Understanding the eye-brain interface during reading: Connecting word skipping and comprehension using parafoveal N400 fixation-related potentials

Understanding the eye-brain interface during reading: Connecting word skipping and comprehension using parafoveal N400 fixation-related potentials | Authorea try { document.documentElement.classList.add('js'); } catch (e) { } var _gaq = _gaq || []; _gaq.push(['_setAccount', 'G-8VDV14Y67G']); _gaq.push(['_trackPageview']); (function() { var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); })(); Skip to main content Preprints Collections Wiley Open Research IET Open Research Ecological Society of Japan All Collections About About Authorea FAQs Contact Us Quick Search anywhere Search for preprint articles, keywords, etc. Search Search ADVANCED SEARCH SCROLL Psychophysiology This is a preprint and has not been peer reviewed. Data may be preliminary. 1 July 2025 V1 Latest version Share on Understanding the eye-brain interface during reading: Connecting word skipping and comprehension using parafoveal N400 fixation-related potentials Authors : Sara Milligan 0000-0001-8600-2768 [email protected] and Elizabeth Schotter 0000-0001-9345-1051 Authors Info & Affiliations https://doi.org/10.22541/au.175134605.50937634/v1 351 views 188 downloads Contents Abstract Information & Authors Metrics & Citations View Options References Figures Tables Media Share Abstract Readers frequently skip words, often when they are expected (Staub, 2015), but sometimes even when they are contextually incompatible (e.g., the high frequency article the ; Angele & Rayner, 2013). Therefore, some skipping behaviors may be automatic (e.g., based on familiarity or length) rather than based on thorough word identification, raising the question of how related word skipping behavior is to comprehension of the meaning of the text. To test this question, we co-registered eye tracking and EEG recordings and measured fixation-related brain potentials (FRPs). We time-locked FRPs to the pretarget word before a three-letter target word embedded in a high constraint sentence, and compared the neural activity between different preview conditions when the target word was subsequently either skipped or fixated (i.e., behavior-contingent analysis ). We used a gaze-contingent display change to manipulate the parafoveal word to be (1) expected, (2) an anomalous orthographic neighbor of the expected word, or (3) an anomalous instance of the , which changed to the expected word during the skip or fixate saccade. We found a parafoveal N400 effect to the anomalous neighbor compared to the expected preview that only occurred when the target was skipped, but not when it was fixated. There was also a parafoveal N400 effect for the anomalous the compared to the expected word that was not skipping-contingent. This pattern demonstrates that word skipping is associated with the depth of parafoveal linguistic processing, but these decisions are initiated prior to complete identification and contextual integration processes, which continue after the eyes have moved on. Understanding the eye-brain interface during reading: Connecting word skipping and comprehension using parafoveal N400 fixation-related potentials Sara Milligan and Elizabeth R. Schotter University of South Florida Author Note The authors declare no conflicts of interest. Work on this study was supported by NSF Grants BCS-2120507 and BCS-2341665, and a USF Strategic Investment Pool Award awarded to Elizabeth Schotter. This study was conducted in partial fulfillment of Sara Milligan’s doctoral dissertation. It was presented as a poster at the Psychonomic Society 65th Annual Meeting (November 21-24, 2024). The data that support the findings of this study, the full sentence stimuli, and the code used for analyses, are available at https://osf.io/mh5ez/. Correspondence concerning this article should be addressed to Sara Milligan, Department of Psychology, University of South Florida, 4202 E. Fowler Ave. PCD 4118G, Tampa, FL, 33620, email: [email protected] Word Count: 8,296 Abstract Readers frequently skip words, often when they are expected (Staub, 2015), but sometimes even when they are contextually incompatible (e.g., the high frequency article the ; Angele & Rayner, 2013). Therefore, some skipping behaviors may be automatic (e.g., based on familiarity or length) rather than based on thorough word identification, raising the question of how related word skipping behavior is to comprehension of the meaning of the text. To test this question, we co-registered eye tracking and EEG recordings and measured fixation-related brain potentials (FRPs). We time-locked FRPs to the pretarget word before a three-letter target word embedded in a high constraint sentence, and compared the neural activity between different preview conditions when the target word was subsequently either skipped or fixated (i.e., behavior-contingent analysis ). We used a gaze-contingent display change to manipulate the parafoveal word to be (1) expected, (2) an anomalous orthographic neighbor of the expected word, or (3) an anomalous instance of the , which changed to the expected word during the skip or fixate saccade. We found a parafoveal N400 effect to the anomalous neighbor compared to the expected preview that only occurred when the target was skipped, but not when it was fixated. There was also a parafoveal N400 effect for the anomalous the compared to the expected word that was not skipping-contingent. This pattern demonstrates that word skipping is associated with the depth of parafoveal linguistic processing, but these decisions are initiated prior to complete identification and contextual integration processes, which continue after the eyes have moved on. Keywords : reading, word skipping, fixation-related potentials, parafoveal N400, behavior-contingent analysis Impact Statement We used co-registration of eye movement and neural (EEG) recordings to provide novel evidence that behavioral decisions to skip a word during reading are determined before semantic information from the word is fully processed. Therefore, oculomotor decisions reflect a balancing act between drives for efficiency and fidelity in information processing, and behavior-contingent neural analysis reveals this relationship better than measuring these data streams simultaneously only to analyze them in isolation. Introduction In the process of understanding text, readers make a sequence of eye movements from one word to the next to directly fixate the words that must be recognized in, more or less, the order in which they were written. But approximately one third of the words in the text are skipped (Rayner, 1998) – not directly fixated before subsequent words in the text – suggesting they may not need foveal input for successful comprehension to proceed. Understanding complex relationships between real time language comprehension processes and oculomotor behavior during reading requires the co-registration of eye tracking and neural measures (i.e., electroencephalography; EEG) to measure reading behavior (e.g., skipping) as well as continuously unfolding language processes (e.g., the N400 response), as well as an analysis approach that allows us to investigate neural responses based on whether a word was skipped or fixated (i.e., behavior-contingent analysis; BCA). Word skipping may reflect cases in which a word was sufficiently identified parafoveally, relieving the reader of the need to obtain clearer (i.e., foveal) perceptual input. For example, skipping rates are higher for words that are high frequency (Angele et al., 2014; Henderson & Ferreira, 1993; Rayner, & Fischer, 1996; Rayner, Sereno, & Raney, 1996) or predictable (Balota, et al., 1985; Drieghe et al., 2005; Ehrlich & Rayner, 1981; Rayner & Well, 1996). Because skipping decisions, necessarily, occur before foveal fixation, any linguistic influences on these decisions must be driven by parafoveal processing. Therefore, some researchers argue that skipping reflects more thorough parafoveal word identification (e.g., Drieghe et al., 2005; Fitzsimmons & Drieghe, 2013; Morrison, 1984). However, the evidence in support of a thorough identification account of skipping is mixed. For example, Balota et al. (1985) found higher skipping rates for expected words compared to nonwords that had low orthographic overlap with the expected word (e.g., “The banker loaned the businessman some more…” money vs. toohz ). On the other hand, nonword previews with high orthographic overlap (i.e., moncg ) were skipped as often as the expected word, suggesting that skipping was driven by only partial identification and a rough mapping of orthographic form to expectations. Subsequent eye tracking studies have replicated the finding that expected words are skipped more than nonwords or anomalous words, but did not find significant differences based on the degree of orthographic overlap with the expected word (Drieghe et al., 2005; Veldre & Andrews, 2017). Drieghe et al. (2005) interpreted the lack of an orthographic similarity effect on skipping rates as evidence that skipping does reflect thorough verification and integration of semantic meaning, given that there was no skipping advantage for orthographic similarity when the preview did not make sense. One phenomenon that adds complexity to the question of why words are skipped is the finding that the word the is more-or-less automatically skipped, regardless of whether it fits in the sentence context, and in fact is skipped more frequently than a semantically correct word (Angele & Rayner, 2013). Abbott et al. (2015) found that having strong expectations for a particular target word, which were violated by the anomalous the, did not reduce the tendency to skip the anomalous the – it was skipped more than the semantically correct word both in low constraint sentences and in high constraint sentences. The is the most frequent word in the English language (Balota, et al., 2007), is short, and semantically shallow, and therefore ticks all the boxes of an easy-to-access skippable word. However, Angele and colleagues (2014) found that, like an anomalous the , anomalous high frequency content words were also more likely to be skipped than lower-frequency semantically correct words. Therefore, the strikingly high rates of the skipping, even when contextually anomalous, is not a qualitatively distinct phenomenon, but rather an effect of extreme lexical frequency. Skipping decisions are clearly associated with the ease of parafoveal lexical processing to some extent, but do not necessarily require complete word recognition or contextual integration. To isolate parafoveal processing in the investigation of reading, researchers use the gaze-contingent display change paradigm (Rayner, 1975), in which a preview word presented in parafoveal vision (i.e., while the reader fixated on the preceding word) changes to a different target word once the reader skips or fixates it. This paradigm allows researchers to dissociate the information readers had prior to the skip/fixate decision from information they obtained during direct fixation. Using this paradigm and asking the participant about the identity of the manipulated word, Schotter and Jia (2016) found that readers reported the parafoveal word 75% of the time when they skipped it, suggesting that they recognized the word they skipped, at least most of the time. But, such self-report measures do not necessarily reflect the online in-the-moment processing surrounding a skipping decision. Using co-registration to measure brain responses time-locked to fixations preceding skipping or fixating behavior has the potential to arbitrate between different accounts of the relationship between linguistic processing and skipping behavior. Neural responses reflect underlying language processing Co-registration of eye tracking and EEG (see Degno et al., 2021; Dimigen et al., 2011; Nikolaev et al., 2016) allows researchers to better understand the neural processes underlying reading by investigating fixation-related brain potentials (FRPs). Studying FRPs, analogous to event-related brain potentials (ERPs), involves averaging brain responses to the onset of a stimulus or response event and measuring deflections in these aggregate neural responses. Instead of time-locking to the moment that the word is displayed on the screen, the response is time-locked to the moment that the reader’s eyes fixate on the word, resulting in an FRP. Furthermore, the neural activity can be time-locked to any fixation during reading, allowing researchers to investigate the response to the foveal word, as well as to parafoveal word by time-locking to the fixation on the word before it (i.e., the pretarget word). The past decades of research on language processing using ERPs has identified and highlighted the N400 component, which is characterized as a negative-going deflection in the ERP signal, maximal at centroparietal electrode sites, that peaks around 400 ms after the presentation of a semantically informative stimulus (Federmeier, 2022; Kutas & Federmeier, 2011). Despite differences in presentation format (i.e., rapid serial visual presentation for ERPs vs. via participant-controlled eye movements for FRPs), the N400 response appears to be similarly sensitive to linguistic variables (Kornrumpf, et al., 2016). The vast literature on the N400 ERP component has revealed that the amplitude of the N400 component is reduced (less negative) for contextually supported or expected words compared to unexpected or otherwise contextually inappropriate words (Kutas & Federmeier, 2011). Interestingly, the N400 amplitude reflects more than just semantic processing and suggests that readers generate expectations not only about the meanings of upcoming words, but also about their forms. For example, the N400 amplitude is reduced for anomalous words that are orthographically similar to a highly expected word (Caliskan et al., 2023; DeLong et al., 2019, 2021; Laszlo & Federmeier, 2009; Ryskin et al., 2021). Expectations may allow readers to proceed with comprehension by retrieving the semantic representation of the expected word despite the subtle orthographic violation. The use of cursory orthographic processing in the presence of expectations may be especially important during parafoveal processing, given the reduced attentional allocation and visual quality, that readers would sometimes misperceive the anomalous neighbor as the expected word. Prior ERP work using the RSVP with flankers paradigm to present words in both foveal and parafoveal vision have reliably found both foveally and parafoveally elicited N400 responses to semantic violations for words that were orthographically unrelated to the expected word (Barber et al., 2010, 2013; Li et al., 2023; Payne et al., 2019; Stites et al., 2017), as well as words that were orthographic neighbors of the expected word (Milligan et al., 2023; Schotter et al., 2023). If so, it is theoretically plausible that readers are able to process parafoveal words deeply enough that the decision to skip or not skip the word would be linguistically informed. However, the parafoveal responses are more sensitive to orthographic violations than semantic violations (Nestor et al., 2025), suggesting that readers have a harder time extracting semantic information from parafoveally perceived words. Furthermore, in RSVP presentations, readers do not have the option to skip a word, so they may adjust attentional strategies to focus on foveal processing because every word will ultimately be presented in foveal vision. In other words, any effort expended on getting a head start on parafoveal processing will not ultimately influence their reading efficiency, so attention allocation strategies may adjust accordingly. A growing number of co-registration studies have demonstrated that effects of predictability (Dimigen et al, 2011; Kretschmar et al., 2015), syntactic and semantic anomaly (Antunez et al., 2022; Loberg et al., 2018; Metzner et al., 2017), and lexical frequency (Milligan, et al., 2023) can be detected in FRPs. The N400 response to an anomaly or semantically unexpected word can also be elicited in response to the fixation preceding the fixation on the word, suggesting that high level linguistic information can be extracted from words parafoveally (Antunez et al., 2021, Kretzschmar et al., 2009, Li et al., 2024, López-Peréz et al., 2016). For example, Antunez and colleagues (2021) found an N400 effect for a comparison between anomalous and plausible parafoveal words, time-locked to the pre-target word in low constraint sentences, suggesting that this effect was due to extraction of meaning from the parafoveal word rather than merely the detection of a violation of expectations. These prior FRP studies that found parafoveally-timelocked effects related to semantic preview manipulations did not specifically separate or control whether the subsequent saccade skipped or landed on the target word. Therefore, the question remains whether parafoveal semantic processing is greater for, or even driven by, trials in which the target word was skipped and the only bottom-up information available for lexical processing was parafoveal. Behavior-contingent analysis (BCA) of FRPs Most co-registration studies report FRP and eye movement measures separately in aggregate, rather than linking the measures on a trial-to-trial basis, prohibiting the ability to link specific behaviors to specific neural activity. Thus far, only two studies have tied specific eye movement decisions to variations in the underlying neuro-linguistic processing by examining behavior-contingent FRP effects in reading (Metzner et al., 2017; Milligan et al, 2024). Metzner et al. (2017) manipulated the semantic and syntactic fit of words in natural sentence reading and found differences in the FRP effects to these manipulations based on regressive eye movements. The P600 component, which has been functionally tied to syntactic processing difficulty and structural reanalysis (e.g., Osterhout & Holcomb, 1992), was larger in response to the syntactic anomaly condition when the reader made a regressive eye movement. This behavioral contingency for FRP responses demonstrated that different eye movement decisions can reflect differences in the underlying neural language processing. Milligan et al. (2024) found an analogous behavior-contingency between parafoveal N400 FRPs and word skipping behavior. They found that skipping behavior was associated with larger pretarget time-locked N400 responses to subtle orthographic violations of parafoveal previews, but only when the target word was expected (i.e., in high constraint but not low constraint sentences). Both of these studies identified distinct patterns of neural responses to experimental manipulations depending on behavioral decisions, demonstrating that language processing and eye movement behavior interact during natural reading. If unexpected or implausible words are skipped while also eliciting a parafoveal N400, it would suggest that readers can access semantics based on parafoveal processing alone. However, it would also suggest that skipping decisions are often made prior to full word identification, which may take more time and occur downstream of saccade programming that was initiated based on incomplete parafoveal word identification. The current study The current study further investigates the nature of word skipping decisions and the extent to which they index parafoveal word identification. More specifically, we aimed to determine whether FRP responses to parafoveally perceived semantic anomalies are larger when readers ultimately skip a word compared to when they fixate it. If this were the case, it would suggest that word skipping decisions are related to more deep lexico-semantic parafoveal processing. Because prior research has shown that