be divided between two separate spatial locations (e.g., Awh & Pashler, 2000; Adamo, Pratt & Pun, 2008; Castiello & Umiltà, 1990, 1992; Gobell et al., 2004; Hahn & Kramer, 1995, 1998; McMains & Sommers, 2004; Muller et al., 2003). Shaw and Shaw (1977) provided some of the earliest evidence that focal attention can be allocated simultaneously to discrete locations across space. In their experiments, they presented observers with a circular display consisting of eight positions at which a target letter could appear. In the critical condition, they varied the probability that the target would appear at each location – some locations were highly likely to contain the target whereas other locations were highly unlikely. In this way, Shaw and Shaw argued, they encouraged observers to deploy attention strategically to some locations and not to others. They found that target detection was  11  enhanced at expected locations and impaired at unexpected locations and they concluded that the focus of attention had been divided and allocated simultaneously to discrete locations. Shaw and Shaw’s (1977) experiment, however, is subject to the argument that data patterns which appear to be the result of a divided focus of attention could also be due to a single focus of attention being rapidly switched from one location to another. It is possible, for example, that observers deployed a single focus of attention most of the time to the most probable locations, and with decreasing probability to the other locations, and that rapid switching between locations enabled them to move focal attention when necessary. This pattern of results would yield the same data patterns as a divided focus of attention. Castiello and Umiltà (1992) took a very different approach to showing that observers can deploy two separate foci of attention that operate simultaneously and independently, and one which is not subject to the rapid-switching argument. They built on the previous finding that efficiency of processing decreases as the spatial extent of a single focus of attention increases – the larger an area one must attend to, the less any object within that area benefits from attention (Barriopedro & Botella, 1998; Castiello & Umiltà, 1992; Castiello & Umiltà, 1990; Egeth, 1977; Eriksen & St. James, 1986). Observers were presented with two boxes, one on either side of fixation and were either instructed to attend to one of the boxes (single-cue condition) or to both of the boxes (double-cue condition). The critical manipulation was that the boxes could vary in size, and could be small, intermediate, or large. The point of this manipulation is that boxes of different sizes should lead to different concentrations of attention. Given this, in the double-cue condition a unitary focus of attention should lead to reaction times which are invariant with box size (since the focus of attention is expanded to encompass both boxes) while a split focus of attention should lead to reaction times being faster to small boxes than to large boxes (since the  12  individual foci must be adjusted to small or large size and hence the attentional resources should vary for each box accordingly). Castiello and Umiltà found this latter pattern of results and concluded that focal attention was divided between the two boxes in the double-cue condition. Although Castiello and Umiltà (1992) concluded from their research that focal attention could be divided between distinct regions of space, because their experiment tested the issue only indirectly, the alternative explanation that observers deployed a single focus of attention that simultaneously encompassed both cued locations still remains viable. Castiello and Umiltà attempted to address this possibility statistically by testing whether reaction times varied overall between single-cue and double-cue conditions. Following the same logic that Castiello and Umiltà employed to test their main hypothesis, if a single focus of attention is deployed, then a larger display (i.e., the double-cue condition, because the unitary focus would encompass both boxes) should result in slower reaction times than a smaller display (i.e., the single-cue condition since only one box must be encompassed). Castiello and Umiltà conducted the appropriate analysis of variance statistical test and found small but not significant differences in reaction times between single- and double-cue conditions with small, intermediate, and large display sizes, tested individually. They concluded, therefore, that focal attention had in fact been divided and that a single focus of attention had not been expanded to encompass both cued boxes. In a commentary on Castiello and Umiltà (1992), McCormick, Johnston, and Klein (1998) argued that Castiello and Umiltà did not have enough statistical power to pick up the difference in RT between the display sizes in this test of whether RTs were faster in the singlecue condition than in the double-cue condition. In order to increase the power of the ANOVA, McCormick, Johnston, and Klein combined data from all three display sizes and found that  13  reaction times varied significantly between the single- and double-cue conditions, just as would be predicted had a unitary focus of attention been deployed. McCormick, Johnston, and Klein (1998) followed up this new, more powerful analysis with two experiments providing a more direct test of the issue. Specifically, they borrowed Castiello and Umiltà (1992) methodology, but presented probes either at the cued boxes or in between them. They found that RTs to the probes were always fast if the probes appeared in the cued boxes. RTs were also fast if the probe appeared between the cued boxes in the double-cue condition, but slow if the probe appeared between the cued boxes in the single-cue condition. This pattern of results would only be observed if a single focus of attention was deployed to encompass both cued boxes. Müller, Palinowski, Gruber, and Hillyard (2003) made use of electrophysiological recording technology to address the question of whether focal attention is unitary or divided. Specifically, they employed steady-state visual evoked potentials (SSVEPs) to determine whether attention can be divided between and maintained at two discrete spatial locations over periods of several seconds. SSVEPs are the electrophysiological response measured at the visual cortex to a rapidly repeating or flickering stimulus. In general, this response emerges as a sinusoidal waveform that has the same frequency as the flickering stimulus. When the flickering stimulus is attended, the amplitude of the waveform increases (Müller et al., 1999). Müller et al. (2003) presented observers with four discrete horizontal patches, each of which flickered at a different rate, the task simply being to monitor two of those patches for target symbols and to make a speeded response to their onset. Observers were required to pay attention to either two adjacent patches in the same hemifield (i.e., patches 1 and 2 or patches 3 and 4), or to divide their attention to two non-adjacent patches (i.e., patches 1 and 3 or patches 2  14  and 4). The latter case was the condition of interest since this condition allowed Müller et al to probe whether the intervening location (2 or 3) was attended or not. A unitary focus would result in the intervening location being attended while a divided focus would result in the intervening location being unattended. Müller et al. (2003) found that when observers were instructed to attend to positions 1 and 3, the SSVEP to the stimulus at position 2 was significantly reduced. Likewise, when observers were attending to positions 2 and 4, the SSVEP to position 3 was reduced. In fact, the amplitude of the SSVEP to the intermediate position (2 or 3) was not significantly different from the SSVEP to the unattended stimulus in the opposite hemifield when adjacent positions were attended (i.e., positions 3 and 4 when positions 1 and 2 were attended, and vice versa). The authors concluded that their results could only be explained by a split focus of attention since a unitary focus would have necessitated an enhanced SSVEP to the intermediate locations. It is also important to note that there was no cost, behavioural or electrophysiological, to dividing focal attention between two separate spatial locations. In general it has been found that the greater the spatial extent of focal attention, the less the encompassed areas appear to benefit from attention – it is as if attention is a “resource” which is spread more thinly across larger areas (Barriopedro & Botella, 1998; Castiello & Umiltà, 1992; Egeth, 1977). This does not appear to be the case when attention is divided, however, at least not with just two locations. It may be, however, that as attention is divided to a greater number of locations each location gradually benefits less from having attention at that location. Kramer and Hahn (1995) followed up on the evidence that focal attention could be deployed simultaneously as separate and independent foci to distinct spatial locations by testing some of the conditions under which this division of focal attention might take place.  15  Specifically, they noted that abrupt onset stimuli tend to capture attention in an exogenous manner that is generally beyond the observers’ control (e.g., Muller & Rabbitt, 1989; Theeuwes, 1991; Yantis, 1993). They suggested that it might be possible to divide focal attention providing that only gradual-onset stimuli appeared in the intervening locations but that if abrupt-onset stimuli appeared in the intervening locations they would be attended, and focal attention would not be divided. In order to test this hypothesis, Kramer and Hahn (1995) employed two critical conditions: onset and non-onset. In the onset condition, a fixation cross appeared, followed by two precue boxes which appeared for 150 ms at two locations around an imaginary circle. Four items then appeared, two targets and two distractors. The targets were letters which appeared in the precue boxes, and the observers’ task was to indicate whether the two target letters matched or mismatched. The two distractors (which could be either the same as or different from the target letters) appeared at locations along the imaginary circle which intervened between the locations of the two targets. Both targets and distractors were present for 60 ms before being replaced by a pattern mask. The non-onset condition was identical to the onset condition except for one critical difference. In this condition, figure eight pre-masks were presented along with the pre-cue boxes. The two targets and distractors were subsequently displayed by removing the required segments, allowing the targets and distractors to be seen without any bottom-up onset signals to capture attention exogenously. Kramer and Hahn found that in the no-onset condition, the intervening distractors did not influence RTs to the targets, regardless of whether or not they matched the targets, suggesting that they were completely unattended and that focal attention had been successfully and completely divided between the two target locations. The onset condition, however, revealed a  16  completely different pattern of results. In this case, reaction times were significantly slowed if the distractor information (same/different) was incongruent with the target information (match/mismatch) and significantly faster if the distractor information was congruent with the target information. The effect of the distractors in the onset but not in the non-onset conditions clearly indicates that the distractors could be ignored and focal attention divided, but only when there was no exogenous information to disrupt this attentional setting. Adamo, Pun, Pratt, and Ferber (2008) took a novel approach to testing whether focal attention is unitary and divided, and went yet one step further by asking whether two separate attentional control sets can be applied and maintained to distinct, separated regions of space. The idea of an attentional control set stems from the literature on attentional capture. It has been well-established that if an irrelevant item which shares features with a task-relevant event appears suddenly, that irrelevant item will capture attention and cause it to be shifted to that location. More concretely, imagine that you are searching for a target, which is a red triangle. The abrupt onset of a small red dot will likely capture your attention and draw it to that location simply because it falls into the “red” category. This form of attentional capture has been termed contingent capture or contingent involuntary orienting (e.g., Folk, Remington, & Johnston, 1992) and depends on the specific attentional control set in operation at any given moment. Adamo et al. (2008) required observers to use two different attentional control strategies for two different locations. They were instructed, for example, to respond to a target on the left only if it were green and a target on the right only if it were blue (or vice versa). Pre-cues then appeared either to the left or to the right that were either congruent or incongruent with the observer’s mental set. Adamo et al. found contingent capture of attention only for those cues which matched the attentional control set applied to that particular region of space. If observers were  17  instructed to respond to green targets on the left and blue targets on the right, RTs were faster if a green pre-cue appeared on the left and slower if a green pre-cue appeared on the right. They were comparably faster if a blue pre-cue appeared on the right and slower if it appeared on the left. Adamo et al. concluded that not only can focal attention can be maintained at two separate spatial locations, but given their findings separate attentional sets can also be maintained at each attended location. In summary, there is clear, strong evidence that focal attention can be divided and deployed to more than one spatial location simultaneously. Shaw and Shaw (1977) provided behavioural evidence, Muller et al. (2003) used steady-state evoked potentials to show that focal attention could be divided, and there is even evidence that when focal attention is divided, different attentional sets can be maintained simultaneously at the two locations (Adamo et al., 2008). Some limits of divided focal attention, however, were shown by Castiello and Umiltà (1992) who demonstrated that external support in the form of physical stimuli were needed before focal attention could be divided. Hahn and Kramer (1995) further showed that focal attention could only be divided in the absence of abrupt-onset stimuli. 1.3 Dissertation Objectives From the research outlined above, it is clear that the focus of attention can be shrunk and expanded so as to optimize performance on the task at hand. Although substantial research has examined the rate at which attention can be shifted from one location to another and the factors that initiate such attentional shifts and influence the rate at which they can be completed, there has been only a single study that investigated the rate at which focal attention expands and no studies at all on the rate at which it contracts. More generally, there has been a dearth of research on the factors that influence the expanding and contracting process. Chapter 2 of this  18  dissertation develops a behavioural procedure to assess the rate at which the focus of attention is contracted. A qualitative model of that process is put forward and tested. Chapter 3 examines factors that modulate the temporal course of attentional narrowing in young healthy adults who presumably can exercise efficient control of attentional processes. In contrast, Chapter 4 examines the effect of reduced attentional control by examining the same process in the natural experiment in which we all participate during the course of our lives as we age. The second goal was to examine the spatial deployment of focal attention – whether it is unitary or divided. There is clear evidence in the literature showing both that attention can be unitary and that it can be divided. These two perspectives are generally viewed as mutually exclusive. An alternative hypothesis, tested in Chapter 5 of this thesis, is that focal attention can be deployed as either a single, unitary focus or divided into multiple foci, depending on the observers mental set and on the task demands. The final chapter (Chapter 6) then combines and compares the findings across all experiments and evaluates how they fit in with current theories of visual attention.  19  References Awh, E., & Pashler, H. (2000). Evidence for split attentional focus. Journal of Experimental Psychology, 26, 834-846. Adamo, M., Pun, C., Pratt, J., & Ferber, S. (2008). Your divided attention please! The maintenance of multiple attentional control sets over distinct regions in space. Cognition, 107, 295-303. Barriopedro, M.I., & Botella, J. (1998). New evidence for the zoom-lens model using the RSVP technique. Perception & Psychophysics, 60, 1406-1414. Benso, F., Turatto, M., Mascetti, G.G., & Umiltà, C. (1998). The time course of attentional f ocusing. European Journal of Cognitive Psychology, 10, 373-388. Castiello, U., & Umiltà, C. (1992). Splitting focal attention. Journal of Experimental Psychology: Human Perception and performance, 18, 837-848. Castiello, U. & Umiltà, C. (1990). Size of the attentional focus and efficiency of processing. Acta Psychologica, 73, 195-209. Eriksen, C.W., & Hoffman, J.E. (1974). Selective attention: Noise suppression or signal enhancement? Bulletin of the Psychonomic Society, 4, 587-589. Eriksen, C.W., Pan, K., & Botella, J. (1993). Attentional distribution in visual space. Psychological Research, 56, 5-13. Eriksen, C.W. & Rohrbaugh, J.W. (1970). Some factors determining efficiency of selective attention. The American Journal of Psychology, 83, 330-342. Eriksen, C.W., & St. James, J.D. (1986). Visual attention within and around the field of focal attention: A zoom lens model. Perception & Psychophysics, 42, 225-240.  20  Eriksen, C.W., & Yeh, Y-Y. (1985). Allocation of attention in the visual field. Journal of Experimental Psychology: Human Perception and Performance, 11, 583-597. Gobell, J.L., Tseng, C-H., & Sperling, G., (2004). The spatial distribution of visual attention. Vision Research, 44, 1273-1296. Hahn, A.F., & Kramer, S. (1995). Splitting the beam: Distribution of attention over noncontinguous regions of the visual field. Psychological Science, 6, 381-386. Heinze, H-J., Luck, S.J., Munte, T.F., Gös, A., Mangun, G., & Hillyard, S. (1994). Attention to adjacent and separate positions in space: An electrophysiological analysis. Perception and Psychophysics, 56, 42-52. Hernandez-Peon, R. (1964). Psychiatric implications of neurophysiological research. Bulletin of the Meninger Clinic, 28, 165-185. James, W. (1950). The Principles of Psychology, Vol. 1. New York: Dover. (Original work published in 1890). Jonides, J. (1980). Toward a model of the mind’s eye. Canadian Journal of Psychology, 34, 103-112. Jonides, J. (1983). Further toward a model of the mind’s eye’s movement. Bulletin of the Psychonomic Society, 21, 247-250. Kiefer, R.J., & Stiple, P. (1987). Spatial constraints on the voluntary control of attention across visual space. Canadian Journal of Psychology, 41, 474-489. Kramer, A.F., & Hahn, S. (1995). Splitting the beam: Distribution of attention over noncontiguous regions of the visual field. Psychological Science, 6, 381-386. LaBerge, D. (1983). Spatial extent of attention to letters and words. Journal of Experimental Psychology: Human Perception and Performance, 9, 371-379.  21  LaBerge, D. (1995). Attentional Processing: The Brain’s Art of Mindfulness. Cambridge, MA: Harvard University Press. LaBerge, D., Brown, V., Carter, M., & Bash, D. (1991). Reducing the effects of adjacent distractors by narrowing attention. Journal of Experimental Psychology: Human Perception and Performance, 17, 65-76. McCormick, P.A., Klein, R.M., & Johnston, S. (1998). Splitting versus sharing focal attention: A commentary on Castiello and Umiltà (1992). Journal of Experimental Psychology: Human Perception and Performance, 24, 350-157. McMains, S.A., & Sommers, D.C. (2004). Multiple spotlights of attentional selection in human visual cortex.  euron, 42, 677-686.  Müller, M.M., Malinowki, P., Gruber, T., & Hillyard, S.A. (2003). Sustained division of the attentional spotlight.  ature, 424, 309-312.  Müller, H., & Rabbit, P. (1989). Reflexive and voluntary orienting of attention: Time course of activation and resistance to disruption. Journal of Experimental Psychology: Human Perception and Performance, 15, 315-330. Murphy, T.D., & Eriksen, C.W., (1987). Temporal changes in the distribution of attention in the visual field in response to precues. Perception & Psychophysics, 42, 576-586. Posner, M.I., Snyder, C.R.R., & Davidson, B.J. (1980). Attention and the detection of signals. Journal of Experimental Psychology: General, 11, 160-174. Shaw, M.L., & Shaw, P. (1977). Optimal allocation of cognitive processes to spatial locations. Journal of Experimental Psychology: Human Perception and Performance, 3, 201-211. Theeuwes, H. (1991). Exogenous and endogenous control of attention: The effects of visual onsets and offsets. Perception and Psychophysics, 49, 83-90.  22  Yantis, S. (1993). Stimulus-driven attentional capture. Current Directions in Psychological Science, 2, 156-161.  23  Chapter 2: Linear Changes in the Spatial Extent of the Focus of Attention Across Time1 Stimuli presented at attended locations are processed faster and more accurately than those presented at unattended locations (Helmholtz, 1866/1962; James, 1890/1950; LaBerge, 1995). A common metaphor to describe this finding is that attention functions like a spotlight: items that fall within the spotlight are processed faster and more accurately than items that fall outside. A characteristic that attentional processes share with a spotlight is that both can be directed at specific locations and can be moved rapidly to new locations. Pursuing this metaphor, Eriksen and colleagues (Eriksen & Yeh, 1985; Eriksen & St. James, 1986) proposed a model in which, in order to optimize performance, the focus of attention can be resized just like a spotlight equipped with a zoom lens. How quickly attention can be moved from one object or location to another has been studied extensively with both behavioural (e.g., Weichselgartner & Sperling, 1987) and electrophysiological (e.g., Müller, Teder-Salejarvi, & Hillyard, 1998) measures. In contrast, studies that investigated changes in the size of the focus of attention have been concerned mainly with the hypothesized inverse relationship between performance and the size of the attended area, with attention becoming more diffuse as the size of the attended area is increased (Barriopedro & Botella, 1998; Egeth, 1977; Eriksen & St. James, 1986; Eriksen & Yeh, 1985; Jonides, 1983; LaBerge, 1983, 1995). Notably, there has been a dearth of studies of the rate at which the focus of attention expands or contracts. An estimate of the time required to expand focal attention has been provided by Benso, Turatto, Mascetti, and Umiltà (1998), who used a pre-cue to draw attention from the center of the 1  A version of this chapter has been accepted for publication. Jefferies, L.N. & Di Lollo, V. Linear changes in the spatial extent of the focus of attention across time. Journal of Experimental Psychology: Human Perception and Performance. 24  screen to a randomly-chosen location, at which a cue was presented. The cue consisted of a ring with a diameter of either 2.5˚ or 7.5˚. After a variable delay, a target was presented within the cued area. The critical assumption was that, upon presentation of the cue, the attentional focus expanded to cover the cued area. The results suggested that the process of expansion was completed within about 33 to 66 ms. One limitation of Benso et al.’s (1998) study was that it was confined to the case in which the attentional spotlight was expanded. For both practical and conceptual reasons, it is equally important to obtain estimates of the rate at which the focus of attention contracts, and that was the main objective of the present work. A phenomenon that can be used for this purpose is the attentional blink (AB). When two targets are presented in rapid succession, correct identification of the second target is impaired. This second-target deficit is most pronounced when the temporal lag between the two targets is in the range of 100 – 500 ms (Raymond, Shapiro, & Arnell, 1992). The AB has been investigated with a paradigm known as rapid serial visual presentation (RSVP), in which two targets (e.g., letters) are inserted in a stream of distractors (e.g., digits). Typically, all the items are displayed sequentially in the same location at a rate of one every 100 ms or so. The present work utilized an aspect of the AB known as Lag-1 sparing (Potter, Chun, Banks, & Muckenhoupt, 1998), in which the AB is much reduced when the second target is presented directly after the first (i.e., at Lag 1), without any intervening distractors, provided that the inter-target stimulus-onset asynchrony (SOA) is about 100 ms or longer (see Bowman & Wyble, 2007; Chun & Potter, 1995; Giesbrecht & Di Lollo, 1998; Maki, Couture, Frigen, & Lien, 1997; Raymond et al., 1992). Visser, Bischof, and Di Lollo (1999) defined Lag-1 sparing as a positive difference between second-target identification accuracy at Lag 1 minus the lowest  25  level of second-target identification accuracy at any other lag. In the present work, the magnitude of Lag-1 sparing was estimated as a positive difference between performance at Lags 1 and 3 on the grounds that in many studies the lowest level of second-target accuracy was reached at Lag 3 or later (e.g., Arnell & Jolicœur, 1999; Chun & Potter, 1995; Duncan, Ward, & Shapiro, 1994; Giesbrecht & Di Lollo, 1998; Raymond, Shapiro, & Arnell, 1992; Seiffert & Di Lollo, 1997; Shapiro, Caldwell, & Sorensen, 1997; Shapiro, Raymond, & Arnell, 1994; Shih, 2000; Spalek, Falcon, & Di Lollo, 2006). A negative difference in second-target accuracy between Lags 1 and 3 was termed Lag-1 deficit.  Lag-1 Sparing Across Space In the present research, we employed Lag-1 sparing and Lag-1 deficit as tools to investigate the temporal dynamics of changes in the spatial extent of attention over very brief intervals of time. In a survey of the literature, Visser et al. (1999) found that Lag-1 sparing never occurs when the two targets are displayed in different spatial locations. Recent examples of this rule have been provided by studies that employed two or more concurrent RSVP streams such that the two targets appear either in the same stream or in different streams (e.g., Dell’Acqua, Pascali, Jolicœur, & Sessa, 2003; Holländer, Corballis, & Hamm, 2005; Juola, Botella, & Palacios, 2004; Kristjansson & Nakayama, 2002; Peterson & Juola, 2000). In these studies, Lag1 sparing was never found when the two targets were presented in different streams. Recently, however, Jefferies, Ghorashi, Kawahara, and Di Lollo (2007) found substantial Lag-1 sparing when the two targets were presented in different streams, but only when the second target fell within the focus of attention. In that study, the size of the attentional focus was manipulated in two main conditions. In one, observers knew which of two streams would contain the first target, causing attention to be focused narrowly on that stream. In the other 26  condition, the first target was presented unpredictably in either stream, leading to a broad focus of attention that encompassed both streams. Of special interest to the present work was the condition in which the two targets appeared in opposite streams. In this case, if attention was narrowly focused on the stream containing the first target, the second target fell outside the attended area and Lag-1 sparing did not occur. If, however, the focus of attention encompassed both streams, the second target fell within the attended area, and Lag-1 sparing occurred. From this perspective, the incidence of Lag-1 sparing can provide a means for achieving the main objective of the present work. Namely, Lag-1 sparing to targets presented in opposite streams can be used to index the location and extent of the focus of attention. The presence of Lag-1 sparing would indicate that the focus of attention was set broadly to encompass both streams; the absence of Lag-1 sparing would indicate that attention was focused narrowly on one stream to the exclusion of the other. Thus, whereas Jefferies et al. (2007) used the extent of the attentional focus to account for the incidence of Lag-1 sparing, in the present work Lag-1 sparing was used to index the extent of the focus of attention. Control of Attention in the Spatiotemporal Domain As noted above, the major objective of the present work was to provide an estimate of the rate at which the focus of attention changes in size. In practice, we used the magnitude of Lag-1 sparing to index the presence of attention at a given spatial location. To monitor changes in the size of the focus of attention over time, we varied the SOA between successive items in the RSVP streams. The importance of SOA becomes evident when one considers the changes in attentional focus that are likely to occur in the course of performing the experimental task.  27  Those changes can be described in terms of an attentional focus the extent of which changes dynamically over time so as to optimize performance on the task at hand. We assume that at the outset of any given trial, the observer is set to optimize performance on the first task, namely the identification of the first target. Support for this assumption comes from the findings of Shih (2000) and Jefferies et al. (2007) who employed a dual RSVP stream paradigm. Jefferies et al. hypothesized that when the observers did not know which stream contained the first target, the focus of attention was set broadly so as to encompass both streams. In contrast, when the location of the first target was known in advance, the focus of attention was hypothesized to be set narrowly on the relevant stream so as to maximize the probability of detecting the first target. This reasoning was supported by the finding that identification accuracy for the first target was significantly higher when the location of the first target was known in advance. From a theoretical standpoint, this finding is consistent with the hypothesis that attention was focused narrowly on the first target's location, resulting in a higher concentration of attentional resources than if attention had been distributed broadly (Barriopedro & Botella, 1998; Castiello & Umiltà, 1990; Egeth, 1977) In the present work we employed two concurrent RSVP streams wherein the first target could appear unpredictably in either RSVP stream. In such a display, the optimal initial strategy would be to set a spatially broad focus of attention so as to encompass both streams. When the first target appeared, the attentional focus would begin to narrow to the stream containing the first target so as to optimize target identification. If the second target then appears in the same stream, it will fall within the focus of attention and its identification will be enhanced, resulting in Lag-1 sparing. If, however, the second target is presented in the opposite stream, it will not be  28  encompassed within the attentional focus, and its identification will hinge on the length of the SOA, as follows. If the SOA is short, there may not have been sufficient time for the focus to fully narrow on the stream containing the first target before the onset of the second target. In this case, both streams may still be encompassed within the focus, allowing the second target to be processed along with the first even when it appears in the opposite stream, resulting in Lag-1 sparing. If, on the other hand, the SOA is long, there is a greater probability that there has been sufficient time for the focus to narrow on the stream containing the first target. In this case, if the second target appears in the opposite stream, it will fall outside focal attention, and Lag-1 sparing will not occur. To examine this hypothesis, we systematically varied the SOA between successive items in the RSVP stream. Six groups of observers were tested, each at a different SOA (53, 66, 80, 100, 118, and 133 ms), and each across three inter-target lags (1, 3, and 9). The changes in the width of the attentional focus expected on the basis of the above hypothesis are illustrated in Figure 2.1. Each segmented box represents the width of the attentional focus at any given combination of SOA and Lag. The extent to which the second target (T2) is encompassed within the segmented box is proportional to the probability that the second target will fall within the focus of attention, thus yielding Lag-1 sparing. Take as an example the top and bottom rows of Figure 2.1, namely, SOAs of 53 and 133 ms, respectively.  Figure 2.1. Schematic illustration of the progressive changes in the spatial extent of the focus of attention (segmented rectangles) as a joint function of SOA and Lag. See text for a detailed description.  29  Consider first an SOA of 53 ms. At Lag 1 (i.e., 53 ms after the onset of the first target, T1), the focus of attention has begun to narrow on the location of the first target, but it still encompasses a substantial portion of the stream containing the second target. This will result in the second target falling within focal attention and being processed along with the first target, thus yielding Lag-1 sparing. By Lag 3 (i.e., 159 ms after the offset of the first target) the focus has further narrowed on the first target’s location, thereby reducing the extent to which the second target is attended. As a result, the accuracy of second-target identification is reduced and an AB deficit ensues. Finally, by Lag 9, 477 ms after the onset of the first target, sufficient time has elapsed for the attentional focus to have expanded so as to again encompass both target locations. 30  At an SOA of 133 ms (Figure 2.1, lowest row) sufficient time has elapsed from the T1 onset for the focus of attention to narrow almost completely on the first target, even at Lag 1, leaving the second target outside the attended area. As a result, the accuracy of second-target identification is reduced, and Lag-1deficit ensues instead of Lag-1 sparing. By Lag 3, 399 ms after the onset of the first target, there has been sufficient time for the focus to widen again so as to encompass most of the second target. Needless to say, by Lag 9 the focus once again fully encompasses both target locations. With the appropriate changes, the contraction and expansion of the focus of attention follow a similar sequence at the intermediate SOAs.  