Abstract. The following problem of Fuchs is considered: relate the abelian groups A and B assuming Tor(^4, G) = Tor(fl, G) for all reduced abelian groups G . A complete characterization is obtained in any set-theoretic universe in which E(k) is valid for a proper class of regular cardinals k .In the well-known list of 100 questions posed in [3], problem 50 asks us to relate the abelian groups A and B assuming Tor(^4, G) = Tor(P, G) for all reduced abelian groups G. There is clearly no loss of generality in assuming A , B, and G are p-groups for some fixed prime p, so hereafter the term "group" will mean "abelian p-group." Hill showed in [4] that A and B share many properties but left unresolved whether they must necessarily be isomorphic. Cutler showed in [1] that this need not be the case. Specifically, groups A and C were found such that C is a direct sum of cyclics, Tor(^4, G) = Tot(A © C, G) for all reduced G, but A ^ A ® C. Is this the only way to construct such examples? In other words, if A and B satisfy problem 50, do they necessarily differ by summands which are direct sums of cyclics?Denote the Ulm function of the group Y by fy . We will say A and B are T-equivalent if (a) fA(ri) = /b(«) for all n < co, and (b) for some direct sums of cyclics X and Y, A® X = B @Y.We show that if A and B are P-equivalent, then they satisfy problem 50 (Theorem 2). Conversely, if A and B satisfy problem 50 and there is a regular cardinal k greater than the ranks of A and B such the E(k) is valid, we show that A and B must be P-equivalent (Theorem 4; we review these set-theoretic terms later). In particular, we have a solution for problem 50 in any set-theoretic universe in which E(k) is valid for a proper class of regular cardinals. This will be true, for example, in the constructible universe. We first review the treatment of Ulm invariants contained in [5]. If Z is a valuated group and a is an ordinal, let kz(a) be the kernel of the map Z(a)/Z(a+ 1) ^ Z(a+ \)/Z(a + 2), so fz(a) is the dimension of kz(a) as a