| |
Abstract:
Tadahisa Kondo, NTT Basic Research Laboratories, Japan,
kondo@av-hp.brl.ntt.co.jp
Shigeaki Amano, NTT Basic Research Laboratories, Japan,
amano@av-hp.brl.ntt.co.jp
In this paper, we first introduce the NTT Kanji Word Reading
Test, and then show how subjects' test scores affect the
variables in the psycholinguistic experiments differently. There
has been little research on directly comparing the adult
subjects' language abilities and their performance levels in
linguistic and psycholinguistic experiments. This test was made
using the word familiarity data from the NTT Lexical Database
(Amano et al, 1995; Kondo et al, 1996). There are two types of
word familiarity rating (on a scale of 1: least familiar to 7:
most familiar) - one is the visual word familiarity rating, and
the other is the auditory word familiarity rating. From the
Database, Kanji words with auditory familiarity ratings above 4.5
but with visual familiarity ratings below 4.0 were selected. In
general, Japanese adult subjects know or at least are familiar
with these words auditorily. Out of these words, 100 words were
randomly selected for the test. The test requires the writing
down of the pronunciations of these 100 words in Hiragana script
within 10 minutes. More than 400 Japanese native speakers (aged
between 18 and 32) were tested and their scores collected. The
mean score was 71.9, and standard deviation was 10.8. There was
no apparent age difference in the subjects' scores.
The relationships between the test scores of the subjects and
their performance levels in the psycholinguistic experiments were
examined. First, we collected word naming data with 1,000
two-character Kanji stimulus words from 20 Japanese adult
subjects (aged between 18 and 28) (Kondo and Wydell, 1997; Wydell
and Kondo, 1997). The subjects were divided into three groups
according to their test scores (poor: 67-75, intermediate: 76-79,
and good reader: higher than 80). The average naming latency was
shorter for the good readers than the poor readers. We found that
the magnitude of the word familiarity effect was stronger for the
poor readers than for the good readers, but that the magnitude of
the consistency effect was stronger for the good readers than for
the poor readers. Second, test scores and the lexical decision
times (Ainsworth-Darnell and Kondo, 1997), and text reading times
(Kondo and Mazuka, in preparation) were also inversely related,
similar to the naming latencies.
We therefore believe that it is worth trying this test before
any linguistic and psycholinguistic experiments are carried out,
so that much better and more precise data can be obtained.
Amano, S., Kondo, T., and Kakehi, K. (1996). "Modality
dependency of familiarity ratings of Japanese words."
Perception and Psychophysics,
57, 5, 598-603.
Kondo, T., Amano, S., and Mazuka, R. (1996, March). "Japanese
lexicon databases for psycholinguistic research." Poster session
presented at the ninth CUNY conference on sentence processing,
New York, NY.
Kondo, T. and Wydell, T. (1997, December). "Nature of naming
latency for Japanese Kanji words." Poster presented at the 8th
International Conference on Cognitive Processing of Asian
Languages, Nagoya, Japan.
Wydell, T. and Kondo, T. (1997, December). "Computation of
phonology in Kanji words: What matters - consistency or On/Kun
difference?" Paper presented at the 8th International Conference
on Cognitive Processing of Asian Languages, Nagoya, Japan.
Ainsworth-Darnell, K. and Kondo, T. (1998, January). "Beyond
orthographic depth: Similarities in the processing of words in
Kanji and Hiragana." Paper presented at Annual Meeting of
Linguistic Society of America, New York, NY.
|
