This paper is concerned with convex subsets of finite dimensional vtctor spaces, over the field of real scalars. As in [10, p. 244] and [20] we say that a compact convex set A, symmetric about the origin, is reducible, if there is a nonsymmetric closed convex set B for which A = B - B. The latter term denotes the set of all differences, {x-y: x, y є B}. Equivalently, A is reducible if, and only if, A =1/2 (B - B) for some B which is not a translate of A. If the identity A = B - B is only possible when B is centrally symmetric, then A is irreducible. If A is symmetric about a point other than the origin, we can say that it is (ir)reducible when it is the translate of an (ir)reducible set which is symmetric about the origin. It is well known that a parallelotope of any dimension is irreducible [6, Hilfssatz 3], that any 2-dimensional convex body other than a parallelogram is reducible [9, p. 217], and that euclidean balls of any dimension (other than one) are reducible [4, Ch. 7]. For more general convex bodies, the determination of reducibility is not a simple problem. Shephard [20] showed that a set is reducible if, and only if, it has an asymmetric summand, and he used this to study reducibility of polytopes. The main purpose of this paper is to give a new condition, necessary and sufficient for a symmetric polytope to be reducible. This condition may be expressed in the form: does a certain finite family of linear equations have a nontrivial solution? Thus, to determine the reducibility of a given polytope, it suffices to find the rank of some matrix. Using our criterion, we are able to describe some large families of irreducible polytopes. For example, every n- dimensional symmetric polytope with 4n – 2 or fewer vertices is irreducible (unless n = 2). We also establish the existence of irreducible, smooth, strictly convex bodies.