Keith Rayner – 1943 2015

Biography

For more than four decades, the psychological investigation of reading was fundamentally defined, reshaped, and propelled forward by the empirical ingenuity and theoretical rigor of Keith Rayner (1943–2015). Prior to Rayner’s seminal contributions in the mid-1970s, the scientific understanding of reading relied predominantly on retrospective, offline assessments—such as post-reading comprehension questionnaires, tachistoscopic word recognition thresholds, and verbal response latencies. While these paradigms revealed critical insights into the structure of the mental lexicon, they lacked the temporal resolution necessary to track the rapid, subconscious processes that unfold dynamically as the human eye traverses a line of printed text. Rayner recognized that eye movements do not merely serve as the passive mechanical transport system of the visual apparatus; rather, they provide an exquisitely sensitive, millisecond-by-millisecond behavioral readout of linguistic decoding, visual attention, cognitive integration, and central executive control.

Through the conception and realization of revolutionary gaze-contingent display technologies—most notably the moving-window and boundary paradigms—Rayner transformed eye tracking from a descriptive recording technique into a precision experimental tool. By dynamically manipulating textual displays in direct synchronization with the reader’s instantaneous oculomotor coordinates, he established the empirical parameters of the human perceptual span, charted the temporal micro-structure of parafoveal information extraction, and dismantled persistent dogmas regarding trans-saccadic visual memory and whole-word apprehension. His insights laid the empirical bedrock for the E-Z Reader model, one of the most durable and computationally precise architectures of visual cognition and eye-movement control ever formulated in cognitive science.

Yet Rayner’s scientific footprint extended far beyond the mechanics of word recognition. His collaborative endeavors bridged visual psychophysics, syntactic parsing, semantic ambiguity resolution, developmental psychology, aging, neuropsychology, scene perception, and visual search across diverse orthographic systems. His institutional leadership at the University of Massachusetts Amherst and the University of California, San Diego established vibrant global hubs of experimental psycholinguistics, while his editorship of premier journals set enduring benchmarks for methodological rigor and theoretical clarity. This comprehensive monograph provides an exhaustive intellectual history and technical appraisal of Keith Rayner’s lifetime of scholarship, charting his trajectory from early behavioral foundations to his status as the foundational architect of modern reading science.

1. Biographical Overview and Academic Trajectory (1943–1974)

1.1 Early Life and Undergraduate Foundations

Keith Rayner was born on June 20, 1943, in the historic coastal town of Dover, England, during the height of the Second World War. In 1949, at the age of six, Rayner immigrated with his family to the United States, settling in Utah. Growing up in the intermountain West, Rayner exhibited an early intellectual curiosity marked by disciplined inquiry, an affinity for empirical observation, and an abiding fascination with language, literature, and human behavior. His formative educational years unfolded within an environment that valued industrious self-reliance, characteristics that would later define his relentless laboratory work ethic and methodological precision.

Rayner entered the University of Utah for his undergraduate education, where he pursued an interdisciplinary course of study focused on psychology and linguistics. During the mid-to-late 1960s, the American psychological establishment was undergoing the profound intellectual transformation known as the cognitive revolution. The long-dominant behaviorist paradigm—which relegated mental processes to an unobservable “black box” and focused exclusively on stimulus-response contingencies—was rapidly giving way to information-processing models that conceptualized the human mind as a sophisticated computational architecture. At Utah, Rayner absorbed both the rigorous operational methodologies of behavioral analysis and the emergent theoretical frameworks of cognitive psychology, information theory, and generative linguistics.

His early intellectual interests crystallized around questions of reading acquisition, educational pedagogy, and the basic cognitive mechanisms underlying visual literacy. He became deeply intrigued by how children learn to map abstract visual orthography onto internal phonological and semantic representations, and why fluent adult reading appears so phenomenologically effortless despite involving extraordinarily complex perceptual and cognitive operations. Recognizing that traditional educational diagnostics and static tachistoscopic experiments could not adequately capture the real-time dynamics of continuous reading, Rayner earned both his Bachelor of Science and Master of Science degrees in Psychology from the University of Utah, assembling a formidable foundation in experimental design, statistical measurement, and psycholinguistic theory that positioned him for advanced doctoral training at the frontier of the discipline.

1.2 Doctoral Training at Cornell University

In 1970, Rayner commenced his doctoral studies in cognitive psychology at Cornell University, an institution renowned for its avant-garde research in developmental psychology, ecological perception, and cognitive science. At Cornell, Rayner was immersed in an intellectually vibrant milieu populated by intellectual giants such as James J. Gibson, Eleanor J. Gibson, and Ulric Neisser. Gibsonian ecological optics challenged simplistic notions of visual passive reception, emphasizing that perception is an active, exploratory act continuous with environmental interaction. Eleanor Gibson’s groundbreaking investigations into perceptual learning and the development of reading principles further stimulated Rayner’s drive to operationalize the perceptual micro-mechanisms of textual decoding.

Crucially, Rayner entered into an intensely fruitful research partnership and mentorship under George W. McConkie, an experimental psychologist who was pioneering the application of real-time computer systems to the investigation of reading behavior. Together, McConkie and Rayner recognized that the nascent field of eye-movement research suffered from severe technological and theoretical limitations. Although researchers had recorded oculomotor behavior since the late nineteenth-century work of Louis Émile Javal and Edmund Huey, previous generations of scientists were constrained to measuring global metrics—such as aggregate fixation counts or total passage reading times—without the ability to experimentally manipulate the visual input that a reader received during the course of an individual fixation.

Rayner’s doctoral dissertation, formally titled The Perceptual Span and Peripheral Cues in Reading (completed in 1974), represented a watershed moment in cognitive psychology. Working within the Cornell laboratory, Rayner set out to answer a question that had eluded psychologists for nearly a century: precisely how much visual, orthographic, and linguistic information does a human reader acquire during a single, discrete fixation of approximately 250 milliseconds? To resolve this question, Rayner and McConkie engineered computer-controlled display paradigms that dynamically altered visual text in contingent response to the real-time position of the reader’s eye. This dissertation laid the foundational hypotheses regarding the coupling of oculomotor control and linguistic processing, demonstrating unequivocally that visual information extraction during reading is strictly bounded, functionally asymmetrical, and tightly synchronized with higher-order cognitive processing.

1.3 Early Career at the University of Rochester

Upon completing his doctorate at Cornell in 1973, Rayner secured his initial faculty appointment as an Assistant Professor in the Department of Psychology at the University of Rochester. Rochester provided an exceptionally fertile environment for an emerging experimentalist, boasting robust institutional strengths in visual psychophysics, optics, neuroscience, and computer science. It was during these foundational Rochester years (1973–1978) that Rayner began constructing his first fully independent oculomotor research laboratory, assembling custom-built electronic hardware, optomechanical eye trackers, and real-time digital minicomputers.

The technical challenges confronting eye-movement researchers in the mid-1970s were immense. Modern plug-and-play, video-based corneal reflection trackers were decades away. Rayner utilized sophisticated Dual-Purkinje-Image (DPI) eye trackers and early scleral reflectance apparatuses. These instruments required intricate optical alignment, continuous analog-to-digital signal calibration, and high-performance interfacing with minicomputers such as the PDP-11. The computer had to sample horizontal and vertical eye positions at high frequencies (typically 1000 Hz, or once every millisecond), execute algorithmic calculations to verify fixation stability or saccade initiation, and instantly update the textual display presented on specialized cathode-ray tube (CRT) monitors coated with fast-decay phosphors—all within an ultra-low latency window of fewer than 5 milliseconds to prevent visual artifacts.

At Rochester, Rayner demonstrated exceptional technical virtuosity and methodological rigor. He published a series of meticulously controlled empirical studies that systematically refined calibration protocols, artifact rejection criteria, and spatial parsing algorithms for eye-movement records. His laboratory became an engine of discovery, examining how typographical variations, letter spacing, and visual degradation influenced ocular behavior. Crucially, these early Rochester years cemented Rayner’s collaborations that led directly to the publication of the seminal gaze-contingent paradigms in 1975. By the time he was recruited to the University of Massachusetts Amherst in 1978, Rayner had already established himself as the premier innovator in the experimental analysis of human eye movements during complex cognitive tasks.

2. Methodological Revolutions: The Gaze-Contingent Paradigms

2.1 The Moving-Window Technique

The introduction of the gaze-contingent moving-window paradigm by George McConkie and Keith Rayner in their landmark 1975 paper, “The span of the effective stimulus during a fixation in reading,” published in Perception & Psychophysics, revolutionized experimental psycholinguistics. The conceptual elegance of the moving-window technique lay in its ability to directly measure the spatial boundaries of visual information extraction by restricting unmanipulated text to a defined, movable window centered precisely on the reader’s point of gaze, while simultaneously masking all text falling outside that window.

Technically, the paradigm required an instantaneous, closed-loop interface between the eye tracker and the display hardware. When the participant fixated on a specific letter within a sentence, the computer maintained a clear “window” of readable characters surrounding that fixation point (for example, eight characters to the left and eight characters to the right). All characters outside this defined window were replaced in real time with visually masking characters, such as uniform strings of Xs or visually similar letter substitutions (e.g., replacing ascenders with ascenders and descenders with descenders), while preserving or destroying space information depending on the experimental manipulation. As the reader initiated a saccadic eye movement to a new location, the eye tracker detected the saccade, and the computer refreshed the display buffer during the saccadic transition—when visual sensitivity is biologically suppressed—so that when the eye landed at its new destination, the clear window had moved instantaneously to encompass the new fixation point.

