Gelareh Mohammadi, Dimitri Van De Ville & Patrik Vuilleumier (2023), Cerebral Cortex 33(12):7993–8010 — DOI-10.1093/cercor/bhad093
Key takeaways
- Design (pp. 7995–7997). “Twenty participants watched 40 emotional clips and rated 119 emotional moments in terms of 32 component features defined by a previously validated componential model” (p. 7993). The ratings came after each scan session, on re-viewing, for pre-selected events “between 1 and 4 events per each video, with mean of 2.9 and duration of 12-s per event” (p. 7996). “This allowed us to ensure ratings corresponded to a precise event, and not more global judgments about the video” (p. 7996). Heart and respiration rate were recorded in the scanner. Electrodermal activity and facial EMG were lost to noise.
- Method (p. 7998). Partial least squares correlation (PLSC) between event-wise whole-brain contrast maps (2,276 events) and the 34 features. Significance by permutation, stability by bootstrap over subsets of subjects.
- Six functional core processes (pp. 8002–8006). LV1 valuation (unpleasantness, standards, wanting to continue or stop, feeling good or bad; vmPFC, OFC, amygdala, VTA vs insula, cingulate). LV2 novelty. LV3 hedonic impact (expression and pleasant, intense feeling; sensorimotor operculum, SMA, posterior insula, ventral pallidum). LV4 goal monitoring. LV5 caring for others or “being moved” (DMN). LV6 approach/avoidance and curiosity.
- Two valences (p. 8005). FCP-1 “may relate to the theoretical construct of “wanting” (or not wanting; Berridge 2009), rather than to the hedonic experience of” pleasure. FCP-3 is “reminiscent of the “liking” rather than “wanting” aspect of valence (Berridge 2009)” (p. 8006).
- No arousal process (p. 8007). “a striking aspect of our results is the absence of a FCP linking behavioral features and brain networks corresponding to the construct of arousal”. The feeling items alone give two latent variables, valence and arousal, and explain less covariance.
- Mixed components. Most processes “include mixed features that do not purely correspond to 1 unique component as traditionally distinguished in CPM” (p. 8005).
Standing in the record
Filed on 2026-10-23 and read for two questions, as set out in NOTE-1107. On measuring potential (QUESTION-268) it gives no synchronization index. Its latent variables are defined per observation, so each 12-second event has a signed score on each process. The paper does not report those scores. They would be a per-episode measure of componential pattern, though not of which ordering wins. On the three-layer operationalization of felt valence (QUESTION-266) it is more useful. It finds in human brain and report the wanting/liking split that THEORY-893 and CLAIM-438 press against the record’s proposal (CLAIM-440). The valuation process carries the action tendencies (continue, stop, undo), and the hedonic one carries feeling and expression.
It is affective neuroscience, not an instruction for machine-learning practice.