Humans routinely integrate affective information from multiple sources. For example, we rarely interpret an emotional facial expression devoid of context. In this paper, we describe the neural correlates of an affective computation that involves integrating multiple sources, by leveraging the ambiguity and subtle feature-based valence signals found in surprised faces. Using functional magnetic resonance imaging, participants reported the valence of surprised faces modulated by positive or negative sentences. Amygdala activity corresponded to the valence value assigned to each contextually modulated face, with greater activity reflecting more negative ratings. Amygdala activity did not track the valence of the faces or sentences per se. Moreover, the amygdala was functionally coupled with the nucleus accumbens only during face trials preceded by positive contextual cues. These data suggest 1) valence-related amygdala activity reflects the integrated valence values rather than the valence values of each individual component, and 2) amygdalostriatal coupling underpins positive but not negative coloring of ambiguous affect.

人类通常会整合来自多个来源的情感信息。例如，我们很少在没有上下文的情况下解释情绪化的面部表情。在本文中，我们通过利用在惊讶的面孔中发现的模糊性和基于特征的微妙效价信号来描述涉及整合多个来源的情感计算的神经关联。使用功能磁共振成像，参与者报告了由正面或负面句子调制的惊讶面孔的效价。杏仁核活动对应于分配给每个上下文调制的面部的价值，活动越大反映负面评价越多。杏仁核活动并没有追踪面部或句子本身的效价. 此外，杏仁核仅在正面情境线索之前的面部试验中与伏隔核在功能上耦合。这些数据表明 1) 与价态相关的杏仁核活动反映了综合价值而不是每个单独成分的价值，2) 杏仁核纹状体耦合支持了积极但不是消极的模糊影响着色。