| |
Abstract:
A number of parser models (e.g., Pritchett, 1992;
Babyonyshev & Gibson, 1999) are based on the idea that
syntactic attachment happens at the verbal head, which gives the
parser information about semantic roles and grammatical relations
of argument noun phrases. Such models predict that
S(ubject)-O(bject)-V(erb) languages are harder to process than
SVO languages, since the parser would have to hold both S and O
until it hits V, as opposed to only holding S in SVO.
However, since there has been no attested difference in reaction
times of SOV and SVO languages for on-line processing, we
hypothesize that SOV languages have strategies to compensate for
the late appearance of the verb. In particular, they may
differ from SVO languages in having fewer sentences with
two-place predicates and both arguments expressed.
To test this hypothesis, we conducted a
comparative corpus study of English (SVO) and Japanese
(SOV). For both languages, root clauses (N=800) were
examined with respect to the frequency of one-place (SV:
intransitives) vs. two-place (SOV for Japanese, SVO for English:
transitives) predicate structures and the overt expression of all
arguments. Four different genres were examined in both
languages: home decoration magazines, mystery novels, books about
Japanese politics, and children's utterances (from
CHILDES). Japanese exhibits a significantly greater use of
one-place predicates than English (for example, 62.9% compared to
the English 36.5% in mystery novels; p<.001 in all genres
except books about Japanese politics). In addition, within
two-place predicates, Japanese uses null pronouns (pro-drop),
thus reducing the number of overt argument noun phrases.
The use of pro-drop with one-place predicates in Japanese is
significantly less than with two-place predicates (p<.05, in
all genres except mystery novels). The differences are
particularly apparent in child language, where Japanese-speaking
children around 3;8 had 21% transitives with 100% pro-drop and
English-speaking children of the same age had 71% transitives
with only 33% pro-drop.
These results suggest that there is an extra cost
associated with the processing of transitive clauses in a
verb-final language. To minimize that cost, Japanese uses a
significantly lower percentage of full SOV
structures. The results thus suggest that SVO and SOV
languages differ in their processing strategies in a systematic
way.
References
Babyonyshev, M., & Gibson, E. (1999).
The complexity of nested structures in Japanese. Language,
75, 423-450.
Pritchett, B. (1992). Grammatical Competence
and Parsing Performance. Chicago: The University of Chicago
Press.
|
