Ординатура / Офтальмология / Английские материалы / Eye Movements A Window on Mind and Brain_Van Gompel_2007

this with a more objective form of assessment. We consider the state of eye-movement research (both in the present and the past) by reporting results from a survey sent to participants of ECEM12 and a journal database search. Based on this combination, our aim is to provide the reader with a synoptic overview of current and past developments across the field in general.

1. Overview of the parts in this book

1.1. History of eye-movement research

Looking backwards is the perfect way to begin a book on eye-movement research. Knowing the genesis and evolution of any subject is critical to understanding the motivation and driving forces (both psychological and technological) that lie behind scientific progress in the area. Our first part therefore examines the historical foundations of eye-movement research; something which has permeated through an extremely rich and diverse range of multidisciplinary empirical topics.

The opening chapter by Wade places the birth of the modern scientific study of eye movements firmly in the eighteenth century, before charting its historical progress via the thoughts and works of many generations of eminent scientists. Central to this was the growth in physiological experimentation, and the examination/exploitation of visual vertigo, torsion, and afterimages in particular. Wade also describes how through the ingenious use of technology, techniques for acquiring eye-movement records and studying scan paths were developed to facilitate reading research and the study of visual perception.

The immediate post-war period is the focus of Westheimer’s historical account. He documents how multidisciplinary scientific work forged under conditions of war later sparked a mini-renaissance in instrumentation and neurophysiological techniques. This naturally fed through into a greater understanding of the oculomotor system and helped establish the investigation of eye movements as a major tool in psychological research. Following Westheimer, Land’s exposition begins in the 1950s but rapidly brings us up to date. Again, he relates how the development of modern technology, choosing head-mounted mobile eye-trackers as his specific example, has led to advances across a range of eye-movement research areas. The flexibility afforded by such devices frees us from many constraints (often literally in the case of head-vices and bite-bars) and opens up the “real” world to greater investigation. Together, these three chapters make it clear that the act of looking backwards results in our forward vision becoming sharper and brighter.

1.2. Physiology and clinical studies of eye movements

The second part of the book deals with physiological and clinical studies of eye movements. First, Munoz, Armstrong, and Coe provide an authoritative overview on how

to use simple eye-movement tasks to diagnose both the developmental state and the possible dysfunctions of the cortical and subcortical oculomotor system. Their accumulator model allows an understanding of performance profiles across a wide range of deficits, such as Parkinson’s disease, Tourette syndrome, and Attention Deficit Hyperactivity Disorder.

In the second Chapter in this part, Ford, Brown, and Everling report an fMRI study which shows that anti-saccade task performance depends on specific brain activation prior to stimulus presentation. Their work describes the relationship between frontal brain activity and the accuracy of a subsequent saccadic decision. It also illustrates how eye-tracking can be combined with high-tech brain imaging methods that have become popular over the last two decades.

The final chapter in this part is by Bour, Van der Meer, and Van Mourik. These authors describe commutative eye rotations in congenital nystagmus (a fixation instability resulting from visual or oculomotor deficits) and a computational model that predicts anatomical aberrations in a patient with congenital nystagmus. Together, the contributions in this part illustrate how a multifaceted clinical and neurophysiological approach, spanning behavioural and brain imaging studies as well as computer simulations and patient studies, can help us understand the functioning of eye-movement control structures in the brain.

1.3. Transsaccadic integration

Part 3 focuses on a topic that has intrigued investigators for many years. As De Graef points out in the first chapter in this part, there has been an ongoing mystery associated with the question of how we “integrate” information across a temporally and spatially discontinuous train of “snapshots”. This mystery has been sufficiently deep to even encourage some to conclude that the question is a mistaken one, and in various important senses we do not in fact integrate information across successive fixations at all. But the chapters in this part suggest that the pendulum is now swinging back. De Graef argues that the bases behind that view are flawed and that there is good reason to consider the mechanisms involved in the integration of information across fixations during the process of scene and object recognition, especially as these relate to the role played by parafoveal preview.

In the second Chapter in this part, Koch and Deubel address the question of how we achieve a stable representation of the location of objects in the world across fixations. That is, how the visual system maps and re-maps the continually changing retinal input into a consistent and stable perceptual frame. They suggest that this is based on the use of “landmarks” in the visual field and demonstrate in a series of experiments that relatively low-level configurational properties of the landmarks drive the relocation process, with the visual system apparently operating on the (perfectly reasonable) assumption that the world is unlikely to have changed significantly while the eyes are being moved.

