Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Reading Comprehension and Eye Movements

Abstract. Reading comprehension unfolds through coordinated visual sampling, word identification, syntactic analysis, contextual inference, discourse-memory updating, and strategic reinspection. This fundamental research monograph examines eye movements as temporally precise yet indirect behavioral evidence about those interacting processes. A critical narrative synthesis draws on gaze-contingent experiments, sentence-level eye tracking, discourse studies, computational accounts of saccadic control, developmental comparisons, individual-differences research, and combined eye-tracking–magnetoencephalography evidence. It evaluates perceptual-span and preview phenomena; lexical frequency and contextual predictability; ambiguity resolution, skipping, regressions and comprehension monitoring; the E-Z Reader and SWIFT explanatory programs; and the evidential relationship between reading fluency and understanding. Studies establish that processing demand alters fixation behavior, parafoveal information facilitates later recognition, and reader goals affect reinspection, while alternative processes often produce overlapping gaze signatures. Recent neural evidence extends tests of parafoveal semantic representation but does not make eye position an unambiguous measure of comprehension. The synthesis specifies measurement limitations, causal tests, cross-reader transfer problems, and research priorities. This is a critical narrative review of previously published work; it reports no original experiment, systematic-review protocol, or meta-analysis.

Keywords: reading comprehension; eye movements in reading; eye tracking; fixation duration; saccades; regressions; parafoveal preview; perceptual span; word skipping; syntactic ambiguity; lexical frequency; contextual predictability; discourse integration; reading strategies; E-Z Reader; SWIFT; gaze-contingent display; reading skill; magnetoencephalography

Introduction: The Reading-Comprehension Problem

Reading comprehension is the construction and updating of a coherent interpretation of written language. The reader must identify visible letter strings, retrieve linguistic information, determine grammatical relationships, integrate propositions, resolve references, infer connections, and preserve relevant information across sentences. Eye movements make this sequence experimentally accessible because the eyes select locations in text while cognitive activity changes the timing and destinations of those selections. The central research problem is therefore to explain how visible sequences of fixations, saccades, skips, and returns relate to the processes that yield a meaningful discourse representation. The same behavior can arise from different cognitive causes, so a theory must specify the mechanisms linking gaze and comprehension rather than equate the two Just & Carpenter (1980) Rayner (1998) Clifton et al. (2007).

This monograph examines reading comprehension and eye movements as one coupled research object: the online construction of meaning under the sensory, attentional, linguistic, and memory constraints of ordinary print reading. Its main questions concern how information becomes available before and during fixation, how word and sentence difficulty influence gaze, how readers recover from misinterpretation, how goals affect rereading, and what can be concluded about comprehension from the resulting records. Eye movements during literary, instructional, and expository reading fall within this subject when they reveal online processing; the acquisition of literacy, the independent theory of eye-movement motor control, and general visual crowding retain their own adjoining scientific domains Hyönä et al. (2002) Kaakinen et al. (2003) Rayner (2009).

The research method is a critical narrative integration of original experimental studies, influential formal models, methodological analyses, and selected newer multimodal findings. The studies are organized by their inferential contribution, from controlled display changes that establish access to visual information, through linguistic contrasts that localize processing demands, to comprehension tasks that evaluate text-level outcomes. Evidence from simulations is evaluated as model fit, from natural reading as behavior under coupled constraints, and from neural recordings as evidence for representational timing. This organization retains the distinction between observed fixations, estimated internal processing, and demonstrated understanding of the text Reichle et al. (1998) Engbert et al. (2005) Wang et al. (2025).

Operational Definitions and Levels of Explanation

A fixation is an interval during which gaze is relatively stable on a text location, while a saccade is a rapid shift to another location. The experimentally defined duration and position of a fixation depend on the eye tracker, sampling, event-detection algorithm, and treatment of blinks and small corrective movements. A first-pass measure includes initial encounter with a region before gaze exits it; gaze duration sums initial fixations on a word; total reading time includes later revisits. A regression is a movement toward earlier text in the writing direction, though its linguistic interpretation depends on its origin and destination. These are measurement categories: their relation to comprehension must be established through experimental contrasts Rayner (1998) Clifton et al. (2007).

Reading comprehension itself admits multiple outcomes. Local lexical identification concerns recognition and interpretation of a particular word; sentence comprehension concerns grammatical and propositional relationships; discourse comprehension includes integration across sentences, causal continuity, referential coherence, situation models, and later recall or inference. A reader might spend little time on a sentence because it is easy, because it is familiar, because it was skimmed, or because it was incompletely processed. A comprehension question, delayed recall, a paraphrase, and a task-relevant judgment therefore measure overlapping but different aspects of constructed meaning. Interpretations of eye traces become stronger when these outcomes are gathered alongside online gaze Hyönä et al. (2002) Kaakinen & Hyönä (2010) Just & Carpenter (1980).

Three explanatory levels must remain explicit. At the visual-oculomotor level, text geometry, acuity, crowding, and landing-position error influence where the eye can move. At the lexical-syntactic level, word frequency, predictability, morphological and grammatical information influence how long language processing takes. At the discourse-strategic level, prior knowledge, reader purpose, and monitoring goals influence integration and revisits. These levels interact: a prolonged fixation can reflect lexical difficulty, syntactic integration, reduced preview, or a strategically imposed pause. A scientifically adequate account jointly models these sources and evaluates them against independent comprehension evidence McConkie & Rayner (1975) Henderson & Ferreira (1990) Kaakinen et al. (2003).

Historical Foundations: From Gaze Records to Online Cognition

The modern experimental history of this field rests on the transition from descriptive eye-movement records to manipulations that intervene during ongoing reading. In the 1970s computer-controlled displays made it possible to change printed material contingent on fixation position or while the eyes moved. McConkie and Rayner established a method for testing the span of useful visual information rather than inferring it from natural fixation behavior alone. Rayner's related early investigations examined peripheral cues and parafoveal identification. This transformed eye-movement research into a program for experimentally determining which visual information guides reading before a word is directly fixated McConkie & Rayner (1975) Rayner (1975a) Rayner (1975b).

Just and Carpenter's 1980 model moved the discussion from visual uptake to linguistic comprehension, interpreting eye-fixation patterns in relation to word recognition, clause processing, and inferential work in passages. A major attraction was that reading provides a continuous behavioral stream aligned with meaningful language. Frazier and Rayner then used temporarily ambiguous sentences to examine initial parsing decisions and revision; Duffy, Morris, and Rayner studied how lexical ambiguity, meaning dominance, and contextual placement changed fixation patterns. These experiments showed why semantic and syntactic manipulations must accompany timing records before gaze can support a psychologically specific explanation Just & Carpenter (1980) Frazier & Rayner (1982) Duffy et al. (1988).

