Despite elusive evidence, it is widely held that code-switching promotes grammatical convergence. This talk puts forward quantitative diagnostics of grammatical similarity and difference by using structural variation in speech. The poster child for convergence has been variable subject pronoun expression in Spanish toward English, which is classified as a non-null subject language (e.g., Heine and Kuteva 2005:70; Otheguy and Zentella 2012). Variation patterns are uncovered in a 300,000-word corpus capturing the spontaneous bilingual speech of members of a long-standing community in northern New Mexico (Torres Cacoullos and Travis 2018). Approximately 10,000 tokens of the variable are extracted from this bilingual corpus, and from comparable monolingual corpora of both Spanish and English. The most direct test of the hypothesis of convergence via code-switching is by comparing bilinguals’ own use of the two languages. Four independent analyses of the linguistic conditioning of variable subject expression show that the bilinguals’ Spanish and English differ from each other and align with their respective monolingual benchmarks. Moreover, comparisons in the presence and absence of code-switching reveal that bilinguals maintain Spanish-particular patterns even in the context of proximate use of English.