关键词: 贝叶斯阅读者模型, N200增强效应, 任务状态, 汉字识别, 事件相关电位

Abstract: Background. How does the brain efficiently recognize visual words? This core question in reading research has prompted inquiry into whether, beyond stimulus-driven feedforward processing, task-state-driven tuning also participates in early orthographic processing. The Bayesian Reader Model proposes that skilled readers dynamically adjust the depth of orthographic processing according to task environments, achieving word recognition with minimal cognitive cost. The N200 repetition enhancement effect, a neural marker of early orthographic processing in Chinese character recognition, provides an opportunity to test whether such task-state-driven tuning exists. However, existing research has focused primarily on stimulus-level manipulations, and whether this effect exhibits dynamic adaptability to task states remains unclear. Drawing on three core parameters of the Bayesian Reader Model, the present study examined how task states modulate the N200 repetition enhancement effect. Method. Two electroencephalography (EEG) experiments employed a visual repetition priming paradigm. Experiment 1 manipulated repetition probability (high vs. low) and task type (lexical decision vs. same-different judgment) to examine the effects of prior probability and task demand on the N200 repetition enhancement effect. Twenty-four participants completed both tasks. Experiment 2 maintained the same level of visual feature repetition at the physical level as the high repetition probability condition in Experiment 1 while manipulating noise level at the lexical level, examining the effect of noise level on the N200 enhancement. Twenty-four participants completed this experiment. Results. Experiment 1 revealed that lexical decision tasks elicited significantly stronger N200 repetition enhancement than same-different judgment tasks. The modulation by prior probability showed task dependency: under high repetition probability, the N200 enhancement disappeared in same-different judgment tasks but remained stable in lexical decision tasks. Experiment 2 demonstrated that, with feedforward processing controlled, lexical-level noise still elicited a significant N200 enhancement effect. Both experiments showed that N400 effects exhibited opposite task modulation patterns, with same-different judgment tasks eliciting stronger N400 effects than lexical decision tasks. Conclusion. The N200 repetition enhancement effect is dynamically modulated by task complexity, prior probability, and noise level, reflecting an optimization mechanism whereby early orthographic processing flexibly adjusts its depth according to task states. These findings extend the core predictions of the Bayesian Reader Model from alphabetic writing systems to Chinese character recognition, providing novel cross-script neural electrophysiological evidence for dynamic optimization mechanisms in early visual word recognition. The results support the view that visual word recognition involves not only stimulus-driven feedforward processing but also a tuning process that flexibly adjusts processing depth in response to task environments.

Key words: Bayesian reader model, N200 enhancement effect, Task state, Chinese character recognition, Event-related potentials