parafoveal N400 FRPs are larger for words in high constraint sentences (Milligan et al., 2024), we presented participants with high constraint sentences that contained sentence-medial target words that were either (1) highly expected, (2) an anomalous word that was an orthographic neighbor (ON; i.e., differed orthographically by only a single substituted character) of the expected word or (3) an anomalous instance of the very high frequency word, the . We used the gaze contingent boundary display change paradigm (Rayner, 1975) to reduce the saliency of the manipulation to the reader and so that, on trials in which they fixated the target word, the foveal input was the same so any differences in process must have been due to the word that was perceived parafoveally. The target words were all three-letters long, in order to maximize skipping rates because word length is the strongest single predictor of skipping probability, accounting for 50% of the variance in contrast to word frequency and predictability, which account for only ~ 4% and Drieghe et al., 2004; Slattery & Yates, 2018). We expected that anomalous orthographic neighbor previews would elicit a parafoveal N400 effect (i.e., more negative amplitude compared to the expected word; see Milligan et al., 2023; Schotter et al., 2023; cf. Nestor et al., 2025). Critically, we expected an interaction between this effect and skipping behavior (i.e., a larger N400 effect when the reader decided to skip versus fixate the target word; see Milligan et al., 2024). If the is frequently skipped by nature of it being short and high frequency, and therefore processed parafoveally to the point of accessing semantic content, we expected a parafoveal N400 effect for the anomalous the condition (i.e., more negative amplitude compared to the expected word). If, like the anomalous ON, the anomalous the N400 effect is behavior-contingent, it would indicate deeper processing and more thorough identification of skipped words. However, if the anomalous the N400 effect is observed regardless of skipping behavior it would suggest that the is a special type of word that might be automatically skipped, regardless of how deeply it is processed (see Abbott et al., 2015; Angele & Rayner, 2013). Participants Sixty-five participants recruited from the Psychology Department’s SONA subject pool at the University of South Florida completed the experiment. Participants were compensated with course credit or recruited through flyers and mailing lists and paid $16/hour for their time. All participants were right-handed native English speakers between ages 18 and 35 with normal or corrected-to-normal vision and no history of reading, learning, or neurological disorders. The study was approved by the Institutional Review Board at the University of South Florida and participants provided informed consent via an online consent form approved by the IRB. Ten participants were excluded from analysis because we discovered they did not meet the inclusion criteria after they had already participated (n = 4), an EEG recording error resulting in inability to synchronize the EEG and eye tracking data (n = 1), or because of poor EEG data quality (e.g., excessive muscle noise, slow drifting, participant movement, disconnected electrodes; n = 5). These exclusions resulted in a final sample of 55 usable participants. Sample size rationale Milligan et al. (2024), who also conducted a skipping-contingent analysis of the parafoveal N400 had 45 participants, but the current experiment contained filler stimuli for a separate experiment for which we wanted a larger sample size, so we aimed for a final sample of 55 participants. We conducted a sensitivity analysis using PANGEA (v0.2; https://jakewestfall.shinyapps.io/pangea/), which showed that with 55 participants and 20 items per condition (conservatively assuming that approximately two thirds of trials would meet the inclusion and behavioral criteria) per skipping behavior (assuming a roughly even split between target skipping and fixating trials), we would be able to detect an interaction effect between the parafoveal word manipulations and skipping behavior with a small-medium effect size (Cohen’s d = 0.352) at a power of .80. Stimuli and design Sentence stimuli consisted of 180 high constraint sentences that made a 3-letter target word predictable. The parafoveal preview was manipulated using the gaze-contingent display change paradigm (Rayner, 1975; Figure 1), with three preview conditions: an identical, expected word (i.e., no display change between preview and target), an anomalous orthographic neighbor (ON) of the predictable target word, and an anomalous the (see example sentence and preview conditions in Figure 1). The target word that was displayed after the eyes crossed the invisible boundary (located immediately after the last character in the pretarget word) was always a predictable, plausible word. Preview condition was counterbalanced across sentence items in three separate lists so that each sentence was presented in every preview condition across the three lists. Each participant saw one list, with each sentence item presented in only one of the preview conditions, resulting in each participant seeing 60 items per preview condition. The anomalous neighbor and identical preview words were matched on average across items on variables have been previously shown to influence eye movements and ERP responses: lexical frequency, orthographic neighborhood size (i.e., the number of real words that share all but one letter with a given word), and semantic diversity (i.e., a measure of the diversity of different contexts in which a word tends to appear in the language; see Table 1). The letter that differed between the expected word and the anomalous ON was evenly distributed across all three letter positions (i.e., first, middle, and final). Item presentation was counterbalanced across all preview conditions, so the pretarget word lexical characteristics did not vary, on average, across preview conditions. Figure 1 Diagram of gaze-contingent boundary display change paradigm and example of experimental sentences and parafoveal preview manipulation In designing the stimuli, we made a deliberate effort to use pretarget words that were not particularly short or high frequency in an attempt to drive down skipping rates of the pretarget words, both because the FRPs of interest were those time-locked to fixations on the pretarget words. The average length of the pretarget words was 6.4 characters (SD = 1.7), the average frequency was 3.1 (SD = 1.1; LgSUBTLWF per million words), and the average orthographic neighborhood size was 3.0 (SD = 4.3). Table 1 Descriptive statistics of lexical characteristics for expected target and anomalous orthographic neighbor preview words Expected Target Word Anomalous Orthographic Neighbor The Frequency 3.2 (0.8) 3.8 (1.0) 17.0 (–) Orthographic Neighborhood Size 18.2 (5.9) 20.0 (5.2) 7 (–) Semantic Diversity 1.7 (.3) 2.0 (0.3) 2.4 (–) Note. Means are reported with standard deviations in parentheses. Frequencies are log SUBTLWF per million words (Retrieved from the English Lexicon Project (https://elexicon.wustl.edu/index.html); Balota, et al., 2007). Sentence norming Sentence stimuli were normed to verify the predictability of the target word and the (im)plausibility of the target and preview words. The mean cloze probability (i.e., proportion of responses that matched the intended target word) was .74 (SD = .23), verifying that the sentence stimuli were at least moderately constraining toward the target word. The cloze probabilities of both the anomalous neighbor and anomalous the were 0. The mean plausibility rating (on a Likert scale from 1 (highly implausible) - 7 (highly plausible)) for the expected word was 5.6 (SD = .5) and for the anomalous ON was 3.0 (SD = .5). We did not collect normative data for the plausibility of the anomalous the out of concern that presenting sentences in a norming study in which the word the repeatedly appeared anomalously might produce unusual judgments from either heightened or dampened sensitivity to the error. However, the always replaced a content word and, based on part of speech alone, was always anomalous in the sentence. Apparatus and recording EEG was recorded from 30 Ag/AgCl active electrodes (extended 10/20-system) using an actiCAP/actiCHamp electrode cap and amplifier system (Brain Products) with a 500 Hz online sampling rate. No online frequency filters were used during recording. Horizontal and vertical electrooculogram (EOG) were recorded from two pairs (bipolar reference) of passive electrodes placed on the outer canthi of each eye and above and below the right eye. The EEG signal was referenced online to the left mastoid and re-referenced offline to the algebraic mean of the left and right mastoids. Impedance values were reduced to 10 kΩ or lower at all electrode sites prior to recording and any noisy or disconnected electrodes were adjusted and reconnected as needed throughout the experiment. Eye movements were recorded using an SR Research Ltd. Eyelink 1000 Plus eye tracking camera in remote desktop mode (sampling rate of 500 Hz). Viewing was binocular, but eye movements were recorded only from the right eye. A three-point horizontal calibration across the central y-position of the screen (where sentences were displayed) was performed at the beginning of the experiment and calibration accuracy had to fall within .3° of visual angle at each point to be accepted. Re-calibration was performed periodically throughout the experiment if accuracy dropped below this level, as determined by an inter-trial drift check. Procedure Participants were seated at a viewing distance of 60 cm from a BENQ XL2540 model LCD monitor with a 240 Hz refresh rate and screen resolution of 1920 x 1080 pixels. They were instructed to read the sentences normally for comprehension and were presented with 5 practice trials to