Later decades connected these questions to formal models, large natural-reading datasets, individual differences, and new neural measurements. The E-Z Reader program related lexical access to the decision to launch and execute saccades; SWIFT explored spatially distributed lexical processing and autonomous saccade timing. Attention shifted toward distinguishing model predictions, interpreting gaze data across populations, and resolving the contribution of upcoming text before fixation. Contemporary multimodal research can detect neural representation during the same reading intervals captured by the eye tracker, but historical evidence still demands careful separation of experimental observation and modeling assumptions Reichle et al. (1998) Engbert et al. (2005) Kuperman & Van Dyke (2011) Wang et al. (2025).

Eye Tracking as an Experimental Measurement System

Eye trackers estimate gaze by using recorded eye features, optics, and calibration. Reading studies additionally require a spatial mapping between sample coordinates and text characters, consistent line boundaries, and algorithms that classify raw samples into fixation and saccade events. The temporal precision of the recording system should not be mistaken for equivalent precision in the inferred mental event. Fixation onset does not imply that attention or lexical processing began at that instant; the parafoveal representation of a word may already be developing. Similarly, a fixation can continue after a lexical decision because a saccade needs time to be prepared and completed Reichle et al. (1998) Schotter et al. (2012) Rayner (2009).

The principal analytic choices include defining regions of interest, handling skipped words, identifying first-pass entry and exit, defining regression paths, and deciding whether very short fixations are merged with adjacent events. Word-based outcomes are often conditioned on the word having been fixated, which selects particular items and reader states. An analysis of gaze duration on fixated words may therefore omit the very cases in which predictability or preview caused a skip. Comparisons can be distorted when long words receive more fixations or when a difficult preceding word changes initial landing position. Measures should be selected for an explicit causal or descriptive question Rayner et al. (2011) Kliegl et al. (2006) Clifton et al. (2007).

Comprehension experiments should control typography, font size, line width, visual angle, reading direction, lexical properties, sentence position, and task demands. Balanced materials and within-reader manipulations help isolate a linguistic feature; independent items and readers require appropriate inferential treatment. A robust analysis reports exclusions, blinks, calibration accuracy, trials with undetected changes, and uncertainty around estimated effects. Statistical significance by itself does not show that longer gaze caused better comprehension. The purpose of the eye-movement record is to constrain models of processing in a specified task, with reading outcome measures supplying complementary validity evidence Slattery et al. (2011) Kaakinen & Hyönä (2010) Kuperman & Van Dyke (2013).

Causal Paradigms: Moving Windows, Boundaries, and Reading Tasks

The moving-window paradigm selectively preserves text around the reader's current fixation and masks or distorts text elsewhere. Systematic changes in window extent reveal how far useful information contributes to fluent reading under a given writing system and display condition. The boundary paradigm instead places an invisible trigger before a target location: an initially displayed preview is replaced with the actual target during the saccade crossing that location. Differences in subsequent target fixation durations can then be attributed to an experimentally controlled preview manipulation, subject to checks on change timing, display artifacts, and awareness. These techniques established much of the field's causal leverage McConkie & Rayner (1975) Rayner (1975a) Angele et al. (2008).

Natural sentence reading adds contextual manipulation. Researchers vary lexical frequency, length, ambiguity, predictability, or syntactic structure while recording eye movements across complete sentences. The continuous context preserves comprehension demands, yet changing one property often changes others: a likely completion may also be shorter or more familiar, and an ambiguous phrase may induce particular prosodic or discourse expectations. Experimental design can cross these properties, use matched alternatives, and model item variation. Eye movements become an informative dependent variable because the local timing of gaze relates to the position of experimental contrasts Duffy et al. (1988) Frazier & Rayner (1982) Staub (2011).

Expository-text and task-manipulation paradigms test a different scope of explanation. Asking a reader to summarize, evaluate material for a particular purpose, proofread, or answer questions changes what counts as relevant and how text is revisited. In these studies the same physical text may generate distinct eye trajectories. Such manipulations directly challenge the idea of a context-free reading-speed metric and allow investigators to relate goals to encoding and later remembering. They also show why findings from short, single-line experimental sentences cannot be generalized automatically to the comprehension of extended educational texts Hyönä et al. (2002) Kaakinen et al. (2003) Kaakinen & Hyönä (2010).

The Geometry of Visual Uptake and Perceptual Span

The written page is a spatially extended signal, but its letters are sampled with highly nonuniform acuity. At each fixation, the foveal region provides the strongest visual detail; eccentric locations contribute progressively coarser information, and crowding and letter position complicate identification. The perceptual span denotes the region from which readers extract information useful for the current reading task, as established by gaze-contingent display manipulations. It is therefore a functional measure under a specified condition rather than a fixed anatomical aperture. The moving-window experiments of McConkie and Rayner demonstrated that masking outside a limited region can impair reading, with the magnitude depending on the available field and its relation to current fixation McConkie & Rayner (1975) Rayner (1975a).

English reading normally involves an asymmetric distribution of useful information extending farther in the direction of progression than backward, although its exact size depends on materials, participant group, display, and the type of information being measured. Rayner's comparison of beginning and skilled readers found a smaller functional span among less experienced readers, linking visual sampling with processing efficiency. This finding does not show that comprehension is determined by peripheral acuity alone. Skilled readers may use information extracted outside the fovea more efficiently, and different language systems and spatial layouts change the balance of character identification and directional attention Rayner (1986) Rayner (1998).

Perceptual span and attention must also be distinguished. The ability to identify a letter at a given eccentricity does not by itself establish when lexical resources are allocated to that letter. Processing difficulty on the currently fixated word can restrict the useful information gathered from upcoming words, supporting a dynamic relation between foveal load and parafoveal uptake. The broader implication is that visual extraction is conditional on ongoing language processing. A useful model must describe how attentional allocation, word difficulty, and spatial sampling jointly influence the next fixation without treating a fixed span as a direct measure of comprehension Henderson & Ferreira (1990) Schotter et al. (2012).

Foveal Difficulty and Parafoveal Preview

Parafoveal preview refers to information collected about a word before the eyes land on it. In the boundary-change paradigm, the region initially contains a controlled preview which becomes the correct target during a saccade. If subsequent fixations are shorter after an informative preview than after an uninformative one, the experiment provides evidence that processing began before direct fixation. Such benefit can arise from partial orthographic identification, phonological information, and, under some conditions, semantic properties. The type of preview and the target language must be specified because a preview benefit is not a single uniform phenomenon Rayner (1975b) Schotter et al. (2012) Schotter et al. (2015).

