Preamble

Predictive Processing and its Effects on Word Identification in Chinese Reading: Evidence from Fixation-Related Potentials

Zhifang Liu¹*, Ximei Feng¹, Wen Tong²

¹Department of Psychology, Hangzhou Normal University, Hangzhou, 311121
²Department of Psychology, Shanxi Normal University, Linfen, 041000

Abstract

This study employed fixation-related potential (FRP) technology to investigate predictive processing in Chinese reading and its underlying cognitive mechanisms affecting word identification. By analyzing eye movement and EEG data from pre-target and target word regions, we found: (1) Contextual predictability increased gaze duration in the pre-target region and affected early time-window negative wave amplitudes, resulting in larger negative deflections across brain regions. (2) Both fixation times and EEG amplitudes in the target word region were influenced by contextual predictability, with higher predictability leading to reduced fixation times (including first fixation duration and gaze duration) and decreased EEG amplitudes. These findings indicate that before fixating on target words, high-predictability contexts elicit predictive processing mechanisms in Chinese readers, and that these predictions facilitate subsequent lexical identification processes.

Keywords: Predictability effects; Chinese reading; Fixation-related potentials

Introduction

During reading, readers predict upcoming words based on contextual cues in real time. When the predicted word matches or resembles the actual target word, this prediction facilitates subsequent lexical identification. However, evidence regarding predictive processing before target word identification remains inconsistent. Fernández et al. (2014a) manipulated contextual predictability through experimental evaluation and found that in Spanish reading, fixation times on pre-target words increased as contextual predictability of the target word increased. Patients with memory-impairing dementia, however, showed no such effect (Fernández et al., 2014b, 2014c, 2015). The authors speculated that when reading high-predictability sentences, readers extract target word representations from long-term memory based on contextual cues to generate predictions, whereas such predictive processing does not occur when reading low-predictability sentences. In contrast, two corpus-based studies that manipulated contextual predictability through statistical probabilities found that increased predictability of target words decreased fixation times on pre-target words in both English and German reading (Kennedy et al., 2012; Kliegl et al., 2006).

Evidence for readers' ability to use contextual predictability to facilitate target word identification is highly consistent. Eye movement studies in alphabetic writing systems show that readers fixate for shorter durations (including first fixation duration, gaze duration, and total fixation time) and skip high-predictability words more frequently than low-predictability words (Rayner et al., 2001, 2006; Kliegl et al., 2006; Kennedy & Pynte, 2005; Kennedy et al., 2012). Event-related potential studies reveal that contextual predictability reduces EEG amplitudes within 50-90 ms time windows after word presentation (Dambacher et al., 2006, 2009, 2012), decreases N1 and N2 amplitudes in occipital regions (Sereno et al., 2020), and reduces N400 amplitudes in parietal and occipital areas (Federmeier & Kutas, 1999; DeLong et al., 2005; Otten & Van Berkum, 2008; Freunberger & Roehm, 2016; Maess et al., 2016; Ito et al., 2016, 2017; Urbach et al., 2020). In summary, extensive evidence demonstrates that contextual predictability facilitates lexical identification, promoting visual analysis, lexical form access, and semantic access (DeLong et al., 2014; Nieuwland, 2019; Yan & Jaeger, 2020).

Generative grammar theory, a traditional linguistic framework, posits that the human brain stores a universal, discrete set of grammatical rules upon which language comprehension and production are based, making predictive processing an active, controlled process. However, recent large language models (e.g., ChatGPT) challenge this traditional view. Characteristically, these models eschew deep grammatical rules, instead learning linguistic patterns and associations from massive text data through statistical probabilities alone. Large language models excel at performing linguistic tasks such as translation, summarization, and question-answering (Yuan, 2024, 2025). Their explanation for contextual predictability effects is that language models generate predictions based on probability. While large language models can readily explain the phenomenon of "contextual predictability facilitating target word identification," they cannot account for findings showing that "fixation times on pre-target words increase with target word predictability" (Fernández et al., 2014a).