acclimate them to the task. The sentence text appeared on the screen in black 12-point Courier New (monospaced) font on a light gray background. One-hundred and twenty sentences (from another experiment, not presented here) containing a constraint manipulation, plausible target words, and no display change served as filler sentences and were randomly interspersed with those from the current experiment. Both experimental and filler sentences were followed by yes/no comprehension questions on 25% of trials, with equal numbers of yes and no correct responses, to ensure that participants read the sentences for meaning. Each trial was initiated by the experimenter after a drift check of the eye tracker. To trigger the presentation of the sentence, the participant had to make a fixation in a black box on the left side of the screen at the location of the beginning of the sentence. Participants were instructed to look at a target sticker on the right edge of the monitor (off screen) once they finished reading for comprehension to ensure that their eyes did not linger on the sentence regions after reading. Each trial was terminated by the participant via a manual button press when they had finished reading. The gaze-contingent boundary paradigm (Rayner, 1975) was used to present the parafoveal preview manipulations; readers therefore only ever directly fixated on the semantically correct expected target word. The invisible boundary was placed immediately after the last letter of the pretarget word, to the left of the space before the target word; prior to fixating the target word, the preview word was visible. The target word replaced the preview within approximately 5-10 ms of the eye tracker detecting that the participants’ eyes had crossed the boundary. After the experiment, participants were asked whether they noticed anything unusual about the sentence display. When prompted, nineteen participants reported noticing a word flicker or change. Eight of these participants were able to report a specific word they remembered changing, but none of them reported seeing display changes or flickering on more than 10 trials. Data cleaning and preprocessing Eye-tracking data were cleaned using the DataViewer program (SR Research Ltd., version 4.3.1) and fixations in any of the predefined interest areas were merged with a neighboring fixation if their duration was below 80 ms and they were within 0.3° horizontally. Trials were excluded from analyses due to eye tracking data loss, display change errors, and pretarget word skipping. We also included only trials with single first-pass fixations on the pretarget word to remove any noise or jitter in the FRPs that might be introduced when the readers made a first pass refixation, following a mislocated initial fixation, for example. EEG data preprocessing was performed using the EEGLAB (v2025.0.0; Delorme & Makeig, 2004), ERPLAB (v12.00; Lopez-Calderon & Luck, 2014) and EYE-EEG (v0.99; Dimigen, et al., 2011) toolboxes in Matlab. The EEG was re-referenced offline to an algebraic mean of the right and left mastoids and band-pass filtered from 0.1-30 Hz, with 0.2 and 19.4 Hz half-power rolloffs, using an IIR Butterworth filter. Ocular artifacts were removed from the EEG using optimized independent components analysis (OPTICAT, version 2020-01-28), following the procedures and recommendations described in Dimigen (2020). The ICA was trained using band-pass filtered (with a passband edge of 3 Hz) training data that over-weighted spike potentials by a factor of 1. Ocular artifact components were automatically flagged and removed using eye tracker-guided eye artifact component identification (Plöchl, Ossandón & König, 2012), using a variance ratio threshold of 1.1. EEG was epoched into segments from 200 ms before to 1000 ms after the start of fixations on the pretarget and target words and baseline corrected by subtracting the mean voltage from -200 to 0 ms for each channel. Epochs containing artifacts were flagged for removal using a moving window peak-to-peak threshold automatic artifact detection algorithm, rejecting epochs with voltage changes of greater than 100 μV within a 200 ms time span, with a 50 ms window-step. The epoched data were also inspected manually to confirm that artifact-contaminated epochs were flagged for exclusion. The resulting dependent variables from the eye tracking and EEG data were exported from their respective processing softwares and were merged at the trial-level for confirmatory analyses in R. After synchronizing and merging the EEG and eye tracking data, 6012 trials were retained that had single fixations on and forward saccades out of the pretarget word. Of these, 541 trials were flagged either for EEG artifacts (433 trials) or eye tracking or display issues (111 trials), including premature or delayed display changes, track loss, or calibration issues, which were identified via manual inspection of the eye tracking data. After these trial exclusions, 5468 trials were retained for the FRP analyses. The average number of trials retained per subject was 33.98 (SD = 7.50) in the identical condition, 33.15 (SD = 7.13) in the anomalous ON condition, and 32.34 (SD = 7.36) in the anomalous the condition. For the eye tracking analyses, we did not exclude trials with only EEG artifacts. However, to ensure that the eye tracking analyses aligned as closely as possible with FRP analyses we did include only trials with single fixations on the pretarget word, forward saccades out of the pretarget word, and with no early display change errors, resulting in a slightly larger number of total trials retained in the eye tracking analyses (5901 trials) compared to the FRP analyses. Transparency and openness The preprocessed EEG and eye tracking data on which these analyses were performed, the R code for plotting and analyses, and the experimental sentence stimuli are publicly available on OSF at https://osf.io/mh5ez/. We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study. No components of this study were publicly preregistered, however the experimental design and analysis plan were proposed to and approved by a dissertation committee of faculty from the University of South Florida and the University of Utah prior to data collection and analysis. Results The primary dependent variable for the FRP analyses was the parafoveal N400 (i.e., the amplitude averaged across a centroparietal ROI (C3, C4, Cz, CP1, CP2) from 300 - 500 ms post-fixation time-locked to single fixations on the pretarget word (i.e., the time point at which the preview was visible parafoveally). The ROI was selected a priori to include a representative selection of electrodes surrounding Cz and including centro-parietal sites because the N400 has typically been described as being maximal centro-parietally (e.g., see Kutas & Federmeier, 2011). In order to understand these responses in the context of eye movement behaviors, we also analyzed target word skipping rates and target single fixation durations. We also analyzed the effect of parafoveal preview condition on SFD on the pretarget word to ensure that there were no statistical differences in fixation durations on the fixations to which we time locked the FRPs. All analyses were performed using (generalized) linear mixed-effects regression models via the glmer() function (i.e., with a logit link for the binary outcome measure of skipping) and the lmer() function (i.e., for the FRP measures) from the lme4 package (version 1.1-12; Bates et al., 2015) within the R Environment for Statistical Computing (version 3.3.1). All analyses included main effects of parafoveal preview condition (expected, anomalous ON, anomalous the ), which were coded as successive differences contrasts comparing the anomalous ON to the expected word and comparing the expected word to the anomalous the . For the FRP analyses splitting by skipping behavior, skipping was entered as a treatment contrast (0, 1) with target skipped as the baseline condition. All models included preview effects as random intercepts for both subjects and items. Skipping rates and fixation durations of (fixated) target words Skipping rate did not differ significantly between the expected word (mean = .49, SD = .17) and the anomalous ON (mean = .47, SD = .18; see Table 2 for analysis results). However, skipping rates in the anomalous the condition (mean = .67, SD = .18) were significantly higher than the expected word condition, replicating the findings from Angele and Rayner (2013; Abbott et al., 2015) that skipping of an anomalous the is more frequent than skipping of plausible or expected words (see Figure 2A). Although skipping rates were higher for the anomalous the , downstream effects on fixation durations (for trials without target word skipping) demonstrate a substantial disruption from the anomalous the preview, indicated by longer single fixation durations (mean = 274, SD = 68) compared to the expected (mean = 226, SD = 36) or anomalous ON (mean = 260, SD = 44) conditions (See Figure 2B). We did not find any significant differences in single fixation durations on the pretarget words (i.e., no parafoveal-on-foveal effects) based on preview condition for the anomalous neighbor (mean = 246, SD = 36) versus the expected (mean = 247, SD = 35) preview comparison (b = 1.26, t = .48) or the expected versus anomalous the (mean = 250, SD = 38) preview comparison (b = 2.59, t = .99).11Although sentence items were counterbalanced across condition, we wanted to ensure that the lexical characteristics of the pretarget words were still equivalent in the retained trials because of potential differences in pretarget fixation probabilities based on item-level variance. We assessed whether the length and lexical frequency of the pretarget words in the retained trials varied by condition and fixation behavior and found no significant differences in length, but that lexical frequency of the pretarget word was significantly higher