Henderson and Ferreira's foveal-load experiments made a crucial causal point: hard processing at the current word constrains uptake from the subsequent location. It follows that an upcoming word cannot be studied independently of the information demand at the presently fixated word. Angele and colleagues further manipulated previews at successive upcoming positions to examine how information from n + 1 and n + 2 can influence behavior. Such designs place restrictions on models that assign attention rigidly to only one word, yet detailed interpretation depends on the manipulation, display visibility, and event timing Henderson & Ferreira (1990) Angele et al. (2008).

Display-change awareness is a methodological hazard because a sudden alteration in the visible text may interfere with reading independently of its intended linguistic content. Changes that occur during saccades are often difficult to detect, but failures of timing, unusual previews, and repeated trials can still create artifacts. Slattery and colleagues studied eye movements and change detection, highlighting the importance of recording whether participants noticed the manipulation. A rigorous preview experiment therefore checks actual eye positions, boundary crossings, display latency, and the extent to which an observed advantage can be attributed to pretarget representation rather than disruption from the experimental apparatus Slattery et al. (2011) Schotter et al. (2012).

Orthographic Coding and Letter Position

Printed words pose an identification problem because the visual system must recognize an ordered letter sequence while accommodating variability in individual letters, eye position, and neighboring symbols. Orthographic processing refers to the representation of these visible letter patterns and their relation to stored word forms. Experiments involving transposed letters show that letter identity and letter position contribute differently to word recognition. White and colleagues examined eye movements during reading of transposed text, with particular attention to initial letters. The findings reinforce that readers exploit constraints beyond independent recognition of isolated letters, while the magnitude of disruption depends on the letters manipulated and the surrounding sentence White et al. (2008).

The interpretation of letter-position effects requires separating orthographic coding from lexical access and contextual compensation. A transposed string might be more recognizable if it resembles a familiar word, if sentence context predicts the intended item, or if its initial letters provide powerful candidate restrictions. A change in fixation time alone cannot distinguish these possibilities. Consequently, contrasts between internal and external transpositions, matched frequencies, and controlled contexts are needed to determine which representation best explains behavior. Eye movements provide continuous timing, but explanatory precision comes from the combination of linguistic manipulation and comparison conditions White et al. (2008) Duffy et al. (1988).

Orthographic effects extend beyond foveally seen target words. A parafoveal preview that preserves useful letter information may facilitate later recognition even if it differs from the final target. Conversely, preview substitutions may introduce orthographic neighbors that compete with the target or induce visual change costs. New combined eye-tracking and magnetoencephalography work has begun to investigate neural responses to parafoveal orthographic and semantic signals during natural reading. Such research can constrain the time course of representation, although it still requires careful source localization, eye-movement artifact control, and independent tests of the proposed functional role of the activity Schotter et al. (2012) Wang et al. (2025).

Lexical Frequency, Word Length, and Landing Sites

Word length and lexical frequency are robust determinants of reading behavior under many task conditions. Longer words differ in visual extent, number of potential fixation locations, and often the probability of receiving more than one initial fixation. More frequent words typically receive shorter processing times on average, although the size and source of that advantage depend on the reader and linguistic context. A central issue is that lexical frequency is correlated with familiarity, contextual distributions, and reading experience. Models that use frequency as a lexical-processing parameter should therefore be evaluated against item-level controls and individual variation rather than only population averages Rayner (1998) Kliegl et al. (2006) Kuperman & Van Dyke (2013).

Landing position determines the visual information available after a saccade. The preferred viewing location for a word is a useful empirical tendency, yet actual landing sites vary with distance, length, launch position, text geometry, and motor error. A fixation at a word edge may lead to a corrective refixation even when lexical difficulty is unchanged. Joseph and colleagues compared length and landing-position effects in children and adults, illustrating how developmental differences in oculomotor targeting can complicate simple interpretations of reading speed. Data on processing and targeting become more informative when analyzed jointly rather than as independent aggregates Joseph et al. (2009) Rayner (1986).

Individual differences further qualify frequency effects. Kuperman and Van Dyke reassessed frequency as a determinant of recognition for readers with different levels of skill, emphasizing that language experience and reading ability modify the role of item properties. Earlier data from their individual-differences program similarly suggested that word-level gaze measures reflect combinations of lexical competence and cognitive capacity. Frequency effects should therefore be reported with information about participants, task, lexical norms, and the distribution of measurements. A single average fixation duration cannot substitute for a theory of the interactions among word properties, perceptual access, and reader-specific linguistic knowledge Kuperman & Van Dyke (2011) Kuperman & Van Dyke (2013).

Word Skipping and Forward Saccade Selection

A word is skipped when an initial forward saccade passes over it without a recorded fixation. Skipping is systematically influenced by length, contextual predictability, and information available before the eyes arrive; very short or expected words are often skipped more readily than visually long or unexpected ones. Yet skipping is not proof that a word was understood, because the motor program may have been initiated before lexical processing was complete. Skipped material can contribute to comprehension through parafoveal processing, linguistic expectation, or later integration, and the consequences depend on whether the reader subsequently returns Rayner et al. (2011) Reichle et al. (1998).

Rayner and colleagues tested length and predictability as contributors to skipping, making explicit that forward movement decisions depend on both text geometry and contextual information. Models differ about whether an early lexical-processing milestone triggers a saccade program, whether attention is distributed across words, and how much of a decision can be revised during preparation. Empirical contrasts should examine skipping probability alongside landing sites, first-pass durations of adjacent words, and comprehension outcomes. Otherwise an apparent reduction in target fixation time may partly reflect the removal of particular trials from the fixated-word sample Rayner et al. (2011) Engbert et al. (2005) Staub (2011).

The general principle is that saccadic choice and comprehension are related through intermediate mechanisms. Eye movement to the next word supplies useful visual material, but the choice of destination is constrained by already accumulated preview, motor limits, and uncertainty. An unfixated word may be partially recognized, inferred, later recovered, or missed; these possibilities call for controlled tests rather than confident classification from the trace alone. A strong reading theory predicts the joint distribution of skips, fixations, regressions, and successful meaning construction over continuous text Kliegl et al. (2006) Schotter et al. (2012) Rayner (2009).

Lexical Ambiguity and Contextual Meaning Selection

An ambiguous written word may be compatible with several lexical meanings even though its physical spelling remains unchanged. Research on lexical ambiguity asks how readers activate, select, and integrate candidate interpretations as a sentence unfolds. Duffy, Morris, and Rayner tested how the relative frequency of meanings and contextual disambiguation affect fixation durations. Their work made lexical ambiguity an especially useful case because the physical word can be held constant while context and meaning distribution vary. Delayed eye movements and later regressions reveal processing consequences of these contrasts, although a prolonged fixation does not directly reveal which candidate meaning was activated at a specific moment Duffy et al. (1988) Clifton et al. (2007).