By systematically varying the horizontal dimensions and symmetry of this unmasked window (e.g., testing symmetrical windows of varying sizes, as well as asymmetrical configurations biased to the left or right of fixation), Rayner and McConkie discovered that reading rate, fixation durations, and saccade lengths remained indistinguishable from natural, unconstrained reading only when the window extended approximately 3 to 4 character spaces to the left of fixation and up to 14 to 15 character spaces to the right of fixation for readers of English. Any further expansion of the window to the right yielded no additional processing benefit, proving definitively that skilled readers do not absorb whole lines of text simultaneously—as speed-reading commercial courses had long claimed—but rather operate within a tightly constrained, functionally asymmetrical spotlight of visual attention.

2.2 The Moving-Mask Paradigm

Complementing the moving-window technique, Keith Rayner and his colleagues developed the moving-mask paradigm (sometimes termed the foveal-mask technique). Where the moving-window paradigm preserves central vision and masks the periphery to determine the outer boundaries of information extraction, the moving-mask paradigm inverts this logic: it places an opaque, gaze-contingent visual mask directly over the foveal region centered on the fixation point (typically covering the fixated word or the central 1 to 2 degrees of visual angle), while leaving the surrounding parafoveal and peripheral text completely clear, intact, and unmasked.

The moving-mask technique was engineered to evaluate the indispensable contribution of central foveal vision to continuous reading and to determine whether peripheral vision alone could sustain any meaningful degree of lexical or linguistic processing. The empirical findings derived from this paradigm were stark and unambiguous. When central foveal vision was masked, even by a tiny patch covering only the central 1 to 3 characters, reading fluency collapsed catastrophically. Readers’ fixation durations lengthened dramatically, saccade amplitudes contracted into disoriented searching behaviors, regressive saccades multiplied, and overall reading rates plummeted to near zero. Parafoveal and peripheral vision, deprived of foveal guidance, proved utterly incapable of sustaining fluent lexical identification or syntactic comprehension.

Crucially, the moving-mask paradigm provided powerful experimental validation for the “eye-mind assumption”—the foundational psycholinguistic premise, prominently articulated by Marcel Just and Patricia Carpenter, that the eye remains fixated on a word for as long as that word is undergoing active cognitive decoding. By demonstrating that readers cannot simply “read ahead” in the parafovea and use foveal fixations merely as arbitrary physical anchors, Rayner established that central foveal identification is the primary, indispensable computational engine of visual reading, with parafoveal processing serving exclusively as an anticipatory, pre-processing staging ground.

2.3 The Boundary Paradigm

In 1975, Rayner published a single-authored study in Cognitive Psychology titled “The perceptual span and peripheral cues in reading,” introducing what would become the most universally celebrated and widely utilized methodological innovation in the history of reading research: the gaze-contingent boundary paradigm. While the moving window established the overall spatial dimensions of the perceptual span, the boundary paradigm was explicitly engineered to dissect the temporal and linguistic micro-structure of parafoveal preview—uncovering the exact nature of the information extracted from an upcoming word before that word is directly fixated.

The operational mechanics of the boundary paradigm represent a masterpiece of experimental chronometry. In an experimental sentence display, an invisible spatial boundary trigger is computationally anchored immediately before a designated target word location (word n+1). Prior to the reader’s eye crossing this invisible boundary—while the gaze is fixated on the preceding launch site word (word n)—a pre-selected preview stimulus occupies the target word slot. This preview can be systematically manipulated by the experimenter to be identical to the target, visually/orthographically similar, phonologically identical (e.g., homophonic), semantically related, or completely unrelated and visually anomalous.

As the reader initiates a forward saccadic eye movement to transition from word n to the target word, the trajectory of the eye crosses the invisible boundary. This boundary crossing instantly triggers the computer hardware to alter the display buffer, replacing the preview string with the true, correct target word. Because this display change is completed within a matter of milliseconds—well within the biological duration of saccadic suppression, during which visual sensitivity is massively attenuated by the central nervous system—the reader has no conscious perceptual awareness that any visual transformation has occurred on the screen.

By measuring the subsequent fixation durations on the target word once the eye lands upon it (recording metrics such as first fixation duration, single fixation duration, and gaze duration), Rayner quantified the magnitude of the parafoveal preview benefit. When the preview matched the target word identically, subsequent fixation durations on that word were shortened by approximately 20 to 40 milliseconds compared to conditions featuring an unrelated preview. By manipulating the nature of the preview string (e.g., presenting “chute” as a preview for “shoot” to isolate phonology, or “table” as a preview for “chair” to isolate semantics), Rayner provided psycholinguists with an unprecedented window into the temporal unfolding of orthographic, phonological, and semantic representations in real time prior to direct ocular inspection.

3. The Nature and Scope of Perceptual Span in Reading

3.1 Asymmetry and Directionality of the Span

The discovery that the human perceptual span during reading is profoundly asymmetrical stands as one of Keith Rayner’s most robust empirical triumphs. In alphabetic orthographies that are read from left to right, such as English, French, and Spanish, the perceptual span extends only 3 to 4 character spaces to the left of the central point of fixation, but reaches up to 14 to 15 character spaces to the right. Early visual theorists had hypothesized that this asymmetry might reflect an innate biological or neuroanatomical hardwiring—perhaps stemming from the structural organization of the visual cortex, hemispheric lateralization of language processing in the left cerebral hemisphere, or differences in retinal ganglion cell distribution across nasal and temporal hemiretinas.

Rayner systematically demolished this biological determinism hypothesis through a series of ingenious cross-linguistic and cross-orthographic investigations. Collaborating with international researchers, Rayner evaluated the perceptual span of skilled readers of Hebrew and Arabic—Semitic orthographies that are written and read strictly from right to left. Utilizing the gaze-contingent moving-window paradigm adapted for right-to-left typography, Rayner and his colleagues demonstrated that the functional asymmetry of the perceptual span completely reversed: Hebrew and Arabic readers exhibited a perceptual span extending 3 to 4 characters to the right of fixation and up to 14 to 15 characters to the left.

Furthermore, studies conducted with bidirectional bilinguals (individuals fluent in both English and Hebrew) revealed that the directionality of the attentional span was exquisitely dynamic and context-dependent. When reading English text, the bilingual reader’s perceptual span immediately expanded to the right and contracted to the left; when switching to Hebrew text within the same experimental session, the span flipped within milliseconds, expanding to the left and contracting to the right. Subsequent investigations involving vertically oriented scripts, such as traditional Chinese or Japanese text printed in vertical columns, revealed that the perceptual span elongated along the vertical dimension, extending downward in the direction of reading progression. These findings demonstrated conclusively that the perceptual span is not constrained by structural retinal or cortical architecture, but represents an exceptionally flexible, learned allocation of visual-spatial attention governed entirely by reading habits, orthographic conventions, and task demands.

3.2 Sub-Divisions of the Perceptual Span

Through decades of fine-grained psychophysical experimentation, Rayner established that the perceptual span is not a monolithic, uniform zone of clear perception. Instead, it is functionally sub-divided into concentric, tiered regions of information extraction, operating under distinct optical, neurobiological, and cognitive constraints. Rayner carefully distinguished between what he designated as the “visual span” (the physical region within which the visual system can resolve high-contrast visual features, dictated largely by retinal acuity and photoreceptor density) and the “cognitive perceptual span” (the functional region from which linguistic information is successfully extracted and integrated during continuous reading).

Within the cognitive perceptual span itself, Rayner identified two distinct sub-zones: an inner zone of letter identification and an outer zone of word boundary and length detection. The zone of letter identification is relatively compact, extending only about 7 to 8 character spaces to the right of fixation in skilled English readers. Beyond this 8-character boundary, readers are fundamentally incapable of recognizing individual letter identities or determining the lexical status of an upcoming string. If letter identities are altered beyond 8 characters to the right of fixation (provided that spaces and gross visual contours remain intact), readers do not notice the alterations, and their reading rates remain completely unaffected.

However, the outer zone of the perceptual span—extending from 8 characters out to approximately 14 or 15 character spaces to the right—remains critically informative. From this distal region, readers reliably extract word length information and word boundary cues, mediated predominantly by the white spaces intervening between words. Rayner proved that this outer-zone information is indispensable for low-level oculomotor programming: the visual system uses these distal word-length cues to identify the location of the next candidate word and calculate the precise landing site for the upcoming forward saccade. When spaces between words were artificially removed or filled with neutral characters (such as hyphens or asterisks), even when letter identities were completely unmasked, reading speeds declined by over 50%, and saccadic targeting accuracy deteriorated into erratic, unguided fixational patterns.

3.3 Factors Modulating Perceptual Span Size

Rayner’s empirical investigations demonstrated that the perceptual span is not an invariant, fixed-capacity biological constant, but rather a dynamic cognitive window that continuously expands and contracts in response to immediate processing demands. Among the most potent modulators of span size identified by Rayner is textual difficulty. When texts exhibit complex syntactical structures, dense conceptual ideas, or archaic lexical vocabulary, the horizontal extent of the perceptual span contracts markedly. Conversely, when reading light, predictable, narrative prose, the span expands to its maximal physical limits.

Typographic legibility and visual layout exert an equally pronounced influence. Rayner and his colleagues conducted extensive evaluations of font design, proportional letter spacing, visual degradation, and contrast attenuation. Degrading text contrast or introducing irregular typographical spacing was shown to compress the perceptual span, forcing the oculomotor system to shorten saccades and execute more frequent, localized fixations to compensate for the impoverished visual signal.