on target skipping trials. To ensure that this frequency difference for skipping trials did not influence the results of the behavior-contingent analyses predicting N400 amplitude by preview condition and skipping behavior (Table 3), we added lexical frequency as a covariate in this model and found no differences in the patterns of effects for the predictors of interest (i.e., skipping behavior and preview condition) on the N400 effect compared to the model reported in Table 3. These supplementary analyses testing for differences in pretarget length and frequency (Table 1A) and controlling for lexical frequency in the primary analysis predicting the N400 effect (Table 2A) are reported in Appendix A. Table 2 Results of generalized linear mixed effects models predicting target word skipping rate and single fixation duration by preview condition Target Skipping Rate Target Single Fixation Duration Predictors Est. SE z p Est. SE t p (Intercept) 1.26 0.14 2.13 0.033 255.61 5.86 43.63 <0.001 Anomalous ON vs. expected 1.10 0.09 1.23 0.219 -34.78 4.30 -8.10 <0.001 Expected vs. anomalous the 2.58 0.25 9.77 <0.001 55.38 5.07 10.93 <0.001 Observations 5901 2607 Note. The skipping rate estimates are reported as the odds ratio from the glmer() analysis. Figure 2 Average skipping rates (panel A) and single fixation durations (panel B) of the target word by parafoveal preview condition Note. Error bars represent standard error. Skipping contingent parafoveal N400 FRP effects The N400 amplitude time-locked to fixations on the pretarget word was significantly more negative for both the anomalous ON and the anomalous the compared to the expected word (as a main effect in target skipping trials, which were coded as the baseline in the model; Table 3). The anomalous ON effect also interacted significantly with target word skipping, with a larger N400 effect when the target was skipped compared to when it was fixated (see Figure 3 for waveforms plotted at selected scalp electrodes and Figure 4A for waveforms in target skipped and target fixated trials, respectively, at the centroparietal ROI used in the N400 analyses), replicating the patterns reported by Milligan et al. (2024). For the anomalous the effect, however, the N400 effect did not significantly differ as a function of skipping behavior. Topographic maps of these N400 effects (Figure 4B) exhibit a canonical centro-parietal scalp distribution regardless of skipping behavior. To better understand the nature of the interaction between the anomalous neighbor and identical preview comparison and skipping behavior on the N400 amplitude, we also performed a follow-up analysis splitting trials by skipping. This analysis showed that the N400 effect for the anomalous neighbor was only significant for skipped trials, while the N400 effect of the anomalous the compared to the identical condition was significant for both skipped and fixated trials (Table 4). Table 3 Results of linear mixed effects regression predicting pretarget time-locked N400 amplitude by preview condition and skipping behavior Parafoveal N400 Predictors Est. SE t p (Intercept) -1.05 0.22 -4.78 < 0.001 Anomalous ON vs. expected 1.64 0.40 4.08 < 0.001 Expected vs. anomalous the -1.19 0.34 -3.46 0.001 Target skipping (fixated vs. skipped) 0.48 0.23 2.13 0.033 Anomalous ON vs. expected x skipping -1.37 0.52 -2.61 0.009 Expected vs. anomalous the x skipping 0.03 0.54 0.06 0.949 Observations 5468 Table 4 Results of linear mixed effects regressions predicting parafoveal N400 amplitude by preview condition split by skipping behavior N400 - Target Skipped N400 - Target Fixated Predictors Est. SE t p Est. SE t p (Intercept) -1.07 0.24 -4.47 < 0.001 -0.53 0.21 -2.54 0.011 Anomalous ON vs. expected 1.64 0.47 3.49 < 0.001 0.23 0.39 0.59 0.555 Expected vs. anomalous the -1.17 0.35 -3.38 0.001 -1.11 0.42 -2.63 0.009 Observations 2978 2490 Figure 3 N400 waveforms across selected (frontal, central, parietal, occipital by left, right, center positions) scalp electrode montage for target skipped (panel A) and fixated (panel B) trials A) B) Figure 4 N400 waveforms (panel A) by preview condition at averaged centroparietal ROI (electrodes C3, Cz, C4, CP1, CP2) and scalp topographic maps (panel B) of preview effects (i.e., condition differences) at averaged N400 time window (300 - 500 ms) split by skipping behavior (top: target skipped, bottom: target fixated) Exploratory analyses: trial-level relationships between preview condition, skipping, and regressive eye movements Because the skipping-contingent FRP analyses revealed that skipped words are identified as anomalous parafoveally, we were interested in whether the disruption from recognizing the anomaly (as demonstrated by the N400 effects) might be apparent in eye movement measures downstream of the skipping decisions. Therefore, we conducted follow-up analyses to investigate whether regression rates out of the post-target region of the sentence were higher when the target word was skipped compared to when it was fixated and whether these regressions were modulated by preview condition. We hypothesized that when readers decided to skip anomalous words, they identified the anomaly parafoveally, but not in time to cancel the skip and that the disruption to comprehension from encountering an anomalous preview might emerge further downstream in regression rates. Indeed, we found that for all preview conditions regression rates out of the end of the sentence were significantly higher (see Table 5) when the target word was skipped versus when it was fixated. There was a significant interaction between skipping behavior and the anomalous the effect on regression rates: there was a larger difference in regression rates between the anomalous the and the expected word when the target was skipped compared to when it was fixated (Table 5; Figure 4). The anomalous the effect was not significant in the simple effects for fixated trials, but it is possible that we did not have the power due to the high skipping rates in the anomalous the condition, leaving fewer fixated trials for this particular comparison. Although the interaction between skipping and the anomalous ON effect was not significant, when splitting by skipping behavior, the effect was significant for skipping trials but not for target fixating trials. Table 5 Results of glmer analyses predicting regression rates out of the post-target sentence region by preview condition and target skipping behavior Target Skipped Baseline Target Fixated Only Predictors Est. SE z p Est. SE z p (Intercept) 0.48 0.06 -5.73 <0.001 0.19 0.03 -10.64 <0.001 Anomalous ON vs. expected 0.62 0.08 -3.63 <0.001 0.76 0.12 -1.82 0.068 Expected vs. anomalous the 2.50 0.33 6.99 <0.001 1.36 0.24 1.73 0.083 Target skipping (fixated vs. skipped) 0.42 0.03 -12.04 <0.001 Anomalous ON vs. expected x skipping 1.35 0.22 1.79 0.074 Expected vs. anomalous the x skipping 0.51 0.09 -3.78 <0.001 Observations 6118 2825 Figure 4 Regression rates out of the post-target sentence region by preview condition and target skipping behavior Overall, these patterns in regression rates align closely with the patterns in the pretarget N400 responses. When readers skipped, they appeared to have gotten further in processing, resulting in larger N400 effects, particularly for the anomalous ON compared to the expected word. However, they did not identify the anomaly quickly enough to terminate the skipping decision even when the N400 response demonstrated that the error was detected. In fact, the anomalous the was skipped more often than an expected, plausible word. But, if we look further downstream in the eye movement record, the disruption in processing caused by the skipped anomalous previews becomes apparent in regression decisions. Discussion The current experiment tested whether word skipping behavior during reading reflects deeper linguistic parafoveal processing of skipped versus fixated words. We presented short three-letter words as parafoveal previews in high constraint sentences that were either the expected word or were contextually anomalous, either subtle orthographic errors or the high frequency function word the . We co-registered eye movements with EEG recordings and separated sentence reading trials based on word skipping behavior to investigate whether language-related neural responses (i.e., the parafoveal N400 FRP) differ based on these eye movement decisions. A parafoveal N400 FRP effect was elicited in response to the anomalous ON only when the word was ultimately skipped, but not when it was fixated. In contrast, the parafoveal N400 FRP response was elicited in response to the anomalous the, regardless of word skipping behavior. The interaction between the anomalous ON comparison and skipping behavior demonstrates that word skipping can be associated with thorough identification of the word parafoveally. However, the lack of an interaction between the anomalous the comparison and word skipping suggests that there are some (potentially unique and extreme) scenarios in which word skipping may be unrelated to ongoing language processing. Behavior-contingency of parafoveal detection of orthographic violations We used the orthographic anomaly preview condition to test whether readers are capable of precisely identifying the orthographic form of upcoming words in parafoveal vision and whether there was a relationship between the precision of this bottom-up decoding and word skipping behavior. The fact that the ON anomaly elicited a larger parafoveal N400 compared to the expected word demonstrates that even when the visual input is lower fidelity and attentional resources must be covertly shifted away from the foveal input, readers are capable of distinguishing between an expected word and a visually similar anomalous word. However, this N400 effect was only present when the reader decided to skip the target word. On trials in which the target word was fixated, the parafoveal preview did not