This article provides additional evidence for the universality of Rizzi's (1990) anaphor agreement effect, under which the ungrammaticality of nominative anaphors in English, Italian, and Icelandic is due to the presence of agreement. Languages without agreement are shown to allow nominative anaphors. Objective anaphors cannot be associated with agreement, unless the agreement is a special anaphoric form. Superficial counterexamples to Rizzi's proposal are shown not to be problematic. The relative merits of two formal accounts outlined by Rizzi (1990) are discussed. Finally, it is suggested that the anaphor agreement effect can be a diagnostic for the presence of covert agreement.