Inherent in the interlanguage hypothesis is the assumption that the language of second language acquirers is systematic. This article reviews the notions of system and variability as applied to interlanguages, especially in the area of syntax. The article summarizes the findings of Huebner (1983a) with respect to variability over time in the forms used to mark referential status of discourse participants, using Bickerton's (1981) proposed universal features for referentiality. It then examines, from the same perspective, new data from the same informant collected approximately 20 months after the end of the initial study. Finally, the new corpus is examined in light of Givón's (1982) Quantity Universal. Three measures are applied to those data to determine topic continuity: degree of referential distance, degree of topic persistence, and degree of importance in discourse. Data from the original study reveal the nonlinear nature of linguistic change in an interlanguage. Both the original and the new data support claims made about the primacy of the specific referent/nonspecific referent distinction (Bickerton 1981). The study also suggests that degree of topic continuity can explain residual data.