|
Abstract:
Languages that allow for word order variation pose a
challenge for language processing models. The question is how the
parser handles the integration of a dislocated constituent into the
syntactic structure. There are two alternative points of view: the
syntax-first and the lexicalist approach. According to the
syntax-first position (Clifton & Frazier, 1989), the parser
interprets a dislocated constituent as filler and searches actively
for a trace or "gap". In contrast, the lexicalist approach
(Pickering & Barry, 1991) claims that the parser does not make
use of traces at all but associates the dislocated constituent
directly with the main verb.
Most studies in this area focus on dislocated direct objects. In
an SVO language like English the base position of the direct object
follows the verb immediately. Thus one may argue that it is not
obvious that any effects observed in the trace position are due to
the trace itself and not to the preceding verb. Clahsen &
Featherston (1998) conducted a study on the processing of German
scrambling sentences. German is an SOV language where the trace
position of the direct object immediately precedes the verb. The
authors found a cross-modal priming effect in the trace position
and interpret this finding in favour of the syntax-first
approach.
However, in our opinion this conclusion is not unequivocal. As
Bader (1996) has demonstrated, listeners of German do not delay
their syntactic analysis until they have encountered the verb. We
suggest that listeners anticipate the position of the verb and
reactivate the candidates for the verb's argument structure in
advance. In order to avoid the possible confusion of trace position
and anticipation of the verb's argument structure we investigated
German ergative constructions with neutral focus which are a
critical case for deciding between gap-filling and direct
association approaches. We compared two kinds of sentences:
(1) Topicalization of the DP marked for nominative:
Der Krug(i) ist [einem jungen Richter des Berliner GeRICHTS t(i)
zerbrochen.]F
the+NOM jug+NOM is a+DAT young+DAT judge+DAT the+GEN Berlin+GEN
court+GEN broken
'The jug is broken by a mistake of a young judge of the Berlin
court.'
(2) Topicalization of the DP marked for dative:
Dem Richter(i) ist [t(i) ein bunter Krug aus schwerem STEINgut
*t(i) zerbrochen.]F
the+DAT judge+DAT is a+NOM colourful+NOM jug+NOM of heavy+DAT
earthenware+DAT broken
'By a mistake of the judge a colourful jug made from heavy
earthenware is broken.'
Whereas in (1) the position before the main verb is a trace
position (t(i)), in (2) it is not (*t(i)). Thus only an effect in
the position before the verb in sentence type (1) would provide
empirical evidence for the syntax-first approach. This conclusion
would not be verified if the effect occurs in both sentence types.
We put this to test in a cross-modal priming experiment measuring
lexical decision times for identical and unrelated targets both in
the position before the verb as well as in a control position.
Bader, M. (1996): Sprachverstehen : Syntax und Prosodie beim
Lesen. Opladen: Westdeutscher Verlag.
Clahsen, H. & Featherston, S. (1998): Antecedent priming at
trace positions: Evidence from German scrambling. Essex Research
Reports in Linguistics, 23.
Clifton, C. & Frazier, L. (1989): Comprehending sentences with
long-distance dependencies. In G. N. Carlson & M. K. Tanenhaus
(eds.): Linguistic structure in language processing. Dordrecht:
Kluwer.
Pickering, M. & Barry, G. (1991): Sentence processing without
empty categories. Language and Cognitive Processes, 6, 229-259.
|
