Interpersonal accuracy is the ability to correctly assess others' traits and states based on the observation of verbal and nonverbal behaviors of our pairs. This skill is usually measured with a general emotion recognition ability test. Some researchers have also assessed it as a more situation-specific skill as it unfolds in a social interaction. The aim of the present study was to test which aspect predicts interaction outcomes best. In a field study, we assessed general emotion recognition accuracy of 60 general practitioners and we asked them to evaluate their patients' preferences for medical communication style after each of 4 to 6 consultations. This evaluation was compared to the patient's actual preference to obtain a measure of situation-specific interpersonal accuracy. Patients also indicated their satisfaction, trust, and enablement. Results show that the situation-specific interpersonal accuracy was related to more positive interaction outcomes whereas the general emotion recognition skill was not.