We studied communication patterns in face-to-face dialogues between people for the purpose of identifying conversation features that can be exploited to improve human-computer interactions. We chose to study the psychological counseling setting as it provides good examples of task-oriented dialogues. The dialogues between two participants, therapist and patient, were video recorded. The participants’ head movements were measured by using head-mounted accelerometers. The relationship between the dialogue process and head nodding frequency was analyzed on the basis of manual annotations. The segments where nods of the two participants correlated were identified on the basis of the accelerometer data. Our analysis suggests that there are characteristic nodding patterns in different dialogue stages.