Infinitive usage in Biblical Gothic

There are at least six positions at which Gothic infinitives may be attached: (1) VI$\sp{\prime\prime},$ the external argument (VI Specifier; see Berard 1993a-b for this and the following); (2) VI$\sp\prime,$ the VI Complement node; (3) A$\sp\prime,$ the AP Complement node; (4) N$\sp\prime,$ the NP Complement node; (5) N$\sb0,$ the node that cojoins to NP; and (6) VI, the head verb node. Gothic permits articularization of subject and, perhaps, object infinitives only as a very exceptional response to unusual length or complexity of expression in situations where the infinitival clause has functional Theme status. VI$\sp\prime$ adjuncts with final semantics regularly take the form (V$\sb{\rm inf}$) when motivated by matrix verbs with motion semantics. Verbs with non- motion semantics regularly take $\lbrack du + {\rm V\sb{inf}}\rbrack$ final adjuncts. The primary motive for exceptions to this rule is the apparent need to avoid (accusative) Default Case Marking of the subject of the infinitive (see Berard 1993c), in which situation final adjuncts are formulated with ei + optative. Almost all AP complement infinitives, the vast majority of VI complement infinitives, and a considerable majority of subject infinitives and infinitives associated with NPs are bare. Except in the case of a periphrastic future, all subject control predicates have bare infinitives. For embedded Ss which are external arguments, there is very often a controller which is the logical subject of the embedded infinitive and which is marked dative. This dative controller is located in the matrix rather than being in the embedded S and attracted by the matrix into the dative case. This construction thus appears to be an example of raising to dative object. An infinitive may be controlled by a covert NP. Strong evidence for a wider usage of the voice-inspecific synthetic infinitive in a passive-voice sense is found in passive complements of predicate APs. Nominative morphology in the NP of which the AP is predicated, combined with expression of the Agent in a PP, does not permit an active-voice interpretation