Two recent papers Wexler, Schütze & Rice (1998) and Rispoli (1999a) reach different conclusions concerning the relationship between finiteness marking and pronoun case errors. Wexler et al. claim they are linked, whereas Rispoli finds little evidence for such a linkage. Wexler et al. discounted data from children who made pronoun case errors in 100% of their attempts at a subject pronoun, whereas Rispoli (1999a) included data from such children. This methodological difference may account for our differing conclusions. Schütze (2001) defends the omission of these data, claiming that children who make a pronoun case error 100% of the time do so because the correct pronoun case form is not in the child's productive inventory. Longitudinal data is presented showing the inadequacy of this assumption. Children may err without variation AFTER a period of variation, indicating that these children have knowledge of the correct form. Apparent methodological differences between the two papers reveal deeper theoretical biases. The paradigm building approach taken by Rispoli views longitudinal variation of this sort as a reflection of lexical retrieval principles at work in a developing paradigm. In contrast, the Agr/Tns Omission Model of Wexler et al. finds such variation an uncomfortable inconvenience.