Experiment 2.1 Experiment 2.1 was designed to test the predictions outlined above and illustrated in Figure 2.1. Method Observers A total of 117 undergraduate students at both the University of British Columbia and Simon Fraser University participated for course credit. All participants were naïve as to the purpose of the experiment and reported normal or corrected-to-normal vision. The observers were allocated randomly to one of six groups (SOA: 53, 66, 80, 100, 118, and 133 ms). Each group was required to have a minimum of 17 observers, but because the data were collected concurrently at the two universities, the final number of observers in each group was 17, 22, 18, 19, 17, and 24, for SOAs of 53, 66, 80, 100, 118, and 133 ms, respectively. Apparatus and Stimuli All stimuli were presented on a computer monitor viewed from a distance of approximately 57 cm. A white fixation cross (0.25˚ by 0.25˚) was displayed in the center of the 31  screen throughout each trial. All other stimuli were white digits (0 - 9) or capital letters (excluding the letters I, O, Q, and Z), each of which subtended approximately 0.9˚ vertically. The luminance of all stimuli was 129 cd/m2, and the luminance of the black background was 2.3 cd/m2. To obtain the six SOAs of 53, 66, 80, 100, 118, and 133 ms, the screen refresh rate was set at 75, 75, 75, 60, 85, and 75 Hz, respectively. Procedure The observers initiated each trial by pressing the spacebar. The trial began with the onset of two synchronized RSVP streams, one centered 1.75˚ to the left and the other 1.75˚ to the right of fixation. Each stream contained an equal number of digit distractors and 0, 1, or 2 letter targets. Both streams contained 8 - 14 leading distractors selected from the digits 0 to 9, with the restriction that each digit differed from the previous two digits and from the digit in the opposite stream. The number of distractors presented before the first target varied randomly between trials, but it was always identical for both streams. A total of two target letters were presented on any given trial. The two letters were never the same on any given trial and appeared with equal probability in either the left or the right stream and in either the same or opposite stream. Each stream terminated with a single digit distractor, which acted as a backward mask for the second target. The second target was presented at one of three inter-target lags: Lags 1, 3, and 9. At Lag 1, the second target was presented directly after the first; at Lag 3, two distractors intervened between the targets; at Lag 9, there were eight intervening distractors. Items continued to be displayed in both RSVP streams until the second target and its mask were presented. Inter-target lags occurred in random order and with equal frequency across trials. The actual amount of time (ms) that elapsed from the onset of the first target to the onset of the second target at each lag 32  depended on the SOA group as follows. The number of milliseconds at Lags 1, 3, and 9 was 53, 106, and 424 for the 53-ms SOA group; 66, 132, and 528 for the 66-ms SOA group; 80, 160, and 640 for the 80-ms group; 100, 200, and 800 for the 100-ms SOA group; 118, 236, and 944 for the 118-ms SOA group; and 133, 266, and 1064 for the 133-ms SOA group. The SOA between successive items in the RSVP stream consisted of approximately two-thirds exposure duration of the stimulus and one-third blank inter-stimulus interval (ISI). The actual proportions were constrained by the refresh rate of the monitor. The ratios of exposure duration to ISI were approximately 26.5:26.5, 40:26, 45:35, 70:30, 71:47, and 80:53 ms for SOAs of 53, 66, 80, 100, 118, and 133 ms, respectively.  Figure 2.2. Schematic representation of the sequence of events within a trial in Experiment 2.1. The first and the second targets (T1 and T2) could appear in either the left or the right RSVP stream and in either the same or opposite streams. Illustrated in the figure is the case in which the second target was presented three frames after the first target (i.e., at Lag 3). At Lag 1, the second target was presented directly after the first, without intervening distractors  33  T1 & T2 in same stream  8  +  3  8  3 H  +  H + 5 5 + 7 + 1 7 + 1 3 + 9 3 + 9 R + 4 R + 4 7 + 2 7 + 2 6 + 3 6 + 3  (T2)  (T1)  T1 & T2 in different streams  The display sequence on any given trial is illustrated schematically in Figure 2.2. The observers’ task was to identify the two target letters presented in each trial. Observers were required to press the appropriate keys on the keyboard in either order at the end of each trial. Results and Discussion Second-Target Accuracy Only those trials in which the first target was identified correctly were included for analysis. This procedure is commonly adopted in AB experiments on the grounds that, on trials in which the first target is identified incorrectly, the source of the error is unknown, and thus its effect on second-target processing cannot be estimated.  34  Figure 2.3. Mean percentages of correct identifications of the second target in Experiment 2.1. Panel (A): Both targets were presented in the same stream, the data have been collapsed across left and right streams. Panel (B): Both targets were presented in opposite streams. Each function represents performance at a different lag across SOAs. The open symbols represent data from Experiment 2.2, in which the spatial separation between the RSVP streams was reduced.  Figures 2.3A and 2.3B illustrate the percentage of correct second-target responses as a function of SOA and Lag, separately for the Same-stream and Different-stream conditions. The data were analyzed in a 2 x 3 x 6 ANOVA consisting of two within-subject factors and one between-subjects factor. The within-subject factors were Lag (1, 3, and 9), and Stream (Same and Different). The between-subjects factor was SOA (53, 66, 80, 100, 118, and 133 ms). The analysis revealed significant effects of Stream, F(1,111) = 84.93, p < .001, ηp2 = .431, Lag 35  F(2,222) = 78.21, p < .001, ηp2 = .418, and SOA, F(5,111) = 24.65, p < .001, ηp2 = .521. There were two significant two-way interaction effects: Lag x SOA, F(10,222) = 5.78, p < .001, ηp2 = .199, and Stream x Lag, F(2,222) = 63.15, p < .001, ηp2 = .362. The three-way interaction amongst Stream, Lag, and SOA was also significant, F(10,222) = 4.96, p < .001, ηp2 = .182. The interaction effect between Stream and SOA was not significant, F(5,111) = .93, p = .46, ηp2 = .041. As expected on the basis of earlier research (Jefferies et al., 2007; Shih, 2000; Visser, et al., 1999), a comparison between the Same-stream and Different-stream conditions revealed Lag1 sparing across all SOAs in the Same-stream condition. This is shown by the greater accuracy of second-target identification at Lag 1 than at Lag 3 in Figure 2.3A. In contrast, in the Different-stream condition, the incidence and magnitude of Lag-1 sparing depended critically on SOA, as illustrated in Figure 2.3B and discussed below. Lag-1 Sparing: Comparing Predicted and Obtained Patterns. Critical to the major objective of the present work was an examination of Lag-1 sparing as a function of SOA in the condition in which the two targets were displayed in different streams. From the model outlined in Figure 2.1, Lag-1 sparing should occur when the SOA is short, but not when it is long. This is because at brief SOAs the second target is hypothesized to be still encompassed within the focus of attention. This prediction, illustrated in the first two columns of Figure 2.1 (Lags 1 and 3), is confirmed by the corresponding results for the differentstream condition in Figure 2.3B. The relationship between the relevant portions of Figures 2.1 and 2.3B is illustrated in Figure 2.4. The vertical bars in Figure 2.4 correspond to the segmented-line rectangles for Lags 1 and 3 in Figure 2.1, and represent the extent of the focus of attention across SOAs. The empirical functions in Figure 2.4 were transposed from Figure 2.3B. 36  Figure 2.4. Comparison between the estimated spatial extent of the attentional focus at Lags 1 and 3 across SOAs (from Figure 2.1) and the corresponding empirical results from Figure 2.3B. The height of each vertical bar in the lower portion of the figure exactly matches the width of the corresponding segmented-line rectangle in  80  Lag 1 3  60  40  20  53  66  80  100  SOA (ms)  118  133  Inferred width of the focus of attention (arbitrary units)  (From Fig. 2.3B) (From Fig. 2.1)  masking  Percentage correct (T2|T1)  Figure 2.1, and represents the extent of the focus of attention across SOAs.  There is a close match between the expected and obtained results in Figure 2.4. In both the expected and the obtained results, the functions for Lags 1 and 3 exhibit a cross-over as SOA is increased. In interpreting this cross-over, it should be emphasized that the height of each bar indexes the probability that the second target will fall within the focus of attention and, therefore, maps directly to the accuracy of second-target identification. The cross-over in the empirical 37  data in Figure 2.4 is consistent with the outcome of the overall statistical analysis which revealed a significant three-way interaction amongst Stream, SOA, and Lag. To confirm this interpretation of the three-way interaction, a subsidiary ANOVA was performed on the empirical data illustrated in Figure 2.4. The analysis was a 2 x 6 ANOVA consisted of one within-subject factor, Lag (1 and 3), and one between-subjects factor, SOA (53, 66, 80, 100, 118, and 133 ms). The analysis revealed a significant effect of SOA, F(5,111) = 12.41, p < .001, ηp2 = .449. Interpretation of this effect, however, is constrained by the significant interaction between Lag and SOA, F(5,111) =7.07, p < .001, ηp2 = .217, confirming the graphical evidence in Figure 2.4 that the accuracy of second-target identification was higher at Lag 1 than at Lag 3 when the SOA was short, but that the reverse was true when the SOA was long. Individual t-tests between performance at Lags 1 and 3, separately for each SOA, were as follows: SOA 53 ms: t(16) = 2.53, p = .02; SOA 66 ms: t(21) = 2.13, p < .05; SOA 80 ms: t(17) = 2.29, p < .04; SOA 100 ms: t(18) = 0.13, p = .90; SOA 118 ms: t(16) = 2.11, p = .05; SOA 133 ms: t(23) = 4.29, p < .001. In essence, the Lag-1 sparing observed at short SOAs turned into a Lag-1 deficit at long SOAs. The Effect of Masking. Figure 2.5 provides a more conventional representation of Lag-1 sparing, and serves to illustrate an incidental yet important factor at work in the present experiment: masking. In Figure 2.5, each line represents second-target performance at a different SOA. Within each line, the first symbol represents performance at Lag 1, and the second symbol represents performance at Lag 3. A negative slope indicates Lag-1 sparing; a positive slope indicates Lag-1 deficit. As may be expected on the basis of Figure 2.1, negative slopes in Figure 2.5 are associated with short SOAs whereas positive slopes are associated with long SOAs.  38  Figure 2.5. Progressive transition from Lag-1 sparing (negative slope) to Lag-1 deficit (positive slope) across SOAs. In each function, the first symbol represents second-target performance at Lag 1, and the second symbol represents second-target performance at Lag 3. The data were redrawn from the data for Lags 1 and 3 in Figure 2.3B  The effect of masking is illustrated by a progressive increment in the mean level of each line in Figure 2.5. Masking becomes progressively stronger – and performance correspondingly less accurate – as the SOA is reduced from 133 to 53 ms. The effect of masking is also seen in an apparent discrepancy between the empirical results illustrated in Figure 2.5 and the predicted patterns in Figure 2.1. Consider two data points in Figure 2.5: Lag 3 at an SOA of 53 ms and Lag 1 at an SOA of 133 ms. Performance is considerably less accurate in the former than in the latter. Reference to the corresponding points in Figure 2.1, however, shows that the extent of the attentional focus is predicted to be exactly the same in the two cases. Were the extent of the attentional focus the only determining factor, accuracy of second-target identification should be 39  the same in both cases. The fact that performance is lower at an SOA of 53 ms, however, is to be expected, based on the well-established principle that the strength of masking is inversely related to the period of time that elapses from the onset of a target to the onset of the trailing mask (Breitmeyer, 1984). That period of time covaried with SOA, being 53 ms for the 53-ms SOA group, and 133 ms for the 133-ms SOA group. It should be noted that the predictions in Figures 2.1 and 2.4 (vertical bars) illustrate the dynamic variation in the extent of the focus of attention independent of masking. The effect of masking is seen in the empirical results in Figure 2.4 in the progressive reduction in overall level of performance as the SOA is decreased from 133 to 53 ms. First-Target Accuracy. Accuracy of first-target identification as a function of SOA in the Same-stream and Different-stream conditions is illustrated in Figure 2.6, separately for each lag. The data were analyzed in a 2 x 3 x 6 ANOVA consisting of two within-subject factors and one betweensubjects factor. The within-subject factors were Lag (1, 3, and 9), and Stream (Same and Different). The between-subjects factor was SOA (53, 66, 80, 100, 118, and 133 ms). The analysis revealed significant effects of Stream, F(1,111) = 23.03, p < .001. ηp2 = .191, Lag F(2,222) = 52.15, p < .001, ηp2 = .337, and SOA, F(5,111) = 48.76, p < .001, ηp2 = .636. There was one significant interaction effect: Stream x Lag, F(2,222) = 30.30, p < .001, ηp2 = .285. No other effects were significant.  40  Figure 2.6. Mean percentages of correct identifications of the first target in Experiment 2.1 as a joint function of Stream (Same or Different), Lag, and SOA.  As was the case for second-target accuracy, the progressive increment in first-target accuracy over SOA seen in Figure 2.7 can be attributed to masking. All functions in Figure 2.7 are parallel – as confirmed by the absence of any statistically significant interactions involving SOA – and overlap substantially with one another, except for the Same-stream condition at Lag 1. This means that the strength of masking at any given SOA was the same across conditions and lags with the single exception of the Same-stream condition in which masking of the first target at Lag 1 was more pronounced across all SOAs. The relatively lower performance in the Same-stream condition was not unexpected and can be explained on the well-established finding 41  that the strength of masking increases as a function of the structural similarity (Fehrer, 1966; Harmon & Julesz, 1973) and/or conceptual similarity (Dux & Coltheart, 2005; Enns, 2004; Intraub, 1981, 1984) between the target and the mask. In the present experiment, the first-target letter was always masked by a digit (relatively low categorical similarity) except in the Samestream condition at Lag 1 in which it was masked by the second target (another letter; relatively high categorical similarity). As a consequence, masking was relatively stronger in that condition. Temporal Dynamics of the Focus of Attention Temporal changes in the spatial extent of the focus of attention can be estimated from the slopes of the functions in Figure 2.5 combined with the model illustrated in Figure 2.1. A negative slope (i.e., Lag-1 sparing) corresponds to a broad attentional focus whereas a positive slope (i.e., a Lag-1 deficit) corresponds to a focus of attention that is set narrowly on the location of the first target, thereby excluding the second target. As illustrated in Figure 2.5, there is a progressive increment in the slope of the functions as SOA is increased. This increment in slope across SOAs reflects the progressive transition from Lag-1 sparing to Lag-1 deficit, and indexes the corresponding changes in the extent of the focus of attention. This relationship provides a basis for estimating the time-course of the changes in the spatial extent of the focus of attention. If the data in Figure 2.5 are to be used to estimate the time-course of the changes in the extent of the focus of attention, it is first necessary to partial out the effects of masking, which caused the mean level of the functions to vary with the SOA. This was accomplished by expressing the slope of each function in Figure 2.5 as the ratio of performance at Lag 3 to performance at Lag 1. By this method, the mean level of performance at each SOA, and therefore the effect of masking, is removed as a determining factor. This makes it possible to 42  express the magnitude of Lag-1 sparing (or the lack thereof, i.e., a Lag-1 deficit) as a single value, independent of masking. To express the magnitude of Lag-1 sparing as positive values and the magnitude of Lag-1 deficit as negative values, we applied Equation (1) to the data in Figure 2.5 Lag-1 sparing value = 100-[(Lag3/Lag1)*100]  (1)  where positive values indicate Lag-1 sparing and negative values indicate a Lag-1 deficit. The Lag-1 sparing values obtained from Equation (1) are illustrated in Figure 2.7. A linear fit through the points in Figure 2.7 by the method of least squares yielded the linear function: y = 60.6 - .62x  (2)  Figure 2.7. Variation in the magnitude of Lag-1 sparing (positive values) and Lag-1 deficit (negative values) as a function of SOA. The effect of masking was partialled out by expressing the magnitude of Lag-1 sparing/Lag-1 deficit as the ratio of second-target accuracy at Lag 3 to that at Lag 1. See text for a detailed description. The closed symbols represent the data from Experiment 2.1; the open symbols represent the data from Experiment 2.2.  43  Figure 2.7 reveals an approximately linear transition from Lag-1 sparing to Lag-1 deficit as the SOA is increased from 53 to 133 ms, mirroring the corresponding changes in the extent of the focus of attention illustrated in Figure 2.1. Equation (2) indicates that, under the conditions of the present study, Lag-1 sparing decreased (or Lag-1 deficit increased) by approximately 2.5% for every millisecond increment in SOA. Experiment 2.2 The main message conveyed by the function in Figure 2.7 is that the spatial extent of the focus of attention, as indexed by the magnitude of Lag-1 sparing, varies linearly with SOA. We interpret this linear relationship as reflecting the time-course of the contraction and expansion of the focus of attention. When the SOA is short there is not sufficient time for the focus of 44  attention to shrink to the location of the stream containing the first target. This causes the second target to remain within the focus of attention with consequent Lag-1 sparing. In contrast, when the SOA is long there is abundant time for the focus to shrink, leaving the second target unattended, resulting in a Lag-1 deficit. Experiment 2.2 was designed to provide a test of this account. This was done by reducing the spatial separation between the two RSVP streams. Bringing the streams closer together should reduce the time required for the attentional focus to shrink to the first-target location. By the same token, closer spatial proximity should reduce the time required for the focus to re-expand so as to again encompass both streams. The hypothesized changes in the extent of the focus of attention as a function of SOA and spatial separation are illustrated in Figure 2.8.  Figure 2.8. Schematic illustration of the progressive changes in the spatial extent of the focus of attention (segmented rectangles) as a joint function of SOA and Spatial Separation of the RSVP streams. See text for a detailed description.  45  Separation Far  Close  SOA (ms) 53  T1  T2  T1  T2  133  T1  T2  T1  T2  An important comparison in Figure 2.8 is the extent to which the second target is encompassed within the focus of attention in the Far- and in the Close-stream conditions at different SOAs. At an SOA of 53 ms, the second target falls within the focus of attention in both the Far and the Close conditions. In contrast, at an SOA of 133 ms, the second target lies outside the focus in the Far condition but at least partly within it in the Close condition. This is because, in the Close-stream condition, the distance between the streams is sufficiently small to allow the focus to shrink to the first-target stream and then to re-expand towards the second-target stream. In practice, this means that approximately the same amount of Lag-1 sparing should be in evidence in both the Close and the Far conditions at an SOA of 53 ms, but Lag-1 sparing should occur only in the Close condition at an SOA of 133 ms.  46  Method Observers A total of 15 undergraduate students at both the University of British Columbia and Simon Fraser University participated for course credit. All participants were naïve as to the purpose of the experiment and reported normal or corrected-to-normal vision. Procedure The procedures in Experiment 2.2 were identical to those in the 53-ms and 133-ms SOA conditions in Experiment 2.1 with the single exception that the centre-to-centre separation between the two RSVP streams was reduced from 3.5º to 0.7º. Therefore, the design of Experiment 2.2 was a 3 (Lags 1, 3, or 9) x 2 (Same-stream, Different-stream) x 2 (SOA: 53 or 133 ms) factorial. Results and Discussion As in Experiment 2.1, only those trials in which the first target was identified correctly were included for analysis. The average accuracy scores for first-target identification, collapsed across lags, were 56.4% (SOA 53 ms, same stream), 59.7% (SOA 53 ms, different stream), 91.0% (SOA 133 ms, same stream), and 91.5% (SOA 133 ms, different stream). The corresponding scores when the streams were far apart (Experiment 2.1) were 50.5%, 54.1%, 84.0%, and 86.2%. This pattern of results is consistent with the hypothesis that first-target detectability was not impaired when the RSVP streams were close together. If anything, firsttarget detectability was better when the streams were close together (Experiment 2.2) than when they were far apart (Experiment 2.1).  47  The percentages of correct second-target responses for SOAs of 53 ms and 133 ms as a function of Lag and Same/Different stream conditions are illustrated as unconnected open symbols in Figures 2.3A and 2.3B. The data were analyzed in a 3 (Lag: 1, 3, or 9) x 2 (Same or Different stream) x 2 (SOA: 53 or 133 ms) within-subject ANOVA. The analysis revealed significant effects of Lag, F(2,28) = 4.47, p = .02, ηp2 = .242 Stream, F(1,14) = 32.528, p < .001, ηp2 = .699, and SOA, F(1,14) = 103.46, p < .001, ηp2 = .881. There was one significant twoway interaction effect between Lag and SOA, F(2,28) = 12.95, p < .001, ηp2 = .48. The threeway interaction effect among Lag, Stream, and SOA was also significant, F(2,28) = 8.10, p = .002, ηp2 = .367. The significance of the three main effects and the two-way interaction is qualified by the significant three-way interaction. As was the case in Experiment 2.1, of particular interest in the present experiment was the condition in which the two targets were presented in opposite streams. For that reason, a separate 3 (Lags: 1, 3, or 9) x 2 (SOAs: 53 ms or 133 ms) factorial analysis was performed on those data. The analysis revealed significant effects of Lag, F(2,28) = 3.51, p = .04, ηp2 = .201, SOA, F(1,14) = 60.7, p < .001, ηp2 = .813, and a significant interaction effect between Lag and SOA, F(2,28) = 17.75, p < .001, ηp2 = .56. The interaction effect reflects the finding that Lag-1 sparing was in evidence at an SOA of 53 ms, t(17) = 2.25, p = .03, but not at an SOA of 133 ms, t(17) = 0.96, p = .35. The important finding is that no Lag-1 deficit was in evidence at an SOA of 133 ms. This contrasts sharply with the corresponding result in the 133-ms SOA condition in Experiment 2.1 which revealed a highly significant Lag-1 deficit, t(23) = 4.28, p < .001. Figure 2.7 permits a direct comparison between the magnitudes of Lag-1 sparing obtained at SOAs of 53 and 133ms in Experiments 2.1 (far streams, filled symbols) and 2 (close streams, open symbols). Spatial separation between the streams made no difference to the 48  magnitude of Lag-1 sparing when the SOA was 53 ms. When the SOA was 133 ms, however, a Lag-1 deficit was in evidence when the streams were far apart (Experiment 2.1) but not when the streams were close together (Experiment 2.2). Indeed, if anything, at an SOA of 133 ms, the results of Experiment 2.2 revealed a small amount of Lag-1 sparing. This pattern of results matches the predictions from the model illustrated in Figure 2.8. The results of Experiment 2.2 also speak to a potential alternative interpretation of the linear relationship between SOA and Lag-1 sparing obtained in Experiment 2.1 (Figure 2.7). It could be suggested that the inverse relationship may stem from a progressive increment in the detectability of the first target as the SOA is increased. The reasoning would be as follows: the first target may be less detectable when the SOA is short than when it is long, perhaps because at shorter SOAs the first-target mask occurs sooner and/or because there is less time to switch processing from the preceding item (Ghorashi, Zuvic, Visser, & Di Lollo, 2003). A less detectable first target would delay the signal that triggers the shrinking of focal attention. Thus, as the SOA is decreased, the trigger signal would be issued correspondingly later. And the later the shrinking is triggered, the longer the second target would remain within the focus of attention. For example, at an SOA of 53 ms the detectability of the first target is low, the trigger is issued relatively late, and the focus remains broadly-set causing the second target to remain within the focus of attention. Lag-1 sparing then follows. In contrast, at an SOA of 133 ms, first-target detectability is high and the trigger is issued promptly, leaving sufficient time for the focus to shrink, causing the second target to lie outside the focus of attention with consequent absence of Lag-1 sparing. The results of Experiment 2.2 disconfirm this interpretation. According to the “firsttarget detectability” hypothesis, the relationship between Lag-1 sparing and SOA should be  49  unaffected by spatial separation. This is because the factors that influence the detectability of the first target (masking, switch costs) are invariant with spatial separation. The prediction that stems directly from this hypothesis is that the slope of the function illustrated in Figure 2.7 should be invariant with spatial separation. This prediction is clearly disconfirmed by the results illustrated in Figure 2.7 showing that spatial separation had a strong effect on the slope of the function relating Lag-1 sparing to SOA. Experiment 2.3 In Experiment 2.1, the stimuli were displayed for approximately 2/3 of the SOA, with the screen remaining blank for the remaining 1/3. This was done to maintain an approximately proportional relationship between stimulus duration and ISI. This procedure, however, might have introduced an unintended variation in the brightness of the stimuli. Because of Bloch’s law, stimuli displayed for shorter durations might have been seen as dimmer than those shown for longer durations (Bloch, 1885). Experiment 2.3 was a control experiment designed to dismiss the option that the results might have been influenced by possible brightness differences. This was done by maintaining a fixed exposure duration for the stimuli and varying the duration of the blank ISI to obtain the required SOA. Method Observers A total of 18 undergraduate students at both the University of British Columbia and Simon Fraser University participated for course credit. All participants were naïve as to the purpose of the experiment and reported normal or corrected-to-normal vision.  50  Procedure The procedures in Experiment 2.3 were identical to those in the 53-ms and 133-ms SOA conditions in Experiment 2.1 with the single exception that the exposure duration of all items in the RSVP streams was 26.5 ms. The balance of the time required to complete the SOA was filled by a blank ISI (26.5 ms for the 53-ms SOA condition, and 106.5 ms for the 133-ms condition). Therefore, the design of Experiment 2.3 was a 3 (Lags 1, 3, or 9) x 2 (Same-stream, Different-stream) x 2 (SOA: 53 or 133 ms) factorial. Results and Discussion As in Experiment 2.1, only those trials in which the first target was identified correctly were included for analysis. The average accuracy scores for first-target identification, collapsed across lags, were 51.1% (SOA 53 ms, same stream), 51.5% (SOA 53 ms, different stream), 78.2% (SOA 133 ms, same stream), and 79.3% (SOA 133 ms, different stream). The corresponding scores in Experiment 2.1, in which the exposure duration of each stimulus was proportional to the total SOA, were 50.5%, 54.1%, 84.0%, and 86.2%. The percentages of correct second-target responses for SOAs of 53 ms and 133 ms as a function of Lag and Same/Different stream conditions are illustrated in Figure 2.9 (solid lines). The data were analyzed in a 3 (Lag: 1, 3, or 9) x 2 (Same or Different stream) x 2 (SOA: 53 or 133 ms) within-subject ANOVA. The analysis revealed significant effects of Lag, F(2,34) = 7.415, p = .002, ηp2 = .304, Stream, F(1,17) = 15.24, p = .001, ηp2 = .473, and SOA, F(1,17) = 36.76, p < .001, ηp2 = .684, There were two significant two-way interaction effects, one between Lag and SOA, F(2,34) = 4.45, p = .02, ηp2 = .21, and one between Stream and SOA, F(1, 17) = 4.17, p = .03, ηp2 = .24. The three-way interaction effect among Lag, Stream, and SOA was also significant, F(2,34) = 5.16, p = .01, ηp2 = .233. The pattern of results of 51  Experiment 2.3 clearly parallels that of Experiment 2.1. This indicates that the same experimental outcome is obtained whether the relationship between stimulus duration and blank ISI is proportional (Experiment 2.1) or fixed (Experiment 2.3).  Figure 2.9. Mean percentages of correct identifications of the second target in Experiment 2.1, in which the exposure duration was proportional (dashed lines), and in Experiment 2.3, in which the exposure duration was fixed (solid lines). Panel A shows the results with a 53-ms SOA; panel B shows the results with a 133-ms SOA.  52  General Discussion From a general standpoint, the present research examined the spatial and temporal dynamics of attentional control, as instantiated in the model illustrated in Figure 2.1. Specifically, we employed the incidence and magnitude of Lag-1 sparing to monitor changes in the spatial extent of the focus of attention. To this end, we used a dual-stream RSVP paradigm and manipulated the SOA between successive items in the stream. Of special interest were those trials in which the two targets appeared in opposite streams. At the beginning of each such trial, the focus of attention was presumed to be set widely so as to encompass both streams. Based on earlier evidence, however, we expected that upon the presentation of the first target, the focus of attention would narrow reflexively onto that location so as to optimize identification of the first target (Jefferies, Ghorashi, Kawahara, & Di Lollo, 2007; Visser, Bischof, & Di Lollo, 2004). When both RSVP streams are encompassed within the focus of attention, the second target is processed accurately along with the first. This occurs when the SOA between successive items in the RSVP stream is too short for the focus to have narrowed on the location of the first target. In this case, the second target falls within the focus of attention, and Lag-1 sparing ensues. At long SOAs, on the other hand, there is sufficient time for the focus to narrow fully onto the location of the first target, leaving the second target outside focal attention and therefore effectively unattended. In this case, a Lag-1 deficit ensues, rather than Lag-1 sparing. The outcomes of both experiments confirmed these expectations. While confirming the relationship between the spatial extent of the focus of attention and Lag-1 sparing (illustrated in Figures 2.1 and 2.4), the data in Figure 2.7 do not permit the rate of change in the extent of the focus of attention to be expressed in spatiotemporal units such as degrees of visual angle/ms. What can be inferred from Figure 2.7, however, is that the extent of  53  the focus of attention varied linearly as a function of time in the manner illustrated in the first column of Figure 2.1. Spatial Dynamics of the Focus of Attention: Analog or Quantal? The continuous, smooth changes in the width of the attentional focus illustrated in Figures 2.1 and 2.7 are consistent with the view that changes in the spatial deployment of attention over time follow an analog course. This contrasts with the view expressed by Weichselgartner and Sperling (1987) and by Sperling and Weichselgartner (1995) that the spatial deployment of attention over time follows a quantal course. These contrasting views pose an obvious question: is attention re-deployed in an analog or in a quantal manner? The present study and that of Weichselgartner and Sperling (1987) lead to ostensibly inconsistent answers. This inconsistency, however, is easily resolved by considering the task differences between the two studies. The observers in Weichselgartner and Sperling’s (1987) study were required to redeploy the focus of attention between two discrete spatial locations. In contrast, in the present experiments the observers were required to monitor both spatial locations concurrently. Given these task differences, the optimal strategy in Weichselgartner and Sperling’s (1987) study was to re-deploy the focus of attention discretely from one location to the other. In contrast, in the present research the optimal strategy was to maintain a wide focus of attention (and, if anything, resist the reflexive narrowing of the attentional focus to the location of the first target) throughout a trial. We are led by this line of argument to the following conclusion: whether the focus of attention is modulated in an analog or quantal manner depends on the task at hand. If the task involves a discrete switch of locations, attention is re-deployed in a quantal  54  manner. If the task requires the maintenance of a wide attentional setting, any shrinking or expanding of the focus of attention is accomplished in an analog manner. The Focus of Attention: Unitary or Divided? There is evidence in the attention literature that separate and independent foci can be deployed simultaneously to discrete locations in space (Awh & Pashler, 2000; Kawahara & Yamada, 2006; Müller, Malinowski, Gruber, & Hillyard, 2003). The results of Experiment 2.1 can be explained equally well by a single focus of attention that expands or contracts (see Figure 2.1) or by two discrete foci, each centered on one RSVP stream. We have seen how the results of Experiment 2.1 can be explained in terms of a unitary attentional focus that shrinks and expands. In order to interpret those results in terms of two discrete foci, it must be assumed that the two foci draw on a single resource pool. The reasoning is as follows. At the outset of a trial, attention is deployed to two separate foci, one at each RSVP stream. Upon detection of the first target, the resources allocated to the stream opposite that of the first target, are gradually redeployed to the stream containing the first target so as to enhance first-target identification. At short SOAs, the reallocation of resources has only just begun. Thus, if the second target appears in the stream opposite the first, sufficient attentional resources remain at that location to process the second target, resulting in Lag-1 sparing. At longer SOAs, the transfer of attentional resources to the first-target stream will have been completed, leaving few or no resources for the second target if it appears in the stream opposite the first. This would result in Lag-1 deficit. Thus, the results of Experiment 2.1 can be explained in terms of two discrete foci. Such a dual-focus account could be mediated by at least two mechanisms. One is that the two foci draw upon a single pool of resources. A single-resource assumption is indicated 55  because, if each focus were to draw upon its own independent resource pool, Lag-1 sparing should always occur, even when the two targets are presented in opposite streams, because the resources initially deployed to the stream which does not contain the first target would remain unchanged. A second possibility is an inhibitory or strategic mechanism of attentional control which would act concurrently on the two foci even if their respective resource pools were independent. The single- and the dual-focus models, however, make different predictions regarding the effect of reducing the spatial separation between the two RSVP streams. According to the single-focus model, as the separation between the two streams is reduced, it should take less time for the focus of attention to shrink to the location of the first target and to re-expand to the location of the second (see Figure 2.8). This would result in reduced Lag-1 deficit at the longer SOAs. In contrast, according to the dual-focus model, the shifting of resources from one focus to the other should be independent of spatial separation. Thus, the magnitudes of Lag-1 sparing and Lag-1 deficit should be invariant with spatial separation. The results of Experiment 2.2 show that at an SOA of 133 ms, the magnitude of Lag-1 deficit was reduced relative to that obtained at the corresponding SOA in Experiment 2.1. In fact, at an SOA of 133 ms, the Lag-1 deficit seen in Experiment 2.1 changed to Lag-1 sparing in Experiment 2.2. This finding is consistent with predictions from the single-focus model, but not with predictions from the dualfocus model. The inability of the dual-focus model to account for the overall pattern of results, however, does not mean that the model is invalid under all circumstances. The evidence supporting the claim that attention can be deployed to several spatial locations concurrently and independently is substantial and convincing (e.g., Awh & Pashler, 2000; Kawahara & Yamada,  56  2006; Müller, Malinowski, Gruber, & Hillyard, 2003; Yamada & Kawahara, 2007). The evidence in favour of a single focus of attention that expands and contracts is equally believable (e.g., Barriopedro & Botella, 1998; Egeth, 1977; Eriksen & St. James, 1986; Eriksen & Yeh, 1985; Jonides, 1983; LaBerge, 1983, 1995). It seems likely, therefore, that both the single-focus and the multiple-independent-foci modes of attentional deployment are valid. Whether one or the other is employed in any given instance depends on the specific demands of the task at hand, the objective being to optimize performance. That this is in fact the case, has been demonstrated in a study by Jefferies and Di Lollo (2008) in which whether observers employed a single unitary focus or two separate foci depended on task demands. To summarize, the present study employed the well-established phenomenon of Lag-1 sparing to examine the spatiotemporal modulations of attention. Systematic variation of the interval between successive items in an RSVP stream revealed that the spatial extent of the focus of attention varies linearly over time.  57  References Allport, A., Styles, E. A., & Hsieh, S. (1994). Shifting intentional set: Exploring the dynamics of tasks. In C. Umiltà & M. Moscovitch (Eds.), Attention and Performance XV, (pp. 421452). Cambridge, MA: MIT Press, Bradford Books. Arnell, K., & Jolicœur, P. (1999). The attentional blink across stimulus modalities: Evidence for central processing limitations. Journal of Experimental Psychology: Human Perception and Performance, 25, 630-648. Awh, E. & Pashler, H. (2000). Evidence for split attentional foci. Journal of Experimental Psychology: Human Perception and Performance, 26, 834-846. Barriopedro, M.I. & Botella, J. (1998). New evidence for the zoom lens model using the RSVP technique. Perception & Psychophysics, 60, 1406-1414. Benso, F., Turatto, M., Mascetti, G.G., & Umiltà, C. (1998). The time course of attentional focusing. European Journal of Cognitive Psychology, 10, 373-388. Bloch, A. M. (1885). Expériences sur la vision. Comptes Rendus de Séances de la Societé de Biologie, Paris, 37, 493-495. Bowman, H. & Wyble, B. (2007). The simultaneous type, token, serial token mode of temporal attention and working memory. Psychological Review, 114, 38-70. Breitmeyer, B. G. (1984). Visual masking: An integrative approach. New York: Oxford University Press. Breitmeyer, B. G., Ehrenstein, A., Pritchard, K. K., Hiscock, M., & Crisan, J. (1999). The roles of location specificity and masking mechanisms in the attentional blink. Perception & Psychophysics, 61, 798-809. Castiello, U. & Umiltà, C. (1990). Size of the attentional focus and efficiency of processing.  