Theoretical explanations differ in how strongly initial access is constrained by context. A context-sensitive account emphasizes the use of preceding semantic and syntactic information in activating a likely interpretation; a multiple-access account permits several meanings to be activated before selection. The research problem requires separating activation from conscious report and selection from final integration. Frequency dominance, contextual strength, subsequent disambiguating words, and reader familiarity can be manipulated to create different predictions. Eye-movement measures become especially informative when an early effect on the ambiguous word is distinguished from a delayed effect on material that makes one interpretation untenable Duffy et al. (1988) Just & Carpenter (1980).

The broader consequence concerns comprehension as a sequence of commitments. A reader can integrate a provisional meaning fluently and then discover, from later text, that revision is required. Some revisions cause an immediate slowdown; others may be resolved with little overt return movement. Absence of a regression therefore provides limited evidence about whether a mistaken interpretation was corrected. The strength of an ambiguity study comes from converging gaze measures, controlled contextual manipulations, and independent tests of the interpretation ultimately constructed Frazier & Rayner (1982) Duffy et al. (1988) Rayner (1998).

Structural Ambiguity and Garden-Path Recovery

Syntactic comprehension entails identifying grammatical relations as words arrive and assigning them to a growing sentence representation. Temporarily ambiguous constructions create a particularly diagnostic problem: early material supports a plausible structural analysis that later words contradict. Frazier and Rayner tracked eye movements as readers encountered and repaired such interpretations, demonstrating that the cost of reanalysis can be temporally and spatially associated with the disambiguating material. Reading-time increases and backward movements supply evidence for processing difficulty, but the precise account of initial preference depends on the grammatical alternatives, the language, and the availability of contextual cues Frazier & Rayner (1982) Clifton et al. (2007).

An initial parse may be influenced by structural economy, lexical category probabilities, referential context, prior discourse, and the range of alternatives licensed by the text. The garden-path concept identifies a trajectory in which an initially workable analysis leads to later revision. It does not specify one universal parsing algorithm. A sentence can also be difficult because of memory interference, rare constructions, or semantically implausible dependencies, even when no discrete initial misparse occurred. Competing explanations should therefore be tested with matched materials in which grammatical structure, lexical familiarity, and interpretive plausibility are varied independently Frazier & Rayner (1982) Just & Carpenter (1980).

Eye movements are often read as evidence for a sequence of syntactic operations, but an inferential bridge is needed. The first fixation on a disambiguating word may reflect detection of inconsistency, while go-past time can accumulate processing on intervening material and regressions. Longer total time may reflect a second reading of material previously recognized. Studies that include final interpretation questions can separate successful reanalysis from persistent misunderstanding, although comprehension questions themselves change reading goals. These considerations motivate models of comprehension that represent initial analysis, error detection, repair, and outcome as distinguishable processes Clifton et al. (2007) Kaakinen & Hyönä (2010).

Predictability, Cloze Probability, and Surprisal

Predictability refers to how expected a word is in its linguistic context. It is commonly estimated by asking respondents to complete a sentence fragment and calculating cloze probability, or by using probabilistic language models to estimate word likelihood. Surprisal is a transformation of probability that represents unexpectedness on a logarithmic scale. These constructs are related but operationally distinct: a cloze sample captures a particular elicitation distribution, while a computational probability depends on the model's training and conditioning context. Both can be correlated with lexical frequency and semantic association. Reading studies must establish which operationalization corresponds to the hypothesized cognitive process Kliegl et al. (2006) Staub (2015).

Contextually expected words often receive shorter fixations and are more likely to be skipped. Yet an expectancy effect does not by itself prove that readers generate one precise lexical candidate in advance. Context may facilitate semantic integration after recognition, narrow the set of candidates, or speed visual identification through partial preactivation. Luke and Christianson examined limits on lexical prediction, adding important evidence to debates about how precisely upcoming words are represented during ordinary reading. Models should specify the conditions under which prediction is sufficiently specific to distinguish alternatives, and the circumstances in which a probabilistic advantage arises without a unique preactivated word Luke & Christianson (2016) Staub (2015).

Experimental prediction research is strongest when contextual support, word frequency, length, and semantic plausibility are carefully disentangled. A highly predictable word may also be a frequent word; a surprising word may be syntactically valid but conceptually inconsistent. Independent norming, within-item manipulations, and distribution-sensitive outcome measures help test competing explanations. Predictability effects also interact with the preceding word and parafoveal access, making a strictly local single-word interpretation incomplete. The processing consequences of prediction should be assessed across targets, spillover regions, skips, and the reader's eventual understanding Kliegl et al. (2006) Staub (2011) Schotter et al. (2015).

Distributional Analysis of Fixation Times

The arithmetic mean of fixation durations compresses a distribution whose shape may change across experimental conditions. Two manipulations can produce the same mean difference while acting on different parts of the distribution, suggesting different processing and decision mechanisms. Staub investigated lexical-predictability effects through distributional analyses rather than relying only on averages. Distributional work can ask whether a contextual advantage shifts most fixations or selectively increases the proportion of especially long durations. Neither pattern uniquely identifies a cognitive mechanism, but each can help distinguish model predictions that are otherwise equivalent at the level of mean reading times Staub (2011) Staub (2015).

A statistical analysis of gaze data must also consider the multiple dependent measures extracted from the same sequence. First-fixation duration, gaze duration, go-past time, regression probability, and total time are correlated and often condition on different trial subsets. Treating every measure as an independent confirmatory test inflates the chance of an apparently persuasive pattern. Models can instead prespecify a primary measure for the hypothesized stage, present complementary measures transparently, and account for variation across readers and lexical items. Distributional comparisons are particularly valuable when alternative theories differ about whether a manipulation changes early processing or later recovery Clifton et al. (2007) Kliegl et al. (2006).

Observed fixation duration also contains motor and visual timing components. A lexical event may finish while the gaze remains stationary because the next movement has not yet been prepared or cannot immediately be executed. Conversely, a movement may begin before all semantic integration is complete. Statistical models are most informative when these temporal constraints are built into psychological predictions rather than treated as measurement noise. Formal comparison of predicted fixation distributions, saccade distances, and regressions with observed records offers a stronger test than fitting one aggregate reading-time outcome Reichle et al. (1998) Engbert et al. (2005) Staub (2011).

