Lo Bianco (Curr Issues Lang Plan 9(2):155-178, 2008) proposed an "ensemble of [three] activities" that comprise language policy: the "textual," "discursive," and "public performance" (p. 157). When expressed in language, the current study proposes that each of Lo Bianco's three activities (text, discourse, performance) becomes a register--a variety of language associated with a given social situation (Biber and Conrad in Register, genre, and style. Cambridge University Press, Cambridge, 2009). Building on a previous study that used corpus-based methods to identify five language ideologies in a 1.4 million word corpus of language policy texts (Fitzsimmons-Doolan in Corpora, 9: 57-82, 2014), this study asks, "Is there variation in the language ideologies expressed in a corpus of institutional language policy texts attributable to language policy register?" Using inferential statistics, groups of texts coded by language policy register (i.e., language policy documents, discourse about language policy, institutional models of language policy, and lists) were compared for expression of each of the five previously identified language ideologies. For four out of five of the language ideologies, there were significant differences among language policy registers with much of the ideological work being done by texts coded as institutional models of language policy. The findings suggest the viability of Lo Bianco's categories and the identification of a new register category of language policy texts-lists. Furthermore, findings suggest the importance of analyzing institutional models of policy for language ideology scholarship and support for a multi-dimensional, layered model of language policy.