For many analysts the concept of projective identi®cation has become a pivotal reference point in their work, for like the older term countertransference with which it overlaps, it addresses the analyst's often confusing clinical experience. To those who have embraced it, it probably represents the most important advance in theory in the last sixty years. To those more sceptical about its explanatory value, it can seem like a Trojan horse that has taken over the analytic establishment. Its importance can be gauged by the number of publications devoted to it; it has been the subject of conferences (e.g. Sandler 1988) and reviews (e.g. Ogden 1979, 1994).