Neural correlates of valence generalization in an affective conditioning paradigm

In case of uncertainty, predictions that are based on prior, similar experiences guide our decision by processes of generalization. Over-generalization of negative information has been identified as an important feature of several psychopathologies, including anxiety disorders and depression, and might underlie biased interpretation of ambiguous information. Here, we investigated the neural correlates of valence generalization to ambiguous stimuli using a translational affective conditioning task during fMRI. Twenty-five healthy individuals participated in a conditioning procedure with (1) an initial acquisition phase, where participants learned the positive and negative valence of two different tones (reference tones) through their responses and subsequent feedback and (2) a test phase, where participants were presented with the previously learned reference tones and three additional tones with intermediate frequency to the learned reference tones. By recording the responses to these intermediate stimuli we were able to assess the participantsí interpretation of ambiguous tones as either positive or negative. Behavioral results revealed a graded response pattern to the three intermediate tones, which was mirrored on the neural level. More specifically, parametric analyses OF BOLD responses to all five tones revealed a linear effect in bilateral anterior insula and SMA with lowest activation to the negative reference tone and highest activation to the positive negative tone. In addition, a cluster in the SMA showed a reverse-quadratic response, i.e., the strongest response for the most ambiguous tone. These findings suggest overlapping regions in the salience network that mediate valence generalization and decision-making under ambiguity, potentially underlying biased ambiguous cue interpretation.