58  Acta Psychologica, 73, 195-209. 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, 21, 109-127. Dell’Acqua, R., Pascali, A., Jolicœur, P., & Sessa, P. (2003). Four-dot masking produces the attentional blink. Vision Research, 43, p. 35-40. Duncan, J., Ward, R., & Shapiro, K. (1994). Direct measurement of attentional dwell time in humans.  ature, 369, 313-315.  Dux, P.E., & Coltheart, V. (2005). The meaning of the mask matters: Evidence of conceptual interference in the attentional blink. Psychological Science, 16, 775-779. Enns, J. T. (2004). Object substitution and its relation to other forms of visual masking. Vision Research, 44, 1321-1331. Eriksen, C.W. & St. James, J.E. (1986). Visual attention within and around the field of focal attention: A zoom lens model. Perception & Psychophysics, 40, 225-240. Eriksen, C.W. & Yeh, Y-y. (1985). Allocation of attention in the visual field. Journal of Experimental Psychology: Human Perception and Performance, 11, 583-597. Fehrer, E. (1966). Effects of stimulus similarity on retroactive masking. Journal of Experimental Psychology, 71, 612-615. Giesbrecht, B. L., & Di Lollo, V. (1998). Beyond the attentional blink: Visual masking by object substitution. Journal of Experimental Psychology: Human Perception and Performance, 24, 1454-1466. Harmon, L.D., & Julesz, B. (1973). Masking in visual recognition: Effects of twodimensional filtered noise. Science, 180, June, 1194-1196.  59  Helmholtz, H. V. (1866). Handbuch der physiologischen Optik (1st ed.). Leipzig: Voss. (Trans. J. P. C. Southall (1962). Handbook of physiological optics (3rd ed.). New York: Dover). Holländer, A., Corballis, M.C., & Hamm, J.P. (2005). Visual-field asymmetry in dualstream RSVP.  europsychologia, 43, 35-40.  Intraub, H., (1981). Rapid conceptual identification of sequentially presented pictures. Boundary extension for briefly glimpsed pictures: Journal of Experimental Psychology: Human Perception and Performance, 7, 604-610. Intraub, H., (1984). Conceptual masking: The effects of subsequent visual events on memory for pictures. Rapid conceptual identification of sequentially presented pictures. Boundary extension for briefly glimpsed pictures: Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 115-125. James, W. (1950). The Principles of Psychology, Vol. 1. New York: Dover. (Original work published in 1890). Jefferies, L.N. & Di Lollo, V. (2008; under editorial review). Is focal attention unitary or divided? It depends on task demands. Jefferies, L.N., Ghorashi, S., Kawahara, J-i., & Di Lollo, V. (2007). Ignorance is bliss: the role of observer expectation in dynamic spatial tuning of the attentional focus. Perception & Psychophysics, 69, 1162-1174 . Jolicœur, P. & Dell’Acqua, R. (1998). The demonstration of short-term consolidation. Cognitive Psychology, 36, 138-202. Jonides, J. (1983). Further toward a model of the mind’s eye’s movement. Bulletin of the Psychonomic Society, 21, 247-250. Joseph, J.S., Chun, M.M., & Nakayama, K. Attentional requirements in a ‘preattentive’  60  feature search task.  ature, 379, 805-807.  Juola, J.F., Botella, J., & Palacios, A. (2004). Task- and location-switching effects on visual attention. Perception & Psychophysics, 66, 1303-1317. Kawahara, J-i., & Yamada, K. (2006). Two noncontiguous locations can be attended concurrently: Evidence from the attentional blink. Psychonomic Bulletin & Review, 13, 594-599. Kristjansson, A., & Nakayama, K. (2002). The attentional blink in space and time. Vision Research, 42, 2039-2050. LaBerge, D. (1983). Identification of the time to switch attention: A test of a serial and a parallel model of attention. In Attention and Performance IV, ed. S. Kornblum. New York: Academic Press. LaBerge, D. (1995). Attentional Processing: The Brain’s Art of Mindfulness. Harvard University Press: Cambridge, MA. Maki, W. S., Couture, T., Frigen, K., & Lien, D. (1997). Sources of the attentional blink during rapid serial visual presentation: Perceptual interference and retrieval competition. Journal of Experimental Psychology: Human Perception and Performance, 23, 13931411. Moore, C. M., Egeth, H., Berglan, L. R., & Luck, S. J. (1996). Are attentional dwell times inconsistent with serial visual search? Psychonomic Bulletin & Review, 3, 360-365. Müller, M.M, Malinowski, P.,Gruber, T., & Hillyard, S.A. (2003). Sustained division of the attentional spotlight.  ature, 424, 309-312.  Müller, M.M., Teder-Salejarvi, W. & Hillyard, S.A. (1998). The time course of cortical facilitation during cued shifts of visual attention.  ature euroscience, 1, 631-634.  61  Nakayama, K. & Mackeben, M. (1989). Sustained and transient components of focal visual attention. Vision Research, 29, 1631-1647. Peterson, M. S., & Juola, J. F. (2000). Evidence for distinct attentional bottlenecks in attention switching and attentional blink tasks. Journal of General Psychology, 127, 6-26. Potter, M. C., Chun, M. M., Banks, B. S., & Muckenhoupt, M. (1998). Two attentional deficits in serial target search: The visual attentional blink and an amodal taskswitch deficit. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 979-992 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, 28, 1149-1162. 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, 18, 849-860. Seiffert, A.E., & Di Lollo, V. (1997). Low-level masking in the attentional blink. Journal of Experimental Psychology: Human Perception and Performance, 23, 1061-1073. Shapiro, K.L., Caldwell, J., & Sorensen, R.E. (1997). Shapiro, K. L., & Raymond, J. E. (1994). Journal of Experimental Psychology: Human Perception and Performance, 23, 501-514. 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, 20, 357-371. Shepard, R.N., & Metzler, J. (1971). Mental rotation of three-dimensional objects.  62  Science, 191, 701-703. Shih, S.I., (2000). Recall of two visual targets embedded in RSVP streams of distractors depends on their temporal and spatial relationship. Perception & Psychophysics, 62, 1348-1355. Spalek, T.M., Falcon, L.J., & Di Lollo, V. (2006). Attentional blink and attentional capture: Endogenous versus exogenous control over paying attention to two important events in close succession. Perception & Psychophysics, 68, 674-684. Sperling, G. & Weichselgartner, E. (1995). Episodic theory of the dynamics of spatial attention. Psychological Review, 102, 503-532. Visser, T. A. W., Bischof, W. F., & Di Lollo, V. (2004). Rapid serial visual distraction: Taskirrelevant items can produce an attentional blink. Perception & Psychophysics, 66, 14181432. Visser, T. A. W., Bischof, W. F., & Di Lollo, V. (1999). Attentional switching in spatial and non-spatial domains: evidence from the attentional blink. Psychological Bulletin, 125, 458-469. Weichselgartner, E., & Sperling, G. (1987). Dynamics of controlled visual attention. Science, 238, 778-780. Yamada, Y. & Kawahara, J-i. (2007). Dividing attention between two categories and locations in rapid serial visual presentation. Perception & Psychophysics, 69, 1218-1229.  63  Chapter 3. Adjusting the Focus of Attention Across Time: Factors which Influence the Rate of arrowing2 Stimuli presented at attended locations are processed faster and more accurately than those presented at unattended locations (Helmholtz, 1866/1962; James, 1890/1950; LaBerge, 1995). A common metaphor to describe this finding is that attention functions like a spotlight: items that fall within the spotlight are processed faster and more accurately than items that fall outside. A characteristic that attention shares with a spotlight is that both can be directed at specific locations and can be moved rapidly from one location to another. Extending this metaphor to include the possibility of focusing attention either narrowly or broadly, Eriksen and colleagues (Eriksen & Yeh, 1985; Eriksen & St. James, 1986) proposed that the focus of attention can be resized like a spotlight equipped with a zoom lens. The present study examines some of the factors that influence the temporal dynamics of changes in attentional focus. To index the spatial extent of attention over time, we followed Jefferies and Di Lollo (in press) in using the phenomenon known as the attentional blink (AB); the identification of the second of two targets is impaired when it is presented shortly after the first (Chun & Potter, 1995; Raymond, Shapiro, & Arnell, 1992). Typically, the two targets (T1 and T2) are inserted in a stream of distractors presented in rapid serial visual presentation (RSVP). The second-target deficit is most pronounced at short inter-target lags and diminishes as the lag is increased to about 700 ms. A description of the Jefferies and Di Lollo (in press) study will set the stage for the present investigation. Participants in that study monitored two concurrent RSVP streams, one on either side of fixation, in order to identify two targets (T1 and T2) that could appear  2  A version of this chapter will be submitted for publication. Jefferies, L.N., Ghorashi, S., Enns, J.T., & Di Lollo, V. Adjusting the focus of attention across time: Factors which influence the rate of narrowing.  64  unpredictably either in the same stream or in opposite streams. Within this procedure, the analysis focused on accuracy of T2 when it occurred immediately following T1 (i.e., at Lag 1). In this temporal position, T2 can either fall prey to reduced accuracy relative to other Lags (i.e., the typical AB) or it can give rise to the seemingly paradoxical finding known as Lag-1 sparing, in which the AB is much reduced (Potter, Chun, Banks, & Muckenhoupt, 1998). Jefferies and Di Lollo (in press) demonstrated that the change in accuracy from Lag-1 sparing to the AB was approximately linearly related to the amount of time that elapsed between successive items in the RSVP stream, and thus proposed that this change in T2 accuracy at Lag 1 was an index of attentional narrowing. In support of using this procedure to monitor the spatial extent of focal attention over time, we can point to the findings of Jefferies, Ghorashi, Kawahara, and Di Lollo (2007), who spelled out the conditions under which Lag-1 sparing occurs when the two targets are presented in opposite streams. Typically, Lag-1 sparing does not occur when the two targets are presented in opposite streams. Lag-1 sparing does occur, however, even though the targets appear in opposite streams, provided that T2 falls within the focus of attention. By systematically varying the stimulus-onset asynchrony (SOA) between successive items in the RSVP stream, the incidence of Lag-1 sparing when T1 and T2 are presented in opposite streams can be used to monitor the rate at which focal attention shrinks and expands to encompass one or both RSVP streams. Jefferies and Di Lollo (in press) modelled the relationship between Lag-1 sparing and dynamic changes in focal attention on the assumption that in dual-stream displays in which T1 appears unpredictably in either stream, the optimal initial strategy would be to set a broad focus of attention so as to encompass both streams. When T1 appears, the attentional focus begins to  65  narrow to the stream containing T1 so as to optimize target identification. If T2 then appears in the same stream, it will fall within the focus of attention and its identification will be enhanced, resulting in Lag-1 sparing. If, however, T2 is presented in the opposite stream, it will not be encompassed within the attentional focus, and its identification will hinge on the length of the SOA, as illustrated in Figure 3.1, redrawn from Figure 2.1 in Chapter 2 (Jefferies & Di Lollo, in press).  Figure 3.1. Schematic illustration of the progressive changes in the spatial extent of the focus of attention (segmented rectangles) as a function of SOA and Lag. See text for a detailed description.  Figure 3.1 illustrates the spatial extent of the focus of attention (segmented boxes) at Lag 1 at each of four SOAs. At the shortest SOA (80 ms) the attentional focus still mostly  66  encompasses both streams because the process of narrowing to the T1 stream has only just begun. As the SOA is increased, there is correspondingly more time for the attentional focus to shrink to the T1 stream before the onset of T2. If T2 then appears in the same stream as T1, it will fall within the focus of attention, and Lag-1 sparing will occur regardless of SOA. If, however, T2 appears in the opposite stream, the probability of Lag-1 sparing will decrease as SOA increases. This will cause a progressive change from Lag-1 sparing to its opposite, Lag-1 deficit, as the attentional focus is progressively withdrawn from the non-T1 stream. As outlined above, it is assumed that the focus of attention does not begin to narrow until T1 appears. The proposed reason for the narrowing process is to enhance identification accuracy of the first target. Considered in tandem, these two assumptions lead to an apparent paradox: in order to narrow to the location of T1 one must know it is the target – on the other hand, once the target has been identified, there should be no need for the focus of attention to narrow to the target’s location. The solution to this paradox lies in the fact that there is a distinction between the classification of an item and its identification. That is, an item may be categorized as a target (i.e., as a letter) before its full identity can be extracted (i.e.,which letter). In one pertinent study, for example, Grill-spector and Kanwisher (2005) demonstrated that an item could be classified as soon as it was detected, but that the process of identification was not complete at this stage and required an additional 65 ms or so. Given this, we may assume that when the target is detected, it is categorized as a target, which initiates the narrowing process. A narrower focus of attention then enables full target identification. The present study examined whether the rate at which attention narrows over time when monitoring dual RSVP streams is modulated by factors that cause attention to be attracted more or less rapidly to target stimuli in other situations. At a first approximation, attention has been  67  shown to be directed either by the goals of the observer (i.e., the intention to process a certain class of information and not other information) or by the visual reflexes of the observer (e.g., the tendency to orient to spatial non-uniformities in simple features such as luminance, shape, and motion) (Egeth & Yantis, 1997; Folk, Remington, & Johnson, 1992; Bacon & Egeth, 1994). During everyday visual functioning, such as occurs when driving a car or crossing a street as a pedestrian, these two ways of directing attention are in a dynamic balance, so as to make the observer adaptive to changing circumstances. This balance prevents the observer from missing unexpected relevant information when pursuing a given goal (i.e., by evoking the visual orienting reflex) or when being guided by the visual orienting reflex (i.e., by continuing to monitor for goal-relevant information). In what follows, we first explore changes in the rate of attentional narrowing that occur when stimuli are more or less relevant to the goals of the observer (Experiments 3.1 and 3.2), before exploring changes in the rate of attentional narrowing that occur when stimuli evoke the visual orienting reflex (Experiment 3.3). We will use the terms attentional pull and attentional capture interchangeably to refer to the relation between stimulus characteristics and the observer’s attentional system. Experiment 3.1 Experiment 3.1 asked whether attentional narrowing was influenced by the relevance of stimulus characteristics to the observer’s main task. Since the targets defined for the observers were letters, we contrasted two types of distractors: a) one class of distractors was digits, which are closely related in features, meaning, and in relative familiarity to letters, and hence should strongly attract and hold attention because of their similarity to the target feature class (see Ghorashi, Zuvic, Visser, & Di Lollo, 2003). This class of distractors was referred to as strong68  pull distractors. b) A second class of distractors consisted of random-dot patterns, which should not attract attention in a goal-relevant way (weak-pull distractors). T1 appeared unpredictably in either the digit stream or the random-dot stream; T2 appeared in either the same stream as T1 or in the other stream. The condition of interest in the present work is that in which the two targets appear in different streams because only in that condition is it possible to assess the extent to which attention has narrowed to the location of T1 and, therefore, the extent to which T2 in the opposite stream is unattended. On the basis of this relative difference in attentional pull, we expected that the focus of attention would shrink more rapidly to the T1 location when T1 appeared in the digit stream (i.e., the strong-pull stream) than when it appeared in the random-dot stream (the weak-pull stream). Conversely, attention can be disengaged from a weak-pull stream (i.e., the random-dot stream) more readily than from a strong-pull stream (i.e., the digit stream). This conceptualization is illustrated in Figure 3.2.  Figure 3.2. Schematic representation of the sequence of events within a trial in Experiment 3.1. The first and the second targets (T1 and T2) could appear in either the digit or the random-dot stream and in either the same or opposite streams. The segmented rectangle represents the spatial extent of the focus of attention. Note that in the Fast-shrink condition, very little time is required to disengage from the weak-pull random-dot patterns. When the second target appears in the opposite stream it therefore falls outside the focus of attention. In the Slow-shrink condition, in contrast, the process of disengaging from the strong-pull digits is relatively slow, and the second target therefore falls within the focus of attention.  69  Item opposite T1 has weak attentional pull “Fast Shrink” (T2)  8 5  Item opposite T1 has strong attentional pull “Slow Shrink”  + 3  8  + D  D  + ....... . R + ...... 9  (T1)  4 5  . + ......  + .......  ...  9 2  . + ......  + ......  + 1  + R  . 4 + ...... 5 + .......  .. .  These ideas have clear implications for the incidence of Lag-1 sparing and Lag-1 deficit. Consider, for example, the spatial extent of the attentional focus (segmented box in Figure 3.2) in the RSVP frame directly following T1. When T1 is presented in the random-dot (weak pull) stream, attention lingers for a relatively long time on the digit (strong-pull) stream thus keeping T2 within the focus of attention; Lag-1 sparing then follows. We refer to this as the Slow-shrink condition. In contrast, when T1 is presented in the digit stream, attention withdraws rapidly from the random-dot stream leaving T2 unattended with consequent Lag-1deficit. We refer to this as the Fast-shrink condition. In the present work we tested these notions by employing four different SOAs (80, 100, 118, and 133 ms). We expected that at the shortest SOA the focus of attention would still 70  encompass both streams, producing Lag-1 sparing of equal magnitude in both the Slow- and Fast-shrinking conditions. The two conditions should differ in respect to Lag-1 sparing, however, at intermediate SOAs. Namely, the T2 location should remain within the focus of attention longer in the Slow- than in the Fast-shrink condition, with Lag-1 sparing continuing to be in evidence in the former more so than in the latter. At the longest SOA, the focus of attention would be expected to have narrowed fully to the T1 location leaving the T2 location unattended with consequent Lag-1 deficit in both conditions. The present experiment was a direct test of these hypotheses.  Method Participants A total of 75 undergraduate students from the University of British Columbia and Simon Fraser University participated in the experiment for course credit. A different group of participants completed each SOA, and hence there were a total of four groups. Each group was required to have a minimum of 14 observers, but because the data were collected concurrently at two universities, the final number of observers in each group was 14, 22, 17, and 22 for the 80, 100, 118, and 133-ms SOA groups, respectively. All participants were naive as to the purpose of the experiment and reported normal or corrected-to-normal vision. Apparatus and Stimuli All stimuli were presented on a computer monitor viewed from a distance of approximately 57 cm. A white fixation cross (0.25° x 0.25°) was displayed in the center of the screen throughout each trial. Two RSVP streams were presented, one 1.75° to the left of fixation, the other 1.75° to the right of fixation. One RSVP stream consisted of white digits (0-9,  71  measuring 0.9° vertically) and the second RSVP stream consisted of random-dot patterns. The random-dot patterns were created by randomly distributing 10 one-pixel dots into a square region of space 0.9° x 0.9°, with the restriction that the dots not overlap one another. The dots were randomly re-arranged on each new frame in the RSVP sequence. The targets were capital letters (excluding the letters I, O, Q, Z), all of which subtended approximately 0.9° vertically. The luminance of all stimuli was 90 cd/m2, and the luminance of the black background was 2.3 cd/m2. Procedure The simplest means of presenting the two RSVP streams simultaneously is to attach an item from each stream to a frame. The frames are then presented in sequential order for 100 ms. Each stimulus is presented for approximately two-thirds of the SOA, followed by a blank interstimulus interval (ISI) for the remaining one-third of the SOA. A fixation cross was present in the center of the screen throughout each trial, and the participants pressed the space bar to initiate each new trial. The trial began with the presentation of two synchronized RSVP streams, one to the left and the other to the right of fixation. One RSVP stream contained digit distractors while the other contained random-dot patterns. On a randomly intermixed half of the trials the digit stream appeared to the left and the random-dot pattern stream to the right of fixation; on the remaining half of the trials the digit stream appeared to the right and the random-dot pattern stream to the left. Both streams contained 8 to 14 leading distractor items prior to the onset of T1. Two non-identical letter targets were presented during the RSVP sequence. Both targets appeared randomly in the digit stream on half the trials and in the random-dot stream on half of the trials. As such, the two targets appeared in the same stream on half of the trials and in opposite streams on the remaining trials. Each target  72  was followed by a digit mask if it appeared in the random-dot stream. The single exception to this was for T1 at Lag-1 when T2 appeared in the same stream as the first, in which case T1 was masked by T2. T2 was presented at an inter-target lag of 1, 3, or 9. At Lag 1, T2 immediately followed the T1; at Lag 3 two frames of distractors intervened between the two targets; at Lag 9 there were 8 intervening target frames. It is worth noting at this point that Lag 9 is of little theoretical interest to the present work since Lag-1 sparing is defined by the difference in T2 identification accuracy at Lag 1 compared to Lag 3. Lag 9 was included in the experiment, however, to maintain the temporal unpredictability of target appearance. The actual amount of time in milliseconds that elapsed from the onset of T1 to the onset of T2 at leach lag depended on the SOA. The number of milliseconds at Lags 1, 3, and 9 was 80, 160, and 640 for the 80-ms group; 100, 200, and 800 for the 100-ms SOA group, 118, 236, and 944 for the 118-ms group, and 133, 266, and 1064 for the 133-ms group. The SOA between successive items in the RSVP stream was comprised of approximately two-thirds stimulus duration and one-third blank inter-stimulus interval (ISI). The precise ratio of stimulus duration to blank ISI was constrained by the refresh rate of the monitor and were as follows: 45:35 (80 ms SOA), 70:30 (100 ms SOA), 71:47 (118 ms SOA), and 80:53 (133 ms SOA). The display sequence is illustrated schematically in Figure 3.2. The observer's task was to identify the two target letters and indicate the identity of the two targets by pressing the appropriate keys on the keyboard at the end of each trial. The order of responses was irrelevant, and responses were coded by whether they correctly identified one of the target letters.  73  Results and Discussion Second-Target Accuracy Only those trials in which T1 was identified correctly were included for analysis. This procedure is commonly adopted in AB experiments on the grounds that, on trials in which T1 is identified incorrectly, the source of the error is unknown, and thus its effect on second-target processing cannot be estimated. As noted in the introduction, the critical conditions for the objectives of the present work are those in which the two targets appeared in different streams. The T2 data for these conditions are illustrated in Figure 3.3 below.  Figure 3.3. Mean percentages of correct identifications of the second target in Experiment 3.1 with the data from each SOA plotted in a separate panel. The open symbols represent data from the Slow-shrink condition; the filled symbols represent data from the Fast-shrink condition.  74  100  SOA = 80 ms  SOA = 100 ms  Percentage correct (T2|T1)  80 60 40 20 [A] 1 3 9 100 SOA = 118 ms  [B]  1  3  9  SOA = 133 ms  80 60 40 20 [C] 1  [D]  3  Slow Shrink Fast Shrink  9 1 3 Inter-target Lag  9  Because the dependent measure of interest in the present work was Lag-1 sparing, statistical analysis was limited to the data for Lags 1 and 3. Lag 9 was of no theoretical interest, and was included in the experiment solely to make the appearance of T2 less temporally predictable. Variations in Lag-1 sparing and Lag-1 deficit across conditions are not immediately evident in Figure 3.3. For that reason, the magnitude of Lag-1 sparing and its converse, Lag-1 deficit, was calculated separately for each of the eight conditions illustrated in Figure 3.3. This was done by subtracting the accuracy score at Lag 3 from the corresponding score at Lag 1. A 75  positive difference indicates Lag-1 sparing; a negative difference indicates Lag-1 deficit. These scores are illustrated in Figure 3.4.  Figure 3.4. Variation in the magnitude of Lag-1 sparing (positive values) and Lag-1 deficit (negative values) as a function of SOA. The difference in overall level of performance at the different SOAs was partialled out by expressing the magnitude of Lag-1 sparing/Lag-1 deficit as the ratio of second-target accuracy at Lag 3 to that at Lag 1. The closed symbols represent the data from the Fast-shrink condition; the open symbols represent the data from the Slow-shrink condition.  B ro a In d f e rr e d w id th o f a tt e n ti o n a lw in N d a o rr w o w  30 n o it r o p o r p a s a d e s s e r p x e g n ir a p S 1 g a L  20  ] 0 0 1 *) 10 1 g a L / 3 0 g a L ([ - -10 0 0 1 -20  Slow-Shrink Fast-Shrink  -30 80  100  118 133  SOA (ms) 76  The filled symbols in Figure 3.4 represent the Lag-1 sparing/deficit scores for the condition in which T1 appeared in the digit stream and T2 in the random-dot stream (Fast-shrink condition). The open symbols represent the corresponding scores for the condition in which T1 appeared in the random-dot stream and T2 in the digit stream (Slow-shrink condition). It is clear from Figure 3.4 that in the Fast-shrink condition, the magnitude of Lag-1 sparing diminished rapidly as the SOA was increased and turned into Lag-1 deficit at the longer SOAs. In contrast, in the Slow-shrink condition the magnitude of Lag-1 sparing remained high at the three shortest SOAs and changed to Lag-1 deficit only at the longest SOA. This pattern of results is consistent with the hypothesis that the strength of attentional pull was greater in the digit stream than in the random-dot stream and that the rate at which focal attention narrowed to the digit (strong-pull) stream was faster than the rate at which it narrowed to the random-dot (weak-pull) stream. This is tantamount to saying that attention was disengaged more readily from the random-dot stream than from the digit stream. The shapes of the two functions in Figure 3.4 differ markedly from one another: the Fast-shrink condition appears linear while the Slow-shrink condition appears quadratic. To examine the difference in shape between the two functions, an analysis of covariance (ANCOVA) was performed on the data in Figure 3.4. The within-subject factor varied at two levels defined by whether T1 appeared in the dot stream followed by T2 in the digit stream (Slow-shrink condition), or the reverse (Fast-shrink condition). Two covariates were considered: SOA and SOA2, to examine the linear and quadratic components, respectively. None of the main effects reached significance (although two approached significance): Fast- vs. Slow-shrink, F(1,71)=3.66, p=.060, SOA, F(1,71)=2.66, p=.108, and SOA2, F(1,71)=3.84, p=.054. Of critical importance to the objective of the present work was the significant interactions between the Fast-  77  /Slow-shrink factor and the linear (SOA) and quadratic (SOA2) covariates: SOA, F(1,71)=4.12, p<.05; SOA2, F(1,71)=4.29, p<.05. The outcome of this analysis confirms the graphical evidence in Figure 3.4 that a relative difference in attentional pull, stemming from the similarity between distractor digits and target letters (in contrast to random-dots and letters) resulted in more rapid narrowing to the digit (strong-pull) stream than to the random-dot (weak-pull) stream. The pattern of results at each SOA is consistent with this interpretation. At the shortest SOA of 80 ms, the magnitude of Lag1 sparing was the same in both the Fast-shrink and Slow-shrink conditions. This suggests that at an SOA of 80 ms, the focus of attention still encompassed both streams. At the two intermediate SOAs (100 and 118 ms), however, the magnitude of Lag-1 sparing was substantially greater for the Slow-shrink condition than for the Fast-shrink condition. This suggests that the focus of attention still encompassed both streams in the Slow-shrink condition (in which attention is withdrawn more slowly from the strong-pull stream), but had narrowed to the location of T1 in the Fast-shrink condition (in which attention is rapidly withdrawn from the weak-pull stream). The rapid withdrawal in the Fast-shrink condition is evidenced in Figure 3.4 by the change from Lag-1 sparing to Lag-1 deficit. To probe the interaction effects, a 2 (Lags 1, 3) x 2 (Fast- vs. Slow-shrink) analysis of variance (ANOVA) was performed separately for each of the four SOAs illustrated in Figure 3.3. The outcomes of the analyses were as follows. At an SOA of 80 ms, the analysis revealed a significant main effect of Lag, F(1,12)=5.30, p=.040, ηp2 =.31, and Fast- vs. Slow-Shrink, F(1,12)=18.00, p=.001, ηp2 = .60. The interaction effect was not significant, F(1,12)<1. This indicates that the magnitude of Lag-1 sparing in the Slow-shrink condition was the same as in the  78  Fast-shrink condition. Within the present theoretical framework, this means that both streams were still encompassed within a focus of attention that has not yet begun to shrink. At an SOA of 100 ms, the analysis revealed a significant main effect of Lag, F(1,21)=4.73, p=.041, ηp2 =.184, and a marginally significant effect of Fast- vs. Slow-Shrink, F(1,21)=3.67, p=.069, ηp2 = .149. The interaction effect was significant F(1,21)=5.95, p=.024, ηp2 = .221, indicating that the magnitude of Lag-1 sparing was greater in the Slow-shrink than in the Fast-shrink condition. This is consistent with the idea that the attentional focus had begun to shrink in the Fast-shrink condition (away from the weak-pull stream towards the strong-pull stream), but not in the Slow-shrink condition. At an SOA of 118 ms, the analysis revealed a significant main effect of Fast- vs. Slowshrink, F(1,16)=5.91, p=.027, ηp2 = .27. The main effect of Lag was not significant, F(1,16)<1. The interaction effect was significant F(1,16)=4.82, p=.043, ηp2 =.232, indicating that the magnitude of Lag-1 sparing was greater in the Slow-shrink than in the Fast-shrink condition. Indeed, the Fast-shrink condition now exhibits Lag-1 deficit, suggesting that the focus of attention has been withdrawn entirely from the Dot stream. Finally, at an SOA of 133 ms, the analysis revealed a significant main effect of Lag, F(1,21)=11.48, p=.003, ηp2 =.353. The main effect of Fast- vs. Slow-shrink was not significant, F(1,21) < 1, nor was the interaction effect, F(1,12) < 1. The lack of a significant interaction indicates that the magnitude of Lag-1 sparing was comparable in the Slow-shrink and Fast-shrink conditions. In terms of the present hypothesis, this means that in both conditions the focus of attention had fully narrowed to the T1-stream, leaving the other stream unattended. In summary, the more rapid change from Lag-1 sparing (broad focus of attention) to Lag1 deficit (narrow focus) in the Fast-shrink than in the Slow-shrink condition supports the 79  hypothesis that attentional pull can influence the rate at which the focus of attention narrows. The most basic interpretation of this finding is that because the stream of digits is automatically processed for meaning whereas the stream of random-dot patterns has no such requirement, the process of disengaging from the digit stream is slower, causing the slower narrowing of the focus of attention. One question not answered by Experiment 3.1, however, is whether the more rapid narrowing of attention to target letters in the context of random dots occurred because the leading streams (items prior to the target) differed in their task-relevance prior to the onset of T1, or whether it was sufficient for T1 to be paired with a low-relevance item at its onset. It is possible that the more rapid narrowing in Experiment 3.1 occurred prior to the onset of T1, in which case the attentional narrowing had already occurred before T1 was detected. However, an alternative hypothesis is that the narrowing only occurred once T1 had been detected. The next experiment was designed to address this question. Experiment 3.2 In Experiment 3.2 both RSVP streams contained digit distractors prior to the onset of T1, making it impossible for observers to narrow their attention in advance of the detection of T1. The attentional pull manipulation was therefore limited to the T1 frame in this experiment. Specifically, on half of the trials, the item which appeared simultaneously with T1 but in the opposite stream was a digit, while on the remaining half of the trials the item was a random-dot pattern (See Figure 3.5). When the item opposite T1 was a digit, the process of narrowing to T1 should be relatively slow (since a strong-pull digit must be disengaged from); when the item opposite T1 was a random-dot pattern, the narrowing process should be relatively rapid (since only a weak-pull random-dot pattern must be disengaged from). 