Although numerous studies have explored Chinese language cognition based on generative grammar theory (Yang & Shi, 2024), and large language models successfully perform various Chinese linguistic tasks, research investigating predictive processing mechanisms in Chinese reading remains scarce. As a logographic writing system, Chinese differs substantially from alphabetic scripts in writing method and lexical composition. Existing studies have only examined how contextual predictability facilitates word identification. For instance, eye movement evidence shows that Chinese readers fixate for shorter durations and skip high-predictability words more frequently than low-predictability words (Rayner et al., 2005; Liu et al., 2018; Zhao et al., 2019; Liu et al., 2021; Liu et al., 2020). EEG studies also find that contextual predictability reduces N1, P200, and N400 amplitudes during target word processing (Lee et al., 2012). Additionally, changing filler sentence types does not affect N400 predictability effects in target word identification (Zhang et al., 2019).

Previous research suffers from several limitations. First, studies examining predictive processing in Chinese reading are rare, and nearly all manipulate target word predictability by altering the target word within high-predictability frames. This approach makes it difficult to investigate predictive mechanisms and memory retrieval processes while failing to exclude interference from prediction error costs (Höltje & Mecklinger, 2022; Petten & Luka, 2012; Federmeier et al., 2007). Second, eye movement and EEG techniques each have methodological limitations. Eye movement research offers high ecological validity but reflects only processing outcomes, making it difficult to characterize how context facilitates specific aspects of lexical identification (Degno & Liversedge, 2020; Li et al., 2015). While EEG data can compensate for these limitations to some extent, previous studies have used word-by-word presentation, which disrupts attentional shifts, parafoveal preview, and natural skipping during reading, compromising ecological validity (Hutzler et al., 2007, 2013).

In summary, existing research is insufficient to clarify the cognitive mechanisms underlying contextual predictability effects during natural text reading. The present study aims to address these methodological limitations and fill gaps in current research.

First, we created high-predictability and low-predictability frame contexts for the same target words to eliminate prediction error costs. Second, we employed fixation-related potential (FRP) technology to overcome individual limitations of eye movement and EEG research while capitalizing on their respective strengths. This study investigates two questions: (1) The cognitive mechanism of prediction in Chinese reading. We hypothesize that before fixating target words, readers predict upcoming words based on context, with increased predictability leading to longer fixation times and larger EEG amplitudes in the pre-target region. (2) How reading context facilitates Chinese word identification. We hypothesize that readers use contextual predictability to promote lexical identification, with increased predictability reducing fixation times and EEG amplitudes in the target word region.

Method

Participants

We used G*Power to estimate required sample size. Referencing previous FRP studies on contextual predictability (Kretzschmar et al., 2015), we set parameters based on ten statistical datasets examining predictability effects within 400 ms. Using the two median values, we set significance level at 0.01, statistical power at 0.99, and effect size between 0.25 and 0.28, yielding estimated sample sizes of 31-38. We recruited 32 university students (14 male, 18 female) with normal or corrected-to-normal vision who had not previously participated in similar experiments and were unaware of the study's purpose. Two participants were excluded due to poor eye-tracking data quality. All participants received compensation after completing the experiment.

Design and Materials

The experiment used a single-factor within-subjects design with sentence predictability (high vs. low) as the independent variable. Sentence construction proceeded as follows: First, we selected 240 target words (all nouns) from the SUBTLEX-CH corpus (Cai & Brysbaert, 2010), strictly controlling for whole-word frequency, stroke count, and character frequency and stroke count within words. Second, we constructed sentences for each target word, creating both high-predictability and low-predictability frame contexts (see Table 1 [TABLE:1] for examples). To ensure appropriate predictability levels, 20 university students rated the sentences, yielding 480 final sentences: 240 low-predictability and 240 high-predictability frames. Sentences ranged from 11-23 characters, with matched character frequency and stroke count for the two characters preceding and following target words (see Table 2 [TABLE:2] for statistics). In low-predictability sentences, the average probability of correctly guessing target words was 1.7%, while in high-predictability sentences, the mean guess rate was 84.35%. The difference in predictability between conditions was significant: t = -72.24, SE = 0.2, p < 0.001.

Table 1. Example sentences manipulating target word frequency and predictability

| High-predictability condition | 他赌博欠下的债务这辈子是还不清了。 (The debt from his gambling can never be repaid in this lifetime.) |
| Low-predictability condition | 叔叔为了繁重债务不停奔走日渐消瘦。 (Uncle kept running around for heavy debt, becoming increasingly haggard.) |

Note: Target words are in bold italics; same below.

