On the Old Japanese Kakari (Focus) Particle koso

1

Transcript of On the Old Japanese Kakari (Focus) Particle koso

Page 1: On the Old Japanese Kakari (Focus) Particle koso

言 語 研 究(GengoKenkyu)127(2005),1~49 1

On the Old Japanese Kakari (Focus) Particle koso: Its Origin and Structure')

Leon A. SERAFIM

(University of Hawaii at Manoa)

Rumlko SHINZATO

(Georgia Institute of Technology)

Keywords: koso, Okinawan, demonstrative, focus particle, gram-

maticalization

1. Introduction

Both formal and functional linguists, both in Japan and elsewhere, have

discovered or rediscovered an Old Japanese (OJ) syntactic agreement

phenomenon called kakari musubi (KM), now no longer extant in Stan-dard Japanese.2)We intend to fill a gap here that was created by the

relatively slower pace of research on the same set of constructions in

Okinawan, a sister language of Japanese, where it has survived, and which

1) In the process of writing this paper, we received invaluable assistance in the discussion of issues, feedback on drafts, and acquisition of rare materials from the following people: Holger Diessel, Karel Fiala, Ichiro Handa, Alice Harris, Atsuko Izuyama, Shigehisa Karimata, Satoshi Kinsui, Frank Lichtenberk, Ta-kashi Nomura, Charles Quinn, Shigeo Tonoike, Akira Watanabe, John Whit-man, and Janick Wrona. In addition, we gratefully acknowledge the invaluable comments of the two anonymous reviewers, without whose feedback this paper would have suffered. Naturally, any remaining faults are attributable entirely to the authors.

2) The following are some representative works: Japanese traditional linguistics: Ono T (1978), Ono (1993), Sakakura (1993), Nomura (1995,2001); Western lin-guistics (functional): Whitman (1997), Quinn (1997), Shinzato (1998), Schaffer (2002); Western linguistics (formal): Hagstrom (1998), Watanabe (2001), Ota (2002). For a useful summary of the approaches, see Kinsui (2002).

Page 2: On the Old Japanese Kakari (Focus) Particle koso

2 Leon A. SERAFIM and Rumiko SHINZATO

will enrich the understanding of OJ kakari musubi.

We first compare OJ koso3) and Old Okinawan (00k) si, establishing

them as cognates based on empirical grounds, such as phonological and

dialectal evidence, as well as on theoretical grounds, such as functional

congruities and grammaticalization theory, and then reconstructing their

Proto-Japonic (PJ) progenitor. We then dispute the generally held ety-

mology of koso as being from the proximal demonstrative ko and the

mesial demonstrative so (Matsushita 1930: 603; Ono T 1978: 158; Ono

1993:4) 120-121; Sakakura 1993: 247). We support and extend instead

a relatively new etymology, first proposed, as far as we are aware, by

Thorpe (1983: 242-243, 251), with comparative data support, as the com-

bination of PJ proximal *ko+PJ nominalizer *swo. Finally we use a broad,

cross-linguistic view of grammaticalization to find the meaning of the cor-

respondence between the focus particle koso and the izen conjugational

form.

2. OJ koso and 00k si in comparative perspective

2.1. Overview of kakari musubi constructions: OJ koso and 00k si

KM employs so-called kakari particles (KP) in mid-sentence to bind with

specific conjugational forms, not the usual finite forms, for rhetorical

effect. A KP highlights the preceding element as the focus of the sentence

in a cleft-like structure (Shinzato 1998; Schaffer 2002; Ota 2002). OJ had

five such KPs, namely, zo namu, ya, ka, and koso, while 00k had three,

ga, du (<*do), and si-syu. The cognation between OJ ka and 00k ga, and OJ zo and 00k du is well known (Yamada 1954, Mamiya 1983, Uchima

3) Earlier forms of Japonic had a distinction between so-called Type-A vowels (characterized by greater tension, and/or by homorganic onglides) and Type-B vowels (characterized by greater laxness, and/or by more central articulation, and possibly offglides). Here we employ the Yale romanization, devised by Samuel E. Martin, for this distinction, with the following eleven sets of vowel types: Type A: yi, ye, wo; Type-B: iy, ey, o (the latter perhaps schwa): neither type: a, u; neutral to Type A or B by position: i, e, o.

4) Unless there is a "T" after this name, it always refers to Ono Susumu.

Page 3: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 3

1994, Nohara 1998, Handa 1999, Serafim and Shinzato 2000, and Shinzato

and Serafim 2003). KM is a highly restricted and yet extremely complex

syntactic phenomenon. The two KPs, namely OJ koso and 00k si-syu,

naturally raise the question of cognation since they are the only ones that

bind with an exceptional izen conjugational form. But no serious study

has been published linking OJ koso and 00k si-syu to reconstruct a PJ

progenitor. As seen in the examples below, it is evident that both OJ koso and

00k si bind with the izen conjugational form, not the usual rentai form. It

should be noted that 00k si has two allomorphs: P«su> as in example (3)

and L T, <<siyo>> as in example (4). The latter is a palatalized variant trig-

gered by the preceding high vowel yi.

Old Japanese 5)

(1) 昔 こそ 外 に も 見 しか

mukasi koso yoso ni mo mi-sika

long ago KP strangely, even think-PAsT(IZE1>)

我 妹 子 が 奥 つ 城 と 思 へ ば は し き 佐 保 山

wagimoko ga oku-tu-ki to omoheba hasiki saho-yama

my love 's sepulcher as when I think dear Mt. Saho `Once I thought Mt. Saho strange, / but time goes by, and all things

change. / Now I hold the mountain dear: / my love laid in its sepul-

cher.' (Man'yoshu 474 [Honda 1967] )6)

5) This is given in the examples in a conventionalized, rather than strictly pho-nological, representation.

6) Note that koso binds with an adnominal form if the final predicate is an ad-jective instead of a verb. In Omoro Sdshi, we were unable to find an adjectival ending; therefore, it is not certain what adjectival form si corresponded to.

Page 4: On the Old Japanese Kakari (Focus) Particle koso

4 Leon A. SERAFIM grid RUmlkO SHINZATO

(2) 夕 さ らば 君 に 逢 は む と 思 へ

yuhu sar-aba kimi ni ah-amu to omohe

evening come-when you to see-INF COMP think

こそ 日 の 暮 る ら く も 嬉 し くあ りけれ

koso hi no kururaku mo uresiku ari-kere

KP day SUB end even happy be-PAST (IZEN) `My dear, you come to me at evensong, / and from the morning for

that hour I long.' (Man'yoshu 2922 [Honda 1967] )

Old Okinawan7.8) は つにしや ま

(3) 初北風 す 待ち居たれ おきとば す

<<hatunisiya su matiyotare okitoba su

faci-nyisya si mac-yu-tari ?ukyi-tuba si

[*atsipi f a si mat f utarl ?ukkituba si

first-north-wind KP await-be-PAST(IZEN) north-wind KP

待ち居たれ

mattyotare>>

mac-yu-tari

mat f utari]

await-be-PAST(IZEN)

「初北 風 こそ待 って い た のだ。北 風 こそ待 って い た のだ。」

`It was the first north wind itself that we awaited . It was the north

wind itself that we awaited.' (Omoro Soshi 899 [Hokama 2000])9)

7) Because Old Okinawan is not well known, we give a transliteration followed by the reconstructed phonological form, then the proposed phonetic form. In addition, because there exists a translation into Japanese, we give it, and then append our English translation of that.

8) The Omoro Sdshi is the oldest Okinawan document of any length, dating from the 16th and 17th centuries, said to represent the language of from about the 10th century to the time it was compiled.

9) It has been noted that some 00k <<su>> correlated to the meirei-kei, impera-tive' form (Hokama 1972). However, the putative imperatives are actually for-mally identical to the realis forms. In other words, they are realis forms.

Page 5: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari(Focus)Particle koso 5

かね つか こ

(4) 金 ぐす く … 大 こ る が 使 い し よ 此 の

《kanegusuku...OOkoru ga tukai siyo kono

kanigusiku ...?ufakuru ga cikayi syu kunu

[kanigusiku...?uΦukuru ga tsikaji ∫u kunu

Kanegusuku...Okoro 's invitation Kp this だしま お

大島 降れたれ

dasima oretare>>

dasyima ?uri-taxi

dafima ?urrtarl]

island descend-PAST(IZEN)

「金 ぐす く...大 ころ...様 の お 招 きが あ っ て こそ,こ の 島 に神 は 降 り

給 うた のだ。」

`It was due, in fact, to the invitation by Kanegusuku Okoro that the

god descended on this island.' (Omoro Soshi 1005 [Hokama 2000])

The Man'yoshu (MYS), the oldest collection of OJ poems, contains about

200 instances of koso, and the Omoro Soshi (OS), the oldest collection of

Old Okinawan songs, contains about 500 instances of si - syu. KM went

out of use in Japanese, and the KPs were lost or became adverbial or sen-

tence particles. The particle koso is an adverbial focus particle in Modern

Japanese. In Modern Okinawan, the KP si survived only in a few fossil-

ized forms.10)

2.2. Evidence to support phonological values of 00k KPH"' <<su>> and

‚µ‚æ<<siyo>>

A rough comparison of OJ koso and OOk si in section 2.1 is sufficient to

suggest their relatedness. However, in order to establish cognation and

reconstruct a progenitor, phonological and derivational details have to be

10) As a transition to the modern language, where the construction is lost, the kumiodori dance-dramas of the 18th century (Ifa 1974, Ryukyu gikyokushu, in Hattori et al. 1974) show a few stylized usages and lexicalizations of this KP.

Probable cognates of 00k si in Northern Kyushu and Western Honshu dialects survive as the nominalizers su and so.

Page 6: On the Old Japanese Kakari (Focus) Particle koso

6 Leon A. SERAFIM and Rumiko SHINZATO

worked out. In contrast to OJ koso, the phonological value of OOk si has

not yet been investigated, so we will first determine its phonological val-

ues based on the spelling clues embodied in the Omoro SOshi.

2.3. Reconstruction of the Proto-Ryukyuan KP *-swo

from 00k (OS) -si~-syu

It is a commonplace that the equivalents to Japanese i and e are

Okinawan i (J i/e :: Ok i), and that the equivalents to Japanese u and o

are Okinawan u (J u/o :: Ok u). To explain the correspondence, one may

claim that some vowels in the proto-language split, according to some

principle, or, alternatively, that they merged. No principle for a split is evident; on the other hand, a simple sound change of raising of mid to

high vowels can readily explain the correspondences. It is straightforward

to hypothesize, then, a generalized raising of mid to high short vowels.

(5) Pre-Ok Ok *i (high)

> i (high) * e (mid)

* u (high) > u (high) *o (mid)

Thus, any seeming exceptions to this large generalization naturally require

explanation. In the case of s-initial syllables in Okinawan, there is indeed

an exception, and it is also quite well known. Japanese su and so typically

correspond to Naha syi and su, respectively:

(6) J su : so :: Nh syi [*i] : su [su]

The language of the Shuri gentry of the latter part of the 20th century gives

a slightly different, and more archaic, correspondence, namely si and su,

respectively.11)

11) Shuri is the old capital, and Naha, its formerly dependent commercial town. Shuri is now part of Naha.