80  Limiting the manipulation to just the T1-frame means that the attentional pull must come not from the items in the leading stream, but only from the item opposite from the first target. There are two reasons to think that the leading stream is not critical for the attentional pull manipulation. First, in Jefferies and Di Lollo’s (in press) study, the narrowing of focal attention appeared to be triggered by the appearance of the first target. Second, a recent study by Kawahara and Enns (2009) showed clearly that although the leading stream influences processing of the first target, it has no influence on the second target or on the magnitude of the attentional blink. We therefore expect that limiting the attentional pull manipulation to the T1frame will still be effective while circumventing the possible confounds of one stream matching the observers mental set and the other stream not.  Figure 3.5. Schematic representation of a trial in Experiment 3.2. The critical difference in this experiment is that both leading streams contain digit distractors and the attentional pull manipulation is implemented only in the T1-frame.  81  Item opposite T1 has weak attentional pull. Fast Shrink condition.  5  4 5  +  ...  8  +  +  D  R  . .. + .. ..  +  1  9  3 5  4 5  +  R  +  +  1  9  8  +  +  D  3  2 Item opposite T1 has strong attentional pull. Slow Shrink condition  . . .  Participants A total of 23 undergraduate students from the University of British Columbia participated in the experiment for course credit. Four participants were eliminated due to having T1 accuracy lower than 55%, leaving a total of 19 participants included for analysis. All participants were naive as to the purpose of the experiment and reported normal or corrected-to-normal vision. Procedure The procedure of Experiment 3.2 was identical to that in Experiment 3.1 with two exceptions. First, whereas in Experiment 3.1 there was one stream of digit distractors and one stream of random-dot patterns, in Experiment 3.2, both RSVP streams contained digit distractors. On a random half of the trials, the digit presented simultaneously with T1 in the opposite stream was replaced with a random-dot pattern (Fast-shrink trial); on the remaining trials the digit was not replaced (Slow-shrink trial). The second difference was that only a single SOA (118 ms) was tested in this experiment. The reason for this is simply that we expected to find the same  82  pattern of results as in Experiment 3.1, and this could be confirmed adequately with only a single SOA. The 118 ms SOA was chosen because it was at this SOA that the largest difference between the slow- and fast-shrink conditions occurred in Experiment 3.1.  Results and Discussion As in Experiment 3.1, only those trials in which T1 was identified correctly were included for analysis. The average first-target accuracy was 80% in the Fast-shrink condition and 79% in the Slow-shrink condition. The results of Experiment 3.2 are illustrated in Figure 3.6, below. Figure 3.6A shows the results for those trials in which T1 and T2 appear in different streams while Figure 3.6B shows the results for those trials in which T1 and T2 appear in the same stream. Only the former condition is of critical interest to the objective of the present research since Lag-1 sparing always occurs when T1 and T2 appear in the same spatial location, and hence it cannot be used to assess the spatial extent of focal attention.  Figure 3.6. Mean percentages of correct identifications of the second target in Experiment 3.2 with the data from the Same-stream and Different-streams conditions plotted in separate panels. The open symbols represent data from the Slow-shrink condition; the filled symbols represent data from the Fast-shrink condition.  83  Percentage correct (T2|T1)  100  Different  Same  80 60 40 20 [A] 1  Slow Shrink Fast Shrink  3  [B]  1 3 9 Inter-target Lag  9  As can be seen from a visual inspection of the graph, Lag-1 sparing is present in the Slow-shrink condition whereas Lag-1 deficit is present in the Fast-shrink condition. A 2 (Condition: Fast- and Slow-shrinking) x 3 (Lag: 1, 3, 9) within-subject ANOVA was conducted on the data illustrated in Figure 3.6. The analysis revealed a significant effect of Lag F (2, 35) = 37.85, p = .001, ηp2 = .717, and a significant Lag x Condition interaction F (2,34) = 7.83, p = .002, ηp2 =.384. The main effect of Condition was not significant F(1,18) < 1. In order to ascertain whether the difference between the Fast- and Slow-shrink conditions is comparable in Experiments 3.1 and 3.2, we plotted the data for Experiment 3.2 alongside the data for Experiment 3.1 (see Figure 3.7 below). As was the case in Experiment 3.1, the data from Experiment 3.2 are consistent with the hypothesis that the strength of the attentional pull influences the rate at which the focus of attention narrows. Specifically, the focus of attention is narrowed more rapidly to goal-relevant stimuli than to non-relevant stimuli.  84  Figure 3.7. Variation in the magnitude of Lag-1 sparing (positive values) and Lag-1 deficit (negative values) as a function of SOA. The filled triangle represents the Fast-shrink condition in Experiment 3.2 and the white triangle represents the Slow-shrink condition.  Experiment 3.3 The goal of Experiment 3.3 was to examine whether the rate of attentional narrowing is modulated by the visual orienting reflex. In Experiments 3.1 and 3.2 we examined the effect of stimuli being goal-relevant or non-relevant. To this end, we manipulated the low-level features of a stimulus to evoke visual orienting. We used target stimuli that were more salient (i.e., brighter and thus of higher contrast from the background) than the distractor stimuli. There are several reasons why brighter targets should induce a more rapid narrowing of the focus of 85  attention. First, reaction-time results show that bright stimuli are processed more rapidly than dim stimuli (Woodworth & Schlosberg, 1954). Second, it is known that in visual search tasks, bright targets are located and identified more rapidly than dim ones; they “pop-out,” drawing attention rapidly to their location (Theeuwes, 1995). Participants A total of 45 undergraduate students from the University of British Columbia participated in the experiment for course credit. The participants were randomly allocated to one of 3 SOA conditions: 53, 80, and 133. Five participants were eliminated due to having T1 accuracy lower than 55%, four in the 53 ms SOA condition, and one in the 80 ms condition. The final numbers of participants in each SOA group were 14, 16, and 15 in the 53, 80, and 133 ms conditions, respectively. All participants were naive as to the purpose of the experiment and reported normal or corrected-to-normal vision. Stimuli and Procedures In Experiment 3.3, two streams of digit distractors were presented, one to the left and one to the right of fixation. The stimuli and procedures in Experiment 3.3 were identical to those described in Jefferies and Di Lollo (in press; Experiment 2.1 in this thesis) with two exceptions. First, only three SOAs were tested, 53, 80, and 133 ms. Second, the two target letters were brighter (i.e., of higher luminance) than the distractor items. The distractor items had a luminance of 60 cd/m2 while the targets were 90 cd/m2. Note that in Experiment 3.1 and 3.2 the earliest tested SOA was 80 ms. This is because processing an item to the level of meaning takes time, and hence differences in attentional pull related to meaning processing should only be evident at intermediate or longer SOAs. In  86  Experiment 3.3, however, the exogenous influences of brightness should be rapidly processed in the visual system and hence we expect and probe for differences at shorter SOAs (53 and 80 ms). Results and Discussion As in Experiment 3.1, only those trials in which T1 was identified correctly were included for analysis. The average first-target accuracy was 74%, 78%, and 91% in the 53, 80, and 133 ms SOA conditions, respectively. The results of Experiment 3.3 are illustrated in Figure 3.8 below.  Figure 3.8. Mean percentages of correct identification of the second target in Experiment 3.3 with the data from each SOA plotted in a separate panel. The open symbols represent data from the Same stream condition; the filled symbols represent data from the Different streams condition.  The data were first analyzed in an overall 3 (SOA: 53, 80, 133) x 2 (Stream: Same, Different) x 3 (Lag: 1, 3, 9) within-subject ANOVA. The analysis revealed significant main  87  effects of Lag, F(2,84)=12.1, p < .001, ηp2 = .248, Stream, F(1,42)=4.27, p = .045, ηp2 = .093, and SOA, F(2,42)=41.44, p < .001, ηp2 = .664. There were also significant interactions between Stream and SOA, F(2,42)=9.38, p < .001, ηp2 = .028, and between Lag and Stream, F(2,84)=7.43, p < .001, ηp2 = .117. Since the factor of interest to the present study was whether the magnitude of Lag-1 sparing varies as a function of SOA, we performed a follow-up analysis in which only the data from Lags 1 and 3 in the Different-streams condition were analyzed. This 2 (Lags 1 and 3) x 3 (SOA) ANOVA revealed significant main effects of Lag, F(1,42)=15.1, p < .001, ηp2 = .264, and SOA, F(2,42)=23.01, p < .001, ηp2 = .524. There was also a significant interaction between Lag and SOA, F(2,42)=4.89, p = .012, ηp2 = .189. This interaction confirms that the magnitude of Lag-1 sparing changes as a function of SOA. Although these statistics indicate that the focus of attention narrows as a function of time, they do not address the primary objective of Experiment 3.3, which was to determine whether the rate at which the focus of attention narrows is reduced for bright targets. In order to determine this, the results of Experiment 3.3 must be compared to the results reported by Jefferies and Di Lollo (in press; Experiment 2.2 of this thesis). As a first step, the results from Experiment 3.3 will be plotted in the same graph as the data from Experiment 2.2 so that they can be directly compared (see Figure 3.9, below). In Figure 3.9, the magnitude of Lag-1 sparing and Lag-1 deficit are plotted as a single value so as to remove the effect of masking. The open symbols represent data from Experiment 3.3 in which the targets were bright; the filled symbols represent data from Experiment 2.1, in which the targets and the distractors were of equal luminance.  88  Figure 3.9. Graph showing the magnitude of Lag-1 sparing and Lag-1 deficit as a function of SOA. The open symbols represent data from Experiment 3.3 whereas the filled symbols represent data from Experiment 2.1 (Chapter 2, Jefferies & Di Lollo, in press).  Summary In order to efficiently extract information from the world, visual attention must be rapidly and flexibly deployed across space and time. An estimate of the rate at which the focus of attention can shrink from broad to narrow was provided by Jefferies & Di Lollo (in press; Chapter 2 of this thesis). The current research was designed to probe factors which might influence the rate at which the shrinking process occurs. 89  The present study examines the dynamics of attentional focus over time by monitoring the patterns of accuracy that occur when participants attempt to identify two targets embedded in simultaneously presented streams of items. The type of distractor item in the streams was manipulated to increase or decrease to extent to which the streams hold attention (referred to as attentional pull) and hence increase or decrease the rate at which the focus of attention narrows. The rate at which the narrowing process occurred was assessed by varying the stimulus onset asynchrony (SOA) between sequential items in the stream; a short SOA left relatively little time for the narrowing process to occur whereas a longer SOA left more time for narrowing. In Experiment 3.1, one stream contained distractors that shared categorical membership with the targets (i.e., they were both members of the alphanumerical character set). As such, those items should exert a strong attentional pull due to their goal-relevant nature. The other stream, in contrast, contained random-dot patterns which were not goal-relevant and therefore should exert a weaker attentional pull. We expected that when the first target appeared in the strong-pull stream, the process of narrowing would be very rapid (Fast-shrink condition); if the first target appeared in the weak-pull stream, in contrast, the narrowing process would be considerably slower. The pattern of results was consistent with our hypothesis. At the shortest SOA (80 ms), the magnitude of Lag-1 sparing was the same in both the Fast- and Slow-Shrink conditions. This suggests that an SOA of 80 ms, the focus of attention still encompassed both streams. At the two intermediate SOAs (100 and 118 ms), however, the magnitude of Lag-1 sparing was substantially greater for the Slow-shrink condition than for the Fast-shrink condition. This suggests that the focus of attention still encompassed both streams in the Slow-shrink condition, but had narrowed to the location of T1 in the Fast-shrink condition. At the longest SOA, there  90  was Lag-1 deficit in both conditions, indicating that sufficient time had elapsed for the focus of attention to narrow to the location of T1, regardless of which the narrowing process was rapid or slow. Experiment 3.2 reduced the featural differences between the two streams by employing digit distractors in both streams and limiting the attentional manipulation to the frame containing the first target. On half of the trial, the item presented at the same time as T1 but in the opposite stream was a digit (strong-pull item); on the remaining trials that item was a random-dot pattern (weak-pull item). Although only a single SOA (118 ms) was probed, the results overlapped almost precisely with the results of Experiment 3.1, indicating that : a) the effect of the attentional pull manipulated does not depend on the leading stream items but only on the item that appears simultaneously with T1, and b) the featural differences between the streams did not skew the results reported in Experiment 3.1. In Experiment 3.3, attentional pull was instantiated as the physical salience of the targets – in this experiment, the targets were brighter than the distractors. As expected, the rate at which the focus of attention narrows to physically salient targets was very rapid. In summary, the research presented in this paper shows clearly that the rate at which the focus of attention is narrowed can be predictably influenced by factors such as goal-relevance, on the one hand, and the visual orienting reflex, on the other. Future research might comparably examine whether internal factors such as mood, ageing, or even circadian rhythms influence the rate at which focal attention expands and contracts as each of these has implications for the efficient functioning of attention in our day-to-day lives.  91  References Bacon, W.F., & Egeth, H.E., (1994). Overriding stimulus-driven attentional capture. Perception & Psychophysics, 55 (5), 485-496 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, 21, 109-127. Eriksen, C.W., & St. James, J.D. (1986). Visual attention within and around the field of focal attention: A zoom lens model. Perception and Psychophysics, 40, 225-240. Eriksen, C.W., & Yeh, Y-Y. (1985). Allocation of Attention in the Visual Field. Journal of Experimental Psychology, 11, 583-597. Folk, C. L., Remington, R. W., & Johnston, J. C. (1992). Involuntary covert orienting is contingent on attentional control settings. Journal of Experimental Psychology: Human Perception and Performance, 18,1030–1044. Ghorashi, S.M., Zuvic, S.M., Visser, T.A., & Di Lollo, V. (2003). Focal distraction: Spatial shifts of attentional focus are not required for contingent capture. Journal of Experimental Psychology: Human Perception and Performance, 29, 78-91. Helmholtz, H. V. (1866). Handbuch der physiologischen Optik (1st ed.). Leipzig: Voss. (Trans. J. P. C. Southall (1962). Handbook of physiological optics (3rd ed.). New York: Dover). James, W. (1950). The Principles of Psychology, Vol. 1. New York: Dover. (Original work published in 1890). Jefferies, L.N., & Di Lollo, V. (in press). Linear changes in the spatial extent of the focus of  92  attention across time. Journal of Experimental Psychology: Human Perception and Performance. Jefferies, L.N., Ghorashi, S., Kawahara, J-i., & Di Lollo, V. (2007). Ignorance is bliss: the role of observer expectation in dynamic spatial tuning of the attentional focus. Perception & Psychophysics, 69, 1162-1174 . Kawahara, J-I. & Enns, J.T. (2009). Selection difficulty and inter-item competition are independent factors in rapid visual stream perception. Journal of Experimental Psychology: Human Perception and Performance, 35, 146-158. LaBerge, D. (1995). Attentional Processing: The Brain’s Art of Mindfulness. Cambridge, MA: Harvard University Press. Potter, M. C., Chun, M. M., Banks, B. S., & Muckenhoupt, M. (1998). Two attentional deficits in serial target search: The visual attentional blink and an amodal taskswitch deficit. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 979-992 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, 18, 849-860. Theeuwes, J. (1995). Abrupt luminance change pops out; abrupt color change does not. Perception & Psychophysics, 57 (5), 637-644 Woodworth, R.S., & Shlosberg, H. (1954). Experimental Psychology. New York: Holt Publishers  93  Chapter 4 Assessing the Rate at which Focal Attention arrows in Older Adults  3  The visual world contains such a rich array of information that there is need for a selective mechanism to limit the information entering the visual system at any given moment. Selective attention serves this function by guiding visual processing to the most relevant information in a scene. In order to enable efficient processing, the focus of attention must be shifted from one location to another and shrunk or expanded in size in order to accommodate a smaller or larger portion of the visual field (e.g., Eriksen & Yeh, 1985; Eriksen & St. James, 1986). Considerable research has been dedicated to examining the time-course of spatial shifts of attention, often employing a paradigm known as spatial cueing (Posner & Cohen, 1984). In a spatial cueing paradigm, observers are instructed to maintain fixation on a central point. After a variable temporal gap, a target appears at a peripheral location, and the observer makes a speeded response to its onset. Prior to the onset of the target, a spatial cue is presented at a peripheral location. If the cue is presented at the location at which the target will subsequently appear, it is referred to as a valid cue; a cue which appears at a different location from the target is referred to as an invalid cue. In general, reaction times (RTs) to the onset of the target are shorter when the cue is valid than when it is invalid. The interpretation of these findings is that attention is drawn rapidly and automatically to the location of the spatial cue, resulting in a RT benefit if the cue is valid. The type of cueing described above is known as direct or peripheral cueing. It has been suggested that attentional shifts that occur in response to peripheral cues are reflexive and 3  A version of this chapter will be submitted for publication. Jefferies, L.N, Roggeveen, A., Enns, J.T., Bennett, P., Sekuler, A., & Di Lollo, V. The creaky attentional gate: Assessing the rate at which focal attention narrows in older adults.  94  automatic in nature (e.g., Jonides, 1983). Reaction time costs and benefits to peripheral cues are evident at very short cue-target-onset-asynchronies (CTOA; e.g., 300 ms), highlighting the rapid nature of these reflexive shifts of attention. A second form of cueing is known as central or symbolic cueing. Symbolic cues always appear at fixation and indicate the location at which the target will appear without themselves appearing at that location. Symbolic cues are typically objects such as arrows which either point to the location at which the target will appear (a valid central cue) or to an incorrect location (an invalid central cue). Symbolic cues must necessarily be processed for meaning and lead to voluntary, controlled shifts of attention. The costs and benefits of symbolic cueing are seen at relatively long CTOAs (e.g., around 1000 ms), suggesting that voluntary attention shifts are slower than automatic ones. If the process of shifting attention or adjusting the spatial extent of its focus is slowed, the needed information will not be extracted from the scene, and visual processing will be inefficient. Considerable research has examined whether the rate at which attention can be shifted is impaired in older adults. The answer as to whether aging negatively impacts attentional shifts is somewhat unclear, however, and depends on the nature of the attention shift. If peripheral cues are used to initiate reflexive shifts of attention, young and older adults perform in a comparable manner. That is, although RTs are slower overall for older adults, the magnitude and time-course of the cueing benefit is comparable, indicating that reflexive shifts of spatial attention do not change with age (e.g., Folk & Hoyer, 1992; Greenwood, Parsuraman, & Haxby, 1993; Lorenzo-Lopez et al., 2002; Tales et al., 2002). The evidence regarding voluntary shifts of attention, on the other hand, is contradictory. Some research shows that aging slows the rate at which voluntary shifts of attention are completed (Greenwood, Parasuraman, & Haxby,  95  1993) while other research suggests that the rate is unchanged (Lorenzo-Lopez et al., 2002; Nissen & Corkin, 1985; Robinson & Kertzman, 1990; Tales et al., 2002). It has been suggested that these contradictory results may stem from the fact that the rate at which information is extracted from a symbolic cue increases with age. Symbolic cues must be interpreted (although they can require more or less interpretation depending on the specifics of the stimulus), and this need to extract information from the cue is the underlying cause of the voluntary attention shift deficit in aging (Folk & Hoyer, 1992). To reiterate, voluntary shifts of attention are impaired only to the extent that older adults have difficulty extracting the information from the cue. The findings discussed above deal exclusively with the effect of aging on the ability to shift the focus of attention from one location to another. In addition to being redeployed, however, the focus of attention can also be adjusted in terms of its spatial extent – that is, it can be shrunk or expanded in size to encompass a larger or smaller region of space. There are estimates in the literature as to the rate at which young adults can shrink or expand the focus of attention (Benso et al., 2002; Jefferies & Di Lollo, in press), but it has yet to be determined whether this component of spatial attention is influenced by the aging process. There is some neurological evidence to suggest that the rate at which focal attention can be shifted or adjusted in spatial extent will slow as a function of aging. It has been established that both frontal and posterior parietal regions of the brain are intimately involved in initiating and governing shifts of attention (Petersen et al., 1991; Posner, 1980; Yantis et al., 2002). It is also known that these same regions are strongly affected by the aging process – frontal regions of the brain exhibit the most marked cell loss with age (e.g., Shefer, 1973) while posterior parietal regions of the brain evidence the greatest age-related decrease in cerebral blood flow (Martin, Friston, Colebatch & Frackowiak, 1991). Both of these findings suggest that the rate at which  96  spatial attention can be adjusted may decline with age. The goal of the present research is, therefore, to determine whether the rate at which focal attention can be shrunk or expanded is reduced as a function of aging, or whether it remains unaffected in the way that reflexive attention shifts are unaffected. In determining whether older adults are slower than young adults to shrink or expand their focus of attention, we employed the methodology used by Jefferies and Di Lollo (in press). In that study, Jefferies and Di Lollo made use of two well-known phenomena to determine the rate at which focal attention shrinks: the attentional blink and Lag-1 sparing. In a typical attentional blink (AB) paradigm, two target letters are presented in a rapid serial visual presentation (RSVP) stream of digit distractors. Although the first target can be reported correctly, identification of the second target is impaired if it appears approximately 100 – 500 ms after the first target (Raymond, Shapiro, & Arnell, 1992). Jefferies and Di Lollo employed a dual-stream paradigm in which two streams of distractor digits were presented, one to the left and one to the right of fixation. The two letter targets appeared unpredictably in either the leftor the right-hand stream and in either the same stream as one another or in opposite streams. Jefferies and Di Lollo (in press) also utilized an aspect of the attentional blink known as Lag-1 sparing. It has been found that the magnitude of the attentional blink is much reduced if the second target is presented directly after the first (in the temporal position known as Lag 1) without any intervening distractors, provided that the inter-target stimulus-onset-asynchrony (SOA) is about 100 ms or longer (see Bowman & Wyble, 2007; Chun & Potter, 1995; Giesbrecht & Di Lollo, 1998; Maki, Couture, Frigen, & Lien, 1997; Raymond et al., 1992). Although Lag-1 sparing typically occurs only if the two targets are presented in the same spatial location (Visser, Bischof, & Di Lollo, 1999), it has recently been reported that Lag-1 sparing can occur to targets  97  that appear in different spatial locations provided that the second target falls within the focus of attention (Jefferies, Ghorashi, Kawahara & Di Lollo, 2007). Given this, the incidence and magnitude of Lag-1 sparing can be used to assess the rate at which the focus of attention narrows, as described below.  Figure 4.1. Schematic illustration of the progressive changes in the spatial extent of the focus of attention (segmented rectangles) as a function of SOA and Lag. See text for a detailed description.  In a dual-stream paradigm, focal attention is assumed to initially encompass both streams but to shrink rapidly to the stream in which the first target appears. A consequence of this narrowing to the first-target stream is that the focus of attention withdraws from the opposite stream. If the second target subsequently appears in that stream, whether or not it falls within the focus of attention will be depend on the stimulus-onset-asynchrony (SOA) between the two targets. If the SOA is very short, there will not have been sufficient time for focal attention to narrow to the first-target stream and withdraw from the opposite stream. The second target will 98  therefore still fall within the focus of attention and Lag-1 sparing will occur. If, on the other hand, there has been sufficient time for the focus of attention to narrow completely on the firsttarget stream, the second target will no longer appear within the focus of attention and Lag-1 deficit (the converse of Lag-1 sparing) will occur (see Figure 4.1). Jefferies and Di Lollo (in press) systematically varied the SOA between successive items in the RSVP stream, testing six different SOAs: 53, 66, 80, 100, 118, and 133 ms. They found a gradual, linear transition from Lag-1 sparing at the shortest SOAs to Lag-1 deficit at the longest SOAs, which is consistent with an analog narrowing of the focus of attention over time. Our goal in the present research is to determine whether the rate at which the focus of attention shrinks (as indexed by the change from Lag-1 sparing to Lag-1 deficit as a function of SOA) changes with age. To this end, we employed the same methodology as Jefferies and Di Lollo (in press) and tested both young and older adults. Experiment 4.1 Method Observers Seventeen young adults (ages 18-27) and 21 older adults (ages 60-74), all with normal or corrected-to-normal vision and naïve as to the purpose of the study, participated in the experiment. The young adults were recruited from the undergraduate population of McMaster University and participated for course credit. The participants for the group of older adults were recruited by newspaper advertisement from the Hamilton, Ontario area and were paid $10 per hour for their participation. All subjects completed visual and general health questionnaires to screen for visual pathology, such as cataract, macular degeneration, amblyopia, etc. Prior to participation in the study, near and far decimal logMAR (logarithm of the minimum angle of 99  resolution) acuities were measured for all subjects with CSV-1000EDTRS eye charts (Precision Vision, LaSalle, Illinois, USA). When measuring visual acuity, subjects wore their normal optical correction for each distance. Older participants were also screened for dementia and scored within the normal range of the Mini Mental Status Examination (Folstein, Folstein, & McHugh, 1975) Apparatus and Stimuli Observers were seated in a dark room approximately 57 cm from a computer monitor, with a small lamp illuminating the computer keyboard. The luminance of all stimuli was 34.3 cd/m2, and the luminance of the black background was 2.3 cd/m2. A white fixation cross (0.25˚ by 0.25˚) was displayed in the center of the screen for the duration of each trial. The stimuli consisted of white digits (0 - 9) and capital letters (excluding the letters I, O, Q, and Z), each subtending approximately 0.9˚ vertically. The screen refresh rate was varied depending on the temporal interval that elapsed between successive items in the display sequence (stimulus-onset asynchrony, SOA; see below). To obtain the three SOAs of 66, 100, and 133 ms, the screen refresh rate was set at 75, 60, and 75 Hz, respectively. Procedure The observers initiated each new trial by pressing the spacebar. At the beginning of the trial, two synchronized rapid serial visual presentation (RSVP) streams of items were presented, 1.75° to the left and right of fixation. Each stream contained 8-14 leading distractor digits prior to the onset of the targets. The digits were chosen randomly for each frame with the restriction that the same digit could never be displayed concurrently in the two streams and that each digit could not be the same as the preceding two digits in that stream. Two different letter targets were presented on each trial. The two targets appeared randomly but with equal probability in 100  either the left or the right stream and could appear in either the same stream or in different streams. The RSVP streams ended with one digit-distractor frame presented directly after the second-target frame. The display sequence on any given trial is illustrated schematically in Figure 4.2. The observers’ task was to identify the two target letters by entering them in the keyboard in either order at the end of the trial.  Figure 4.2. Schematic representation of the sequence of events within a trial in Experiment 4.1. The first and the second targets (T1 and T2) could appear in either the left or the right RSVP stream and in either the same or opposite streams.  T1 & T2 in same stream  8  3  8  3 H  +  H + 5 5 + 7 + 1 7 + 1 3 + 9 3 + 9 R + 4 R + 4 7 + 2 7 + 2 6 + 3 6 + 3  (T2)  (T1)  +  T1 & T2 in different streams  The SOA between successive items in the RSVP stream was 66, 100, or 133 ms. The three SOAs were presented in separate blocks of trials, presented in random order to the participants. In every case, the SOA consisted of two parts: first, the item itself (whether distractor or target) was displayed for approximately two-thirds of the SOA. Second, a blank 101  inter-stimulus interval (ISI) was inserted for the remaining one-third of the SOA. The actual duration of the two parts depended on the SOA, as follows. For SOAs of 66, 100, and 133 ms, the ratios of exposure duration to ISI were approximately 40:26, 70:30, and 80:53 ms respectively. The second target was presented at one of three inter-target lags: 1, 3, or 9. At Lag 1, the second target was presented in the RSVP frame directly following the first target. At Lag 3, two distractor frames were inserted between the two targets. At Lag 9, eight distractor frames were inserted between the two targets. The three inter-target lags occurred in random order and with equal frequency across trials. Results Only those trials in which the first target was identified correctly were included for analysis. This procedure is commonly adopted in AB experiments on the grounds that, on trials in which the first target is identified incorrectly, the source of the error is unknown, and thus its effect on second-target processing cannot be estimated. Figure 4.3 illustrates the percentage of correct second-target responses as a function of Same-stream/Different-stream conditions and Lag, separately for each SOA for both young and older participants.  102  Figure 4.3. Mean percentages of correct identifications of the second target in Experiment 4.1. Data is plotted separately for the Young and Older adults at each SOA. The filled circles represent data from trials in which the targets were presented in the same stream; empty circles represent data from trials in which the targets were presented in different streams.  