This dissertation advances our understanding of the historical evolution and grammatical structure of English do-support through the application of novel historical data to this classical problem in historical syntax. Do-support is the phenomenon in English whereby a pleonastic auxiliary verb do is inserted in certain clause types. The phenomenon is characteristic of the modern language, and there is robust evidence that it emerged beginning in roughly the year 1500. The fine quantitative details of this emergence and the variation it engendered have been an object of study since Ellegård (1953). From the standpoint of generative grammar, Roberts (1985), Kroch (1989), and many others have treated the emergence of do-support as a closely-following consequence of the loss of V-to-T raising in the 15th and 16th centuries. Taking a cross-linguistic perspective, I show that though the totality of English do-support is uncommon in other languages, the phenomenon may be seen as the combination of several discrete building blocks, each of which is robustly attested. From this perspective, a question is raised about the genesis of English do-support: given that the present-day phenomenon is evidently composed of several separate subcases, why should its cause be attributed solely to the loss of V-to-T raising? I argue that the earliest emergence of do-support in English is in fact attributable to a different source: a usage of do as a marker of external arguments. This explanation addresses the following points, which under earlier accounts were unexplained: * The different behavior of do-support across argument structure types. * The appearance of do-support in affirmative declaratives at a peak rate of 10%, much more than can be attributed to emphatic assertions. * The emergence of do-support from a Middle English causative. This "intermediate do" spread through the language until roughly 1575, when the loss of verb raising triggered an abrupt reanalysis which transformed the argument-structure marking do into its modern form.