description, and in particular whether it is a. Here the second step would not follow, for the actual referent of b need not be such that it is contingently identical to a. In [5], Harold Noonan considers argument (A). Although its conclusion would be embraced by Kripke and followers, Noonan resists. He offers Gibbard's example of Lumpl, a lump of clay, and Goliath, a statue composed of that clay at all times at which they both exist [2]. Noonan maintains that Lumpl and Goliath are identical: after all, they occupy exactly the same position through exactly the same time. But they might not have been identical - Lumpl could have been formed into a ball and not destroyed, while Goliath could not. Thus he concludes that there are cases of contingent identity, and any argument to the contrary must be flawed. Noonan seeks to fault argument (A) by locating a characteristic of modal predicates which renders the inference invalid. They are what he labels 'Abelardian': predicates 'whose reference is affected by the subject term to which they are attached' ([5], p. 188). As a non-modal case, take Abelard's own example of 'was made by Socrates'. We would say that Goliath, the statue, was made by Socrates but deny that Lumpl was made by Socrates - for he did not make the clay constituting it. But, Noonan goes on, when preceded by a term invoking a statue, the relevant predicate means 'was made by Socrates to be a statue', whilst it means 'was made by Socrates to be a lump of clay' when the subject term is 'Lumpl'. The identity claim is not threatened. Sentences predicating the same Abelardian predicate of different subject terms, 'a' and 'b', can effectively be saying different things about a and b (whether or not a = b), and its applicability to a but not to b does not establish the non-identity of a and b.