Parafoveal Semantics and New Neural Evidence

The scope of meaningful information extracted before fixation remains an active issue. Orthographic preview benefits are well established, but semantic preview effects have varied across experimental configurations, languages, and analytic choices. It is useful to distinguish the proposition that upcoming words can influence processing from the stronger proposition that a complete semantic representation is available before the target is directly seen. Schotter and colleagues studied how contextual constraint changes parafoveal processing, providing a way to test the relation between expectation and pretarget information. Their framework reinforces the importance of carefully selected preview conditions and interpretation of later target fixations Schotter et al. (2015) Schotter et al. (2012).

Wang and colleagues reported in 2025 that combined eye tracking and magnetoencephalography detected temporally ordered orthographic and semantic signals associated with parafoveal information during natural reading. The research is important because it relates a word's spatial status to neural representation without requiring the observer to report a separate preview decision on each trial. Nonetheless, a neural classifier's ability to distinguish information conditions is not identical to proof that a reader consciously identified the upcoming word or used its meaning to make a particular saccade. Representational decoding and comprehension remain separate explanatory questions Wang et al. (2025).

Integrating behavioral preview studies with neural timing can create stronger converging tests. Researchers can ask whether an informative preview alters a specific neural response, whether that response predicts later fixation behavior, and whether a controlled intervention changes comprehension. These designs must handle the special artifacts produced by moving eyes in electrophysiological recording and the dependence between neighboring words in natural language. New neural results should therefore sharpen the hypotheses tested by gaze-contingent manipulation rather than replace it. The aim is a multilevel account of what is processed, when, where, and with what consequences for meaning Wang et al. (2025) Angele et al. (2008) Rayner (2009).

Discourse Cohesion and Integration Across Sentences

Comprehending a passage involves building relations among propositions that are distributed across sentences. A reader identifies referents, interprets temporal and causal connections, and updates a representation of the topic as new information arrives. A local word can be identified correctly while the meaning of the passage remains fragmented. This distinction explains why eye tracking at the word level should be combined with measures of inference, summary quality, and memory for relations. In expository reading, readers must also choose how much processing to allocate to explanatory details, topic transitions, and statements that serve as integration points. Eye movements offer a temporally ordered record of attention to those textual locations Just & Carpenter (1980) Hyönä et al. (2002).

Discourse cohesion includes explicit linguistic links, such as repeated entities and connectives, as well as relationships that the reader must infer. Integration becomes challenging when a pronoun has several plausible antecedents, when a causal step is implicit, or when a later statement revises an earlier interpretation. These operations may produce longer fixations or return movements, but their absence in gaze records does not establish that no integration occurred. A reader can infer a connection from memory without directing the eyes backward. Researchers should therefore measure understanding of the relational content in addition to recording rereading behavior Kaakinen et al. (2003) Clifton et al. (2007).

One methodological consequence is that meaning construction is distributed across text regions and time. An effect observed on a sentence may originate in information introduced much earlier; a late return may signal a newly recognized inconsistency rather than poor encoding at the original location. Time-locked analysis can track where integration-related processing is expressed, yet causal inference requires manipulating the availability, relevance, or coherence of antecedent information. Extended passages and naturalistic reading allow these mechanisms to be investigated, provided that stimulus differences and reader goals are controlled Hyönä et al. (2002) Kaakinen & Hyönä (2010).

Reading Goals, Perspective, and Strategic Allocation

Reading behavior changes when the purpose of reading changes. A reader studying for a summary may allocate additional attention to topic sentences and integrative claims, whereas a reader locating a specific fact may skim surrounding text and slow sharply at potentially relevant regions. Hyönä, Lorch, and Kaakinen studied individual differences in reading expository text for summarization, linking eye fixation patterns to strategies for selecting and organizing information. These findings locate gaze control within an active task system: eye movements reflect the pursuit of a goal through a text as well as the demands imposed by its words Hyönä et al. (2002).

Kaakinen, Hyönä, and Keenan examined how prior knowledge, working-memory capacity, and the relevance of information affected fixation patterns in expository reading. Task relevance can alter attention allocation even with the same reader and physical material, and background knowledge can change what is difficult or informative. These interactions complicate comparisons between participants who differ in educational experience or domain expertise. A long fixation may indicate difficulty for a novice but deliberate evaluation for a knowledgeable reader. Analysis should therefore include experimentally specified task purposes and an assessment of whether readers achieved those purposes Kaakinen et al. (2003).

Task effects also limit the assumption that a universal fixation threshold can classify comprehension. Kaakinen and Hyönä showed that reading instructions and goals influence eye movements, emphasizing that the observed trace depends on context. A useful theory distinguishes default word identification during continuous reading from strategic control when the reader must verify evidence, locate inconsistencies, or prepare a report. This distinction has educational implications because reading for recognition, reading for explanation, and reading for critical evaluation place different demands on the same visual and linguistic systems Kaakinen & Hyönä (2010) Hyönä et al. (2002).

Return Movements, Reinspection, and Comprehension Monitoring

Backward-directed saccades and later revisits provide evidence that readers sometimes reallocate visual attention to previously encountered material. They can occur after syntactic ambiguity, semantic inconsistency, memory uncertainty, or a shift in the task. Their interpretation depends on what region is revisited, where the return originates, and what the reader subsequently does. A regression to a disambiguating phrase may be consistent with syntactic repair; a return to an earlier name may support referential resolution; reinspection of a heading can help organize an expository summary. A regression count alone collapses these distinct explanatory possibilities Frazier & Rayner (1982) Hyönä et al. (2002).

Comprehension monitoring is the process of evaluating whether the evolving interpretation is sufficiently coherent for the reader's purpose. Some errors become visible when later text conflicts with an initial understanding; others persist unnoticed. Return movements can mark successful detection of a problem, incomplete processing, or deliberate checking, and they are not uniformly signs of low comprehension. Conversely, fluent forward reading can coexist with a confidently maintained mistake. Strong designs pair gaze measures with interpretation questions, repair decisions, and the specific location at which an inconsistency becomes identifiable Duffy et al. (1988) Frazier & Rayner (1982).

Analysis of go-past time and regression path can capture processing costs incurred before the eyes move beyond a region. These measures are useful for ambiguous sentences but include time spent on earlier areas, so they are not localized clocks for one syntactic process. Total reading time combines initial encounter and subsequent visits, blending extraction with strategic reinspection. When models predict late repair, researchers can examine not only whether regressions occur but whether their destinations contain useful evidence for correcting the interpretation. This permits a more functional assessment of rereading behavior Clifton et al. (2007) Just & Carpenter (1980).

