Although positive and negative affect are assumed to be highly distinct, recent work has shown that facial valence of positive and negative situations may be highly confusable, especially when the emotions are intense. However, previous work has relied exclusively on static images, portraying a single peak frame of the emotional display. Dynamic expressions, on the other hand, convey a far broader representation of the emotional reaction, but are they diagnostic of the situational valence? Participants (N = 245) watched videos portraying reactions to real-life highly positive situations and evaluated the affective valence of the target. Video information was controlled by: (a) truncating the movies after 5, 10, or 20 seconds from the start, and by (b) digitally manipulating the videos such that only the face was visible with no context, only the context was visible with no face, or the face appeared in context. Results indicate that during real-life intense positive situations, facial expressions alone were rated as negative and failed to convey diagnostic information about the positive situational valence even at the most extended presentation durations. By contrast, when contextual information appeared alone or with the face, participants accurately rated the target as feeling positive, and this positivity increased with extended viewing duration. These findings suggests poor coupling between facial valence and felt emotions, supporting the notion that when emotions run high, the diagnostic power of facial expressions is reduced. Conversely, the findings demonstrate an inherent role for contextual information in the recognition of real-life intense faces. (PsycINFO Database Record (c) 2019 APA, all rights reserved).