| |
Abstract:
Cross-serial dependencies in Dutch involving three verbs (1)
lead to major processing difficulty compared to constructions
involving only two verbs (2,3) (Bach et al 1996), even if the
number of NPs is controlled for (3), cf. (Dickey & Vonk, 1996).
(1) NP1 NP2 NP3 V1 V2 V3
(2) NP1 NP2 V1 V2
(3) NP1 NP2 NP V1 V2
These facts can be accounted for both in an interference model
as proposed by Lewis (1996, 1998), and a discourse-based complexity
metric as proposed by Gibson (1998). These two accounts can be
distinguished by manipulating the nature of the NPs involved:
according to Gibson's metric, integrating the verbs with their
subjects in triple cross-serial dependencies should become easier
to process when NP2 and NP3 are inherently given NPs, in particular
first and second person pronouns compared to discourse-new NPs,
such as proper names. The nature of the NPs should have less effect
on double dependencies. Lewis, on the other hand does not predict
such an interaction.
An offline rating and a self-paced reading study were conducted
comparing sentences with the structures given below:
(4) [P NP ] aux pronoun1 pronoun2 V1 V2
(5) [P NP ] aux Proper_Name1 Proper_Name2 V1 V2
(6) NP1 aux pronoun2 pronoun3 V1 V2 V3
(7) NP1 aux Proper_Name2 Proper_Name3 V1 V2 V3 (8)
As expected, 2-verb sentences (4,5) were judged to be easier
than 3-verb sentences (6,7). Sentences containing pronouns were
easier than sentences with proper nouns in the 2-verb condition
only. This is likely to be due to differences before the verb:
pronouns were also read faster than proper names before the verbs
in the 2-verb conditions, suggesting that the sentence initial PP
is retained in memory and the cost of this retention is modulated
by the (discourse) nature of the intervening NPs in terms of
Gibson's model.
At V2, sentences with pronouns were read faster than those
containing proper names for 3-verb sentences (6,7) but not for
2-verb sentences (4,5). This is in accordance with Gibson (1998),
and in contrast to the interference model proposed by Lewis (1996,
1999).
Finally, reading times at position between the aux and V1
increase with each NP, replicating results from Dickey &Vonk.
This can be accounted for in terms of on-line reanalysis from a
transitive to a ditransitive to a double verb construction.
Bach, E., Brown, C. & Marslen-Wilson, W. (1996). Crossed and
Nested Dependencies in German and Dutch. A Psycholinguistic Study.
Language and Cognitive Processes, 1, 249-262.
Dickey , M. & Vonk, W. (1996). Center Embedded structures in
Dutch: An on-line study. Poster Presented at the 9th CUNY
conference.
Gibson, E. (1998). Linguistic complexity: Locality of syntactic
dependencies. Cognition, 68, 1-76.
Lewis, R. (1996). Interference in short-term memory: The magical
number two (or three) in sentence processing. Journal of
Psycholinguistic Research, 25, 93-115.
Lewis, R. (1998). Interference in working memory: retroactive and
proactive interference in parsing. Paper presented at the 11th CUNY
conference.
|
