Littérature scientifique sur le sujet « Eye movements in reading »

Despite a large and established literature, studies of eye movements during skilled reading generally assume uniformity at the participant level. However, there is growing evidence that individual differences in reading proficiency among skilled adult readers modulate the early stages of lexical processing. Measures of lexical knowledge have also been shown to be more predictive of differences in eye movement patterns than many word- or sentence-level variables that have traditionally been the focus of research in the field. In five experiments, large samples of skilled readers were assessed on measures of reading and spelling ability in order to test the hypothesis that individual differences in proficiency, and, particularly, the precision of a reader’s lexical representations, modulate eye movements during reading. Gaze-contingent display change paradigms were used to manipulate parafoveal information during sentence reading, tapping the early stages of word identification. The combination of high reading and spelling ability, i.e., lexical expertise, was found to consistently predict both the spatial extent and depth of parafoveal processing. Lexical experts made use of a wider perceptual span and were more likely to extract lexical information from upcoming words than readers with imprecise lexical knowledge. Lexical expertise was also associated with more effective integration of parafoveal and foveal information and more immediate comprehension. These results support the lexical quality hypothesis of reading skill and challenge the assumption that skilled adults all read in essentially the same way. The results also provide insight into the role of parafoveal processing during reading and provide opportunities for the refinement of computational models of eye movement control.

Miscue analysis and eye-movement recording technology are combined in this dissertation to explore the reading processes of adult, skilled readers. The combination of approaches forms a new reading research methodology termed Eye Movement Miscue Analysis, or EMMA, that provides a powerful view of the reading process. Miscue analysis, the psycholinguistic analysis of unexpected responses in a reader's oral text, provides a verbal dimension of data for reading research. Similarly, eye-movement recording, which shows precisely where in a text a reader looks, provides a visual dimension of data. When these two research approaches are combined, both verbal and visual data are analyzed, resulting in a powerful, multi-dimensional view of the reading process. This dissertation focuses on adult readers' eye movements made during the production of miscues and other oral reading phenomena. Patterns of eye movements relative to substitutions, omissions, insertions, partials, and repetitions are described, analyzed, and compared. Results of the analysis are discussed in terms of whether current causal explanations of miscues are augmented or refuted. Original conceptions about the reading process formed as a result of this research are developed and placed in existing theoretical frameworks. Major findings include that the eye movements relative to different types of miscues and other oral reading phenomena exhibit different patterns, and both eye movements and miscues, and the relationship between them, are functions of comprehension. Also, contrary to conventional wisdom, most miscued words are examined, and examined thoroughly, before the miscue is produced; miscues are not caused by careless or reckless reading, or visually skipping words. Implications for theories and models of the reading process are discussed, and areas of needed research are described.

It sometimes happens that we finish reading a passage of text just to realize that we have no idea what we just read. During these episodes of mindless reading our mind is elsewhere yet the eyes still move across the text. The phenomenon of mindless reading is common and seems to be widely recognized in lay psychology. However, the scientific investigation of mindless reading has long been underdeveloped. Recent progress in research on mindless reading has been based on self-report measures and on treating it as an all-or-none phenomenon (dichotomy-hypothesis). Here, we introduce the levels-of-inattention hypothesis proposing that mindless reading is graded and occurs at different levels of cognitive processing. Moreover, we introduce two new behavioral paradigms to study mindless reading at different levels in the eye-tracking laboratory. First (Chapter 2), we introduce shuffled text reading as a paradigm to approximate states of weak mindless reading experimentally and compare it to reading of normal text. Results from statistical analyses of eye movements that subjects perform in this task qualitatively support the ‘mindless’ hypothesis that cognitive influences on eye movements are reduced and the ‘foveal load’ hypothesis that the response of the zoom lens of attention to local text difficulty is enhanced when reading shuffled text. We introduce and validate an advanced version of the SWIFT model (SWIFT 3) incorporating the zoom lens of attention (Chapter 3) and use it to explain eye movements during shuffled text reading. Simulations of the SWIFT 3 model provide fully quantitative