Evaluating the reliability of functional near-infrared spectroscopy data in the context of a reasoning paradigm

Abstract Functional near-infrared spectroscopy (fNIRS) is a portable, motion-tolerant neuroimaging method particularly well suited for developmental and naturalistic research. To evaluate the utility of fNIRS for studying individual differences and longitudinal changes, we measured activation and functional connectivity during a relational reasoning task in young adults (N = 73). We sought to (1) establish whether fNIRS captures frontoparietal activation patterns consistent with prior fMRI studies using similar paradigms, (2) assess the effect of the amount of data (number of task blocks) on signal strength and precision, (3) assess the paradigm’s measurement properties in the form of intra- and interindividual stability of activation and functional connectivity within and across testing sessions, and (4) examine whether grouping channels into anatomical regions of interest (ROIs) conferred benefits to the above. We observed robust task-evoked activation across lateral prefrontal and parietal cortices, with effect sizes on par with prior fMRI studies. Generally, we observed diminishing returns in effect size and measurement precision beyond ∼7 minutes. Internal consistency and test-retest reliability varied across metrics; while they were very low for a specific task contrast, they were extremely high for functional connectivity, confirming the robustness of channel- and ROI-level connectivity as a stable marker of functional architecture. Exploratory analyses supported prior observations of lower signal quality in participants with darker skin tones and hair, underscoring the need for inclusive methodological strategies. Together, these findings highlight key design considerations for optimizing longitudinal and individual-differences research on higher-level cognition, particularly in diverse and developmentally variable populations. Highlights We measured within- and between-session reliability of fNIRS metrics Collecting more data yielded diminishing returns in effect size and precision There were tradeoffs to aggregating channel data into regions of interest General task activation was more reliable than a specific task contrast Functional connectivity showed extremely high test–retest reliability