Two recent proposals link the use of nonagreeing "don't" to the Root Infinitive (RI) Stage. Guasti & Rizzi (2002) argue for a misset parameter involving how agreement is spelled out. Schütze (2010) proposes that Infl is underspecified in child language and that "do" surfaces to support the contracted clitic/affix "n't." Both proposals obtain partial support from the Sarah corpus (Brown 1973), yet neither proposal drew on Sarah's parents' use of nonstandard, nonagreeing "don't" as a possible explanation for Sarah's production. In this article I argue that much of the nonagreeing "don't" produced by Sarah is not part of the RI Stage. Once Sarah's data are removed from the analysis, the remaining data support Schütze's proposal.