Using fMRI data and experience sampling data to map patterns of ongoing thought onto brain activity during movie-watching. Left to Right - One sample of participants was scanned while watching movies (Sample 1), and a different set of participants responded to experience sampling probes (Sample 2) while watching the same movies in the laboratory. Decomposition of mDES data into low-dimension experiential patterns using principal component analysis (PCA) produced a set of dimensions that describe experience during movie watching (a “thought-space” within which the dynamics of the movie-watching experience unfold). Word clouds illustrate how the experience sampling questions map onto each dimension that describes this space. In these word clouds, the font size describes their importance (bigger = more important), and the colour describes their polarity (red = positive, blue = negative).

The relationship between how patterns differ across genres of movies and relate to comprehension. Left to Right – The 3D scatterplot shows the average location of each film on three of the four PCA dimensions, “Episodic Knowledge,” “Verbal Detail,” and “Sensory Engagement.” The bar graphs show the average loading on each dimension, with the error bars showing the 95% Confidence Intervals. The plots on the right illustrate the relationship between the mDES dimensions and memory for information in the film. The top barplot shows the average comprehension score on each film with 95% Confidence Intervals error bars. The scatter plots below show the association between mDES components and comprehension. The scatter plot on the left shows the negative linear relationship between the “Intrusive Distraction” thought and memory. The plot on the right shows a positive association with “Sensory Engagement.” The blue line represents the best-fit line, and the shaded area shows the 95% Confidence Intervals.

Group-level neural activation patterns associated with each of the dimensions of thought identified in a voxel space analysis. Left to Right - Regions in red are associated with activity corresponding to reports of “Episodic Knowledge,” green regions are associated with “Intrusive Distraction,” areas in purple are associated with “Verbal Detail,” and the regions in orange represent activity associated with “Sensory Engagement.” The bar plot illustrates the directionality of each parameter estimate with error bars representing 95% Confidence Intervals. Corresponding word clouds for each thought pattern are presented on the right for reference (Top to Bottom: “Episodic Knowledge,” “Intrusive Distraction,” “Verbal Detail,” and “Sensory Engagement”).

Brain regions are associated with multiple experiential features during movie watching. A region of superior temporal cortex is associated with positive reports of thoughts like “Sensory Engagement” and negative reports of thoughts like “Verbal Detail” (coloured green). A region of dorsal visual cortex was associated with both thoughts reported like “Sensory Engagement” and “Episodic Detail” (coloured red). The word clouds in the middle panel show the results of a Neurosynth meta-analysis of the regions, highlighting the most likely functions associated with these regions. The font size describes their importance (bigger = more important), and the colour describes their polarity (darker = positive). The panel on the right shows the results of seed-based functional connectivity analysis of these regions of overlap from a separate resting-state study.

Comparison of the locations of each moment across the movie clip in the (top row) “Thought Space” and the “Brain Space.” Left to Right – 3D scatterplots of the coordinate locations of each thought pattern (“Episodic Knowledge,” “Verbal Detail,” and “Sensory Engagement) and gradients 1-3 (Gradient 1 (Associated – Primary), Gradient 2 (Visual – Somato-motor), Gradient 3 (Frontoparietal – Default) during Citizenfour, Little Miss Sunshine, and 500 Days of Summer. Observations in blue occur earlier during the film, and observations in red occur later in the film. The gradient maps (1-3) and thought pattern word clouds are presented on the right for reference.

Comparison of State Space- and Voxel-based analyses of “Sensory Engagement” and “Episodic Knowledge” with Gradients. Left to right – The barplots illustrate the associations for the significant models using Gradients 1-5 as explanatory variables and the thought patterns, “Sensory Engagement” and “Episodic Knowledge” as dependent variables. We performed two spin tests to formally compare these results to those using the voxel space analysis (permutation = 2500). The spin tests revealed the location of the cluster of voxels associated with “Sensory Engagement” are located within the sensory regions of Gradient 1, unlike to have occurred by chance, p = .018. In contrast, the location of the cluster of voxels associated with Episodic Knowledge on Gradient 4 was within the null distribution, p = .251. The locations of the relevant clusters in gradient parcel space are presented in the scatter plots (red points indicate the location of parcels from the relevant comparison).