Figure 1.

 A. Predictive conceptual framework of emotion-driven plasticity in cortical representations of face ID. Prior to conditioning, we expect no overlapping activation of neurons coding for front (F) and quarter (Q) views among the cortical population. Each presentation of a face elicits a view specific representation. Throughout emotional signals mediated by amygdala feedback, we hypothesize that the representation of a face in front-view will not only be strengthened after being paired with threat (blue doted line), but also “enlarged” to become associated with another view (e.g. quarter) of the same face (yellow dotted line). After emotion learning, we predict that these new connections will lead to overlapping activation of F and Q neurons coding for the same face, leading to the emergence of face ID priming across different viewpoints. N.B. Each circle corresponds to one hypothetical neuron and lines represent connections between them within a cortical population. Filled circles indicate active neurons whereas empty circles indicate inactive neurons in response to either F or Q face presentation. B. Experimental Design. Familiarization session: Full-front and ¾-views (left or right) images of two neutral faces were presented in a rapid succession, always separated by a scrambled visual mask. Participants were asked to judge on each trial whether the two faces depicted the same or different individuals, allowing them to form robust representations of identity (ID) for different views of the same face. Aversive conditioning session: One of the two IDs used in the familiarization session in full-front view (familiar and conditioned stimulus, famCS+) was paired with a negative emotional experience (unpleasant loud sound), using a classic Pavlovian conditioning procedure with a 60% reinforcement. The other ID (familiar but not conditioned, famCS-) plus a series of filler faces (all unfamiliar) were paired with a neutral sound (soft beep noise). Importantly, all faces were always presented in front view only. Participants were instructed to look carefully at the faces and memorize them. Filler faces were added to increase encoding load and credibility of the memory task. Main repetition priming experiment: This session tested for any repetition priming effect when the same face ID (famCS+, famCS−, and new unknown faces) was presented in pairs of images, across different viewpoints. The famCS+, famCS-, and new unfamiliar faces (new) were presented in a rapid succession of two different images. The first image always depicted a full-front view (like during aversive conditioning), allowing for the generation of perceptual predictions, while the second image always depicted a ¾-view (either left or right deviated), and the identity of these two faces was either the same or different (repeated or non-repeated ID) in order to assess cross-view priming. Participants had to indicate whether the second face looked toward the left or right side. N.B. Faces used in the figure are home-made avatars created using FACsGen software for illustrative purpose only. Karolinska Face dataset 66 was used in the experiment.