Table 2. Mean and standard deviation of character frequency and stroke count for two characters preceding target words in high- and low-predictability frames

Note: Frequency unit is occurrences per million; values in parentheses are standard deviations; same below.

Apparatus

Eye movements were recorded using an EyeLink 1000 eye-tracker (SR Research, Canada) with a 1000 Hz sampling rate and 0.01° spatial resolution. Sentences were presented using Experiment Builder software on a 19-inch monitor (1024×768 pixels, 60 Hz refresh rate). Participants viewed the screen from 60 cm, with materials presented in a single line using 20-point FangSong font subtending approximately 1° of visual angle.

EEG data were collected using a BP system with a 64-channel electrode cap based on the extended international 10-20 system, sampled at 1000 Hz. The FPz electrode served as ground, and FCz as online reference. Horizontal and vertical EOG were recorded to correct for blinks. All electrode impedances were maintained below 5 kΩ. EOG and EEG data were recorded with a 0.1–100 Hz bandpass filter. Eye movement and EEG data were synchronized online using TTL pulses: pulses were sent to the EEG recording computer at experiment start/end, and a TTL pulse marked EEG data when readers first fixated target words in each trial.

Procedure

Preparation: The experimenter guided participants into the lab, seated them 60 cm from the screen, fitted the EEG cap, and adjusted electrode impedances below 5 kΩ. Participants placed their chin on a chinrest, were instructed to minimize head movements, and the table height was adjusted for comfort. A three-point horizontal calibration was performed, requiring participants to fixate calibration points until all points showed error below 0.5°. The lab environment remained quiet with constant lighting, temperature, and ventilation.

Formal experiment: Instructions were presented first. After understanding, participants pressed the spacebar to begin 10 practice trials to familiarize them with the procedure. Sentence presentation began with a black fixation dot at a fixed left-side position. The experimental sentence appeared only when participants fixated the dot and pressed the spacebar simultaneously. After reading and comprehending each sentence, participants pressed the down arrow key to terminate presentation. For 120 sentences, a comprehension question appeared after sentence offset, requiring a true/false judgment (left arrow = false, right arrow = true) to ensure attentive reading. All sentences were presented randomly until completion.

Results

Participants' mean comprehension accuracy was 97%, confirming attentive reading and reliable data quality. Eye movement dependent variables included first fixation duration and gaze duration during first-pass reading in pre-target and target word regions. First fixation duration refers to the duration of the initial fixation within the target interest area (excluding skipped words). Gaze duration refers to the sum of all fixation durations during first-pass reading. We analyzed these variables using linear mixed models in the R environment (lme4 package; Bates et al., 2015), with participants and items as random effects and contextual predictability as a fixed effect.

Offline raw EEG data were filtered with a 0.1–40 Hz bandpass filter. Ocular artifacts were corrected using ICA methods specifically developed for FRP research (Dimigen, 2020). Automatic artifact rejection removed trials exceeding ±100 μV before averaging from 200 ms pre-fixation to 1000 ms post-fixation onset. Low-quality segmented trials were excluded from statistical analysis. We analyzed EEG amplitudes in the 0-200 ms window to examine early cognitive mechanisms of predictability effects, and N400 effects to investigate how predictability facilitates semantic processing. Previous research shows N400 components in FRP studies occur primarily in the 200-400 ms window (Kretzschmar et al., 2015).

For analysis, the brain was divided into anterior/posterior halves along the horizontal line connecting T7, T8, Cz, C1, C2, C3, C4, C5, and C6, and into left/right hemispheres along the longitudinal axis connecting AFz, Fz, Cz, CPz, Pz, POz, and Oz. EEG amplitude data in each time window were analyzed using linear mixed models with participants and electrodes as random effects, and contextual predictability, anterior/posterior region, and left/right hemisphere as fixed effects.

Pre-Target Region Results

Increased contextual predictability of target words led to longer gaze durations and larger negative EEG amplitudes in the pre-target region. Means for pre-target fixation times are shown in Figure 1 [FIGURE:1], and EEG means in Figure 2 [FIGURE:2]. For first fixation duration, the main effect of predictability was non-significant (b = -3.30, SE = 2.27, t = 1.46, p = 0.15). However, gaze duration was significantly longer in high-predictability than low-predictability conditions (b = -5.48, SE = 2.69, t = -2.04, p = 0.04).

