Traditional fMRI data analysis relies on pre-defined BOLD reference functions derived from task paradigms, potentially overlooking transient brain processes linked to task performance. Additionally, active brain areas identified through general linear model (GLM) analysis exhibit highly correlated time courses, limiting effective connectivity analysis. To address these issues, we applied independent component analysis (ICA) to extract independent BOLD processes and identify task-related components. This approach preserves the time courses of selected independent BOLD components, enabling connectivity analysis without the constraints of high correlation inherent in GLM-based methods. More importantly, ICA allows for single-trial segmentation of BOLD time courses, facilitating single-trial analysis. By leveraging this methodology, we assessed functional connectivity at the single-trial level and examined its correlation with the difficulty of individual Chinese relative clause sentences. This novel approach enhances our ability to investigate dynamic neural processes and their relationship to cognitive complexity, providing deeper insights into sentence comprehension mechanisms.