The unfolding dynamics of the vocal expression of emotions are crucial for the decoding of the emotional state of an individual. In this study, we analyzed how much information is needed to decode a vocally expressed emotion using affect bursts, a gating paradigm, and linear mixed models. We showed that some emotions (fear, anger, disgust) were significantly better recognized at full-duration than others (joy, sadness, neutral). As predicted, recognition improved when greater proportion of the stimuli was presented. Emotion recognition curves for anger and disgust were best described by higher order polynomials (second to third), while fear, sadness, neutral, and joy were best described by linear relationships. Acoustic features were extracted for each stimulus and subjected to a principal component analysis for each emotion. The principal components were successfully used to partially predict the accuracy of recognition (i.e., for anger, a component encompassing acoustic features such as fundamental frequency (f0) and jitter; for joy, pitch and loudness range). Furthermore, the impact of the principal components on the recognition of anger, disgust, and sadness changed with longer portions being presented. These results support the importance of studying the unfolding conscious recognition of emotional vocalizations to reveal the differential contributions of specific acoustical feature sets. It is likely that these effects are due to the relevance of threatening information to the human mind and are related to urgent motor responses when people are exposed to potential threats as compared with emotions where no such urgent response is required (e.g., joy).