Crucially, Rayner demonstrated that reader skill and developmental maturity dictate the operational dimensions of the span. Developing child readers possess significantly narrower perceptual spans than skilled adults, often extending only 5 to 6 characters to the right of fixation during early primary school. The span gradually expands across developmental schooling as decoding becomes automated and cognitive resources are freed for anticipatory parafoveal processing. Similarly, readers with dyslexia or atypical cognitive profiles exhibit severely constricted perceptual spans.

Underpinning all of these dynamic modulations is the principle that Rayner formulated as the “foveal load effect.” Rayner demonstrated that when the fixated word (word n) is cognitively challenging to process—due to low lexical frequency, ambiguous syntactic category, or contextual implausibility—the immediate foveal processing load consumes the lion’s share of available central visual-attentional capacity. Under conditions of heavy foveal load, visual attention cannot be spared to project outward into the parafovea; consequently, the effective perceptual span temporarily shrinks. Only when the fixated foveal word is processed with rapid, low-effort fluency does the attentional spotlight project forward into the parafoveal zone, facilitating rich preview extraction and enabling longer forward saccades.

4. Parafoveal Processing and Lexical Information Extraction

4.1 Orthographic and Phonological Preview Effects

The boundary paradigm afforded Keith Rayner and his research group an unprecedented laboratory methodology to systematically unpack the micro-temporal dynamics of parafoveal information extraction. In a long series of rigorous experimental studies, Rayner sought to delineate the precise taxonomy of codes—orthographic, phonological, and semantic—that are activated prior to a word receiving direct ocular fixation. By contrasting different experimental preview conditions against baseline controls, Rayner proved that skilled readers pre-activate both low-level orthographic features and abstract letter identities within the parafovea.

When the parafoveal preview string shared initial orthographic letters with the target word (e.g., presenting a preview such as “chart” before displaying the target “chest”), readers experienced a substantial reduction in subsequent fixation duration on the target word once fixated. Rayner showed that the magnitude of this orthographic preview benefit was highly sensitive to letter position: previewing initial letters yielded dramatic processing advantages, whereas previewing medial or trailing letters yielded minor or negligible benefits. This finding established that parafoveal orthographic decoding proceeds in a strictly left-to-right, letter-position-dependent trajectory in alphabetic scripts, reinforcing the primary role of word-initial cues in lexical access.

Beyond visual-orthographic form, Rayner made pioneering contributions to resolving the contentious theoretical debate over the role of phonology in visual reading. Proponents of direct-access models had asserted that skilled readers map visual letters directly onto semantic concepts without activating intermediate speech sounds. Rayner and his collaborators, particularly Alexander Pollatsek, deployed the boundary paradigm to evaluate whether phonological codes are generated parafoveally. Utilizing homophonic previews that were visually dissimilar but phonetically identical to the target (e.g., using “chute” as a preview for the target word “shoot”, contrasted against visually matched non-homophonic controls such as “suite”), Rayner and colleagues demonstrated a clear, statistically robust phonological preview benefit.

Readers fixated significantly shorter on the target word when the parafoveal preview had activated its correct phonological representation, even though the orthographic surface form differed. This phonological preview effect was observed across diverse alphabetic scripts and was later extended into logographic orthographies, such as Chinese, where homophonic characters yielded pronounced preview advantages. These findings demonstrated that phonological code activation is not a secondary, post-lexical artifact of silent reading, but a rapid, automatic, and foundational component of early lexical identification that initiates while the word still lingers in parafoveal vision.

4.2 The Semantic Preview Debate

While the empirical reality of orthographic and phonological preview benefits was rapidly embraced by cognitive psychology, the question of whether skilled readers can extract semantic information from the parafovea sparked one of the most vigorously contested scientific debates of Rayner’s career. In his original 1975 boundary experiment, Rayner tested semantically related previews (e.g., presenting “dog” as a parafoveal preview for the target word “cat”) and observed no measurable preview benefit in subsequent target fixation durations; fixation times on the target following a semantically related preview were virtually identical to those following completely unrelated word previews.

For more than two decades, Rayner maintained a rigorous skepticism toward claims of parafoveal semantic processing in English. He argued that given the steep drop-off in visual acuity across the parafovea and the strict attentional bottlenecks governing human cognition, readers simply do not possess sufficient time or computational bandwidth within a 200–250 millisecond fixation on word n to completely resolve the orthography, phonology, and semantic identity of word n+1 prior to ocular execution. In Rayner’s serial-processing framework, lexical access for word n must achieve near completion before attentional resources can shift forward to begin processing word n+1; hence, while early visual and phonological cues could be registered, full semantic extraction could rarely occur parafoveally under standard conditions in English.

However, cross-linguistic investigations in the late 1990s and early 2000s began producing divergent findings. Researchers conducting boundary paradigm experiments in non-alphabetic languages—most notably logographic Chinese—consistently uncovered robust semantic preview benefits. In Chinese, where individual characters visually compress high-density semantic information into compact spatial units and where words typically comprise only one or two characters, parafoveal semantic extraction proved highly reliable. Similarly, studies involving morphologically transparent compound words in German revealed subtle semantic preview effects under specialized syntactic configurations.

Rayner approached these challenges with exemplary scientific openness, collaborating extensively with international teams to replicate and understand these discrepancies. The integration of high-density event-related potentials (ERPs) with gaze-contingent eye tracking (eye-fixation-related potentials) eventually provided the critical empirical bridge. While classical behavioral fixation durations in English often failed to show statistically reliable semantic preview benefits due to ceiling effects and saccadic execution deadlines, neural markers such as the N400 component revealed subtle, early signatures of semantic priming originating from the parafovea. Rayner ultimately reconciled these findings: while semantic preview extraction is indeed constrained by severe spatial and temporal bottlenecks in sprawling alphabetic orthographies like English, it can and does occur when orthographic density is high, when foveal processing load is low, and when contextual constraints strongly pre-activate the semantic network of the upcoming word.

4.3 Integration Across Saccades: Trans-Saccadic Memory

Every time a reader executes a saccadic eye movement—a ballistic motion lasting roughly 20 to 40 milliseconds—the retinal projection of the visual world sweeps violently across the photoreceptor mosaic at speeds exceeding several hundred degrees per second. Despite this continuous retinal disruption, human beings perceive an entirely stable, coherent, and uninterrupted visual environment. In the 1970s and 1980s, a prominent theoretical hypothesis asserted that the visual system achieves this continuity through “trans-saccadic fusion”—a hypothesized mechanism whereby the sensory image from the pre-saccadic fixation is retained in iconic memory and physically overlaid, pixel-by-pixel, onto the sensory image received during the post-saccadic fixation, like aligning two photographic transparencies.

Keith Rayner conducted definitive experimental studies that decisively refuted the trans-saccadic fusion hypothesis. In a series of brilliant display-change experiments, Rayner manipulated the visual case of text across saccades. For example, a reader fixating on word n would see the upcoming word printed in alternating lowercase letters (e.g., tAbLe). During the execution of the saccade toward that word, the computer instantly altered the display so that the word was rendered in alternating uppercase letters (e.g., TaBlE). If trans-saccadic memory relied on iconic visual fusion, the physical superposition of tAbLe and TaBlE would generate a garbled visual composite of overlapping letter segments, severely disrupting lexical identification and inflating fixation durations.

The results were startlingly clear: readers exhibited absolutely no disruption whatsoever. Fixation durations, reading fluency, and comprehension were identical regardless of whether visual case was preserved or altered across the saccade. Rayner demonstrated that what is preserved and integrated across saccadic jumps is not an iconic visual image or photographic bitmap, but abstract orthographic identities and spatial-coordinate representations. The human brain rapidly strips away superficial visual characteristics (such as exact font style, retinal size, and case geometry) during the pre-saccadic fixation, transferring only abstract letter identities (e.g., the concept of the letter /T/, whether printed as ‘T’ or ‘t’) into trans-saccadic working memory. This discovery permanently reshaped cognitive psychology’s understanding of visual memory, proving that visual perception across eye movements is conceptually abstract rather than sensory-pictorial.

5. The E-Z Reader Model: Foundations and Theoretical Architecture

5.1 Core Premises and the Eye-Mind Link

By the late 1990s, the vast corpus of empirical discoveries generated by Keith Rayner and his collaborators demanded a unified, mathematically formal computational architecture. In 1998, Keith Rayner, along with cognitive scientists Erik D. Reichle, Alexander Pollatsek, and Donald L. Fisher, published a landmark theoretical treatise in Psychological Review introducing the E-Z Reader model. The model, which was subsequently refined, expanded, and comprehensively articulated in their 2003 paper in Behavioral and Brain Sciences, stands as the most influential and rigorously tested computational model of eye-movement control in reading.

The E-Z Reader model was explicitly designed to operationalize and refine the classic “eye-mind hypothesis” originally formulated by Marcel Just and Patricia Carpenter in 1980. Just and Carpenter had proposed an immediate, tightly synchronized link between the locus of eye fixation and the locus of mental processing, positing that the eye remains locked on a word until all levels of lexical and conceptual processing are completed. While revolutionary, Just and Carpenter’s strict formulation could not account for several critical empirical phenomena uncovered by Rayner, including parafoveal preview benefits, word skipping, and the physiological latency required for the brain to program and execute an oculomotor saccade (typically 150 to 200 milliseconds).

