The paper [6] contains a few errors in the basic assumptions as well as in the formula of Corollary 0.2. First of all, it should have been made clear at the outset that the regular model E for the elliptic curve E must be the minimal regular model, and X the complement of the origin in the regular minimal model. Similarly, the tangential base-point b must be integral, in that it is a Z-basis of the relative tangent space e * T E/Z . It could also be an integral two-torsion point for the arguments of the paper to hold verbatim.The most significant error is in the contribution of the local terms at l = p, that is, Lemma 1.2. The problem is that a point that is integral on X may not be integral on a smooth model over a field of good reduction. As it stands, the lemma will only apply to points that are integral in this stronger sense.However, to get immediate examples, one can replace the lemma by
Lemma 1.2 . Suppose the Neron model of E has only one rational component for each prime. (Equivalently, the Tamagawa number is one at each prime.) Then the mapis trivial for every l = p.Therefore, for the functionconstructed via the refined Massey product, we get The assumption can be easily verified if the elliptic curve has square-free minimal discriminant, since the Neron model will then have only one geometric component in the special fiber. We point out that the integral j-invariant hypothesis is no longer necessary in this version.