100  Older  Young  80  6 6 m s S O A  60 40  ) 20 1 T | 100 2 T (t c 80 e rr o 60 c e g 40 a t n e cr 20 e P100  1 0 0 m s S O A  80  1 3 3 m s S O A  60 40 Same stream Different streams  20 1  3  9  1  3  9  Inter-target Lag  The data were first analyzed in an overall 2 x 2 x 3 x 3 ANOVA consisting of three within-subject factors and one between-subjects factor. The within-subject factors were Lag (1, 3, and 9), Stream (Same and Different), and SOA (66, 100, and 133). The between-subjects 103  factor Group (Young adults, Older adults). The analysis revealed significant main effects of Lag, F(2,72) = 34.78, p < .001, ηp2 = .491, Stream, F(1,36) = 70.01, p < .001, ηp2 = .661, Group, F(1,36) = 52.58, p < .001, ηp2 =.594, and SOA, F(2,72) = 94.2, p < .001, ηp2 = .724. The following interactions were also significant: Stream x Group, F(1,36) = 23.03, p < .001, ηp2 = .39, Lag x Group, F(2,72) = 9.52, p < .001, ηp2 = .209, Stream x Lag, F(2,72) = 7.21, p < .001, ηp2 = .543, SOA x Lag, F(4,144) = 7.22, p < .001, ηp2 = .167, SOA x Lag x Group, F(4,144) = 10.24, p < .001, ηp2 = .21, SOA x Group x Lag, F(4,144) = 3.15, p < .049, ηp2 = .222, and SOA x Stream x Lag, F(4,144) = 11.37, p < .001, ηp2 =.240. The interpretation of these interactions was constrained, however, by the significant four-way interaction among Lag, Stream, Group, and SOA, F(4, 144) = 6.78, p < .001, ηp2 = .158. The significance of this overall interaction in conjunction with the graphical evidence in Figure 4.3 clearly indicates that the magnitude of Lag-1 sparing and Lag-1 deficit varied as a function of age group and SOA, justifying further statistical examination. In order to determine whether the pattern of responses differs for the Young and Older adults at each SOA, three separate 2 (Group: Young, Older) x 2 (Stream: Same, Different), x 3 (Lag: 1, 3, 9) ANOVAs were conducted. For an SOA of 66 ms, that ANOVA revealed a significant main effect of Stream, F(1,36) = 23.77, p < .001, ηp2 = .398, and Group, F(1,36) = 34.5, p < .001, ηp2 = .489. The interaction between Stream and Group was also significant, F(1,36) = 8.16, p = .006, ηp2 = .191, and the interaction between Stream and Lag approached significance, F(2,72) =2.67, p = .07, ηp2 = .069. Critically, the three-way interaction between Stream, Lag, and Group was not significant, F(1,36) =2.21, p = .116. This overall pattern of significance indicates that although the overall level of performance was lower for the Older than for the Young adults, the pattern of responding did not differ. That is, with an SOA of 66 ms, 104  there was no difference in the magnitude of Lag-1 sparing or Lag-1 deficit for the young and the older adults. For an SOA of 100 ms, that same 2 (Group: Young, Older) x 2 (Stream: Same, Different), x 3 (Lag: 1, 3, 9) ANOVA revealed significant main effects of Stream, F(1,36) =44.39, p = .001, ηp2 = .552,, Lag, F(2,72) =27.81, p = .001, ηp2 = .436, and Group, F(1,36) =43.98, p < .001, ηp2 = .55. There were also three significant two-way interactions: Stream x Group, F(1,36) =20.13, p < .001, ηp2 = .359, Lag x Group, F(2,72) =12.45, p < .001, ηp2 = .257, and Stream x Lag, F(2,72) =31.36, p < .001, ηp2 = .466. Most critically, the interaction between Stream, Lag, and Group was significant, F(2,72) =4.11, p = .02, ηp2 = .103. This pattern of significant shows that the overall level of performance was higher for the young than for the older adults. The significant three-way interaction however, also indicates that the overall pattern of responding differs. This difference is most likely driven by the differences at Lags 1 and 3, at which the Older adults show Lag-1 sparing in both the Same- and Different-stream conditions while the Young adults show Lag-1 sparing in the Same-stream condition, but Lag-1 deficit in the Different-stream condition. This interpretation was confirmed by a follow-up ANOVA testing only the data from Lags 1 and 3 in the Different-streams condition for the Young and the Older adults at 100 ms. This 2 (Lag: 1, 3) x 2 (Group: Young, Older adults) ANOVA revealed a main effect of Group, F(1,36) =64.8, p < .001, ηp2 = .643, and a significant interaction between Lag and Group, F(1,36) =5.5, p = .025, ηp2 = .133. The significant interaction confirms that the magnitude of Lag-1 sparing differed for the Young and Older observers. Finally, the results were compared when the SOA was 133 ms in a 2 (Group: Young, Older adults) x 2 (Stream: Same, Different), x 3 (Lag: 1, 3, 9) ANOVA. There were significant 105  main effects of Stream, F(1,36) = 35.39, p < .001, ηp2 = .496, Lag, F(2,72) = 58.19, p < .001, ηp2 = .618, and Group, F(1,36) = 46.56, p < .001, ηp2 = .564. There were also three significant twoway interactions: Stream x Group, F(1,36) =7.49 , p = .01, ηp2 = .172, Lag x Group, F(2,72) = 32.37, p < .001, ηp2 = .473, and Stream x Lag, F(2,72) = 47.79, p < .001, ηp2 = .570. Finally, the three-way interaction between Stream, Lag, and Group was also significant, F(2,72) = 11.18, p < .001, ηp2 = .237. As with the 100 ms SOA condition, we followed this analysis with an ANOVA intended to probe whether the difference in the magnitude of Lag-1 sparing/Lag-1 deficit between the Young and the Older adults with an SOA of 133 ms is significant. In doing this, only the data from Lags 1 and 3 were included. The 2 (Lag: 1, 3) x 2 (Group: Young, Older adults) ANOVA revealed a main effect of Lag, F(1,36) =4.6, p < .04, ηp2 = .113, and a significant interaction between Lag and Group, F(1,36) = 11.75, p = .002, ηp2 = .246, confirming that the magnitude of Lag-1 sparing differed for the young and older adults. Two final ANOVAs were conducted to confirm that the magnitude of Lag-1 sparing changes across SOA for the young observers but not for the older observers. Since these analyses were intended to specifically probe whether Lag-1 sparing changes to Lag-1 deficit as a function of increasing SOA, only the data from Lags 1 and 3 in the different-streams condition were included for analysis. For the young adults, this 2 (Lag: 1, 3) x 3 (SOA: 66, 100, 133) ANOVA showed a significant main effect of SOA, F(2,32) =15.07, p < .001, ηp2 = .485, and Lag, F(1,16) =10.62, p = .005, ηp2 = .399. Critically, the interaction between SOA and Lag was significant, F(2,32) =8.22, p = .001, ηp2 = .340. This confirms the graphical evidence that Lag-1 sparing becomes Lag-1 deficit as a function of increasing SOA for Young adult observers. The interpretation of these findings, which is consistent with the findings reported by Jefferies and Di 106  Lollo (in press), is that when the first target appears, the focus of attention begins to rapidly narrow to that stream, resulting in Lag-1 deficit for the second target if it appears in the stream opposite that containing the first target. For young adults, this narrowing process is clear by 100 ms after the presentation of the first target. The same 2 (Lag: 1, 3) x 3 (SOA: 66, 100, 133) ANOVA was conducted on the data from the older adults and revealed no significant main effects and, most critically, the interaction between SOA and Lag was not significant, F(2,40) < 1. This confirms the graphical evidence that Lag-1 sparing doesn’t change across the tested SOAs for the Older adults. The interpretation of these findings is that the Older adults have not yet begun to adjust the spatial extent of their focus of attention within the first 133 ms after the appearance of the first target and their focus of attention is of constant width over the three SOAs. Discussion We now consider the theoretical implications of the results. Visual inspection of the graphs in Figure 4.3 reveals several important differences between the young and the older adults. First, the overall level of performance is higher for the young than for the older adults, at all SOAs. One notable exception to this is performance at Lag 1 in the Same-stream condition, a difference which is discussed in greater detail below. A second clear difference between the results of the young and the older adults is that the overall magnitude of the attentional blink is greater for the older adults. This is consistent with previous research showing that the magnitude of the attentional blink increases with age (e.g., Lahar, Isaak, & McArthur, 2001; Maciokas & Crognale, 2003). As noted above, identification accuracy at Lag 1 in the Same-stream condition at SOAs of 100 and 133 ms is atypically high for the older adults – it is comparable to the accuracy of the 107  young observers. In fact, accuracy is even higher at Lag 1 than at Lag 9, which is typically outside the period of the attentional blink. One key to understanding this atypically high accuracy comes from considering the result for T1. As can be seen in Table 4.1, the older adults are relatively poor at identifying T1 at Lag 1 in the Same-stream condition, the same condition in which they identify T2 so accurately. This poor T1 accuracy suggests that somehow the representation of T2 is replacing that of T1 and is therefore being identified accurately to the detriment of T1. One possible explanation for why T2 might be identified accurately at the expense of T1 depends on a non-metacontrast masking effect. It is known that when a brief target is followed directly by a brief pattern-mask (ISI = 0), the two displays become temporally integrated. Under these conditions, it is often possible to see the contours of the target through the contours of the mask (Coltheart & Arthur, 1972). Although it seems intuitive that increasing the duration of the blank ISI should only make it easier to identify the target, this is in fact not the case. In fact, as the blank ISI between the target and the mask is increased, it becomes more and more difficult to identify the target. This trend continues until the ISI is around 50 ms. Beyond that point, identification of the target becomes easier (Bachmann & Allik, 1976; Purcell & Stewart, 1970). The impression of an observer in this case is that at the optimal ISI (i.e., around 50 ms blank ISI) the target appears to morph into the mask (hereafter simply referred to as morphing). In an RSVP paradigm such as is utilized here, at Lag 1 T2 serves as a mask for T1. Since both the target (T1) and the mask (T2) are of short duration, if the ISI is also short, T1 will appear to morph into T2. An ISI of around 50 ms leads to the strongest perception of morphing for young adults, and this can be clearly seen in the reduced level of performance by the young adults when the  108  SOA is 66 ms. In addition, when the SOA is 66 ms, T1 identification accuracy is substantially impaired when compared to accuracy at 100 or 133 ms. Although an ISI of approximately 50 ms leads to the strongest perception of morphing for young adults, this interval is most likely longer for older adults, and the effect of morphing can still be seen when the SOA is 100 and 133 ms. At these longer SOAs at Lag-1, identification accuracy of T2 is still very high while T1 is relatively low. Since the perception of T1 is impaired by its perceived morphing into T2 at appropriate SOAs, one might ask why the perception of T2 is not equally clouded by its morphing into the following digit distractor. It is known that the strength of masking increases as a function of not only the structural similarity (Fehrer, 1966; Harmon & Julesz, 1973), but also the conceptual similarity between the target and the mask (Dux & Coltheart, 2005; Enns, 2004; Intraub, 1981, 1984). Hence, a letter will be masked by another letter more than by a digit. Similarly, a letter should be more readily perceived as morphing into another letter than into a digit. This may explain why although T1 appears to morph into T2, T2 does not appear to morph as readily into the following digit distractor.  109  Table 4.1. Table listing first-target identification accuracy for Young and Older adults as a function of Condition (Same-stream, Different-stream), Lag, and SOA.  Lag 3  Lag 9  66 ms SOA, Same 66 ms SOA, Different 100 ms SOA, Same 100 ms SOA, Different 133 ms SOA, Same 133 ms SOA, Different  Lag 1 Young Adults 54.4 73.2 85.0 90.0 87.1 90.9  77.9 79.7 95.9 92.6 91.5 91.2  75.6 74.11 92.9 90.6 92.6 90.3  66 ms SOA, Same 66 ms SOA, Different 100 ms SOA, Same 100 ms SOA, Different 133 ms SOA, Same 133 ms SOA, Different  Older Adults 38.0 50.4 53.5 68.3 57.5 73.9  55.7 58.5 77.8 79.8 78.9 76.6  58.9 63.3 76.3 73.3 77.2 74.6  The Rate of arrowing Focal Attention In the current research, the magnitude of Lag-1 sparing or Lag-1 deficit is being used to assess the rate at which the focus of attention narrows in young and older adults. Since the young and older adults clearly differ in their overall level of performance, such estimates must be calculated independently of overall level of performance to enable a direct comparison. To this end, we calculated the difference between second-target identification accuracy at Lag 1 and Lag 3 to arrive at a single value. If positive, this value indexes the magnitude of Lag-1 sparing; if  110  negative, it indexes the magnitude of Lag-1 deficit. Those values are plotted in a single graph for both young and older adults as a function of SOAs in Figure 4.4, below. The filled circles represent data from the young observers while the empty circles represent data from the older adults. It is clear that there is a monotonic progression from Lag-1 sparing to Lag-1 deficit across SOAs for the young observers. These results closely match those reported by Jefferies and Di Lollo (in press), and provide a replication and confirmation of their findings. Quite a different pattern is evident for the older adults – namely, there is very little change in the magnitude of Lag-1 sparing across SOAs of 66, 100, and 133 ms. This strongly suggests that very little if any adjustment has been made in the spatial extent of focal attention up to 133 ms after the presentation of the first target.  Figure 4.4. Variation in the magnitude of Lag-1 sparing (positive values) and Lag-1 deficit (negative values) as a function of SOA. The filled symbols represent the data from the young adults; the open symbols represent the data from the older adults.  111  -10 -20 -30 -40  Older adults Young adults  66  100  Narrow  Lag-1 Sparing expressed as a proportion 100 - [(Lag3/Lag1)*100]  0  133  Inferred width of attentional window  Broad  10  266  SOA (ms)  One interpretation of this finding is that the older adults experience a delay before initiating the process of narrowing to the location of the first target. Once the process of narrowing is initiated, however, it takes place at approximately the same rate as for young observers. This interpretation meshes neatly with the endogenous cueing results reported by Folk and Hoyer (1992), who found that older adults were slower at extracting meaning information from a central cue. Once the meaning was extracted, however, the process of shifting attention was unimpaired. Comparably in the present experiment, older adults may have been slower to extract meaning from the items presented in the RSVP stream and as such are slower to identify the target letter as distinct from the digit distractors. This slower identification 112  would delay the trigger to narrow the focus of attention to the location of the first target, and would produce results comparable to those shown in Figure 4.4. This conjecture cannot be entirely supported from the results of Experiment 4.1, however, because at the SOAs tested the process of narrowing had not yet begun. In order to confirm that the older adults are delayed only in initiating the narrowing process and that they are in fact able to narrow to the location of the first target, a longer SOA must also be tested. Experiment 4.2 Experiment 4.1 tested Young and Older adults at three SOAs: 66, 100, and 133 ms. The results illustrated in Figure 4.4 show clearly that when the SOA was 133 ms, the Older adults had not yet initiated the process of narrowing their focus of attention. In order to confirm that the older adults are in fact able to narrow their focus of attention in this paradigm and that the process is simply delayed, Experiment 4.2 employed a much longer SOA of 266 ms. We expected that given this additional 133 ms in which to narrow to the location of the first target, the older adults would be able to complete the narrowing process, and Lag-1 deficit should be observed when the second target then appeared in the stream opposite that containing the first target. Observers Twenty older adults (ages 60-74) participated in the experiment. The older adults were recruited and screened as in Experiment 4.1. Stimuli and Procedure The stimuli and procedure were identical to those of Experiment 4.1 with two exceptions. First, only a single SOA of 266 ms was tested. Second, we did not use a young adult comparison group because young observers would be at ceiling at all lags with an SOA of 266 ms. 113  Results and Discussion As in Experiment 4.1, only those trials in which the first target was identified correctly were included for analysis. Figure 4.5 illustrates the percentage of correct second-target responses as a function of Same-stream/Different-stream conditions and Lag.  Figure 4.5. Mean percentages of correct identifications of the second target in Experiment 4.2. The filled circles represent data from trials in which the targets were presented in the same stream; empty circles represent data from trials in which the targets were presented in different streams.  A 2 (Condition: Same-stream, Different-Streams) x 3 (Lag: 1, 3, 9) within-subject ANOVA was conducted on the data illustrated in Figure 4.5. The analysis revealed significant main effects of Lag, F(2,38) = 45.34, p < .001, ηp2 = .705, and Condition, F(2,38) = 46.88, p < .001, ηp2 = .712. The interaction between Lag and Condition was also significant, F(2,38) = 40.0, p < .001, ηp2 = .678. As can be seen graphically in Figure 4.5 and Figure 4.4, and 114  confirmed by these analyses, there is a substantial Lag-1 deficit in the Different-streams condition. This is consistent with the interpretation that with this longer SOA, the older adults have narrowed their focus of attention to the stream containing the first target, leaving the second-target unattended when it appears in the opposite stream, with consequent Lag-1 deficit. Experiment 4.3 The process of aging affects the peripheral sensory organs as much, or perhaps even more, than it does the brain. It is well known, for example, that aging almost invariably leads to a clouding of the vitreous humour, which reduces the amount of light information transmitted to the retina. In addition, aging reduces the dilation capability of the pupil, which consequently reduces the retinal illuminance. In fact, it has been estimated that there is a three-fold reduction (i.e., approximately a 0.5 log reduction) in retinal illuminance from age 20 to age 60 (Weale, 1961, 1963). Of this 0.5 log unit loss, it has been estimated that increased opacity of the lens and the clouding of the vitreous humour accounts for approximately 0.2 log units and the reduced dilation capability of the pupil for the 0.3 log unit balance (Elliott, Whitaker, & MacVeigh, 1990). Given this drastic reduction in the amount of information reaching the retina, it can be difficult to dissociate whether performance and cognitive changes as a function of age are due to these changes in the sensory organs or to changes in brain function. In an effort to ensure that the effects we report in Experiments 1 and 2 are not due solely to changes in the information entering the visual system through aging sensory organs, Experiment 4.3 is a control experiment in which the computer monitor was covered with a neutral density filter to reduce the retinal illuminance in the young adults by 0.5 log units, thereby mimicking the effect of ageing on the eyes. If once these filters are in place, the pattern of results yielded by young adults is more 115  similar to the young adult results in Experiment 4.1 than to the results of the older adults in that same experiment, we can conclude that changes in the eye cannot be the cause of the differences observed between young and older adults in narrowing focal attention. Observers A new group of 17 undergraduate students from McMaster University participated in Experiment 4.3 for course credit. Stimuli and Procedure The stimuli and procedure were identical to those of Experiment 4.1 except that neutral density filters were placed in front of the monitors. These filters reduced the luminance of the display by 0.5 log units, but left the contrast of the display intact.  Figure 4.6. Mean percentages of correct identifications of the second target. The data plotted in the left-hand column is from Experiment 4.3 in which a filter was used to reduce the luminance of the displays and equate the young and the older adults in terms of the information entering the eye from the stimuli. The data plotted in the right-hand column are the data from Experiment 4.1. The filled circles represent data from trials in which the targets were presented in the same stream; empty circles represent data from trials in which the targets were presented in different streams.  116  Results and Discussion As in Experiment 4.1, only those trials in which the first target was identified correctly were included for analysis. Average first-target accuracy was 74.5% collapsed across Lag and SOA. Figure 4.6 illustrates the percentage of correct second-target responses as a function of Same-stream/Different-stream conditions and Lag. In order to ascertain whether the retinal illuminance differences caused by the neutral density filter changed the pattern of results, we compared the data in this experiment (left-hand 117  panel in Figure 4.6) to the data for the young adults in Experiment 4.1 (right-hand panel in Figure 4.6). The data were analyzed in a 2 (Experiment: filter, no-filter) x 3 (SOA: 66, 100, 133) x 2 (Stream: Same, Different) x 3 (Lag: 1, 3, 9) ANOVA. There was one between-subjects factor, Experiment, and three within-subjects factors. The analysis revealed significant main effects of SOA, F(2,64)=186.38, p<.001, ηp2 = .853, Stream, F(1,32)=31.27, p<.001, ηp2 = .494, and Lag, F(2,64)=38.16, p<.001, ηp2 = .544. The main effect of Experiment (filter, no-filter) was not significant, F(1,32) < 1. There were significant two-way interactions between SOA and Lag, F(4,128)=4.64, p=.004, ηp2 = .127, SOA and Stream, F(2,64)=5.09, p=.009, and Lag and Stream, F(2,64)= 33.12, p<.001, ηp2 = .51. There was also a significant three-way interaction among SOA, Stream, and Lag, F(4,128)=9.31, p<.001. Notably none of the interactions with Experiment (filter, no-filter) were significant: Experiment x SOA, F(2,64)<1, Experiment x Lag, F(2,64)<1, Experiment x Stream, F(1,32)< 1, Experiment x SOA x Lag, F(4,128)<1, Experiment x SOA x Stream, F(2,64)<1, Experiment x Lag x Stream, F(2,64)<1, and Experiment x SOA x Stream x Lag, F(4,128)<1. The lack of a significant main effect of Experiment (filter, no-filter) as well as any significant interactions with Experiment indicated that whether or not a filter was employed had no effect on the pattern of results. That is, placing a filter over the computer monitor to reduce the limit and mimic the effects of age on the sensory organs did not influence the pattern of results. It seems clear, therefore, that the difference between the young adults and the older adults in Experiment 4.1 was due to changes in attentional processing that occur as a function of age, and not due to changes in the external sensory organs.  118  Conclusions For efficient visual functioning, the focus of attention must be shifted rapidly from one location to another or be expanded or contracted to encompass a larger or smaller region of the visual field. Previous research has investigated whether the efficiency of attentional shifts is impaired with age. The objective of the present research was to determine whether the ageing process influences the rate at which the focus of attention can be narrowed. In addressing this question, we employed a dual-stream attentional blink paradigm to monitor changes in the spatial extent of focal attention. Specifically, we built on the fact that Lag-1 sparing occurs only if the second target falls within the focus of attention. As such, the incidence and magnitude of Lag-1 sparing can be used to index the width of focal attention as follows. We assume that since which stream will contain the first target is unpredictable, the observers will initially employ a broad focus of attention that encompasses both streams. Once the first target appears, however, the focus of attention will narrow to the location of the first target. If the second target then appears in the opposite stream, whether or not it will fall within the focus of attention depends on the SOA between successive items in the stream. If the SOA is short, the focus of attention will not have narrowed to the location of the first target and therefore the second target will still fall within the focus of attention and Lag-1 sparing will result. On the other hand, if the SOA is long, there will have been sufficient time for the focus of attention to narrow to the first-target stream, leaving the second target in the other stream unattended, resulting in Lag-1 deficit. By indexing the magnitude of Lag-1 sparing with a single value, it is possible to track the change from Lag-1 sparing (indicating a broad focus of attention) to Lag-1 deficit (indicating a narrow focus) across SOAs, as can be seen in Figure 4.4. Changes in the slope of this function indicate changes in the rate at which the focus of attention narrows. 119  In Experiment 4.1, we found that young adults showed a relatively linear change from Lag-1 sparing to Lag-1 deficit as a function of increasing SOA, consistent with a gradually narrowing focus of attention. Older adults, however, had a pattern of results that was consistent with a focus of attention that did not begin to narrow for at least the first 133 ms. It seems that the older adults are slower in disengaging from the broad setting and in initiating the narrowing process. In Experiment 4.2, we tested a longer SOA of 266 ms, and found that with this increased time to disengage and initiate the narrowing process, the older adults were able to fully narrow the focus of attention to the first-target stream by 266 ms. Considered together, the results of Experiments 1 and 2 suggest that although the older adults are slow in initiating the narrowing process, once it has begun it is as efficient as that of the young adults. Experiment 4.3 confirmed that the different response patterns shown by young and older adults were due to changes in attentional process, and not to impoverished information reaching the brain through age-impaired sensory organs. Hemifield Effects A recent report by Verleger et al. (2009) examined visual hemifield effects in a dualstream attentional blink paradigm. They presented two RSVP streams, one in each visual field. They found that T2 identification was significantly more accurate when it appeared in the left visual field. The authors attributed this improved T2 identification accuracy to the fact that the right hemisphere is better able to single out targets which are presented rapidly over time. Since the paradigm used in the present research closely matches that employed by Verleger et al. (2009) we examined hemifield effects in our data a) to provide confirmation of Verleger et al.’s findings, and b) to determine whether hemifield effects would vary between the young and the older adults. 120  In probing for hemifield effects, we considered those trials from Experiment 4.1 in which the first and second targets appeared in different streams. As expected, there was a hemifield effect: when the second target appeared in the left visual field, performance was more accurate at all inter-target lags. This was equally true for both young and older adults. The hemifield effect was also evident at all SOAs, although it was most pronounced at an SOA of 66 ms. The average difference between T2 identification accuracy in the left and right hemifields for young adults was 19.7% at an SOA of 66 ms, 8.1% at an SOA of 100 ms, and 8% at an SOA of 133 ms. The corresponding hemifield effects for older adults were 19.2%, 10.3%, 7.1%, and 9.3%, for SOAs of 66, 100, 133, and 266 ms, respectively . These findings confirm the report by Verleger et al. (2009) that targets presented in RSVP are identified better in the left visual field. Furthermore, this effect appears to remain intact for older adults, indicating that the right hemisphere maintains its advantage in terms of the temporal precision required to extract items presented in rapid temporal sequence.  121  References Bachmann, T., & Allik, J. (1976). Integration and interruption in the masking of form by form. Perception, 5, 79-97. Benso, F., Turatto, M., Mascetti, G.G., & Umiltà, C. (1998). The time course of attentional focusing. European Journal of Cognitive Psychology, 10, 373-388. Bowman, H. & Wyble, B. (2007). The simultaneous type, token, serial token mode of temporal attention and working memory. Psychological Review, 114, 38-70. 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, 21, 109-127. Coltheart, M., & Arthur, B. (1972). Evidence for an integration theory of visual masking. Quarterly Journal of Experimental Psychology, 24, 262-269.]. Dux, P.E., & Coltheart, V. (2005). The meaning of the mask matters: Evidence of conceptual interference in the attentional blink. Psychological Science, 16, 775-779. Elliott, D. B., Whitaker, D., & MacVeigh, D. (1990). Neural contributions to spatiotemporal contrast sensitivity in healthy ageing eyes. Vision Research, 30, 541-547. Enns, J. T. (2004). Object substitution and its relation to other forms of visual masking. Vision Research, 44, 1321-1331. Eriksen, C.W., & Yeh, Y-Y. (1985). Allocation of Attention in the Visual Field. Journal of Experimental Psychology, 11, 583-597. Eriksen, C.W., & St. James, J.D. (1986). Visual attention within and around the field of focal attention: A zoom lens model. Perception and Psychophysics, 40, 225-240. Fehrer, E. (1966). Effects of stimulus similarity on retroactive masking. Journal of 122  Experimental Psychology, 71, 612-615. Folk, C.L., & Hoyer, W.J. (1992). Aging and shifts of visual spatial attention. Psychology and aging, 7, 453-465. Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). ‘‘Mini-mental state’’. A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatry Research, 12(3), 189–198. Giesbrecht, B. L., & Di Lollo, V. (1998). Beyond the attentional blink: Visual masking by object substitution. Journal of Experimental Psychology: Human Perception and Performance, 24, 1454-1466. Greenwood, P.M., Parasuraman, P. & Haxby, J.V. (1993). Changes in visuospatial attention over the adult lifespan.  europsychologia, 31, 471-485.  Harmon, L.D., & Julesz, B. (1973). Masking in visual recognition: Effects of twodimensional filtered noise. Science, 180, June, 1194-1196. Intraub, H., (1981). Rapid conceptual identification of sequentially presented pictures. Boundary extension for briefly glimpsed pictures: Journal of Experimental Psychology: Human Perception and Performance, 7, 604-610. Intraub, H., (1984). Conceptual masking: The effects of subsequent visual events on memory for pictures. Rapid conceptual identification of sequentially presented pictures. Boundary extension for briefly glimpsed pictures: Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 115-125. Jefferies, L.N., & Di Lollo, V. (in press). Linear changes in the spatial extent of the focus of attention across time. Journal of Experimental Psychology: Human Perception and Performance.  123  Jefferies, L.N., Ghorashi, S., Kawahara, J-i., & Di Lollo, V. (2007). Ignorance is Bliss: The role of observer expectation in dynamic spatial tuning of the attentional focus. Perception & Psychophysics, 69, 1162-1174. Jonides, J. (1983). Further toward a model of the mind’s eye’s movement. Bulletin of the Psychonomic Society, 21, 247-250. Lahar, C.J., Isaak, M.I., & McArthur, A.D. (2001). Age Differences in the Magnitude of the Attentional Blink, Aging, europsychology, and Cognition, 8(2), 149 – 159. Lorenzo-Lopez, L., Doallo, S., Vizoso, C., Amenedo, E., Rodriguez Holguin, S., & Cadaveira, F. (2002). Covert orienting of visuospatial attention in the early stages of aging. euroreport, 13, 1459-1462. Maciokas, J.B. & Crognale, M.A. (2003). Cognitive and Attentional Changes with Age: Evidence from Attentional Blink Deficits, Experimental Aging Research, 29, 137 – 153 Maki, W.S., Couture, T., Frigen, K., & Lien, D. (1997). Sources of the attentional blink during rapid serial visual presentation: Perceptual interference and retrieval competition. Journal of Experimental Psychology: Human Perception and Performance, 23, 13931411. Martin, A.J., Friston, K.J., Colebatch, J.G., & Frackowiak, R.S.J. (1991). Decreases in regional cerebral blood flow with normal aging. Journal of Cerebral Flood Flow and Metabolism, 11, 684-689. Nissen, M., & Corkin, S. (1985). Effectiveness of attentional cueing in older and younger adults. Journal of Gerontology, 40, 185-191. Posner, M. I., & Cohen, Y. (1984). Components of visual orienting. In H. Bouma & D. G.  124  Bouwhuis (Eds.), Attention and performance X: Control of language processes (pp. 531– 556). Hillsdale, NJ: Erlbaum. Purcell, D. G., & Stewart, A. L. (1970). U-shaped backward masking functions with nonmetacontrast paradigms. Psychonomic Science, 21, 361-363. 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, 18, 849-860. Robinson, D.L., & Kerttzman, C. (1990). Visuospatial attention: Effects of age, gender, and spatial reference.  europhyschologia, 28, 291-301.  Shefer, V.F., (1973). Absolute number of neurons and thickness of the cerebral cortex during aging, senile and vascular dementia and Pick’s and Alzheimer’s diseases.  euroscience  and Behavioral Physiology, 6, 319-324. Tales, A., Muir, J.L., Bayer, A., & Snowden, R.J. (2002). Spatial shifts in visual attention in normal ageing and dementia of the Alzheimer type.  europsychologia, 40, 2000-2012.  Verleger, R., Sprenger, A., Gebauer, S., Fritzmannova, M., Freidrich, M., Kraft, S., & Jaśkowski, P. (2009). On why left events are the right ones: Neural mechanisms underlying the left-hemifield advantage in rapid serial visual presentation. Journal of Cognitive eurosicence, 21, 474-488. Visser, T. A. W., Bischof, W. F., & Di Lollo, V. (1999). Attentional switching in spatial and non-spatial domains: evidence from the attentional blink. Psychological Bulletin, 125, 458-469. Weale, R. A. (1961). Retinal illumination and age. Transactions of the illuminating engineering Society, 26, 95-100.  125  Weale, R.A. (1963). The Aging Eye . London: Lewis Publishers Yantis, S., Schwarzback, J., Serences, J., Carlson, R.L., Steinmetz, M.A., Pekar, J.J, & Courtney, S.M. (2002). Transient neural activity in human parietal cortex during spatial attention shifts.  ature euroscience, 5, 995-1002.  126  Chapter 5. Is focal attention unitary or divided? It depends on task demands  4  Stimuli presented at attended locations are processed faster and more accurately than those presented at unattended locations (LaBerge, 1995). There are two ostensibly incompatible accounts of how focal attention is deployed in space: (a) a single unitary focus that expands and contracts so as to optimize performance on the task at hand (e.g, Barriopedro & Botella, 1998; Castiello & Umiltà, 1990; Eriksen & Yeh, 1984; Heinze et al., 1994; Jonides, 1983), and (b) multiple foci deployed to different locations simultaneously (e.g., Awh & Pashler, 2000; Kawahara & Yamada, 2007; Müller et al., 2003). Here we test the possibility that either a unitary or a divided mode of attentional control can be implemented, depending on task demands. To answer this question, we employed two well-established paradigms: the attentional blink and Lag-1 sparing. The attentional blink (AB) refers to impaired identification of the second of two rapidly-sequential targets (T1, T2). This deficit is most pronounced at short intertarget lags, and vanishes at lags beyond about 500 ms (Chun & Potter, 1995; Raymond, Shapiro, & Arnell, 1992). Lag-1 sparing refers to the somewhat paradoxical finding that identification of the second target is unimpaired when it appears directly after the first target in the same spatial location (Potter, Chun, Banks, & Muckenhoupt, 1998; Visser et al., 1999). Lag-1 sparing also occurs when the two targets are presented in different spatial locations, but only when T2 falls within the focus of attention (Jefferies & Di Lollo, in press; Jefferies et al., 2007). From this it follows that the incidence of Lag-1 sparing can be used to determine whether or not a particular spatial location falls within the focus of attention. 