3 of his poem Shànggǔ zhī shé.(1 / 3)

(8) 囝音蹇, 閩 俗呼 子 為 囝.

Jiǎn yīnjiǎn, mǐn sú hūzǐwéijiǎn

(word) soundjianMincustomcallsonas jian

‘The sound of this character is jiǎn, the Min usually call “son” jiǎn’

The more general case of semantic change from ‘child’ to diminutive morpheme is well attested in other languages of the world, for example, in Jurafsky (1996) and Heine et al.(1993: 38) while the use of diminutives with probable source morphemes in sexspecific ‘son’ is characteristic of Sinitic (for more data, see Huang 1996). The Arte provides the hard evidence for this semantic change into a diminutive suffix, affecting the morpheme ‘son’ in Southern Min (see also Chappell 2000). Given the widespread occurrence of the first type of conceptual shift crosslinguistically, I conclude that while this more semantically specific case may be a shared development in Sinitic languages, it only partially characterizes it typologically.

[...]

4.3COMPLEMENTIZERS

In Taiwanese Southern Min, a complementizer similar in function to English that has grammaticalized out of the verb ‘to say’ kóng 講. Matisoff (1991: 398400) describes this path of grammaticalization as an example of the general category of verbs developing into verb particles in SouthEast Asian languages, represented by Thai, Khmer, and Lahu. Like these three languages, the Southern Min verb ‘say’ is also used at the end of a nonfinal clause and before the intonation break to introduce the complement clause. It is not fully grammaticalized since it may be omitted. Moreover, it forms a kind of verb complex with the preceding matrix verb which must belong to one of the following verb classes: speech act, cognition, or perception, and it directly introduces the embedded clause, as in (11):

(11) Taiwanese Southern Min:

遐個 敵對的 武將共 笑講

Hia ê 〈MC:díduì〉 ê búchiòng kā chhiò kóng,

that CL opposingL general PRETR laugh SAYthat

這是 號作 <J: 猴麵冠者>

che sì hōtsò <J: Sarumen Kanja>.

this be name.as monkey.face youngster

‘Those generals who opposed him mocked him (General Toyotomi) as the one who should be called “monkeyface boy”.’ (Japanese tales 62930) (Note:MC = Mandarin Chinese insert; J = Japanese insert)

In this first stage of grammaticalization, when ‘say’ verbs are used as quotative markers, the lexical meaning is not completely bleached. Examples such as chhio kóng could still be rendered as ‘laughed (at him) saying’ while in the second stage where kóng is used with cognitive verbs such as siūn‘think’, its literal meaning is less plausible: ‘think saying’. The putative path of development is outlined in Chappell (forthcoming e) in addition to other grammaticalized or partially grammaticalized uses of kóng as a metalinguistic marker of explanation; an evidential marker of hearsay; a component of a compound conditional marker; a topic introducer and as a clausefinal marker of assertions and warnings. It has not yet developed a purposive function, which may indicate that certain of its several grammaticalization pathways are relatively ‘young’ (Bernd Heine, p.c.).

There has been only very little study of this phenomenon in typological work on Sinitic languages to date. In Chappell (forthcoming e), I show that this development has proceeded as far as the quotative stage in some Yue and Wu dialects and less far in standard Mandarin. For the Yue dialect of Cantonese, ample evidence can be found of the use of wa6‘to speak’ in conversational and narrative texts where it functions as such a quotative marker with speechact verbs. Note, however, that wa6 does not form a verb complex with the preceding speechact verb: this is clear in that it can be separated from the verb by a noun denoting the direct object:

(12) Cantonese:

讚哩 個 男仔 話……

jaan3 lei5 goh3 laam4jai2 wa6...

praise this CL young.man say...

‘(she) praised this young man saying ...’

Although a verb complex with ‘say’ as V2 is not a possible strategy for introducing complement clauses in standard Beijing Mandarin, or pǔtōnghu (as opposed to such a use for quotations), it is in the regional variety known as Taiwanese Mandarin. It is striking that Taiwanese Mandarin does not choose the cognate verb for kóng, which is jiǎng in Mandarin, to create the new syntactic calque but instead makes use of its functional equivalent, the high frequency verb shuō 說, in the configuration SUBJECT-VERB1- shuō + CLAUSE:

(13) Taiwanese Mandarin:

那 我希望說這 個 願望

nà wǒxīwàng shuō zhègeyuànwàng

CONJ 1sg hope SAYcompthisCLwish

很快 就到 了

hěnkuài jiù dào le

very quickly then arrive PFV

‘So I hope that this wish will be realized very soon.’

(14) Beijing Mandarin:

*我希望說

*wǒxīwàng shuō

1SGhopesayHowever, this does not provide supporting evidence just for the northsouth divide for Sinitic languages: it appears that Sinitic is encircled by language families and language isolates (such as Japanese and Korean) that all possess complementizers which have developed from verbs of saying. This feature has been described in the relevant literature for individual languages belonging to TibetoBurman, TaiKadai, HmongMien, Indie, Dravidian, and Altaic (see Matisoff 1991, Saxena 1988).

Since this semantic change is also crosslinguistically well attested (it occurs widely in various language families of Africa—see Frajzyngier 1996 for Chadic, Amberber 1995 for Amharic, Heine et al. 1991: 21617, 2467, Heine et al. 1993:1908 for a larger sample of languages), it seems that the grammaticalization of kóng into a complementizer in Taiwanese Southern Min is most likely a language internal development. It has simply drawn on its own resources (Dixon 1997) to recreate a syntactic device which was in fact available in Classical and Middle Chinese, as attested in the written register.

Indeed, earlier periods of written Chinese made use of verbs of saying such as yuē 曰 (Classical Chinese) and dào 道 (Medieval Chinese) as quotative markers, although not as fully fledged complementizers (described in Chappell forthcoming e). This means that not only does Sinitic have its own inherited language internal devices upon which to analogize but it also has access to patterns and processes which can be imitated from surrounding unrelated language families.

It seems that this has taken place in recent times for sister languages within Sinitic, the case in point being the calquing of the Taiwanese Southern Min complementizer into Taiwanese Mandarin. This is an unusual development in terms of the direction of metatypy from a less prestigious to a more prestigious language, and note that there are many other examples of Taiwanese Southern Min constructions which have been borrowed into the Taiwanese variety of Mandarin (see Kubler 1985). This probably reflects linguistic creativity in transferring favoured syntactic forms and devices into Mandarin where gaps exist, rather than a negative description in terms of interference from the first language.

Further research on dialect materials would be in order to show irrefutable evidence for the view that the development of a complementizer in Taiwanese Southern Min is a purely independent innovation, triggered however by a combination of factors: a conducive environment in terms of areal typological features and the existence of appropriate languageinternal characteristics.

Unlike the case for negative existential verbs, the existence of a complementizer in Southern Min and some Wu and Yue dialects tallies well with Matisoffs inclusion of Southern Sinitic in the SouthEast Asian linguistic area. The theoretical problem remains, however, of distinguishing between areal diffusion and a putative language universal for the development of complementizers from verbs of saying, given the right typological preconditions.

4.4ADVERSATIVE PASSIVES

Matisoff (1991) points out that verbs of giving typically develop into causatives and benefactives in SouthEast Asian languages. In Southern Sinitic languages, verbs of giving are also used to form the passive construction. For example, most Hakka dialects use the high frequency verb pun44 ‘to give’ as both the passive and the benefactive marker, while Cantonese does the same with bei2 < ‘give’.

A further characteristic feature of passives which unites Sinitic is that the colloquial forms are both adversative and agentful. This appears to be an unusual development for ‘give’ (compare this with data in Heine et al. 1993: 97103). Such a description applies to standard Mandarin as well where only the bèi passive has an agentless form although it has lost its adversative feature in some contexts. Note that the bèi passive belongs to more formal discourse, in contrast to the agentive colloquial passives formed by jiào ‘make’ and ràng ‘let’ (see Chappell 1986).

