|
Abstract:
Normal silent reading is usually accompanied by a process of
phonological coding. As a result of phonological coding, written
sentences are not only assigned a syntactic structure but also a
phonological and prosodic structure. Earlier work (Bader, 1994)
has shown that revision processes in garden-path sentences are
constrained by prosodic information computed during phonological
coding. In this paper I will argue that the computation of a
prosodic structure during reading is an important factor for
certain first-pass parsing decisions as well. Experimental
evidence for prosodic influences on syntactic processing during
reading comes from reading experiments (self-paced reading,
speeded grammaticality judgements) investigating ambiguous
sentences differing with respect to the location of sentence
accent.
A first series of experiments investigated the effect of focus
particles on the processing of German extraposed relative clauses
which could associate either to the subject or the object (cf.
(1)).
(1) ... (focus particle) subject object verb
relative-clause
Sentences without focus particles exhibit a strong tendency
for associating the relative clause to the object which is the
most recent host and which bears default sentence stress. This
tendency is reflected in a garden-path effect for sentences in
which syntactic factors force the relative clause to associate to
the subject. Introducing a focus particle in front of the subject
had the effect that association to the subject was facilitated
whereas association to the object became more difficult. Since a
focus particle directs sentence accent on the constituent
immediately following, this finding indicates that a forced
accentuation of the subject increased the likelihood of subject
associations during first-pass parsing.
A second series of experiments investigated the lexical
ambiguity of the German word "mehr" which has either a temporal
("anymore") or a comparative meaning ("more"). In its comparative
meaning, "mehr" can fill an argument slot of a verb whereas it is
an adverbial in its temporal meaning. Theories positing a
preference for argument attachment over adverbial attachment
would predict a preference for the comparative reading of "mehr."
Contrary to this prediction, we found a strong preference for the
non-argument temporal reading. This finding can be explained with
reference to a prosodic difference between the two readings of
"mehr": "mehr" must be stressed in its comparative meaning but
unstressed in its temporal meaning. Given the prior finding that
function words are preferentially read unstressed, a preference
for the temporal reading of "mehr" follows under the assumption
that prosodic information can constrain first-pass parsing..
Bader, M. (1994). Sprachverstehen: Syntax und Prosodie beim
Lesen. Doctoral Dissertation, University of Stuttgart.
|