support for the ‘mindless’ and the ‘foveal load’ hypothesis. They moreover demonstrate that the zoom lens is an important concept to explain eye movements across reading and mindless reading tasks. Second (Chapter 4), we introduce the sustained attention to stimulus task (SAST) to catch episodes when external attention spontaneously lapses (i.e., attentional decoupling or mind wandering) via the overlooking of errors in the text and via signal detection analyses of error detection. Analyses of eye movements in the SAST revealed reduced influences from cognitive text processing during mindless reading. Based on these findings, we demonstrate that it is possible to predict states of mindless reading from eye movement recordings online. That cognition is not always needed to move the eyes supports autonomous mechanisms for saccade initiation. Results from analyses of error detection and eye movements provide support to our levels-of-inattention hypothesis that errors at different levels of the text assess different levels of decoupling. Analyses of pupil size in the SAST (Chapter 5) provide further support to the levels of inattention hypothesis and to the decoupling hypothesis that off-line thought is a distinct mode of cognitive functioning that demands cognitive resources and is associated with deep levels of decoupling. The present work demonstrates that the elusive phenomenon of mindless reading can be vigorously investigated in the cognitive laboratory and further incorporated in the theoretical framework of cognitive science.<br>Beim Lesen passiert es manchmal dass wir zum Ende einer Textpassage gelangen und dabei plötzlich bemerken dass wir keinerlei Erinnerung daran haben was wir soeben gelesen haben. In solchen Momenten von gedankenverlorenem Lesen ist unser Geist abwesend, aber die Augen bewegen sich dennoch über den Text. Das Phänomen des gedankenverlorenen Lesens ist weit verbreitet und scheint in der Laienpsychologie allgemein anerkannt zu sein. Die wissenschaftliche Untersuchung von gedankenverlorenem Lesen war jedoch lange Zeit unzureichend entwickelt. Neuerer Forschungsfortschritt basierte darauf gedankenverlorenes Lesen durch Selbstberichte zu untersuchen und als ein Phänomen zu behandeln das entweder ganz oder gar nicht auftritt (Dichotomie-Hypothese). Hier stellen wir die ‚Stufen der Unaufmerksamkeit’-Hypothese auf, dass gedankenverlorenes Lesen ein graduelles Phänomen ist, das auf verschiedenen kognitiven Verarbeitungsstufen entsteht. Wir stellen zudem zwei neue Verhaltensparadigmen vor um verschiedene Stufen von gedankenverlorenem Lesen im Augenbewegungslabor zu untersuchen. Als erstes (in Kapitel 2) stellen wir das Lesen von verwürfeltem Text vor als ein Paradigma um Zustände von schwach gedankenverlorenem Lesen experimentell anzunähern, und vergleichen es mit dem Lesen von normalem Text. Die Ergebnisse von statistischen Augenbewegungsanalysen unterstützen qualitativ die ‚Unaufmerksamkeits’-Hypothese, dass kognitive Einflüsse auf Augenbewegungen beim Lesen von verwürfeltem Text reduziert ist, und die ‚Foveale Beanspruchungs’-Hypothese, dass die Reaktion der zoom lens visueller Aufmerksamkeit auf lokale Textschwierigkeit beim Lesen von verwürfeltem Text verstärkt ist. Wir stellen eine weiterentwickelte Version des SWIFT Modells (SWIFT 3) vor, welches die zoom lens der Aufmerksamkeit implementiert, und validieren dieses Modell am Lesen von verwürfeltem und normalem Text (Kapitel 3). Simulationen des SWIFT 3 Modells unterstützen die ‚Unaufmerksamkeits’ und die ‚Foveal Beanspruchungs’-Hypothese in einem vollständig quantitativen Modell. Zudem zeigen sie, dass die zoom lens der Aufmerksamkeit ein wichtiges Konzept ist um Augenbewegungen in Aufgaben zum Lesen und gedankenverlorenen Lesen zu erklären. Als zweites (Kapitel 4) stellen wir den sustained attention to stimulus task (SAST) vor um Episoden von spontaner externer Unaufmerksamkeit (also Entkopplung der Aufmerksamkeit oder Abschweifen der Gedanken) in einem Paradigma über Verhaltensparameter wie das Übersehen von Fehlern im Text und Signal-Detektions-Analysen von Fehlerentdeckung zu messen. Augenbewegungsanalysen im SAST decken abgeschwächte Einflüsse von kognitiver Textverarbeitung während gedankenverlorenem Lesen auf. Basierend auf diesen Befunden zeigen wir, dass es möglich ist Zustände von gedankenverlorenem Lesen online, also während dem Lesen, aus Augenbewegungen vorherzusagen bzw. abzulesen. Dass höhere Kognition nicht immer notwendig ist um die Augen zu bewegen unterstützt zudem autonome Mechanismen der Sakkadeninitiierung. Ergebnisse aus Analysen von Fehlerdetektion und Augenbewegungen unterstützen unsere ‚Stufen der Unaufmerksamkeit’-Hypothese, dass Fehler auf verschiedenen Textebenen verschiedene Stufen von Entkopplung messen. Analysen der Pupillengröße im SAST (Kapitel 5) bieten weitere Unterstützung für die ‚Stufen der Unaufmerksamkeit’-Hypothese, sowie für die Entkopplungs-Hypothese, dass abschweifende Gedanken eine abgegrenzte kognitiver Funktionsweise darstellen, welche kognitive Ressourcen benötigt und mit tiefen Stufen von Unaufmerksamkeit zusammenhängt. Die aktuelle Arbeit zeigt, dass das flüchtige Phänomen des gedankenverlorenen Lesens im kognitiven Labor mit strengen Methoden untersucht und weitergehend in den theoretischen Rahmen der Kognitionswissenschaft eingefügt werden kann.

