MRC Cognition and Brain Sciences Unit, Cambridge, UK.
Neuroimage. 2012 Apr 2;60(2):1462-77. doi: 10.1016/j.neuroimage.2012.01.061. Epub 2012 Jan 16.
We usually feel that we understand a familiar word "immediately". However, even basic aspects of the time-line of word recognition are still controversial. Different domains of research have still not converged on a coherent account. An integration of multiple sources of information would lead to more strongly constrained theoretical models, and help finding optimal measures when monitoring specific aspects of word recognition impairments in patient groups. In our multimodal approach--combining fast behavioral measures, ERPs and EEG/MEG source estimation--we provide converging evidence for the latencies of earliest lexical and semantic information retrieval in visual word recognition. Participants performed lexical and semantic decisions (LD, SD) in a Go/NoGo paradigm. We introduced eye-blink latencies as a dependent variable, in order to measure behavioral responses that are faster and less variable than traditional button presses. We found that the earliest behavioral responses distinguishing stimulus categories can occur around 310 ms. Ex-Gaussian analysis of behavioral responses did not reveal reliable differences between LD and SD. The earliest ERP differences between Go and NoGo conditions occurred around 160 ms for both LD and SD. Distributed source analysis of combined EEG/MEG data estimated neuronal generators for the lexicality effect around 200 ms in the left anterior middle temporal lobe. Thus, behavior and brain responses provide coherent evidence that the brain starts retrieving lexical and semantic information near-simultaneously within 200 ms of word onset. Our results support models of word recognition that assume a continuous accumulation of task-related information from the stimulus, which might be described by Bayesian principles.
我们通常会觉得自己“立刻”理解了一个熟悉的单词。然而，即使是单词识别时间线的基本方面仍然存在争议。不同的研究领域尚未就一个连贯的解释达成一致。整合多种信息来源将导致更受约束的理论模型，并有助于在监测患者群体的特定单词识别障碍方面找到最佳的测量方法。在我们的多模态方法中——结合快速行为测量、ERP 和 EEG/MEG 源估计——我们提供了视觉单词识别中最早的词汇和语义信息检索的潜伏期的一致性证据。参与者在 Go/NoGo 范式中执行词汇和语义决策 (LD、SD)。我们引入了眨眼潜伏期作为一个因变量，以便测量比传统按钮按压更快且更可变的行为反应。我们发现，最早能够区分刺激类别的行为反应可以在 310 毫秒左右发生。对行为反应的 Ex-Gaussian 分析并未揭示 LD 和 SD 之间可靠的差异。对于 LD 和 SD，最早的 ERP 差异在 Go 和 NoGo 条件之间大约在 160 毫秒时出现。组合 EEG/MEG 数据的分布式源分析估计了左前颞中回的词汇效应的神经元发生器，大约在单词起始后的 200 毫秒。因此，行为和大脑反应提供了一致的证据，表明大脑在单词出现后大约 200 毫秒内开始同时检索词汇和语义信息。我们的结果支持了假设从刺激中连续积累与任务相关的信息的单词识别模型，这些信息可能可以用贝叶斯原理来描述。
