In the third of his excellent papers on the borderline character, Kernberg focuses on issues of therapy. He feels that for most borderline patients neither classical analysis nor supportive therapy is ideal; rather, a special form of analytic psychotherapy is indicated. A full transference neurosis (or psychosis) is not developed, and resolution does not occur via interpretation alone. Special attention is paid to the negative transference, and to the primitive defense patterns seen in borderline patients. Of the latter, Kernberg particularly focuses on splitting, primitive idealization, projective identification, denial, omnipotence, and devaluation. Therapy is generally conducted in the face-to-face position, and hospitalization is used as needed. The author does feel that there is a small group of borderline patients, particularly narcissistic characters, in whom classical analysis is the treatment of choice.