Page 7: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 7

(7) J su : so :: Sr si [si] : su [su]

Putting the two sets of correspondences together, we have:

(8) J su : so :: Sr si [si] : su ::

Nh syi [*i] : su

From these correspondences, we may hypothesize the following changes

in the history leading up to the Shuri/Naha dialects: (a) *su > (Shuri) si >

(Naha) syi, and (b) *so > su. The set of changes (a) had to start no later than the change (b). Or, to put it another way, high back vowels central-

ized in post-sibilant position, and they did so before the generalized rais-

ing occurred. 12)

(9a) *su > *si > (OS) si > (Shuri) si > (Naha) syi

(9b) *so > *so > su > su > su

or:

(10a) *su > (OS) si > (Shuri) si > (Naha) syi

(10b) *so > su > su > su

Thus it should be no surprise that the spellings <<su>> and <<so>>

(‚») are distinct in the Omoro-SOshi language, corresponding cleanly to

Shuri si = [si] / Naha syi = [*i] and Shuri/Naha su [su], respectively, as in

(11a):

12) All of Japonic seems to have a tendency to centralize (and depending on the

dialect, to front) high back vowels after sibilants. This is seen also in, e.g., stan-

dard Japanese, in the phonetic value of <<su>> as [si], a more centralized vowel

than the more back (and generally weakly rounded) [us] generally seen after

non-sibilants. Okinawan has also had this centralization, which then led to a

fronting of *i=* [i] to (non-palatalizing) i=[i] -instantiated in, e.g., le and Shuri

dialect mini=[mizi] 'water'- then to (palatalizing) yi=[(j)i] -instantiated in, e.g.,

Naha dialect mizyi=[mi‚ëi] `(id.)'. (For the sake of ease of dialect comparison, zi

and zyi are distinguished, as required for Shuri; but zyi is a redundant spelling

in Naha and in the Okinawan common dialect.)

Page 8: On the Old Japanese Kakari (Focus) Particle koso

8 Leon A. SERAFIM and Rumiko SHINZATO

(11a) Spelling: <su> : <<so>> Omoro-Soshi language: Ph

onological value: si : su

Shuri: si : su

Naha:syi:su

The spelling <<su>> (‚·) must be /si/ = [si], because it is distinct from both

<<se>* (‚¹) =[si] =[sI]and <<si>> (‚µ) = /syi/ = [*i], as well as the above-men-

tioned <<so>> (‚») = /su/ = [Su].13) See (11b):

(11b) A sibilant-plus-high-vowel set

<si> <<se>> <<su>> <<so>>

syi Si Si su

[fi] [sI] [si] [Su]

Thus, we hypothesize that one allomorph of the OS KP particle, <<su>> (‚·)

is -si = [si], and .the other, <<siyo>> (‚µ‚æ) is -syu = [fu].

Next, we need to establish the pre-OS source for the allomorph,

-syu . In order to do so, we have to take into consideration an important

phonological fact: a consonant is not progressively palatalized (PPed) if

followed by *u (Hattori 1978:97b). Thus OS -si (< *-su, through Sibilant-

based Unrounding) - see (12) - cannot be the source of the allomorph OS

_Syu .14)

13) One ready way to see the distinction among the members of the set <<si> , se> , and <<su>> is to compare the large number of instances of each of three

words that begin with the same kana, then one of the three sibilant initial kana above. They are: for <si> , <<kusikawa>> - <<gusikawa>> 'Gushikawa', a place name,

either gusyikawa [gufikawa] or-judging from the modern dialect version of the word gusyicyaa-OS gusyikyawa; for <se> , <<kuse>> `(eulogistic prefix)' kusi-

[kusi] (surely related to, but different in form from, the J shiku-adjective kusi-); for <su> , <<kusuku>> - <<gusuku>> `fortress', gusiku [gusiku]. In general, <...e>>- ending syllables, where distinguishable from <<...i>> -ending syllables, are i=[i],

and therefore for <...u» of <su > the only available vowel space that differs among the high vowel positions is i=[i], and such a position accords with the

known evolution of such syllables in Japonic, leading to the conclusion that <<su>> is necessarily si=[si].

14) "X" refers to changes that did not take place or that were impossible.

Page 9: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 9

(12) x{... yiCu *... yiCyu}

However, what we write provisionally as *o, not being high, does not

block PP, as in the case of another KP, *-do:

(13) *... yi-do > *... yi-dyo > OS ... yi-dyu [jid‚ëu].

Otherwise:

(14) *... V-do > OS ... V-du [du].

The facts of PP followed by raising") suggest the obvious move, to

hypothesize an allomorph *-so as the source of the OS -syu. As for the

immediate source (Pre-OS, but not Proto-Ryukyuan) for the OS -si, we

hypothesize it to be *-su. Observe (15):

(15) Two-Height Allomorphy

high-vowel allomorph * -su OS -s i

mid-vowel allomorph * -so > *-syo > OS -syu

The reconstruction so far leaves two questions unanswered: (1) Does

the original Proto-Ryukyuan (= PR) KP from which the two allomorphs

(*-su - *-so) derived have a high or mid vowel? (2) If it is mid, and on the

assumption that sound change is regular, shouldn't the vowels of pre-OS

"* -so" and the other KP, "*-do," differ from each other? This question

naturally arises because there is no allomorphic alternation for the KP

"* -do" in OS arising from pre-OS height allomorphy, unlike "*-so". (That

is, "*-do" did not give rise to an allomorph OS x-zi, nor to Sr x-zi = [zi] or

Nh x-zyi = [‚ëi].) And if they do, what is the proper characterization of that

difference? Let us deal with question (1) first.

If we had started with a high vowel, as in (16), then *-su > *-su ~ *-so

> *-su ~ *-syo > *-si~*-syu (where V in the table stands for any vowel

other than *yi):

15) We will later call this "Raising 2."

Page 10: On the Old Japanese Kakari (Focus) Particle koso

10 Leon A. SERAFIM and Rumiko SAINZATo

(16) Putative Lowering PR *... V-su ~ *... yi-su

Lowering * ... V-su ~ * ... yi-so

PP *... V-su - *... yi-syo

Unrounding/Raising OS ... V-si -OS ... yi-syu

But what could have caused a Lowering, followed by a Raising? Could the

preceding vowel, * yi, have caused it? This is the gambit -height dissimila-tion-that Hattori (1976c: n26) opts for. He sees PJ *wu16) as lowered by

a neighboring high vowel. His example is *iswo rocky shore; crag jutting

from water', supposedly from *iswu, as in (17):

(17) *iswu > OJ iswo rocky shore'

But Hattori (1976b: 29-30) also claims that PJ *wo and *ye17) raised to

OJ u and yi, respectively. This sometimes-lowering-sometimes-raising

approach cries out for simplification. We need what we will call "Raising

1" anyway (see below), so why not start with mid vowels?") Since Hayata

(1998) handles word-final cases as original mid vowels (such as in iswo) that did not raise because they were in final position (therefore *iswo),

we don't need to account for them as lowerings. Thus, having determined

that setting up a mid vowel constitutes a simpler hypothesis, we can set up

the sequence-which includes Raising 1-leading to the OS form -si, as in

(18):

(18) OS -si < Pre-OS *-su < PR *-swo.

Some positions are prone to raising and some are not, so some *wo and *ye raise, and some don't. A vowel *ye or *wo in the position just before the end of the quadrigrade-class verb stem is prone to raising in Ryukyuan

if there is no preceding high vowel: e.g., (19):19)

16) Here respelled from his *u. 17) Respelled here from his *o and *e, respectively. 18) For a similar line of argument contra Hattori's lowering-and-raising ap-

proach, see Hayata 1998.

Page 11: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 11

(19) PR *aswob- play, dance' > pre-OS *?asub- > OS ?asib- > Shuri ?asib- > Naha ?asyib-

The verb base for `play' in the OS is, similarly, ?asib-, written, again, with

su> (t) _ /si/ (OSJS 32a-33a, 338c-339a) '20) apparently unlike OJ, where

significant numbers of Type-A mid vowels occur in stem-final position (cf.

the cognate OJ aswob- play' in (20)).

(20) OJ aswob- 'play':: OS ?asib-

Note the parallel to OS -si (~-syu) under discussion above.

Now let us deal with question (2) raised above. To reiterate: On the

assumption that sound change is regular, shouldn't the vowels of pre-OS "*-so" and the other KP, "*-do," differ from each other? And if they do,

what is the, proper characterization of that difference? The solution for the differential raising of the vowel of "*-so" and that of "*-do" is that

the vowels do indeed differ from each other: they are, respectively, *-swo

and *-do. Yamada (1954) already suggested the cognation of OJ so and its

free variant zo to Okinawan (and also OS) -du-see (21)-which uncon-

troversially comes from older *-do, that is, a mid back vowel. This shows

that PJ had to have a Type-B *o here:

(21) OJ so - zo :: Ok/OS -du (< *-do) < PJ *Co

and that glide-onset (=Type A) vowels undergo raising earlier ("Raising

1"=R1 below) than non-glide-onset (=Type-B) vowels ("Raising 2"=R2

below). Either one of the following orders of sound changes in (22) will

work; the end result of the sound changes affecting the reconstructed *-swo

and *-do represents the forms of the OS, the allomorphy of which can only

be predicted by the differing vowel qualities of the reconstructed forms:21)

19) Note that in the environment after a sibilant, the PR vowel *wo can change all the way to Modern Naha yi. Note further that the Type-B vowels did not

participate in this raising, Raising 1. 20) Also Shuri ?asib-, OGJ 1963: 125a, not x?asyib- or x?asub-. 21) Takahashi (1982: 223) homes in on the fact that the only example of PP with

Page 12: On the Old Japanese Kakari (Focus) Particle koso

12 Leon A. SERAFIM and Rumiko SHINZATO

(22a) PR *... yi-swo~*... V-swo| *... yi-do- *... V-do R1 *... yi-swo *... V-su | *... yi-do- *... V-do

PP *... yi-sywo- *... V-su| *... yi-dyo- *... V-do

merger *... yi-syo- *... V-su |*... yi-dyo- *... V-do

SUa *... yi-syo- *... V-si| 1 *... yi-dyo- *... V-do

R2 ... yi-syu- ... V-si | ... yi-dyu- ... V-du a Sibilant-based Unrounding

(22b) PR *... yi-swo *... V-swo |*... yi-do- *... V-do R1 *... yi-swo *... V-su| *... yi-do- *... V-do

merger *... yi-so- *... V-su | *... yi-do-- *... V-do

PP *... yi-syo- *... V-su| *... yi-dyo- *... V-do

SU *... yi-syo- *... V-si| *... yi-dyo- *... V-do

R2 ... yi-syu-- ... V-si | ... yi-dyu- ... V-du

The above arguments serve to account for the differing allomorphies

of the two KPs of the OS. The next section will address the differences

between PR *-swo and OJ *koso, and how they may be reconciled, in

order to reconstruct the PJ form.

2.4. Reconstruction of Proto-Japonic form *ko#swo

The reconstruction so far has hypothesized a PR KP *-swo. Starting from

this *-swo, a proportional analogy can predict an expected OJ form, which

can be compared with the actual form. Going back to parallels between

the verb stem PR *aswob- and PR *-swo, the following proportion in (23)

predicts *-swo for X:

the syllable <su> -i.e., *su-is the KP <<su>> . His observation allows us to start

off phonologically from the interesting exceptionality of this allomorphy, which

requires two alternating high vowels after a sibilant, i and u. This alternation in

turn requires the PR vowel *wo. Here we only argue for *swo, and therefore do

not fully develop a theory of PR *wo and *ye that sometimes undergo Raising 1

and sometimes Raising 2. For that, see Serafim 2003.

