| |
Abstract:
In a corpus study we investigated the effect of some
structural and lexical variables in governing the preference to
attach a relative clause to one of two Noun Phrases (NP's). The
phrase-structures were of the type: "NP 1 - Preposition - NP 2 -
Relative Clause (RC)", such as "Someone shot the servant of the
actress who was on the balcony". In sequences of this kind the RC
(who was on the balcony) can be attached to the first noun
(servant) or the second noun (actress). To disambiguate the
sentence it is necessary to determine whether the modifier applies
to the first or the second potential host. Several theoretical
proposals have been made concerning the resolution of ambiguities
of this kind. Some of the most important are the Garden-path Theory
(e.g. Frazier, 1987), the Construal Theory (Frazier & Clifton,
1996), the Tuning Hypothesis (Cuetos and Mitchell, 1991), the
constraint based theories (e.g. Macdonald, Pearlmutter and
Seidenberg, 1994), the Predicate-Proximity /Recency Theory (e.g.
Gibson , Pearlmutter, Canseco-Gonzalez and Hickok, 1996) and
Prosodic Phrasing (Fodor, 1998). The research following these
theories revealed a number of factors which suggest that attachment
preferences can vary depending on particular features of the RC and
the characteristics of the NP hosting the RC. In our corpus
analysis we looked into Dutch journal and magazine articles for
phrases with the critical "NP1 - preposition - NP2 - RC" structure.
Consequently we considered the degree to which the broad number of
factors that were related, co-occurred with the particular NP1 or
NP2 attachments in our corpus. The variables diverged from lexical
to structural. An example of a lexically inspired variable is the
relative frequency occurrence of the nouns in Dutch. In the
constraint based framework distributional information of this kind
is considered important. Structural inspired factors were derived
from research affiliated with the Construal Theory or the Tuning
Hypothesis. In the last theory several proposals were made for more
fine grained subsets in which attachments seem to be based on
frequency of occurrence of the RC attachments in the language (e.g.
Desmet, Brysbaert and Mitchell, 1999). In order to check up on
those proposals variables as animacy and concreteness were
introduced in our analyses. Finally an example of a variable
checking for the influences of characteristics of the RC was
inspired by the Prosodic phrasing hypothesis which expects an RC
attachment to differ across constructions as functions of the
lengths of the modifier and its host configuration (Fodor, 1998).
To shed a light on this proposition, the length of the NP's and the
RC's were considered in the study. In summary we can conclude that
the investigation of a great deal of important features of the NP's
and RC's in this corpus analysis, reveals an important insight in
the way the NP1 - Preposition - NP2 - RC structure is used in
naturally occurring texts. Furthermore it enables us to evaluate
some theoretical claims made in the recent past.
Cuetos, F., and Mitchell, D.C. (1991). Cross-linguistic
differences in parsing: Restrictions on the use of the Late Closure
strategy in Spanish. Cognition, 30, 73-105.
Desmet T., Brysbaert M. and Mitchell D.C. (1999). Modifier
Attachment in Dutch: Assessing the merits of the Tuning Hypothesis.
Poster presented at AMLAaP-99, Edinburgh, UK, September, 1999.
Fodor, J.D., (1998). Learning to parse? Journal of
Psycholinguistic Research, 27, 285-319.
Frazier, L. (1987). Sentence Processing. In M. Coltheart (Ed):
Attention and Performance XII. Hillsdale, LEA.
Frazier, L. and Clifton, C. , Jr. (1996) Construal. Cambridge. MA:
MIT Press.
|