The core theoretical breakthrough of the E-Z Reader model was its formal decoupling of covert visual attention from overt ocular motor programming. The model asserts that while visual attention is strictly serial—allocating its processing spotlight to only one word at a time—the signal to initiate a saccadic eye movement is triggered at an early stage of lexical decoding, well before the cognitive system has completed full lexical access. By establishing a two-stage architecture for lexical processing and interfacing it with a two-stage oculomotor engine, E-Z Reader provided an elegant mathematical solution to the paradox of how the brain navigates saccadic latencies while maintaining dynamic cognitive control over the eyes.

5.2 Computational Stages: L1, L2, and Saccadic Programming

The architectural engine of the E-Z Reader model is organized around four sequentially coordinated processing stages: two linguistic/cognitive stages ($L_1$ and $L_2$) and two oculomotor programming stages ($M_1$ and $M_2$), operating downstream of early visual sensory encoding ($V$). When the eye lands on a new word (word n), early visual processing ($V$) commences, transmitting raw sensory information through the visual pathways.

Once visual features are extracted, lexical processing begins with Stage 1 Lexical Access ($L_1$), conceptually defined as a rapid, preliminary “familiarity check.” During $L_1$, the cognitive system computes a fast probabilistic assessment of whether the visual input matches a known entry in the mental lexicon. The mathematical duration of $L_1$ is heavily modulated by two primary psycholinguistic variables: word frequency (how often the word occurs in the natural language) and contextual predictability (how easily the word can be anticipated given the preceding discourse context). Words of high frequency and high predictability require markedly fewer computational cycles to complete $L_1$.

The completion of $L_1$ serves as a critical biological signal: it indicates to the oculomotor system that word n is familiar enough that the eye should prepare to depart. Crucially, the completion of $L_1$ immediately triggers the first phase of oculomotor programming: Stage 1 Motor Programming ($M_1$). The $M_1$ phase represents a labile (modifiable and cancelable) motor program aimed at moving the eye to the next target word (word n+1). Because $M_1$ is initiated prior to full lexical identification, the oculomotor system gains a head start on overcoming physical saccadic latencies.

Meanwhile, the cognitive system immediately proceeds to Stage 2 Lexical Access ($L_2$), defined as complete lexical identification. During $L_2$, the brain resolves full phonological coding, semantic activation, and morphological binding. The critical theoretical event in E-Z Reader occurs at the exact millisecond that $L_2$ terminates: completion of $L_2$ causes the covert spotlight of visual attention to immediately decouple from word n and shift forward to word n+1. This covert attentional shift occurs while the eyes are still physically fixated on word n! Consequently, the cognitive system begins pre-processing word n+1 in parafoveal vision during the remaining fixation duration of word n.

While parafoveal processing of word n+1 is initiating, the oculomotor system completes the non-cancelable Stage 2 Motor Programming ($M_2$), which represents the ballistic execution of the neuromuscular command. The eye then physically launches its saccade toward word n+1. Through this intricate computational choreography, E-Z Reader quantitatively simulated with unprecedented accuracy how word frequency, predictability, and length dictate empirical fixation durations across thousands of experimental sentences.

5.3 Explaining Complex Oculomotor Phenomena

The true genius of the E-Z Reader architecture lies in its ability to account for complex, seemingly anomalous oculomotor phenomena through the natural interaction of its computational stages. A prime example is the mechanism of word skipping. In skilled reading, approximately 20% to 30% of all words in a sentence are skipped entirely, never receiving direct fixation. Function words (such as “the,” “and,” “of”) are skipped upwards of 65% to 75% of the time, whereas long content words are rarely skipped.

E-Z Reader explains word skipping through the temporal dynamics of the labile $M_1$ stage. When word n is short, highly frequent, or highly predictable, its $L_1$ and $L_2$ stages complete exceptionally fast. Visual attention shifts covertly to word n+1 while the eye is still fixated on word n and while the oculomotor program to move to word n+1 is still in its labile $M_1$ phase. If word n+1 is also short or familiar, its own familiarity check ($L_1$ of word n+1) can complete before the $M_1$ program for word n transitions into the non-labile $M_2$ phase. When this occurs, the cognitive system cancels the original motor program targeting word n+1 and immediately instantiates a new motor program targeting word n+2. The physical eye movement then sails directly over word n+1, landing squarely on word n+2. Thus, skipping emerges as an emergent consequence of serial attention shifting ahead of saccadic execution.

Furthermore, E-Z Reader provides an elegant computational account of refixation behaviors and optimal landing positions. Due to oculomotor motor noise and optical constraints, saccades frequently land slightly off-center—away from the “preferred landing position” (typically located halfway between the beginning and center of a word). When a saccade lands in a suboptimal visual location (such as the far trailing edge of a long word), visual feature extraction is degraded, slowing the completion of $L_1$. The model naturally accounts for when and why the oculomotor system cancels forward saccades and executes a rapid, corrective refixation within the same word.

Finally, E-Z Reader captures the computational triggers for regressive saccades (backward eye movements, which constitute roughly 10% to 15% of all saccades in skilled reading). If lexical processing at Stage 2 ($L_2$) encounters an insurmountable failure—due to a morphological anomaly, an unresolved lexical ambiguity, or an integration failure at the syntactic interface—the ongoing forward motor programming is aborted. The system initiates an emergency, central executive interrupt that directs the oculomotor engine to launch a backward saccade to the source of processing difficulty, operationalizing the cognitive mechanics of comprehension repair.

6. Theoretical Debates: E-Z Reader versus Parallel Models (SWIFT)

6.1 The SWIFT Model and Parallel Distributed Processing

The dominant serial-processing framework embodied by Rayner’s E-Z Reader did not go unchallenged. The most formidable theoretical alternative emerged from the German psycholinguistic school led by Ralf Engbert, Reinhold Kliegl, and their colleagues, who formulated the SWIFT model (Synchronized Walk-in-Field Model; Engbert et al., 2002, 2005). The fundamental philosophical divergence between E-Z Reader and SWIFT concerns the core nature of visual-spatial attention: is attention deployed strictly to one word at a time in serial order, or is attention distributed simultaneously across multiple words in a parallel, spatially graded fashion?

SWIFT is a parallel distributed processing model that conceptualizes the attentional spotlight as a continuous, asymmetrical gradient spanning an entire field of characters. Within this parallel gradient, multiple adjacent words—specifically the fixated word n, the upcoming word n+1, and occasionally word n+2 or the preceding word n-1—are actively undergoing lexical processing concurrently. In SWIFT, lexical activation is governed by an ongoing “random walk” stochastic process across all words covered by the attentional field, with activation rates determined by the relative visual acuity of each word’s retinal eccentricity and lexical difficulty.

A second radical departure in SWIFT involves its oculomotor decoupling mechanism. Unlike E-Z Reader, where the completion of a specific cognitive milestone ($L_1$) directly triggers the forward saccadic motor program, SWIFT incorporates an autonomous internal biological “clock.” An internal neural pacemaker continuously generates saccadic commands at semi-regular intervals, largely independent of immediate cognitive processing. Cognitive processing modulates the system by either inhibiting this autonomous saccadic timer (slowing down saccade execution when foveal processing is difficult) or influencing spatial target selection (determining which word currently exhibits the highest level of activation demand). Saccades and regressions, in SWIFT’s parallel framework, represent the emergent competitive dynamics of multiple lexical representations vying for ocular foveation.

6.2 Rayner’s Defense of Serial Processing

Keith Rayner vigorously defended the serial-attention architecture of E-Z Reader against parallel models, spearheading a multi-decade theoretical debate that elevated the mathematical and empirical standards of cognitive psychology. Rayner’s fundamental objection to parallel distributed models rested on the profound computational bottleneck inherent in human lexical access. While the early visual cortex undeniably processes sensory features in parallel across the visual field, Rayner maintained that the human language system possesses an architectural bottleneck that restricts full lexical identification and syntactic binding to a strictly serial queue.

To substantiate this serial claim, Rayner pointed to an array of empirical phenomena that are difficult to reconcile with true parallel processing. Most prominent was the sequential nature of word frequency and predictability effects. Rayner and his colleagues conducted meticulously controlled experiments demonstrating that the processing difficulty of word n+1 does not exert an immediate, concurrent influence on the fixation duration of word n under standard conditions. In a true parallel architecture, where words n and n+1 are identified simultaneously within a distributed gradient, high processing difficulty or low frequency on word n+1 ought to routinely “drain” processing resources away from word n, causing fixation durations on word n to lengthen systematically.

Rayner’s extensive datasets revealed that the cognitive difficulty of word n almost invariably “spills over” to affect the subsequent fixation on word n+1 (the ubiquitous spillover effect), but reverse effects (parafoveal difficulty affecting the current foveal fixation) are extraordinarily rare, fragile, and typically attributable to methodological artifacts. Furthermore, Rayner demonstrated that in gaze-contingent boundary experiments where words n and n+1 were presented under varying conditions of difficulty, processing latencies strictly mirrored serial queueing rather than simultaneous parallel capacity sharing. Serial processing, Rayner argued, provides the simplest, most parsimonious, and neurobiologically plausible mechanism for maintaining correct word order during sentence comprehension.

6.3 The Parafoveal-on-Foveal Effect Controversy

The battleground between serial models (E-Z Reader) and parallel models (SWIFT) ultimately centered on the contentious issue of parafoveal-on-foveal (PoF) effects. A parafoveal-on-foveal effect is operationally defined as any statistical phenomenon wherein properties of the upcoming, unfixed word (word n+1)—such as its lexical frequency, orthographic regularity, or semantic plausibility—directly influence the duration of the current fixation on word n.

