Experimental evidence shows that many physiological processes abide by nonlinear dynamics and evidence chaos. Novel conceptual models postulate chaotic phenomena in psychological processes as well. The challenge in empirically testing these models is to develop measures that are sufficiently precise to permit nonlinear dynamical analyses and that also say something meaningful about psychological phenomena. Toward this end, we examined the spontaneously occurring autonomic activity of a patient during psychotherapy. Phase portraits were constructed of the patient's heart rate data for each of several therapy sessions. The flows within these phase spaces were visually segmented and assigned to one of four previously defined trajectory classifications. These trajectories were quantitatively analyzed, and the corresponding clinical material was assessed qualitatively. The trajectories in the physiological phase space were recognizable, recurrent, and robust; they appeared coupled to ongoing psychological processes. Possible refinements in this typology are discussed and suggestions are made for potentially using such analyses to track and navigate among psychophysiological states.