This thesis discusses the development in Proto–Indo–European, Latin and Romance of a word–formation pattern which the most adequate terminology in use dubs ‘verbal government compounds with a... Show moreThis thesis discusses the development in Proto–Indo–European, Latin and Romance of a word–formation pattern which the most adequate terminology in use dubs ‘verbal government compounds with a governing first member’; I use the shorthand ‘pickpocket compounds’. The first member of such compounds derives from a verb, while the second mostly represents its direct object: thus English pickpocket. Most English examples are functionally agent–nouns, referring to the agent of the implied verbal act. Nevertheless, they lack a suffix indicating this. By contrast, the more prolific type of compound agent–noun, represented by truck–driver, has the deverbal member second and carries an agentive suffix, ‑er. Pickpocket compounds are attested in early strata of Greek and Indo–Iranian and in medieval strata of Germanic, Slavic and Romance. Latin has around a dozen examples. The scholarly debate, continued in this thesis, has been centred round two issues: (1) the morphological make–up of the type and (2) its historical origin. (1) is preliminarily assessed in chapter 2, reviewing the basics of nominal composition and providing an account of the fundamental difference between pickpocket and truckdriver compounds. It contains a cursory discussion of Homeric Greek, Indo–Iranian and Germanic representatives. This comparative perspective on the morphology of the type is continued in chapters 3–5 on Latin and Romance. An important conclusion, and an answer to one of the most–discussed questions in the debate, is that these compounds cannot be considered univerbated imperative clauses. (2) is likewise assessed preliminarily in chapter 2, offering an outline of the history of the Greek and Indo–Aryan material. Taking the history of the Latin/Romance material up for revision, chapters 3–5 conclude that Latin offers no evidence in support of a Proto–Indo–European type. The Romance type may have an origin in Latin; however, the theoretical considerations laid out in Chapter 2 suggest that given the relevant typological conditions, new compositional types may arise spontaneously. Finally, the medieval Germanic and Slavic types result from Romance adstrate and have no bearing on Proto–Indo–European; see the summary discussion in 6.3. Show less