Norman (1982: 245) observes that these two Northern Chinese passives formed with the causative verbs jio ‘make’ and rng ‘let’ are unique amongst Sinitic languages, as opposed to the use of verbs of giving. He argues that this is not an independent development in Mandarin but rather is due to Manchu superstrate influence on Chinese. In Manchu and other Altaic languages the same structure can be used for both passive and causative meanings. In support of this view, an earlier study by Hashimoto (1987: 46) contrasts standard Mandarin with Mandarin dialects on the periphery of the Northern Chinese zone which continue to use verbs of giving as passive markers. This suggests that ‘give’ verbs as passive markers are an older feature.

The adversative feature appears to be an areal feature as not only do SouthEast Asian languages such as Thai and Vietnamese show this restriction, but also Japanese (see Shibatani 1994). Hence there are different allegiances for each of these features: some evince the northsouth divide in Sinitic (verbs of giving versus causative verbs used as passive exponents), some are relevant to the SouthEast and East Asian area (the adversative feature), while this particular development for ‘give’ is possibly specific to Southern Sinitic within the Asianzone, and is quite rare crosslinguistically (Bernd Heine, p.c.).

10.3思考和練習

1. 請根據以下的例子,分析以下幾種語言在詞素形態上屬於哪種語言類型(孤立語、屈折語、黏著語和複綜語)。

甲:ev ler imde

houseplmyin

‘in my houses’

乙:eatayanaone na yalewa

PAST hitthe child the girl

‘the girl hit the child’

丙:chhatrara monojog diye likhchhilo

studentPLattentionwithwritePROGPST3

‘students are writing attentively.’

丁:wak akyatawitsher ú:n:

PAST 1SG dressmake

‘I made a dress.’

2. 瀏覽《語言結構的世界地圖集》在線版的語音特征庫(https:\/\/wals.info\/feature),觀察中國方言在語音方麵具有哪些共性和差異,列舉幾個你的發現。

3. 根據你對中國方言的理解,有哪些詞素形態和語法方麵的特征可以作為對中國方言分區的參考標準?

4. 語言類型學研究的發展與語言描摹密不可分。結合本章的幾篇文章,討論:類型學研究的發展對語言描摹工作有什麼啟示?一個具有類型學意義的語言描摹應具備哪些要素?

10.4延伸閱讀

Comrie, B. 1989. Language Universals and Linguistic Typology: Syntax and Morphology (Second Edition). Chicago: University of Chicago Press.

Croft, W. 1990. Typology and Universals. New York: Cambridge University Press.

Greenberg, J. (ed.).1966. Universals of Language (Second Edition). Cambridge, Mass.: The MIT Press.第十一章曆史語言學第十一章曆史語言學

11.1導 引

●引言

●曆史比較語言學

●語言譜係

●語言演變的類型

●語言演變的過程

●語言演變的動因

11.1.1引言

曆史語言學主要研究三個方麵的問題:(1) 構擬語言的原始形式;(2) 確定語言的親屬關係;(3) 發現語言演變的規律。曆史語言學研究始於曆史比較法,曆史比較法奠定了曆史語言學的研究方法以及對語言演變的一些基本認知的基礎,因此本章將首先闡釋曆史比較法,再介紹近期關於語言演變的重要理論和熱點問題。

11.1.2曆史比較語言學

曆史比較語言學是曆史語言學的一個重要組成部分,它奠定了曆史語言學發展的根基。曆史比較語言學發展於19世紀的歐洲,其核心是曆史比較法。

要了解曆史比較法,首先要談談曆史比較語言學的一個理論預設。我們知道,意大利語、西班牙語、葡萄牙語和法語等歐洲語言雖為不同的語言,但是在詞彙、句法上卻高度地相似,原因是這些語言都是由通俗拉丁語演變而來的。羅馬帝國的統治曾使拉丁語通行於歐洲,但大約在公元1世紀,口頭拉丁語在不同的地域出現了分化;隨著時間的推移,這些分化日漸累積,不同區域間語言的差別越來越顯著,最終形成了今天歐洲的語言格局。

拉丁語和歐洲語言的發展曆程體現的正是曆史比較語言學的一個理論預設,即一些語言可能具有同一個祖先,它們發源於同一個早期的原始母語(protolanguage);在人口的遷移和其它因素的作用下,這種原始母語在不同的地方會發展出新的特征;久而久之,這些特征累積形成了語言和語言的差別。基於這樣的假定,曆史比較語言學家認為,具有親屬關係的語言(由同一原始母語演變出來的不同語言)很有可能仍保留著原始母語的一些特征,通過比較和分析這些特征便可以確立這些語言的親屬關係——這是曆史比較法的第一步;然後,在第一步的基礎上,根據同譜係語言共有的特征,可以對它們共同的祖先——原始母語進行構擬。曆史比較語言學采用比較分析的方法實現兩個研究目標:建立譜係關係和構擬原始母語。

諸拉丁語言的曆史發展似乎過於顯而易見,有充分的曆史文書見證和記錄了這一過程。對於那些沒有發展出文字記錄的語言,或者發生在史前的語言演化,曆史比較的方法能夠發揮更加重大的作用。

18世紀,當歐洲國家開始向海外擴張,歐洲人開始接觸到遠在東方的印度文明和各種用梵語書寫的經典,有些人發現,這種在遙遠東方的語言竟然和他們熟悉的希臘語和拉丁語有驚人的相似之處。1786年“亞洲學會”上,一名供職於東印度公司的英國人William Jones宣讀論文,指出梵語和眾多歐洲語言的相似之處,提出它們同出一源的猜想;這在當時被許多人視為無稽之談。但正是由於歐洲語言與梵語在此時的邂逅,引發了語言學家的思考和討論,催生出了曆史比較語言學及其係統的研究方法,致使“印歐語係”及世界幾大語係的提出(見圖11.2)。

11.1.2.1主要著作

曆史比較語言學的幾篇重要奠基之作,始於對印歐語係語言的分析。曆史比較語言學發展的第一階段以波普、拉斯克和格裏姆為代表。他們開創了曆史比較法,提出並論證了有關印歐語種共同發源的設想,樹立了一些曆史比較語言學的基本觀念。

德國的葆樸(Franz Bopp)有兩部重要的著作:《論梵語動詞變位係統,與希臘語、拉丁語、波斯語和日耳曼語相比較》(ber das Conjugationssystem der Sanskritsprache in Vergleichung mit jenem der griechischen, lateinischen, persischen und germanischen Sprache,1816)、《比較語法》(Vergleichende Grammatik des Sanskrit, Zend, Griechischen, Lateinischen, Litthauischen, Gothischen und Deutschen,1833),尤其是其中第一本,被認為標誌著曆史比較語言學的元年。在這兩本著作中,通過比較這些語言中具有同樣語法功能的變位——第一本主要集中在動詞的變位,而第二本則擴展到不同的語法類別,如:名詞的變位(性、數、主格、賓格、所有格等)、形容詞的格(比較級、最高級)和數詞,葆樸主張把這些詞語分成“詞根詞綴”兩部分看待,並著重分析詞綴的演變。通過比較分析法,波普不僅構擬了詞綴的原始形式,提出詞綴來源的假設;並且分析了這些變位詞綴的語音變化的規律。波普討論的曆史語音變化規律很大程度上印證了格裏姆提出的“格裏姆定律”(Grimms Law)。

圖11.1四種語言中的與格(dative case)變位圖片來源:Bopp, F., Eastwick, E. B. 1885. A Comparative Grammar of the Sanskrit, Zend, Greek, Latin, Lithuanian, Gothic, German, and Slavonic languages. London, UK: Williams. p. 262。

格裏姆(Jacob Grimm)對曆史比較語言學最大的貢獻是他在《德語語法》(Deutsche Grammatik)(1822)中提出的“格裏姆定律”。受到波普(1816)和拉斯克(1818)拉斯克(Rasmus Rask),丹麥語言學家,主要研究語言為冰島語,著有《古代北方語或冰島語起源研究》。此書成書於1814年,最後於1818年出版。拉斯克與波普不謀而合,在此書中采用了曆史比較法,把冰島語和峨特語、斯拉夫語、立陶宛語、拉丁語和希臘語等作比較,找出了許多詞彙上的對應,並整理出一個比較語法。拉斯克對古音演變的規律有很細密的考察,這點對格裏姆有尤其大的啟發。研究中不同語言的啟發,以及他對日耳曼語語音演變的觀察,格裏姆提出印歐語言的語音演變是係統的音變,並且具有一定的規律。比如在輔音方麵:

