Recognition judgments to the non-antecedents of a repeated-noun anaphor are slower and less accurate after than before the processing of the anaphor. Disagreement exists as to whether this pattern of performance reflects a bias shift carried out by a memory process associated with the recognition of a word that has previously occurred in the discourse or an accessibility reduction carried out by a comprehension process specifically tied to the resolution of anaphors. The results of 3 experiments supported the memory-based account. (Contains 5 tables.)