4  A version of this chapter has been submitted for publication. Jefferies, L.N., Enns, J.T., & Di Lollo, V. Is focal attention unitary or divided?: It depends on the task.  127  In pursuing our objective, we adapted the methodology used by Kawahara and Yamada (2007), who employed two concurrent streams of digit-distractors, one on either side of fixation, separated by a spatial gap. A first pair of letter-targets (T1-pair) appeared simultaneously, one within each stream (In-Stream condition). A second pair of letter-targets (T2-pair) appeared either in-stream or in the space between the streams (Between-Stream condition). In that experiment, Lag-1 sparing occurred when the T2-pair appeared in-stream, but not when it appeared between streams. Kawahara and Yamada concluded that the space between the streams was unattended and, therefore, that focal attention was split, disconfirming the single-focus hypothesis. Here we test the hypothesis that whether focal attention is unitary or divided depends on the observer's mental set. On this view, Kawahara and Yamada's (2007) results stemmed from the fact that the T1-pair always appeared in-stream. On the assumption that attention is initially deployed so as to optimize identification of the T1-pair, this encouraged observers to divide attention into two separate foci, one on each stream (cf. Jefferies & Di Lollo, in press). In the present study we employed two different conditions to encourage observers to adopt either a unitary or a divided focus of attention. In one condition (T1-Pair Predictable), the T1-pair always appeared in-stream, as in Kawahara and Yamada's study, encouraging the deployment of two separate attentional foci. The other condition (T1-Pair Unpredictable) was designed to encourage the deployment of a single broad attentional focus that encompassed all potential target locations. To that end, we presented the T1-pair unpredictably either in-stream or in the space between the two streams (see insets in Figure 5.1). The labels "Predictable" and "Unpredictable" apply only to the location of the T1-pair. In both conditions, the T2-pair appeared randomly and with equal probability either in-stream or  128  between the streams. The trials of interest in the present work are those in which the T2-pair appeared between the streams. This is because the deployment of attention to the space between the streams depends on whether the focus of attention is unitary or divided. Specifically, if focal attention is unitary, the central region is attended, and Lag-1 sparing will occur to T2-pairs in that are presented in that region. If, on the other hand, attention is divided into two separate foci, the central region will be unattended, and Lag-1 sparing will not occur. In brief, we expected to replicate Kawahara and Yamada's findings (no Lag-1 sparing when the T2-pair appears between the streams) in the T1-pair predictable condition, but to find substantial Lag-1 sparing in the unpredictable condition. Such a pattern of results would suggest that focal attention is divided when the location of the T1-pair is predictable and unitary when it is unpredictable. Experiment 5.1 Methods Participants Twenty-six undergraduate students at Simon Fraser University (16 females, ages 19-24) were randomly allocated to two groups of observers: Fifteen observers in the T1-Pair Predictable and eleven observers in the T1-Pair Unpredictable group. All had normal or corrected-to-normal vision and were naïve as to the purpose of the experiment. Stimuli and Procedure All stimuli were displayed on a computer monitor set to a refresh rate of 75 Hz. Observers were seated in a dimly-lit room and viewed the displays binocularly from a distance of approximately 60 cm. The luminance of all stimuli was 129 cd/m2, and the luminance of the black background was 2.3 cd/m2. A small white fixation cross was displayed centrally on a 129  black background. Two synchronized rapid serial visual presentation (RSVP) streams of stimuli were presented 1.75˚ to either side of fixation (centre to centre). Each stream contained 8-14 randomly-chosen leading white digit-distractors (approximately 0.9˚ in width and height), appearing one every 100 ms. The digits were selected randomly with replacement from the set 19, with the restrictions that the same digit could not appear in both streams simultaneously and that the same digit could not be appear in two sequential frames. The targets were two pairs of white letters (T1-pair, T2-pair) drawn from the letters of the English alphabet, excluding I, O, Q, and Z. The two letters in the T1-pair were the same as one another on half the trials, and different on the other half of the trials. The two types of trials were randomly intermixed. The two letters in the T2-pair always differed from one another. The observers made three responses on each trial: the first was to indicate whether the letters in the T1-pair were the same as one another or different, by pressing one of two labeled keys on the keyboard. The second and third responses were to indicate the identity of the letters in the T2-pair by pressing the corresponding keys on the keyboard. For one group of observers – the T1-pair Predictable group – the T1-pair was always presented in-stream. For the other group – the T1-pair Unpredictable group – the T1-pair was presented with equal probability either in-stream or in the space between the streams. For both groups, the T2-pair could appear either in-stream or between the streams and could therefore appear either in the same location as the T1-pair or in the different location (see insets in Figure 5.1). As noted above, the latter is the condition of interest because it permits an assessment of the spatial extent of focal attention. The T2-pair followed the T1-pair at one of three inter-target lags: Lag 1(100 ms; the T2pair was presented directly after the T1-pair), Lag 3(300 ms; two distractors intervened between  130  the two target pairs), and Lag 9 (900 ms; eight distractors intervened between the two target pairs). The three lags were presented randomly with equal frequency across trials. Each target was masked by a trailing digit except when the T2-pair followed directly after the T1-pair (Lag 1) in the same location, in which case the T2-pair masked the T1-pair. The total number of trials in both the T1-pair predictable and T1-pair unpredictable conditions was 384.  Figure 5.1. Mean percentage of correct responses for the T2-pair. Insets illustrate the display sequences at Lag 1. 1A: T1-pair Predictable, 1B: T1-pair Unpredictable (trials in which T1-pair appeared in-stream), 1C: T1-pair Unpredictable (trials in which T1-pair appeared between-streams). The shaded areas represent the hypothesized spatial extent of focal attention  131  Results and Discussion Only those trials in which the response to the T1-pair was correct were included for analysis. This procedure is commonly adopted in AB experiments on the grounds that, on trials in which the first target is identified incorrectly, the source of the error is unknown, and thus its effect on second-target processing cannot be estimated. The two responses to the T2-pair were scored independent of the order in which they were reported. We followed Kawahara and Yamada's (2007) practice of averaging the two responses to the T2-pair to calculate a single T2pair score.  132  Accuracy of T1-pair responses, averaged over lags and T2-pair location was 61.7% for the T1-pair Unpredictable group, and 65.3% for the T1-pair Predictable group. Although seemingly counterintuitive, the somewhat higher accuracy in the Unpredictable condition is consistent with the finding that a unitary attentional focus (T1-pair Unpredictable) leads to more efficient target processing than does a divided focus (T1-pair Predictable; Castiello & Umiltà, 1992). Figure 5.1 illustrates the percentage of correct T2-pair responses for the T1-pair predictable (panel A) and T1-pair unpredictable condition (panels B and C). Because the dependent measure of interest in the present work was Lag-1 sparing, all statistical analyses were limited to the data for Lags 1 and 3. Lag 9 was of no theoretical interest, and was included in the experiment solely to make the appearance of the T2-pair less temporally predictable. We estimated the magnitude of Lag-1 sparing as the positive difference between T2-pair accuracy at Lag 1 and Lag 3. A negative difference was termed Lag-1 deficit. Because the data in Figure 5.1 inherently cannot form a complete factorial design, we performed two separate analyses, one for the T1-pair Predictable group (Figure 5.1A), the other for the T1-pair Unpredictable group (Figures 5.1B and 5.1C). This was followed by a third analysis on the data in Figures 5.1A and 5.1B, which represent matching conditions in the T1pair Predictable and T1-pair Unpredictable conditions. In both of these conditions, the T1-pair was presented in-stream, with the T2-pair presented either in-stream or between the streams. This last analysis is the critical one for the objective of the present work: it addresses the issue of whether the focus of attention is divided in the T1-pair Predictable condition, but unitary in the T1-pair Unpredictable condition.  133  The first analysis was a 2 (Lag: 1, 3) x 2 (T2-pair Location: in-stream, between-streams) within-subject ANOVA for the T1-pair Predictable group (Figure 5.1A). The analysis revealed a significant effect of T2-pair location, F(1,14)=14.02, p=.002, η2 = .50, and a significant Lag xT2Pair Location interaction, F(1,14)=4.85, p=.045, η2 = .257. No other effects were significant. This pattern of results strongly suggests that when the location of the T1-pair was predictable, attention was divided into two separate foci, leaving the central region unattended. If the T2-pair then appeared in that central region, Lag-1 deficit ensued. In contrast, if the T2-pair appeared instream, Lag-1 sparing was in evidence. As expected, these results replicate those of Kawahara and Yamada (2007). The second analysis was a 2 (Lag: 1, 3) x 2 (T1-pair Location: In-stream, betweenstreams) x 2 (T2-pair location: In-stream, between-streams) within-subject ANOVA for the T1pair Unpredictable group (Figure 5.1B and 5.1C). The analysis revealed a significant effect of T2-pair location, F(1,10)=7.17, p=.023, η2 = .42, and a significant Lag xT2-Pair Location interaction, F(1,10)=5.75, p=.037, η2 = .37. No other effects were significant. The finding that Lag-1 sparing occurred when the T2-pair was presented between the streams indicates that the region between the two streams was encompassed within the focus of attention. This, in turn, strongly suggests that, when the location of the T1-pair was unpredictable, attention was deployed as a single unitary focus. The third analysis was a mixed ANOVA with one between-subjects factor and two within-subject factors (Figures 5.1A and 5.1B). The between-subjects factor was at two levels (T1-pair Predictable or Unpredictable). The within-subject factors were Lag (2 levels: 1 and 3) and T2-pair location (2 levels: in-stream, between-streams). The analysis revealed significant main effects of Lag, F(1,24)=7.27, p=.013, η2 = .23, and T1-pair location, F(1,24)=121.37,  134  p=.001, η2 = .84. There were also two significant two-way interactions: T1-pair location x T2Pair location, F(1,20)=15.45, p=.001, η2 = .39, and Lag x T1-pair location, F(1,24)=6.62, p=.017, η2 = .22. Finally, the Lag x T1-pair location x T2-pair location interaction was also significant, F(1,24)=8.11, p=.009, η2 = .25. No other effects were significant. This last analysis is the critical one for the objective of the present study. The comparison of interest is between the two segmented lines for Lags 1 and 3 in Figures 5.1A and 5.1B. These data are for the conditions in which the T2-pair appeared in the region between the streams. According to the rationale outlined in the Introduction, the occurrence of Lag-1 sparing (Figure 5.1B, segmented line) is indicative of a unitary focus of attention whereas the occurrence of Lag-1 deficit (Figure 5.1A, segmented line) is indicative of a divided focus of attention. This means that the focus of attention was divided in the T1-pair Predictable condition (Figure 5.1A, replicating Kawahara and Yamada, 2007) but was unitary in the T1-pair Unpredictable condition (Figure 5.1B). This is consistent with the present hypothesis that attention can be deployed either as a unitary focus or divided into separate foci, depending on task demands. With reference to the issue of whether focal attention is unitary or divided, we conclude that both modes of attentional deployment are possible. Whether one or the other is employed in any given instance depends on the specific demands of the task at hand, pointing to a greater degree of flexibility in the deployment of human spatial attention than in the extant theories proposed to account for it. Experiment 5.2 Experiment 5.2 showed that the focus of attention can be flexibly deployed either as a unitary or divided focus and specifies one set of conditions under which this flexible division takes place. Experiment 5.2 was designed to test whether there is a default mode of attentional 135  deployment and whether the focus of attention can morph from unitary to divided over the course of completing a task. Given that a unitary focus of attention has been shown to be more efficient than a divided one (Heinze et al., 1994), we expected that if there is default mode, it will be unitary. In order to test the hypothesis that the focus of attention is initially deployed as a unitary focus and then divided over time if the task so requires, we replicated Experiment 5.1, but varied the SOA (70 or 100 ms) between successive items in the RSVP stream, thereby manipulating the amount of time between the target pairs. If attention is initially deployed as a unitary focus and then gradually divided into two separate foci across the duration of the task, it should be possible to show these two different states within the same task. When the SOA is relatively long (such as the 100 ms SOA used in Experiment 5.1), there may have been sufficient time for the focus of attention to divide when the location of the T1- pair was predictable. If the T2-pair then appears in the blank regions between the two streams, it will not fall within the focus of attention and Lag-1 deficit will result. If, on the other hand, a shorter SOA of 70 ms is employed, the process of dividing a unitary focus of attention may not yet be complete and that central region may still be partially attended. If the T2-pair appears in that central region, therefore, some Lag-1 sparing may still be evident. Method Observers A total of 20 undergraduate students at the University of British Columbia and Simon Fraser University participated for course credit, 10 in the Unpredictable condition, 10 in the Predictable condition. All participants were naïve as to the purpose of the experiment and reported normal or corrected-to-normal vision.  136  Apparatus and Stimuli All stimuli and procedures were identical to those described in Experiment 5.1, except that each participant completed one block in which the SOA was 70 ms and one block in which the SOA was 100 ms, in counterbalanced order.  Results and Discussion As in Experiment 5.1, only those trials in which the first target was identified correctly were included for analysis. In the Predictable condition, the T1-pair always appeared in-stream whereas in the Unpredictable condition, the T1-pair could appear either in-stream or betweenstreams. In order to meaningfully compare these two groups, only those trials in which the T1pair appeared in-stream will be analyzed. Similarly, the data presented in Figure 5.2 represents only those trials in which the T1-pair appeared in-stream. Visual inspection of the data in Figure 5.2 indicates that when the T1-pair was unpredictable (encouraging the deployment of a unitary focus) and the T2-pair appeared between the streams, there was Lag-1 sparing in both the 70- and 100-ms SOA conditions. This suggests that for the Unpredictable condition: (a) a single unitary focus of attention that encompassed both streams and the intervening locations was deployed, and (b) this distribution of attention did not change as a function of time during the task. A striking difference as a function of time, however, was evident in the T1-pair Predictable condition. In the critical condition in which the T1-pair was predictable and the T2-pair subsequently appeared between-streams, there was substantial Lag-1 deficit in the 100-ms SOA condition, as was found in Experiment 5.1. In the 70-ms SOA condition, on the other hand, there was a significant amount of Lag-1 sparing, indicating that the central region was attended to some degree in this condition. This pattern of  137  results – Lag-1 sparing when the SOA is 70 ms and Lag-1 deficit when the SOA is 100 ms – is consistent with a unitary focus of attention that is divided into two separate foci as the task progresses. This interpretation was confirmed by analyzing the magnitude of Lag-1 sparing in the different-stream condition at 70 and 100 ms for both the Predictable and Unpredictable conditions. A 2 (Lag: 1, 3) x 2 (SOA: 70, 100) within-subjects ANOVA was conducted on the data illustrated in Figure 5.2 for the Predictable condition. The analysis revealed no main effects, but a significant interaction between Lag and SOA, F(1,20)=4.47, p = .04, ηp2 = .31. This interaction confirms that there was Lag-1 sparing in the 70 ms condition and Lag-1 deficit in the 100 ms condition, buttressing the conclusion that at 70 ms there is a unitary focus while at 100 ms there is a divided focus. A second 2 (Lag: 1, 3) x 2 (SOA: 70, 100) within-subjects ANOVA was conducted on the data illustrated in Figure 5.2 for the Unpredictable condition. This analysis revealed an interaction between Lag and SOA that was not significant, F(1,20) = 1.07, p = .31 , which supports the visual interpretation that the magnitude of Lag-1 sparing did not vary as a function of SOA – the focus of attention is unitary at both 70 and 100 ms. In summary, Experiment 5.2 suggests that the default mode of attentional deployment is unitary, and that a focal of attention can change from a unitary mode to a divided mode over approximately 100 ms as a task progresses. This experiment provides a clear demonstration that both modes of attentional deployment can operate within the same task, changing only as a function of time.  138  Figure 5.2. Mean percentage of correct responses for the T2-pair for trials on which the response to the T1-pair was correct. Insets illustrate the display sequence for Lag 1, in which the T2-pair was presented directly after the T1-pair. The shaded areas represent the hypothesized spatial extent of focal attention: a unitary focus when the location of the T1pair was unpredictable, a divided focus when the location of the T1-pair was predictable. The data in the graphs illustrate only the critical condition in which the T1-pair appeared in-stream. The solid lines represent performance when the T2-pair appeared in-stream; the segmented lines represent performance when the T2-pair appeared between-streams.  139  Although intriguing, the conclusion that the “default” setting of attention is unitary and that attention is divided over time if the task demands so require is not merited in a strong form from these data alone – further testing and confirmation from additional experimental paradigms and with a more complete testing of the time course would first be required. The findings do, however highlight the extent to which the deployment of attention across space is flexible and dynamic, fluctuating with both time and task demands. It is clear from Experiments 5.1 and 5.2 that focal attention is not exclusively unitary or exclusively divided – if the conditions are right, either mode of attentional deployment can be employed. Furthermore, there is suggestive evidence that focal attention is initially deployed as a single, unitary focus and then divides over approximately 100 ms into two separate and independent foci.  140  References Awh, E. & Pashler, H. (2000). Evidence for split attentional foci. Journal of Experimental Psychology: Human Perception and Performance, 26, 834-846. Barriopedro, M.I. & Botella, J. (1998). New evidence for the zoom lens model using the RSVP technique. Perception & Psychophysics, 60, 1406-1414. Castiello, U, & Umiltà, C. (1990). Size of the attentional focus and efficiency of Processing. Acta Psychologica, 73, 195-209. Castiello, U., & Umiltà, C. (1992). Splitting focal attention. Journal of Experimental Psychology:Human Perception and performance, 18, 837-848. 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, 21, 109-127. Eriksen, C.W. & Yeh, Y-y. (1985). Allocation of attention in the visual field. Journal of Experimental Psychology: Human Perception and Performance, 11, 583-597. Heinze, H-J., Luck, S.J., Munte, T.F., Gös, A., Mangun, G., & Hillyard, S. (1994). Attention to adjacent and separate positions in space: An electrophysiological analysis. Perception and Psychophysics, 56, 42-52. Jefferies, L.N., & Di Lollo, V. (in press). Linear changes in the spatial extent of the focus of attention across time. Journal of Experimental Psychology: Human Perception and Performance. Jefferies, L.N., Ghorashi, S., Kawahara, J-i., & Di Lollo, V. (2007). Ignorance is bliss: The role of observer expectation in dynamic spatial tuning of the attentional focus. Perception & Psychophysics, 69, 1162-1174. 141  Kawahara, J-i. & Yamada, Y. (2006). Two noncontiguous locations can be attended concurrently: Evidence from the attentional blink. Psychonomic Bulletin and Review, 13, 594-599. LaBerge, D. (1995). Attentional Processing: The Brain’s Art of Mindfulness. Harvard University Press: Cambridge, MA. Müller, M.M, Malinowski, P.,Gruber, T., & Hillyard, S.A. (2003). Sustained division of the attentional spotlight.  ature, 424, 309-312.  Potter, M. C., Chun, M. M., Banks, B. S., & Muckenhoupt, M. (1998). Two attentional deficits in serial target search: The visual attentional blink and an amodal task-switch deficit. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24, 979-992. 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, 18, 849-860. Visser, T. A. W., Bischof, W. F., & Di Lollo, V. (1999). Attentional switching in spatial and non-spatial domains: evidence from the attentional blink. Psychological Bulletin, 125, 458-469.  142  Chapter 6: General Discussion Summary The principal objective of this dissertation was to probe the spatiotemporal dynamics of visual attention. There were two specific goals: a) to probe the spatiotemporal dynamics of the contraction of the focus of attention, and b) to examine whether the focus of attention can be deployed flexibly as either a unitary or a divided focus according to task demands. The overview presented in Chapter 1 shows that research on the spatiotemporal dynamics of visual attention has centered primarily on shifts of attention from one location to another. Relatively little heed has been paid to the fact that the focus of attention can also be adjusted in spatial extent. In fact, although the rate at which attention is shifted from one location to another has been the topic of many studies, only a single study has attempted to provide an estimate of the rate at which the focus of attention can be expanded. At this point, there is still no estimate of the rate at which attention can be narrowed or an examination of the factors that can initiate or modulate the rate of the expanding and contracting processes. Just as the existing research has focused on attentional shifts while ignoring the process of expanding and contracting, previous research has also become mired in a debate as to whether the focus of attention is unitary or divided. Although there is substantial research to support both positions, these alternatives have generally been viewed as mutually exclusive. The alternative hypothesis – that the focus of attention can be deployed flexibly as either a unitary or divided focus according to moment-tomoment task demands – has not yet been tested. The bulk of this dissertation was devoted to answering these questions.  143  The first study (Chapter 2) was devoted to examining the time-course of changes in the spatial extent of focal attention. In pursuing this goal, we combined the Attentional Blink (impaired perception of the second of two targets) and Lag-1 sparing (the seemingly paradoxical finding that second-target accuracy is high when the second-target immediately follows the first) in a dual-stream paradigm to assess the spatiotemporal changes in focal attention. It is known that Lag-1 sparing only occurs to targets in different spatial locations if the second target falls within the focus of attention. As such, the incidence and magnitude of Lag-1 sparing can be used to track whether the location at which T2 appears is encompassed by the focus of attention at any particular moment. In this paradigm, focal attention is assumed to initially encompass both streams but to shrink rapidly to the first-target stream, thus withdrawing from the secondtarget if it appears in the opposite stream. The time available for the focus to shrink before second-target onset was manipulated by varying the SOA between successive items in the stream. There was a progressive transition from Lag-1 sparing to its converse (Lag-1 deficit) as the SOA was increased. This transition was related linearly to SOA, which suggests that the spatial extent of focal attention varies linearly over time. A rough estimate of the rate at which this contraction occurs is provided along with a descriptive model of the changes in focal attention in Chapter 2. Experiment 2.2 tests the model by reducing the spatial separation between the RSVP streams, and thereby reducing the distance over which focal attention must contract. Predictable changes in the rate at which Lag-1 sparing changed to Lag-1 deficit across SOA were obtained. The second series of experiments (Chapter 3), was designed to examine whether the rate at which the focus of attention narrows is modulated by those attributes of a stimulus that cause attention to be drawn to them more or less rapidly. We tested stimuli that draw attention due to  144  their goal-relevant nature (Experiments 3.1 and 3.2) as well as those which capture attention reflexively due to low-level stimulus features such as brightness (Experiment 3.3). We found that the rate at which the focus of attention narrows varied predictably with these manipulations. Since Chapter 3 examined whether stimulus characteristics influenced the rate at which the focus of attention was narrowed, the third study (Chapter 4) probed instead whether organismic variables modulate the rate at which the spatial extent of focal attention is adjusted. Specifically, in Experiment 4.1 we tested whether the aging process influences the principles of shrinking and expanding the focus of attention. We found that older adults were delayed in beginning the narrowing process – in fact, the narrowing process did not begin for at least 133 ms after the appearance of the first target. When a longer SOA (266 ms) was tested in Experiment 4.2, however, the focus of attention had been fully narrowed to the location of the first target, indicating that once the narrowing process begins, it is as rapid and efficient as for young adults. In Experiment 4.3 we confirmed that the changes in the narrowing process associated with age are in fact due to changes in the brain, and not to limitations in the peripheral sensory organs. Chapter 5 addresses the issue of whether the focus of attention is deployed as a unitary or a divided focus of attention. Mutually-exclusive theories posit that spatial attention is allocated either as single focus or as multiple foci. In this experiment we show instead that the mode of attentional deployment changes based on task demands. We used two RSVP-streams and two target pairs (T1-pair,T2-pair) to probe whether a single focus or dual foci were employed. For one group of observers, the T1-pair appeared predictably within the streams (encouraging separate-foci strategy); for a second group of observers it appeared unpredictably on-stream or between the streams (encouraging single-focus strategy). When the T2-pair appeared between  145  the streams, Lag-1 deficit occurred in the T1-pair predictable condition (consistent with a dualfocus) and Lag-1 sparing occurred in the T1-pair unpredictable condition (consistent with a single focus). Thus, attention can be deployed either as a single-focus or as multiple-foci. Whether one or the other is employed in any given instance depends on the specific demands of the task at hand, the objective being to optimize performance.  arrowing the Focus of Attention to the Location of the First Target The theoretical model outlined in Chapter 2 is predicated on the assumption that when the first target appears, the focus of attention narrows to that location. There is much evidence in the literature that when an item appears that matches the observer’s mental set (i.e., it shares characteristics with the target), it captures attention and draws attention to its location to the detriment of other items. Consider, for example, the case of continent capture as outlined by Folk, Leber, & Egeth (2002). Observers were presented with a single RSVP stream of letters and were instructed to look for a red letter target. One or more frames prior to the onset of the letter target, four distractor items in the form of number signs (#) were presented in a square surrounding the distractor letter. On half of the trials, all four of the number signs were black. These items did not capture attention and target identification was very accurate. On the remaining trials, however, one of the number signs was red (and hence shared a defining characteristic with the target). When this red item was present, identification of the letter target was poor, suggesting that the irrelevant distractor item had captured attention and drawn it reflexively to its location. Comparably, in the paradigm employed in the present series of experiments, when the first target appears, it reflexively captures attention, which narrows to its location.  146  In addition to evidence that items matching the observer’s mental set reflexively capture and draw attention, there is also evidence to show that the spatial distribution of attention strongly influences the accuracy with which targets can be identified. For example, when the focus of attention has a broad spatial extent, perceptual reports of targets are slowed and less accurate than when the focus of attention has a relatively narrow spatial extent (e.g., Barriopedro & Botella, 2002; Castiello & Umiltà, 1990; Egeth, 1970). Most directly to the point for the present experiments involving dual RSVP streams, the benefit of narrowing the focus of attention to the location of the first target has been demonstrated by Jefferies, Ghorashi, Kawahara, & Di Lollo (2007). In that research, one group of observers were told which stream would contain the first target (the Known condition). In order to optimize first-target identification accuracy, therefore, the ideal strategy of the observers would be to narrowly center their focus of attention on the stream known to contain the first target. A second group of observers had no advance knowledge of which stream would contain the first target – it could appear randomly in either the left stream or the right stream (the Not-Known condition). Given that the location of the first target was unknown, the observers’ best strategy would be to deploy a broad focus of attention that encompassed both streams. Jefferies et al. found that identification accuracy for the first target was approximately 10% higher in the Known condition than in the Not-Known condition, indicating that there is a processing-benefit advantage to narrowing focal attention to the location of the first target. In summary, there is clear evidence in the literature to support the hypothesis that the focus of attention narrows to the location of the first target when it appears.  147  Comparing Results Across Experiments In the following section, the data reported in Experiments 2.1, 2.2, 3.1, and 3.3 are plotted in a single graph so that they can be compared directly. Each of these experiments tested the rate at which the focus of attention shrinks, and probed different factors that influence the rate at which this shrinking takes place. Direct comparison of the results of these experiments allows for a better understanding of the process of narrowing and expanding the focus of attention. The theoretical perspective on the process of expanding and contracting focal attention that was adopted in this thesis is as follows. In essence, it is posited that because the observer never knows which stream will contain the first target, the focus of attention will initially be broad in order to encompass both streams. The initial focus is expected to be broad regardless of the manipulation being implemented in the experiment since the process of narrowing is not triggered until the appearance of the first target. Once the first target appears, the focus of attention begins to narrow to that location so as to optimize identification of the first target. This narrowing appears to be a linear process that occurs gradually over time. Concomitant with the narrowing to the first-target-stream is a gradual withdrawal from the opposite stream. It is known that Lag-1 sparing occurs to targets that appear in different spatial locations only if they fall within the focus of attention (Jefferies et al., 2007). Given this, the Lag-1 sparing that occurs when the focus of attention encompasses both streams will gradually change to Lag-1 deficit as the focus of attention narrows to the location of the first target, leaving the opposite stream unattended. The incidence and magnitude of Lag-1 sparing can, therefore, be used to index the width of the focus of attention: the presence of Lag-1 sparing indicates a broad focus while the presence of Lag-1 deficit indicates a narrow focus centered on the opposite (T1) 148  stream. As shown in Experiment 2.1, there is an overall trend for Lag-1 sparing to turn into Lag1 deficit as the SOA is increased, mirroring the linear narrowing of the focus of attention across time. The magnitude of Lag-1 sparing or Lag-1 deficit as a function of SOA is plotted for each experiment in Figure 6.1 below.  Figure 6.1. Variation in the magnitude of Lag-1 sparing (positive values) and Lag-1 deficit (negative values) as a function of SOA for Experiments 2.1 (filled circles), 3.1 (black and white triangles for the Slow-shrink and Fast-shrink conditions, respectively), and 3.3 (filled squares).  