For parallel processing architectures, the existence of PoF effects is an absolute theoretical necessity: if attention is distributed across words n and n+1 simultaneously, the linguistic properties of word n+1 must inevitably affect the global processing dynamics of that fixation. For serial models like E-Z Reader, robust lexical PoF effects represent an existential threat: if attention is strictly focused on word n until its lexical familiarity ($L_1$) is verified, the linguistic identity of word n+1 cannot penetrate central cognitive processing in time to modulate the current foveal fixation duration.

Rayner undertook exhaustive methodological and empirical audits of published PoF claims. In a series of highly influential critical papers, Rayner demonstrated that the vast majority of reported PoF effects were statistical artifacts resulting from oculomotor mislocation errors, overlapping saccadic distributions, or uncontrolled typographic variables. Specifically, Rayner proved that when a reader’s eye launches a saccade toward word n but accidentally lands near the trailing space—close to word n+1—or when the eye fixates very close to the rightmost boundary of word n, characters from word n+1 fall directly into central foveal vision. In these instances, what appears to be a “parafoveal” effect is actually an unacknowledged foveal processing effect.

When Rayner implemented rigorous methodological controls—excluding fixations falling within two characters of the word boundary, filtering out oculomotor landing artifacts, and employing high-power linear mixed-effects modeling—lexical and semantic PoF effects vanished in English alphabetic reading. While Rayner conceded that low-level visual properties of word n+1 (such as word length or total orthographic density) can exert subtle, early oculomotor influences via low-level retino-tectal pathways, he established that higher-order lexical and semantic PoF effects do not exist with sufficient reliability in alphabetic reading to overthrow the serial processing foundation of E-Z Reader.

7. Eye Movements in Syntactic Parsing and Cognitive Ambiguity

7.1 Garden-Path Sentences and the Two-Stage Model

Keith Rayner’s intellectual contributions extended far beyond isolated word identification; he fundamentally reshaped the landscape of higher-order psycholinguistics, sentence processing, and structural parsing. In an enduring and profoundly influential scientific partnership with linguist Lyn Frazier at the University of Massachusetts Amherst, Rayner operationalized the study of human syntactic parsing using high-resolution eye tracking, formulating what is universally known as the Garden-Path Theory of sentence processing (Frazier & Rayner, 1982).

The Garden-Path Theory posited that the human sentence parsing mechanism is structured as a two-stage, modular information-processing system. The human language parser, operating under severe short-term memory constraints, must construct structural syntactic trees in real time as each successive word is encountered. Frazier and Rayner hypothesized that the initial syntactic stage is governed by universal, computationally economical heuristics that rely strictly on phrase-structure rules, completely blind to non-syntactic information such as contextual plausibility, semantic meaning, or real-world pragmatic knowledge.

Frazier and Rayner identified two primary universal parsing heuristics:

• Minimal Attachment: The structural heuristic that directs the parser to attach each newly incoming word into the phrase-marker structure being constructed using the fewest possible syntactic nodes consistent with the grammar.
• Late Closure: The structural heuristic dictating that, if grammatically permissible, the parser preferentially attaches newly encountered linguistic elements into the phrase or clause currently under active processing.

To test these theoretical principles, Rayner and Frazier recorded participants’ eye movements as they read temporary syntactic ambiguities, famously exemplified by sentences such as: “While Mary was mending the sock fell off her lap.” Under the Late Closure principle, the parser automatically and erroneously attaches “the sock” as the direct object of the preceding transitive verb “mending.” However, when the eye encounters the disambiguating matrix verb “fell,” the structural analysis collapses: “the sock” must actually serve as the subject of the main clause, not the object of the subordinate clause.

Rayner’s eye-movement records provided sensational, indisputable empirical evidence for this processing breakdown. Readers exhibited normal, rapid fixation durations through the ambiguous phrase, confirming that the initial misanalysis proceeded effortlessly. But upon fixating the disambiguating word “fell”—termed the syntactic disambiguation zone—readers experienced massive oculomotor disruption: first-pass fixation durations inflated dramatically, reading stalled, and the eyes launched immediate, high-velocity regressive saccades back to the ambiguous region (“the sock”) to re-parse the syntactic structure. This empirical methodology permanently established eye tracking as the gold standard for dissecting the temporal architecture of syntactic parsing.

7.2 Constraint-Based vs. Modular Parsing Perspectives

The publication of Rayner and Frazier’s two-stage modular model ignited one of the fiercest and most productive debates in the history of cognitive science, pitting their modular architecture against emergent constraint-based interactive models championed by researchers such as Michael Tanenhaus, Mark Seidenberg, Maryellen MacDonald, and Ken McRae in the 1990s. Constraint-based theorists argued that the human mind does not maintain a modular syntactic parser that operates in isolation from context. Instead, they claimed that sentence comprehension is governed by a parallel, interactive constraint-satisfaction network in which all available informational sources—syntactic rules, lexical frequency of argument structures, thematic fit, visual context, and semantic plausibility—are evaluated simultaneously and interactively from the very first millisecond of word encounter.

Rayner entered this theoretical arena with characteristic empirical rigor, defending the temporal priority of syntactic heuristics while establishing the precise methodological criteria required to adjudicate between modular and interactive accounts. Rayner emphasized the vital necessity of differentiating between first-pass reading measures (metrics reflecting the initial processing pass through a sentence, such as first fixation duration, single fixation duration, and gaze duration) and second-pass or total reading measures (metrics reflecting re-reading, regression paths, and downstream comprehension repair, such as regression-path duration and total viewing time).

In a series of landmark empirical studies, Rayner, Frazier, and their colleagues demonstrated that when sentences featured strong semantic or pragmatic biases that favored the structurally non-preferred parse (e.g., in sentences where thematic plausibility should theoretically prevent the garden path), readers’ eyes still exhibited initial oculomotor disruption at the disambiguating region in first-pass measures. Even when real-world knowledge screamingly indicated that an inanimate noun (e.g., “the evidence”) could not perform an active agent role (as in “The evidence examined by the lawyer…”), the parser’s immediate first-pass reflex still attempted minimal attachment, generating subtle but reliable early oculomotor hiccups before semantic plausibility intervened moments later to guide syntactic re-analysis. Rayner argued that while constraint-based mechanisms undeniably dominate downstream comprehension and total reading times, the earliest temporal slice of syntactic processing remains structurally encapsulated, driven by rapid, modular heuristics.

7.3 Semantic and Pragmatic Anomaly Resolution

Beyond abstract syntactic phrase structures, Rayner explored how the cognitive architecture detects, processes, and resolves semantic and pragmatic anomalies during natural reading. By employing the eye-tracking apparatus to monitor the reading of semantic violations—ranging from stark categorical violations (e.g., “The boy kicked the mountain”) to subtle violations of selectional restrictions (e.g., “The vegetarian ordered a steak”)—Rayner mapped the chronological time course of semantic integration.

His empirical findings demonstrated that the brain’s sensitivity to semantic implausibility is breathtakingly fast. When an anomalous target word is fixated, eye movements register the disruption almost immediately: gaze durations on the anomalous word are significantly inflated within 200 to 250 milliseconds of visual contact. If the anomaly is extreme or bizarre, readers rarely finish reading the sentence; instead, the oculomotor engine abruptly halts its forward progression and executes immediate regressions back to preceding context words, seeking to determine whether an error in perceptual identification had occurred.

Rayner extended this paradigm into the study of non-literal language, metaphors, idioms, and discourse-level incongruities. In investigations of idiomatic expressions (e.g., “kicking the bucket”), Rayner and his collaborators showed that familiar idioms undergo expedited lexical processing, exhibiting shorter fixation durations and higher skipping rates than literal phrases matched for length and word frequency. However, if an idiom is modified or presented in an incongruous discourse context, processing costs escalate rapidly, manifest as extensive regressions and inflated second-pass reading times. These studies established that working memory resources are dynamically allocated during reading: early oculomotor measures capture automatic lexical-semantic retrieval, whereas late oculomotor measures capture the heavy, deliberative cognitive labor of discourse-level pragmatic integration and conceptual repair.

8. Visual Cognition Beyond Reading: Scene Perception and Visual Search

8.1 Eye Movements in Complex Natural Scenes

While Keith Rayner’s international fame was anchored predominantly in the psycholinguistics of reading, his theoretical vision encompassed the entirety of visual cognition. Throughout his career, and with increasing intensity from the 1990s onward, Rayner applied his formidable methodological toolkit to the study of scene perception, visual exploration, and object identification in naturalistic environments. He recognized that while reading text and inspecting complex scenes both rely on alternating fixations and saccades, their computational goals and visual dynamics differ in profound ways.

Through systematic comparative analyses, Rayner identified fundamental psychophysical differences between reading and scene perception:

• Fixation Durations: Average fixation durations in scene perception are markedly longer ($\approx$ 260 to 330 ms) than in skilled reading ($\approx$ 200 to 250 ms), reflecting the higher visual complexity and multidimensional feature analysis required by natural scenes.
• Saccadic Amplitudes: Saccades during scene exploration are substantially longer and more variable, frequently spanning 4 to 10 degrees of visual angle, compared to the tightly constrained 2-degree (7 to 9 letter spaces) saccades observed in reading.
• Spatial Constraints: Scene perception lacks the rigid, linear left-to-right constraints of alphabetic typography, requiring the visual system to navigate two-dimensional space through top-down cognitive schemas and bottom-up visual saliency.

