This paper presents a longitudinal study on the acquisition of first, second, and third person pronouns in twelve French-speaking and twelve English-speaking children. Comprehension and production data were collected every two months, beginning when the subjects were aged 1;6 and ending once pronouns were fully acquired. Three hypotheses concerning the rules children develop in learning pronouns were tested: (1) the person-role hypothesis (Charney, 1980), (2) the speech-role hypothesis (Clark, 1978), and (3) the name hypothesis (Clark, 1978). An analysis of children's pronominal confusion when they were addressed listeners as well as when they were non-addressed listeners was performed. The results indicated that the mastery of pronouns did not follow the developmental sequence predicted by the speech-role hypothesis; they provided evidence for the person-role hypothesis only when children were speakers, and partially supported the name hypothesis. The data also suggested that pronominal confusion is not a rare phenomenon among children tested in a non-addressee context. Finally, effects of child gender and native language were observed. Possible interpretations of the data discussed.