60 patients admitted to the Geneva University Hospitals after a suicide attempt were video-recorded during an interview with a psychiatrist. After each interview, the therapist was asked to assess the patient's suicide risk on a 4-point scale. During follow-up appointments two years later, we identified 11 patients who had made subsequent suicide attempts ('repeaters'). These patients were matched with 11 'non-repeaters' from the original sample with respect to gender, age and number of previous suicide attempts. To code the therapist's facial behavior during the original interviews, we used Ekman and Friesen's 'Facial Action Coding System' (FACS) and analyzed the ways in which her behavior differed between repeaters and non-repeaters.

Using traditional statistical analysis, we found significant differences between the two groups that allowed 82-91% of the patients to be correctly classified as either repeaters or non-repeaters. In particular, there were differences in the average activation of some coded units, such as peri-ocular activation, and in the duration of the therapist's gaze straight at the patient. By contrast, the therapist's written prediction was erroneous: only 23% of the patients were correctly classified.

These findings reflect the therapist's accurate perception of risk without her being aware of it. In the future, after replication, these findings may lead us to new methods for assessing suicide risk. New analyses with Theme should help us to better understand the interaction between therapist and patient, and may reveal which cues made the therapist able to distinguish between repeaters and non-repeaters - albeit non-verbally and subconsciously.

Paper presented at Measuring Behavior 2002 , 4^th International Conference on Methods and Techniques in Behavioral Research, 27-30 August 2002, Amsterdam, The Netherlands

© 2002 Noldus Information Technology bv