Hierarchical Bayesian analysis reveals complex neural dynamics of inhibitory control

Thumbnail Image



Journal Title

Journal ISSN

Volume Title


The Ohio State University

Research Projects

Organizational Units

Journal Issue


Cognitive control has been of interest to psychologists and neuroscientists because of its contribution to understanding individual differences, impulsivity, addiction, and obsessive-compulsive disorder. Two tasks used to test cognitive control are the Go/No-Go (GNG) and Stop-Signal (SS) tasks. In the GNG task, subjects are given a cue to respond or withhold a response at the beginning of a trial. The SS task extends this basic paradigm by including the possibility that a “Go” cue may switch to a response-withholding cue. Behavioral and functional magnetic resonance imaging (fMRI) data, extracted for twenty- four regions of interest (ROIs), were collected from eleven subjects who completed both the GNG and SS tasks. In this study, blood oxygenation level-dependent (BOLD) responses were fit using a hierarchical Bayesian analysis to five increasingly complex models of the trial-wise neural activation to improve the signal-to-noise ratio and explore differences in neural activation between response (Go trials) and response inhibition (No-Go/Stop trials). We found that constructing a hierarchy, or adding multiple levels to the model, greatly constrained the predicted BOLD signal by systematically removing outliers. Additionally, increasing model complexity elucidated brain regions that played a role solely in carrying out a response (Go trials). We next replicated these results using the more complicated SS task. We found, from adding a hierarchical structure, that some brain areas showed less activation after a stop signal than during either a Go or No-Go trial. Our results suggest hierarchical modeling is a useful tool in interpreting often noisy fMRI data.



hierarchical Bayesian, neuroscience, fMRI, response inhibition, cognitive control