Between-movie variability severely limits generalizability of “naturalistic” neuroimaging
Presented During:
Friday, June 27, 2025: 11:30 AM - 12:45 PM
Brisbane Convention & Exhibition Centre
Room:
P2 (Plaza Level)
Poster No:
1041
Submission Type:
Abstract Submission
Authors:
Simon Leipold1, Rajat Ravi Rao1, Jan-Mathijs Schoffelen1, Sara Bögels1, Ivan Toni1
Institutions:
1Donders Institute for Brain, Cognition, and Behaviour, Nijmegen, The Netherlands
First Author:
Co-Author(s):
Jan-Mathijs Schoffelen
Donders Institute for Brain, Cognition, and Behaviour
Nijmegen, The Netherlands
Donders Institute for Brain, Cognition, and Behaviour
Nijmegen, The Netherlands
Introduction:
"Naturalistic imaging" paradigms, where participants watch movies during fMRI, have gained popularity over the past two decades. Many movie-watching studies measure inter-subject correlation (ISC), which refers to the correlation between participants' neural activation time series. Previous research has focused on explaining ISC differences during movie-watching based on individual states and traits, such as social distance, personality, and political orientation. For example, friends show higher ISC than strangers while watching movies. However, movies are not natural categories but cultural artifacts that evoke varying levels of ISC depending on content, directing style, or editing methods. This raises questions about how much trait- or state-like differences in ISC depend on the specific movies chosen, potentially limiting the generalizability of findings across different movies.
·Overview.
Methods:
Here, we used an fMRI dataset of 112 participants watching eight animated movies to (a) quantify between-movie variability in ISC across the brain and (b) assess the implications for the generalizability of trait- or state-like effects on ISC. Specifically, we investigated the spatial distribution of ISC variability by quantifying the between-movie variability in ISC values across all parcels of the cortex. We conducted a repeated-measures ANOVA with Movie as the within-subject factor for each of the 210 cortical parcels defined by the Brainnetome atlas.
Results:
We found substantial between-movie variability in ISC, with this variability differing across brain regions. Crucially, brain regions with the highest ISC exhibited the greatest variability, indicating that trait- or state-like differences in ISC from one movie may not generalize to others.
Specifically, we found a statistically significant main effect of Movie (family-wise error (FWE)-adjusted p < 0.05) for 96 of the 210 cortical parcels (45.7%). The strongest variability was found in the primary and associative visual cortex, as well as the primary and associative auditory cortex. However, there was also substantial variability in higher-order areas, such as the precuneus and inferior parietal lobule.
Furthermore, we discovered that brain regions showing high variability, such as the superior temporal cortex, also had relatively higher levels of ISC. Conversely, brain regions with low variability, such as the parahippocampal gyrus, had lower levels of ISC. To quantify this relationship, we calculated the Pearson correlation coefficient between the level of ISC and between-movie variability of ISC across the 210 brain regions. ISC levels and between-movie variability were strongly positively correlated (r = 0.70, t(208) = 14.21, p < 0.001, R2 = 0.49), indicating that brain regions with stronger ISC showed higher between-movie variability in ISC. Thus, the brain regions that synchronize most across subjects when viewing movies are the most strongly affected by variability in ISC between movies.
Specifically, we found a statistically significant main effect of Movie (family-wise error (FWE)-adjusted p < 0.05) for 96 of the 210 cortical parcels (45.7%). The strongest variability was found in the primary and associative visual cortex, as well as the primary and associative auditory cortex. However, there was also substantial variability in higher-order areas, such as the precuneus and inferior parietal lobule.
Furthermore, we discovered that brain regions showing high variability, such as the superior temporal cortex, also had relatively higher levels of ISC. Conversely, brain regions with low variability, such as the parahippocampal gyrus, had lower levels of ISC. To quantify this relationship, we calculated the Pearson correlation coefficient between the level of ISC and between-movie variability of ISC across the 210 brain regions. ISC levels and between-movie variability were strongly positively correlated (r = 0.70, t(208) = 14.21, p < 0.001, R2 = 0.49), indicating that brain regions with stronger ISC showed higher between-movie variability in ISC. Thus, the brain regions that synchronize most across subjects when viewing movies are the most strongly affected by variability in ISC between movies.
·Variability in inter-subject correlation across movies.
Conclusions:
We conclude that variability between movies limits the generalizability of trait- or state-like ISC differences. Using a specific movie in neuroscience should be treated similarly to using a particular task, requiring a comparable characterization of the constituent cognitive elements. Broad generalizations about "naturalistic imaging" or "movie watching" are not warranted.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 1
Perception, Attention and Motor Behavior:
Perception: Auditory/ Vestibular
Perception: Visual 2
Keywords:
FUNCTIONAL MRI
1|2Indicates the priority used for review