The group of homotopy classes of homotopy equivalences from a space X to itself, which we write as G(X), has been studies in recent years by Arkowitz and Curjel [1,2], the author [6,7] and several others. It is a natural quotient of the associative H-space made from homotopy-equivalcnces from X to itself under composition, and it has the same position, in the homotopy category, as the group of automorphisms of a group in the category of groups. It also has the practical use of determining how much the choice of k-invariants overdetermines the homotopytype of a space. G(X) is usually a complicated, non-abelian group, but we often have strong information -for example, we know in important cases, that it is finitely-presented (see [7]).Many simple questions are unfortunately not answered. For example: When does G(X) have torsion? When is G(X) finite? One basic problem, due to M.Arkowitz, asked whether there is a reasonable space X with G(X)=Z, the group of integers. The only case where we have the full picture is the fact that
G(K(~, n)) = Aut (g).Our goals here are the following:1) We give a characterization of when G(X) is finite, when X is a reasonable space with finite k-invariants. Our results cover H-spaces and generalize results in [1] and [2].2) We construct a space, X, with non-trivial rational homology, for which G(X) is trivial. This shows that there is no reasonable analogue of ~-theory for G(X), even when ~ is a collection of finite groups. It also shows that one cannot expect to deduce the result of 1) above from the corresponding theorem for a product of a finite number of K(GI, 2ni + 1) spaces. We also indicate how one may get a finite complex with the same properties as X.3) We construct a reasonable space X, which has four non-vanishing homotopy groups, for which G(X)= Z.4) We prove two easy propositions which show that the situation in 2) or 3) above cannot happen in the usual stable cases.5) Finally, we indicate various open questions in this area, which should be within reach of present methods.