Page 13: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 13

(23) Proportion: OS ?a si b-:: OS -si

PR *a swo b-:: PR *-swo

OJ a swo b-:: OJ X X = -swo

That is, "OS ?asib-/PR *aswob- is to OS -si/PR *-swo as OJ aswob- is to

OJ X, and X is -swo." But we don't get OJ x-swo. Instead we find the KP

of equivalent function, koso. Let us now compare the expected (but unat-

tested) OJ form x-swo with the actual OJ form koso, as in (24):

(24) Actual and expected forms: actual - k o - s o

expected - s wo

The proposed relationship is between the two actually existing KPs, so we

array forms in (25):

(25) Parallels between PJ *aswob- and PJ *-swo, and their related OJ and OS forms:

OS ?as i b - 'play' **- s i '(KP)'

PR * *aswob - '(id.)'** - s wo '(KP)'

OJ *aswob - `(id.)' k o - s o '(KP)'

PJ **aswob - '(id.)' * * -swo KP '

The reconstructed PJ candidate forms *(ko)-swo%so would seem to raise

two questions: (1) The reconstructed KP portion PR *-swo with type-A *wo is at odds with the OJ portion -so, with its type-B vowel o, tradition-ally reconstructed as PJ *-so. (2) So far it is unclear what effect the lack of

an equivalent piece in Okinawan to OJ ko- will have on the reconstruc-

tion of the overall PJ form. The problems associated with question (1) will

be resolved in favor of PJ *-swo, meaning `one' or `thing' (i.e., having a

nominalizer function), and, as for question (2), we will show that PJ had *ko--thus , the morpheme sequence will turn out to be *ko#swo, with

Page 14: On the Old Japanese Kakari (Focus) Particle koso

14 Leon A. SERAFIM and Rumiko SHINZATO

an internal word boundary. We propose to explore the question in detail

below.

In order to explain how a proposed PJ *ko#swo can be the ancestor of

OJ koso, we need to consider the notion of word-internal vowel harmony,

which naturally leads to Hattori's work on the topic from the 1970s. Now,

Hattori (1976a,b) explains some cases of alternation as due to change of

a vowel for the purposes of preserving vowel harmony within a word.

Examples he gives22) are as in (26): *wuko- foolish'>*wuko *wukwo-

>OJ 'ukwo- (Hattori 1976a); and *puto- `fat'>*pwuto-->*pwutwo->OJ

putwo- (Hattori 1976b). The vowel-harmonic process works left-to-right.

(26) Hattori's proposed harmonic shifts:

Pre-OJ * wuko- (*[wika]) `foolish' * puto- (*[pita]) `fat' * wuko- (* [wuko]) * p wuto- (* [puta] )

* wukwo- (* [wuko]) * pwutwo- (* [puto] )

OJ 'ukwo- ([(?)uko]) putwo- ([puto])

The trigger in both the morphemes is a purely phonetic shift in which labi-

als (e.g., *[w] and *[p]) plus type-B (that is, non-round) * u = * [i] result in

syllables with type-A (that is, round) *wu = * [u]:

(27) Vowel-color changes after labials: * [wi] > * [wu]

*[pi] > *[pu]

The change results in rounded (*wu = *[u]) and unrounded (*o = *[a])

vowels coexisting in the same morpheme, a violation of vowel harmony

(see below), and the automatic application of a left-to-right harmoniza-tion, in these two cases, of *o = *[a] to *wo = *[o]. In addition, the OJ

form 'ukwo- represents a neutralization between older syllable. *wwu = * [wu] and *'wu = *[(?)u] (that is, between wa-gyo and a-gyo syllables). Hattori's formulation has the effect of obeying the inviolable Arisaka's

22) Respelled here.

Page 15: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 15

First Law (Arisaka 1963; and see below), which prohibits tautomorphe-

mic type-A (*)wo and type-B (*)o.23)

Now, let us return to the question of the form of the reconstructed

KP of PJ. Here, as represented in (28), we assume the same process in a

restructuring of the lexicalized compound of type-B *ko = * [ka] this' and

type-A *-swo = *[so] one, (nominalizer)': *wo changed to (*)Q to agree

with the leftward vowel in the unitary morpheme of which it had become

a part.24)

(28) Lexicalization and harmonization of the sequence of two mor-

phemes: PJ * ko#swo = B # A

OJ koso = B B

Our proposed formal and semantic changes are as in (29):

(29) *ko `(proximal deictic)' + *swo '(nominalizer)' -* *ko#swo this one' - koso '(KP)'

Our positing of *swo (A) instead of so (B) for the prehistory of the sec-

ond syllable of the particle koso counters the traditional etymology for

koso as the proximal ko + the mesial so, since the mesial morpheme so is

independently attested as so (B) `that'. See (30):

(30) Dueling etymologies: Traditional etymology: *ko this (proximal)'+*so that (mesial)'

Proposed etymology: *ko this (proximal)'+*swo one (nominalizer)'

Vowel values known from OJ texts are bolded.

23) The rule as given by Arisaka is no more general than that-it is not a gener- alized round/unround tautomorphemic constraint.

24) While the vowel-harmony condition invoked here is the same as Hattori's (i.e., application from left to right), it results in a type-B word, as opposed to

Hattori's type A examples in (26). This is straightforward, given left-to-right application, which is commonly seen in Northeast Asian languages with vowel

harmony.

Page 16: On the Old Japanese Kakari (Focus) Particle koso

16 Leon A. SERAFIM and Rumiko SHINZATO

As already stated, we view *swo as a nominalizer, meaning `one' or `thing'.25> We noted in the introduction that the hypothesis of *'this thing'

was originally formulated in Thorpe (1983). In the space of a few lines, he

claims that the OJ form is from PJ *ko- this' and *su, his form for what

he glosses as `thing'. He notes the existence of (his) *su in the J dialects

that we point out and exemplify below, and in OJ lexicalized forms, as

given (with differences in detail) in our preceding note. He states that Ryukyuan does not have this type of KM, apparently unaware of the

purport of the KP -si - -syu.26) He also points out an Oita-dialect kosu

25) Does OJ evince this nominalizer at all? Yes, but only as a lexicalization onto some animal (mostly bird?) names, from PJ times, and is thus found in OJ as

well: birds: karasu `crow, raven'<*kara-swo 'black-one', kyigyisi 'pheasant'< *kyinkyi-swo-yi 'pheasant-one-NOM', ugupisu `bush-warbler'<*unkupi-swo 'bush-warbler-one' (or perhaps 'dace-one', named after the fish called the dace,

ugupi < *unkupi); fish: kisu `a sillaginoid fish'; also the generalized words refer- ring to male and female animals: wosu<*wo-swo 'male-one', myesu<*mye-swo

'female-one'. Apparently this ending has been generalized to -su-this fact sug- gests that Japanese has treated suffixes as a special category, not subject to the rule that retains mid height at the end of words. Instead, the mid height is re- tained by the portion of the word that precedes it: *mye-swo>mye-su, not Xmyi-

swo. Apparently a preceding high vowel is unable to block raising in the case of the use of *-swo as a suffix in the line leading to OJ: *unkupi-swo>ugupisu, not

Xugupiswo. (The vowel of *iswo does not raise, but this is a case of a monomor- phemic word, whereas the items treated here have a suffix *-swo.)

26) Thorpe (1983: 251) in treating uses of the izen form in PR, briefly alludes to an entry in Hokama (1972: 108), on <su> , the KP at issue in our paper, as

follows: "... Hokama ... describes an odd literary Okinawan example of 1711 showing linkage with an emphatic *so that is otherwise known only in Japanese

(OJ so - zo) and is there associated with attributive verbs." By "an emphatic *so," Thorpe meant OJ KM so-zo, not KM koso, from which we start our re-

construction. Thorpe apparently lacked a first-hand familiarity with premodern Okinawan texts. He may have misconstrued the phonological value of <<su>> as

"*so," not si, though this is unclear from what he writes. Hokama (1972: 108) lists the lexical item under <su> , but in his discussion of it, he points out its

allomorphy in its palatalized variants with both <<...o>>- and < ... u>>-ending spell- ings. Thorpe does not mention how he has come upon his reconstruction of his

*so, which for him has no Type-A or -B status, since his PR reconstruction does not recognize the distinction-it is not possible to determine even whether the

reconstruction is for the language of the OS, or whether it is for PR. Further, Thorpe has wrongly tied his Okinawan *so to OJ so~zo (our so~zo). These

OJ alternants, he correctly notes, bind with the rentai form, while Hokama

Page 17: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 17

with this function. He sees (his) PJ *su as having changed to OJ so by

application of Arisaka's Second Law (Thorpe 1983:226), which states that

tautomorphemic o and u are relatively few in number in the OJ lexicon.

The entire etymology is tacked onto the list of lexicalized forms alluded

to above, and that list in turn supports his etymology of *su thing' in the

negative OJ -azu, and its relation to the postverbal PR nominalizer *su

a (with his vocalic reconstruction). While our hypothesis on the origin of OJ koso incorporates most

of Thorpe's points, it adds the 00k KP -si~-syu to the argument, and

supports the etymology in far greater detail than in Thorpe's very brief

remarks. It also postulates a different vowel for Thorpe's *su (our *swo),

and discusses in detail the original system of semantics, pragmatics, and

grammaticalization that this construction is only one part of. Four pieces of evidence support the claim of *swo as a nominalizer,

the first two empirical, the last two theoretical, suggesting that positing

the so of koso as a mesial is problematic.

First and foremost, the Konkokenshu (Hyoj of o 1711, Hokama 1970,

Ikemiya 1995), the oldest Okinawan dictionary, compiled in 1711 for the

purpose of reading the Omoro soshi, clearly states as in (31) that the KP <<siyo>> = syu (one allomorph of <<su>> = si) and the nominalizer <<koto>>

_ kutu 'thing/one' (here voiced to <<goto>> = -gutu; Japanese koto%goto) are

synonymous.27) See (31):

(31)「『 だ り しよ』 『だ に ご と』 二 つ 共 され ば こそ と云 心」混 効 験 集1711`<<darisiyo> <<danigoto>> futatsu tomo sarebakoso to iu kokoro'

`Both daryisyu and danyigutu mean "sure enough"' Konkokenshu

clearly states that the OS KP <<su>> is equivalent functionally to the J KP koso, and that <<su>> binds with an izen conjugational form. This fact apparently es-

caped his attention. It should become clear from the arguments that we present in this paper that Thorpe's linking of his *so to OJ KM so -. zo is a mistake in

terms of the predicate forms that bind with the OS KP -si - -syu, as is his linking of it to OJ single-syllable forms that have the Type-B vowel o.

27) The forms <<dari> = daryi and <<dani>> = danyi are variants meaning `truly, re-ally'.

Page 18: On the Old Japanese Kakari (Focus) Particle koso

18 Leon A. SERAFIM and Rumiko SHINZATO

This quotation from the KKKS (Konkokenshu) is cited by Ifa, to

semantically link (-si -.) -syu to both koto and koso. Ifa (1910 [original],

1942 [revised], Hokama, ed., 2000: 373 [reprinted]) clearly believes that

the two si of his own Okinawan speech (the Shuri dialect of about a cen-

tury ago), that is, the nominalizer -si `one' and the (obsolete) KP -si (.--syu), share the same origin, and that they, too, mean `thing'. He approv-ingly cites Kanazawa's (1923: 337-338) statement about the origin of the

two Japanese KPs. Kanazawa in turn implies that he believes that both OJ

koso and so are related to Korean kes 'thing'. 21) Kanazawa assumes that

so is simply a form shortened through loss of the first syllable, though he

does not mention how that might have come about in a monomorphemic

lexical item.

Where we wish to differ is in: (1) Kanazawa's linking to Korean kes,

which has generally been linked, instead, to Japanese koto `thing (etc.)';

(2) Kanazawa's notion that a one-morpheme string of two syllables arbi-trarily shortens to get a one-syllable (new) KP, so. We show elsewhere

that, in any case, koso and so in Old Japanese are not related, so we will

not repeat those arguments here. The same holds true for the bi-morphe-

mic origin of koso. Here we only need underscore the fact that having a

bi-morphemic string lose one of its morphemes is simpler as a hypothesis

(quite apart from other arguments that have also been brought to bear) than having a monomorphemic lexical item arbitrarily be chopped in two,

and have one of those parts be deleted.

