Movement and Locality in Sundanese Wh-Questions

Movement and Locality in Sundanese Wh-Questions

By Eri Kurniawan and William Davies
Nominal wh-questions (non-prepositional arguments) in Indonesian-type languages have received two principal analyses in recent years. Saddy (1991) for Indonesian and Cole & Hermon (1998) for Malay have proposed in situ, long distance movement, and partial movement analyses for the Malay sentences in (1-3), respectively. Madurese apparently has the same three structural options, but Davies (2003) has proposed that there is no interclausal wh-movement in Madurese and that all nominal wh-arguments are generated in situ. Sundanese appears to have the same array of structures as these other languages (4-6), so it is of some interest to determine whether either analysis can adequately account for the facts. We will conclude that there is no interclausal wh-movement for Sundanese, and that the apparent cases of movement, e.g. (5-7) are best analyzed in other ways.
We first lay out what the in situ and movement structures entail, adopting the analysis in Cole & Hermon (1998) of unselective binding of the in situ wh-element by a matrix Q feature. We then focus on the long distance movement structure. As is true of both Malay/Indonesian and Madurese, any verb that intervenes between a fronted wh-nominal and its apparent thematic position in an embedded clause must not be in the active voice, thus (8) is ill-formed. In addition, neither can there be a voice neutral verb in Sundanese, (9). Any verb intervening between a fronted wh-nominal and its embedded thematic position obligatorily occurs in the passive (6,7), a fact that a movement analysis must account for.
An appropriate analysis must also take into consideration the fact that not all complement-taking predicates allow all of the structures. For instance, control predicates are incompatible with long distance wh-movement. With embedded nominals, object control predicates countenance only in situ questions (10). Given standard analyses of control structures as having embedded CP structure, this is unexpected. Any adequate account of the facts must include an explanation for this inability of wh-nominals to move out of dependents of control predicates. Two other classes of matrix predicates present slightly different structural patterns. With verbs such as percaya ‘believe’, sangka ‘think’ and others, the embedded wh-nominal may occur in sentence-initial position (11a) (the long distance structure) or in situ in the embedded clause (11b). With verbs such as carita ‘say’, téléla ‘mention’, and others, the wh-nominal may occur in embedded position (12a) or sentence-initial position (12b). However, when the whnominal is sentence initial, the matrix verb obligatorily occurs with the applicative suffix -keun. When this affix is absent, the sentence is ungrammatical (12c).
We propose an analysis for the movement structure which includes no interclausal whmovement, which is consistent with the analysis of cleft structures. We show how the restriction against interclausal wh-movement explains the impossibility of long-distance movement in control clauses and the peculiarities with the percaya-type vs. carita-type matrix predicates, as well as the requirement that intervening verbs occur in passive voice. We also address additional facts concerning plural verbal marking and a subject restriction on moved wh-nominals.
Please contact erikurn@gmail.com for further reading.