·送氣濁塞音變濁塞音或擦音

·濁塞音變清塞音

·清塞音變擦音

波普、拉斯克、格裏姆等人的發現由此在歐洲掀起了一股曆史比較語言學的風潮,詞源研究開始興起,關於“重建”(reconstruction)的研究也開始出現。波特(August Friedrich Pott)的《印度日耳曼係語言領域內的詞源研究》(“Etymologische Forschungen auf dem Gebiete der Indogermanischen Sprachen”,1833)是一個較早期的詞源研究。施萊赫爾(August Schleicher)的《印度日耳曼係語言比較語法綱要》(“Compendium der Vergleichenden Grammatik der Indogermanischen Sprachen”,1861\/62)也是第一次使用“重建”的方法構擬了古印歐語的研究,其構擬的內容包括元音、輔音、詞根、詞幹結構、名詞變格等。

所謂“重建”,就是就所比較的語言材料,用曆史統計的方法為每個形式、每個詞構擬出一個對每種語言來說都適合的“一般曆史性的公分母”,用來代表最原始的形式,並表明各有關的個別語言以後的演變。構擬出來的原始形式(一般前加*表示)的準確性是無法驗證的,它隻是一種基於理論的猜測,很難代表當時的準確發音。

曆史比較語言學第二個階段的發展大概可以從1870年開始算起。在這個階段,比較語言學家開始認識到一切語音的演變都是有規律的,並且沒有例外——如果有不能被解釋的例外,隻是因為還沒有找到它的規律,這就是“維爾納定律”(Verners Law)。維爾納定律的提出為新語法學派(The Neogrammarians)提供了有力的理論依據,進一步推動了曆史比較語言學的發展。

新語法學派中特別值得一提的是法國語言學家梅耶(Antoine Meillet)。他的《曆史語言學中的比較方法》(La Méthode Comparative en Linguistique Historique,1925)一般被認為是曆史比較法最好的書。本章節選了其中幾節,展示曆史比較法基本的操作方法。

在此必須指出,語音形式在曆史比較語言學研究中始終占據著核心的地位。首先,兩個語言間親屬關係的確認必須由同源詞(cognates)是否存在對應的規律的語音形式來驗證語音形式的係統對應通常僅在語言的基本詞彙中確立。一般詞彙通常不作考慮,因為很多是從非親屬語言中借來的。;其次,詞源的構擬也是基於其後代語言的語音的。曆史比較法通常不會在構詞形態和基本語序上使用——前者極其不穩定,而後者更容易受到其它語言的影響而徹底改變。雖然早前波普的分析是基於變位詞綴,但根據波普的推斷,這些詞綴演變自代詞等基本詞彙。

11.1.2.2 漢語古音的構擬

20世紀初,瑞典漢學家高本漢(Bernhard Kalgren)把曆史比較語言學的方法引入傳統的漢語音研究中。高本漢根據《切韻》的音類、現代漢語方言和域外漢字讀音,利用曆史比較法構擬了中古音,後來趙元任、李方桂、羅常培等學者又對高本漢的構擬做出了修改和訂正。高本漢對於漢語中古音的構擬發表於他的博士論文《中國音韻學研究》(“tudes sur la Phonologie Chinoise”,1915)中。本章的第二篇選文總結了迄今為止包括高本漢在內的中古音構擬的一些重要研究。

11.1.3語言譜係

曆史比較語言學的一個巨大貢獻,就是推動了世界語言語係的建立。19世紀的曆史比較法從不同的方麵論證了橫跨歐亞大陸的印歐語係(IndoEuropean)的存在。漢藏語係(SinoTibetan)也早在19世紀就有學者提出。除此以外,世界的幾大語係中,使用人數較多的語係包括:分布於西亞\/北非的亞非語係(AfroAsiatic,舊稱含閃語係),廣布於南太平洋、印度洋、南亞諸島的南島語(Austronesian),分布於非洲西南部的的尼日爾剛果語係(NigerCongo)。位於亞洲大陸上的幾個主要語係除了漢藏語係,還包括: 主要分布於印度中南部的達羅毗荼語係(Dravidian),亞洲南部的侗台語係(TaiKadai)和南亞語係(Austroasiatic),以及亞洲北部的阿爾泰語係(Altaic)。

圖11.2世界語係分布圖 https:\/\/upload.wikimedia.org\/wikipedia\/commons\/e\/ed\/Primary_Human_Language_Families_Map.png.

11.1.3.1譜係樹模型

前麵提到,兩種語言之間的親屬關係可以通過比較法建立。但是一個語係中龐雜的關係應該如何表示呢?施萊赫爾受到達爾文進化論物種演變的啟發,以譜係樹(Stammbaum)的模型,把不同語言分為語支和語族,以表示同一個語係內不同語言的分化和共同進化(圖11.3)。

譜係樹的模型及其反映的語言演變觀對曆史語言學的影響深刻,以至於有學者提出了與譜係樹理論相關的“落葉模型”,試圖來解釋目前還無法厘清關係的漢藏語係的譜係結構(圖11.4)。“落葉模型”的提出者,荷蘭語言學家George van Driem解釋,這種模型並不是反對譜係樹模型,而是反映了現階段語言學家對漢藏語的認識——我們近期才了解了漢藏語大部分語言的全貌;就像看到一棵樹滿地的落葉一樣,還無暇去抬頭觀察樹的枝幹。但或許通過觀察樹幹投射在地上的影子,我們能逐漸看到枝幹的走勢……圖11.3施萊赫爾構建的印歐語譜係樹 圖片來源:http:\/\/lingulist.de\/pyjs\/slides\/img\/schleicher1861.jpg。

圖11.4漢藏語係的“落葉模型”van Driem, G. 2003. TibetoBurman v.s SinoTibetan. In B. Bauer & G.J.Pinault (eds.) Language in Time and Space. Berlin: Mouton de Gruyter. p. 111.

11.1.3.2波浪模型

因為語言演化的實際情況並不是能簡單地用譜係樹結構來解釋的。即使在曆史比較法盛行的時期,人們也已經意識到語言之間可能相互影響:互借詞彙,甚或影響另一個語言的結構。德國的施密特(Johannes Schmidt)另外提出了“波浪模型”,認為原始印歐語已存在方言,這些不同方言的特點像波浪一樣向四周傳播擴散,互相影響,形成了印歐語各語族語支中的某些共同點。

11.1.4語言演變的類型

在這一小節,我們開始跳出曆史比較語言學的框架,進入廣義的曆史語言學研究。現代的、廣義的曆史語言學研究的內容,除了是對早期語言的一種靜態構擬,更多的是一種動態的曆時(diachronic)研究,其關注的是語言演化的過程,以及語言演化中呈現的普遍規律和趨勢。曆史語言學也因此往往需要采用一個跨語言的視角。在這點上,它與類型語言學非常相似,兩者也經常有方法上的交叉。

11.1.4.1語音變化

從新語法學派時期起,語音變化就被認為是規律、係統發生的現象。某個音素的合流、分化、增加、脫落、替換不會孤立地出現在一個或幾個詞裏,而是會普及到語言的整個語音係統中。例如在漢語中,從中古漢語到現代普通話,就有“濁聲母清化”“知、莊、章三組聲母合流”“精、見二組聲母分化”“韻尾合流歸並”“韻頭由繁到簡”,以及平上去入聲調的變化等發生在整個語音係統的語音變化現象。而在印歐語係的構擬中,則能發現演變過程中元音或輔音在不同位置的增加、脫落、替換和音色的改變(延長、縮短、增強、弱化、顎化等等)。這些音變的發生有時與語流音變有關,或者與人發音和聽音的生理基礎有關,有時與語言的省力原則有關;但更多的時候,語音變化是一種隨機發生、無法預測其方向的現象,所以印歐語言才會出現各種各樣的分化。

在語音變化中,有一種特殊的現象,稱為鏈變(chain shift),指的是一個語音係統中一係列相互聯係的音位全部改變其語音現實的複雜音係變化。鏈變的基本假設是,一個語音中的所有音位構成了一個平衡的語音係統,隻要其中一部分發生了改變,就會影響整個係統發生改變。音位的移動意味著原語音係統出現空槽,因而產生了填補空槽的需求,這時會由另一個音位的移動來填補,這就是所謂的鏈變現象。格裏姆定律所指向的印歐語曆史上發生的係統音變本質上就是一種鏈變。

