 |
SPOKEN WORD ACCESS PROCESSES (SWAP)May 29-31, 2000 |
 |
|
SPOKEN WORD ACCESS PROCESSES (SWAP)May 29-31, 2000 |
|
A word familiarity database (Amano & Kondo, 1999) was developed for
about 70,000 spoken words to provide a basis for a psycholinguistic research
in Japanese. Using this database, analyses and experiments were conducted
on two lexical competitor sets, the neighborhood and the cohort, to examine
their validity in lexical processings of Japanese spoken words. Firstly,
the characteristics of the neighborhood and the cohort were analyzed in
terms of their descriptive variables such as density (i.e., size), mean
familiarity, maximum familiarity, sum of familiarity, and the uniqueness
point as a function of a word familiarity. Correlation between the descriptive
variables were also analyzed. Secondly, to investigate effects of the neighborhood
and the cohort on Japanese spoken word recognition, a lexical decision
experiment and a word recognition experiment were conducted using two sets
of 100 spoken words which were selected from the database. The familiarity
scores of the word stimuli ranged between middle (4.0) and high (7.0).
These words and 100 nonwords were randomly presented to 30 Japanese subjects
through headphones in the lexical decision experiment. The subjects were
instructed to press one of two keys as soon as possible according to their
judgment of whether they heard a word or a nonword. Reaction times were
measured from the beginning of the stimulus. Partial correlation analyses
between there action time and their descriptive variables excluding a factor
of target word familiarity showed that there are no significant neighborhood
effects or cohort effects on the reaction time. In the word recognition
experiment, the words were randomly presented to 40Japanese subjects without
noise or with noise at signal-to-noise ratios of -5, -2.5, 0, and 2.5 dB
through headphones. The subjects were instructed to type what they heard
using a computer keyboard. Recognition scores were obtained by dividing
the number of correct answers by the total number of answers. Partial correlation
analyses showed that there are significant neighborhood effects but no
cohort effects on recognition scores. The results of two experiments showed
that only the neighborhood affects the spoken word recognition in Japanese
at least in a kind of 'unspeeded' task. It is suggested that the neighborhood
is more plausible than the cohort as a lexical competitor set, and that
some amount of time is necessary for the neighborhood to be activated and
effective in lexical processing of spoken word recognition.
Bibliographic reference. Amano, Shigeaki / Kondo, Tadahisa (2000): "Neighborhood and cohort in lexical processings of Japanese spoken words", In SWAP-2000, 91-94.