extracta mathematicae Vol. 15, N´ um. 2, 391 – 420 (2000) Sobczyk’s Theorems from A to B elix Cabello S´ anchez , Jes´ us M.F. Castillo , David Yost Dpto. de Matem´aticas, Univ. de Extremadura, 06071-Badajoz, Spain e-mail: fcabello@unex.es, castillo@unex.es, dyost@unex.es AMS Subject Class. (1991): 46B03, 46B20, 46M10 1. Sobczyk’s theorem and how to prove it Sobczyk’s theorem is usually stated as: Every copy of c 0 inside a separable Banach space is complemented by a projection with norm at most 2. Never- theless, our understanding is not complete until we also recall: and c 0 is not complemented in . Now the limits of the phenomenon are set: although c 0 is complemented in separable superspaces, it is not necessarily complemented in a nonseparable superspace, such as . The history of complemented and uncomplemented subspaces of Banach spaces is traced back in another article of this volume [48]. It is probably worth mentioning that it starts with two propositions: Every closed subspace of a Hilbert space is complemented by a norm one projection and 1 contains uncomplemented subspaces. The first result easily follows by proving that the metric projection onto a closed subspace acts linearly; the second result holds since the kernels of quotient maps 1 X are necessarily uncomplemented when X has not been previously chosen a subspace of 1 (and recalling that all separable Banach spaces are quotients of 1 ). A more interesting question for us is: why should one suspect that c 0 is complemented inside separable superspaces? A previous result in this dir- ection had been proved by Phillips [47]: Every copy of inside a Banach space is complemented by a norm one projection. In other words, the spaces (Γ) are injective. Since it was well known (and can be easily proved) that every Banach space is isometric to a subspace of some (Γ) it is clear that the injective spaces are precisely the (Γ) spaces and their complemented subspaces. In order to determine all the injective spaces the story starts with * Supported in part by DGCYT project PB 97-0377. 391