Reading Skill, Vocabulary, and Individual Differences

Readers vary in vocabulary, word-recognition efficiency, language experience, attention control, and ability to maintain information across a passage. These differences affect the same eye-movement measures often interpreted as properties of text. Kuperman and Van Dyke investigated how verbal skills relate to fixation patterns, showing the importance of reader-level variables for explaining sentence reading. A participant's response to a rare word may depend on individual familiarity, and an effect attributed to predictability may depend on comprehension skill. Theories built on narrow samples should therefore be tested against independent readers with diverse levels of literacy and language exposure Kuperman & Van Dyke (2011) Kuperman & Van Dyke (2013).

Reading skill also changes the relation between visual sampling and understanding. A skilled reader may identify words with fewer or shorter fixations while still allocating considerable time to reasoning about difficult concepts. A less experienced reader may devote more resources to local decoding, leaving fewer resources for discourse integration. However, the apparent relationship between speed and comprehension is conditional on the task and text; rapid scanning for one target can be efficient while slow rereading of a complex argument can be adaptive. Eye-tracking research must distinguish efficiency of visual word recognition from quality of the resulting interpretation Rayner (1986) Hyönä et al. (2002).

Individual-differences studies are particularly vulnerable to confounding between reader characteristics and task behavior. Vocabulary, reading experience, educational background, motivation, and familiarity with a genre may be correlated. Cross-sectional associations cannot independently establish that improving one of these variables will change eye movements or comprehension. Strong evidence comes from multiple measurement instruments, sufficiently diverse samples, explicit models of correlated skills, and experimental manipulation where feasible. The aim is to determine which observed gaze differences are explanatory, which are consequences of a shared cause, and which depend on the demands of the text Kuperman & Van Dyke (2011) Kaakinen et al. (2003).

Development, Educational Reading, and Generalizability

Beginning readers often exhibit different patterns of fixations, refixations, and forward movements from experienced adult readers. Rayner's work on the perceptual span in beginning and skilled readers established an important link between growing reading proficiency and useful information extraction. Later child–adult comparisons studied how much visual information is retained during fixations and how landing locations and word length affect behavior. Blythe and colleagues tested visual information capture in children and adults, while Joseph and colleagues examined length and landing position. These paradigms help locate developmental differences in the coordination of visual attention, targeting, and linguistic processing Rayner (1986) Blythe et al. (2009) Joseph et al. (2009).

Developmental findings must be interpreted against changing text difficulty. The same word can be familiar to an adult and novel to a child; age differences in word frequency effects may therefore express unequal experience with the items. Children and adults also differ in reading goals, tolerance of complex sentence structures, and strategies for checking comprehension. Educational experiments are strongest when text level is calibrated, language knowledge is measured, and comprehension outcomes accompany eye tracking. The methodological goal is to identify the interaction between reader development and task requirements without treating adult eye-movement patterns as the universal definition of successful reading Joseph et al. (2009) Kuperman & Van Dyke (2013).

Generalization also depends on writing system, language, screen layout, accessibility conditions, and genre. The direction and character spacing of a script influence oculomotor planning, while morphology and word segmentation influence lexical processing. An English alphabetic-reading result is informative within its experimental conditions and provides hypotheses for broader comparison; it does not automatically establish equivalent visual spans or skipping patterns in every script. Future research should integrate cross-language measures with the same levels of comprehension, explicit reporting of reading experience, and study designs that test whether a mechanism transfers across materials rather than assuming it does Rayner (2009) Schotter et al. (2012).

E-Z Reader and Sequential Attention Models

The E-Z Reader family of models explains a substantial portion of eye-movement behavior by linking lexical processing milestones to saccadic programming. Reichle, Pollatsek, Fisher, and Rayner developed the original account to derive where and when the eyes move during continuous reading from assumptions about processing of words and programming of movements. The framework distinguishes an earlier recognition-related signal that can begin a forward saccade program from subsequent processing and shifts of attention. Because motor programming takes time, fixation duration is not a transparent measure of the instant a word becomes understood. This distinction lets the model explain how word difficulty influences a sequence of eye movements rather than one independent dwell time Reichle et al. (1998).

Sequential allocation of attention has predictive consequences. If a word is difficult, the initiation or timing of the next saccade may be delayed; if a short upcoming word is sufficiently processed before landing, it may be skipped. Preview benefit arises because information is obtained before the word becomes the currently fixated item. The model's explanatory power depends on whether one set of assumptions can account jointly for frequency effects, predictability, skipping, refixations, and the distribution of fixation durations. A fit to mean gaze duration alone is too weak a test, since alternative parameterizations could reproduce a similar average Reichle et al. (1998) Rayner et al. (2011).

Any comprehensive model must also specify the boundary between word recognition and comprehension. The E-Z Reader program primarily formalizes eye-movement control during reading; discourse integration, syntactic repair, and strategic rereading require further mechanisms or extensions. It is therefore possible for a simulation to reproduce typical saccadic patterns while leaving important determinants of meaning construction underspecified. These limitations become especially apparent with extended narrative, purposeful study, and readers who return to earlier material for reasoning. The research opportunity is to integrate a testable lexical-control theory with an independently specified account of sentence and discourse comprehension Just & Carpenter (1980) Frazier & Rayner (1982) Reichle et al. (1998).

SWIFT and Spatially Distributed Processing

SWIFT offers a contrasting formalization of reading control in which saccade timing has a more autonomous component and lexical processing can occur across a spatially distributed field. Engbert, Nuthmann, Richter, and Kliegl constructed a dynamical model to explain fixations and saccades in continuous reading, including effects of word difficulty and the processing of neighboring text. Its emphasis on parallel lexical activity raises different empirical questions from models that enforce more strictly sequential attention shifts. The important comparison concerns the predicted behavior under controlled preview, foveal-load, and neighboring-word manipulations, not a verbal distinction between strictly parallel and strictly serial cognition Engbert et al. (2005) Kliegl et al. (2006).

The SWIFT framework links graded processing activity with where the eyes are likely to move and how long the current fixation lasts. It can represent the fact that visual information enters from several words, although lexical activity and attentional selection are constrained by acuity and task. Different distributions of processing across positions can yield similar overall reading rates, so distinguishing mechanisms requires close attention to spillover, launch-site effects, parafoveal preview, and fixation-duration distributions. A model is scientifically useful when it makes quantitative commitments that can be contradicted by data collected in experiments it was not specifically tuned to reproduce Engbert et al. (2005) Staub (2011).

Neither formal family should be equated with a complete account of reading comprehension. Their principal outputs concern the dynamics of gaze and lexical processing, whereas understanding passages also entails proposition formation, memory updating, causal inference, and deliberate evaluation. The models can be connected with these higher-level processes through explicit mappings: sentence difficulty may modify lexical processing demand, while a monitoring goal may influence the destination of a return saccade. These mappings should be validated rather than inferred solely from the qualitative resemblance of simulated and observed eye trajectories Engbert et al. (2005) Kaakinen et al. (2003) Clifton et al. (2007).