appear to be fully identified as an anomaly and the N400 amplitude was comparable for the expected word and ON violation. This relationship between the oculomotor decisions and neural index of linguistic processing provides compelling evidence that skipping decisions are determined, at least in part, by the depth of linguistic parafoveal processing. The presence of the parafoveally elicited skipping-contingent N400 response demonstrates that the readers detected the difference between the expected word and the orthographically similar anomalous previews, but the fact that their skipping rates were roughly equivalent suggests that the semantic violation was not identified quickly enough to abort the skipping decision. Therefore, we argue that the linguistic influences on skipping decisions are limited to successful early stages of linguistic processing that satisfy a cursory confidence threshold (e.g., Pollatsek et al., 2006; Schotter et al., 2014) or a “hedged bet” Schotter (2018) that that parafoveal word identification will likely be successful. This linguistic influence on skipping is not, however, due to the completion of full semantic access and integration. It should be noted that, in the current study, all of the sentences were moderately to highly constraining to the target word. Therefore, the linguistic influence on skipping in this case may result from contextually preactivated lexical representations (see Caliskan et al., 2023; Laslo & Federmeier, 2009). It is possible that having two out of three of the letters in the orthographic neighbor map onto the expected word was enough to trigger a skipping saccade. Alternatively, in the cases in which the reader chose not to skip the target word (roughly half the time for the ON condition), the downstream neural response of the N400 demonstrated reduced sensitivity to the parafoveal preview manipulation. Therefore, we argue that in these cases, the reader was not able to achieve extensive enough parafoveal linguistic processing and that the failure to reach a sufficient identification threshold resulted in the reader deciding to fixate the word directly to achieve comprehension. Finally, in cases where the anomaly was skipped, regression rates were higher than they were in the expected preview condition. We interpret these regression patterns as indicating that despite sufficient linguistic processing to trigger a skipping decision, readers did experience a disruption to comprehension once they realized that the word they skipped didn’t make sense. Therefore, skipping decisions appear to be influenced by the success of early stages of linguistic processing achieved parafoveally, but the information extracted parafoveally “percolates” over time and higher-level semantic processing lags behind the behavioral decision. Relatively automatic skipping of short, high frequency articles (i.e., the ) Based on high skipping rates of an anomalous the , Angele and Rayner (2013; Abbott et al., 2015; Angele et al., 2014) proposed that short, high frequency articles are skipped automatically, regardless of their contribution (or lack thereof) to a higher level semantic representation. The current study replicates their findings in eye movement behavior with the anomalous the being skipped almost 70% of the time when a different three-letter content word was highly predictable. Our parafoveal N400 FRP data further clarify this phenomenon. The anomalous instance of the appears to have been identified as anomalous (eliciting an N400 response) irrespective of skipping behavior, suggesting that parafoveal identification of high frequency function words is relatively ubiquitous. It is worth noting that because the skipping rates of the were so high, the number of trials entering into the FRP analysis for fixated trials was numerically lower than for skipped trials, however the N400 effect of the preview violation in these cases still reached significance. The fact that the anomalous the , just like the ON anomaly, was skipped even when it was detected as anomalous suggest that a shallow rapid assessment of the upcoming word’s familiarity, rather than thorough word identification and comprehension, has a strong influence on skipping decisions. The extreme familiarity of function words such as the , and perhaps their relatively shallower semantic richness, allow them to be easily and rapidly identified parafoveally, eliminating the need for subsequent foveal processing. So, if readers detected these contextual violations parafoveally, why did they still decide to skip? Based on our follow-up analyses looking at the relationship between skipping and regressive eye movements, it is clear, particularly for the anomalous the previews, that comprehension was disrupted by the anomalous previews. When the target was skipped, the anomalous parafoveal previews resulted in higher regression rates out of the end of the sentence. Based on this relationship between preview condition, skipping, and downstream regressions, we propose that early eye movement decisions, like skipping, are valid indices of the success of parafoveal linguistic processing, but that they are often executed prior to full recognition and integration of the word into the meaning of the sentence. Therefore, an initial familiarity check (Reichle et al., 1998; Reichle et al., 2011; Reichle et al., 2012) or a check for compatibility of orthographic form with a predicted word form (in the case of the ON anomalies) can prompt the reader to initiate a skipping saccade even when the skipped word makes no sense. Skipping decisions are therefore linguistically driven, but not necessarily indicative of deep comprehension. If even anomalous instances of high frequency articles are skipped automatically, why did readers sometimes fixate the anomalous the ? One possibility is that target-fixated cases in the anomalous the preview condition may have been simply due to oculomotor error, leading to mislocated fixations (Nuthmann, et al., 2005), in which the reader fixated on a word that was intended to be skipped (Engbert & Nuthmann, 2008), which may be as high as 30% of the time for three-letter words (Krugel & Engbert, 2010). Implications for theories of eye movement control and language comprehension Some researchers have argued that decisions about how long to spend fixating a word are strongly influenced by linguistic processing, but that decisions about whether to skip it are largely explained by low-level visual properties of the text (Rayner & McConkie, 1976). However, models of eye movement control during reading (e.g., E-Z Reader, Reichle et al., 2006; SWIFT, Engbert, et al., 2005; OB-1 Reader, Snell et al., 2018) coalesce around the idea that covert attention can be directed to upcoming words before the eyes move to fixate them, allowing for pre-processing of parafoveal linguistic information. The premise that skipping is influenced by linguistic processing hinges on the question of whether words, and their semantics, can be fully identified parafoveally. Our existing knowledge of the relationship between skipping and linguistic processing, however, is limited to behavioral evidence. The current study, therefore, provides a valuable piece of the puzzle by providing neural evidence of an association between skipping decisions and linguistic processing. The N400 component has been tied to semantic expectancy (e.g., see Kutas & Federmeier, 2011) and, at the very least, is sensitive to orthographic expectancy violations (Laszlo & Federmeier, 2009). Therefore, the fact that we find a relationship between the N400 response to parafoveal expectancy violations and skipping rates provides convincing evidence that skipping decisions do reflect the success of parafoveal linguistic processing. The question of whether readers’ oculomotor systems regularly drive them to move their eyes past words that have not been fully identified also has implications for the nature of language comprehension more generally. It has previously been proposed that the process of language comprehension is well adapted to dealing with incomplete or noisy input (Shannon, 1948). This noisy channel framework provides an explanation for how comprehenders are able to deal with typographical and word choice errors (Ryskin et al., 2021), anomalous word transpositions (Hossain & White, 2023), inserted or missing function words (Staub et al., 2019), and comprehending speech in auditory noise (Van et al., 2022), for example. From this perspective, that comprehenders use rational inference to comprehend imperfect linguistic signals, it may be that word skipping represents an example of the system’s tolerance for incomplete bottom-up input. Our findings that skipped words have often been partially, but not fully, identified bottom-up aligns with an account of word skipping as a cognitively and linguistically informed hedged bet in service of information processing efficiency. Implications for ERP and co-registration research Our study has important implications for the field of ERPs and psycholinguistics. For decades, limitations to ecological validity associated with the use of the RSVP paradigm have been accepted as a necessary obstacle. However, ERPs to serially presented single words may not fully reflect the neural processes required for the task of language comprehension during reading. If not all of the phenomena observed during fixed-gaze paradigms are generalizable to natural reading, then any theories of language processing that are based exclusively on RSVP findings would need to be re-evaluated and replicated using co-registration. Our finding that (some) neural responses during reading are behavior-contingent (see also Metzner et al., 2017; Milligan et al., 2024) suggests that merely replicating RSVP ERP studies with co-registration may not be sufficient to evaluate whether past findings are generalizable. Directly tying behavioral decisions to neural responses during