Second, the nominalizer -not the deictic -origin of -so (< * #swo) in

koso finds corroboration in dialect-geographical data. Extreme northeast-

ern Kyushu and extreme western Honshu have nominalizers su and so,

respectively, both < *swo, cognate to the modern Shuri Okinawan nomi-

nalizer -si,29) identical in origin to the KP, 00k si. See (32):

28) It is not entirely clear from the passage whether he also believes in a genetic relationship among them.

29) Cf., for example, OGJ -si (our -si) '(nominalizer)'.

Page 19: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 19

(32) Nominalizer shapes in Okinawan, NE Kyushu, and W Honshu: Shuri NE K W H

-Si [si] su so/ho

*-su < *-swo (raising) *swo (raising) *swo (raising blocked)

Examples (33a) and (33b) are from an extreme western Honshu dia-

lect, Ebara.30

(33a)モ トノ ミチ ー デ ラ レダ ッタ チ ュ-

moto-no mici-i de-rare-d-atta cyuu

original road-to go-out-able-not-PAST APPOSITIVE

ソ ガ

so ga

NOMINALIZER SUB

キ ツネ ガ … バ カ シ タ ソ デ ア リマ シ ョー

kicune ga bakasita so de ar-imas-yoo

fox SUB changed-shape NOMINALIZER COP-POLITE-INFERENCE `The fact is , you see, they couldn't go back on the road that they

had come on, and the fox changed into human form (and accosted

them).'

(33b)… カ ン テ ン ノ ニ ル ソ ガ サ ンボ ン[「煮 て使 う寒 天」の こ と。]

... kanten no niru so ga san-bon

... agar-agar GEN boil NOM SUB three-sticks ...three sticks of agar-agar that you boil' (lit., `a boil-one of agar-

agar, three sticks')

Now we move on to theoretical evidence: Thirdly, positing the so

of koso as a mesial demonstrative like the other OJ KP so/zo, as in Ono

(1993: 120-121), would make it difficult to explain why in this case so/zo, which as a KP itself normally calls for a rentai form, does not agree with a

rentai form in this case, but rather with an izen form. See (34).

30)NHK 1967:232,243:Yamaguchi-ken Mine-gun Akiyoshi-ch6 B eppu Ebara

(山 口県 美 祢 郡 秋 芳 町別 府江 原).

Page 20: On the Old Japanese Kakari (Focus) Particle koso

20 Leon A. SERAFIM and Rumiko SHINZATO

(34) Clashing agreements: OJ so '(KP)' = rentai ending

OJ koso '(KP)' = izen ending

Fourthly, in historical pragmatics, it is well known that demonstra-

tives grammaticalize as focus particles (Luo 1997; Diessel 1997; Heine and

Kuteva 2002), but, outside of this accepted Japanese etymology, we know

of no other example of a combination of proximal and distal deictics

together being grammaticalized as a single focus marker. This is because

having two different deictics focusing one item is somewhat like saying, `X, this and that, is Y,' and thus it is self-contradictory, obfuscating the

very function of pointing to an entity as a focus. See (35):

(35) X`X, this and that, is Y'

Now, was the starting point in Proto-Japonic *ko#swo, or was it just *swo? All else being equal, it would be simpler to choose *swo alone-but all else is not equal. While *ko#swo means `this one', *swo alone only

means `one'. When the KP is appended to a phrase as an appositive, it

must point to it, as it does with the PJ distal deictic *ka yonder (one)' and

the mesial deictic *so that (one)'. Thus there should be a deictic, at least.

Since the Ryukyuan form has no deictic, it must have lost the first mor-

pheme at a time when the boundary between the two morphemes was still

present-see (36). This means that the boundary still existed in PR, since without it there would have existed no reasonable way to drop the piece *ko. Furthermore, the vowel of *swo would certainly have changed due to

vowel harmony to yield *...so if the boundary had been deleted. Thus we

hypothesize (36):

(36) PJ *ko#swo > PR *ko#swo -> *swo.

Based on the arguments above, we dispute the traditional and popu-

lar double-deictic origin hypothesis for OJ X koso and instead hypoth-

esize the developmental paths given in (37):

Page 21: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 21

(37) OJ X koso... < pre-OJ *X ko#swo *'X, this thing/one, ...' 00k X si... < pre-OOk *X swo *`X, (the) thing/one...'

< PR *X ko#swo *`X, this thing/one...'

To wrap up: This section has argued that the original form in Proto-

Japonic of the ancestor of the kakari particle is *ko#swo this one'. In

the line leading to OJ, the form was fully lexicalized as koso, assuming

the status of a KP. In the line leading to Okinawan through PR, the form

lost its first morpheme, since the second one sufficed functionally. This

left only the portion *-swo, which then underwent a series of changes to

give the allomorphs seen in the OS, namely -si (plain alternant) and -syu

(PPed alternant), as in (38).

(38) OS: -si (plain alternant) and -syu (PPed alternant)

The portion that both lines of development share is *swo one'.

3. On the meaning of the correlation between koso and

the izen conjugational form

3.1. Issues concerning the izen conjugational form31>

Within Japanese, it strikes all students of KM that all KPs correlate with

the rentai form with the exception of koso, which requires the izen form.

In order to understand this off-norm correlation, we believe we have to

take into account the existence of a subtype of KP ka KM, in which ka

cooccurs with the inference/intention auxiliary -(a)m- in early Japanese.

(The cognate of ka co-occurs exclusively with the cognates of the auxiliary -(a)m- in Ok, and in the Hachijo dialect.)

(39) KP Auxiliary ka - (a) m-

31) In order to avoid confusion, we use Japanese terms such as wizen, rentai, and izen where we refer to formal (i.e., conjugational) properties, but use realis and irrealis when we refer to semantic properties to be discussed in 3.3.

Page 22: On the Old Japanese Kakari (Focus) Particle koso

22 Leon A. SERAFIM and Rumlko SHINZATO

In this comparative context, we suggest that the izen conjugational form

must be approached in two ways: First, it must be approached formally,

i.e., as a paradigmatic form (as an izen form vis-a-vis a rentai form).

(40) KP Form ka RENTAI

k s IZEN

See 3.2 for this formal characterization in detail. Second, it must be

approached semantically, i.e., in its semantic character of 'realis', which is

embodied in all izen conjugational forms, including the izen form of the

auxiliary -(a)m-, namely -(a)m-ey. Our characterization of the izen form -(a)m-ey as 'realis', as opposed to the 'irrealis' denoted by the auxiliary -(a)m- in the rentai form, namely -(a)m-u, may be counterintuitive to some readers for rightful reasons, and therefore the issue will be further

elaborated in section 3.3.

(41) Auxiliary Form Semantics - (a) m-u RENTAI IRREALIS

-(a)m-ey IZEN REALIS

We are aware that the inference/intention auxiliary -(a)m- does also cooc-

cur with KP ko(so) in its realis (i.e., izenkei) form, -(a)m-ey (> MJ -(a)m-

e). However, -(a)m-ey, with an izen form, is different from -(a)m-u, with a

rentai form, the ending called for by KP ka, in that the izen-form -(a)m-ey

expresses a strong assertion by the speaker, while the rentai-form -(a)m-

u does not. Thus, even if the same inference/intention auxiliary should

appear as the ending for both KP ka and ko (so), their epistemic standings

are quite distinct.

(42) KP Auxiliary Form Semantics Epistemic stance ka -(a)m-u RENTAI IRREALIS lack of strong assertion

by speaker ko(so) -(a)m-ey IZEN REALIS strong assertion by speaker

Page 23: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 23

The traditional pure-formalist approach obfuscates the izen form's se-

mantic importance.

3.2. From the formal point of view: rentai vs. izen

It is often suggested that KM is syntactically a cleft construction,

having a focus (cf. it is - in English clefts), and a presuppositional relative

clause (that -). For instance, Harris and Campbell (1995: 153) characterize

clefts as follows: "The cleft (a) consists of a superordinate clause (S) and

a subordinate (S2), (b) the former containing a copula, and (c) the latter

having the structure of a relative clause." Clefts are often assumed to have

originated in equational sentences, in which two nominals, or nominalized

clauses (or relative clauses) were connected by a copula (Heine and Reh

1984; Harris and Campbell 1995; and Harris 2001). As a natural course of

development, the two sentences become amalgamated syntactically as one

sentence, the result of which is the cleft construction. For instance, Harris

(2001) describes the development of clefts in literary Dargi, a language of the North East Caucasian family, as in (43a, b):

(43a) x'o sa j-ri uzbi arku-si 2SG.ABS.FM COP-CM-2SG] brothers ABS bring PART SG

`YOU brought the brothers .' `It is YOU, having brought the broth-

ers.'

(43b) x'o-ni sa j-ri uzbi arku-si 2SG. ERG. FM [COP] brothers bring PART. SG

`YOU brought the brothers .'

(43a) represents a biclausal structure in which x'o `you' is the absolutive subject of the copula, and the verb arku-si `bring' is a participle, used as a

relative clause. Here, x'o sa j-ri `it is you' and uzbi arku-si `brought [the]

brothers' are two separate constituents, and the latter constituent cannot

be broken up by the insertion of an element from the former. (43b) shows

characteristics of monoclausal structure with ergative-case x'o-ni `you',

Page 24: On the Old Japanese Kakari (Focus) Particle koso

24 Leon A. SERAFIM and Rumiko SHINZATO

showing it is the subject of the transitive verb, arku-si `bring'. In addi-

tion, the copula, reanalyzed as a focus marker, no longer agrees with the

focused constituent, X'o-ni you'.

Quinn (1997:78-80) proposes a similar developmental scheme for

Japanese KM. For instance, the sentence, nani ka sayareru `What is it that

is in the way?' in (44),

(44) Nani ka sayareru what KP interfering `What is it that is in the way?'

has evolved according to Quinn (1997:79), from a biclausal structure as in

(45),

(45) *Nani ka - sayarer-u what cop interfere-ADN `What is it? -the one in the way.'

in which the nominalized clause, formed by a predicate with a rentai end-

ing, is added as an afterthought topic. Quinn views ka as originally the

distal deictic, but here already functioning as a copula, combining a nomi-

nal focus (i.e., nani) and a nominalized clause (i.e., sayareru). When the

afterthought clause came to be perceived as part of the sentence, then the

original biclausal structure was grammaticalized as a KM construction.")

See (46) for the development of the focus particle:

(46) *ka `(distal deictic)' > (copula)' > '(KP)'

32) In Japanese traditional linguistics also, a similar biclausal-to-monoclausal hy-

pothesis (nibun renchi setsu) has been proposed by Nomura (1995, 2001). Just

like Western linguists, Nomura takes the clause following a KP to be a nominal-

ized clause acting almost like an unbreakable word allowing no intervention

within it. (cf. Literary Dargi example above.) His hypothesis is substantiated

by two facts: (1) that the word order is fixed as ... ha (= Topic) KP ... no/ga (=

Subject)... and (2) the ending part (no/ga ... rentai) behaves like a nominalized

clause (˜A‘Ì‹å). His hypothesis has been regarded as a promising alternative to

Ono's (1993) inversion hypothesis and Sakakura's (1993) insertion hypothesis.

Page 25: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 25

In order to view KM as a cleft construction, the ending must be a rentai,

i.e., nominalized, form. Therefore, the izen conjugational form has tended

to be regarded as somewhat odd.") In the present paper, however, we

argue that there is nothing strange about the izen conjugational form.

Like other, rentai, forms, the izen conjugation form, too, is a nominalized

form. That is, koso-KM also forms a cleft construction, just like other KM

constructions.