In interpreting Figure 6.1, the critical fact to note is that at any particular magnitude of Lag-1 sparing, the focus of attention has narrowed to a specific width. This is true regardless of 149  which experiment that data is drawn from, and hence cross-experiment comparisons are valid. In comparing results within this graph, we will limit our considerations to the data reported in Chapter 3. The reason for this is that precise predictions can be made regarding how the data from Experiments 3.1 and 3.3 should compare to one another in terms of the rate at which the focus of attention contracts. Experiment 3.2 is a control experiment and will not be considered here. As discussed above, if one chooses a single magnitude of Lag-1 sparing, this is equivalent to choosing a single width of focal attention. By determining what SOA that width is reached at in any given experiment, one can evaluate whether the rate of narrowing was rapid or slow. Consider the dashed rectangle inscribed in Figure 6.1. It is anchored at a single magnitude of Lag-1 sparing (about 8%), and therefore indicates a specific width of focal attention. The three symbols which lie inside the rectangle represent conditions in which the same magnitude of Lag-1 sparing has been obtained. One can infer, therefore, that the focus of attention has narrowed the same amount in each of the three conditions. Notice that the SOA associated with these three data points differs, indicating that different SOAs were required for the focus of attention to narrow the same amount. This suggests that the rate at which the focus of attention narrowed differed in the three conditions. The relative SOA at which the same magnitude of Lag-1 sparing (and hence the same width of focal attention) is reached in each condition can be predicted directly from our conceptual framework. In considering the three points inside the dashed box, the black square is located towards the left of the rectangle, at the 53 ms SOA point. This symbol represents data from Experiment 3.3, in which the targets were of higher luminance than the distractors. It is known that physically salient stimuli are processed more rapidly than dim stimuli (Woodworth &  150  Scholsberg, 1954) and tend to “pop out” in visual search tasks (Theeuwes, 1991), and it was therefore expected that they would also induce a relatively rapid narrowing of focal attention. Since the focus narrowed the specified amount in only 53 ms, bright targets did indeed result in a very rapid narrowing of focal attention, consistent with our expectations. To the right of the black square are black and white triangles, representing data from the Slow-shrink and Fast-shrink conditions of Experiment 3.1, respectively. In Experiment 3.1, the relative attentional pull of the two RSVP streams was manipulated by varying the degree to which the streams were processed for meaning. When items which must be processed for meaning (strong-pull items; digits) are in competition with items that do not need to be processed for meaning (weak-pull items; random-dot patterns), the process of narrowing to the strong-pull stream away from the weak-pull stream will be rapid (Fast-shrink condition) while the reverse process of narrowing to the weak-pull stream away from the strong-pull stream will be slow (Slow-shrink condition). Consistent with this expectation, the width of focal attention indicated by the dashed rectangular box is reached in the Fast-shrink condition at an SOA of 80 ms, but is not reached in the Slow-shrink condition until the SOA is 118 ms. In other words, it takes approximately 38 ms longer to narrow the specified amount in the Slow-shrink condition than in the Fast-shrink condition. Incorporating ew Data into Current Models of Visuospatial Attention Models and theories provide a tangible means of probing our understanding of cognitive functions and allow us to test that understanding by exploring the predictions which the models make. As such, they play a critical role in the process of science. As much as models and theories are useful, they limit our thinking as much as guide it, and we must be ever-vigilant to ensure that they continue to accurately describe the existing data. The utility of any given model 151  can, therefore, be assessed by the extent to which it can accommodate new experimental findings. Current models of visuospatial attention have been developed primarily to accommodate a single focus of attention that shifts from one location to another. As is made clear by the research reported in Chapters 2-5 of this dissertation, however, in order to accurately represent the spatiotemporal dynamics of visual attention, these theories must also be able to accommodate a focus of attention that: a) expands and contracts over time, b) is deployed as either a unitary or a divided focus depending on task demands, and c) can morph from unitary to divided over time. In the following sections, three current theories of visual attention will be evaluated with regards to how readily they can accommodate these two facets of the spatiotemporal dynamics of visual attention. As discussed earlier, modern theories of visuospatial attention have been developed primarily within a unitary-focus perspective, and most are ideally suited to explain a single focus of attention that shifts rapidly from one location to another while a task is being completed. It is clear, however, from the data presented in this dissertation that the spatiotemporal dynamics of attentional control are extremely flexible and dynamic. Models of attention can no longer operate solely within a “unitary” framework. Although the issue of task-dependent division of focal attention is a complex one, it is an issue which any viable theory of visual-spatial attention must be able to address. In the following sections of this review, three modern theories of visuospatial attention will be examined in detail: The Human Attentional Network model (Posner et al., 1980), the Episodic Theory of attention (Sperling & Weichselgartner, 1995), and the Activity Distribution model (as detailed in LaBerge, 1995). The theories will be reviewed in light of the extent to which they can incorporate: a) a focus of attention that expands and  152  contracts as well as shifts, and b) a focus of attention that can be flexibly deployed as either a unitary or a divided focus, and which can change rapidly from one to the other.  The Human Attentional etwork Model One influential theory of attention is the Human Attentional Network model, developed by Posner and colleagues (Posner & Petersen, 1990; Posner, Petersen, Fox, & Raichle, 1998; Posner, Sheese, Odludaş, & Tang, 2006; Posner, Snyder, & Davidson, 1980). The most fundamental premise of the Human Attentional Network model is that while complex cognitive tasks are not performed by any single brain area and hence cannot be localized in the cortex, the elementary operations that underlie more complex tasks are strictly localized in the cortex. A set of these distributed elementary operations must function in harmony in the form of a network to allow the completion of more complex tasks (Posner, Petersen, Fox, & Raichle, 1998). Posner suggests that by reducing complex tasks into their more basic components and by examining the brain with neuroimaging, single-cell-recordings, or lesion studies for areas which show the characteristic enhancement in neural firing which is the neural hallmark of attention, we can uncover the networks responsible for a variety of cognitive tasks. Posner has developed a rather extensive model of the human attentional network which, he holds, is comprised of at least four primary functional networks: an alerting network, an orienting network, a pattern recognition network, and a network for executive control. Since the aim of this section is to examine the extent to which the Human Attentional Network model can account for: a) a focus of attention which can be expanded and shrunk as well as shifted, b) a focus of attention that can be flexibly deployed in either a unitary or a divided manner, and c) a  153  focus of attention that can morph from unitary to divided over time, discussion will, therefore, be limited to the Orienting network. Orienting and the Posterior Attention etwork Orienting, according to Posner, is the “selective allocation of attention to a particular part of the visual field” (Fernandez-Duque & Posner, 1997, p. 477). Essentially, orienting is about the allocation and re-allocation of visual attention over space, and as such is the network responsible for all aspects of the spatiotemporal dynamics of attention. In the following sections, Posner’s approach to attention shifts will be outlined along with the evidence to support his theory. How those same mechanisms may be involved in expanding/contract or dividing focal attention will then be discussed. Orienting in the Human Attention Network model is primarily about shifting a single focus of attention from one location to another. Consider a simple task which would require a shift of focal attention – a first target appears on the left side of a computer monitor, followed rapidly by a second target on the right side of the screen. The two targets are letters drawn randomly from the set B, G, C, D, P, Q, and O, and the task is to indicate as rapidly as possible which letters were presented. This is a relatively difficult task, and in order to accurately identify the targets, focal attention must be shifted from the first target to the second target when it appears. Posner suggests that there are three separate stages involved in a shift of attention. First, after the first target has been processed, attention must be disengaged from that target so that it may be shifted to the second target. The shift itself must then be completed, and finally attention must be engaged on the new target so that it can be processed and identified. In Posner’s conception, then, an attentional shift has three components: disengage, shift, and 154  engage. There is substantial evidence for the dissociation between the disengage, shift, and engage components of attention shifts, and as the Human Attentional Network model predicts, each component is localized in a different part of the cortex: disengaging in the posterior parietal cortex, shifting in the superior colliculus, and engaging in the pulvinar nucleus of the thalamus. The first component of an attention shift is disengaging from the current area of attentional engagement. Evidence for the role of the parietal cortex in the disengagement of attention is provided by data from patients with parietal lobe lesions. Consider a simple reaction time experiment in which a cue is presented and followed by a target. The target can appear either in the same location as the cue (a valid trial) or in a different location (an invalid trial). Cues and targets can also appear in the contralesional or ipsilesional visual hemifield. Parietal lesion patients did showed a dramatic increase in reaction times to targets that appeared after an invalid cue in the contralesional hemifield (Posner, Walker, Friedrich, & Rafal, 1984; Rafal & Posner, 1987). In other words, they were unable to disengage from the invalid cue in order to respond to the target, and this provides strong evidence that the role of the parietal lobe is primarily one of attentional disengagement. Converging evidence for the role of the parietal cortex in attentional disengagement is provided by neuroimaging studies with normal individuals, which showed activity in the posterior parietal lobe in tasks involving disengaging attention (Corbetta et al., 2000). Returning to the simple cueing paradigm introduced at the beginning of this section, after attention has been disengaged from the cue, the next step is to shift the focus of attention. There is compelling evidence that the shift component involves the Superior Colliculus, a midbrain structure, which has been strongly implicated in the planning and execution of eye movements, and which seems also to be intimately involved in the control of covert attention  155  shifts. Specifically, patients with a degenerative disorder known as supranuclear palsy suffer from a gradual deterioration of the superior colliculus. These patients are either much slower or completely unable to shift their attention from one location to another, as evidenced by slowed reaction times to spatial cues. This deficit appears to be shift-specific in that shifting attention is impaired regardless of whether attention was already engaged at a location, ruling out the possibility that the deficit lies in disengaging attention (Posner, Choate, Rafal, & Vaughan, 1985). The third and final component of the attention shift is engaging attention on the new target. There is substantial evidence to suggest that the pulvinar nucleus of the thalamus is responsible for attentional engagement. Rafal and Posner (1987) tested patients with localized thalamic lesions in a spatial cueing task. They found that although the patients were able to shift attention readily once the shift was completed, they were unable to restrict their focus of attention to the target if the target appeared in the contralesional hemifield. In other words, they were unable to engage attention at the location of the target. The role of the pulvinar in engaging attention at a new location or object can also be demonstrated in normal adults who are asked to restrict their attention to a target that appears amongst distractor items. PET studies have shown that in this case, the pulvinar is strongly activated (LaBerge & Buchsbaum, 1988) whereas the parietal lobe and the superior colliculus are not. In summary, there is substantial evidence supporting the Human Attentional Network’s proposal that a network of three brain areas underlines the orienting of visual-spatial attention. Specifically, there is convincing evidence that orienting is a three-part process involving disengaging (mediated by the posterior parietal cortex), shifting (mediated by the superior colliculus), and engaging (a process governed by the pulvinar nucleus of the thalamus).  156  Incorporating a Focus of Attention that Expands and Contracts Although the Human Attentional Network model does not deal specifically with a focus of attention that expands and contracts according to task requirements, it can easily accommodate such a perspective. In fact, if one restructured the hypothesis that a shift of spatial attention involves a disengage-shift-engage process to the hypothesis that changes in spatial attention require a disengage-change-engage process, the process of expanding and contracting would fit in very neatly. In this case, one would predict that parietal activation should be seen prior to the narrowing process. In Experiment 2.1, for example, if one were to measure brain activity using a temporally-sensitive technique such as Magnetoencephalography (MEG), which also has high spatial precision, one would expect to see a burst of Parietal activity linked to the appearance of the first target – in essence this would be a neural correlate of the disengagement required prior to shrinking the focus of attention. Similarly, one might predict that the superior colliculus would be active during the narrowing process. The experiments presented in Chapter 4 provide some evidence to support a disengagenarrow-engage process involved in contracting focal attention. Although young adults appear to show a constant, linear narrowing of focal attention (e.g., see Figure 2.7 in Chapter 2), older adults show a delay before the narrow process begins. This delay of around 133 ms may reflect a delay in disengaging attention from a “broad” state so that it may be narrowed. It also seems that once the older adults begin to narrow the focus of attention, the rate at which the narrowing process occurs is approximately equal to that at which the young adults narrow the focus of attention. The fact that the narrowing process is unimpaired with age while the process of disengaging appears to be delayed supports the idea that the disengage-shift-engage process  157  posited by the Human Attentional Network model can be applied equally to a focus of attention that narrows and expands. Incorporating a Flexibly Divided Focus of Attention The Human Attentional Network model was designed specifically to feature a unitary focus of attention, and Posner has maintained that focal attention is unitary and provided substantial evidence to support his position. Although a unitary focus is integral to the existing conception of the model, there is nothing inherent in the Human Attentional Network model that prevents focal attention from being divided. That is, there is nothing in the proposed mechanisms of attention or in the underlying neural basis that prevents more than one focus of attention from being deployed simultaneously, each being subject to a disengage-shift-engage process when moved. Similarly, there is nothing inherent to prevent a single focus of attention being split into multiple foci. One critical characteristic of a focus of attention which is divided according to task demands is that it takes a certain amount of time for that division to take place (less than 100 ms, as described in Experiment 5.2). The Human Attentional Network model provides a simple explanation for this delay in dividing focal attention. Specifically, in the same way that a single focus of attention is subject to a disengage-shift-engage process when being moved from one location to another, it is also plausible that when focal attention is divided, a disengage-splitengage process is involved. This required disengagement process is the source of the 100-ms delay. As discussed with reference to a focus of attention which expands and contracts, if dividing focal attention involves a disengage process, then one would expect the posterior parietal lobe to be active during the 100-ms period prior to the emergence of a divided focus of 158  attention. This could be tested quite readily by combining a paradigm such as that described in Experiment 5.1 with neuroimaging techniques such as MEG, which features both temporal and spatial precision. In summary, the Human Attentional Network model provides an excellent explanation of some aspects of divided focal attention; in particular it provides an explanation of the delay that occurs before a single focus is divided into multiple foci. Providing one accepts the evidence showing that focal attention can indeed be divided and assumes the existence of multiple foci, this model provides a solid and well-supported model of attention shifts whether of a single focus or of multiple foci. The model, however, does not provide a mechanism or a rationale for the task-dependent division of focal attention and hence does not fully capture the flexibility of the deployment of focal visual attention.  The Episodic Theory of Attention The second theory of attention to be considered is Sperling and Weichselgartner’s (1995) Episodic Theory of the dynamics of spatial attention, which developed from the debate as to whether the movement of the focus of attention is analog or quantal. This debate hinged on the issue of whether in the process of shifting focal attention from one location to another, the focus moves smoothly over the intervening areas (an analog shift) or whether it “jumps” from one region to another (referred to as a quantal or discrete shift). In an analog shift, the total shift time is dependent on the distance to be travelled and intermediate areas benefit from attentional processing whereas in a quantal shift, shift-time is independent of distance and intermediate regions are unattended.  159  In an analog shift, all that changes as a function of time is the spatial location of the spotlight – the degree of illumination and the spatial extent of the spotlight remain unchanged. In a quantal shift, on the other hand, there is a gradual “closing” of the spotlight at one location accompanied by a simultaneous gradual “opening” at the new location, with no attentional resources distributed to the intervening areas. Due to the gradual nature of the turning on and turning off processes in the quantal shift case, there is a brief period of time at which residual activity is present at both the initial and the subsequent spatial locations. This fact will be of considerable importance in assessing this model and the extent to which it can incorporate a flexibly divided focus of attention. According to Sperling and Weichselgartner (1995), attention is “the modifiable component of an observer’s location-dependent readiness to respond to a stimulus” (p. 504). This state has a specific time-frame, referred to as an episode (Ei) and is a function of both time (t) and space (which can be two- or three-dimensional, but for the sake of simplicity and in order to enhance comparability with the other theories of visual attention, will only be considered as the two-dimensional x, y). Unlike previous spotlight models (e.g., Eriksen & Yeh, 1985; Eriksen & St. James, 1986; Posner & Petersen, 1990; Posner, Snyder, & Davidson, 1980), which posited the existence of a single spotlight of attention which shifted from one location to another in order to process the relevant information in the visual field, Sperling and Weichselgartner’s proposal is quite different. Instead of a single spotlight of attention, the episodic theory proposes that the visual field is covered by a grid of spotlights. These spotlights can differ from one another in size and shape as well as in the total amount of light they produce (i.e., the degree to which the information at a specific area in the visual field is enhanced by attention). The primary constraint on the functioning of this grid of spotlights is that each spotlight can be moved only  160  with great difficulty, and cannot be moved at all while it is illuminated. Hence the primary mechanism underlying shifts of attention is not the movement of the spotlight per se. Instead, Sperling and Weichselgartner propose that the “movement” of focal attention is in fact more like illusory motion, caused by a sequence of spotlights being turned on and off in sequence. A shift of attention from one location to another can, therefore, involve either a single quantal change from one spotlight to another or multiple quantal changes depending on the size of the spotlights and the distance to be covered during the shift. Although the process of turning a spotlight on or off is theoretically instantaneous, Sperling and Weichselgartner (1995) suggest that it is in fact a gradual process. Speaking metaphorically, in the same way that when a strong light is turned off, there is still some energy in the form of heat that remains at that location and dissipates gradually over time, even though attentional spotlights can be turned off instantaneously, there is still some residual attention at that location which fades gradually over time. Practically, then, switches from one spotlight to another are not instantaneous, and there will inevitably be some overlap in time at which both spotlights have a certain amount of energy emanating from them. Sperling and Weichselgartner argue that this monotonic change from on to off (or vice versa) can be modelled by a transition function [written as G(t-ti)], which describes a gamma function that rises monotonically from 0 to 1. Attentional episode 1 at time f0, centered at a specific coordinate (x, y) is gradually replaced by attentional episode 2 at time f1, centered at a new (x, y) coordinate5. The transition function G(t-ti) describes the timecourse of this change without necessarily specifying the spatial coordinates that focal attention is shifted from or to (in essence, it models the decrease in attention at one location and simultaneous increase at another location,). A so-called density  5  Reeves and Sperling (1986) provided a formal model of the process by which attention turns on and off at a given location, and their model provides the basis for the transition function outlined here.  161  function can be derived from the transition function and corresponds to the rate of attention shifting. Although the precise mathematical details of this model are not necessary for the present discussion, the Episodic Theory provides precise predictions regarding the speed at which focal attention can be shifted from one location to another, and the precise manner in which focal attention is removed from one location and allocated to another. Incorporating a Focus of Attention that Expands and Contracts The Episodic Theory has a built-in flexibility component in that it is proposed that the grid of spotlights is not uniform – the spotlights can differ in size, shape, extent, and “strength.” If such variability is to be effective, of course, the grid of spotlights must be adjustable based on task demands, changing on a moment-to-moment basis. Such flexibility is the key to a realistic representation of how focal attention operates, and allows the model to explain many of the experimental findings. In positing that the spotlights within the grid can differ independently from one another in terms of their spatial extent, the model provides a basis for a focus of attention that expands and contracts on a moment-to-moment basis so as to optimize task performance. The model does not provide, however, any specific mechanism by which this flexible deployment might take place, nor does it outline the circumstances which might initiate the narrowing or expanding process. One can, of course, invoke top-down inputs from some executive process, but this is vague at best and limits the descriptive power of the model. Incorporating a Flexibly Divided Focus of Attention As is the case with the Human Attentional Network model, the Episodic Theory of attention was developed within a unitary framework. Unlike the Human Attentional Network model which assumes a single focus of attention that is shifted from one location to another, the 162  Episodic theory assumes a grid of spotlights covering the entire visual field. Only one spotlight can be active at a time, however, and hence the model takes an essentially unitary perspective. The Episodic theory is, however, less strongly unitary than is the Human Attentional Network model. The model already incorporates multiple spotlights and, since it takes time to turn one spotlight off and another one on, more than one spotlight is lit at a time, at least for some interval. In fact, if one considers that the time-course of shifts described by the transition or density functions can take up to 300 ms to complete (Weichselgartner & Sperling, 1987), multiple spotlights may be illuminated simultaneously for several hundred milliseconds. Given this, there is nothing inherent in the model that would prevent more than one spotlight being lit simultaneously for sustained periods of time. In addition to enabling focal attention to be flexibly deployed as either a unitary or a divided focus depending on task demands, models of attention must also account for the delay before focal attention is divided. The transfer function built into the Episodic model provides a natural mechanism to explain this delay. In the same way that a shift is accomplished by focal attention gradually decreasing at one location while gradually increasing at a different location, morphing from a unitary focus to a divided one would also be a gradual process governed by the transition function – and this neatly mirrors the observed data. Although the Episodic Theory provides a flexible model of attention that could easily accommodate a divided focus of attention, and although it can readily explain the delay while focal attention is divided, it is subject to the same limitation as Posner’s Human Attentional Network model. Specifically, although it can accommodate a flexibly deployed focus of attention, there is no mechanism by which this flexible deployment can take place. The model provides no means by which multiple foci would be deployed rather than a single focus, and no  163  insight into the circumstances which would result in each being deployed. In addition, there is no specified mechanism by which the splitting process would take place. One can, of course, invoke top-down inputs from some executive process, but this is vague at best and limits the descriptive power of the model.  The Activity Distribution Model In the Activity Distribution model (La Berge, 1995; La Berge & Brown, 1989) it is proposed that the deployment of focal visual attention is determined by the interaction of several different components of attention. Specifically, it is proposed that two main components of attention, selective attention and preparatory attention, fluidly interact to determine where focal attention is allocated. Preparatory attention is represented as a “peaked distribution of preparatory attentional activity”, or a so-called activity distribution. The activity distribution is a conceptual representation of the amount of neural activity associated with particular objects or locations in space. The spatial extent of activity distributions and the locations at which they form are determined by the interaction of exogenous and endogenous factors, including current stimuli, prior experience, and expectation. La Berge and Brown (1989) suggest that activity distributions can be best represented as three-dimensional Gaussian (normal) distributions with the maximum height of the distribution coinciding with the center of focal attention and with a gradual decline in height at locations of increasing distance from the attended area (see Figure 6.2).  164  Figure 6.2. A representation of the primary components of the Activity Distribution model of Attention. The combined Gaussian functions in the bottom portion of the figure represent the overall activity distribution of preparatory attention. The shallower Gaussian represents residual, inter-trial activation while the steeper Gaussian represents top-down inputs from knowledge and expectation. The selective channel is represented by the gray cylindrical volume.  Activity distributions of preparatory attention are theorized to build up in the spatial location maps based in the posterior parietal cortex, and are derived from two primary sources: The first source is residual activation from recent target locations, stored in the short-term memory store. This residual activation is the primary determinant of the spatial extent (but not the height) of the activity distribution. Furthermore, the shape of this memory-based component of the activity distribution remains relatively constant across trials in an experiment since no individual trial or target onset strongly influences it. Finally, the memory-based distribution is relatively enduring since it depends on the memory of recently activated locations – information which builds up across trials. The second source of preparatory attention is top-down goal- and 165  expectation-related activation. Activity from this component governs the overall height of the activity distribution and is very slow to decay. In addition to the preparatory attention component represented by the activity distribution, attention also contains a selective component, which is instantiated in the form of a selective channel of attentional resources (see Figure 6.2). The selective channel opens and closes on a moment-to-moment basis at locations within the area of preparatory attention defined by the activity distributions. This narrow channel is the source of the focused attentional resources required for the information at one specific spatial location to reach higher levels of processing.  The Deployment of Focal Attention According to the Activity Distribution Model The Activity Distribution model (LaBerge & Brown, 1989; LaBerge 1995) provides a quite different explanation of the mechanisms of attention shifts. Instead of assuming a single focus of attention that is moved from one location to another in either an analog or a discrete manner as had been assumed by proponents of the spotlight metaphor and its descendants [e.g., Eriksen & Yeh’s (1985) zoom lens model; Posner, Snyder, & Davidson (1980); and Sperling & Weichselgartner; 1995; but see also Shulman, Remington & MacLean (1979); Downing & Pinker’s (1985) gradient model, among others], the Activity Distribution model proposes that an attention shift involves a single channel of selective attention closing at one location and simultaneously opening in another6. The speed at which this redeployment of the selective  6  Note that this is similar to the proposal in the Episodic Theory of attention that focal attention shifts in a quantal manner by closing at one location and simultaneously opening at another. In the Episodic theory, however, the shift is not absolutely discrete, and there is a delay in the closing and opening processes (as described mathematically by the transition function) which results in both locations being partially attended simultaneously. In the Activity Distribution model, the opening of the selective channel at a new location may also not happen simultaneously, although for a different reason. Specifically, the height of the activity distribution must reach some threshold level  166  channel can occur and what objects or locations are most likely to trigger the redeployment of attention can be understood in terms of a fluid interaction between preparatory and selective attention. It may be best to consider how the Activity Distribution model explains attention shifts by means of an illustratory paradigm. Specifically, consider a cueing paradigm in which the observer was presented with a warning cue, followed by a target to be identified. In this particular paradigm (as described in LaBerge & Brown, 1989), the warning cue was a string of alternating digits with a single central letter. The observer was instructed to attend in a highly focused manner to that central letter and to identify it and keep it in mind until the end of trial as the final response depended on the identity of the letter. For example, the observer might be presented with a horizontal series of alternating 5’s and 8’s, with a central “S” (e.g., 585858S585858). Immediately following the warning cue, a second target appeared at one of five possible locations along the horizontal display: center, middle left or right, or extreme left or right. The second target was always a letter triplet (e.g., VRY). Although the flanking distractor letters varied, the central letter was always one of three letters: R, K, or P. These particular letters were chosen because they can only be differentiated under conditions of highly focused attention – their correct identification depends on the conjunction of two critical features. The overall task of the observer was to press a response key if and only if the first target was an “S” and the second target was an “R”. In order to be able to correctly respond to this “S - > R” sequence (in which the S was always presented centrally and the R always presented in one of five horizontal locations), focal attention must necessarily be shifted away from the before the selective channel can open at that location – if the preparatory attention levels at the area the channel is being shifted to are below threshold, there will be a delay while the activity distribution builds to threshold at that location. Unlike the episodic theory, however, this does not result in two locations being attended at the same time.  167  central cue to the location of the second target. If the second target appeared at the extreme left or right, the distance to be shifted was relatively large whereas if the second target appeared in the middle left or middle right, the distance was relatively short. If the second target also appeared at the central location, this was designated a control, “no shift,” condition. According to the Activity Distribution model, all components of attention play a role in this attention-shifting paradigm. At the most basic level, there is a broad, relatively shallow distribution of preparatory attention derived from memory-based inputs of previous target locations. Since the second target can appear anywhere along the horizontal string of elements, this component of the preparatory attention distribution would broadly encompass the entire length of the string. The second component of preparatory attention is the distribution based on the expectancies in each particular trial. The observer knows that the first target, contained within the warning cue, will always appear at the central location. Given this, a distribution of preparatory attention based on this expectation will build rapidly at the central location in anticipation of the target onset. The longer the duration of the warning cue, the more time is available for attention to be intensified to a high level at that location. The bulk of this preparatory activity will build at the central location (where the first target is expected to appear). However, because the activity distribution is Gaussian in nature and therefore has a gradual rather than an abrupt edge, the locations adjacent to the center will also receive some preparatory activity, although to a much letter extent. When the warning cue actually appears, a channel of selective attention opens at the central location so that the “S” can be identified. Once this highly-focused, narrow channel has opened at the first-target location, target identification is enabled and the first part of the task is complete. This, however, is merely the allocation of  168  attention, not its shifting – shifting is only instigated once the first target has been identified and the second target appears. In order for the second target (i.e., the central letter in the triplet) to be identified accurately, the channel of selective attention must be shifted from the location of the first target to the location of the second target (except, of course, in those trials in which the second target appears in the central location). According to the Activity Distribution model, the selective channel can open anywhere within the area of preparatory attention and the process of shifting is the process of simultaneously closing the channel at one spatial location and opening it in a new location. There is, however, a cost to this shifting process in some situations because in order for the selective channel to open, the underlying preparatory attention must be at a threshold level. Given this there is a distance-dependent cost to attention shifts in that the closer in space the second target is to the first, the greater the amount of preparatory attention at that location, and the more rapidly the selective channel will reach that threshold level. Effectively, this means that there will be a gradual increase in reaction time and decrease in identification accuracy as the second target is moved further away from the first. With regards to the LaBerge and Brown (1989) paradigm, this will result in a V-shaped curve in which reaction times are fastest at the central location and slowest both extreme locations, with intermediate reaction times to intermediate locations. At a more basic level, the differences in speed of processing or responding indicated by the V-curve are caused by the amount of time required for the dorsolateral prefrontal cortex to project the required amount of activity to the activity distribution building in the posterior parietal cortex. Another factor which influences the distance-dependent effects of spatial attention shifts is the relative target-onset time. Specifically, if the time between the onsets of the two targets is  169  long, the preparatory attention peak will decay, negating the advantage for second targets that appear in close spatial proximity to the first target. The decay of the preparatory attention distribution is reflected in the time-course of RTs in spatial cueing paradigms. At relatively short stimulus-onset-asynchronies (SOA), there is a strong facilitation effect on RTs for targets that appear in the same spatial location as the preceding spatial cue. The effect fades as the SOA increases, however, and ultimately even changes to inhibition such that responses to targets that appear at the same location as the cue are actually slower than responses to targets that appear at uncued locations (Klein, 1988, 2000; Klein & McInnes, 1999; Maylor, 1985; Maylor & Hockey, 1985; Posner & Cohen, 1984). Several strong tests of the Activity Distribution model have been made by LaBerge and colleagues (e.g., LaBerge & Brown, 1986; 1989; LaBerge, Carlson, Williams, & Bunney, 1997) as well as by other researchers (e.g., Pratt & Quilty, 2002; Turk-Browne & Pratt, 2005). LaBerge et al. (1997) used a three-target paradigm in order to pit the Activity Distribution model against the spotlight model. The paradigm they used was similar to that described above. There was an initial warning signal consisting of a horizontal series of items and a central item that differed from the others. The warning signal was followed by the first target, which was comprised of alternating letter pairs with a central target (e.g., GQGQGQGQGQ). The second target was a letter triplet (e.g., VRV) which could appear in one of three locations: central, left, or right. This second target was followed by a third target, which was also a letter triplet, and which could appear at one of five locations: central, left (same as second-target), right (same as second-target), or in the left or right space between the second target and the central location. In this paradigm, the activity distribution and moving spotlight models make distinct predictions regarding the fate of the third target.  