The predictive value for spontaneous improvement in individuals suffering from medically unexplained symptoms (MUS) was explored of (1) anxiety and depression obtained from questionnaires, (2) negative affective states obtained from experience-sampling, and (3) ambulatory-assessed real-life physiological recordings. Sixty-seven individuals with MUS and 61 healthy controls were included. Twenty-four hour ambulatory recordings of cardiac autonomic activity, respiration, end-tidal CO2 and saliva cortisol were combined with experience-sampling of somatic complaints and mood. Complaints were assessed again after one year. Although a reduction in symptoms (25%) was found, this could not be predicted from initial anxiety and depression. Improvement was somewhat related to relatively low diary reports of fatigue, especially in the late-afternoon and evening (3% variance explained). From the physiological measures only relatively high PetCO2 values in the morning predicted improvement (5% explained). It was concluded that spontaneous recovery from MUS is hard to predict from self-reported distress and ambulatory physiological recordings.