|
Abstract:
The canonical word order for German sentential negation is
Object-Before-Negator (1):
(1) weil er den Rasen nicht betrachtete.
because he the lawn not looked at
Hauptmann (1993) argues that when the negator NICHT precedes the
object, the sentence is unambiguous and must be interpreted as
constituent negation, a fact which is due to the structural
position of the functional phrase NegP:
(2) weil er nicht den Rasen betrachtete, sondern das Haus
because he not the lawn looked at but the house
Zimmermann and Stromswold (1999) argue against this analysis and
against the existence of NegP. They show that the negator in (2) is
an adjunct which can adjoin to the object-NP or to VP making a
sentential negation interpetation possible although less preferred.
Moreover, it should be possible to override the constituent
negation bias by manipulating the verb lexical semantics (3):
(3) weil er nicht den Rasen maehte, sondern heimging.
because he not the lawn mowed but went home
The constituent negation interpretation is dispreferred because
people don't ever mow things other than lawns.
We have tested this prediction in two off-line experiments in
which we manipulated the verb lexical semantics, e.g., (2) vs. (3).
In Experiment 1, 16 German subjects completed the sentence
fragments of the type (2) and (3) in which they were supposed to
fill in the rejoinder fragment (sondern ___________). If Z&S's
hypothesis is correct, then subjects should complete (3) with a
VP-continuation and (2) with a NP-continuation. This will be
evidence that not only the sentential negation interpretation of
the Negator-Before-Object sentences is possible but it can be
preferred due to verb lexical bias. The results of Experiment 1
confirmed the Z&S analysis: the VP-completions for fragments
(3) were significantly higher (77%) than for fragments (2) (49%),
F1(1,14)=18,92, p=.001.
In Experiment 2, 30 subjects rated the acceptability of
completed sentences like (2) and (3), as well as (2) completed with
a VP. We predicted that there should be a hierarchy of ratings,
from the most acceptable to the least acceptable: (2) sondern_NP
>> (3) >= (2) sondern_VP
Type (2) sondern_NP should be rated the highest because both
verb semantics and word order favor the constituent negation
interpretation. The (2) sondern_VP should be rated lower or the
same as (3) depending on which cue, the verb lexical bias or the
word order, is stronger. The results of Experiment 2 confirmed
these predictions (4.3 vs. 3.3 vs. 3.2, on the scale from 0 to
5).
Taken together, the results of Experiments 1 and 2 show that
Type (2) sentences are initially interpreted as constituent
negation, but if the verb lexical bias forces the sentential
negation interpretation, as in Type (3) sentences, a successful
reanalysis occurs, makiing the full sentence acceptable as a
sentence in which reanalysis was not necessary (Type (3)). A
self-paced reading experiment is in progress which is designed to
pinpoint the exact location of reanalysis and will allow us to
compare the Garden-Path theory (increased RTs on the
VP-continuation in (2) sondern_VP and in (3)) andthe
Constraint-Based theory (increased RTs on the main verb).
|