reading at a single trial level (i.e., behavior-contingent analysis) has the potential to maximize the benefits of the co-registration method above and beyond treating eye movements and FRPs as separate dependent measures. Behavior-contingent analysis offers a unique opportunity to connect disparate and sometimes contradictory literatures that have studied similar aspects of language processing using different methods (i.e., ERPs and eye tracking). Conclusion Our study demonstrates that readers can perform a substantial amount of lexical processing parafoveally because parafoveally perceived anomalies elicit N400 effects, even when the violation is orthographically subtle. More importantly, we demonstrated that word skipping behavior modulates the magnitude of these N400 effects, demonstrating that skipping behavior reflects deeper processing of the word during parafoveal preview. However, we also found that readers skip anomalous high frequency function words relatively easily and automatically, leading to a parafoveally elicited N400 effect for an anomalous the preview regardless of skipping behavior. As a whole, our findings demonstrate that oculomotor decisions during reading, and skipping decisions in particular, reflect a push and pull between the need to (1) accurately extract semantic information embedded in symbolic visual language forms and (2) rapidly move the eyes forward to perceive and attend to new information. CRediT Statement Sara Milligan : Conceptualization (lead), Data curation, Formal analysis, Investigation, Methodology (lead), Software, Supervision (equal), Validation, Visualization (lead), Writing - original draft (lead) Elizabeth R. Schotter : Conceptualization (supporting), Funding acquisition, Methodology (supporting), Project administration, Resources, Supervision (equal), Visualization (supporting), Writing - original draft (supporting) Acknowledgments We thank Casey Stringer, Ayah Elaboudi, and Milca Jaime Brunet for assistance with experiment programming and preparation and Bekhzodbek Moydinboyev for assistance with the development of data processing scripts for successfully aligning EEG and eye tracking recordings. We also thank Milca Jaime Brunet, Snezana Trendova, Nadija Sulcaj, Nitya Thakkar, Brooklyn Olesen, Kaniz Angel, Ayah Elaboudi, and Rhea Mancha for their dedicated assistance with participant recruitment and data collection. Finally, Sara would like to thank the members of her dissertation committee, Ruthann Atchley, Peter Clayson, Brennan Payne, and Sandra Schneider, for their valuable advice and feedback on this project. References Abbott, M. J., Angele, B., Ahn, Y. D., & Rayner, K. (2015). Skipping syntactically illegal the previews: The role of predictability. Journal of Experimental Psychology: Learning, Memory, and Cognition, 41(6), 1703. Andrews, S., & Veldre, A. (2019). What is the most plausible account of the role of parafoveal processing in reading?. Language and Linguistics Compass, 13 (7), e12344. Angele, B., & Rayner, K. (2013). Processing the in the parafovea: Are articles skipped automatically?. Journal of Experimental Psychology: Learning, Memory, and Cognition, 39 (2), 649. Angele, B., Laishley, A. E., Rayner, K., & Liversedge, S. P. (2014). The effect of high-and low-frequency previews and sentential fit on word skipping during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40 (4), 1181. Antúnez, M., Milligan, S., Hernández‐Cabrera, J. A., Barber, H. A., & Schotter, E. R. (2022). Semantic parafoveal processing in natural reading: Insight from fixation‐related potentials & eye movements. Psychophysiology, 59 (4), e13986. Balota D. A., Pollatsek A., Rayner K. (1985). The interaction of contextual constraints and parafoveal visual information in reading. Cognitive Psychology, 17, 364–390. Balota, D.A., Yap, M.J., Cortese, M.J., Hutchison, K.A., Kessler, B., Loftis, B., Neely, J.H., Nelson, D.L., Simpson, G.B., & Treiman, R. (2007). The English Lexicon Project. Behavior Research Methods, 39, 445-459. Brysbaert, M., & Vitu, F. (1998). Word skipping: Implications for theories of eye movement control in reading. In Eye guidance in Reading and Scene Perception (pp. 125-147). Elsevier Science Ltd. Clifton Jr, C., Ferreira, F., Henderson, J. M., Inhoff, A. W., Liversedge, S. P., Reichle, E. D., & Schotter, E. R. (2016). Eye movements in reading and information processing: Keith Rayner’s 40 year legacy. Journal of Memory and Language, 86, 1-19. Coltheart, M., Rastle, K., Perry, C., Langdon, R., & Ziegler, J. (2001). DRC: a dual route cascaded model of visual word recognition and reading aloud. Psychological Review, 108 (1), 204. Cutter, M. G., Drieghe, D., & Liversedge, S. P. (2017). Reading sentences of uniform word length: Evidence for the adaptation of the preferred saccade length during reading. Journal of Experimental Psychology: Human Perception and Performance, 43 (11), 1895. Degno, F., Loberg, O., & Liversedge, S. P. (2021). Co-Registration of Eye movements and fixation—related potentials in natural Reading: Practical issues of experimental design and data analysis. Collabra: Psychology, 7 (1), 18032. Delorme A & Makeig S (2004) EEGLAB: an open-source toolbox for analysis of single-trial EEG dynamics, Journal of Neuroscience Methods 134 :9-21. Dimigen, O. (2020). Optimizing the ICA-based removal of ocular EEG artifacts from free viewing experiments. NeuroImage, 207, 116117. Dimigen, O., Sommer, W., Hohlfeld, A., Jacobs, A. M., & Kliegl, R. (2011). Coregistration of eye movements and EEG in natural reading: analyses and review. Journal of Experimental Psychology: General, 140 (4), 552. Drieghe, D., Brysbaert, M., Desmet, T., & De Baecke, C. (2004). Word skipping in reading: On the interplay of linguistic and visual factors. European Journal of Cognitive Psychology, 16 (1-2), 79-103. Drieghe, D., Rayner, K., & Pollatsek, A. (2005). Eye movements and word skipping during reading revisited. Journal of Experimental Psychology: Human Perception and Performance, 31 (5), 954. Ehrlich, S. F., & Rayner, K. (1981). Contextual effects on word perception and eye movements during reading. Journal of Verbal Learning and Verbal Behavior, 20 (6), 641-655. Engbert, R., & Nuthmann, A. (2008). Self-consistent estimation of mislocated fixations during reading. PLoS ONE, 3 (2), e1534, 1–6. Eskenazi, M. A., & Folk, J. R. (2015). Reading skill and word skipping: Implications for visual and linguistic accounts of word skipping. Journal of Experimental Psychology: Learning, Memory, and Cognition, 41 (6), 1923. Fitzsimmons, G., & Drieghe, D. (2013). How fast can predictability influence word skipping during reading?. Journal of Experimental Psychology: Learning, Memory, and Cognition, 39 (4), 1054. Gautier, V., O’Regan, J. K., & Le Gargasson, J. F. (2000). ‘The-skipping’revisited in French: programming saccades to skip the article ‘les. Vision Research, 40 (18), 2517-2531. Heilbron, M., van Haren, J., Hagoort, P., & de Lange, F. P. (2023). Lexical processing strongly affects reading times but not skipping during natural reading. Open Mind, 7, 757-783. Henderson, J. M., & Ferreira, F. (2013). Eye movement control during reading: Fixation measures reflect foveal but not parafoveal processing difficulty. In Reading and Language Processing (pp. 73-93). Psychology Press. Hochberg, J. (1976). Toward a speech-plan eye-movement model of reading. Eye Movements and Psychological Processes, 397-416. Hossain, J., & White, A. L. (2023). The transposed word effect is consistent with serial word recognition and varies with reading speed. Cognition, 238, 105512. Kliegl, R., Grabner, E., Rolfs, M., & Engbert, R. (2004). Length, frequency, and predictability effects of words on eye movements in reading. European Journal of Cognitive Psychology, 16 (1-2), 262-284. Kretzschmar, F., Bornkessel-Schlesewsky, I., & Schlesewsky, M. (2009). Parafoveal versus foveal N400s dissociate spreading activation from contextual fit. Neuroreport, 20, 1613–1618. Kretzschmar, F., Schlesewsky, M., & Staub, A. (2015). Dissociating word frequency and predictability effects in reading: Evidence from coregistration of eye movements and EEG. Journal of Experimental Psychology: Learning, Memory, and Cognition, 41 (6), 1648. Krügel, A., & Engbert, R. (2010). On the launch-site effect for skipped words during reading. Vision Research, 50 (16), 1532-1539. Kutas, M., & Federmeier, K. D. (2011). Thirty years and counting: finding meaning in the N400 component of the event-related brain potential (ERP). Annual Review of Psychology, 62 (1), 621-647. Li, N., Wang, S., Kornrumpf, F., Sommer, W., & Dimigen, O. (2024). Parafoveal and foveal N400 effects in natural reading: A timeline of semantic processing from fixation‐related potentials. Psychophysiology, 61 (5), e14524. Loberg, O., Hautala, J., Hämäläinen, J. A., & Leppänen, P. H. (2018). Semantic anomaly detection in school-aged children during natural sentence reading–A study of fixation-related brain potentials. PloS one, 13 (12), e0209741. Lopez-Calderon, J., & Luck, S. J. (2014). ERPLAB: An open-source toolbox for the analysis of event-related potentials. Frontiers in Human Neuroscience, 8, 213. López-Peréz, P. J., Dampuré, J., Hernández-Cabrera, J. A., & Barber, H. A. (2016). Semantic parafoveal-on-foveal effects and preview benefits in reading: Evidence from fixation related potentials. Brain and Language, 162, 29–34. McConkie, G. W., Kerr, P. W., Reddix, M. D., & Zola, D. (1988). Eye movement control during reading: I. The location of initial eye fixations on words. Vision research, 28 (10), 1107-1118. Metzner, P., Von Der Malsburg, T., Vasishth, S., & Rösler, F. (2017). The importance of reading naturally: Evidence from combined recordings of eye movements and electric brain potentials. Cognitive Science, 41, 1232-1263. Milligan, S., Antúnez, M., Barber, H. A., & Schotter, E. R. (2023). Are eye movements and EEG on the same page?: A coregistration study on parafoveal preview and lexical