170  The predictions of the spotlight model of attention are as follows. First and foremost, since the spotlight is moved to the location of the second target so that target may be accurately and rapidly identified, it is assumed that identification of the third target will be fastest if it appears in the same location as the second target, regardless of where in the display the second target appeared. The fate of the third target when it appears at a location different from that of the second depends on whether one assumes the spotlight moves in an analog or a discrete manner. If one assumes an analog spotlight (e.g., Eriksen & Murphey, 1987; Shulman, Remington, & MacLean, 1979; Tsal, 1983), then the further away from the second target the third target appears, the slower should be the response to the third target. For example, if the second target appears to the left of the display, there should be a linear increase in reaction times as the third target is presented further and further to the right, and vice versa if the second target appears to the right. If, on the other hand, one assumes a discrete (quantal) spotlight (e.g., Briand & Klein, 1987; Weichselgartner & Sperling; 1985), reaction times to the third target should be fastest at the location of the second target, and then equally slower at all other locations. In contrast to the spotlight model, which predicts that reaction times to the third target should be fastest when it appears in the same location as the second target, the Activity Distribution model predicts that the shortest reaction times will be to third-targets that appear in the central location, regardless of where the second target appears. The reasoning behind this prediction pivots on the fact that the channel of selective attention is not the determining factor for response-speed since when the channel closes in one location it re-opens simultaneously in a new location (regardless of whether that location is near or far). Since distance between the where the selective channel is currently (i.e., at the location of the second target) and where the  171  channel is moving to (i.e., the location of the third target) is irrelevant, there is no advantage to having the third target appear in the same location as the second. The factor which governs the speed of responses to targets is the distribution of preparatory attention. The simple reason for this is that it is the height of the preparatory distribution at any particular location that determines whether or not the channel of selective attention, once opened there, reaches threshold levels immediately, or whether more activation must first build up before the threshold can be reached (a process which takes time to occur). Given that the preparatory distribution builds such that its height is greatest at the location of the first target (the central location), all targets appearing at this location will have the fastest reaction times throughout the trial. The further away from the central location any target appears, the less preparatory attention will be at that location, and the longer it will take for the selective channel to build to threshold levels. Although the description thus far predicts a perfect V-shaped response function with RTs increasing equally as the distance to either the left or the right of centre increases, there is one small addition which must be made in order to properly represent the model’s predictions. Specifically, because there is relatively little time between the onsets of the second and the third targets, there will still be some residual activity from the second target to influence the speed of reactions to the third target. Hence, if the second target appeared to the left-hand side of the display, reaction times to third-targets on the left will be somewhat faster than third-targets on the right. This influence, however, is relatively slight, and the fundamental determinant of the response speed to the third target is distance from center. The predictions from the Activity Distribution model and the analog and discrete spotlight models are clearly distinct from one another and hence this paradigm provides a strong test between these hypotheses. LaBerge et al. (1997) found a clearly V-shaped set of response  172  curves which closely match the predictions from the Activity Distribution model. The results cannot be explained by a spotlight model of attention, whether movement is assumed to be discrete or analog, unless one assumes that the speed at which the spotlight moves varies with the proximity of the targets: the speed of movement is faster for close targets and slower for far targets. Since there is no strong evidence for this proposal7, LaBerge et al. conclude that their results provide strong evidence for the Activity Distribution model of attention shifts. Incorporating a Focus of Attention that Expands and Contracts Activity distributions are simply representations of the amount of neural activity corresponding to a particular region of the visual field. As such, if the area of stimulation encompasses a larger region of the visual field, then the corresponding representation in the cortex will also be larger and the activity distribution will be broader. To that extent, activity distributions can be unitary or multiple and broad or narrow in their spatial extent. LaBerge directly states that residual activation determines the spatial extent of activity distributions. Residual activation is the activation from recently attended locations which are stored in the short-term memory store. As such, it is not a feasible mechanism for underlying an expanding and contracting focus of attention – it is based on past experiences, not the task being completed at the moment. In fact, given LaBerge’s conception of how visual attention functions, it is the selective channel which must vary in spatial extent, not the underlying activity distribution. To this end, one may make an assumption about the relationship between the activity distribution and the selective channel which would enable a close modeling of a focus of attention that expands and contracts. Namely, one must assume that the selective channel  7  Note that this argument has been put forward by several researchers including Remington and Pierce (1984) and Kwak, Degenbach, and Egeth (1991), both of whom proposed that attentional movement time is invariant with distance because the attentional mechanism adjusts its velocity in proportion to the distance to be moved.  173  changes in size along with the spatial extent of the underlying activity distribution. That is, if there is a broad underlying distribution, the selective channel will have a large diameter whereas a narrow underlying distribution will lead to a narrow selective channel. La Berge and Brown (1990) have provided data which supports this interpretation, although they do not explicitly state this assumption. Incorporating a Flexibly Divided Focus of Attention The Activity Distribution model can readily incorporate a flexible focus of attention that is either unitary or divided depending on task demands. Activity distributions are simply representations of the amount of neural activity corresponding to a particular region of the visual field. As such, if neural activity builds simultaneously in response to more than one visual stimulus, then multiple activity distributions will also build simultaneously. According to the Activity Distribution model, once a distribution of attention reaches threshold levels, the selective channel is deployed to that location, enabling the item at that location to be processed. Consider a situation in which two locations are sufficiently salient and expected that the activity distributions at both locations exceed threshold levels – in this case, the stage is set for the division of focal attention, and two selective channels will be deployed simultaneously, one to each activity distribution. If on the other hand, only one location exceeded threshold levels (or, perhaps if two locations exceed threshold levels but there is a large difference in the extent to which they surpass threshold), only one selective channel will be deployed and focal attention will be unitary. The balance between the relative height of the activity distributions (which themselves stem from both top-down and bottom-up activity) provides a mechanism for modelling the task-dependent division of focal attention.  174  Despite the ease with which the task-dependent division of focal attention can be modelled, it is important to note that the Activity Distribution model can usefully accommodate a divided focus of attention if and only if one assumes the existence of multiple selective channels. Without that assumption, the selective channel becomes a limiting factor equivalent to the “spotlight of attention,” and the activity distributions serve only to guide the spotlight’s deployment to the most salient locations. Simply modeling this process is, of course, no small feat since the formation of these distributions stems from the combination of both top-down and bottom-up factors, a fact which no other model of attention has incorporated so naturally and easily. In order for this model to accommodate a truly divided focus of attention one must assume that multiple selective channels can be deployed simultaneously, an assumption which in no way compromises the model. Continuing on the assumption that there can be multiple selective channels, consider how the Activity Distribution model can account for the findings reported in Experiments 5.1 and 6.1. Observers were presented with two streams of distractor items, one to the left and one to the right of fixation. Half of the observers knew that the first-target-pair would appear in the streams (predictable condition) while the remaining observers knew that the first-target-pair would appear unpredictably either in the streams on in the region between the streams (unpredictable condition). The second-target-pair always appeared equally but unpredictably in the streams or between them. There are two “base” components to all activity distributions: a component based on inter-trial effects (i.e., where the targets appeared across the previous trials) and a component based on expectation or top-down knowledge. In addition to these is a bottom-up component that builds rapidly when an item actually appears. Since this component is constant between  175  conditions, it will not be discussed here. In the unpredictable condition, the targets appeared equally at all of the locations across trials, and hence the component of the activity distribution relating to inter-trial effects should be flat in this condition. The component relating to knowledge and expectation should also be flat since the observer knows that the targets will appear equally and unpredictably at all of the locations. Since the overall activity distribution is flat with all locations exceeding threshold, a single broad selective channel will be deployed (note that in order for the model to explain these data, one must make the assumption spelled out in the section above that the width of the selective channel matches the spatial extent of the underlying activity distribution). In the predictable condition, the inter-trial component is flat as it was in the unpredictable condition. The component relating to expectation and knowledge, however, now has a bimodal distribution (or alternatively, two expectation components build) since the observers know that the first-targets will appear in the streams, expectation builds at those locations and not in the central regions. Since the activity distribution is now bimodal and only the two peaks exceed threshold levels, two selective channels will be deployed and focal attention will be “divided.” Clearly the Activity Distribution model can readily account for why focal attention might be unitary in some situations and divided in others – in fact, it not only accounts for it, but also provides a mechanism by which this occurs. One limitation of the model, however, is that it cannot readily explain why even under conditions which make a divided focus of attention the most efficient, focal attention is first deployed as a single unitary focus and then divided within approximately 100 milliseconds. In Experiment 5.1, for example, in the predictable condition, observers knew in advance that the two first targets would always appear simultaneously in the RSVP streams – never in the region between the two streams. Given that the residual activity from previous trials would be greatest  176  at the two streams (because both T1 and T2 could appear in the streams while only T2 could appear in the area between the streams) and the activity stemming from observer expectation would also be greatest at the two outside locations, a bimodal distribution would result. Given this distribution of attention, two selective channels should open at the two locations that exceed threshold, and this should be an immediate process – there is no reason why a single large selective channel should open and then subsequently divide. Activity distributions are inherently sensitive to time because they are a representation of neural activity which fluctuates on a moment-to-moment basis. The activity distributions, therefore, should be represented as continually building and decaying over time, especially the components based on bottom-up input from the visual field. This, however, is not sufficient to explain why focal attention is initially deployed as a single focus and then is gradually divided in about 100 milliseconds. Despite the fact that the Activity Distribution model does not provide a clear rationale for why focal attention is initially deployed as a unitary focus and divided within about 100 ms when the task-demands require, it clearly provides the most flexible model of attention to date. Critically, it provides a mechanism by which focal attention can be either unitary or divided based on task demands, and can readily account for most of the requirements of a divided focus of attention. Models have played an important role in attention research, indeed in all cognitive research. As fast as models are developed, however, new data is provided which challenges those models and requires them to add new components or to change existing ones. In the previous sections for this chapter, I have outlined in some detail three current theories of visuospatial attention and considered whether they can account for the evidence showing that a) the focus of attention can be rapidly shrunk and expanded, b) that the focus of attention can be  177  flexibly deployed as either a unitary or divided focus, and c) focal attention can change from unitary to divided over the duration of a task. Although each of the three models provides critical insights into visuospatial attention and each uniquely explains at least one component of attention shifts, each also has its limitations. In general, the Activity Distribution model appears to be the most flexible model to date, and with some untested assumptions (as described in previous sections) being made, can quite readily accommodate a focus of attention that shrinks/expands and is deployed as a unitary or divided focus depending on task demands. Future research, however, is needed to confirm these assumptions before they can be accepted.  Contributions and Implications A description of the spatiotemporal dynamics of attention is critical to our understanding of this most basic human cognitive function and is important for both conceptual and practical reasons. Conceptually, a deeper understanding of these dynamics may lead to more comprehensive and realistic theories of visuospatial attention. As is shown in the earlier sections of this chapter, theories of visual attention cannot yet accommodate a focus of attention that is flexibly deployed as either a unitary or divided focus according to task demands. Even less can they account for a focus of attention which is initially deployed as a unitary focus and then is divided as the task progresses. The mechanisms the models proposed to account for shifts of a unitary focus of attention cannot appropriately be applied to the spatiotemporal dynamics of attention as described in this thesis, and these mechanisms must either be adjusted or new mechanisms be proposed. At a broader level, the research presented in this thesis underscores the flexible, dynamic nature of the human visual attention system. As shown in Experiments 5.1 and 6.1, the focus of 178  attention can be deployed either as a unitary or a divided focus, depending on task demands. In addition, the focus of attention can morph from one mode (unitary) to the other (divided) as a task progresses. Clearly, there is need in attention research to consider not only the ultimate goal in a task, but also the intermediate steps and to appreciate that attention changes fluidly on a moment-to-moment basis. Future research on visual attention must take into account that it is critical to test attention as a function of time – it is a highly dynamic process and in understanding its functioning, a snapshot is of less use than a video clip. There are many strengths of the method developed in Chapter 2 for assessing the rate at which the focus of attention is adjusted across space and time. First, it is a very simple behavioural methodology that is well suited to a range of populations. It has already been used to assess the rate of narrowing focal attention in older adults (Chapter 4), and since the attentional blink can be obtained with picture stimuli, it is possible to develop a language-free version of this task for young children or others with language impairments (e.g., individuals with dyslexia, specific language impairments (SLI), or autism). Overall, the behavioural procedure developed in Chapter 2 provides a very precise methodology for determining the time-course of changes in visuospatial attention. The rate at which the focus of attention narrows/expands or divides can be tested with very fine incremental control over the space and time continuums. Experiment 2.1 shows clear differences, for example, between 53, 66, and 80 ms SOAs, which are separated by a mere 12 or 13 milliseconds.  Possible Limitations The main paradigm used in this research to assess the spatiotemporal dynamics of visual attention is the attentional blink (AB). A further, independent component, Lag-1 sparing, is also 179  essential for tracking changes in the focus of attention in the paradigm developed in Chapter 2. The generalizability of this behavioural procedure therefore depends on certain limits of the attentional blink paradigm. For example, in normal adults, once the SOA between successive items in the RSVP stream exceeds a certain length, the attentional blink no longer occurs as observers are at ceiling in identifying the target. Hence, this paradigm can be used to probe only relatively short time-frames, although within those time-frames it provides a very sensitive test. A second limit on the attentional blink paradigm which would influence this paradigm is the type of stimuli that can be used to generate an attentional blink. Thus far it is clear that simple stimuli such as letters, digits, and patterns yield an attentional blink (e.g., Chun & Potter, 1995; Jefferies, Ghorashi, Kawahara, & Di Lollo, 2007; Raymond, Shapiro, & Arnell, 1992; Visser, Bischof, & Di Lollo, 1999). The attentional blink can also be obtained with pictures (Potter, Staub, Rado, & O’Connor, 2002), faces (Marois, Yi, & Chun, 2004), words (Davenport & Potter, 2005) targets identified by colour (Dell'Acqua, Sessa, Jolicœur, & Robitaille, 2006), and even with a search array as the second target (Ghorashi, Enns, Klein, & Di Lollo, submitted). Although the range of stimuli that can be used to generate the attentional blink is broad, there are limits. For example, it has been shown that Lag-1 sparing does not occur if there are changes in two or more dimensions between the targets (Visser, Bischof, & Di Lollo, 1999). Since Lag-1 sparing is essential to tracking changes of focal attention in this paradigm, this would limit the range of situations in which this paradigm would be of use. A final limitation on this paradigm stems from the need to process rapidly sequential items. This taxing procedure may not be feasible for many special populations of individuals. In this case, even though they may be able to efficiently shift or expand the focus of attention, this particular procedure will be unusable.  180  Future Directions The results of the studies described in this dissertation open a number of avenues for future research. First, the research has direct implications for the neurophysiological mechanisms that mediate the spatiotemporal dynamics of visual attention. Therefore, brain imaging and neurophysiological measure should be coupled with behavioural measures to provide a more comprehensive understanding of the rules that govern the spatiotemporal dynamics of attention. Specifically, future research should be directed towards providing electrophysiological and brain-imaging evidence to confirm the descriptive model outlined in Chapter 2. By using electrophysiological tools such as electroencephalography (EEG) or the more specific recording of event-related potentials (ERPs), a very temporally precise mapping of the expanding and contracting of focal attention could be provided. In addition, ERPs could confirm the conditions under which a unitary or a divided focus is deployed. Specifically, comparing the magnitude of the P1 ERP component to probe stimuli appearing at attended locations (i.e., in the RSVP streams) to probe stimuli appearing between the streams (as in Experiment 5.1) should result in components of equal magnitude if a unitary focus is deployed and reduced P1 magnitude for the intervening location if a divided focus is deployed. This methodology could be used to better establish that focal attention is initially deployed as a unitary focus that subsequently divides over approximately 100 ms. Second, future research should be directed at probing the possibility that expanding/contracting and/or dividing focal attention involves a disengage-change-engage process, as outlined in the section above discussing the Human Attentional Network model. If this is indeed the case, very specific predictions can be made about the expected patterns of activation. For example, if one were to combine Experiment 2.1 with a neuroimaging technique 181  such as magnetoencephalography (MEG), which is both spatially and temporally precise, one would expect to see a burst of Parietal activity immediately upon the appearance of the first target. This would indicate that once the first target appears, the focus of attention is disengaged from its broad focus setting so that the narrowing process can begin. Similarly, if disengagement is required prior to dividing a unitary focus of attention, there should be parietal activation prior to the division of focal attention. Another avenue for future research lies in understanding the attention deficits present in developmental disorders such as autism. It is known, for example, that individuals with autism are less able to rapidly and flexibly adjust their focus of attention (Allen & Courchesne, 2001; Belmonte, 2000; Burack, 1994; Townsend et al., 2001; Townsend, Courchesne, & Egaas, 1996). Given this, whereas normal adults may rely on their ability to flexibly adjust focal attention according to momentary task demands (as demonstrated in Experiments 5.1 and 6.1), this may not be the case for individuals with autism. In this case, when faced with a task like that in Experiment 5.1 and 6.1, individuals with autism, who cannot readily adjust their focus of attention, may simply deploy a divided focus from the outset or may alternatively deploy a unitary focus that never divides. Attentional deficits are a great liability in developmental disorders – the quality of information entering the brain determines the content that the brain has available to work with. As such, if attention is shifted, adjusted, or deployed inefficiently, there will inevitably be behavioural consequences, and a precise mapping and understanding of attention deficits in disorders such as autism are critical for both our understanding of the disorder and in terms of developing treatment strategies.  182  References Allen, G., & Courchesne, E. (2001). Attention functions and dysfunctions in autism. Frontiers in Bioscience, 6, 105-119. Belmonte, M. (2000). Abnormal attention in autism shown by steady-state visual evoked potentials. Autism, 4, 269-285. Briand, K.A., & Klein, R. (1987). Is Posner’s ‘beam’ the same as Treisman’s ‘glue’? On the relation between visual orienting and feature integration theory. Journal of Experimental Psychology: Human Perception and Performance, 13, 228-241. Burack, J. (1994). Selective attention deficits in persons with autism: Preliminary evidence of an inefficient attentional lens. Journal of Abnormal Psychology, 103, 535-543. Castiello, U., & Umiltà, C. (1992). Splitting focal attention. Journal of Experimental Psychology: Human Perception and performance, 18, 837-848. 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, 21, 109-127. Corbetta, M., Kincade, J.M., Ollinger, J.M., McAvoy, M.P., & Shulman, G. (2000). Voluntary orienting is dissociated from target detection in human posterior parietal cortex. ature euroscience, 3, 292-297. Davenport, J.L., & Potter, M.C. (2005). The locus of semantic priming in RSVP target search. Memory & Cognition, 33, 241-248. Dell’Acqua, R., Sessa, P., Jolicœur, P., & Robitaille, N. (2006). Spatial attention freezes during the attentional blink. Psychophysiology, 43, 394-400. Downing, C. J., & Pinker. S. (1985). The spatial structure of visual attention. In M. I. Posner & 183  O. S. M. Marin (Eds.), Attention and performance XI (pp. 171-187). Hillsdale, NJ: Erlbaum. Eriksen, C.W., & Murphey, T.D. (1987). Movement of attention across the visual field: A critical look at the evidence. Perception and Psychophysics, 42, 299-305. Eriksen, C.W., & St. James, J.D. (1986). Visual attention within and around the field of focal attention: A zoom lens model. Perception and Psychophysics, 40, 225-240. Eriksen, C.W., & Yeh, Y-Y. (1985). Allocation of Attention in the Visual Field. Journal of Experimental Psychology, 11, 583-597. Fernandez-Duque, D., & Posner, M.I. (1997). Relating the mechanisms of orienting and alerting.  europsychologia, 35, 477-486.  Ghorashi, S.M., Enns, J.T., Klein, R., & Di Lollo, V. (submitted). Spatial selection is separable from identity extraction in visual search. Grill-spector, K., & Kanwisher, N. (2005). Visual recognition: As soon as you know it is there, you know what it is. Psychological Science, 16, 152-160. Heinze, H-J., Luck, S.J., Munte, T.F., Gös, A., Mangun, G., & Hillyard, S. (1994). Attention to adjacent and separate positions in space: An electrophysiological analysis. Perception and Psychophysics, 56, 42-52. Jefferies, L.N., Ghorashi, S., Kawahara, J-i., & Di Lollo, V. (2007). Ignorance is bliss: the role of observer expectation in dynamic spatial tuning of the attentional focus. Perception & Psychophysics, 69, 1162-1174 . Jefferies, L.N., Wright, R.D., & Di Lollo, V. (2005). Inhibition of return to an occluded object depends on expectation. Journal of Experimental Psychology: Human Perception and  184  Performance, 31, 1224-1233. Klein, R. M. (1988). Inhibitory tagging system facilitates visual search. ature, 334, 430–431. Klein, R. M. (2000). Inhibition of return. Trends in Cognitive Sciences, 4, 138–147. Klein, R. M., & MacInnes, W. J. (1999). Inhibition of return is a foraging facilitator in visual search. Psychological Science, 10, 346–352. LaBerge, D. (1995). Attentional Processing: The Brain’s Art of Mindfulness. Cambridge, MA: Harvard University Press. LaBerge, D., & Brown, V. (1986). Variations in size of the visual field in which targets are presented: An attentional range effect Perception and Psychophysics, 40, 188-200. LaBerge, D., & Brown, V. (1989). Theory of attentional operations in shape identification. Psychological Review, 96, 101-124. LaBerge, D., & Buchsbaum, M.S. (1988). Attention filtering and the pulvinar: Evidence from PET scan measures. 29th annual meeting of the psychonomic society, p. 4, abstract. LaBerge, D., Carlson, R.L., Williams, J.K., & Bunney, B.G. (1997). Shifting attention in visual space: Tests of moving spotlight models versus an Activity Distribution model. Journal of Experimental Psychology: Human Perception and Performance, 23, 1380-1392. Marois, R., Yi, D-J., & Chun, M.M. (2004). The neural fate of consciously perceived and missed events in the attentional blink. Neuron, 41, 465-472. Maylor, E., & Hockey, R. (1985). Inhibitory component of externally controlled covert orienting in visual space. Journal of Experimental Psychology: Human Perception and Performance, 11, 777-787. Maylor, E. (1985). Facilitatory and inhibitory components of orienting in visual space. In M. I.  185  Posner & O. S. M. Marin (Eds.), Attention and performance XI: Mechanisms of attention (pp. 189–203). Hillsdale, NJ: Erlbaum. Petersen, S.E., Robinson, W.J., & Keys, W.J. (1985). Pulvinar nuclei of the behaving rhesus monkey: Visual responses and their modulation. Journal of europhysiology, 54, 867886. Posner, M.I., (1992). Attention as a cognitive and neural system. Current Directions in Psychological Science, 1, 11-14. Posner, M.I., Choate, L., Rafal, R.D., & Vaughan, J. (1985). Inhibition of return: Neural basis and function. Cognitive europsychology, 2, 211-228. Posner, M. I., & Cohen, Y. (1984). Components of visual orienting. In H. Bouma & D. G. Bouwhuis (Eds.), Attention and performance X: Control of language processes (pp. 531– 556). Hillsdale, NJ: Erlbaum. Posner, M.I., & Peterson, S.E. (1990). The attention system of the human brain. Annual Review of euroscience,13, 25-42. Posner, M.I., Petersen, S.E., Fox, P.T., & Raichle, M.E. (1988). Localization of cognitive operations in the human brain. Science, 240, 1627-1631. Posner, M.I., Sheese, B.E., Odludaş, Y., & Tang, Y.Y. (2006). Analyzing and shaping human attentional networks.  eural etworks, 19, 1422-1429.  Posner, M.I., Snyder, C.R.R., & Davidson, B.J. (1980). Attention and the detection of signals. Journal of Experimental Psychology: General, 11, 160-174. Posner, M.I., Walker, J.A., Friedrich, F.J., & Rafal, R.D. (1984). How do the Parietal lobes direct covert attention?. Journal of euroscience, 25, 135-145. Potter, M.C., Staub, A., Rado, J., & O’Conner, D.H. (2002). Recognition memory for briefly  186  presented pictures: The timecourse of rapid forgetting. Journal of Experimental Psychology: Human Perception and Performance, 28, 1163-1175. Pratt, J. & Quilty, L. (2002) Examining the Activity Distribution model of visual attention with exogenous cues and targets. The Quarterly Journal of Experimental Psychology, 55, 627-641. Rafal, R.D., & Posner, M.I. (1987). Deficits in human visual spatial attention following thalamic lesions. Proceedings of the ational Academy of Science, 84, 7349-7353. 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, 18, 849-860. Shulman, G.L., Remington, R., & MacLean, J.P. (1979). Moving attention through visual space. Journal of Experimental Psychology: Human Perception and Performance, 5, 522-526. Sperling, G. & Weichselgartner, E. (1995). Episodic theory of the dynamics of spatial attention. Psychological Review, 102, 503-532. Taylor, M. M. & Creelman, C. D. (1967). PEST: Efficient estimates on probability functions. The Journal of the Acoustical Society of America, 41, 782-787. Townsend, J., Courchesne, E. & Egaas, B. (1996) Slowed orienting of covert visual-spatial attention in autism: Specific deficits associated with cerebellar and parietal abnormality. Development and Psychopathology, 8, 563-584. Townsend, J., Westerfield, M., Leaver, E., Makeig, S., Jung, T-P., Pierce, K. & Courchesne, E. (2001). Event-related brain response abnormalities in autism: Evidence for impaired cerebello-frontal spatial attention networks. Cognitive Brain Research, 11, 127-145.  187  Tsal, Y. (1983). Movements of attention across the visual field. Movements of attention across the visual field. Journal of Experimental Psychology: Human Perception and Performance, 9, 523-530. Turk-Browne, N.B., & Pratt, J. (2005). Attending to eye movements and retinal eccentricity: Evidence for the Activity Distribution model reconsidered. Journal of Experimental Psychology: Human Perception and Performance, 31, 1061-1066. Visser, T. A. W., Bischof, W. F., & Di Lollo, V. (1999). Attentional switching in spatial and non-spatial domains: evidence from the attentional blink. Psychological Bulletin, 125, 458-469. Weichselgartner, E., & Sperling, G. (1987). Dynamics of automatic and controlled attention. Science, 238, 778-780.  188  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0067342/manifest

Comment

Related Items