Uncovering the Neural Substrates of Physics Problem Solving: A New Paradigm with Behavior Correlates

Bartley JE, Riedel MC, Salo T, Boeving ER, Odean R, Bravo E, Laird R, Pruden S, Brewe E, Sutherland MT, Laird AR, HBM (2017).


Physics education research examines how students think about physics to facilitate learning, which is often measured via physics problem solving (PPS) abilities. Behavioral studies have focused on understanding the sequential, cognitively diverse processes and factors influencing PPS success [1]. For example, science anxiety negatively impacts PPS performance [2], whereas visuospatial skills are positively associated with PPS [3]. Currently, no neuroimaging study has characterized the neural correlates of PPS, although investigations of numeric tasks have begun to map brain dynamics across problem stages, noting differential fronto-parietal contributions across time [4]. To provide insight into the neural mechanisms of PPS, we developed a new fMRI paradigm, probed brain activity across PPS stages, and evaluated brain-behavior correlates with anxiety and visuospatial skills.


We acquired fMRI data from 56 healthy right-handed participants (29 female; aged 18-24). Participants were Florida International University undergraduates enrolled in college-level introductory physics. Students completed generalized and science anxiety inventories and WAIS behavioral testing. Students underwent fMRI immediately after physics course conclusion wherein they solved physics problems based on the Force Concept Inventory (FCI); a widely used test of physics conceptual understanding [5]. FCI (and ‘Control’) questions were presented via three self-paced screens: S1 consisted of text and picture descriptions of a physical scene, S2 posed a physics question, and S3 listed possible answer choices (Fig 1ab). Analyses were performed in FSL and activation maps (FCI - Control) were thresholded at P<0.05 (clusters z>3.09) for S1-3 concatenated and S1, S2, and S3 separately. We assessed correlations between resultant S1-3 ROI parameter estimates and WAIS (total and subtests), general and science anxiety scores.


PPS was linked with bilateral activity in the prefrontal cortex, frontal poles, anterior cingulate, inferior parietal lobules (PL), temporal gyri, and cerebellum (Fig 1c). Screen-wise analyses confirmed distinct activation during sequential PPS stages (Fig 1c). S1 evoked bilateral insula, superior frontal sulci, inferior frontal gyri, superior PL, postcentral sulci, posterior cingulate (PCC), thalamus, and occipitotemporal activity. Limited S2 activations occurred in right intraparietal sulcus, angular gyrus, bilateral superior PL, and superior frontal sulci. S3 engaged a left-lateralized frontal-parietal network with PCC and middle temporal activation. High WIAS Visual Puzzles subtest performance was linked to increased PPS-related activity in the left frontal pole, left inferior PL, and right supramarginal gyrus. We observed negative correlations in the same ROIs for science anxiety. Full WAIS and generalized anxiety scores did not correlate with PPS-related activity.


FMRI data from college physics students identified fronto-tempo-parietal activity during PPS. Differential activation occurred across PPS stages, suggesting system-level brain dynamics support stepwise problem solving. PL, frontal, and salience network recruitment in S1 may detect semantic conflicts in text [6], transient S2 PL activity might guide attentional visual focus [7], and cooperation between default mode and frontal areas in S3 is consistent with reasoning involving autobiographic events [8]. PPS-related activity positively correlated with visuospatial reasoning but negatively correlated with science anxiety in a frontopolar area, involved in top-down emotion reappraisal during cognition [9], the supramarginal gyrus, which discriminates between spatial cues [10], and inferior PL, which supports episodic memory-based reasoning [8]. These results suggest students who experience science anxiety may have reduced capacity to mobilize cognitive resources when faced with negative emotional responses across problem solving stages.