This document examines the writing system of Old Hiragana and questions the traditional view of the relationship between graphemes and allographs. It finds that in Old Hiragana, contrasts at the sub-mora level may constitute distinct graphemes rather than just allographic variations. Evidence from Jesuit missionary presses in Japan shows some sub-mora characters had unique associations with phonetic values. This implies their independence in the writing system and supports analyzing them as distinct graphemes rather than allographs. Introducing the concept of grapheme classes allows reanalyzing contrastiveness criteria and the division of grapheme types in the writing system.
2. Agenda
• Reviewing a theory of grapheme–allograph
structure in detail
• Examining the theory with the character
structure of Old Hiragana
• Introducing grapheme class to clarify the
responsibilities for grapheme and allograph
2
3. On Some Terms
(Rogers )
• An is a member of a grapheme,
which is not contrastive to the other
allographs
• A is a contrastive unit in a writing
system
3
4. On Some Terms
(Rogers )
• A is a system for graphically
representing the utterances of a language
• A is a general term for a writing
system without regard for its structural
nature
4
5. What is Allograph?
• A and a are allographs of the grapheme A in
that replacing them does not alter the
meaning of the word, as calm and cAlm,
except oddness
• Note: Studies of Chinese characters treat
allograph as relation of graphemes which
descend from the same origin, despite a
description found in Rogers ()
5
6. Allograph Class
• Allographs constitute classes over grapheme
(Rogers, )
• CAPITAL and minuscule
• Sans-serif and serif
• Roman, Bold and Italic
• Gill Sans and Avenir and so on…
6
7. Grapheme and
Phoneme
• Grapheme, as the name suggests, is defined
parallel to phoneme, and allograph to
allophone
• It is also referred to morpheme/allomorph,
hereafter omitted
• Note that Rogers () does not insist
that writing system is completely parallel
to phonology
7
8. Grapheme and
Phoneme
• Minimal pair plays a central role in
determining a phoneme, but similarity is also
convincing
• Conversely grapheme is solely determined
by usage, not by graphical likeness
• No outsiders would understand that γ
and Γ are of the same grapheme
8
9. A HoTEL Experiment
• To a certain extent, to know a writing
system is to know allograph relations
• In the brain we process a written word
without regard to allograph variation
• Consequently nonetheless HoTEL and
hotel are visually different we can read
both /həʊˈtɛl/ (Dehaene, )
9
10. A HoTEL Experiment
• ‘Wait! Can we assume that both HoTEL and
hotel suffer same process?’
• That both HoTEL and hotel go the same
process has little implication to the structure
of grapheme
• For instance it is not obvious that either H
and h are unified then processed or
separately processed
10
11. A HoTEL Experiment
• No one will argue against allograph itself
• Still there is room for an argument against
grapheme–allograph structure
• In other words, there is some doubt that
linguistic contrast can fully capture a
structure of a writing system
11
12. Contrastiveness
Criterion
• Whether instances make a contrast is not so
straightforward in the case of writing system
• colour and color
• beber and vivir (Spanish, b and v are not
distinguished)
• Once contrast happened, it will guarantee
the other contrastiveness in writing system
12
13. Contrastiveness
Criterion
• Degree of difference made with each
allograph class is not slight
• In Latin script, the case class seems the most
differentiated class
• No other classes make a variation like Q
and q, R and r
13
14. Contrastiveness
Criterion
• Supposed that both the case distinction and
stylistic differences make up equally
allograph classes, how to illustrate the
speciality of the case class?
• They do, actually, make a contrast, don’t
they?
