Anaphors, such as pronouns, are linguistic devices which allow an author to abbreviate subsequent mentions of a concept. This paper reports our preliminary findings of the effect on retrieval performance of replacing anaphors with the terms to which they refer. Documents retrieved by queries were edited so that all anaphors were resolved. In a post-retrieval experiment, the resolved and unresolved versions of the documents were placed in order of predicted relevance and compared with the users' relevance judgments. The results are mixed. For the majority of queries, there was no difference in the rankings of the two sets of documents. For a few queries, resolving anaphors had a beneficial effect on rankings, while for other queries the effect was detrimental.