Eye movements reveal process of feature integration

With rapid development of neuropsychology science, studies on eye movements made more fateful achievements since eye movements are regarded as an important tool for studying visual information processing. Further, studies have showed that models of eye movement reflect human's psychological changes. The current study aims to build classifiers using eye movements while subjects freely viewed segmented stimulus (2 or 30 bars, coherent or incoherent). Saccade rate, saccadic duration, and saccade amplitude were picked as parameters and Fisher linear discriminant analysis (LDA) and Support Vector Machine (SVM) were applied on eye movements to build classifiers. For both types of classifiers, the classifier used saccade amplitude produced highest accuracy which was even better than that used the combination all parameters. Comparing between two types of classifiers, the SVM showed advantages over the Fisher LDA. Furthermore, classification accuracy of discriminating incoherent-2bars condition from other conditions was highest, consistent with the results of statistical analysis. These results suggest that saccade amplitude might play an important role during the process of feature integration.