Integrating Cognitive, Oculomotor, and Neural Explanations

Reading is a coordinated process in which language-specific computation interacts with continuous perceptual sampling and physical eye movements. A unifying explanation must distinguish the visual information received at a fixation, the linguistic processes triggered or continued there, the motor command selecting another region, and the longer-term meaning representation created from the accumulated text. These stages overlap in time and share resources. Evidence from moving-window studies constrains the spatial uptake of information; ambiguity experiments constrain the costs of constructing and revising analyses; formal models constrain allowable temporal trajectories. A satisfactory synthesis should preserve their different dependent measures while showing how each limits the others McConkie & Rayner (1975) Frazier & Rayner (1982) Reichle et al. (1998).

Neural recordings can add an independently measured time series to this framework. Wang and colleagues combined eye tracking with magnetoencephalography to investigate orthographic and semantic processing of parafoveal material in natural reading. Their findings strengthen the case for studying pretarget representation with methods beyond gaze behavior. The theoretical challenge is to determine how a neural information signal relates to attention, word identification, and subsequent comprehension. A neural response that distinguishes orthographic or semantic categories may reflect access to information; establishing its causal role in successful integration requires additional interventions and behavioral outcomes Wang et al. (2025) Schotter et al. (2012).

A multilevel model can be evaluated through a hierarchy of predictions. At the lowest level it predicts where and when the eyes move when text length, visibility, or landing site changes. At a lexical level it predicts changes caused by frequency and preview. At a sentence level it predicts difficulties and revisions under ambiguity or violated expectation. At a discourse level it predicts return movements and understanding when evidence must be integrated across sentences. Strong validation requires adequate performance across all these levels with a limited, explicit set of mechanisms, rather than a separate free parameter introduced after observing each result Kliegl et al. (2006) Duffy et al. (1988) Hyönä et al. (2002).

Evidence Appraisal, Reproducibility, and Open Problems

The evidence base is strongest for some relationships and more conditional for others. Gaze-contingent manipulations demonstrate that useful information is acquired outside direct fixation and that access to such information affects subsequent behavior. Repeated lexical contrasts establish reliable effects of frequency, length, and context on reading times, while syntactic ambiguity provides causal evidence that some comprehension demands alter gaze trajectories. However, these findings should not be compressed into the claim that longer fixations uniformly indicate poorer comprehension. A fixation can become longer because a reader is successfully elaborating an inference, because a display change disrupted processing, or because eye-movement execution was delayed McConkie & Rayner (1975) Henderson & Ferreira (1990) Frazier & Rayner (1982).

Some widely discussed effects require especially careful interpretation. Semantic preview benefits vary across experimental choices; parafoveal-on-foveal effects may depend on lexical properties, landing positions, and attentional assumptions; and word-skipping analyses are susceptible to conditioning on words that were actually fixated. Predictive-language findings do not always identify whether readers formed a specific word-level prediction or benefited from postrecognition facilitation. Formal models may fit the same aggregate features using different mechanisms, especially when evaluated on datasets close to those used during development. Independent replication, adversarial model comparisons, and distributional analyses can reduce these ambiguities Schotter et al. (2015) Luke & Christianson (2016) Rayner et al. (2011) Staub (2015).

The largest open problems concern explanation across levels and populations. Researchers need to establish when neural preview activity changes meaningful reading decisions, how strategic rereading supports durable comprehension, which mechanisms transfer across writing systems and age groups, and how individual language experience changes formal model parameters. A future agenda would combine preregistered linguistic contrasts, controlled display interventions, participant-level reading assessments, passage-level comprehension measures, open analytic specifications, and out-of-sample model evaluation. The Artificial Era adds a parallel methodological opportunity: computational text-processing systems can be evaluated against human gaze and comprehension benchmarks while preserving the difference between predicting a fixation sequence and demonstrating the human experience of understanding Kuperman & Van Dyke (2011) Engbert et al. (2005) Wang et al. (2025).

Conclusion

Reading comprehension and eye movements form a coherent scientific subject because understanding written language depends on active selection and processing of visible text over time. The experimental record establishes that readers gather information both before and during fixation, allocate gaze according to lexical and contextual demands, and sometimes revisit earlier material to resolve structural, referential, or discourse problems. These findings emerged from a research tradition that combines moving-window and boundary-change interventions with natural sentence reading, discourse tasks, formal modeling, and, increasingly, neural measurements. Each method constrains a different level of explanation, and the strongest inferences arise when their results converge McConkie & Rayner (1975) Just & Carpenter (1980) Rayner (2009).

The critical synthesis identifies two principles for the interpretation of gaze. First, eye position and fixation duration are behavioral manifestations of interacting visual, linguistic, motor, and strategic systems. They index opportunities and costs of processing under defined conditions, while comprehension must also be assessed through the reader's interpretation, inference, recall, or task performance. Second, scientific models need to account for the joint pattern of fixations, skips, regressions, and meaningful outcomes. E-Z Reader and SWIFT have clarified important control mechanisms; ambiguity and discourse research reveal the additional processes a full explanatory architecture must incorporate Reichle et al. (1998) Engbert et al. (2005) Hyönä et al. (2002).

Future progress depends on experiments that distinguish perceptual access from lexical prediction, recognition from integration, and strategic checking from error-driven repair. Better comprehension measures, more diverse reader samples, cross-script comparisons, transparent analysis choices, and direct tests of rival model predictions will refine the conclusions that eye tracking can support. Research on parafoveal neural representation contributes a new source of evidence but remains most powerful when joined to controlled linguistic manipulations and comprehension outcomes. The resulting scientific program treats eye movements as part of an experimentally testable explanation of reading rather than a self-sufficient substitute for understanding the text Schotter et al. (2012) Kaakinen & Hyönä (2010) Wang et al. (2025).