In the final chapter, Melcher and Morrone address questions concerning the integration of information related to both the location and the form of objects across fixations. Their

experiments provide evidence that visually detailed but temporally inexact information is retained across saccades, and they suggest that our subjective impression of a stable visual world is derived from a combination of a fast-acting saccadic re-mapping, reliant, again, on the assumption of invariance over time, and a more durable scene-based memory.

1.4. Modelling of eye movements

Modelling of eye movements has been one of the clearest growth areas in eye-movement research since the late 1990s (see section 2.3). It seems clear that great advances have been made in this area, at least in so far as this relates to reading, but this is certainly not an area without controversy. The nature of these controversies and some important issues arising from them are well documented in the introductory chapter by Radach, Reilly, and Inhoff. These authors provide a very useful overview and classification of model types and then proceed to highlight some of the important questions that models of eyemovement control in reading need to address. One of these is the role played by attention, but as Radach et al. point out, this concept itself is frequently underspecified and models need to be more explicit about the characteristics of attention that are assumed. They conclude the chapter with some proposals regarding model testing and some challenges for the modelling enterprise.

In the following chapter, Reichle, Rayner, and Pollatsek explore aspects of the E-Z Reader model in the context of lexical ambiguity. This serves two primary purposes: It allows additional investigation of the distinction proposed in the model between the “familiarity check” (L1) and the completion of lexical processing (L2) and it uses lexical ambiguity as a tool to investigate the way in which the model handles the effects of higher-level sentence context on word identification. The authors incorporate changes to E-Z Reader that closely resemble the assumptions made by the reordered access model of lexical ambiguity resolution, and thus this represents a step further along the path that they have noted for some time as a necessary extension to this model.

Yang and Vitu then concentrate on the modelling of saccade targeting procedures during reading. In contrast to the many models which propose a uniform, generally word-based, oculomotor targeting process, they show how well a model postulating a combination of targeting strategies performs. They suggest that strategy-based saccades of a relatively constant length are common at early time intervals during a fixation, while visually guided (word-based) saccades only become frequent after longer latencies. They conclude that eye guidance in reading is the result of a dynamic coding of saccade length rather than the sort of cognitively based aiming strategies related to the requirements of lexical identification and the shift of attention that are assumed by many of the current models.

In the final chapter of this part, Engbert, Nuthmann, and Kliegl discuss a model that does assume word-based targeting strategies – SWIFT – but they focus on the question of when those strategies do not produce the desired outcome and result in “mislocated fixations”. They use simulations with the SWIFT model as a way of checking an algorithm

that iteratively decomposes observed landing position distributions into mislocated and well-located cases, and conclude that “mislocated fixations” occur remarkably frequently, especially when the intended target is a short word. They conclude by discussing the link between mislocated fixations and the inverted optimal viewing position (IOVP) effect.

1.5. Eye movements and reading

One of the most productive areas in eye-movement research is the investigation of reading processes. This part begins with a comprehensive review of the research into human sentence processing that has arisen from studying eye movements whilst people read. Clifton, Staub, and Rayner have compiled a list of 100 key articles in this area of language processing. Beginning with word recognition, they proceed to summarise the results of studies investigating the effects of syntactic, semantic, and pragmatic factors on sentence comprehension.

Word recognition, or rather decomposition, is the focus of the chapter by Juhasz. She examines the role of semantic transparency in English compound words, embedded in neutral-context sentences. Her results suggest that there is an early pre-semantic decomposition of compound words during reading and that decomposition occurs for both transparent and opaque compounds. In their chapter, Bertram and Hyönä also investigate the reading of compound words, but in Finnish sentences. Using an eye-contingent paradigm (altering what is displayed on a computer screen depending on where the eye is looking or when it is moving), they manipulated the letter information available in the parafovea for words with long or short first constituents and conclude that there is no morphological preview benefit: morphological structure is only determined in the fovea.

