The goal of this study is to observe the use of referential expressions used by patients with Alzheimer's disease (AD) at both discourse and syntactic levels. Discourses contain both new information and old, given information. One important aspect to produce a coherent discourse is to select an appropriate form to introduce new information and to refer to old or given information. Given information is typically encoded via pronouns while new information via full lexical NPs. The challenge for the speaker is to decide to what extent a referent is accessible to the hearer (Meyer & Bock, 1999). The speaker then has to select the appropriate form accordingly, ranging from a reduced form (pronoun) when the referent is assumed to be highly accessible or given, to a more explicit lexical form when it is assumed to be inaccessible (Givón, 1983). In addition, there is a strong tendency in conversational speech for given, accessible information to occur in subject position (Chafe, 1994). AD patients have been reported to show deficits in discourse production. Indeed their speech contains a high proportion of empty utterances and overuses pronouns (Almor et al. 1999, Berrewaerts et al; 2003). However some studies do not replicate those findings (Ska & Duong, 2005; Chapman et al., 1995). Following previous work, we hypothesize that AD patients should produce a higher proportion of third person pronouns and more pronouns with ambiguous reference. We also expect both groups to produce more pronouns in subject position. Finally we predict that because AD patients tend to shift conversational topics more often than controls (Berrewaerts et al., 2003), distances between pronouns and antecedents should be shorter. Twenty participants diagnosed with Alzheimer's disease and twenty healthy controls matched for age, sex, level of education and socioeconomic status participated in the study. Spontaneous speech data were elicited by asking participants about biographical details. The data were transcribed and the following variables were examined: the proportion of third person pronouns over the total number of NPs, the syntactic position of each pronoun, the proportion of ambiguous pronouns, the distance between the pronoun and its antecedent. Our hypotheses are partially validated: contrary to previous findings, we do not observe an overuse of pronouns by AD patients. Our other hypotheses show significant differences between AD and controls. Finally we address some methodological concerns that we claim should be carefully taken into consideration in order to allow for comparisons across studies.