Composite reaction time and variability correlate with whole-brain white-matter characteristics

The relationship between processing speed and brain network characteristics has been widely studied, yet the results remain inconsistent. While many studies have linked processing speed to the microstructure of white matter, discrepancies arise due to differences in the tasks used, behavioral measures assessed (based on raw reaction time or modeled processing speed), and specific white-matter tracts considered. To address these challenges and clarify any relationship between individual differences in speed and white-matter brain networks, we present a pre-registered analysis using a large (N = 159) dataset, incorporating state-of-the-art MRI data acquired from a high-gradient 3T Connectom scanner. We combine data from three reaction-time tasks to create composite measures of cognitive performance, mitigating the limitations of experiment-specific analyses. Alongside classic behavioral measures of mean reaction time, reaction-time variability, and accuracy, we applied the drift-diffusion model to derive the common metric of modeled processing speed, drift rate, as well as accompanying parameters of boundary separation, and non-decision time. Using general linear models, we explored the relationship between these parameters and the whole-brain and task-specific structural networks of the brain, weighted by volume-normalized streamline counts and myelin water fraction. Our results revealed negative associations between the global efficiency of streamline-weighted networks and both mean reaction time and reaction-time variability (β = -0.18/-0.21, p = 0.025/0.01 for whole-brain and β = -0.18/-0.18, p = 0.028/0.022 for the task-specific network). Effect sizes were small, consistent with other pre-registered assessments of brain-behavior correlations. These effects were not captured by decision model parameters signaling a note of caution for the assumed interpretation of these parameters. The significant association with reaction-time variability was robust to controlling for age, while age captured significant variance in the association with mean reaction time. This may imply that physiological changes associated with age would be an avenue for research to uncover mechanisms relating structure to reaction time. In sum, we attempted a state-of-the art clarification of whether structural brain organization is associated with speed in common cognitive tasks, and we found a small association with reaction-time variability and mean reaction time (and age).