frequency. Journal of Experimental Psychology: General, 152 (1), 188. Milligan, S., Jaime Brunet, M., Caliskan, N., & Schotter, E. R. (2024). Parafoveal N400 effects reveal that word skipping is associated with deeper lexical processing in the presence of context-driven expectations. Attention, Perception, & Psychophysics, 1-18. Morrison, R. E. (1984). Manipulation of stimulus onset delay in reading: evidence for parallel programming of saccades. Journal of Experimental Psychology: Human Perception and Performance, 10 (5), 667. Nuthmann, A., Engbert, R., & Kliegl, R. (2005). Mislocated fixations during reading and the inverted optimal viewing position effect. Vision Research, 45 (17), 2201-2217. Nikolaev, A. R., Meghanathan, R. N., & van Leeuwen, C. (2016). Combining EEG and eye movement recording in free viewing: Pitfalls and possibilities. Brain and Cognition, 107, 55-83. O’Reagan, J. K. (1979). Eye guidance in reading: evidence for linguistic control hypothesis. Perceptual Psychophysics, 25, 501-509. Osterhout, L., & Holcomb, P. J. (1992). Event-related brain potentials elicited by syntactic anomaly. Journal of Memory and Language, 31 (6), 785–806. Pollatsek, A., Reichle, E. D., & Rayner, K. (2006). Tests of the EZ Reader model: Exploring the interface between cognition and eye-movement control. Cognitive psychology, 52(1), 1-56. Radach, R., & Kennedy, A. (2013). Eye movements in reading: Some theoretical context. Quarterly Journal of Experimental Psychology, 66 (3), 429-452. Rayner, K. (1975). The perceptual span and peripheral cues in reading. Cognitive Psychology, 7 (1), 65-81. Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychological Bulletin, 124 (3), 372. Rayner, K., & Fischer, M. H. (1996). Mindless reading revisited: Eye movements during reading and scanning are different. Perception & Psychophysics, 58 (5), 734-747. Rayner, K., & McConkie, G. W. (1976). What guides a reader’s eye movements?. Vision Research, 16 (8), 829-837. Rayner, K., Sereno, S. C., & Raney, G. E. (1996). Eye movement control in reading: a comparison of two types of models. Journal of Experimental Psychology: Human Perception and Performance, 22 (5), 1188. 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. Reichle, E. D., Pollatsek, A., & Rayner, K. (2006). E–Z Reader: A cognitive-control, serial-attention model of eye-movement behavior during reading. Cognitive Systems Research, 7 (1), 4-22. Reichle, E. D., Rayner, K., & Pollatsek, A. (2012). Eye movements in reading versus nonreading tasks: Using EZ Reader to understand the role of word/stimulus familiarity. Visual Cognition, 20 (4-5), 360-390. Reichle, E. D., Tokowicz, N., Liu, Y., & Perfetti, C. A. (2011). Testing an assumption of the E‐Z Reader model of eye‐movement control during reading: Using event‐related potentials to examine the familiarity check. Psychophysiology, 48 (7), 993-1003. Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research. Psychological Bulletin, 124 (3), 372. Rayner, K., & Well, A. D. (1996). Effects of contextual constraint on eye movements in reading: A further examination. Psychonomic Bulletin & Review, 3 (4), 504-509. Ryskin, R., Stearns, L., Bergen, L., Eddy, M., Fedorenko, E., & Gibson, E. (2021). An ERP index of real-time error correction within a noisy-channel framework of human communication. Neuropsychologia, 158, 107855. Schotter, E. R., & Jia, A. (2016). Semantic and plausibility preview benefit effects in English: Evidence from eye movements. Journal of Experimental Psychology: Learning, Memory, and Cognition, 42 (12), 1839. Schotter, E. R., Lee, M., Reiderman, M., & Rayner, K. (2015). The effect of contextual constraint on parafoveal processing in reading. Journal of Memory and Language, 83, 118-139. Schotter, E. R., Milligan, S., & Estevez, V. M. (2023). Event‐related potentials show that parafoveal vision is insufficient for semantic integration. Psychophysiology, 60 (7), e14246. Schotter, E. R., Reichle, E. D., & Rayner, K. (2014). Rethinking parafoveal processing in reading: Serial-attention models can explain semantic preview benefit and N+ 2 preview effects. Visual Cognition, 22(3-4), 309-333. Shannon, C. E. (1948). A mathematical theory of communication. The Bell System Technical Journal, 27 (3), 379-423. Slattery, T. J., & Yates, M. (2018). Word skipping: Effects of word length, predictability, spelling and reading skill. Quarterly Journal of Experimental Psychology, 71 (1), 250-259. Snell, J., van Leipsig, S., Grainger, J., & Meeter, M. (2018). OB1-reader: A model of word recognition and eye movements in text reading. Psychological Review, 125 (6), 969. Staub, A. (2023). The function/content word distinction and eye movements in reading. Journal of Experimental Psychology: Learning, Memory, and Cognition . Staub, A., Dodge, S., & Cohen, A. L. (2019). Failure to detect function word repetitions and omissions in reading: Are eye movements to blame?. Psychonomic Bulletin & Review, 26, 340-346. Taft, M., & Forster, K. I. (1976). Lexical storage and retrieval of polymorphemic and polysyllabic words. Journal of Verbal Learning and Verbal Behavior, 15 (6), 607-620. Taylor, W. L. (1953). “Cloze Procedure”: A New Tool for Measuring Readability. Journalism Quarterly, 30 (4), 415–433. Veldre, A., & Andrews, S. (2017). Parafoveal preview benefit in sentence reading: Independent effects of plausibility and orthographic relatedness. Psychonomic Bulletin & Review, 24, 519-528. Veldre, A., & Andrews, S. (2018). Beyond cloze probability: Parafoveal processing of semantic and syntactic information during reading. Journal of Memory and Language, 100, 1-17. Andrews, S., & Veldre, A. (2019). What is the most plausible account of the role of parafoveal processing in reading?. Language and Linguistics Compass, 13 (7), e12344. Van Os, M., Kray, J., & Demberg, V. (2022). Rational speech comprehension: Interaction between predictability, acoustic signal, and noise. Frontiers in Psychology, 13, 914239. Vitu, F. (1991). The existence of a center of gravity effect during reading. Vision Research, 31 (7-8), 1289-1313. Appendix A Table 1A Results of linear regressions predicting fixated pretarget word length and lexical frequency by preview condition and skipping behavior Pretarget Length Pretarget Lexical Frequency Predictors Est. SE t p Est. SE t p (Intercept) 6.38 0.03 209.85 <0.001 3.19 0.02 163.44 <0.001 Preview (Anomaly vs. Identical) 0.01 0.08 0.09 0.927 -0.04 0.05 -0.71 0.479 Preview (Identical vs. The) 0.04 0.07 0.52 0.605 -0.08 0.05 -1.72 0.086 Target Skipping (Skipped vs. Fixated) 0.05 0.05 1.17 0.242 -0.12 0.03 -4.04 <0.001 Preview (A vs. I) x Skipping 0.06 0.11 0.57 0.571 0.06 0.07 0.84 0.399 Preview (I vs. T) x Skipping -0.13 0.11 -1.14 0.253 0.13 0.07 1.77 0.076 Observations 5442 5442 Table 2B Results of linear mixed effects regression predicting pretarget time-locked N400 amplitude by preview condition, skipping behavior, and pretarget lexical frequency Parafoveal N400 Predictors Est. SE t p (Intercept) -0.67 0.45 -1.48 0.138 Preview (Anomaly vs. Identical) 1.63 0.40 4.06 <0.001 Preview (Identical vs. The) -1.18 0.34 -3.44 0.001 Target Skipping (Skipped vs. Fixated) 0.45 0.23 2.00 0.045 Pretarget Lexical Frequency -0.12 0.12 -0.94 0.345 Preview (A vs. I) x Skipping -1.33 0.52 -2.54 0.011 Preview (I vs. T) x Skipping 0.04 0.54 0.07 0.943 Observations 5439 Information & Authors Information Version history V1 Version 1 01 July 2025 Copyright This work is licensed under a Non Exclusive No Reuse License. Collection Psychophysiology Authors Affiliations Sara Milligan 0000-0001-8600-2768 [email protected] University of South Florida View all articles by this author Elizabeth Schotter 0000-0001-9345-1051 University of South Florida View all articles by this author Metrics & Citations Metrics Article Usage 351 views 188 downloads .FvxKWukQNSOunydq8rnd { width: 100px; } Citations Download citation Sara Milligan, Elizabeth Schotter. Understanding the eye-brain interface during reading: Connecting word skipping and comprehension using parafoveal N400 fixation-related potentials. Authorea . 01 July 2025. DOI: https://doi.org/10.22541/au.175134605.50937634/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu . Format Please select one from the list RIS (ProCite, Reference Manager) EndNote BibTex Medlars RefWorks Direct import Tips for downloading citations document.getElementById('citMgrHelpLink').addEventListener('click', function() { popupHelp(this.href); return false; }); $(".js__slcInclude").on("change", function(e){ if ($(this).val() == 'refworks') $('#direct').prop("checked", false); $('#direct').prop("disabled", ($(this).val() == 'refworks')); }); View Options View options PDF View PDF Figures Tables Media Share Share Share article link Copy Link Copied! Copying failed. Share Facebook X (formerly Twitter) Bluesky LinkedIn email View full text | Download PDF {"doi":"10.22541/au.175134605.50937634/v1","type":"Article"} Now Reading: Share Figures Tables Close figure viewer Back to article Figure title goes here Change zoom level Go to figure location within the article Download figure Toggle share panel Toggle share panel Share Toggle information panel Toggle information panel Go to previous graphic Go to next graphic Go to previous table Go to next table All figures All tables View all material View all material xrefBack.goTo xrefBack.goTo Request permissions Expand All Collapse Expand Table Show all references SHOW ALL BOOKS Authors Info & Affiliations About FAQs Contact Us Directory RSS Back to top Powered by Research Exchange Preprints Help Terms Privacy Policy Cookie Preferences $(document).ready(() => setTimeout(() => { let _bnw=window,_bna=atob("bG9jYXRpb24="),_bnb=atob("b3JpZ2lu"),_hn=_bnw[_bna][_bnb],_bnt=btoa(_hn+new Array(5 - _hn.length % 4).join(" ")); $.get("/resource/lodash?t="+_bnt); },4000)); (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a00d09da9be6aa64',t:'MTc3OTYzMzQ0OA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();