WORDS WRITTEN IN KANA ARE NAMED FASTER THAN THE SAME WORDS
WRITTEN IN KANJI*
Laurie Be Feldman+ and Me T
colors wri tten in Kanj i, a
y0 Although colors
a
are more
i form and although Kanji are
more
of words in general, latency to
vocalization was consistentl y less for the Kana. This superiority
is attributed to the closer relation of Kana to phonology and,
therefore, to
The demonstrated
facility for naming
Kana accords with observations in the literature that very familiar
visual
are consistently named faster when they conform to
pr
than when
do not.
The evolution
is characterized by a trend away from
many
units towards representing a more
restricted set 0 abstract
ical uni ts. The characters of the oldest
systems
and situations. These pictographs and semasiographs
did not
Their iconic
made them visually distinctive,
but they could
to only
few concrete
ects and common ri tuals. As
these
s became more conventionalized and their resemblance to
ects diminished, the 1
value of the character as the sy~bol for a
could
any word, logographs
word was enhanced
provided for ex
written communication,
however,
number
, usually according
to a mor
ical
, semanticall y related
words were often visually similar
a common radical
Their
howev
in the written form. The
o
zation (Gelb,
many level
, words that sounded alike
the same
their meanings were unrelated
were
, but the same principle could be
These were substitutions for the
The
from a logography and represented
applied
stent
by which
deliberate
the
consistently
derived from the Chinese
al
NICHD Grant
(1
)]
195
alphabet obtained.
While the arbitrary pairs were initially better, after
practice the sound-correlated orthography proved superior in terms of shorter
latencies to vocalizationo When Brooks (1977) exaggerated the visual interaction within the forms by combining the component parts into a glyphic pattern,
he found that this enhanced visual
also facilitated naming
In
subsequent studies, he introduced controls both by expanding the stimulus
vocabulary and
other artificial orthographies, but the reliance on
contrived
and extensiv
leaves lingering fears about
the application of these results to skilled reading of natural orthographies
0
0
The structure of the two
permits a natural
language variation on the Brooks latenc
ization procedure. Kana is a
syll
in which the
specification of each
(more precisely
mora) is
character
virtue of this sound-referencing or
phonographic orthography, similar sound
words look alike o In contrast, the
Kanj i script is
is no structure internal to the whole
character that denotes pronunciatione Moreover, where Kana are generally used
to designate tense, prepositions, new words and
terms, Kanji characters are used for nouns, verbs and
ectives0 Finally, the Kanji tend to
be
and square whereas the Kana tend to be a horizontal arrangement of
discrete curved
By
with Brooks (1977), we compared latency
to vocalization for
color names written in Kana and in Kanjie
Phonographic writing
the sounds of speecho
Given the
maj or outcome to Brooks? ex
should ex
the latency of naming
to be shorter for Kana than for
ainst this expectation, however, are
the
Forster
Chambers (1
) demonstrated a strong
po tive correlation between
y of
ish words and naming timee
Based on this evidence, we
that because color names in Japanese
1 iterature appear more
i than in Kana, naming the colors
written in
i should be
than nam
the colors lrlritten in Kana.
Second, Brooks demonstrated, as
abov
that
were named
more
y than their discrete counter
Therefore, we might expect
shorter nam
latencies for the somewhat
i forms than for the
discrete Kana form 0
color
Stimuli consisted of
in
y from
) corpus
000
ables
names who
ish
urrences based on the
word had between
y in
alents
s
and four
i and
and half
ucture
0
197
Table 1
Summary of stimulus-item structure
Japanese
Words
English
Equivalents
Number of
Characters
Number of
Syllables
Number of
Strokes
Frequency of
Occurrence
Kuro
black
2
11
203
Midori
green
3
14
116
Chairo
brown
2
3
9 + 6
176
Hairo
gray
2
4
6 + 6
80
Shuriro
vermilion
3
6
3
Kuriiro
chestnut
4
10 + 6
5
2
Two native Japanese served as subjects~ They were instructed to read as
y as possible the stimulus words handwritten on slides displayed in two
fields of a Scientific Prototype Model GB Tachistoscope ~
Each item was
exposed for 500 msec and followed by a dark interval of about a second ~ The
signal to light the display also triggered a timer that stopped at the onset
of vocalization~ In the course of three sessions, the two orthographic forms
(Kanj i/Kana) of the six color names were each presented 100 times in a
randomized order.
In summary, the experimental design consisted of sUbjects' vocalizations
of two orthographic forms (script) of each of six color names (stimulus items)
presented in three sessions. Each session was composed of six trials per item
where each trial was the average of approximately five observations, and data
were then averaged over the six trials~
and
scussion
An anal
S
condition
1,10)
across all six stirnul us i terns in each
ect revealed
ficant main effects for
43.77
.001, and
66.88 p < .001, session, F(2,20)
1 , 10)
25 02 - p < .001
The
x session
\'laS
,20)
8
- < 01
As evident in Table ,the
itation
198
0
variance
for each
Table 2
Individual word latencies as a function of writing system and session
Session I
Session II
Session III
Word
Kana
i
1•
Kuro
458
470
423
440
409
424
2.
