Techniques for classification of a posture state of a patient using multiple posture state definitions, and for delivering therapy according to the patient's classified posture state. Detected parameter values describing a patient's posture state are compared to posture state definitions. This comparison is used to determine similarity values describing how similar the patient's posture state is to each of the posture states described by the posture state definitions. Weighting factors may be determined from the similarity values and used to weight therapy parameter values that are associated with each of the posture state definitions. The resulting weighted therapy parameter values may be used to derive a blended therapy parameter value for use in delivering therapy to the patient. The patient's posture state may be expressed in terms of a blending of the multiple posture state definitions.