This study explores the effectiveness of statistical feature extraction from functional near-infrared spectroscopy (fNIRS) signals in assessing English reading comprehension. Brain signals were recorded from 15 participants while reading 30 passages, followed by multiple-choice comprehension tests. Nine statistical features were extracted, and up to three-feature combinations were formed, resulting in 1,290 feature sets. The k-nearest neighbor (k-NN) classifier was utilized for classification, achieving an average accuracy of 73.48%. Among the statistical features, kurtosis was the most frequently selected, appearing 66 times, while skewness was the least selected, appearing 8 times. The highest and lowest classification accuracies were 76.30% and 71.85%, respectively. A unique dataset was collected by implementing an original experimental procedure in this study. This study contributes to the NeuroIS field by offering a novel, brain-based approach to evaluate reading comprehension, a key competency in multilingual information technologies teams and global digital collaboration environments.