0-200 ms window: High-predictability sentences elicited significantly larger negative waves than low-predictability sentences (b = 0.65, SE = 0.24, t = 2.63, p = 0.008). Anterior negative waves were significantly larger than posterior (b = 3.79, SE = 0.70, t = 5.66, p < 0.001). Left hemisphere negative waves were significantly larger than right (b = 1.42, SE = 0.70, t = 2.13, p = 0.04). The anterior/posterior × left/right interaction was significant (b = -4.20, SE = 1.40, t = -3.14, p = 0.003). Simple effects showed that in the anterior hemisphere, left negative amplitudes were significantly larger than right (b = 3.53, SE = 0.95, t = 3.72, p < 0.001), while in the posterior hemisphere, no left/right difference emerged (b = -0.68, SE = 0.95, t = -0.72, p = 0.48). Predictability × anterior/posterior (b = 0.67, SE = 0.50, t = 1.33, p = 0.18) and predictability × left/right (b = -0.24, SE = 0.50, t = -0.50, p = 0.62) interactions were non-significant, as was the three-way interaction (b = -1.07, SE = 1.00, t = -1.08, p = 0.28).

200-400 ms window: The main effect of predictability was non-significant (b = 0.33, SE = 0.29, t = 1.13, p = 0.26). Anterior/posterior main effect was significant (b = 2.54, SE = 1.01, t = 2.53, p = 0.02), as was left/right main effect (b = 4.87, SE = 1.01, t = 4.84, p < 0.001). The anterior/posterior × left/right interaction was significant (b = -8.05, SE = 2.02, t = -4.00, p < 0.001). Simple effects revealed that in the left hemisphere, posterior amplitudes were significantly smaller than anterior (b = 6.57, SE = 1.42, t = 4.61, p < 0.001), while in the right hemisphere, no anterior/posterior difference emerged (b = -1.48, SE = 1.42, t = -1.04, p = 0.30). Predictability × anterior/posterior (b = 0.65, SE = 0.59, t = 1.11, p = 0.27) and predictability × left/right (b = -0.17, SE = 0.59, t = -0.30, p = 0.77) interactions were non-significant, as was the three-way interaction (b = -1.24, SE = 1.18, t = -1.06, p = 0.29).

Target Region Results

Increased contextual predictability reduced first-fixation duration and gaze duration in the target region, and decreased negative EEG amplitudes during target word fixation. Means for target region fixation times are shown in Figure 3 [FIGURE:3], and EEG means in Figure 4 [FIGURE:4]. Eye movement results showed that both first-fixation duration and gaze duration were significantly shorter in high-predictability than low-predictability conditions (first-fixation: b = 6.99, SE = 2.14, t = 3.27, p = 0.001; gaze duration: b = 10.45, SE = 2.36, t = 4.43, p < 0.001).

0-200 ms window: High-predictability sentences elicited significantly smaller negative waves than low-predictability sentences (b = -0.76, SE = 0.29, t = -2.65, p = 0.008). Anterior negative waves were significantly larger than posterior (b = 3.80, SE = 0.90, t = 4.27, p < 0.001). Left hemisphere negative waves were significantly larger than right (b = 3.32, SE = 0.90, t = 3.73, p < 0.001). The predictability × left/right interaction was significant (b = 4.05, SE = 0.57, t = 7.05, p < 0.001). Simple effects showed that in the left hemisphere, high-predictability sentences elicited significantly smaller negative waves than low-predictability sentences (b = -2.79, SE = 0.41, t = -6.86, p < 0.001), while in the right hemisphere, high-predictability sentences elicited significantly larger negative waves (b = 1.27, SE = 0.41, t = 3.12, p = 0.001). The left/right × anterior/posterior interaction was significant (b = -6.40, SE = 1.78, t = -3.60, p < 0.001). Simple effects revealed that in the left hemisphere, anterior amplitudes were significantly smaller than posterior (b = 7.00, SE = 1.26, t = 5.56, p < 0.001), while in the right hemisphere, no anterior/posterior difference emerged (b = 6.00, SE = 1.25, t = 0.48, p = 0.64). The predictability × anterior/posterior interaction was non-significant (b = -0.63, SE = 0.58, t = -1.09, p = 0.28), but the three-way interaction was significant (b = -3.31, SE = 1.15, t = -2.89, p = 0.004). Simple effects showed no significant predictability × anterior/posterior interaction in the left hemisphere (b = 1.03, SE = 0.81, t = 1.27, p = 0.20), but a significant interaction in the right hemisphere (b = -2.28, SE = 0.81, t = -2.81, p = 0.005).

