The authors examine the ways in which college students, in non-clinical experiments, judged different interactions by therapists as indicating the degree to which therapists understood their patients. Therapist interactions which predicted future patient behavior were judged to exhibit the greatest degree of understanding. Those which restated the patient's problems (reflective) or admitted not understanding the patient (naïve) were judged not to differ from each other in revealing the therapist's understanding. On the basis of this experiment, the authors question the usefulness of the latter two types of therapist interactions.