|
Abstract:
Tadahisa Kondo, NTT Basic Research Labs,
kondo@av-hp.brl.ntt.co.jp
In a language such as Japanese, the presence of scrambling
allows the order of arguments to be relatively free. Thus,
although the principle word order of Japanese is SOV, sentences
with OSV word order are also grammatical.
Linguistically, the presence of scrambled sentences has been
used to argue for a flat (non-configurational) structure for
Japanese (Hale, 1980; Farmer, 1980). Alternatively, the facts
that some grammatical phenomena, such as pronominalization, scope
interpretation, and weak crossover are affected by scrambling
have been used to argue for a configurational structure of
Japanese (e.g., Saito, 1985, Hoji, 1986;1992; Saito and Hoji,
1983).
If Japanese sentences have configurational structure, and
scrambled sentences are derived from an SOV structure, then
scrambled OSV sentences should have more complex structure than
non-scrambled SOV sentences. Previously, however,
psycholinguistic studies have not found increased processing cost
associated with scrambled sentences. Nakayama (1995) found no
difference between scrambled and non scrambled sentences in
probe-recognition experiments. Yamashita (1997) also found no
difference either in a self-paced reading experiment or in a
lexical decision task. Yamashita argued that the Japanese
listener may not be computing the configurational structure of a
sentence on-line.
In this study, scrambled and non-scrambled sentences are
studied in four experiments -- two types of questionnaire
studies, an eye movement monitoring experiment, and a self-paced
reading experiment. Five types of sentences are studied.
(1) Mary-Nom John-Acc saw.
(2) Mary-Acc John-Nom saw.
(3) Mary-Nom [Relative clause] John-Acc saw.
(4) [Relative clause] Mary-Acc John-Nom saw.
(5) Mary-Acc [Relative clause] John-Nom saw.
The results showed that scrambled sentences indeed show an
increased processing load associated with them. Japanese speakers
rated scrambled sentences (2) and (5) more difficult and more
misleading compared to non scrambled counterpart (1) and (3)
respectively. However, sentence (4), in which the scrambled NP
had the relative clause and thus did not have a center-embedded
structure was significantly easier than (5). Eye movement data
and the self-paced reading time data showed in general an
increased processing cost for scrambled sentences. Critically,
the second argument in a scrambled sentence (John-Nom in (2)) had
significantly longer reading time than the second argument of the
unscrambled sentence (John-Acc in (1)).
The results will be discussed in light of the linguistic
complexity of the scrambled sentences as well as in terms of
models of sentence comprehension in Japanese.
Nakayama, M. (1995) "Scrambling and probe recognition." In R.
Mazuka and N. Nagai, eds.,
Japanese Sentence Processing
257-273.
Yamashita, H. (1997) "The effects of word-order anc case marking
information on the processing of Japanese."
Journal of Psycholinguistic Research
26:163-188
|