11.1.4.2詞彙變化

詞彙變化在這裏有兩個方麵的含義。一種詞彙變化指的是詞義的變化,也就是既有詞彙在曆史的演變中詞義發生了改變。詞義變化的類型可分為詞義擴大、詞義縮小、詞義轉移,以及詞義褒化和貶化。比如古漢語(文言文)中的“行”、“走”以前的詞義是“走”和“跑”,這種現象屬於詞義轉移。而在對某些特殊人群的指稱中,我們經常會看到一個中性含義的詞經曆詞義貶化的現象,比如中文的“小姐”(原是對年輕女性的稱呼,後指從事色情服務行業的女子)、“同誌” (原是社會主義體製下對他人的敬稱,後指同性戀者),英語的mistress (原義為“女主人”,後指“情婦”)、madam (原是對成年女性的尊稱,後在某些情形下指“老鴇”)。

另一種詞彙變化則指新詞彙的發明及舊詞彙的淘汰。伴隨著社會的興替,新事物的湧現和舊觀念的式微造就了這個意義上的詞彙變化。計算機時代的到來產生了許多新事物,也創造了許多新詞來描述這些現象:電腦\/計算機、芯片、內存、鍵盤、博客、雲計算、彈幕等等,不一而足。還有一類新詞,比如網絡流行語言“也是醉了”、“十動然拒”則是一種更類似於俚語、詼諧幽默的社會方言。

11.1.4.3形態句法變化

還有一類語言變化是語法規則的變化。在具有屈折形態的語言中,可能體現為詞素形態的變化;在孤立語中,可能體現為新句型(詞語的組合序列)的出現;或者是在任何語言中的語序的變化。我們把這一類統稱為形態句法變化。形態句法變化往往伴隨著詞義的變化以及語音的變化。中古漢語六朝之後發生的語序變化,以及由此誘發的動補結構的出現、體標記的產生,可視為形態句法變化的範疇。

語法化(grammaticalization)是形態句法變化中的一個常見的現象,指的是實詞演變成語言中具有語法功能成分的一種語言變化。漢語的很多虛詞就是從實詞語法化而來的,比如體標記“在”、“過”、“著”、“了”都是從動詞變化而來的。在漢語這樣的孤立語中,語法化之後的成分仍然可視為一個詞;但在屈折語言中,實詞經曆語法化後可能會失去詞的地位,而變成附著於詞根的詞綴,甚至消失。語法化的成分通常讀音會變弱;在具有詞素形態的語言中,其長度會縮短。比如,英語中的have表完成時的時候,可以縮寫為 ve, 在發音上也從一個音節縮為一個輔音[v]。這種語義從實到虛,形態從有到無的變化——也就是語言變化的“單向性”(unidirectionality),曾被認為是語法化的普遍真理,但隨著越來越多反例的出現,這種觀念已經受到了撼動。當然,語法化的單向性是一個較為普遍的現象,因此語義的虛實和形態特征在某些情形下能幫助我們判斷比較語言發展的階段。

如果僅用虛實之分來描述語法化成分的詞義變化未免過於簡單。實際上,語法化成分的語義變化就如同我們的語言係統一樣,具有一定的隨意性。同樣語義的實詞,在不同語言中可能會演變成不同功能的語法成分;或者說,不同語言中具有類似功能的語法成分,是從不同的實詞演變而來的。

但語法化成分的語義變化同時也具有一定的規律性。我們大概會以為,漢字的“把”字句非常特殊(特別是“把”從“握”、“持”的動詞義演變為處置構式的一部分),但實際上,在尼日爾剛果語係中的語言Twi和Nupe中,都存在具有“拿”義的動詞演變成使動結構一部分的例子。上文提到的英文用表“擁有”的動詞have表示事情的發生和完成,這種用法在中國的某些方言裏也能看到,比如受到閩南話影響的台灣普通話“你有沒有看到他?”又比如,用個人的客觀能力來表示事情發生的可能性,用身體部位來表示方位,用方位詞來表示時間等等,都是雖不絕對但是極為普遍的語法化現象更多的例子可參閱 Heine, B. & Kuteva, Tania, 2002. World Lexicon of Grammaticalization, New York, N.Y., USA: Cambridge University Press。 。當這種規律的語義變化同時出現在沒有親屬關係、地理上也不毗鄰的語言中時,唯一的解釋便是人類共有的認知能力和認知傾向造就了這種相似的、有規律的語言變化。

語法化發生的一個重要前提,就是人類喻象思維(metaphorical thinking)的能力。喻象思維,在這裏並不限於作為文學創作修辭手法“比喻”,而是更廣泛地指向人類跨越概念空間,以一個概念替代另一個概念的一種思維能力。時間和空間屬於不同的概念空間,主觀意願和客觀事實也屬於不同的概念空間;但人類運用喻象思維,就可以把兩個不同概念空間的概念關聯起來——人們為跨空間概念A和B建立的關聯性,使他們可以利用原來專門用於描述A的語言來描述B。而由於人類的生活體驗(人周遭的時間、空間,人的生理結構,人的行為,人與人之間的關係)大部分是類似的,人類利用喻象思維建立的關聯也是相似的——或者至少是可以被別人理解的。人類的喻象思維是語言創造性的重要來源。

近幾年在語法化理論的基礎上,有學者結合構式語法理論提出了語言變化的“構式化”(constructionalization)理論。這種理論認為,詞義變化以及形態句法的變化都是在“構式”(construction)的框架內進行的。一個構式的組成成分組成了一個整體,決定了構式整體的語義(語用)以及其中各個成分的語法關係和含義。

11.1.5語言演變的過程

如果從語言的係統性角度來考慮,語言變化的開端總是始於某處的,比如某個具體的搭配、句子或場景。由於語言的規律和係統性特征,一些語言使用者可能會對幾個偶然發生的語言變化做出分析,發現它們變化的規則,逐漸推導應用到其它的符合規則的場景中,最終擴及到整個語言係統。類推(analogy)是語言演變發生的一個重要機製。無論是語音變化、詞義變化還是形態句法變化,類推都決定了語言變化在整個語言係統的擴散。

如果從語言的社會性這個角度來考慮,語言的變化也是始於某些使用者,然後逐漸擴散到社會的絕大部分使用者。一個新的語言現象之所以能被理解,是因為說話者之間有語用推理(pragmatic inference)的能力。即使是從沒接觸過的一個新的語言現象,母語者仍能在語境中猜測出它的意思。我們在兒童時期學習語言,甚至在網絡時代學習新的流行語言,依靠的都是語用推理的能力。

因此,語言演變的過程是一個既有突變,又有漸變的過程。突變指的是說話者的語言創新以及聽者理解的瞬間;漸變指的是這種變化從孤立、個別的現象到係統、普遍的存在的過程。總的來說,語言演變從開始走向完成,是一個使用頻率增加、使用場景增加和使用人數增多的過程。

11.1.6語言演變的動因

語言演變的動因大致可分為內因和外因。內因是指引起語言演變的語言係統內部因素。以上提到的印歐語係語音係統的鏈變、漢語語序變化而產生的語法變化,包括語言使用者的類推能力、語用推理能力,我們都可以視為內部因素。

在討論語言演變的外部因素時,通常會考慮到語言使用的社會生態,比如語言接觸、語言特征作為社會標記,以及來自官方的語言規劃的影響。地理上毗鄰語言的相互接觸勢必會造成一定程度的語言交融,例如,中國各異的方言在一定程度上是曆史上漢族人南下遷徙與當地語言(楚國、百越國)交融的結果。但是,接觸的兩個語言哪一方對另一方的影響更大,則可能與經濟實力和文化勢力有關。中國在建國之後推廣普通話,許多方言都開始向普通話靠攏,這與普通話帶來的各種就業、教育、醫療等社會資源不無關係。另一方麵,我們也可以把這類語言變化視為一種社會標記,也就是說,不同群體使用“普通化”的方言或本土味的方言,有可能是在試圖呈現一種對本地文化的身份認同。來自官方的語言規劃對語言變化的作用則更加不可忽視。建國以來,我國對文字書寫、發音、甚至語法使用都做出了規範,遏製了一些可能正在進行中的隨機語言變化,但這並不意味著語言變化一定按照官方的意誌和方向進行。來自上層和底層的語言變化,最終由一代又一代的語言使用者決定其是去是留。