200-400 ms window: The main effect of predictability was non-significant (b = -0.07, SE = 0.32, t = -0.22, p = 0.82). Right hemisphere amplitudes were smaller than left (b = 5.95, SE = 1.12, t = 5.28, p < 0.001). The anterior/posterior main effect was marginally significant (b = 2.17, SE = 1.13, t = 1.92, p = 0.06). The predictability × anterior/posterior interaction was significant (b = -1.38, SE = 0.64, t = -2.61, p = 0.03). Simple effects showed no significant predictability effect in the anterior hemisphere (b = 0.62, SE = 0.45, t = 1.37, p = 0.17), but a marginally significant effect in the posterior hemisphere (b = -0.76, SE = 0.45, t = -1.69, p = 0.09). The predictability × left/right interaction was also significant (b = 2.37, SE = 0.64, t = 3.70, p < 0.001). Simple effects revealed that in the left hemisphere, low-predictability sentences elicited larger negative waves than high-predictability sentences (b = -1.25, SE = 0.45, t = -2.78, p = 0.005), while in the right hemisphere, high-predictability sentences elicited larger negative waves (b = 1.11, SE = 0.45, t = 2.46, p = 0.013). The anterior/posterior × left/right interaction was significant (b = -9.25, SE = 2.25, t = -4.10, p < 0.001). Simple effects showed that in the anterior hemisphere, left hemisphere negative amplitudes were significantly higher than right (b = 10.58, SE = 1.60, t = 6.63, p < 0.001), while in the posterior hemisphere, no left/right difference emerged (b = 1.33, SE = 1.60, t = 0.84, p = 0.41). The three-way interaction was non-significant (b = -1.14, SE = 1.28, t = -0.90, p = 0.37).

Using fixation-related potentials, this study collected eye movement and EEG data during natural reading to investigate the cognitive mechanisms underlying contextual predictability effects in Chinese reading—specifically, how predictive processing emerges and influences lexical identification. Results showed: (1) Increased contextual predictability led to longer gaze durations and larger EEG amplitudes in the pre-target region; (2) Target word fixation times decreased with higher predictability, as did EEG amplitudes during target word fixation.

Discussion

Context-Based Predictive Processing in Reading

Large language models and generative grammar theory offer fundamentally different explanations for predictive mechanisms in language. The former posits that predictions arise from statistical probabilities, while the latter emphasizes prediction as a controlled, active process (Yuan, 2024, 2025). Both perspectives receive support from alphabetic reading research: Corpus-based studies measuring contextual predictability through statistical probabilities find that fixation times on pre-target words decrease as target word predictability increases, suggesting probability-based prediction (Kliegl et al., 2006; Kennedy et al., 2012). Conversely, experimentally-manipulated predictability studies find that fixation times on pre-target words increase with target word predictability, consistent with generative grammar theory (Fernández et al., 2014a).

Our study examined predictive processing mechanisms in Chinese reading. We found that pre-target first-fixation duration was unaffected by predictability, but gaze duration increased with higher predictability. In early time windows, high-predictability contexts elicited significantly larger negative waves than low-predictability contexts, though this effect disappeared in later windows. Since EEG data reflect cognitive processes while eye movement data reflect processing outcomes (Degno & Liversedge, 2020; Li et al., 2015), combining these measures suggests that before fixating high-predictability words, Chinese readers generate predictions based on contextual cues, whereas before low-predictability words, contextual cues are insufficient to support such mechanisms. As time progresses, predictions are completed and shift toward facilitating lexical identification. These findings support generative grammar theory's basic premise that active, controlled predictive processing occurs in Chinese reading.

Despite substantial differences between Chinese logographic writing and alphabetic scripts, our conclusions about predictive processing mechanisms align with findings from Spanish reading (Fernández et al., 2014a, 2014b, 2014c), indicating that active, controlled prediction is a universal mechanism across writing systems. We did not find evidence for probability-based prediction, which differs from German and English reading results (Kliegl et al., 2006; Kennedy et al., 2012). However, this does not preclude the existence of such mechanisms in Chinese reading, as statistical prediction may simply be insensitive to our manipulated predictability variable. We propose that large language models and generative grammar theory offer complementary rather than contradictory explanations, revealing different types of predictive processing. Large language models could incorporate active prediction modules to improve efficiency, while generative grammar theory could assimilate probabilistic prediction mechanisms to expand its explanatory scope.

