(implicit) Argument Introduction, Voice And Causatives
This dissertation explores the syntactic and semantic properties of implicit arguments in various voice constructions, such as active and passive voice, applicatives, causatives and impersonals, using mainly Sason Arabic (SA) and Turkish as empirical starting points. I add to the typology of null arguments, further demonstrating that they do not form a homogeneous category (e.g. Williams 1985; Rizzi 1986; Bhatt and Pancheva 2017; Landau 2010). My investigation reveals (at least) four types of implicit arguments in languages under investigation in terms of their semantic properties and syntactic visibility: (i) an existentially closed passive agent, (ii) a full DP, (iii) a free variable, and (iv) an impersonal pronoun. Establishing a distinction in Turkish between two constructions with identical morphology, i.e., passive and impersonal, I show that the implicit agent of passive is unprojected, whereas the null impersonal pronoun is fully projected. I also demonstrate that purported ‘passives of passives’ in Turkish are in fact impersonals of passives, and passives cannot iterate. This follows from an analysis of passive as a subtype of Voice, the head that introduces the external θ-role (following Legate 2014). I compare the null impersonal with the overt impersonal insan ‘human’ in Turkish, indicating that they exhibit distinct behavior. I also provide a syntactic analysis of the passive that confirms and captures the generalization that passive cannot iterate (Perlmutter and Postal 1977). The approach to passive adopted in the dissertation predicts that an active-passive-like alternation should be available to other functional categories, such as ApplP or CauseeP. Accordingly, I investigate several morphological and periphrastic causative constructions from SA and Turkish, arguing that this prediction is borne out. While all the causatives embed a second VoiceP, the behavior of this VoiceP varies across causative constructions: some are like the canonical, agentive VoiceP, whereas the behavior of others warrants identifying them as distinct categories, specifically VoicecauseeP or CauseeP. Furthermore, the investigation of ‘make’ causatives in SA reveals that the embedded agent may be present (i) as a free variable on thematic, active Voice head (à la Heim 1982) without needing a specifier or (ii) as a full DP, which is separated from its licensor by a phase domain and needs to Ā-move to be (Case)-licensed.