Excellent, Carrol.
Since I am currently pursuing a line of investigation on this very subject
at the suggestion of Dr. Kess, I shall ask him for further clarification.

He has been content at this point to let me pass on what I have found from
Pound, Fenellosa et al. (good old al.) on this subject.

There is a further line of pursuit having to do with sensory input vs
sensory closure here to which your correspondent has not spoken.

We are looking not so much at the understanding or meaning here so much as
at what happens at the percetual level, which is a related but different


-----Original Message-----
From: Carrol Cox
To: [log in to unmask]
Sent: 12/31/03 3:34 PM
Subject: [Fwd: Re: Query on Japanese characters]

I have just received the post below from a Japanese friend, in response
to a query I had sent her re recent discussion of "ideograms" on this


-------- Original Message --------
Subject: Re: Query on Japanese characters
Date: Wed, 31 Dec 2003 16:41:44 -0500
From: Yoshie Furuhashi <[log in to unmask]>
To: Carrol Cox <[log in to unmask]>
References: <[log in to unmask]>

>Yoshie, the text below is from a post on the TSE list. Could you
>comment on its accuracy.
>It might help to elaborate on just what was involved in Pound's
>study of the congi (sp?) ideogram (as used by the Japanese, though
>Chinese in origin). Pound's study involved the realisation that the
>ideograms are not words but representations of thought processes,
>ie.: the effects of perception.
>As a Japanese student of mine once managed to squeeze through my
>thick brain, the Japanese do not think in words, but, so to speak,
>in pictures.

The idea that the Japanese think in pictures sounds like Orientalism run
amok. There seem to be differences between Kanji and alphabetic
processing, but the differences are not as radical as Pound thought they

Joseph F. Kess and Tadao Miyamoto
Department of Linguistics
University of Victoria
Victoria, B.C.
Canada V8W 2Y2
Occasional Paper #15
November 1997

. . . 9. Conclusion

Japanese research into the processing dimensions of a mixed orthography
sheds light on the basic questions of word recognition and lexical
access research in psycholinguistics. Our purpose in this paper has been
to introduce this rich psycholinguistic paradigm, and to show how which
considerations affect the path by which lexical access becomes word
recognition in processing written Japanese when presented in exclusively
syllabary, Chinese kanji characters, or mixed scripts. The picture for
Japanese lexical access is obviously not a simple one, but it is
certainly an interesting one because of the complexity of the writing
system. Certainly the picture is not so clear as to allow us to choose
between one simple, thorough-going explanation which places logographic
scripts on one side and alphabetic scripts on the other. It is obvious
that the mental lexicon has a complex structure which allows some
complex kanji to be retrieved either phonetically or semantically,
enabling Japanese readers to figure out the meanings of unknown words by
using the kanji lexicon in concert with compounding schemata, world
knowledge, and contextual information. A better way of looking at the
problem might be to suggest that kanji processing can employ either of
two processing routes in accessing the specific properties of a lexical
item presented in kanji script. In fact, this notion of a double-route
is not limited to logographic systems using hanji or kanji, but it can
apply to access strategies in alphabetic or syllabic systems which are
phonologically based.

We are not, however, saying that kanji processing is the same as
alphabetic or syllabary processing, especially in the earliest stage of
processing. It seems reasonable to assume that pattern recognition
processes are likely to be different for stimuli of the logographic type
and stimuli of the alphabetic or syllabary type, with logographic
stimuli having a greater dependence on visual pattern-matching stimuli.
And, of course, there is a vast array of experimental literature using a
variety of experimental tasks which suggests a contributory role of
graphemic information (see Miyamoto and Kess, 1995; Kess and Miyamoto,

But, by the same token, we cannot support the equally simplistic view
that kanji processing has a single route, with a cognitive leap from
Orthography to Semantics which ignores the contribution of Phonological
information. The most plausible cognitive model may mix its basic tenets
in this respect. That is, depending upon the contextual setting for a
given kanji, and its specific features of familiarity, frequency, and
complexity, one of two processing routes may be taken. Both processing
routes ultimately access semantic information, but one route is a
sound-mediated route and the other route is a grapheme-mediated route.
For many processing tasks that involve natural language, kanji symbols
are like alphabet or syllabary symbols in that they must invoke
phonological properties as the decoder searches through the mental
lexicon. Tasks that are not simple pattern-matching maneuvers take the
decoder from Grapheme through Phonology to Semantics. We suggest that
phonological properties are automatically accessed in most analytical
tasks that are not pattern-matching or category-matching in nature.

