Value spillover: How contextually irrelevant values influence choice and vmPFC activity in humans

Abstract

Objects we choose between often have multiple features. Research has shown that decision-making is guided by attentional selection of contextually-relevant features, while ventromedial prefrontal-cortex(vmPFC) represents the expected outcomes associated with these features. Yet, if this selective value retrieval is not entirely perfect, irrelevant features, and values they carry in other contexts, will influence neural value representations as well as choice. We tested this idea by utilizing a contextdependent random-dot motion paradigm. Forty humans made decisions between two clouds of moving dots, each consisting of two features (motion direction and dot color). First, participants learned to associate each color and motion with specific rewards. During subsequent decision making, a context cue indicated the trials relevant feature-type (color/motion) and choices led to outcomes associated with the relevant feature. In line with our hypothesis, the more values of the irrelevant features agreed with the relevant, the faster participants reacted. fMRI analyses showed parametric modulation of the vmPFC/OFC signal by both (1) the relevant feature value and (2) the value difference between irrelevant features. These results indicate that contextually-irrelevant features influence value representation and suggest that the brain’s decision system computes values in the presence of partial activation of irrelevant context or task states.