| |
Abstract:
Parsing performance is widely recognized as being determined
by the surface distance (or locality) of linguistic entities.
Hawkins (1994) and Gibson (1998), among others, proposed
locality-based metrics to estimate on-line processing cost during
sentence perception. According to Hawkin's EIC principle (Early
Immediate Constituents), processing suffers from establishing the
final immediate constituent of a phrase marker being delayed.
Gibson's (1998) account of on-line integration cost, on the other
hand, is based on the number of new discourse entities intervening
the item to be integrated and its dependents.
I will present a series of corpus, off-line and on-line
(self-paced reading and eye-tracking) studies to establish
processing cost differences in sentences containing extraposed and
non-extraposed Relative Clauses variing in a. length of the RC (2),
and b. distance of the RC and its host in case of extraposition
(1).
(1) RC-position, RC-distance
Lisa hat das Geld _RCpos1_ a. abgehoben _RCpos2_, ...
b. von ihrem Konto abgehoben _RCpos2_, ...
c. von ihrem Konto bei der Sparkasse
abgehoben _RCpos2_, ...
Lisa has the money _RCpos1_ a. withdrawn _RCpos2_, ...
b. from her account withdrawn _RCpos2_, ...
c. from her account at the Sparkasse
withdrawn _RCpos2_, ...
"Lisa has withdrawn the money {from her account ...}, ... "
(2) Relative Clause
a. ..., das sie gespart hat
..., that she has saved
b. ..., das sie fuer ein neues Auto gespart hat
..., that she has saved for a new car
c. ..., das sie seit drei Jahren fuer ein neues Auto geespart hat
..., that she has saved for a new car since three years
While the corpus data and the off-line acceptability study
revealed strong effects of locality as expected by both Hawkins and
Gibson, a divergent picture emerges from the on-line data.
Locality-based integration cost predicts the processing time for 1.
The verb, (i) to increase with the length of the adverbial
locative, (ii) to be higher if it follows the RC, and (ii) to
increase with the length of the RC that it follows. Secondly,
integration cost for the relative pronoun should be higher if (i)
the RC is extraposed and (ii) the extraposition distance increases
with the length of the adverbial locative. While both self-paced
reading and eye-tracking were sensitive to subtle lexical and
positional effects, none of the locality based predictions could be
confirmed.
Taken together, the results suggest that locality does in fact
influence sentence production preferences (Wasow, 1996), whereas
integration cost in sentence perception seems unaffected. I will
suggest modifications to Gibson's approach compatible with the
data.
Gibson, E. (1998). Linguistic complexity: locality of syntactic
dependencies. Cognition, 68, 1-76.
Hawkins, J. A. (1994). A performance theory of order and
constituency. Cambridge Studies in Linguistics, CUP
|
