André Gallois’ (1993) modified account of restrictedly rigid designators (RRDs) does indeed block the objection I made to his original account (Gallois 1986; Ramachandran 1992). But, as I shall now show, there is a deeper problem with his approach which his modification does not shake off. The problem stems from the truth of the following compatibility claim: (CC) A term’s restrictedly rigidly designating (RR-designating) an object x is compatible with it designating an object y in a world W where x (...) exists but is distinct from y.1 It follows from (CC) that the necessary (contingent) truth of a sentence of the form “α is identical with β”, where “α” and “β” are RRDs of objects x and y respectively, does not require the necessary (contingent) identity of x and y. This is borne out by Gallois’ original example (see 1986, p. 58-63). Taking W to be the actual world, we have: (1) “Mary is identical with Mary” is necessarily true; yet “Mary” RR- designates Mary and Alice, which are only contingently identical. 1 I leave it to the reader to check that in Gallois’ own example (1986, p. 58), on the view he defends (pp. 62-63), and despite his modified characterisation of restricted rigidity, RDC# (1993, p. xx), “Mary” RR-designates Alice (as well as Mary) in W, but designates Mary and not A/ice in W1. Gallois has accepted this in correspondence. In light of (CC), while I agree with Gallois (1993, p. 153) that: (7) (a=b & ◊(a≠b)) → (∃x)(∃y)(x=y & ◊(x≠y)) is a theorem given RDC#, I dispute his defence of it (on p. 153). For, if (CC) is correct, the antecedent of (7) could be true even though a and b are identical in every world (where either exists)! (shrink)