The detection of conflict cues in call centre interactions may be related to the quality assessment of the services provided, since these cues reveal both speakers’ emotional states and positioning as expressed through complaining and identifying problematic issues on the one hand, and managing requests or resolving problems on the other hand. This paper describes a set of emotional and conversational cues associated to conflict as well as a machine learning approach to classify emotional speech units occurring in a call centre dataset by employing emotion labels as well as automatically extracted acoustic and additional context-related features.