Analysis of intelligent visual processes by behavioral observation and MEG measurement

本研究的目的是在行为观察和脑活动测量的基础上，勾勒出感知和识别日文字符和符号的智能视觉过程。对于行为观察，我们使用了视觉形式的认知任务，其中测量了反应时和正确操作的百分比。作为大脑活动，我们测量了受试者在执行任务时的脑磁图反应。本研究的主要结果如下：1.在两项认知任务中，被试对元音和词形特征检测的反应时无显著差异。然而，根据脑磁图分析，在字符开始后150-300ms处的局部神经活动的不同方式不同，在元音检测中，左侧腹枕区域选择性地激活，而在形态检测中，腹侧至顶枕区区域活跃。这些结果表明，视觉纹外区的皮质神经活动可能会通过自上而下的机制发生改变，这取决于个体识别书面字符的策略，即更一般的书面句子。三字词分类任务的反应时与单字元音检测任务的反应时无差异。然而，在符号的形态检测任务中，RT显著缩短，这表明这一tadk是容易的。对这些刺激和任务的神经活动通常在150-300ms的枕内侧至腹枕区观察到。此外，在单词任务中选择性地观察到左侧枕颞部活动，但在字符和符号任务中没有观察到。考虑到前人对各种视觉形式加工的研究，我们认为左侧枕枕区对特定的视觉文字加工和/或复杂视觉形式的一般加工具有辅助作用。

The objective of this research project is to delineate the intelligent visual processes to perceive and recognize Japanese characters and symbols, based on the behavioral observation and the measurement of brain activity. For the bbehavioral observation, we used a cognition task of visual forms, where the reaction time and the percent correct performance were measured. As the brain activity, we measured MEG responses while subjects performed the task. The main results are summarizedbelow :1. The RT measured in two cognition tasks of single Kanacharacters to detect vowels and morphological features was not signijicantly different each other. However, the focal neural activities, as located from the MEG analysis, at 150-300 ms after the character onset differed in a manner that the left ventraloccipitotemporal region is selectively active across subjects in the vowel detection, but the areas from the ventralto parietaloccipital region were active in the morphology detection. These results suggest that the cortical neural activity in the visual extrastriate areas would be modidied, through topdown mechanisms, depending on the individual strategy to recognize written characters, i.e., more generally written sentences.2. The RT measured ia a categorization task of threecharacter words and that in a vowel detection task of single characters was not different. However, RT was significantly shorter in a morphology detection task of symbols, which indicates the easiness of this tadk. Neural activites for these stimuli and tasks were commonly observed at 150-300 ms in the medialoccipital to ventraloccipital areas. In addition to these, the left occipitotemporal activity was selectively observed in the word task, but not character and symbol tasks. Taking into consideration of previous studies on various visual form processing, it is suggesed that the left ventraloccipitotemporal area subserves specific visual word processing and/or general processing of complex visual forms.