11.2選讀

●梅耶(1925)《曆史語言學中的比較方法》選讀

●Jerry Norman(1988)“The Methodology of Middle Chinese Reconstruction”選讀

●梅祖麟(1991)《從漢代的“動、殺”、“動、死”來看動補結構的發展》選讀

●CavalliSforza et al. (1988)“Reconstruction of Human Evolution”選讀梅耶(1925)《曆史語言學中的比較方法》選讀 梅耶著,岑麒祥譯,《曆史語言學中的比較方法》,《國外語言學論文選譯》,語文出版社1992年版。

◆ 作者簡介

安東尼·梅耶圖片來源:法國維基百科https:\/\/fr.wikipedia.org\/wiki\/Antoine_Meillet。(Antoine Meillet, 1866—1936),法國曆史語言學家。梅耶師從索緒爾,1891年任巴黎高級研究學院教授,1906年任法蘭西學院印歐語比較語法教授。他的研究領域涉及希臘語、拉丁語、日耳曼語、凱爾特語、波羅地語、吐火羅語等諸多領域,編寫過斯拉夫語教材。1924年他與科恩同編《世界語言》一書,一生出版專著24種、論文540篇,其中《印歐係語言比較研究導論》、《曆史語言學中的比較方法》對曆史比較語言學產生了重大影響。梅耶構擬出原始印歐語的語音語法係統,但認為它不可能完全重建。

◆ 正文節選

一、 比較方法的定義

進行比較工作有兩種不同的方式:一種是從比較中揭示普遍的規律,一種是從比較中找出曆史的情況。這兩種類型的比較都是正當的,又是完全不同的。

世界上各處地方都有一些關於動物的故事:動物與人的相似之點是這樣明顯,所以把一些人類固有的行為加到動物身上,來表明一些不容易直接使人了解的事情,是很自然的。我們可以比較這些故事,確定它們的形式、性質和應用範圍,從而建立一個關於動物的故事的一般理論。我們在其中所發現的相同之點,是由共同的人類心理所造成的;其中的不同之點,是由文明的類型和程度的差別所造成的。用這樣的方法,我們可以知道人類的共同性質,但是一點也不能知道人類的曆史。

假如我們和一位法國青年學者杜美西爾(Dumézil)先生一同去研究那些關於“長生水”的印度歐羅巴神話,所得的結果就完全不同了。這種以為有一種水可以使人長生的想法實在太自然了,所以不是什麼民族的特點。可是,在每個印歐語民族中,都或多或少的可以遇到這種在一個巨桶裏釀造“長生水”的傳說,這個傳說還加上假未婚妻的故事或神仙和惡魔相鬥的說法,可見這裏麵是包含著許多特殊的事情的。這些特殊的事情之間並沒有什麼聯係,所以它們的結合決不是偶然的。

假如語言所表達的意思和那些用以表示這意思的聲音之間有一種或鬆或緊的自然聯係,就是說,假如語言符號可以撇開傳統,單用它的音值本身可以使人想到它所表達的概念,那末,語言學家所能采用的就隻有這種一般的比較方法,任何語言的曆史也就都不會有了。

但是事實上語言的符號是任意規定的:它隻有靠傳統的力量才能有意義。如果在法語裏,大家用un, une來表示“一”,用deux來表示二”,……那並不是因為un,une和deux等詞本身和“一”、“二”等意思有什麼關係,而隻是因為說法語的人教給學法語的人的習慣是這樣的。

隻是因為語言符號具有這種完全任意的性質,所以才能有現在所要研究的這種曆史比較方法。

比方法語、意語、西班牙語的數詞是這樣的:

法語意語西班牙語

一un, une uno, unauno, una

二deuxdue dos

三troistre tres

四quatrequattrocuatro

五cingcinquecinco

六six sei seis

七septsette siete

八huitotto ocho

九neuf nuovenueve

十dix diecidies

二十vingt venti veinte

三十 trentetrenta treinta

四十quarantequarantacuarenta

一百centcento ciento

這裏麵的相符之點決不是出於偶然的;其所以不是偶然的,是因為從一種語言與另一種語言的那些差異中,可以找出一些確定的對應規律來。比方“huit”, “otto”, “ocho”“八”等詞間的差異初看起來很大,但是這個差異並不是偶然的,因為這樣的對應還有很多,如法語的nuit“夜” ,意語的notte, 西班牙語的noche;或法語的cuit“煮”,意語的cotto;以至法語的lait“乳”,意語的latte,西班牙語的leche;法語的fait“事實”,意語的fatto,西語的hecho等等。這些顯而易見的相符之點固然給我們指出了應當遵循的途徑,但是可以利用的卻隻是那些語音對應的規律。

顯而易見的相似之點給我們指出了很好的途徑的地方,常常有些特殊的細節予以證實。比方un和une有陰陽性之分,而其他的數詞卻都沒有,這一點是很重要的。

因此,我們可以假定,法、意、西班牙三種語言的數詞出於同一個來曆。在這種情形之下,經驗告訴我們,隻有兩個可能的來曆:一、這三組數詞同出一源;二、這三組中有兩組的形式是由另外一組借來的。在這個例子裏,第二種假設是不可能的,因為在這三種語言中,我們不能用任何一種語言的形式來解釋另一種語言的形式。法語的huit不能出於意語的otto或西班牙語的ocho,意語的otto不能出於法語的huit或西班牙語的ocho,西班牙語的ocho 也不能出於法語的huit或意語的otto。由此可以證明法、意、西班牙這三種語言的數詞有一個共同的出發點,而這出發點不是法語,不是意語,也不是西班牙語。

在以上所舉的例子中,相符之點又多又完備,對應的規律又極容易認識,連門外漢也可以馬上看得出來,用不著語言學家才看得到它的證明價值。假如我們考察一些在時間上和空間上相隔較遠的語言,如梵語、古雅典希臘語、拉丁語和古典亞爾明尼亞語等,相符之點就沒有這樣顯著,對應的規律也就比較難於確定。

梵語古雅典希臘語

一ekah, ék,ékamhēs, mia, hen

拉丁語亞爾明尼亞語

ūnus, ūna, ūnummi

以上三個形式中,第一個是陽性的,第二個是陰性的,第三個是中性的;亞爾明尼亞語沒有這種語法上的性的分別。

梵語希臘語拉丁語亞爾明尼亞語

二d(u)vdyoduoerku

“二”這個數詞是有性的分別的,這裏隻舉出陽性的形式;“三”和“四”也是如此。

梵語希臘語拉丁語亞爾明尼亞語

三 tryahtrēstrēserek‘

四catvrahtéttares quattuorcˇork‘

五pnˉcapéntequinquehing

六stheks sexvec

七saptheptseptemewt‘n

八st 在輔音前的吠陀梵語形式,在元音前變成stv。óktō octōut‘n

九nvaennéanoueminn

十 dadékadecemtasn

除了“一”這個數詞暫且不論,可以說希臘語、拉丁語以至梵語間的對應大部分是顯而易見的,可是亞爾明尼亞語和其他語言間的對應就沒有這樣明顯了。

但是我們把亞爾明尼亞語的各種事實詳細考察一下,就可以看到這些相符之點的證明價值。

比方亞爾明尼亞語的erku“二”和拉丁語的duo等是不相似的;不過,一些其他的對應表明erk可以與其他語言的*dw相當,例如:希臘語有一個詞根dwi表示“怕”的意思,亞爾明尼亞語也有一個erki(erkiw “怕”);希臘語有一個很古老的形容詞dwrón,表示“長久”,亞爾明尼亞語也有一個erkar“長”(參看中譯本第26頁。)從這個相符之點可以得出一條一般的對應規律:古代的*dw在亞爾明尼亞語變成了erk。