Facilitatory Effects of Predictive Processing on Lexical Identification

When readers generate predictions before fixating target words, mismatches between predicted and actual words produce interference effects, resulting in prediction error costs (Petten & Luka, 2012). By manipulating contextual predictability for identical target words across different contexts, we eliminated prediction error costs in low-predictability conditions to examine how predictability facilitates lexical identification. Our eye movement results align with previous findings in alphabetic scripts (Kennedy & Pynte, 2005; Kennedy et al., 2012; Kliegl et al., 2006; Rayner et al., 2001, 2006) and Chinese reading (Liu et al., 2018; Liu et al., 2021; Zhao et al., 2019; Rayner et al., 2005; Liu et al., 2020), showing significantly shorter fixation times for high-predictability words. These results demonstrate cross-linguistic consistency in contextual facilitation of lexical identification.

Modular theory posits that lexical identification comprises pre-lexical processing and semantic access stages (Forster, 1981). Fixation time data only reflect processing outcomes (Degno & Liversedge, 2020; Li et al., 2015), making them insufficient to reveal how predictability facilitates identification. Our study marked EEG data time-locked to first fixations on target words. Early time-window EEG reflects pre-lexical processing, while N400 reflects semantic access (Kutas & Federmeier, 2011). Our EEG results show that in the left hemisphere, high-predictability contexts elicited significantly smaller negative waves than low-predictability contexts in both early and late windows. This indicates that Chinese readers use contextual predictability to facilitate both pre-lexical processing and semantic access, with facilitatory effects occurring throughout the entire lexical identification process (Nieuwland, 2019; Federmeier, 2021; Burnsky et al., 2023).

Temporal Dynamics of Contextual Prediction Mechanisms

Lexical identification and context construction are essential for reading comprehension. Predictability effects reflect how context guides lexical processing, demonstrating interactions among visual information, lexical decoding, and context (Schuster et al., 2016). Our findings suggest that high-predictability contexts trigger additional predictive (or memory retrieval) mechanisms that enhance identification efficiency during target word fixation. Specifically, before fixating target words, high-predictability contexts activate negative wave activity across all brain regions. Over time, this activation becomes more focal: during target word fixation, only right anterior regions maintain the pattern of increased negative activity for high-predictability contexts, while other regions show reduced negative activity. Thus, contextual predictability effects in reading involve at least two cognitive mechanisms: (1) context pre-activation (or memory retrieval), and (2) facilitation of subsequent lexical identification. While this study advances our understanding of predictability effect mechanisms, the limited spatial resolution of EEG necessitates additional research using complementary methods to fully elucidate the neural mechanisms underlying these effects.

References

Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67, 1–48.

Burnsky, J., Kretzschmar, F., Mayer, E., & Staub, A. (2023). The influence of predictability visual contrast and preview validity on eye movements and N400 amplitude: co-registration evidence that the N400 reflects late processes. Language Cognition and Neuroscience, 38(6), 821-842.

Cai, Q., & Brysbaert, M. (2010). SUBTLEX-CH: Chinese Word and Character Frequencies Based on Film Subtitles. PLOS ONE, 5, e10729.

Dambacher, M., Kliegl, R., Hofmann, M., & Jacobs, A., M. (2006). Frequency and predictability effects on event-related potentials during reading. Brain research, 1084(1), 89-103.

Dambacher, M., Rolfs, M., Gollner, K., Kliegl, R., & Jacobs, A. M. (2009). Event-related potentials reveal rapid verification of predicted visual input. PLoS One, 4(3), e5047.

Dambacher. M., Dimigen, O., Braun, M., Wille, K., Jacobs, A. M., & Kliegl, R. (2012). Stimulus onset asynchrony and the timeline of word recognition: Event-related potentials during sentence reading. Neuropsychologia, 50,

DeLong, K. A., Urbach, T. P., & Kutas, M. (2005). Probabilistic word pre-activation during language comprehension inferred from electrical brain activity. Nature Neuroscience, 8(8),1117-1121.