References

  • Angele, B., Slattery, T. J., Yang, J., Kliegl, R., & Rayner, K. (2008). Parafoveal processing in reading: Manipulating n + 1 and n + 2 previews simultaneously. Visual Cognition, 16(6), 697–707. https://doi.org/10.1080/13506280802009704
  • Blythe, H. I., Liversedge, S. P., Joseph, H. S. S. L., White, S. J., & Rayner, K. (2009). Visual information capture during fixations in reading for children and adults. Vision Research, 49(12), 1583–1591. https://doi.org/10.1016/j.visres.2009.03.015
  • Clifton, C., Jr., Staub, A., & Rayner, K. (2007). Eye movements in reading words and sentences. In R. P. G. van Gompel, M. H. Fischer, W. S. Murray, & R. L. Hill (Eds.), Eye movements: A window on mind and brain (pp. 341–371). Elsevier. https://doi.org/10.1016/B978-008044980-7/50017-3
  • Duffy, S. A., Morris, R. K., & Rayner, K. (1988). Lexical ambiguity and fixation times in reading. Journal of Memory and Language, 27(4), 429–446. https://doi.org/10.1016/0749-596X(88)90066-6
  • Engbert, R., Nuthmann, A., Richter, E. M., & Kliegl, R. (2005). SWIFT: A dynamical model of saccade generation during reading. Psychological Review, 112(4), 777–813. https://doi.org/10.1037/0033-295X.112.4.777
  • Frazier, L., & Rayner, K. (1982). Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14(2), 178–210. https://doi.org/10.1016/0010-0285(82)90008-1
  • Henderson, J. M., & Ferreira, F. (1990). Effects of foveal processing difficulty on the perceptual span in reading: Implications for attention and eye movement control. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16(3), 417–429. https://doi.org/10.1037/0278-7393.16.3.417
  • Hyönä, J., Lorch, R. F., Jr., & Kaakinen, J. K. (2002). Individual differences in reading to summarize expository text: Evidence from eye fixation patterns. Journal of Educational Psychology, 94(1), 44–55. https://doi.org/10.1037/0022-0663.94.1.44
  • Joseph, H. S. S. L., Liversedge, S. P., Blythe, H. I., White, S. J., & Rayner, K. (2009). Word length and landing position effects during reading in children and adults. Vision Research, 49(16), 2078–2086. https://doi.org/10.1016/j.visres.2009.05.015
  • Just, M. A., & Carpenter, P. A. (1980). A theory of reading: From eye fixations to comprehension. Psychological Review, 87(4), 329–354. https://doi.org/10.1037/0033-295X.87.4.329
  • Kaakinen, J. K., Hyönä, J., & Keenan, J. M. (2003). How prior knowledge, WMC, and relevance of information affect eye fixations in expository text. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29(3), 447–457. https://doi.org/10.1037/0278-7393.29.3.447
  • Kaakinen, J. K., & Hyönä, J. (2010). Task effects on eye movements during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(6), 1561–1566. https://doi.org/10.1037/a0020693
  • Kliegl, R., Nuthmann, A., & Engbert, R. (2006). Tracking the mind during reading: The influence of past, present, and future words on fixation durations. Journal of Experimental Psychology: General, 135(1), 12–35. https://doi.org/10.1037/0096-3445.135.1.12
  • Kuperman, V., & Van Dyke, J. A. (2011). Effects of individual differences in verbal skills on eye-movement patterns during sentence reading. Journal of Memory and Language, 65(1), 42–73. https://doi.org/10.1016/j.jml.2011.03.002
  • Kuperman, V., & Van Dyke, J. A. (2013). Reassessing word frequency as a determinant of word recognition for skilled and unskilled readers. Journal of Experimental Psychology: Human Perception and Performance, 39(3), 802–823. https://doi.org/10.1037/a0030859
  • Luke, S. G., & Christianson, K. (2016). Limits on lexical prediction during reading. Cognitive Psychology, 88, 22–60. https://doi.org/10.1016/j.cogpsych.2016.06.002
  • McConkie, G. W., & Rayner, K. (1975). The span of the effective stimulus during a fixation in reading. Perception & Psychophysics, 17(6), 578–586. https://doi.org/10.3758/BF03203972
  • Rayner, K. (1975a). The perceptual span and peripheral cues in reading. Cognitive Psychology, 7(1), 65–81. https://doi.org/10.1016/0010-0285(75)90005-5
  • Rayner, K. (1975b). Parafoveal identification during a fixation in reading. Acta Psychologica, 39(4), 271–281. https://doi.org/10.1016/0001-6918(75)90011-6
  • Rayner, K. (1986). Eye movements and the perceptual span in beginning and skilled readers. Journal of Experimental Child Psychology, 41(2), 211–236. https://doi.org/10.1016/0022-0965(86)90037-8
  • Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychological Bulletin, 124(3), 372–422. https://doi.org/10.1037/0033-2909.124.3.372
  • Rayner, K. (2009). Eye movements and attention in reading, scene perception, and visual search. Quarterly Journal of Experimental Psychology, 62(8), 1457–1506. https://doi.org/10.1080/17470210902816461
  • Rayner, K., Slattery, T. J., Drieghe, D., & Liversedge, S. P. (2011). Eye movements and word skipping during reading: Effects of word length and predictability. Journal of Experimental Psychology: Human Perception and Performance, 37(2), 514–528. https://doi.org/10.1037/a0020990
  • Reichle, E. D., Pollatsek, A., Fisher, D. L., & Rayner, K. (1998). Toward a model of eye movement control in reading. Psychological Review, 105(1), 125–157. https://doi.org/10.1037/0033-295X.105.1.125
  • Schotter, E. R., Angele, B., & Rayner, K. (2012). Parafoveal processing in reading. Attention, Perception, & Psychophysics, 74(1), 5–35. https://doi.org/10.3758/s13414-011-0219-2
  • Schotter, E. R., Lee, M., Reiderman, M., & Rayner, K. (2015). The effect of contextual constraint on parafoveal processing in reading. Journal of Memory and Language, 83, 118–139. https://doi.org/10.1016/j.jml.2015.04.005
  • Slattery, T. J., Angele, B., & Rayner, K. (2011). Eye movements and display change detection during reading. Journal of Experimental Psychology: Human Perception and Performance, 37(6), 1924–1938. https://doi.org/10.1037/a0024322
  • Staub, A. (2011). The effect of lexical predictability on distributions of eye fixation durations. Psychonomic Bulletin & Review, 18(2), 371–376. https://doi.org/10.3758/s13423-010-0046-9
  • Staub, A. (2015). The effect of lexical predictability on eye movements in reading: Critical review and theoretical interpretation. Language and Linguistics Compass, 9(8), 311–327. https://doi.org/10.1111/lnc3.12151
  • Wang, L., Frisson, S., Pan, Y., & Jensen, O. (2025). Fast hierarchical processing of orthographic and semantic parafoveal information during natural reading. Nature Communications, 16, Article 8893. https://doi.org/10.1038/s41467-025-63916-y
  • White, S. J., Johnson, R. L., Liversedge, S. P., & Rayner, K. (2008). Eye movements when reading transposed text: The importance of word-beginning letters. Journal of Experimental Psychology: Human Perception and Performance, 34(5), 1261–1276. https://doi.org/10.1037/0096-1523.34.5.1261