The heterogeneity of somatizing patients influences outcomes, especially in unselected samples in primary care. A cluster analysis was performed as secondary analysis on an existing data set of 127 somatizing patients included in a randomized controlled clinical trial. Anxiety and depression (HADS), number and intensity of physical symptoms (SOMS), physical and emotional functioning (short form of the SF-36 Health Survey), health beliefs (KKU-G), and psychological distress (General Health Questionnaire) were used for clustering. Outcome, treatment satisfaction, and diagnosis were calculated and compared for the clusters. We differentiated three groups from this analysis: one with elevated emotional and physical stress, one in which emotional stress dominated, and one with low emotional and physical stress. The three groups did not differ in diagnoses of somatoform disorders. The high-stress groups improved over time, whereas the depression and emotional-functioning scores in the low-stress group deteriorated. All patients were satisfied with the treatment provided. Deterioration in the scores of the low-stress group may be a result of a clinically valuable change process, in that patients who were initially in denial were able to open up and admit their problems. The increased satisfaction with treatment supports this interpretation. This so-called response shift must be taken into account in the planning of studies.