作為複合詞的第一個成分,希臘語有一個dwi,亞爾明尼亞語有一個erki。所以有這組特殊的相符之點是不容許我們存任何懷疑的(參看中譯本第91頁)。

亞爾明尼亞語erek‘ “三”和c'ork‘ “四”等詞的形式和希臘語的trēs,téttares相差很遠;但是它們至少有一部分可以用同類的對應來解釋。並且有一種特殊的細節,就是在梵語和希臘語中,“三”和“四”都有正規的格變形式,而自“五”以後的數詞卻是不變的;在亞爾明尼亞語中,“三”和“四”也有正規的格變形式,特別是語尾k‘是亞爾明尼亞語多數體格(nominatif)的記號,而這個記號在其他各數詞裏就找不著了。

由此看來,這些數詞形式的相符之點,在梵語、希臘語、拉丁語和亞爾明尼亞語中,初看起來雖沒有像在法語、意語和西班牙語中那麼明顯,其實是同樣可靠的。

這些相符之點不能用各語言間互相借用來解釋,可以假設它們有一個共同的來源。但是我們還要用一種係統化的方式來解釋,這就是比較曆史語言學的目的。

二、 共同語言

[……]

三、 所用的證明

我們要確定一種古代共同語的存在,必須在所比較的語言中盡量找出這種古代語言的那些被保存下來的特性。因此應當研究這種語言的各個成分的作用是怎樣的,因為它們並不是以同等的程度保存下來的,也不是以同樣的方式保存下來的。任何語言都包含有三個不同的係統,彼此之間有一定的聯係,但是大體上這三個係統可以各自獨立發生變化。這三個不同的係統就是:形態、語音和詞彙。

[……]

可以作為確定“共同語”和後代語間的連續性的證據的,隻有那些表現形態的特殊規則。比方大家都知道兩個名詞的領屬關係,可以用一個前置的虛詞來表示,如法語的de,可以用一個後置的成分來表示,如英語的s。但是這個虛詞用de的形式還是用s的形式,這件事卻是有特殊性的;因為這種關係用任何音素來表示都是可以的,隻要傳統沒有決定用別種音素來表示它。所以我們可以說:用放在補語前麵的de來表示這種關係,是一種法國土語的特性;用放在補語後麵的s來表示這種關係,卻是一種英國土語的特性。

這一類的特殊事實常常是很穩固的。發音盡管起變化,詞彙盡管改變,而這些特性卻是不變的。比方在現在法國北部的那些土語中,本地的詞的形式起了變化來適合法語的形式,詞彙有了革新,人們一般地有依照共同法語的習慣來說話的趨勢。最後留存的就隻有一些形態學上的地方特點;比方陽性與陰性的分別,共同法語說il dit(他說),elle dit(她說),他們卻說i dit, a dit。這一類的特殊事實,是從小就學到的,不知不覺地成了習慣,其他一切盡管起變化,這些特殊事實卻可以不變。

因此,一種形態繁雜的語言,包含著很多的特殊事實,它的親屬關係自然比較容易得到證明;反過來,一種形態簡單的語言,隻有一些一般的規則,如詞的次序,要找出有力的證據就很不容易了。我們差不多用不著去證明一種語言是印歐係的:隻要碰到一種大家還不認識的印歐係語言,如最近發現的吐火羅語(tokharien)或喜低特語(hittite).我們略加考釋就可以看出它的印歐語的特性。反過來,遠東的那些語言,如漢語和越南語,就差不多沒有一點形態上的特點,所以語言學家想從形態的特點上找出一些與漢語或越南語的各種土語有親屬關係的語言,就無所憑借,而想根據漢語、西藏語等後代語言構擬出一種“共同語”,是要遇到一些幾乎無法克服的阻力的。

兩種語言之間相符的事實愈特殊,這個相符之點的證明力量就愈大。所以例外的形式是最適於用來確定一種“共同語”的形式。

[……]

在語音方麵,我們也要把一個古代的係統和新的係統互相對照,這些係統間的差別可以是非常之大的。

但是係統之間的差別並不是亂來的。要認識兩種同源的語言之間的有規則的對應,雖然常常是不可能的,然而共同語和從它演變出來的語言之間的對應卻有一定的規則。我們可以把它們列成許多確定的公式。這就是所謂“語音規律”。

日耳曼語的f,p,x(以後變為h),以及在某些條件之下的,,γ和印歐語的p,t,k相對應;日耳曼語的p,t,k和印歐語的b,d,g相對應;日耳曼語的b,d,g(在二元音之間則為,,γ)和印歐語的bh,dh,gh相對應。日耳曼語的這套有規則的對應係統,就是所謂“輔音轉化”,或格林姆(Grimm)規律(在格林姆之前不久,大部分已經為拉斯克Rask所發現,不過到了格林姆現多譯為“格裏姆”,後文不再一一說明。——編者注才把它定為規律)。這種原始“共同語”和各個後代語言之間語音對應的規律性,表現出發音的變化並不是孤立的影響某一個詞或某一形式,而是影響到整個語音係統的。

根據方法上的原則,可見對應的規則可以在原始語和它的每一種後代語之間擬定,而不能在各種出於同一共同語的後代語之間擬定。我們可以擬定印歐語的首音p在希臘語和梵語為p,在歌特語現多譯為“哥特語”,後文不再一一說明。——編者注為f,如希臘語的patēˇr,梵語的pit1a和歌特語的fadar(父親),印歐語的首音kw後麵跟著一個o的時候在希臘語為p,在梵語為k,在歌特語為hw,如希臘語的póteros, 梵語的katarh,歌特語的hwapar(二者之一)。實際上觀察到的這兩個對應:

希臘語p=梵語p=歌特語f

希臘語p=梵語k=歌特語hw

隻有在和那個為了解釋它們而擬定的印歐語的形式對照之下,才可以理解。

我們比較同族語言時所注意的並不是形式上的相似,而是對應的規律。上文第5麵已經說過,亞爾明尼亞語的數詞“二”是erku,和古印歐語的*dwō(或 *duwō)對應。這個對應初看來好像很奇怪。但是亞爾明尼亞語的erk和歐印語的dw相對應,是遵守一個一般的規則的。因為此外我們知道還有兩個別的例子。印歐語曾有一個詞根*dwei(怕),在荷馬的希臘語裏是廣泛的以一些動詞形式如*dedwoa(寫作deidō) ,dedwoike(寫作deidoike)(我怕)或一些名詞形式如*dweos(寫作deos)(怕)來表現的;亞爾明尼亞語與它對應的就有erkiwl(怕),erkeay(我曾經怕)。印歐語曾有一個形容詞*dwro(長),在希臘語裏還顯著的保存著,而亞爾明尼亞語也有一個形容詞erkar(長)。所以這個規律有三個明顯的相近的例子作為基礎。如果我們想到已知的以*dw,*duw為首音的印歐語的詞數目非常少,就可以看出這三個符合的例子同時並存有證明的價值了。

這個對應還可以有解釋。舌尖輔音後麵跟一個w而造成的音組,在亞爾明尼亞語裏變成了舌根音: tw變為k‘如k‘o(你的),與希臘語的twe(古雅典語變為se)之類的形式相對應。這個清音k的來源是這樣的:在亞爾明尼亞語中和在日耳曼語中一樣,從前曾經有過一個由濁塞音變為清塞音的變化,如d變為t,g變為k。前麵的那個r,就是這個音組中首音的古代濁輔音性質的一個痕跡 詳情參看格拉蒙 M. Grammont在《巴黎語言學會專刊》第20種,第252頁的解釋。,隨後這個r又引出一個首音e,erku。可是如果那個詞中間本來有一個r ,如krkin(兩倍),這個首音r就不能產生了(krkin是由古代的*kirkin或*kurkin變來的)。所以在亞爾明尼亞語的erk與*dw這個非常奇特的對應中,一切都是由亞爾明尼亞語的結構所造成的。任何對應,假如不能得到這樣的解釋,就是可疑的。

[……]

詞彙雖然常常是不穩固的,但是在比較各種語言時,最先引人注意的卻是那些詞彙上的相符,常常甚至於隻是處理詞彙。其所以這樣做,或者是由於對所考察的那些語言知道得太少,隻有從詞彙上來找證據;或者是由於所研究的語言語法非常簡單,如在遠東的各種語言;或者是由於存留下來的形態建立得太晚,是在所擬定的共同時期之後。因此我們要特別細心考察怎樣才能證明詞彙之間的相符。