White re-examines the hypothesis that foveal load (i.e. processing difficulty) modulates the amount of parafoveal (pre)processing that takes place. She finds that manipulating localised foveal load, in the form of orthographic regularity and misspellings, does not influence the probability of the next word being skipped. However, there is still an effect on reading times and so she concludes that this may be evidence for different, independent mechanisms influencing eye-movement control. The final chapter also exploits misspellings, but this time in the parafovea itself. Johnson explores non-adjacent transposedletter effects through the use of an eye-contingent boundary change of display. She shows that transposed-letter effects continue to occur even when the manipulated letter positions are non-adjacent and that they arise similarly across vowels and consonants.

1.6. Eye movements as a method for investigating spoken language processing

Part 6 concerns a relatively new area of eye-movement research; the use of eye-tracking as a tool for investigating spoken language processing. Research in this area took off following the seminal article by Tanenhaus, Spivey-Knowlton, Eberhard, and Sedivy (1995). What has come to be known as the visual world method – recording eye movements

directed to objects and scenes while the participant listens to spoken text is becoming increasingly popular. In his overview chapter, Tanenhaus, a pioneer in this field, provides an overview of the most important developments since the publication of their article. He also discusses data analysis techniques and assumptions underlying the link between fixation behaviour and language processing models. Finally, he compares the results from reading studies with studies on spoken language comprehension using the visual world method, and thereby also provides a link with Part 5.

The remaining chapters in Part 6 all report research using the visual world method. The results reported by Dahan, Tanenhaus, and Salverda provide important evidence against the view that participants in visual-world experiments covertly name the pictures prior to word onset. Furthermore, they show that fixations made to pictures of words that are auditorily presented are affected both by the amount of time that people can inspect the picture before the word is presented and by the position of the pictures in the display. Bailey and Ferreira provide an excellent example of how the visual world method can be used to investigate issues that are specific to spoken (rather than written) language comprehension by investigating the effects of disfluencies (uhh) on sentence comprehension. They show that disfluencies had no effect on both online and offline parsing preferences. However, there was evidence that listeners identified ambiguities during disfluencies.

Recently, researchers have also started to use the visual world method to investigate speech production. Wheeldon, Meyer, and Van der Meulen use this method to investigate speech planning when speakers make anticipation errors (e.g., they say whistle instead of ladder, whistle). They show that the order of fixating the objects is the same, regardless of whether the words are named in the correct order or whether an anticipation error occurred. This suggests that anticipation errors are not due to premature shifts in visual attention. In the last chapter of this part, Knoeferle employs the visual world method to investigate how information provided by the scene interacts with linguistic information provided in the sentence. Her experiment shows that sentence interpretation in German word order ambiguities is closely time-locked to the point in time when the combination of both linguistic and visual information disambiguates the sentence. Hence both linguistic and visual information crucially affect sentence comprehension.

1.7. Eye movements as a method for investigating attention and scene perception

The study of eye movements has been an extremely valuable tool for investigating attention and scene perception, and this part illustrates some of the recent research in this area. The part starts with a critique of the currently fashionable concept of “visual saliency” by Henderson, Brockmole, Castelhano, and Mack. They show that the pattern of eye movements during visual search in real-world scenes cannot be accounted for by visual saliency. This work is in conflict with the currently popular approach of predicting eye behaviour on the basis of image statistics, an approach that is illustrated in the following chapter by Underwood, Humphreys, and Cross. They report effects of semantic congruency between target and background, visual saliency, and the gist of a scene on the

inspection of objects in natural scenes. From these chapters it seems clear that the relative contribution of low-level visual factors and high-level cognitive factors to eye-movement control remains a topic of hot debate.

The following two chapters in this part examine the classical visual search task where observers look through a set of items in a display to determine whether a predefined target is present or absent. The two visual search studies clearly illustrate the benefits of adding eye-movement recording to this established line of research into covert attention and visual cognition. Hooge, Vlaskamp, and Over recorded fixation durations as a function of the difficulty of both the previous and the current item, as well as the expected difficulty of the search task, while the chapter by Shen, Elahipanah, and Reingold documents an important distinction between local context effects and higher-level congruency effects on eye-movement control in visual search. Both chapters resonate with the familiar theme that both top-down and bottom-up factors have an impact on eye behaviour. In the concluding chapter, Gareze and Findlay report that scene context had no effect on object detection and eye gaze capture in their study, thus making the case for a predominance of lower-level factors. This message conflicts with the results reported by Underwood et al., and the authors suggest that factors such as the eccentricity of objects from fixation and their size are important experimental variables that need to be explored in future research on eye-movement control during scene perception.