Rayner investigated the perennial theoretical conflict between bottom-up saliency models (such as the computational saliency maps of Laurent Itti and Christof Koch, which argue that gaze is driven strictly by low-level image properties such as local luminance, color contrast, and edge orientation) and top-down cognitive guidance models. In classic experiments manipulating scene semantics—such as placing an anomalous object into a coherent scene context (e.g., placing a farm tractor inside a modern kitchen display)—Rayner and his colleagues revealed the astonishing speed of top-down conceptual processing.

Rayner demonstrated that the human brain extracts the “gist” of a complex natural scene within an initial, single fixation lasting fewer than 100 milliseconds. Once this global scene gist is established (e.g., recognizing that the setting is a kitchen), top-down cognitive expectations immediately guide subsequent oculomotor programming. Fixations do not mindlessly follow low-level luminance peaks; instead, the eyes rapidly target semantically anomalous or task-relevant objects far faster than can be predicted by pure saliency models. Gaze durations on semantically incongruous objects were heavily inflated, proving that visual cognitive processing in scenes, like reading, is dominated by the continuous interplay of central semantic representations and peripheral sensory cues.

8.2 Visual Search and Object Recognition

Keith Rayner made foundational contributions to the experimental literature on visual search, examining how the human visual system locates specific targets hidden within cluttered, high-density visual arrays (e.g., searching for a specific item in a grocery store aisle or identifying a concealed object in a complex visual array). Rayner brought his celebrated gaze-contingent moving-window methodology directly into the visual search paradigm, engineering a dynamic “search window” that moved in real time with the participant’s gaze across visual scenes.

By systematically shrinking or expanding the diameter of this gaze-contingent search window, Rayner measured the functional “visual field” during search tasks. He discovered that the effective field of view during visual search is highly dependent on target distinctiveness and distractor crowding. When searching for targets characterized by simple, low-level features (e.g., searching for a bright red circle among blue squares), the functional visual span was broad, facilitating rapid, pre-attentive guidance. However, when searching for complex, conjunctive targets or objects requiring structural identification, the search span contracted sharply, forcing participants to execute dense clusters of short-amplitude saccades.

Rayner used these findings to construct detailed comparative models contrasting lexical search in reading with visual search in natural environments. He proved that while both domains utilize saccadic suppression and parafoveal preview extraction, reading benefits from an invariant, predictable spatial structure (spaces separating words in a linear row), whereas visual search requires continuous, active spatial re-mapping. In visual search, parafoveal processing is deployed to evaluate peripheral candidate objects for target-like features, allowing the central executive to program ballistic saccades directly toward potential matches while inhibiting visual return to previously visited locations (inhibition of return).

8.3 Face Perception and Visual Expertise

Rayner’s broad intellectual curiosity led him to examine the specialized cognitive architectures underlying face perception and domain-specific visual expertise. Face processing represents one of the most evolutionarily sophisticated feats of human visual cognition, relying heavily on holistic, configural processing rather than isolated feature decomposition. Rayner and his collaborators recorded eye-tracking scanpaths during facial identification, emotion discrimination, and unfamiliar face matching tasks.

His research demonstrated that human scanpaths during face processing are characterized by highly stereotyped, triangular geometric configurations, with fixations overwhelmingly concentrated on the eyes, the bridge of the nose, and the mouth. Rayner showed that the eyes receive the highest density of fixations and longest dwell times, functioning as the primary informational hub for both identity verification and emotional state attribution. Furthermore, by employing the gaze-contingent moving-window paradigm on human faces, Rayner demonstrated that masking the foveal region completely abolishes configural processing, reducing facial perception to a clumsy, error-prone analysis of isolated features and proving that holistic face recognition requires the simultaneous integration of foveal and parafoveal visual cues.

In his explorations of visual expertise, Rayner investigated the oculomotor profiles of highly trained professionals, including board-certified radiologists inspecting complex chest X-rays, master chess players evaluating tactical boards, and professional artists viewing compositional canvases. Across all domains of expertise, Rayner identified common oculomotor signatures of high-level cognitive mastery:

• Expanded Functional Field of View: Experts exhibit a significantly larger perceptual span within their domain of expertise, allowing them to rapidly extract holistic structural configurations without needing to execute dense, exhaustive local fixations.
• Reduced Time to First Fixation on Critical Targets: Radiologists detecting lung nodules or master chess players spotting tactical threats launch forward saccades directly to critical anomaly zones within their very first two or three fixations, guided by rapid parafoveal pattern recognition.
• Suppression of Irrelevant Saliency: Novices are repeatedly trapped by salient, low-level visual distractions (such as high-contrast boundaries), whereas experts completely ignore bottom-up distractions, their gaze driven entirely by sophisticated, top-down internal schemas.

9. Special Populations: Development, Aging, and Atypical Reading

9.1 Development of Reading Expertise in Children

Throughout his scientific career, Keith Rayner maintained an active research commitment to understanding how reading expertise is systematically assembled across childhood development. He recognized that tracing the developmental trajectory of eye movements provides an objective, non-invasive, millisecond-accurate behavioral index of how the reading brain matures from awkward, effortful decoding into rapid, automated fluency.

In extensive longitudinal and cross-sectional studies comparing beginning primary school readers (grades 1–3) against advanced elementary children and skilled adults, Rayner and his colleagues documented the profound transformation of oculomotor metrics across development:

• Fixation Durations: Young novice readers exhibit extraordinarily long fixation durations, averaging between 300 and 400 milliseconds (and frequently exceeding 500 milliseconds on unfamiliar words), compared to the brisk 200 to 250 milliseconds typical of skilled adults.
• Saccade Lengths: Beginning readers execute very short saccades, typically advancing only 3 to 5 letter spaces per jump, reflecting their restricted ability to preview upcoming text.
• Regression Rates: Novices launch regressive saccades on 25% to 35% of all fixations, compared to the streamlined 10% to 15% regression baseline of proficient adult readers.

Crucially, Rayner utilized the gaze-contingent moving-window paradigm with children to chart the developmental expansion of the perceptual span. He discovered that for children in the early primary grades, the perceptual span is remarkably constricted, extending only 4 to 6 character spaces to the right of fixation and virtually zero characters to the left. At this developmental stage, the immense cognitive effort required for phonological decoding consumes all available central attentional bandwidth, leaving zero capacity for parafoveal preview. As children achieve reading automaticity—transitioning from slow, letter-by-letter phonological assembly to direct orthographic sight-word recognition—the cognitive load attenuates dramatically. This liberation of cognitive resources allows visual attention to cascade forward into the parafovea, causing the perceptual span to expand systematically until it reaches the adult dimensions of 14 to 15 character spaces by late primary or middle school.

9.2 Reading in the Aging Brain

In the later decades of his career, Keith Rayner pioneered the systematic experimental study of reading across the adult lifespan, establishing the cognitive and oculomotor hallmarks of the aging reading brain. As healthy adults age, the visual and neurological systems experience unavoidable structural declines: senile miosis and nuclear cataracts reduce retinal illuminance, contrast sensitivity deteriorates across high spatial frequencies, photoreceptor and ganglion cell densities decline, and central cognitive processing speeds experience generalized slowing.

Rayner investigated how older readers (typically aged 65 to 85) adapt to these biological constraints to maintain reading comprehension. His eye-tracking investigations revealed a fascinating, counterintuitive compensatory strategy. Rather than slowing down uniformly or adopting an overly conservative reading strategy, older skilled readers frequently adopt what Rayner termed a “risky” reading strategy. Rayner showed that older readers actually display substantially higher word-skipping rates for short- and medium-length words than young college-age adults, fixating fewer words overall on the initial pass through a sentence.

Because their parafoveal visual acuity is degraded by optical and neural aging (resulting in a demonstrably narrower physical perceptual span), older readers cannot cleanly resolve letter identities in the parafovea. To compensate, their cognitive systems lean heavily on high-level linguistic contextual constraints, linguistic redundancy, and top-down predictive inference to “guess” upcoming words and skip them outright. However, because this predictive strategy is inherently risky, older readers encounter frequent comprehension failures when an upcoming word violates their expectations. Consequently, Rayner documented that older readers execute far more regressive saccades and exhibit dramatically inflated second-pass re-reading times. Their reading behavior is thus characterized by a trade-off: fast, speculative forward jumping punctuated by frequent, extensive re-reading passes to verify meaning, demonstrating the extraordinary plasticity and strategic resilience of the aging cognitive architecture.

9.3 Dyslexia and Oculomotor Function

One of Keith Rayner’s most profound and socially impactful scientific contributions was his decisive role in deconstructing a widespread, deeply entrenched clinical myth: the erroneous belief that developmental dyslexia is fundamentally caused by an oculomotor or visual-motor defect. For decades, a prominent school of thought in ophthalmology, optometry, and fringe educational therapy had asserted that individuals with dyslexia suffer from unstable eye tracking, erratic saccadic coordination, or binocular tracking failure, which allegedly causes letters to “jump,” “reverse,” or blur on the page. Based on this flawed premise, thousands of struggling children were subjected to expensive, unvalidated “eye-muscle training” regimens and colored lens therapies designed to “cure” their oculomotor tracking problems.

Rayner published devastating empirical refutations of this visual-motor hypothesis. Through exquisitely controlled laboratory experiments, Rayner recorded the eye movements of children and adults with developmental dyslexia under two contrasting conditions: reading continuous written text versus performing non-linguistic visual search tasks matched for spatial complexity (such as tracking a moving dot across a screen or searching for geometric symbols in a horizontal array).