For evidence to claim the izen conjugational form as a nominalized

form, we incorporate the following -see (47):34)

(47) OJ -(ur)ey < *-(ur)agi < *-(ur)ua -gi/ *-(ur)au -gi

IZEN < *IZEN < *ADN NOM

Hayata (1998: 31) hypothesizes that the OJ izen form -(ur)ey comes from

Pre-OJ *-uragi, itself perhaps from either *-urua-gi or *-urau-gi. He views *-gi as a nominalizer. He sees the formation of the izen ending as coming

about through the combination of the original rentai ending (either *-urua

or *-urau, according to him) with the nominalizer *gi, which, Hayata

suggests, still exists in unchanged form in the dialect of Saga, northern

Kyushu. This dialect forms a mood of supposition by joining its rentai

form with a nominalizer -gi, as in (48):

(48) tab-uru + -gi -* tab-uggi eat-ADN + NOM *eat-SUPPOSITIONAL

`supposing one eats'

This form, consisting of the same original elements but with the output

33) For example, Ono (1993), who otherwise always opts for the inversion hy- pothesis, in the case of koso-type KM, opts for the insertion hypothesis. He sees

the inversion as requiring a rentai form, but he does not treat the izen form in the same way, suggesting that he does not view the izen and rentai forms as hav-

ing the same status. 34) We have slightly modified the proposal by adding u before r-this in no way

alters Hayata's main point.

Page 26: On the Old Japanese Kakari (Focus) Particle koso

26 Leon A. SERAFIM and Rumiko SHINZATO

differing, is exactly equivalent to OJ tabure < * tab-urey < * tab-urua-gi.

The suppositional function is one of the functions of the izen form. If

Hayata's hypothesis is correct, then the izen is indeed a nominalized form,

and koso-type KM is far more similar in character to the other KM types

than previously realized, since it, too, forms a cleft structure.

3.3. From the semantic/pragmatic point of view:

irrealis vs. realis endings

The irregularity of the izen conjugational form in the KM construction

has generally been discussed in syntactic terms, as pointed out in the pre-

vious section. However, its semantic nature has largely been neglected.

This is because the koso-type KM has never been contrasted with the ka-

type KM. Two striking parallels in these two constructions have escaped

the attention of researchers: (1) their KPs, ka and ko(so) originated in

demonstratives (see section 2 above); and (2) they both have epistemi-

cally-laden endings (irrealis vs. realis).35)

To begin with, let us clarify the terms irrealis and realis. The distinc-

tion between 'irrealis' and 'realis', or their rough terminological equiva-

lents, mizen and izen is often considered aspectual: `not yet actualized'

vs. `actualized'. However, the principal distinction between irrealis and

realis that we adopt here is epistemic: it is a matter of whether or not the

speaker commits her/himself to the truth of the proposition. In the case

of irrealis, the speaker is uncertain about the likelihood of the described

event, while s/he is certain in the case of realis (cf. Akatsuka's (1985)

epistemic scale). Taking Central Pomo (Mithun 1995: 370), for instance,

future (i.e., not actualized) could be marked with either irrealis or realis.

The difference is that the future marked with realis indicates a greater

expectation of the event taking place (ibid: 379) than the one marked with

35) OJ so/zo and Ok du are yet another pair, this one derived from the mesial demonstratives. However, since the adnominal ending that they take does not

show any epistemic coloring, this pair has not been put in the scope of compari- son in this paper.

Page 27: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 27

irrealis. Closer to the languages at issue here, in Miyara (Ishigaki) dialect

of Ryukyuan, the izen form is used for a projected future event which is

certain to occur. According to Izuyama (2003:91), the following question

is uttered when the speaker knows for sure that someone will write.

(49) ta: du kak ja who FOCUS write-STATIVE-REALIS

`Who is going to write?'

We believe that a similar parallel can be applied to so-called future

auxiliary -(a)m-: The rentai form with which KM ka binds expresses less

certain projection of future events, while the izen form with which KM

koso correlates indicates a higher likelihood. We view the rentai of -(a)m-,

i.e., -(a)m-u, to be semantically irrealis, while the izen-marked -(a)m-ey is

realis. This contrast is clearly depicted in the following syntactic minimal

pair, in which both predicates have the auxiliary -(a)m-, co-occurring with a wh-word, but without a KP: the rentai form -(a)m-u shows the speaker

wondering as to the location where her love might succor her in case of

an unexpected and unfavorable turn of events, while the izen -(a)m-ey

expresses the speaker's strong conviction that her husband would not

leave her behind.36)

(50a) 大き海を さもらふ 水門 事 しあらば

ohoki umi wo samorahu minato koto si ar-aba

big ocean Do watch harbor accident EMPH happen-if

いつへゆ 君は 我を 率 しのがむ

idu he yu kimi ha wa wo wisinog-am-u

where to my love TOP me DO take-INFERENCE-RENTAL `In the harbor/peopled by lofty ships/ should misadventure overtake

us, /whither will you succor me?' (MYS 1308 [Honda 1967])

36) Since -(a)m- is identical in its rentai and shushi forms, it is uncertain whether the forms are indeed rentai. There are some controversies over this issue (see

Oshika 1991, inter alia), but it is not of concern here, because -(a)m-u in (50a) shows less certainty than -(a)m-ey in (50b).

Page 28: On the Old Japanese Kakari (Focus) Particle koso

28 Leon A. SERAFIM and RUmlkO SHINZATO

(50b) 我が背子を いつち 行かめと さき竹の

Waga seko wo iduti yuk-am-e to... Sakitake no

My love SUB where 9n-INFERENCE-T7FN OT PW

そがひに 寝 しく 今 し 悔 しも

sogahi ni ne-si-ku ima si kuyasi-mo

back-to-back sleep-PAST-ing now EMPH regrettable-sr `Little knowing that my husband would / be early taken from

me, / I refused him in our bed, / and now, too late, I grieve.'

(MYS 1412 [Honda 1967])

The semantic difference between irrealis and realis is also reflected in

00k mizen vs. izen forms. As explained below, 00k mizen is all that

remains of the auxiliary -(a)m-u. In the following, (a) is an example of

mizen, while (b) is an example of izen -(a)m-e. Just like the OJ examples,

the mizen form expresses lesser certainty than the izen form. 17) かぐらお てお あす よ

(51a) 神座居て 手折 り遊び し居 らい

kaguraote teoriasubi siyorai>>

kagura-wuti tiwuryi-?asibyi syi-yur-a-yi

[kagurawuti tiwurri?asibii Sijuraji] heaven-being arm bending-sacred dance dO-PROGRESSIVE-INFER-

ENCE/MIZEN-QP

「天上の神座に居て神遊びをしているのだろ うか。」

`In heaven, could he be doing a sacred dance with his hands?'

(OS 1501 [Hokama 2000])

37) Some readers may find our placement of morpheme boundaries in the Oki- nawan examples perplexing. We have, in fact, erred on the side of historical

rather than synchronic accuracy in our morpheme cuts, in order to facilitate the reader's interpretation of the equivalents at different historical stages and in different languages. A fully synchronic analysis would be unlikely to adopt the

morpheme breaks used here. The reader may prefer merely to consider them markers of correspondence.

Page 29: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari(Focus)Particle koso 29

あやみや あまごあ もど

(51b) 綾宮の大ころ 眼合わちへ 戻らめ

<<ayamiyano OOkoro amako awatihe modorame>>

?aya-miya-nu ?ufu-kuru ?amaku ?awa-cyi mudur-am-i

[?aj amJij a-nu ?u(Dukuru ?amaku ?awat f i mudurami] splendid-enclosure-'s big-men eyes lock-ing return-INFER-

ENCE-IZEN

「綾庭の長老,男たちは眼を合わせて,顔を合わせて,戻るだろう。」

`The elders at the splendid enclosure will (surely) lock eyes and

return.' (OS 1411 [Hokama 2000])

Given the above, we semantically characterize the OJ rentai -(a)m-u and 00k mizen forms as irrealis (i.e., uncertainty) as

opposed to the OJ and 00k izen forms as realis (certainty).

On the basis of this irrealis vs. realis contrast, let us examine ka-type

KM and koso-type KM. The co-occurrence of the auxiliary -(a)m- with

KP ka and their semantic affinity have been noted in previous studies

(Sakakura 1993; Yamaguchi 1989; Serafim and Shinzato 2000). Accord-ing to our previous paper, Serafim and Shinzato (2000), the occurrence of -(a)m- with ka amounts to 72.5% of the entire occurrence of ka-type KM

sentences in the Man'yoshu, as in (52):

(52) ka ... (a)m- = 72.5% of ka-type KM in MYS

We found a similar tendency in Taketori Monogatari, the earliest novel,

in a slightly later period. As is well-known, the KM with the KP ka forms

a self-addressed question. (See Sakakura 1993, Quinn 1997, Serafim and

Shinzato 2000, Nomura 2001, and Shinzato and Serafim 2003.) Observe

examples (53) and (54).

Page 30: On the Old Japanese Kakari (Focus) Particle koso

30 Leon A. SERAFIM ariCl RU1111kO SHINZATO

(53)こ も り くの 一泊 瀬 の 山 の 山 の 際 に

komorikuno hatuse no yama no yama no ma ni

Pw Mt. Hatsuse 's mountain's brow at

い さ よふ 雲 は 妹 に か あ らむ

isayohu kumo ha imo ni ka ar-am-u

drift clouds TOP maiden COP KP be-INFERENCE-RENTAI `The vapor hanging over Mt. Hatsuse / is she, is she?' (MYS 428

[Honda 1967] )

(54)我 妹子 に 猪名野 は みせつ 名次山

wagimoko ni winano ha mise-tu nasugiyama

my wife to Inano TOP show-PERF Mt. Nasugi

角 の 松原 いつ か 示さむ

tuno no matubara itu ka simes-am-u

Tsuno 's pines when KP show- INFERENCE-RENTAI `Now have I shown my love Inano Field, / but when will she see Mt.

Nasugi, and the pines at Tsuno?' (MYS 279 [Honda 1967] )

Just like its Japanese counterpart, the cognate ga-type KM of Oki-

nawan also forms a self-inquiry as seen in the next two examples, (55) and

(56). .

(55) Old Okinawanた と よ た う よ

誰 が 取り居ら 誰 が 打ち居ら

to ga toriyora to ga utiyora>

to ga tur-yur-a to ga ?uc-yur-a

[ta ga turrura to ga ?ut f ura] Who KP take-be-MIZEN Who KP beat-be-MIZEN

「誰 が 取 り,誰:が 打 つ の で あ ろ う。」(OS 1157[Hokama 2000])

J wonder who is taking it, who is beating it.' (Referring to a hand-

drum.)

Page 31: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 31

(56) Okinawan ?aree nama syumucyi ga yu-door-a

he.TOP now book KP read-PROG-MIZEN `I wonder if he is reading a book now.'

The difference is that the Okinawan ga-type KM exclusively binds with

the mizen form. However, in our previous study (Serafim and Shinzato

2000), we hypothesize that this mizen conjugational form originated in the

auxiliary *-(a)m-u (rentai), from which *...m-u was dropped, as in (57).

(57) *-am-u > -a

Based on these comparative grounds between Old Japanese and Oki-

nawan, we reconstruct a subtype of ka-type KM in Proto-Japonic as fol-

lows, in (58):

(58) *... ka ... -(a)m- -wo KP INFERENCE RENTAL

Corroborating this reconstruction is the fact that in the Hachijo

dialect, the direct descendant of the Eastern dialect of the Old Japanese

period, the KP ka also obligatorily calls for the auxiliary noo, a corol-lary of -(a)m,38) as in the following example, (59), taken from Kaneda

(1998:70):39)

(59) ダ イ カ ヨキ ー タ ン ノ ウ?