1.8. Eye movements in natural environments

With the development of new eye-trackers that allow participants to freely move around in their environment, one area that has seen a recent upsurge in interest is the investigation of eye movements in natural environments. In their chapter, Hayhoe, Droll, and Mennie provide an overview of these developments. They argue that fixation patterns are learnt behaviour: People have to learn the structure and dynamic properties of the world, the order in which a task is carried out, learn where to find the relevant information in the scene, and learn where to look to balance competing demands of the task. Therefore, as their literature review indicates, the nature of the task critically influences fixation behaviour. Pelz and Rothkopf provide one demonstration of this in their chapter. In their experiment, the extent to which people fixate on the path in front of them depends on whether they walk on pavement in a man-made environment or on a dirt path in a wooded environment. Furthermore, when standing still, fixation durations were longer in the wooded than in the man-made environment, although there was no difference during walking. Patla, Tomescu, Greig, and Novak also investigated fixation behaviour during walking. Their data suggest that there is a tight link between fixation behaviour and route planning. They present a model of route planning that is consistent with fixation behaviour. Finally, Tatler and Kuhn used eye-tracking to investigate how magicians misdirect observers. Their results show that observers follow the magician’s gaze during the trick, so their fixations are directed away from the disappearing object. Fixation behaviour was also affected by whether observers had previously seen the trick. But,

interestingly, even when observers watched the trick a second time and spotted it, they usually did not fixate on the disappearing object, suggesting that detection is linked to covert visual attention.

2. Questionnaire study and journal database search

The above overview gives an impression of current developments in eye-movement research, while the chapters in the remainder of this book describe those developments in more detail and refer to the latest research in these areas. However, any book is likely to underrepresent some areas of research while overrepresenting others. In order to gain a more objective overview across the area in general, in terms of both the current state of eye-movement research and the past developments, we carried out a questionnaire study and searched a database of journal articles.

For the questionnaire study, we sent out an e-mail to all attendants of ECEM 12 asking for their comments and impressions in response to a number of questions. This questionnaire was sent between March and April 2006. The attendants were asked the following four questions:

1.In what area (or areas) of eye-movement research do you work? (They could choose one of the parts in this book or specify a different area.)

2.List up to 3 eye-movement papers (in order) that you consider to have been most influential in the development of the field.

3.What do you consider to be the best journal(s) for keeping up to date with developments in eye-movement research?

4.What do you consider to be the most important development(s) in eye-movement research in the past 5 years (please use 1 or 2 sentences)?

Forty-four attendants (out of the approximately 200 who received the questionnaire) responded. Twenty-one of the respondents were authors of one of the chapters in this book.

For our database search, we used ISI Web of Knowledge (http://wos.mimas.ac.uk/), which contains articles from over 8600 journals published since 1970, of which 1790 journals are in the social sciences. This allowed us to search for journal articles with ‘eye(-)movement(s)’ in the title or keywords and for citation counts for selected eye-movement articles. Web of Knowledge also provides citation counts for selected

for ‘eye(-)movement(s)’ in the title and keywords. The searches were run on 12 and 15 May 2006.

Table 1 shows the number of respondents working in eye-movement research areas that correspond to the parts in the book. The majority of respondents were involved in research in either reading (21) or attention and scene perception (18). Both areas have traditionally been very productive in eye-movement research, and this is reflected here and in the chapters (Parts 5 and 7) in this book. There were also many respondents (15) who indicated that they were involved in the modelling of eye movements, an area which

Table 1

Number of respondents to the questionnaire working in the different eye-movement research areas

Note: Number of researchers: Number of respondents to the questionnaire who specified that they worked in each of the eye-movement research areas N = 44 .

has seen important developments in recent years, as can be seen in Part 4 of this book. There was also reasonable representation in the areas of physiology and clinical studies of eye movements (10) and spoken language processing (9). This latter area has grown dramatically since the mid-1990s, and the relatively high number of respondents working in the area reflects that recent growth. Physiology and clinical studies of eye movements has traditionally been a very strong area within eye-movement research in terms of number and impact of publications, and is therefore possibly somewhat underrepresented here. Fewer people indicated that they worked on transsaccadic integration (5), eye movements in natural environments (5), and the history of eye-movement research (2). This seems a fair representation of the number of people working in these areas, although of course that does not necessarily reflect on the importance of the research. Seven respondents indicated that they conducted a different type of research; eye-movement behaviour during driving was mentioned twice. A high number of respondents (24) indicated that they worked in more than one area, showing the highly cross-disciplinary nature of eye-movement research (and perhaps this bias in those who attend ECEM). Overall, these data suggest that the questionnaire results are likely to be fairly representative of the views of a broad cross section of researchers working in the area of eye-movement research.

