On the nature of the stress function for thermally activated flow