上麵已經說過,有效的語源上的符合決不是根據一些語言形式間的相似而確定的,而是根據一些對應的規律:我們之所以能夠拿亞爾明尼亞語的erku來和俄語的dva相比較,並不是因為這兩個形式相似:在語音方麵,它們是毫無共同之點的;而是因為那些對應的規則容許我們這樣比較,印歐語的ō在斯拉夫語變成了a, 在亞爾明尼亞語變成了u,印歐語的duw,在俄語變成了dv,在亞爾明尼亞語變成了erk。

對應中的不規則的情形,假如不能用某個詞的特殊情況來解釋,那就或者是由別的語言借來的,或者是由於語源不合。比方古拉丁語的ca,在法語變成了cha,che,chè,例如campum變成了champ(田),carrum變成了char(車),caballum變成了cheval(馬),carum變成了cher(親愛的) ,等等;至於有個camp(營地)和campum相當,那是因為這個詞並不屬於法語的舊傳統,實際上是由意大利語借來的,並且事實上還知道它是在什麼時候、為什麼“借來”的。可以與日耳曼語的b相對應的是拉丁語的f,如拉丁語的flōs,flōris和德語的Blume(花);所以德語的Feuer和法語的feu(火)毫無關係:想一想各個羅馬族語言中那些與法語的feu相對應的詞,如意大利語的fuoco,西班牙語的fuego,就可以知道feu與Feuer的相似是不相幹的。所以我們進行比較時隻能用一些精密的對應公式——並且要小心避開那些借用的成分。

[……]

如果所考察的是一些真正由“共同語”直接變來的詞,那麼就應當構擬出這種“共同語”的一個各方麵都確定的詞,而不要滿足於比較一些詞根上的細小成分。錯誤的危險性既然很大,所以我們必須用些精密的證據來斷定觀察到的相符不是偶然的。

大家所同意的第一點(如果不是在原則上同意,至少是在事實上同意)就是:一個詞源,隻有確實符合語音對應規則,或者雖有歧異之點,也能用一些嚴格確定的特殊情況來解釋,這樣的一條詞源研究才是有效的。

不消說,相對應的語音成分數目愈大,相符之點的偶然性的危險就愈小。

[……]

意義方麵的相符也應當與語音方麵的相符(根據語音的對應規則)同樣正確,同樣精密。這並不是說,意義相符的程度應當比語音相符的程度大;而隻是說,如果有意義上的分歧,就不應該隻用一些含胡的、一般的“可能性”來解釋,而必須用一些特殊的情況來解釋。法語的ouaille是由拉丁語的ouicula(羊)變來的,ouaille這個詞在現代法語裏僅指禮拜堂裏的教士的忠實信徒而言,這條語源決不會因此變得可疑。

我們知道, 一所基督教禮拜堂的忠實信徒是常常被人比作這所禮拜堂的牧師所牧養的羊群的,由於這個原因,這個意義上的比較便得到說明了。此外,ouaille 這個詞在法國某些土語裏也還有作“羊羔”解的,這一點更可以證明這個詞源的正確性。

[……]

我們要構擬一種原始“共同語”時,必須注意到我們對於某個詞所擁有的證據有多少。兩種語言之間的相符,如果不是全部的,就會保不定不是偶然的。但是如果相符之點擴充到三種、四種或五種很不相同的語言,偶然的可能性就大大減少了。雖然古波斯語的rdiy(因為)和斯拉夫語的radi(因為)在他種語言裏找不出來,我們還是毫不遲疑地拿它們來比較,因為這兩個詞在形式上、意義上和用法的細節上都是完全一致的。 除了這一類的例子以外,假如我們隻在兩種印歐係的語言裏找到一組相符的詞,而在別的語言裏卻找不到使它們失掉這個語詞的特殊情況,這種相符就是可疑的了。

無論那種語言,必定要全體的密切相符之點,確定了所比較的詞之所以相似不可能是出於偶然的,才能說詞源得到了證明。

[……]

無論在形態方麵語音方麵還是詞彙方麵,我們決不能忽視一條原則,就是:所有的比較,隻有合乎嚴格的規則,才是有效的。語言學家愈是粗心大意,他的比較愈是任意,他的證明也就愈不可靠。Jerry Norman(1988) “The Methodology of

Middle Chinese Reconstruction”選讀 Norman, J., 1988. Chinese, Cambridge; New York, NY: Cambridge University Press.◆ 作者簡介

羅傑瑞 圖片來源:https:\/\/upload.wikimedia.org\/wikipedia\/en\/4\/4a\/Jerry_Lee_Norman_%281936%E2%80%932012%29.jpeg。(Jerry Lee Norman, 1936—2012),美國漢語言學家,華盛頓大學亞洲語言文學係教授。他早年師從趙元任,後與橋本萬太郎、餘靄芹等提出普林斯頓假說,主張擺脫《切韻》的框架,先構擬漢語各方言的原始語,在其基礎上再構擬原始漢語。在諸多漢語方言中,羅傑瑞對閩語的研究最深。其博士論文題為《閩方言的特征描述》。

本文出自羅傑瑞1988年出版的漢語通論性著作《漢語概說》,這本書也是英語世界漢語語言學教程最通用的教材。

◆正文節選

2.4The methodology of Middle Chinese reconstruction

In his Compendium of Phonetics in Ancient and Archaic Chinese (1954) Kalgren gives a succinct description of the methodology used in reconstructing Middle Chinese. Karlgren considered Middle Chinese to be the language reflected in the categories of the Qièyùn as they have come down to us in its various redactions. The Qièyùn categories in most cases are interpreted in light of the rhyme tables; even though Karlgren admitted that the Qièyùn zhǐzhǎngtú was based on a later stage of Chinese, he felt that it was still useful in interpreting the Qièyùn, since it was much closer to it in time than the modem dialects. Second in importance to the rhyme tables were the socalled Sinoxenic dialects (a term created by Samuel Martin to refer to the systems of pronouncing Chinese characters in Japan, Korea and Vietnam). In all three of these countries Chinese has been studied and used extensively for many centuries, and each country has its own distinctive manner of reading Chinese texts aloud. Karlgren, while recognizing that these foreign dialects were “corrupt” in some respects, considered that they were of paramount importance in reconstructing the language of the Qièyùn. Third in importance in Karlgrens view were the native Chinese dialects.

The entire process of reconstructing Middle Chinese was of course vastly complicated, but some simple examples will suffice to illustrate the principles which were employed. A comparative study of the Gungyùn fǎnqiè formulas reveals that there was a distinct initial consonant in the language represented by eight fǎnqiè upper characters pronounced (in Modem Chinese) tā, tuō, tǔ, tù, tōng, tiān, tái, tāng; this initial corresponds to the rhyme table initial tòu, and tòu is one of the shétóu sounds. Each of the shétóu sounds corresponds to a distinctive set of fǎnqiè upper characters in the Gungyùn. The only way phonetic substance can be given to these categories is by comparing them to actual pronunciations in modern Chinese dialects and in the Sinoxenic dialects of Japan, Korean and Vietnam. Table 2.2 gives the readings of several common characters from each of the shétóu initials in several Chinese and Sinoxenic dialects. A glance at the table shows that a large majority of the forms are either dental or alveolar stops and nasals. The initial ní may be safely reconstructed as a nasal on the basis of the forms given. The initials duān and tòu are voiceless dental (or alveolar) stops in all the dialects; with the exception of Kanon, which has merged the two series. The remaining dialects have an unaspirated stop for duān and an aspirated stop for tòu, it is reasonable to suppose that Middle Chinese possessed the same sort of contrast. The only dialect which clearly distinguishes the initial dìng from duān and tòu is Sūzhōu in which dìng corresponds to a voiced stop contrasting with the voiceless correspondences for duān and tòu. Provisionally at least, the initial dìng can be considered some kind of