The results were conclusive and definitive:

• When reading written text, individuals with dyslexia indeed exhibited highly anomalous, chaotic scanpaths—characterized by prolonged fixation durations (often exceeding 400–600 ms), tiny, fragmented saccades, and massive clusters of backward regressions.
• However, when the exact same individuals performed non-linguistic visual search or dot-tracking tasks, their oculomotor control was completely indistinguishable from typically developing readers. Their saccadic precision, smooth pursuit, fixational stability, and landing accuracy were entirely normal.

Furthermore, Rayner proved that when skilled, non-dyslexic adult readers were presented with highly degraded, syntactically bewildering, or archaic foreign texts, their scanpaths instantly fractured into the exact same chaotic, regressive, short-saccade patterns exhibited by dyslexic readers. Rayner established that erratic eye movements in dyslexia are the downstream cognitive consequence, not the physical cause, of severe reading failure. Dyslexia is primarily a central neurocognitive deficit in phonological decoding, orthographic mapping, and language processing. When the brain cannot decode a word, the oculomotor engine stalls, hesitates, and loops backward. Rayner’s authoritative scientific stance directly altered clinical guidelines worldwide, steering educational resources away from bogus vision therapies and toward evidence-based, phonics-grounded literacy interventions.

10. Cross-Linguistic Studies and Non-Alphabetic Scripts

10.1 Reading in Chinese and Logographic Systems

A hallmark of Keith Rayner’s scientific leadership was his relentless drive to formulate a truly universal cognitive theory of eye-movement control that transcended the peculiarities of Western alphabetic scripts. To test the cross-linguistic generalizability of his models, Rayner forged deep, sustained collaborations with cognitive psychologists across East Asia, conducting groundbreaking investigations into the reading of logographic Chinese.

Chinese typography represents a profound, fascinating contrast to alphabetic English. Chinese is written as continuous strings of square-shaped box characters without any intervening spaces to delineate word boundaries. Furthermore, Chinese words are compact: while an English word averages 5 to 6 letters in length, Chinese words typically comprise only one, two, or occasionally three characters, with each individual character containing a complex interior configuration of visual strokes packing dense semantic and morphemic information.

Rayner’s Chinese eye-tracking research resolved fundamental theoretical puzzles regarding word segmentation and saccadic targeting. In English, word boundaries (white spaces) dictate saccade programming: the visual system targets the center of the upcoming blank-delimited word. How does the oculomotor system target saccades in Chinese when spaces do not exist? Rayner and his colleagues demonstrated that skilled Chinese readers rapidly perform covert, pre-attentive morphological and lexical segmentation in the parafovea. Rather than launching saccades character-by-character, the oculomotor system targets multi-character lexical units as integrated words. Saccade lengths in Chinese average approximately 2 to 3 characters per jump, which precisely corresponds to the average length of Chinese compound words!

Moreover, Rayner evaluated the perceptual span in Chinese using the gaze-contingent moving-window paradigm. Because of the extraordinary visual and semantic density of Chinese characters, the physical span is dramatically more compact than in English: extending only 1 character to the left of fixation and approximately 2 to 3 characters to the right. Beyond 3 characters to the right, information extraction drops to zero. Finally, Rayner and his team successfully extended the computational architecture of the E-Z Reader model to Chinese reading, proving that despite profound orthographic, structural, and typographic differences, the fundamental core principles of serial attention, two-stage lexical access ($L_1 to L_2$), and decoupled motor programming represent universal, species-wide mechanisms of human visual reading.

10.2 Bi-Directional Scripts: Hebrew and Arabic

Keith Rayner’s investigations into Semitic orthographies—principally Hebrew and Arabic—provided vital empirical laboratories for dissecting the interactions between reading directionality, morphological architecture, and eye-movement guidance. As established in Section 3.1, Rayner’s early work in Hebrew proved that the asymmetry of the perceptual span is governed by reading direction, flipping to extend predominantly to the left of fixation.

However, Semitic orthographies presented Rayner with a much richer cognitive landscape than mere directional inversion. Both Hebrew and Arabic feature non-concatenative root-and-template morphology, wherein words are formed by interlocking a typically three-consonant semantic root into an abstract morpho-phonological vowel-and-affix template. Furthermore, these scripts are written in two distinct orthographic formats: “voweled” (shallow) orthography containing diacritical vowel markers (utilized predominantly in children’s books, liturgical texts, and poetry) and “unvoweled” (deep) orthography, where vowel diacritics are omitted entirely and fluent adults must infer vocalic phonology through high-level contextual constraints.

Rayner and his Israeli and international collaborators recorded eye movements across both voweled and unvoweled texts. They discovered that contrary to pedagogical assumptions that diacritics make reading easier, skilled adult readers actually fixate longer and read slower when vowel diacritics are added to familiar text. The unnecessary visual clutter of diacritics imposes an extraneous visual load that retards saccadic programming. In unvoweled scripts, Rayner demonstrated that readers utilize early parafoveal previews of consonant roots to pre-activate lexical candidate families, allowing for rapid contextual disambiguation.

Furthermore, Rayner studied the remarkable cognitive flexibility of cross-script bilingual readers who navigate both English (left-to-right, alphabetic, non-concatenative) and Hebrew or Arabic (right-to-left, root-based) within the same linguistic repertoire. Rayner demonstrated that these bilinguals do not suffer from cognitive interference; their oculomotor control engines recalibrate instantly upon detecting script identity, switching directionality, adjusting saccadic amplitudes, and altering perceptual span geometries within a single fixation.

10.3 Agglutinative and Morphologically Rich Languages

To further test the universal boundary conditions of eye-movement control, Keith Rayner directed investigations into morphologically rich, agglutinative languages, with a specialized focus on Finnish, Turkish, and German long compound constructions. Agglutinative orthographies present a unique challenge to computational models of reading because their words can become colossal typographic units. In Finnish, for example, complex inflectional, derivational, and clitic morphemes are sequentially appended to a root stem, generating single words that can easily exceed 20 to 30 characters in length (e.g., lentokonesuihkuturbiinimoottoriapumekaanikkoaliupseerioppilas).

In standard English reading, words exceeding 10 letters rarely fit comfortably within the fovea, typically triggering a single refixation. How does the human eye navigate words in Finnish or German compound words (e.g., Donaudampfschiffahrtsgesellschaftskapitän)? Rayner and his European collaborators discovered that in agglutinative scripts, the visual system does not treat long words as monolithic lexical entities. Instead, eye guidance is intimately synchronized with internal morphological constituent boundaries.

When Finnish or German readers encounter long words, their oculomotor systems systematically target the first morphological sub-constituent (the initial stem). Fixation durations on this initial constituent are dictated by its specific sub-lexical frequency. The eye then executes an intra-word saccade directly targeting the junction of the second morphological constituent. Rayner demonstrated that morphological segmentation occurs rapidly during the initial fixation, with constituent boundaries acting as internal “pseudo-spaces” that guide subsequent landing sites. These findings led to critical expansions of the E-Z Reader model, incorporating sub-lexical processing loops that demonstrated that the cognitive architecture parses complex words via sequential morphological stages, preserving the universality of serial eye-movement control across all morphological typologies.

11. Institutional Leadership, Editorial Impact, and Mentorship

11.1 Building World-Class Laboratories (UMass and UCSD)

Beyond his extraordinary individual intellectual output, Keith Rayner was an institutional visionary who built two of the most vibrant, productive, and internationally renowned experimental laboratories in the history of cognitive science: first at the University of Massachusetts Amherst (1978–2008) and subsequently at the University of California, San Diego (2008–2015).

When Rayner arrived at UMass Amherst in 1978, he established the University of Massachusetts Eye-Tracking Laboratory. Over the course of three decades, Rayner transformed UMass into the undisputed global epicenter of research on eye movements, psycholinguistics, and visual cognition. Together with brilliant colleagues such as Alexander Pollatsek, Lyn Frazier, and Charles Clifton Jr., Rayner created an extraordinarily collaborative, intellectually rigorous ecosystem. The “UMass Eye Lab” attracted visiting scholars, postdoctoral fellows, and graduate students from every continent. Rayner invested tremendous energy in developing open, robust technical methodologies, publishing calibration routines, open-sourcing parsing algorithms, and freely distributing customized experimental software that became the international standard for gaze-contingent research.

In recognition of his towering scientific contributions, the University of Massachusetts Amherst appointed Rayner as a Distinguished University Professor in 1991—the highest academic honor bestowed by the institution. In 2008, after thirty glorious years at UMass, Rayner accepted a prestigious appointment at UC San Diego, where he was named the Atkinson Family Endowed Chair in Psychology and Cognitive Science. At UCSD, Rayner constructed a state-of-the-art research center, expanding his investigations into neuroimaging co-registration, cross-linguistic typologies, and multimodal cognitive science, cementing his laboratory as a dynamic nexus of discovery until the final days of his life.

11.2 Editorial Stewardship of Premier Journals

Keith Rayner’s profound impact on psychological science was magnified through his extraordinary stewardship as an editor of the field’s most prestigious, tier-one peer-reviewed academic journals. Rayner possessed an unmatched capacity for intellectual synthesis, a legendary work ethic, and an unyielding commitment to empirical integrity, fairness, and methodological reproducibility.

