Abstract. If C is a small category enriched over topological spaces the category 5" c of continuous functors from C into topological spaces admits a family of homotopy theories associated with closed subcategories of C. The categories ?T c, for various C, are connected to one another by a functor calculus analogous to the <8>, Horn calculus for modules over rings. The functor calculus and the several homotopy theories may be articulated in such a way as to define an analogous functor calculus on the homotopy categories. Among the functors so described are homotopy limits and colimits and, more generally, homotopy Kan extensions. A by-product of the method is a generalization to functor categories of E. H. Brown's representability theorem.Introduction. The subject of this investigation is the kind of homotopy theory that can be done in categories 9"c of functors with values in the category of topological (more properly, separated compactly generated) spaces. The index category C is itself supposed to be enriched over 9" so that such categories as, for example, that of G-spaces where G is a topological monoid or group, are subsumed.There are two points which are to be taken into account. First, in such a functor category there are always several notions of homotopy. Two of these have been extensively considered [3,13, 14]; a third, provided by the obvious homotopy congruence, has perhaps been thought too trivial to warrant much interest. In fact there are many more-one, indeed, for each closed subcategory of the index category. Second, these functor categories are related to one another by a functor calculus formally analogous to the " tensor product" and " hom" calculus of module theory. In this calculus appear, in particular, the functors induced by composition as well as their adjoints, the Kan extensions.The objective then is to articulate these notions of homotopy and the functor calculus into a comprehensive system. As usual this is mediated by the introduction of notions of fibration and cofibration. The several "standard" forms in which this may be done-most notably Quillen's "model category" structures [10, cf. also 4]-are perhaps not adequate here. It has, rather, seemed appropriate to associate to each notion of homotopy two each of fibration and cofibration: a kind of bifurcation of function not unfamiliar in such instances as this of generalization.There are to be then weaker and stronger notions of fibration and cofibration. It is the stronger ones which will interact well with the functor calculus. To show that