Dai ka yokiita n noo.

who KP sent NOM INFERENCE

(Hachij o dialect, while opening a box)

38) This is a direct descendant of the Eastern dialect variant -nam- of the Cen- tral OJ inference auxiliary -ram-, in its adnominal form, i.e., -namu: thus, -namu

> noo. See Serafim and Shinzato 2000:85. 39) "Regardless of whether an addressee is present, it is used as in the manner of

self-inquiry. The addressee is not expected to respond."

Page 32: On the Old Japanese Kakari (Focus) Particle koso

32 Leon A. SERAFIM grid RUmlkO SHINZATO

だ れ が よ こ し た の か な あ?

Dare ga yokoshita no ka naa.

who SUB sent NOM KP SP

(Standard Japanese equivalent) `I wonder who sent it to me. 140)

The foregoing leads to the following correspondence: ka-type KM ::

irrealis.

Let us now turn to koso KM. Both OJ koso-type KM and 00k

su-type KM express strong assertion, as seen in examples (1)-(4). This

semantic characteristic of strong assertion, or certainty, still holds even if

the ending is the inference/intention auxiliary, though in this case, it is in

the izen form, -(a)m-ey, not the rentai form, -(a)m-u. In 00k, for instance,

in example (60), the second sentence, the one with the KP syu correlating

with the future auxiliary -(a)m-e, is given as a reason that lamenting is

not warranted (see also omoro 31 and duplicate omoros 631, 877, 1491,

and 1510). The second sentence can be interpreted as the reason for the

speaker's firm belief in the fact that lamenting is not warranted, namely

that Senkimi will protect him. This strong conviction is the reflection of

the realis nature of the izen form, -(a)m-e. あ な きよ なげ せ きみ し

(60)吾 が成さい子… 嘆 くな 精ん君 しゆ 知 りよわめ

<<aganasaikiyo nagekuna seNkimisiyu siriyowame>>

a-ga nasayi-kyu nagiku-na si-N-kyimyi-syu syiryi-yuwa-m-i41)

[gaga nasajikku nagikuna sqkJimii f u f irJijuwaml] I-'s father-child lament-not spiritual.power-'s- know-Ex-

prlestess-KP INFERENCE-IZEN

40) In the original Japanese text, the word by word glosses were not available. The English translation was added by the authors.

41) The exalting auxiliary is irregular, and in this allomorph, -yuwa- is all part of the stem.

Page 33: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari(Focus)Particle koso 33

「我 が … 父 な る 方 よ,嘆 き 給 う な。精 ん 君 神 女 こ そ,守 護 し 給 う

で あ ろ う。」42)

My lord, do not lament; none other than the priestess Senkimi will

protect you.' (OS 710 [Hokama 2000])

In the next example and others (omoro 36, more in omoro 853 below),

the verb + the izen form of the inferential -(a)m-e is paired with the verb

in its izen form, but without the inferential -(a)m-, in a series of discon-

nected parallel lines. This is suggestive of their synonymity: Just as the

izen forms without the auxiliary -(a)m- express the speaker's certainty, so,

too, does the izen form -(a)m-e also express the speaker's commitment to

the proposition. うら かず きみ まぶ うら かず

(61)浦 の数 君君43)しよ 守れ 浦の数

<<ura no kazu kimikimi siyo mabure ura no kazu

?ura-nu kazi kyimyi-gyimyi-syu mabur-i ?ura-nu kazi

[?uranukazi kJimJigiimii f u maburi ?uranukazi village-'s no. priestess-es-Kp protect-IZEN village-'s no.きみ まぶ

君君しよ 守れ

kimikimi siyo mabure>>

kyimyi-gyimyi-syu mabur-i

kJimiigiimJi f u maburi]

priestess-es-KP protect-IZEN「村 ご との 神 女 た ち こそ船 を 守 り給 うの だ。」(Repeat.)

`It is the priestesses of the inlet's villages who will protect (the ship

in its voyage).' (Repeat.)

42)Hokama gives the same translation in(1510). In(631)it's守 護 し 給 う こ と で

あ る.In(1491)it's守 護 し給 う の だ. We believe that the final two translations

are preferable.

43)The ditto mark shaped like an elongatedくis used in all these examples for

the second instance of the morpheme.

Page 34: On the Old Japanese Kakari (Focus) Particle koso

34 Leon A. SERAFIM and Rumiko SHINZATO

首里杜 君君しよ 守らめ

<<siyorimori kimikimisiyo maburame

syiyuryi-muryi kyimyi-gyimyi-syu mabur-am-i

[ f ij uriimurJi kJimJigiimJi f u maburami Shuri-grove priestess-es-KP protect-INF-Iz

真玉杜 君君しよ 守らめ

madamamori kimikimisiyo maburame>>

madama-muryi kyimyi-gyimyi-syu mabur-am-i

madamamurri kiimiigiimJi f u maburami

True.j ewel-grove priestess-es-KP protect-INF-Iz

「首里杜の神女たちこそ守るであろう。真 玉杜の神女たちこそ守る

であろう。」

`It is the priestesses of the Shuri Sacred Grove, of the Madama Sa- cred Grove, who will protect it.' (OS 853 [Hokama 2000] )

Furthermore, it is noteworthy that two thirds of all the examples of the

izen -(a)m-e are appended to verbs which denote `govern', `rule', and `protect', in praise of guardian deities, kings, and rulers. In these contexts, the most natural interpretation of the izen -(a)m-e is not the speaker's

vacillation, but rather his/her strong advocation for the powers of the gods

and the rulers.

(62) Verbs co-occurring with -(a)m-e a. <<maburame>> `protect'

b. <<maburiyowame>> `protect (exalting)'

c. <<sirame>> `govern'

d. <<siriyowame>> `govern (exalting)'

e. <<miyowame>> `govern (exalting)'

Based on the foregoing, we obtain the following correspondence: koso-

type KM:: realis

As noted above, the semantic affinity between the KP ka and irrealis

(i.e., the rentai -(a)m-u) has been noted in previous studies (Sakakura

Page 35: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 35

1993; Yamaguchi 1989; Serafim and Shinzato 2000). In a surprising con-

trast, the affinity between ko and the semantically realis (i.e., izen) end-

ing has never been brought up. We believe that this is probably due to

the conventional etymology of koso as a proximal ko + mesial so. If our

hypothesis of decomposing koso as proximal ko + nominalizer *swo is

correct, then there is indeed a similar affinity between ko and the semanti-

cally realis (i.e., izen) ending. That is, the OJ (00k) KM represents such

ties as distal-ka(ga)-irrealis and proximal-ko-realis. We lay this out as fol-

loWS:44)

(63) * ko#swo `this one near me' :: realis * ka `yonder (one)' :: irrealis45)

44) Hashimoto (1982) sees the OJ deictic system as being divided into two cat- egories: kankakuteki shiji 'sense-based deictics', and kannenteki shiji 'cogni-

tion-based deictics'. The proximal deictic ko and the distal deictic ka belong to the former, and the mesial deictic so belongs to the latter. He further asserts

that OJ ka is derived from OJ ko, primarily based on the fact that there are few instances of ka in MYS, which he takes as suggesting the underdevelopment

of the distal deictic in OJ, and on the hypothesis that ka was derived from ko because the a:: o alternation was common in OJ. However, Quinn (1997:66, n. 11) finds his claim rather puzzling, that "the sense of epistemic familiarity of ko/

ka `this' was basic for historical ka, and its spatial that over there' use derived from this." Given the grammaticalization data in the world's languages, it seems rather unlikely. Diessel (p.c., 2002.9) says that he is unaware of "any language

in which a distal demonstrative directly derived from a proximal demonstra- tive." He further states that he knows of languages losing their distal/proximal

opposition, after which they developed a new two way system by addition of adverbial elements. He ends by noting that, "given this scenario, I wouldn't be surprised if there are languages in which a proximal demonstrative eventually

turns into a distal demonstrative." It should be noted that the pathway to distal from proximal that he envisions is (1) only a possibility, and (2) based on a dif-

ferent observed pathway from one meaning to another than that envisioned by Hashimoto. Diessel's observations, in any case, are quite different from the case

where a proximal remains, but develops a distal element from it through vowel alternation, as proposed by Hashimoto.

45) We are aware that this match is not perfect, since one fourth of OJ ka-type KM examples bind with the -(a)m-u-less rentai form. However, it is also true

that -(a)m-u is obligatory for various Okinawan dialects and for the Hachijo dialect. Thus, we believe this correspondence should not be discounted.

Page 36: On the Old Japanese Kakari (Focus) Particle koso

36 Leon A. SERAFIM and Rurniko SHINZATO

Whether this correspondence is accidental or not will be elaborated in the

context of the grammaticalization of demonstratives into focus particles.

Under the hypothesis that the KPs, ka and ko of koso were respec-

tively derived from distal and proximal demonstratives (developed in

section 2) and in view of the two types of KM as cleft structures, with

their KPs as focus particles, what is portrayed here is the development

of focus particles, ka and ko (so) from the demonstratives ka `(distal)'

and ko `(proximal)' respectively. The evolution of focus meaning from

demonstratives is not idiosyncratic to Japanese. In fact, such a change

has been well documented and recognized as a common path of gram-

maticalization. Heine and Kuteva (2002: 111) state, as in (64): "There is a

cross-linguistic grammaticalization chain-DEMONSTRATIVE>PERS-

PRON>COPULA>FOCUS ...-that can be held responsible, with or

without an intermediate PERS-PRON stage, for the fact that focus

markers can ultimately be traced back to, and may be polysemous with,

demonstratives. "46)

(64) DEMONSTRATIVE > PERS-PRON > COPULA > FOCUS ...

In the same vein, Diessel (1999: 149) illustrates cases in which focus mark-

We are also aware that these correspondences and the explanation to follow do not marshall overwhelming positive evidence. However, given the fact that both

Ok and 00k only had KM of deictic origins (there is no attestation of namu, and the counterpart of OJ ya is a sentence-final particle), KM of deictic origins should be stressed more. We also believe that it is a legitimate question in any

theory, whether it is inversion, insertion, or nibun renchi setsu, to ask whether the KM particles are simply randomly selected to express the rhetorical effects

of doubt, assertion, and strong assertion, etc. If it is not a random selection, then we believe our analysis provides the first working hypothesis. Further, Popper's requirement for a strong and falsifiable hypothesis (Popper 1962),

which stresses the importance of negative evidence, is also met by our hypoth- esis, since evidence to the contrary will falsify (i.e., refute) it, and its withstand-

ing repeated attempts at falsification will strengthen it. 46) This formula is reminiscent of Quinn's characterization of the KPs ka and so

as having quasi-copula function. He also cites Li and Thompson's (1977) cross- linguistic study in this regard, which includes Mandarin, Hebrew, Palestinian

Arabic, and Wappo.

Page 37: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 37

ers were developed directly from what he calls identificational demonstra-

tives.47)

What Old Japanese reveals is the correspondence of proximal-to-

realis and distal-to-irrealis. We believe that this correspondence is in no

way accidental, but rather reflects the transfer of the deictic center from a

spatial domain to an epistemic domain. This is because such a correspon-

dence is cognitively sound, and seems to find corroboration in grammati-

calization in world languages.

For instance, in Ambulas, two demonstrative identifiers ken `proxi-

mal' and wan `distal' have been grammaticalized as focus markers. Since

Wilson (1980) did not give a detailed account as to the difference between

ken and wan, the data are still tentative rather than definitive. However,

it seems that it is possible for wan to appear in a question, while ken has

been reserved for a strong assertion, as below, in (65a) and (65b).