• There seems a need to elaborate the
criterion
14
15. Contrastiveness in
Modern Latin
• Contemporary Latin writing system, which is
not authorised one, some distinguish vowel
i, u from consonantal j /j/, v /w/, the
others not (partial application is also found)
• Originally the writing system of Latin lacks
these vowel/consonant distinction whose
distributions are purely complimentary
(Marotta, )
15
16. Contrastiveness in
Modern Latin
• In the older writing system we neutralise the
distinction in writing and decode in reading
• Practically those subtle differences ease
distinctions over i, u and j, v, and let them be
one time distinct graphemes, the other time
allographs
16
18. Old Hiragana
• Used from around to
• Has over kanas for core morae
(adopted from Sproat, )
• later fused into , but maintained ancient
‘category’ (Frellesvig, )
18
19. Old Hiragana
• Cursivised from Kanji, which was borrowed
from Chinese writing system to represent
Japanese morae
• Not closed system
• Hereafter ‘Kana’ refers to ‘Old Hiragana’
19
20. Kanji to Kana
• When Kanji (hàn zì in Chinese) was utilised
to represent Japanese, there were mainly
two ways:
• Borrowing its sound
• Utilising the first sound of correspondence
Japanese word
• Both ways ignore what the word means
20
21. Kanji to Kana
• Most kana borrowed the sound
• In borrowing some simplification took place
as Middle Chinese syllable structure is more
complex than Old Japanese
• As a result large amount of homophonous
application occurred
• /ka/ in Kojiki (): 加可珂賀何訶
21
22. Kanji to Kana
• In developing Kana the use by public servant
give a direction
• Nearly ignored so-called seidaku
distinction
• Consulted few Kanji for a core mora
• Tended to write cursively
22
23. Contrast of Kana
• Two level contrast shall be distinguished
• Mora level contrast (kana category)
• あ, い, う, え, お…
• Sub-mora level contrast
• /ha/: は, は, は, は…
23
24. Contrast of Kana
• Mora level contrast is no doubt contrastive
• Sub-mora level contrast is said to make no
contrast
• Whether one writes かは or かは does
not contribute to the representation of a
word
24
25. Contrast of Kana
• Non-contrastiveness does not immediately
lead to the conclusion that they are
allographs
• There is a possibility of distinction as they
are mostly derived from different Kanji and
also a degree of cursiveness differentiate the
shape largely
25
26. Seidaku and Grapheme
• Sub-mora level contrast is not contrastive
on core morae
• However Seidaku contrast is not mentioned
26
27. Seidaku and Grapheme
• As noted, Kana lacks distinction for seidaku
• Seidaku, which nowadays is a voiced/
unvoiced contrast, was formerly contrast
with prenasalisation (supposed to have
changed gradually within Middle Japanese
period)
27
28. Seidaku and Grapheme
• If mora level contrast is grapheme
distinction in Kana, allographs, namely, sub-
mora level contrast will be used freely over
Seidaku
• If sub-mora level contrast make a contrast in
usage of Seidaku even slightly, it shall include
grapheme distinction, or even question
grapheme/allograph structure
28
29. A Case of f/b/p
Distinction
• In Late Middle Japanese (–) current
/h/ was /ɸ/
• By a convention later transcribed as f
• /ɸ/ descended from /p/ in the
environment of word-initial
• /w/ was for word-medial and final
29
30. A Case of f/b/p
Distinction
• Dakuon for /p/ can be reconstructed as
/ᵐb/, which fused into /b/ no later than
Late Middle Japanese
• After losing its place, however, [p] seemed
to have remained as an allophone for the
environment of such as geminate consonant
and onomatopœia
30
31. A Case of f/b/p
Distinction
• f/b/p share characters in writing
• By moraic nature they are combined with
vowel
• は, ひ, ふ, へ, and ほ correspond to /fa/,
/fi/, /fu/, /fe/ and /fo/, respectively
• Do they, in fact, make no contrast?