Rayner served as the Editor-in-Chief of two flagship journals of the American Psychological Association (APA):

• Journal of Experimental Psychology: Learning, Memory, and Cognition (JEP:LMC) from 1990 to 1995.
• Psychological Review—universally regarded as the premier theoretical journal in scientific psychology—from 2004 to 2010.

Under his editorial leadership, both journals maintained rigorous empirical standards. As Editor of Psychological Review, Rayner championed formal computational and mathematical modeling, insisting that theoretical architectures must generate quantitative, falsifiable predictions rather than vague qualitative metaphors. He had zero tolerance for methodological shortcuts, small-sample underpowered studies, or selective data reporting. Yet despite his exacting standards, Rayner was revered by authors for his lightning-fast turnaround times, his deeply constructive, insightful editorial letters, and his exceptional ability to nurture high-risk, groundbreaking theoretical ideas from junior researchers. Through his editorial pens, Rayner actively curated and elevated the intellectual trajectory of cognitive psychology for more than two decades.

11.3 Pedagogy and Generational Mentorship

Perhaps Keith Rayner’s most enduring and fertile legacy lies in his extraordinary human mentorship. Rayner was not merely a brilliant researcher; he was a generous, dedicated, and deeply beloved mentor who trained several generations of the world’s leading cognitive psychologists, psycholinguists, and vision scientists. The academic genealogy descending directly from Keith Rayner represents an astonishing “who’s who” of modern cognitive science.

Among his numerous illustrious doctoral students and postdoctoral advisees are world-class scholars such as Erik D. Reichle, Barbara Jochimsen Juhasz, Julie A. Kirkby, Simon P. Liversedge, Denis Drieghe, Robin K. Morris, Tessa Warren, Monica Castelhano, Jane Ashby, and Elizabeth Schotter, among scores of others. Rayner fostered a laboratory culture that was intellectually demanding yet deeply supportive, devoid of toxicity or territorial arrogance. He treated graduate students as genuine intellectual colleagues from their first week in the laboratory, instilling in them his personal ethos: respect the raw data above personal pride, master the technical hardware down to the millisecond, write with absolute conceptual clarity, and maintain relentless intellectual honesty.

Rayner’s mentorship extended to the broader international scientific community. He was a central organizing pillar of the European Conference on Eye Movements (ECEM), held biannually across Europe, where he served as a warm, accessible mentor to young researchers from around the globe. He organized foundational workshops, international symposia, and edited multiple definitive volumes that served as the primary textbooks for generations of students entering the field. His academic family continues to populate leading university faculties across North America, Europe, Asia, and Australasia, perpetuating his high standards of rigorous empirical psycholinguistics.

12. Keith Rayner’s Enduring Legacy and Future Directions

12.1 Methodological Synthesis: Co-Registration Techniques

In the final phase of his illustrious career, Keith Rayner actively championed the methodological frontier that represents the future of reading neuroscience: the concurrent co-registration of eye movements with neuroimaging and electrophysiological recordings. For decades, the methodologies of eye tracking and electroencephalography (EEG) were artificially separated due to severe technical incompatibilities. In traditional EEG reading experiments, participants were forced to sit completely motionless and fixate a single central cross while sentences were flashed one word at a time via Rapid Serial Visual Presentation (RSVP)—an unnatural paradigm that completely obliterates parafoveal preview, eliminates normal oculomotor targeting, and suppresses word skipping and regressions.

Conversely, classical eye tracking provided exquisite micro-chronometric behavioral data, but could only infer the underlying neural circuitry. Rayner spearheaded efforts to synchronize high-speed eye tracking with concurrent high-density EEG, giving birth to the paradigm of Eye-Fixation-Related Potentials (EFRPs). By developing sophisticated mathematical algorithms to remove ocular muscle artifacts (electro-oculograms) from the continuous EEG signal, Rayner and his colleagues succeeded in recording real-time neural voltage fluctuations (such as the P1, N170, and N400 event-related components) time-locked precisely to the onset of individual, natural saccadic fixations during continuous reading.

This methodological synthesis provided breathtaking confirmation of Rayner’s theoretical models. EFRP recordings validated that parafoveal preview codes directly modulate early neural responses in the visual cortex within 100 milliseconds of landing on a target word. Furthermore, the combination of eye tracking with functional magnetic resonance imaging (fMRI) has mapped the structural neural networks—bridging the occipito-temporal visual word form area, the superior temporal gyrus, and inferior frontal regions—that execute the computational stages formalized in Rayner’s E-Z Reader model, elevating his chronometric findings into the mainstream of modern cognitive neuroscience.

12.2 Artificial Intelligence and Modern Computational Psycholinguistics

The dawn of twenty-first-century artificial intelligence, natural language processing (NLP), and deep learning has triggered an extraordinary resurgence of interest in Keith Rayner’s work. As modern computer scientists develop massive Large Language Models (LLMs) based on transformer architectures (such as GPT-4, LLaMA, and BERT), a critical scientific challenge has emerged: evaluating whether these artificial neural networks process human language in a manner that mirrors human cognitive plausibility.

To resolve this challenge, computational linguists rely heavily on vast, open-access human eye-movement reading corpora—such as the Dundee Corpus, the Provo Corpus, and the multi-lingual datasets directly inspired by and modeled upon Rayner’s empirical protocols. In contemporary computational psycholinguistics, a model’s cognitive plausibility is routinely evaluated by its ability to accurately predict human fixation durations, skipping probabilities, and regression trajectories on a word-by-word basis. The metric known as surprisal (the negative log-probability of a word given its preceding context, derived from information-theoretic models and deep transformers) correlates remarkably with the empirical fixation durations charted by Rayner decades earlier.

Furthermore, Rayner’s E-Z Reader model has itself undergone robust modernization. Computer scientists and cognitive modelers have integrated E-Z Reader with deep learning front-ends, replacing simplified word-frequency tables with dynamic neural language model embeddings. These modern extensions enable E-Z Reader to simulate human eye-movement patterns across massive, diverse natural text corpora, confirming that Rayner’s core architectural insights—serial attentional queuing, two-stage lexical access triggers, and decoupled oculomotor motor programming—remain the gold standard for cognitively plausible computational implementations of reading.

12.3 Final Assessment of Scientific Stature

Keith Rayner passed away on January 21, 2015, in San Diego, California, at the age of 71, leaving behind an intellectual legacy of titanic proportions. Over his extraordinary 41-year academic career, Rayner published nearly 500 peer-reviewed journal articles, dozens of book chapters, and authored or edited 10 definitive scholarly volumes. His work has accrued tens of thousands of scientific citations, making him one of the most highly cited cognitive psychologists in the history of science.

Among his definitive summary works, two monumental reviews stand as towering monuments of scientific synthesis:

• Rayner, K. (1978). “Eye movements in reading and information processing.” Psychological Bulletin, 85(3), 618–660.
• Rayner, K. (1998). “Eye movements in reading and information processing: 20 years of research.” Psychological Bulletin, 124(3), 372–422.

His 1998 review alone has accrued well over 5,000 academic citations, serving as the foundational reference manual for virtually every eye-tracking laboratory on Earth. Rayner formulated fundamental empirical laws of human reading that will endure for centuries: the exact physical boundaries of the asymmetric perceptual span, the quantitative magnitude of parafoveal preview benefits, the abstract non-photographic nature of trans-saccadic memory, and the intricate timing of lexical access and oculomotor programming. He transformed reading research from an imprecise, subjective field into a rigorous, micro-chronometric natural science.

Conclusion: The Architecture of the Reading Mind

When reflecting upon the history of cognitive psychology, Keith Rayner occupies a rare pantheon alongside figures such as George Miller, Donald Broadbent, and Ulric Neisser. Before Keith Rayner, the act of reading was largely a phenomenological mystery—a silent, subjective experience that defied direct empirical observation. By wedding technological virtuosity with deep theoretical insight, Rayner turned the human eye into an open, readable instrument of the human mind.

His creation of the gaze-contingent moving-window and boundary paradigms demonstrated how computer technology could be harnessed to dismantle complex cognitive behaviors into their atomic temporal components. Through these paradigms, Rayner taught the scientific world that human reading is an astonishing evolutionary improvisation: an intricate, high-speed biological ballet wherein visual sensations, abstract orthographic symbols, phonological representations, and semantic concepts are dynamically assembled within a temporal window of two hundred milliseconds, hundreds of times per minute.

Keith Rayner’s life was characterized by exemplary scientific integrity, profound personal generosity, and a joyful, unyielding devotion to discovery. His theoretical models, empirical baselines, and philosophical rigor continue to guide experimental psycholinguists, vision scientists, educational policy makers, and artificial intelligence researchers across the globe. As science continues to map the boundless intricacies of the human brain, Keith Rayner’s foundational discoveries will forever illuminate our understanding of the reading mind.

References

• Engbert, R., Longtin, A., & Kliegl, R. (2002). A dynamical model of saccade generation in reading based on spatially distributed lexical processing. Vision Research, 42(5), 621–636. https://doi.org/10.1016/S0042-6989(01)00301-7
• 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
• 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
• 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. (1975). 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. (1978). Eye movements in reading and information processing. Psychological Bulletin, 85(3), 618–660. https://doi.org/10.1037/0033-2909.85.3.618
• 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., & Pollatsek, A. (1989). The Psychology of Reading. Prentice-Hall.
• 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
• Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The E-Z Reader model of eye-movement control in reading: Comparisons to other models. Behavioral and Brain Sciences, 26(4), 445–476. https://doi.org/10.1017/S0140525X03000104