Given a certain commutative diagram of groups and monomor-phisms, does there necessarily exist a group in which the given diagram is essentially a diagram of subgroups and inclusions? In general, the answer is negative, but J. Corson, and Gersten and Stallings have shown that in the case of a "non-spherical triangle" of groups the answer is positive. This paper improves on these results by weakening the non-sphericality requirement.