We also suggest that there is a cognitive routing that can travel a
grapheme-mediated route. This is the only way that we can account for
how some tasks access information about, as well as make decisions on,
kanji logographs that do not require phonological mediation. Moreover,
Japanese kanji and Chinese hanji will employ a direct route especially
in cases where hanji exhibit high frequency and high familiarity. There
are, of course, examples in alphabetic systems like English where the
processing route travelled is a direct route. For example, the English
lack of a perfectly transparent sound-letter correspondence is
overlooked in cases of morphophonemic identity such as the plural <-s>,
the past tense <-ed>, the alternation /haws > hawz-/ in houses, and so
forth. This is certainly the case in repeated instances of highly
idiosyncratic spellings; these quickly become immune to phonological
analysis and their spellings are soon ignored.

Words like Ubyssey in British Columbia, Liliuokalani in Hawaii, Thames,
Gloucester, and the admirable Crichton in Great Britain, and well as
common words like thyme, are forms of this type. There is, of course,
considerable experimental support for this. For example, in two
experiments using a vocalization task, Seidenberg (1985) has shown that
very frequent words in English are recognized visually, without
phonological decoding, just as they are in Japanese and Chinese.
Infrequent or newly-coined words were accessed by referring them to the
process of phonological decoding, whereas high frequency words and
characters were recognized visually without phonological mediation. 13

Thus, it may not be an all-or-none hypothesis we should entertain in our
explanatory model. A number of critical factors enter into the question
of what will be the most efficient strategy for achieving the task at
hand, If this expectation is valid, then we surmise that the claims for
the absolute uniqueness of logographic systems of Japanese kanji and
Chinese hanji are considerably weakened. The grapheme-mediated primary
route would be unique to neither Chinese nor Japanese, but is a matter
of degree, tied to how often this route is activated as the primary
route. Such a dual route notion, with its suggestion of two possible
routes to lexical representation, one a phonological route and the other
a direct route, is congruent with experimental results arising from work
on lexical access, word recognition, and reading in other languages.

Although the current philosophy of science inexorably draws our
attention to the question of universal constraints on how the mental
lexicon is searched, language-specific considerations of correspondence
regularity, frequency, and familiarity exert an influence as well.
Holding these factors constant, the analytical task type may drive the
choice of the most efficient route for turning lexical access into word
recognition. The three types of orthography, alphabetic, syllabic, and
logographic, certainly differ in their representational basis, in being
either phonologically based or morphologically based in principle. But
they will not be inherently different in their processing nature, in
that graphemic properties and phonological properties will be both
processed, but to varying degrees in different tasks. For example, it is
obvious that search procedures treat kana as more than strings to be
phonologically recoded. Orthographically familiar words in kana script
are named faster than both visually unfamiliar words and non-words,
suggesting that such words bypass the orthographic code to direct
lexical access. Thus, it is not the type of orthography that determines
processing time; rather it is familiarity with the frequency of the
orthographical shape which has an effect on lexical access. We must
conclude that the degree to which we employ the two processing routes
may differ across languages, but the fact of their availability will not
vary across these languages.

13 For Seidenberg's Chinese subjects, phonetic compounds were read more
quickly than non-phonetic compounds when the characters were of a low
frequency. In Seidenberg's experimental results, the interactive
relationship of Chinese hanji compounds with low frequency may have
exploited phonetic activation as the most effective processing strategy.
This is also what Leong, Cheng, and Mulcahy (1987) conclude after
analyses of variance underscored the individual contributions of reader
ability, frequency of hanji, and complexity of the hanji to vocalization
latencies in reading Chinese lexical items.
<>  *****

* Bring Them Home Now! <>
* Calendars of Events in Columbus:
<>, & <>
* Student International Forum: <>
* Committee for Justice in Palestine: <>
* Al-Awda-Ohio: <>
* Solidarity: <>