2.1. Most influential and most cited publications

Table 2 shows the eye-movement publications that our respondents considered most influential in the development of the field. The eye-movement model of reading proposed by Reichle, Pollatsek, Fisher, and Rayner (1998) received most points, highlighting its importance for both reading and modelling research. However, it was not the paper that was most often mentioned as most influential. Both Findlay and Walker’s (1999) model of saccade generation and Rayner’s (1998) overview on eye movements during reading

Table 2

Eye-movement publications considered to have been most influential in the development of the field by respondents to the questionnaire

Note: Points: For each respondent N = 44 , the most influential paper counted as 3 points, the second most influential paper as 2 points, and the third most influential paper as 1 point.

Times mentioned as most influential: Number of respondents who indicated that this publication has been most influential (i.e. it received 3 points).

Citation count: Number of citations (including incorrect citations) in all databases in Web of Knowledge, searched on 12 May 2006.

Area: Area according to the classification in Table 1.

and information processing were mentioned more frequently as the most influential paper. Tanenhaus, Spivey-Knowlton, Eberhard, and Sedivy (1995) was equal to Reichle et al. (1998) in terms of this measure of importance. Most of the publications in Table 2 are concerned with reading research (5) or attention and scene perception (3); perhaps not surprisingly, given that many respondents worked in these areas. In contrast, the history of eye movements, physiology and clinical studies, and transsaccadic integration are not represented in the table, and this may also be related to the background of the respondents.

It is interesting to see whether those publications considered influential by specialists in the field of eye movements are also the ones most often cited. Or to put it another way, are the most-cited publications also considered to be most influential? Citation counts are often used as a means of quantifying the impact of publications, so we should expect to see a good correlation between these two measures.

In order to determine whether the publications that were considered to be most influential also had high citation counts, we conducted a search in Web of Knowledge, using all available databases: science, social sciences, and arts and humanities. In this citation count, we included incorrect citations where, for example, the exact title of the book was wrong, as long as the authors’ surnames and year of publication were correct. As shown in Table 2, all publications in the top 10 received a fairly high number of citations, but there is no clear correlation between the citation count and the rankings.

Note: Points: For each respondent (N = 44), the most influential paper counted as 3 points, the second most influential paper as 2 points, and the third most influential paper as 1 point.

Times mentioned as most influential: Number of respondents who indicated that this publication has been most influential (i.e. it received 3 points).

Citation count: Number of citations (including incorrect citations) in all databases in Web of Knowledge, searched on 12 May 2006.

Area: Area according to the classification in Table 1.

Of the 10 publications in Table 2, the two that were considered most influential have the fewest citations, while Yarbus’s (1967) book, which was ranked sixth, had by far the most citations (1049).

To further explore this question, we have listed in Table 3 the 10 most cited publications from all those mentioned by at least one of the respondents to the questionnaire. Four of the publications that were in the top 10 in Table 2 are also in Table 3, suggesting at least some correspondence between citation counts and publications that specialists considered influential. However, the other six publications were only mentioned by one respondent, including Robinson’s (1963) article on using a scleral search coil for measuring eye movements, which received by far the highest number of citations. Furthermore, the two most influential articles in Table 2 (Findlay & Walker, 1999; Reichle et al., 1998) do not even reach the top 10 in Table 3.

Table 4 confirms that the most cited publications are not always seen as most influential. In this table, we have listed the 10 most cited journal articles (according to Web of Knowledge) that have ‘eye(-)movement(s)’ in their title or keywords. Of course, this search is somewhat limited, since some eye-movement articles do not have ‘eye(-)movement(s)’ in their title or keywords. However, it does enable us to determine whether articles that are well cited are considered to be influential by specialists in the area. Clearly, this is not the case: of the 10 most cited articles, only one – by Duhamel, Colby, and Goldberg (1992) – was mentioned, and this article was only mentioned as the third most influential by a single respondent.