Most instances of grammaticalization have been shown to arise in restrictive contexts (cf. Bybee et al. 1994). The persistence (Hopper 1991) of linguistic contexts raises theoretical and methodological issues for historical corpus research. What is the appropriate unit of linguistic context? How long do contexts remain relevant in the history of specific constructions? In quantitative work should “bridging contexts” (Heine 2002) and “critical contexts” (Diewald 2002) that enable grammaticalization be counted after grammaticalization has set in? I argue that ambiguous contexts (‘co-texts’ broadly defined to include prior and following discourse), if attested, should be counted (contra Eckardt 2006), because they persist as part of the ecology of a newly grammaticalizing construction and should therefore be considered an integral component of diachronic corpus research. Data discussed involve the development of motion-with-a-purpose BE going to V into an auxiliary of the future as evidenced by the Early Modern English part of Helsinki Corpus, and by the first fifty years of Proceedings of the Old Bailey.