Although Verb-Object (VO) is the basic unmarked constituent order of predicates in Present-Day English, in earlier stages of the language Object-Verb (OV) is the preferred pattern in some syntactic contexts. OV predicates are significantly frequent in Old and Middle English, and are still attested up to 1550, when they “appear to dwindle away” (Moerenhout & van der Wurff 2005: 83). This study looks at OV in Early Modern English (EModE), using a corpus-based perspective and statistical modelling to explore a number of textual, syntactic, and semantic/processing variables which may account for what by that time had already become a marked, though not yet archaic, word-order pattern. The data for the study were retrieved from the Penn-Helsinki Parsed Corpus of Early Modern English (1500–1710) and the Parsed Corpus of Early English Correspondence (c.1410–1695), the largest electronic parsed collections of EModE texts. The findings reveal a preference for OV in speech-related text types, which are less constrained by the rules of grammar, in marked syntactic contexts, and in configurations not subject to the general linearisation principles of end-weight and given-new. Where these principles are complied with, the probability of VO increases.