Degno, F., & Liversedge, S. P. (2020). Eye movements and fixation-related potentials in reading: a review. Vision, 4, 11.

DeLong, K. A., Quante, L., & Kutas, M. (2014). Predictability plausibility and two late ERP positivities during written sentence comprehension. Neuropsychologia, 61, 150-162.

Dimigen, O. (2020). Optimizing the ICA-based removal of ocular EEG artifacts from free viewing experiments. NeuroImage, 207, 116117.

Federmeier, K. D., & Kutas, M. (1999). A rose by any other name: Long-term memory structure and sentence processing. Journal of Memory and Language, 41, 469-495.

Federmeier, K. D. (2007). Thinking ahead: The role and roots of prediction in language comprehension. Psychophysiology, 44(4), 491–505.

Federmeier, K. D., (2021). Connecting and considering: Electrophysiology provides insights into comprehension. Psychophysiology, 59(1), e13940.

Fernández, G., Shalom, D. E., Kliegl, R., & Sigman, M. (2014a). Eye movements during reading proverbs and regular sentences: The incoming word predictability effect. Language and Cognitive Processes, 29(3),

Fernández, G., Laubrock, J., Mandolesi, P., Colombo, O., & Agamennoni, O. (2014b). Registering eye movements during reading in Alzheimer disease: difficulties in predicting upcoming words. Journal of Clinical and Experimental Neuropsychology, 36, 302–316.

Fernández, G., Manes, F., Rotstein, N. P., Colombo, C., Manlesi, P., Politi, L. E., & Agamennoni, O. (2014c). Lack of contextual-word predictability during reading in patients with mild Alzheimer disease. Neuropsychologia, 62, 143-151.

Fernández, G., Castro, L. R., Schumacher, M., & Agamennoni, O. E. (2015). Diagnosis of mild Alzheimer disease through the analysis of eye movements during reading. Journal of Integrative Neuroscience, 14(1), 121-133.

Forster, K. I. (1981). Priming and the effects of sentence and lexical contexts on naming time: evidence for autonomous lexical processing. Quarterly Journal of Experimental Psychology, 33, 465-495.

Freunberger, D., & Roehm, D. (2016). Semantic prediction in language comprehension: evidence from brain potentials. Language Cognition and Neuroscience, 31(9), 1193–1205.

Höltje, G., & Mecklinger A. (2022). Benefits and costs of predictive processing: How sentential constraint and word expectedness affect memory formation. Brain Research, 1788, 147942.

Hutzler, F, Braun, M., Võ, M. L., Engl, V., Hofmann, M., Dambacher, M., Leder, H., & Jacobs, A. M. (2007). Welcome to the real world: validating fixation-related brain potentials for ecologically valid settings. Brain Research, 1172, 124–129.

Hutzler, F., Fuchs, I., Gagl, B., Schuster, S., Richlan, F., Braun, M., & Hawelka, S. (2013). Parafoveal x-masks interfere with foveal word recognition: evidence from fixation-related brain potentials. Frontiers in Systems Neuroscience, 7, 33.

Ito, A., Corley, M., Pickering, M. J., Martin, A. E., & Nieuwland, M. S. (2016). Predicting form and meaning: evidence from brain potentials. Journal of Memory and Language, 86, 157-171.

Ito, A., Martin, A. E., & Nieuwland M. S. 2017. How robust are prediction effects in language comprehension? Failure to replicate article-elicited N400 effects. Language Cognition and Neuroscience, 32(8). 954-965.

Kennedy, A., & Pynte, J. (2005). Parafoveal-on-foveal effects in normal reading. Vision Research, 45(2), 153–168.

Kennedy, A., Pynte, J., Murray, W. S., & Paul, S. A. (2012). Frequency and predictability effects in the Dundee corpus: An eye movement analysis. Quarterly Journal of Experimental Psychology, 66(3), 601-618.

Kliegl, R., Nuthmann, A., & Engbert, R. (2006). Tracking the mind during reading: the influence of past present and future words on fixation durations. Journal of Experimental Psychology General, 135(1), 12–35.

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 & Cognition, 41(6), 1648-1662.

Kutas, M., & Federmeier, K. D. (2011). Thirty years and counting: Finding meaning in the N400 component of the eventrelated brain potential (ERP). Annual Review of Psychology, 62, 621–647.

