Figure 3
Conceptual model of relative contributions of neuroimaging and physiological measures of affect induction based on a theory of individual differences. Patterns of neural activation are represented as gray circles. Normative valence scores of the stimuli associated with these neural activations are presented as labels within the circles’ boundaries. Single subject perceived valence scores of the stimuli are depicted as gray numbers above and to the right of the circles. (A) Theoretical Case 1: a subject’s perceived valence of the stimuli exactly agrees with the normative valence scores. The resulting direction of prediction for fixed effects of heart rate deceleration (blue arrow) and predictions of a fit support vector machine (red arrow) mutually agree. (B) Theoretical Case 2: a subject’s perceived valence of the stimuli (as indicated by shifts of neural activations and associated perceived valence scores) disagrees with the normative valence scores, thereby increasing the relative predictive effect of heart rate deceleration. When image stimuli are polar-extreme, perceived affect better aligns with the stimulus set’s normative affect scores (panel A); therefore, measures of both neural activation (which is fit to, and, therefore, predicts normative affect scores) and HR change (which, in theory, measures perceived affective valence) predict along the axes of normative valence scores (the measure of interest in this experiment). However, when image stimuli are neutral, disagreements between the subjects’ perceived affect and the normative valence scores emerge. Measures of neural activation generalize to these individual differences during the learning process whereas HR change continues to weakly, but reliably, predict perceived affect, yielding relatively lower predictive effect size on the measure of interest.