(65a) wan samu bene y-o (<wan `that') FOCUS what YOU(D) do-PR

`What is it that you two are doing?' (Wilson 1980:172-3)

(65b) ken wunat kaperedi waasa kaperedi waasa FOCUS to.me very.bad dog very.bad dog

naadaka (<ken `this')

they.say.and `It is to me that they say, "very bad dog, very bad dog," and...'

(Wilson 1980: 334-335)

This distribution of ken and wan is certainly reminiscent of Old Japanese

ka and ko(so), which form inquiry and strong assertion respectively.

A language like Ambulas is difficult to find, since in the majority of

languages, only one demonstrative, instead of two, is grammaticalizaed as

47) Diessel (1999: 147) distinguishes demonstrative pronouns from identifica- tional demonstratives: the former is a complex free form, while the latter is a

monosyllabic enclitic.

Page 38: On the Old Japanese Kakari (Focus) Particle koso

38 Leon A. SERAFIM and Rumiko SHINZAT0

a focus particle. However, there are languages in which a pair of demon-

stratives were grammaticalized as tense-marker copulas (cf. Heine and

Kuteva's grammaticalization path). For instance, in Kilba, according to

Diessel (1999: 147-148), identificational demonstratives developed into

copulas that mark different tenses: proximal =na has reanalyzed as a pre-

sent tense copula, while distal =nda as a past tense as seen in examples

(66a) and (66b) below (Schu 1983: 321; Diessel 1999: 147):

(66a) u'sman haba" na Usman Kilba is 'Usman is a Kilba (Usman is present).'

(66b) Usman haba' nda Usman Kilba was 'Usman was a Kilba.'

Similarly, Gildea (1993) reports that in Napare, two demonstratives have

grammaticalized as tense markers. The proximal deictic, kej, which rep-resents something within the sphere of perception, was grammaticalized

as a present or immediate future marker. In contrast, the distal deictic,

nej became a past or distant future marker. See Gildea's examples below

(ibid: 60-61):

(67a) maestro kej mej teacher ANIM:PROX ANIM:VISIB `This guy is a teacher here'

[He (PROX) is (PROX) a teacher.]

(67b) maestro nej mej teacher ANIM:DIST ANIM:VISIB `This guy was a teacher here'

[He (PROX) is (DIST) a teacher.]

He (ibid: 63) offers the following figure (X marks the location of the

speaker, and the circle encompasses the sphere of the speaker's perception).

Page 39: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 39

Figure 1. Adding a time line to the sphere of perception

Gildea's figure is reminiscent of Fleischman's (1989: 5-6) layout of

the relationship between temporal and modal (i.e. our epistemic) distance

as shown next:

Figure 2. Conceptual/cognitive extensions of temporal distance

The epistemic distance here should be interpreted as the strength of

the commitment the speaker feels towards the proposition. There is an

inverse relationship between the strength of the speaker's commitment

and the temporal distance. She argues that the further the temporal dis-

tance in either direction, the harder it is for the speaker to vouch for the

truth of the proposition. Thus, what we see in front of our eyes right at

this moment is undoubtedly true. In contrast, what awaits in the future is

impossible to vouch for. Likewise, even what happened in the past is less

easily vouched for compared to the events in the present time frame.

Page 40: On the Old Japanese Kakari (Focus) Particle koso

40 Leon A. SERAFIM and Rumiko SHINZATO

One of the examples she uses to portray the relationship between

temporal distance and modal distance is a set of conditional utterances,

which are traditionally distinguished by the decreasing likelihood of the

proposition in the protasis: probable, improbable, and impossible. Exam-

ples in (68) show a cross-linguistic tendency in the use of tense forms in these three conditional types: correlation of present to `probable' (68a),

past to `improbable' (68b), and pluperfect to `impossible' (68c).48)

(68a) If I have time, I'll write to you Si j'ai [PR] le temps, je t'ecrirai [FUT] (French)

Si tengo [PR] tiempo, to escribo [PR] (Spanish)

(68b) If I had time, I would write you. Si favais [PAST/IMP] le temps, je t'ecrirais [FUT-OF-PAST/COND]

Si tuviera [IMP SUBJ] tiempo, to escribirla [FUT-OF-PAST/COND]

(68c) If I had had time, I would have written you. Si j'avais eu [PLUP] le temps, je t'aurais ecrit [FUT-PERF-OF PAST/

COND PERF]

Si huviera tenido [PLUP SUBJ] tiempo, to habria escribido [FUT-

PERF-OF-PAST/COND PERF]

Based on these consistent cross-linguistic patterns, she (ibid: 6) claims

that "the lesser the likelihood ascribed by the speaker to the situation, the

further in the direction of past will be the tense used to represent it...."

This view is echoed in Heine, Claudi, and Hunnemeyer (1991: 178) as "...the greater the distance from reality, the more remote the past tense,

which is likely to be used to represent epistemic distance."

What emerges from Gildea's and Fleischman's figures put together

is the close relationship among spatial, temporal and epistemic distance.

The realization of this relationship is on the one hand reflected as the

alignment of `here', `now' and 'certainty/strong commitment', and on the

48) All examples are just as in the original.

Page 41: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 41

other hand, it associates there', 'Past/future' and 'uncertainty/weak com-

mitment'. The interrelatedness of these three types of distances, space,

time and epistemics, is, in a way, not surprising, since they are all mea-

sured from the same deictic center, here', `now', and `I' (Biiler 1934).

Based on the foregoing discussion, it is quite reasonable to view the

correlation between the KP markers and their respective conjugational

forms as being well-grounded in this deictic and epistemological principle:

The closer things are to us, the more real they are felt to be .41) The distal

demonstrative ka represents `there', and thus it is suited for expressing

uncertainty. The rentai inference/intention auxiliary (in the case of OJ

and HachijO dialect) or mizen conjugational form (in the case of Oki-

nawan) naturally expresses uncertainty, and it is in this context that the "doubt" that ka... -(a)m-u KM expresses should be understood. On the

other hand, the proximal demonstrative ko denotes `here', and thus is fit-

ting to express certainty, which undoubtedly associates with the assertive

tone that koso ... izen exerts. Both cases should be viewed as connect-

ing spatial and epistemic characteristics. In this context, it comes as no

surprise that koso expresses a stronger assertion than another KP, so/zo,

originating in a mesial demonstrative, as noted in Ono (1993),50) since ko

is proximal, whereas so/zo are mesial.

4. Conclusion

This has been yet another attempt to show how Ok data are valuable for

a more comprehensive analysis of KM. Without Ok data, it was impos-

49) What Givon (1982: 44) calls the proximity hierarchy is illuminating in this regard. It is one of the four hierarchies, i.e., proximity, personal/deictic, sensory/

source, and directness, on which Givon claims subjective certainty is founded. According to his proximity hierarchy, what is near the scene is considered

subjectively more certain than what is away from the scene. This claim is cogni- tively sound and corroborated by the indisputable fact that the accuracy of our

vision decreases as the distance to the object increases. 50) The same semantic differentiation is noted by Hokama (1972; 108) for "<<su> -

lineage particles" as opposed to "<<to> -lineage particles," respectively: "`Su'-kei- retsu no j oshi wa 'to'-keiretsu no j oshi yori, kakari j oshi to shite no shokuno wa

kydsei de aru."

Page 42: On the Old Japanese Kakari (Focus) Particle koso

42 Leon A. SERAFIM and Rumiko SHINZATO

sible to dispute the generally held theory of the bi-deictic origin of koso,

without which it is impossible to see the correlation existing among space,

time, and epistemics in the process of grammaticalization of the KPs,

especially ka and koso.

We believe that the present study contributes to studies of grammati-

calization of demonstratives by offering new data from Japanese.

What was not discussed in this paper is why 00k si was lost, while

Ok ga and du survived. This will be an interesting avenue to pursue, since

the cause of the loss of KM is a controversial area awaiting further inves-

tigation.

Abbreviations

COMP = complementizer PP = progressive palatalization

cop = copula PROG = progressive

DO = direct object PPed = progressively palatalized

EMPH = emphasis Pw = pillow word

EX = exalting QP = question particle

GEN = genitive QT = quotative particle

INF = inference R1 = Raising 1

INF-IZ = inference-izen R2 = Raising 2

io = indirect object SP = sentence particle

KP = kakari particle SUB = subject

NOM = nominalizer TOP = topic

PERF = perfective

Page 43: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 43

References

Akatsuka, Noriko (1985) Conditionals and the epistemic scale. Language 61(3):

625-639.

Arisaka, Hideyo (1963 [1957]) Kodai Nihongo ni okeru onsetsu ketsugo no hosoku

[Syllable-combination rules in Old Japanese]. In: Arisaka. Hideyo, Kokugo

on'in-shi no kenkyu [Studies in the phonological history of Japanese], Revised

and enlarged edition, 103-117. Tokyo: Sanseido.

Biiler, Karl L. (1982 [1934]) The deictic field of language and deictic words. In: R.J.

Jarvella and W. Klein (eds.), Speech, place and action: Studies in deixis and

related topics, 9-30. New York: John Wiley.

Diessel, Holger (1997) The diachronic reanalysis of demonstratives in cross-linguis-

tic perspective. CLS 33: 83-98.

(1999) Form, Function, and Grammaticalization (Typological Stud-

ies in Language 42). Amsterdam and Philadelphia: John Benjamins.

Fleischman, Suzanne (1989) Temporal distance: A basic linguistic metaphor. Stud-

ies in Language 13.1: 1-50.

Gildea, Spike (1993) The development of tense markers from demonstrative pro-

nouns in Panare (Cariban). Studies in Language 17.1: 53-73.

Givon, Talmy (1982) Evidentiality and epistemic space. Studies in Language 6.1:

23-49.

Hagstrom, Paul (1998) Decomposing Questions. Unpublished doctoral dissertation,

MIT.

Handa, Ichiro (1999) Ryukyugo jiten [Dictionary of Ryukyuan]. Tokyo: Daigaku

Shorin.

Harris, Alice C., and Lyle Campbell (1995) Historical Syntax in Cross-linguistic

Perspective. Cambridge: Cambridge University Press.

(2001) Focus and universal principles governing simplification

of cleft structures. In: J.T. Faarlund (ed.) Grammatcial relations in change,

159-170. Amsterdam: John Benjamins Publishing Co.

Hashimoto, Shiro (1982) Shijigo no shiteki tenkai [The historical development of

deictics]. In: Kawabata Yoshiaki (ed.) Kdza Nihongogaku 2, 217-240. Tokyo:

Meiji Shoin.

Page 44: On the Old Japanese Kakari (Focus) Particle koso

44 Leon A. SERAFIM and Rumiko SHINZATO

Hattori, Shiro, Seizen Nakasone, and Shuzen Hokama (eds.) (1974) Ifa Fuyu

zenshu [Complete works of Fuyu Ifa]. Tokyo: Heibonsha.

Hattori, Shiro (1976a) Jodai Nihongo no boin onso wa muttsu de atte yattsu de wa

nai [Old Japanese has six vowels, not eight]. Gekkan gengo 57(5.12): 69-79.

(1976b) Jodai Nihongo no boin taikei to boin chowa [Vowel system

and vowel harmony in Old Japanese]. Gekkan gengo 51(5.6): 2-14. Part of

series, Boin chowa o saguru [Investigating vowel harmony].

(1976c) Ryukyu hogen to hondo hogen [The Ryukyu dialects andthe mainland dialects]. In: Ifa Fuyu Seitan Hyakunen Kinenkai (eds.) Oki-

nawagaku no reimei: Ifa Fuyu seitan hyakunen kinenshi [The dawning of Oki-

nawan studies: A volume commemorating the hundredth anniversary of the

birth of Ifa Fuyu]. Tokyo: Okinawa Bunka Kyokai.