This thesis investigates new methods for understanding eye movement behavior in reading based on the use of eye tracking corpora and data-driven modeling. Eye movement behavior is characterized by two basic, generally unconscious, decisions: where and when to move the eyes. We explore the idea that empirical eye movement data carries rich information about the processes that guide these decisions. Two methods are investigated, each addressing a different aspect of eye movements in reading. The role of prediction in eye movement modeling is emphasized, and new evaluation methods for assessing the predictive accuracy of models are proposed.  The decision of where to move the eyes is approached using standard machine learning methods. The model proposed learns where to move the eyes under different conditions associated with the words being read. Applied to new text, the model moves the eyes in ways it has learnt, showing characteristics similar to human readers. Furthermore, we propose the use of entropy to measure the similarity between observed and predicted eye movement behavior on held-out data. The main contribution is a flexible model, with few fixed parameters, that can be used to investigate decisions about where the eyes move during reading.      The decision of when to move the eyes is approached using time-to-event modeling (survival analysis). The model proposed learns the timing of eye movements under different conditions associated with the words being read. Applied to new text, the model estimates the probability that a fixation survives for any given length of time. We propose an entropy-related measure to assess the probabilistic temporal predictions of the model. The main contribution is the use of Cox hazards modeling to address questions about the strength, as well as the timing, of processes that influence the decision of when to move the eyes during reading.

The primary aim of this thesis was to identify the mechanism under-pinning the word-predictability effect, while a secondary aim was to investigate whether words are processed in serial or parallel. In five experiments, adults’ eye-movements were monitored as they read sentences for comprehension on a computer screen. In Experiments 1 and 2, a critical target-word that was either of high- or low-frequency and either predictable or unpredictable was embedded in experimental sentences. The nature of the preview of the target word was manipulated such that it was either identical to the target or was misspelled (the misspelling was more severe in Experiment 2). Predictability effects were apparent in the identical preview condition in both experiments, whilst they were only apparent in the misspelled condition of Experiment 1. This outcome is compatible with early Guessing Game type models of reading which propose that readers predictions about up-coming words using contextual parafoveal information. When taken together, the results of Experiments 1 and 2 also suggested that frequency and predictability exert additive effects on fixation durations.In Experiment 3, four levels of word-predictability were employed. The function relating word-predictability and word-processing time was strictly monotonic: word-processing time decreased as predictability increased. This outcome was consistent with a word-prediction account of predictability in which there is no penalty for incorrect guessing. Experiment 3 also showed that processing time on the pre-target word increased as the predictability of the up-coming increased. This outcome replicated an effect obtained by Kliegl, Nuthmann and Engbert (2006) who claim that it arises as a result of memory retrieval processes cued by prior sentence context Experiment 4 replicated the manipulation in Experiment 3 but included additional condition in which the preview of the target word was masked while in parafoveal vision, using a pixel scrambling technique. The target-predictability effect was again a graded one, and did not depend upon the availability of initial information, providing evidence against the word-prediction theory. Additionally, there was no pre-target predictability effect in the unmasked condition. There was a pre-target effect in an direction in the masked condition, although this appeared to be a consequence of the mask. Experiment 5 replicated Experiment 4, but replaced the masked condition with a non-predictable but semantically related word, and the results showed no pre-target effects at all. It was concluded that inverted pre-predictability effects are more likely to be related to higher-level sentential processing.

