- Woolford, E. 1991. 'VP internal subjects in VSO and Non-configurational Languages', Linguistic Inquiry 22:3, summer. 503-540.
Abstract: "Since the influential work of Emonds (1979; 1980) and Sproat (1983; 1984; 1985), it has been a fairly standard assumption that verb-subject-object (VSO) languages are derived from an underlying subject-verb-object (SVO) structure by means of verb movement. However, recent work such as that of Kuroda (1988) and Koopman and Sportiche (1988) has introduced the possibility of generating the subject inside the VP, in an otherwise configurational structure. In this article I will present evidence that at least four VSO languages have VP-internal subjects: Jacaltec, Niuean, Chamorro, and Breton. However, whereas Kuroda (1988) and Koopman and Sportiche (1988) generate subjects of SVO languages in the Spec of VP, there is strong evidence in Jacaltec, and somewhat weaker evidence in Niuean and Chamorro, that the thematic subject of at least some VSO languages is generated as a sister to the verb in a structure like (2)."
Conclusion: "Evidence from binding, word order, NP-movement, and control constructions in Jacaltec, Niuean, Chamorro, and Breton indicate that these VSO languages have VP-internal subjects in an otherwise configurational structure like that of English. Moreover, for at least one of these languages (Jacaltec), binding evidence shows that the subject is generated as a sister to the verb. The data from Niuean and Chamorro are consistent with either a V'-internal subject or a subject in the Spec of VP. Breton shows evidence of subject-object asymmetries that are generally viewed as demonstrating that the subject asymmetrically c-commands the object (although they also hold of double object constructions), and these data may indicate a Spec of VP subject position for Breton. This configurational VP-internal subject approach to VSO languages is not contradicted by the arguments of Emonds (1979) and Sproat (1985) that VSO languages have an underlying SVO structure. Those arguments were primarily directed against a completely flat, nonconfigurational approach, and they either are not relevant to deciding on the location of the subject within a configurational structure or they have been undermined by changes in the theory. The so-called nonconfigurational languages Papago and Warlpiri are also compatible with a configurational structure with a VP-internal subject. This removes much of the motivation for a parameter of configurationality in Universal Grammar to accommodate completely flat, nonconfigurational languages."