(1978) Nihon Sogo ni tsuite [On Proto-Japanese], 8, Ryukyu Hogen

to Jodai Nihongo to Nihon Sogo [The Ryukyuan Dialects, Old Japanese, and

Proto-Japanese], 4.4. Gekkan gengo 10:94-103.

Hayata, Teruhiro (1998) Jodai Nihongo no onsetsu kozo to o-retsu kO otsu no

betsu [Syllable structure and the distinction between o1 and o2 in Old Japa-

nese]. Onsei Kenkya 2.1: 25-33.

Heine, Bernd, and Mechthild Reh (1984) Grammaticalization and Reanalysis in

African Languages. Hamburg: Helmut Buske.

and Tania Kuteva (2002) World Lexicon of Grammaticalization.

Cambridge: Cambridge University Press.

Ulrike Claudi, and Friederike Hunnemeyer (1991) Grammatical-

ization: A Conceptual Framework. Chicago and London: The University of

Chicago Press.

Hokama, Shuzen (ed.) (1970) Konkokenshu: Kohon to kenkyu [The Konkokenshu:

Redaction and interpretation]. Tokyo: Kadokawa Shoten.

(ed.) (1972) Omorogo jiten: Okinawa no kojisho Konkokenshu

[The Konkokenshu, an old Okinawan dictionary]. Tokyo: Kadokawa Shoten.- and Nobutsuna Saigo (eds.) (1972) Omoro soshi [Book of

omoros]. Nihon shiso taikei 18. Tokyo: Iwanami Shoten.

(ed. & trans.) (2000) Omoro Soshi [The book of omoros] Iwanami

Page 45: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 45

Bunko, yellow series, 142-1,2. Tokyo: Iwanami.

(ed.) (2000) Kdchu Ko-Ryukyu [Ancient Ryukyu, revised and

annotated]. Iwanami Bunko, blue series, 102-1. Tokyo: Iwanami Shoten.

Honda, H. H. (1967) The Manyoshu: A New and Complete Translation. Tokyo:

Hokuseido Press.

Hyojojo (eds.) (1711) Konkdkenshu [The Konkdkenshu], black-and-white fac-

simile edition, 1984.2.15. Afterword by Okinawaken Kyoiku Iinkai [Okinawa

Prefectural Board of Education]. Naha: Okinawaken Kyoiku Iinkai.

Ikemiya, Masaharu (1995) (Ryukyu kogo jiten) Konkdkenshu no kenkyu [A study

of the Konkdkenshu (a dictionary of the archaic Ryukyuan language)]. Nanto

bunka sosho 17. Tokyo: Daiichi Shobo.

Izuyama, Atsuko (2003) Studies on Luchuan Grammar. ELPR Publications Series

A4-024.

Kanazawa, Shozaburo (1923) Nihon bunpo shinron [A new theory of Japanese

grammar]. Tokyo: Waseda Daigaku Shuppanbu.

Kaneda, Akihiro (1998) Gendai Nihongo no naka no kakari musubi: Hachijojima

hogen no rei o chUshin ni [Kakari musubi in modern Japanese, centered on

examples of Hachijojima dialect]. Gengo 27.7: 67-73. Tokyo: Taishukan.

Kinsui, Satoshi (2002) Nihongo bunpo no rekishiteki kenkyu ni okeru riron to

kijutsu [Theory and description in historical studies of Japanese grammar].

Nihongo Bunpd vol. 2 no. 2. also available at: http://www.let.osaka-u.ac.jp/

knsui/kls/kinsui02. htm.

Kokuritsu Kokugo Kenkyujo (eds.) (1963) Okinawago Jiten [A dictionary of the

Okinawan language]. Tokyo: Okurasho.

Li, Charles N. and Sandra A. Thompson (1977) A mechanism for the development

of copula morphemes. In: Charles N. Li (ed.) Mechanism of Syntactic Change,

419-444. Austin: University of Texas Press.

Luo, Cheng (1997) Iconicity or economy? Polysemy between demonstratives,

copulas and contrastive focus markers. CLS 33: 273-286.

Mamiya, Atsushi (1983) Omoro Sdshi no kakari musubi ni tsuite [On the kakari

musubi of the Omoro Soshi]. Okinawa Bunka 61: 6-18.

Matsushita, Daizaburo (1930) Kaisen hyojun Nihon bunpo [Revised grammar of

Page 46: On the Old Japanese Kakari (Focus) Particle koso

46 Leon A. SERAFIM and Rumiko SHINZATO

standard Japanese]. Tokyo: ChUbunkan.

Mithun, Marianne (1995) On the relativity of irreality. In: Joan Bybee, et. al (eds.),

The Evolution of Grammar: Tense, Aspect and Modality in the Languages of

the World, 367-368. Chicago: University of Chicago Press.

NHK (eds.) (1967) Zenkoku hogen shiryo, 5, Chugoku . Shikoku-hen [All-Japan

dialect materials, 5, volume on the Chugoku and Shikoku regions]. Tokyo:

Nippon Hoso Kyokai.

Nohara, Mitsuyoshi (1998) Shinpen Ryukyu hogen joshi no kenkyu [A study of

the particles of the Ryukyuan dialects, revised]. Okinawa: Okinawagaku

Kenkyujo.

Nomura, Takashi (1995) Ka ni yoru kakari musubi shiron [Prolegomena to a

theory of kakari musubi with the particle ka]. Kokugo Kokubun 64.9: 1-27.

(2001) Ya ni yoru kakari musubi no tenkai [The development of kakari musubi with ya]. Kokugo Kokubun 67.1: 1-34.

OGJ. See Kokuritsu Kokugo Kenkyujo (eds.)

Ono, Susumu (1993) Kakari musubi no kenkyu [A study of kakari musubi]. Tokyo:

Iwanami.

Ono, Toru (1978) Nihongo no sogenteki kenkyu [An etymological study of the

Japanese language]. Tokyo: Takayama Honten.

Oshika, Tadahisa (1991) Man'yoshu ni okeru futeigo to futei no gimon [Indefinite

words and indefinite questions in the Man'yoshu]. Kokugogaku 165: 53-66.

Ota, Kaoru (2002) Kakarimusubi and focus structure. In: Noriko Akatsuka and

Susan Strauss (eds.) Japanese/Korean linguistics 10, 293-305. Stanford: CSLI.

Popper, Karl R. (1962) Conjectures and refutations: The growth of scientific knowl-

edge. New York: Basic Books.

Quinn, Charles J., Jr. (1997) On the origins of Japanese sentence particles ka

and zo. In: John Haig and Ho-min Sohn (eds.) Japanese/Korean linguistics 6,

61-89. Stanford: CSLI.

Sakakura, Atsuyoshi (1993) Nihongo hyogen no nagare [The history of Japanese

expression]. Tokyo: Iwanami.

Schaffer, Wolfram (2002) Kakari musubi, noda-construction, and how gram-

maticalization theory meets formal grammar. In: Noriko Akatsuka and Susan

Page 47: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari (Focus) Particle koso 47

Strauss (eds.) Japanese/Korean Linguistics 10, 320-333. Stanford: CSLI.

Schu, R.G. (1983) Kilba equational sentences. Studies in African Linguistics 14:

311-326.

Serafim, Leon A. (2003) When Did Japonic Language Enter the Ryukyus? -A

Critical Comparison of Language, Archeology, and History. In: Alexander

Vovin & Toshiki Osada (eds.) Nihongo keitoron no genzai/Perspectives on

the origins of the Japanese language, Nichibunken Sdsho 31: 465-478. Kyoto:

Kokusai Nihon Bunka KenkyU Senta.

, and Rumiko Shinzato (2000) Reconstructing the Proto-Japonic

kakari musubi, *-ka ... -(a)m-wo. Gengo kenkyu 118:81-118.

Shinzato, Rumiko (1998) Kakari Musubi: Its Functions and Development. In: D.

Silva (ed.) Japanese/Korean linguistics 8, 203-216. Stanford: CSLI.

and Leon A. Serafim (2003) Kakari musubi in comparative

perspective: Old Japanese ka/ya and Okinawan -gal-i. In: Patricia Clancy (ed.)

Japanese/Korean Linguistics 11, 189-202. Stanford: CSLI.

Takahashi, Toshizo (1982) Omoro soshi ni okeru kogaika ni tsuite [On progressive

palatalization in the Omoro soshi]. In Nakasone Seizen Sensei Koki Kinen Ronshu Kankd Iinkai (eds.) Ryukyu no gengo to bunka [The language and

culture of Ryukyu], 219-242. Tokyo: Ronshu Kankd Iinkai.

Thorpe, Maner Lawton (1983) Ryukyuan language history. Unpublished disserta-

tion, University of Southern California.

Uchima, Chokujin (1994) Ryukyu hogen joshi to hydgen no kenkyu [A study of the

particles and expression in Ryukyuan dialects]. Tokyo: Musashino Shoin. Watanabe, Akira (2001) Loss of overt wh-movement in Old Japanese and demise

of 'Kakarimusubi.' Proceedings of the COE International Symposium, 37-57.

Chiba: Kanda University of International Studies.

Whitman, John (1997) Kakarimusubi from a comparative perspective. In: J. Haig

and H. Sohn (eds.) Japanese/Korean Linguistics 6, 161-178. Stanford:CSLI

Wilson, Patricia R. (1980) Ambulas grammar. Ukarumpa, Papua New Guinea:

Summer Institute of Linguistics.

Yamada, Yoshio (1954) Naracho bunpdshi [A history of Nara-Period grammar].

Tokyo: Hdbunkan.

Page 48: On the Old Japanese Kakari (Focus) Particle koso

48 Leon A. SERAFIM and Rurniko SHINZATO

Yamaguchi, Gyoji (1989) Gimon hyogen no suiryogo [Inferentials in interrogative

sentences]. Kokugo to Kokubungaku 66.7: 42-56.

Page 49: On the Old Japanese Kakari (Focus) Particle koso

On the Old Japanese Kakari(Focus)Particle koso 49

古代 日本語 の係 り結 びコソ-そ の起源 と構造-

レオン A.セ ラフィム

(ハワイ大学,マ ノア校)

新里瑠美子

(ジョ「ジ ア工 科大学)

古代沖縄語 のス=siの 係 り結び と古 代 日本語 のコ乙ソ乙=kosoの 係 り結び は,已 然

形 で結 ぶ強調表現 であるところが類似す るが,積 極的 に比較研究 されることは少な

か った.本 稿においては,両 者を比較 し,日 本祖語 におけ る原形を*ko#swo(指

示 詞 の*コ 乙+形 式名詞 の*ソ 甲)と 再構す る.そ して,有 坂第 一.法則 の適用 で,

甲類のswoが,先 行す る乙類の Ωに母 音調和 した結果,古 代 日本語 では,kosoと

なった と仮説 す る.ソ 甲の部 分が 甲類 で,形 式 名詞 であった との見解 は,従 来 の

近 称の コ乙+中 称 のソ乙との語源 と相容れ ないが,そ の裏づ けとして,① 沖 縄最古

の辞 書『 混効験 集』 の知見,② 西 日本方言 に痕 跡を留め る形式名 詞のス・ソ,③

機 能論,文 法化理論の観点からの論 証を挙 げる.更 に,コ 乙ソ乙(沖 縄 ス=si)と カ

(沖縄 ガ)の 係 り結 びを対 照 させ,両 者 の結 びがrealisとirrealis(古 代 日本 語 は

多 くが推量 の助 動詞-(a)m-)に 対 応す る意味を認知論 的に考察 し,ま た指示詞か

ら係助詞 のようなfbcus particleへの移行 は世 界の言語の文法化 のデータにも合致

す る点を指摘す る.

(受理 日 2004年4.月20日 最終原稿受理 日 2005年1月8日)