In Freud's thought repetition was an obstacle to change and a reason to be cautious about what analysis might accomplish. Indeed, since it is hard to know what transference is if it is not repetition, the compulsion to repeat is the central event in an analytic treatment. Since Freud, however, repetition has become a vague and often harmless notion. It could regain clinical vitality if it were understood as a patient's being caught in natural, linear causality and thereby cut off from the possibilities the analytic therapies try to present to patients. These possibilities can be seen as features of a uniquely human experience of time, different from natural time and implying something reminiscent of Freud's death instinct. By rethinking repetition through Kierkegaard's use of that category, change in treatment comes into focus as an unpredictable, largely uncontrollable, and sometimes whimsical willingness to try something new much more than it is a function of the analyst's interventions or the therapeutic relationship. These certainly matter, but less than we might wish, and they are never decisive. A case study and other clinical material illustrate the argument.