When we read a text, we obtain information at different levels of representation from abstract symbols. A reader’s ultimate aim is the extraction of the meaning of the words and the text. The reserach of eye movements in reading covers a broad range of psychological systems, ranging from low-level perceptual and motor processes to high-level cognition. Reading of skilled readers proceeds highly automatic, but is a complex phenomenon of interacting subprocesses at the same time. The study of eye movements during reading offers the possibility to investigate cognition via behavioral measures during the excercise of an everyday task. The process of reading is not limited to the directly fixated (or foveal) word but also extends to surrounding (or parafoveal) words, particularly the word to the right of the gaze position. This process may be unconscious, but parafoveal information is necessary for efficient reading. There is an ongoing debate on whether processing of the upcoming word encompasses word meaning (or semantics) or only superficial features. To increase the knowledge about how the meaning of one word helps processing another word, seven experiments were conducted. In these studies, words were exachanged during reading. The degree of relatedness between the word to the right of the currently fixated one and the word subsequently fixated was experimentally manipulated. Furthermore, the time course of the parafoveal extraction of meaning was investigated with two different approaches, an experimental one and a statistical one. As a major finding, fixation times were consistently lower if a semantically related word was presented compared to the presence of an unrelated word. Introducing an experimental technique that allows controlling the duration for which words are available, the time course of processing and integrating meaning was evaluated. Results indicated both facilitation and inhibition due to relatedness between the meanings of words. In a more natural reading situation, the effectiveness of the processing of parafoveal words was sometimes time-dependent and substantially increased with shorter distances between the gaze position and the word. Findings are discussed with respect to theories of eye-movement control. In summary, the results are more compatible with models of distributed word processing. The discussions moreover extend to language differences and technical issues of reading research.<br>Wenn wir einen Text lesen, erfassen wir Informationen auf verschiedenen Repräsentationsebenen anhand abstrakter Symbole. Das oberste Ziel des Lesers ist das Erfassen der Bedeutung der Worte und des Textes. Die Erforschung der Blickbewegungen beim Lesen umfasst verschiedene Verarbeitungsebenen, die von Warhnehmung über motorische Prozesse bis hin zu Kognition auf übergeordneter Ebene reichen. Das Lesen geübter Leser verläuft zum großen Teil automatisch, ist aber gleichzeitig ein komplexes Phänomen interagierender Teilprozesse. Die Untersuchung von Blickbewegungen beim Lesen eröffnet die Möglichkeit, kognitive Prozesse bei der Ausübung einer alltäglichen Aufgabe anhand von Verhaltensmaßen zu untersuchen. Der Leseprozess ist nicht beschränkt auf das direkt fixierte (oder foveale) Wort, sondern umfasst auch umgebende (oder parafoveale) Wörter, insbesondere das Wort rechts der Blickposition. Obgleich dies nicht notwendigerweise bewusst geschieht, ist die parafoveale Information dennoch wichtig für effizientes Lesen. Es wird darüber diskutiert, ob die Verarbeitung des nächsten Wortes die Wortbedeutung (Semantik) oder nur oberflächliche Eigenschaften umfasst. Um ein besseres Verständnis zu erhalten, ob die Bedeutung eines Wortes bei der Verarbeitung eines anderen Wortes hilft, wurden sieben Experimente durchgeführt. In diesen Studien wurde ein Wort im Satz während des Lesens ausgetauscht. Der inhaltliche Zusammenhang zwischen einer parafoveal präsentierten Vorschau und dem anschließend fixierten Zielwort wurde experimentell manipuliert. Außerdem wurde der zeitliche Verlauf der Bedeutungserfassung aus parafovealen Wörtern mit zwei Ansätzen untersucht, einem experimentellen und einem statistischen. Als primärer Befund zeigte sich, dass die Fixationszeiten durchweg kürzer waren, wenn ein semantisch verwandtes Wort als Vorschau präsentiert wurde, verglichen mit einem Wort ohne Verwandtschaft. Mit der in dieser Arbeit verwendeten experimentellen Vorgehensweise konnte zudem der zeitliche Verlauf des Verarbeitens und Integrierens von Bedeutung ermittelt wurde. Dabei ergaben sich kürzere Fixationszeiten auf dem Zielwort bei ähnlichen Wortbedeutungen und längere Fixationszeiten bei unterschiedlichen Wortbedeutungen. Die Ergebnisse zeigten sowohl leichtere als auch schwerere Verarbeitung in Folge der Ähnlichkeit von Wortbedeutungen. In einer natürlicheren Lesesituation war die Wirksamkeit der Verarbeitung nachfolgender Wörter teilweise abhängig von der Dauer der Vorschau, und sie war deutlich größer bei kürzerer räumlicher Distanz zwischen der Blickposition und der Vorschau. Die Befunde werden mit Blick auf Theorien der Blickbewegunskontrolle diskutiert. Die Ergebnisse sind stärker mit Modellen verteilter Wortverarbeitung vereinbar. Die Diskussion erstreckt sich außerdem auf Sprachunterschiede und technische Aspekte der Leseforschung.