Functional near-infrared spectroscopy (fNIRS) may be suited for functional monitoring during swallowing as it is comparatively immune to body movement. However, still fNIRS measurement on swallowing poses a technical problem that it may often involve motion artifacts. Although there is no single way to solve this problem, technical insights have been available form related studies in the past. Here we introduce two examples for analyzing data rich in motion artifacts putting emphasis on temporal structures of the data. The first is about fNIRS assessment of language function during overt naming tasks. Since data were temporally continuous, we adopted a general linear model with regression to a canonical hemodynamic response function to extract cortical activations related to overt naming tasks. The second example is about fNIRS assessment on go/no-go task performance with or without methylphenidate administration in Attention Deficit Hyperactivity Disorder (ADHD) children. Since data were disrupted by unexpected motion artifacts, we simplified temporal data structures by averaging to extract only robust signals. Thus, we indicated that the optimum analytical strategy varies depending on the temporal structures of the data.