Midori
429
445
401
436
401
424
3
Chairo
495
488
444
466
434
454
4.
Hairo
478
487
430
447
425
443
5.
Shuriro
488
507
460
480
443
468
6
Kuriiro
532
539
456
486
468
501
ana
of Kana relative to Kanj i increases ov
interaction was significant, F(2 20)
Kana
i
Kanji
sessions
The subject x session
45, -E < 00 1
G
G
G
G
When subjects' data were pooled, only script was significant, F( 1 , 1) ::
192015, -E < 046. Stimulus items
significance, F(5,5)
4.48, -E <
063.
G
A significant facilitation of vocalization for the sound-referencing Kana
orthography relative to the logographic Kanji orthography obtained for almost
all stimulus words throughout all sessions
Naming latencies to the Kana
averaged 18 msec faster than to the Kanj i.
(Any comparison of specific
stimulus items must be mad cautiously, as the acoustics of differing initial
segments may have
ered the timer at different points in the
utterance 0) This result is
it violates documented effects of
word structure related both
usage
word
y, and to
visual
of d
1
convention,
color words are usually
i, but the familiarity
of this
to be 0 no
In addition enhanced
visual
characterized
and demonstrated
) to be
discrete linear forms (such
Kana) d
the
y to
i
faster than
G
9
199
Japanese
i has been ci ted as an example of a script that does not
contain information about phonology and recrui ted as evidence that readers
must be able to access the lexicon visually in order to obtain a phonological
specification
Another per
e on the same issue is the role of the
lexicon in providing phonological codes for tasks such as naming.
The
structure of
i would seem to imply that such mediation is mandatory. In
contrast the lexical mediation of phonology may be optional in Kana, given
its
characters
0
At thi point, it is perhaps useful to appreciate orthographic structure
conditions in an attempt to account for the continued
relevant to
facilitation for read
aloud of Kana relative to Kanji.
There is some
dev
evidence that reflects this influence of orthographic structure
on lexical per formance .
and Yamad a (1978) found that among threeand
the relative difficul ty of learning Kana symbols far
exceeded
i word s" Sakamoto (in press) reports that while a small
set of
character
is systematically introduced by grade in the school
curriculum 9
to read in Kana is completed in a relatively short period
once the child
ins to read
e
Evidence
of selective
and hemispheric superiority in word
also
distinction in processing the two Japanese
On both a v isual
and a wri ting task (Sasanuma,
imura, 1971), apraxic aphasics make more errors on the
1974; Sasanuma &
i while
aphasics perform comparably on Kanji, but
Kana than on
make fewer error s on Kana" It seems that the Kana specification of phonology
apraxic e
One interpretation (Sasanuma & Fujimura
not ex
i
o the apraxic
renders
1 1)
that
forms as particular phonological
the
treated
the phonological processor in
terns
Since
forms
are more vulnerable to left hemisphere
order to be identified,
which can be directly identified without any
than a
recognition by normals
a
different
for Kana and for Kanj i.
Hatta
(1977)
for recognition of Kanji words
tha t
and Koba yashi (1977) find ing 0 f
left
A nonsignificant right hemisphere
effect for
(Sasanuma et al., 1977) may reflect differences in stimulus
structure between these two ex
Where Hatta used individual Kanj i
characters Sasanuma et ale used random
s of characters, but the combination of
characters will often determine the semantic and phonological
inter
f each character (Martin, 1972)
a Kana
nonwords may be
character
In
200
These resul ts represent an extension of the Brooks (1977) finding e The
mora-sized graphemes of Kana are
to the phoneme-sized graphemes of
an artificial al
both adhere to a phonographic principle. In a
of a phonographic script relative to a logographic
naming task, the adv
script is
manifest
To conclude, it seems that a delineation of
strategies appropriate for a read
task such as naming must consider the
particular
of the
as well as the specific task, and
that it is the
fication of
intrinsic to its orthographic form
that accounts for the facilitation of Kana relative to Kanjie
e
words: Use and acquisition
In
(Ed ), Basic processes in reading:
Hillsdale,
Lo Erlbaum, 19776
correspondence without trying to
0
0
Baron,
learn
Baron, J 0' &
reading
••
55-706
and word-specific knowledge in
Human
In Ao Reber &
identification
Proceedings
of
----=Erlbaum t 1977 ..
ournal of
and naming time .
627-635.
of Chicago 1952.
the left and right visual
6
For
Gelb t
Hatta, T.
fields.
t
Reciprocity in
Attention and
observations
ear and
0
In
Patterson,
cod
1977 t
Sakamoto
and
Sasan
c~
_
transient
nonKana
than
201