Lee, C.Y., Liu, Y. N., Tsai, J. L. (2012). The time course of contextual effects on visual word recognition. Frontiers in Psychology, 3, 285.

Li, N., Niefind, F., Wang, S., Sommer, W., & Dimigen, O. (2015). Parafoveal processing in reading Chinese sentences: Evidence from event‐related brain potentials. Psychophysiology, 52(10), 1361–1374.

Liu, Y., Guo, S., Yu, L., & Reichle, E. D. (2018). Word Predictability Affects Saccade Length in Chinese Reading: An Evaluation of the Dynamic-Adjustment Model. Psychonomic Bulletin & Review, 25, 1891-1899.

Liu, N., Wang, X., Yan, G., Paterson, K. B., & Ascensión, Pagán. (2021). Eye movements of developing Chinese readers: effects of word frequency and predictability. Scientific Studies of Reading, 25, 234-250.

Liu, Z., Tong, W., Zhang, Z., & Zhao, Y. (2020). Effects of contextual predictability on lexical processing during reading: Eye movement evidence. Acta Psychologica Sinica, 52(9), 1031-1047.

Maess, B., Mamashli, F., Obleser, J., Helle, L., & Friederici, A. D. (2016). Prediction signatures in the brain: semantic pre-activation during language comprehension. Frontiers in Human Neuroscience, 10, 591.

Nieuwland, M. S. (2019). Do "early" brain responses reveal word form prediction during language comprehension? A critical review. Neuroscience & Biobehavioral Reviews, 96, 367-400.

Otten, M., & Van, Berkum, J. J. A. (2008). Discourse-Based Word Anticipation During Language Processing: Prediction or Priming? Discourse Process, 45(6), 464-496.

Petten, V. C., & Luka, B. J. (2012). Prediction during language comprehension: benefits costs and ERP components. International Journal of Psychophysiology, 83(2), 176-190.

Schuster, S., Hawelka, S., Hutzler, F., Kronbichler, M., & Richlan, F. (2016). Words in context: the effects of length frequency and predictability on brain responses during natural reading. Cerebral Cortex, 10, 1-16.

Sereno, S. C., Hand, C. J., Shahid, A., Mackenzie, I. G. A., & Leuthold, H. (2020). Early EEG correlates of word frequency and contextual predictability in reading. Language Cognition and Neuroscience, 35(5), 625-640.

Rayner, K., Binder, K. S., Ashby, J., & Pollatsek, A. (2001). Eye movement control in reading: Word predictability has little influence on initial landing positions in words. Vision Research, 41(7), 943–954.

Rayner, K., Li, X., Juhasz, B. J., & Yan, G. (2005). The effect of word predictability on the eye movements of Chinese readers. Psychonomic Bulletin & Review, 12, 1089-1093.

Rayner, K., Reichle, E. D., Stroud, M. J., Williams, C. C., & Pollatsek, A. (2006). The effect of word frequency word predictability and font difficulty on the eye movements of young and older readers. Psychology and Aging, 21(3), 448–465.

Urbach, T. P., DeLong, K. A., Chan, W. H., & Kutas, M. (2020). An exploratory data analysis of word form prediction during word-by-word reading. Proceedings of the National Academy of Sciences, 117(34),

Yan, S., & Jaeger, T. F. (2020). (Early) context effects on event-related potentials over natural inputs. Language cognition and neuroscience, 35(5), 658-679.

Yang, Y., & Shi, D. (2024). The development and innovation of Chinese linguistic theory under the influence of generative grammar. Foreign Language Teaching and Research, 56(2), 201-212.

Yuan, Y. (2025). Description or explanation: Reflecting on the two goals of linguistics through ChatGPT. Language Strategy Research, 55(1), 62-74.

Yuan, Y. (2024). Challenges and warnings from large language models like ChatGPT to linguistic theory. Contemporary Rhetoric, 1, 1-17.

Zhang, W., Chow, W. Y., Liang, B., & Wang, S. (2019). Robust effects of predictability across experimental contexts: Evidence from event-related potentials. Neuropsychologia, 134, 107229.

Zhao, S., Li, L., Chang, M., Xu, Q., Zhang, K., Wang, J., & Paterson, K. B., (2019). Older adults make greater use of word predictability in Chinese reading. Psychology and Aging, 34(6), 780–790.