31
32. Jesuit Mission Press in
Japan
• Jesuit Province of Japan printed Japanese
textbooks with movable type (–)
• Their publication includes in Latin script and
in Japanese script
• Originally made by European hand, from
they renewed their movable type on
their own
32
33. Jesuit Mission Press in
Japan
• They had used metal movable type, not
wooden type, in order to recast same
character again and again
• It is important that in which character it is
written is very clearer than hand-written
materials
33
34. f/b/p Distinction in
Missionary Press
• Early Japanese script publication (EJ)
• Late Japanese script publication (LJ)
• EJ lacks a digraph for p; both have a
digraph for b, but often omitted
• Examining more than two characters per a
core mora
34
35. f/b/p Distinction in
Missionary Press
• EJ (Okada, , modified)
• は(者, 波, 者゛, 八, 和/wa/)
/fa/ /ba/ part. ba /pa/ /wa/ part. wa /Cw-/ Total
22 18 3 43
1 2 3
49 49 98
1 96 172 23 292
10 10
35
36. f/b/p Distinction in
Missionary Press
• EJ (Okada, , modified)
• へ(部, 遍, 部゛, 衣/e/, 恵/e/)
/fe/ /be/ /pe/ /e/ part. e total
79 9 88
9 6 9 6 30
130 1 131
4 4
28 28
36
38. f/b/p Distinction in
Missionary Press
• Mora level distribution does not relate
simply to phoneme
• Rather each character has their own relation
• 八 goes mostly to /ba/, or /wa/
• 者 goes to /fa/, /ba/, or /pa/
• 部 to /e/, 遍 to /fe/, /be/, or /pe/
38
39. Allograph in Missionary
Press
• There was some unification before, both
look alike:
• /fa/: 八 A B
• /fe/: 部 A B
39
40. Allograph in Missionary
Press
• Attestation of 部B is too scarce ( times)
• Both 八A () and 八B () relate to
/wa/ in the environment of word-medial
and final
40
41. Allograph in Missionary
Press
• Attestation of 八 confirms that they did not
receive distinction
• 部 implies that the difference was so subtle
as hardly to work independently
• Allograph in Kana has such a difficulty in
distinguishing them in shape and usage
41
42. Theoretical Implication
• What we called mora level contrast so far is
rather weak category regarding core morae
• The fact that some sub-mora level contrasts
have unique relation to phoneme shows
their independency in the writing system and
may name them graphemes
42
43. Theoretical Implication
• In sub-mora level contrast, shape unlikeness
is a source of distinction
• Unique sound relation in some of them
shows every shape unlikeness has a potence
to have it
• Therefore sub-mora level contrast is a
graphemic level difference
43
44. Theoretical Implication
• Mora level contrast is more abstract than
grapheme
• To generalise it can be named as
grapheme class
44
45. Theoretical Implication
• Graphemes of Kana also gives a ground for
dividing capital and minuscule into separate
graphemes
• Introducing grapheme class makes
contrastiveness criterion weaken than
Rogers ()
• This alternate intends to limit allograph to
stylistic one
45
46. Conclusion
• By introducing grapheme class
responsibilities for grapheme and allograph
become lighter
• Old Hiragana is a good example to present
a grapheme class–grapheme–allograph
structure
46
47. Conclusion
• A is a weak category which
comprises similar sound graphemes
• A is a unit which has
independence in identification and sound
correspondence
• An is what is subjected to parent
grapheme
47
48. Thank you for listening!
k-oada@let.hokudai.ac.jp
48
49. Literature
• Dehaene, S. (). Reading in the brain: The science and evolution of a human
invention. New York: Viking
• Frellesvig, B. (). A history of the Japanese language. Cambridge: Cambridge
University Press
• Marotta, G. (). The Latin syllable. Hulst, H. & Ritter, N. (eds.) The syllable: Views
and facts. Berlin & New York: Mouton de Gruyter
• Okada, K. (). A development of Kana movable type by Jesuit mission press in
Japan: With special reference to On Baptism and Preparation for Death (日本イエズ
ス会版における日本語活字の開発: 『病者を扶くる心得』の仮名活字組版か
ら). Unpublished ms. [in Japanese]
• Rogers, H. (). Writing systems: A linguistic approach. Malden, MA: Blackwell
• Shirai, J. (). Kana glyph usage in Jesuit Mission Press (キリシタン版の仮名文字
遣). Kuntengo to Kunten-shiryō [in Japanese]
• Sproat, R. (). A computational theory of writing systems. Cambridge: Cambridge
University Press
49