Table 2.2Dialectal reflexes of the Shétóu initials

PekingSūzhōuXimèn GungzhōuKanonSino

VietnameseSino

Koreanduān端多tuo1tu1to1to1tada1ta刀tau1t1to1tou1todao1to短tuant3tuantyntandoantntòu透他t‘a1t‘1t‘a1t‘a1tatha1t‘a天t‘ien1t‘ix1t‘ien1t‘in1tenthien1ch’ǒn鐵t‘iet‘i7t‘iet7t‘it7tetsuthiet7ch‘ǒldìng定弟ti5di6ti6tai6teide6che頭t‘ou2dY2t‘2t‘au2tōdu2tu豆tou5dY6t6tau6tōdao6tuní泥內nei5nE6lui6noi6dainoi6nae年nien2nir2lian2nin2nennien2yǒn農nu2no2lo2nu2nōno2nongvoiced dental stop in Middle Chinese. Karlgren argued that dìng and the other quánzhuó initials were voiced aspirates, whereas the present consensus is that at the time of the Qièyùn they were voiced unaspirated stops. Once one set of the rhyme table initials has been reconstructed, the meaning of the terms quánqīng, cìqīng, quánzhuó and cìzhuó becomes clear, and this information can then be applied to other sets of initials. The above example illustrates in a very simple way how the Qièyùn and rhyme table categories, which viewed in isolation were just so many abstract formulas, were given phonetic substance by comparing them with corresponding words in the Modern Chinese dialects and Sinoxenic reading systems. Table 2.3 shows the inventory of Middle Chinese initials as reconstructed by Karlgren and modified by F. K. Li (1971).Table 2.3Middle Chinese initials

BilabialspphbmDentalstthdnlSibilantststshdzszRetroflexestthdRetroflex sibilantststshdzszPalatal sibilantsts′ts′hdz′nz′s′z′VelarskkhgngxGlottals()j(zero)The reconstruction of the finals was carried out in similar fashion: the děngyùn categories which were applied to the finals were compared with actual dialect forms, and reconstructed values were postulated. In the case of the finals, however, the rhyme tables do not actually distinguish all the different and distinct rhymes of the Qièyùn, so that even after all the rhyme table distinctions are reconstituted, there still remain a considerable number of Qièyùn categories which must be dealt with if all the Middle Chinese rhymes are to be distinguished. As indicated above, the most difficult of the rhyme table categories applied to thefinals is the notion of děng or ‘division’. In Table 2.4 the actual pronunciation of a representative sample of words from table 25 of the Yùnjìng is shown. (The examples for divisions Ⅰ, Ⅱ and Ⅲ are píng tone; a qù tone word is used to exemplify the fourth division, since it is a more common word than the corresponding píng tone form.) Each of the words is in a different division; they all share the same initial and offglide, so the significance of děng must be sought in the remainder of the syllable—in either the medial or the main vowel, or perhaps in both simultaneously.Table 2.4The four divisions in the dialects

DivisionⅠⅡⅢⅣdialect高交驕叫PekingkautiautiautiauSūzhōukktitiXiménkokaukiaukiauGungzhōukoukaaukiukiuKanonkaukaukeukeuSinoVietnamesecaogiaokiêukiêuSinoKoreankokyokyokyoUsing the forms in Table 2.4 as examples, we can now outline the main lines of Karlgrens thinking concerning the reconstruction of the four divisions: (1)The offglide in all the forms is u. (2) The main vowels in divisions Ⅰ and Ⅱ must be some sort of low vowel like a. (3) The four divisions differ not only in the presence or absence of a medial i, but also in having several different main vowels in several of the dialects (Ximén, Gungzhōu, SinoVietnamese); hence the divisions differed both by medial and main vowel. (4) Division Ⅰ is clearly the least palatalized; a low back  [] is postulated as the main vowel. (5) Division Ⅱ causes palatalization of the preceding velar in some dialects but not in others; this would make sense only if it had some sort of front vowel without an intervening medial i (which, as we will see below, is the distinguishing characteristic of divisions Ⅱ and Ⅳ). A low front vowel a [a] is reconstructed as the main vowel of this division. (6) Divisions Ⅲ and Ⅳ clearly both had a palatal medial of some sort; and, judging from Kanon and Sino Vietnamese, the nucleus was some kind of mid vowel. (7) The most difficult question to answer is how the third and fourth divisions differed from one another; in the present sample all the dialects have merged them. To resolve this problem, Karlgren had recourse to SinoKorean where, in a few cases (not illustrated in Table 2.4), fourth division words have a medial y where the corresponding thirddivision words lack this medial: jin ‘to see’, a fourth division word, is kyǒn [kin] in SinoKorean, where jiàn ‘to construct’, a third division word, is kǒn [kn]. On the basis of cases like this, Karlgren concluded that the fourth division had a longer and more vocalic medial i, as opposed to a shorter, more consonantal i in the third division. The stronger and more vocalic i survives in SinoKorean (in some cases at least), whereas the weaker and more consonantal thirddivision medial i drops out. Finally, since third and fourthdivision finals are different rhymes in the Qièyùn, they must be reconstructed with different main vowels in order to preserve their rhyme distinction. Karlgren proposed that the fourth division had a more strongly palatal e [e] as its main vowel, which matched its strongly palatal medial i; by.similar reasoning, then, the third division had a somewhat less palatal mainvowel to match its more weakly palatal medial; Karlgren reconstructed this vowel as  []. As a result of this analysis we arrive at the following Middle Chinese forms:

DivisionⅠⅡⅢⅣ

Reconstructionku kaukiukieu

A solution to the problem of the nature of the děng was of course an immense help in reconstructing the Qièyùn, finals, but it did not solve all the problems by any means. Although one could easily classify all the Qièyùn rhymes as to division, in some cases the Qièyùn zhǐzhǎngtú treated distinct Qièyùn rhymes as if they were identical; such rhymes had doubtless merged at an early date, and it was difficult to find evidence for distinguishing them. Generally Karlgren foundhis solution to this problem in marginal distinctions maintained in one or more of the native or foreign dialects which he relied upon. In two cases he was unable to find any evidence for distinguishing two important Qièyùn rhymes; in one case he reconstructed two separate rhymes as i; F. K. Li (1971) has proposed that these two rhymes be distinguished as i versus , on the basis of their separate origins in Old Chinese. Likewise, Karlgren has reconstructed two distinct Qièyùn rhymes both as ai; Li distinguishes them as ai and a.

Karlgrens Middle Chinese vowel system is highly complex; it consists of sixteen distinct vowels and four medials, two of which can also function as offglides; his medials are i, i, u and w; the difference between i and i is discussed above; u, in contrast to w, is used when the kāi and hé variants of a final are put into different Qièyùn (or Guǎngyùn) rhymes. Since his transcriptional system is a bit strange to most modern linguists (it was based on J. A. Lundells Swedish dialect alphabet), we will examine all his vowels and medials, describing the phonetic interpretation of each and listing the finals in which each occurs. Bracketed symbols are taken from the International Phonetic Alphabet.

(1) i. A high front unrounded vowel, [i]: (j)i, (j)wi. The first of these two finals actually represents two contrasting rhymes, as we pointed out above. Karlgren writes these finals with the j after labials, velars and gutturals; otherwise he generally omits the j.

(2) e. An upper mid, front unrounded vowel, [e]: iei, iwei, ieu, iem, iep, ien, iwen, iet, iwet, ieng, iweng, iek, iwek.

(3) ě. A shorter variety of vowel (2): iěu, iěn, iwěn, iět, iěu, iwět.

(4) e. Karlgren writes this [e] when it functions as a glide and the preceding or following i is to be taken as the main vowel of the final: (j)ei, (j)wei,(j)ie, (j)wie.

(5) . A lower mid, front unrounded vowel, []: ii, iwi, iu, im, ip, in, iwn, it, iwt, ing, iwng, ik.

(6) . A front unrounded vowel somewhat lower than vowel (5), []: in, it, ng, wng, k, wk.

(7) a. A low front unrounded vowel, [a]: a, wa,ia, ai, a, wai, wa, au, am, ap, an, wan, at, wat, iang, iwang, iak, iwak.

(8) . A shorter variety of vowel (7): i, wi, m, p, n, wn, t, wt.

(9) . A mid, central unrounded vowell, []: im, ip, n, un, ut,in, iun, it, iut, ng, wng, ing.

(10) . A short (nonsyllabic) variety of vowel (9):u, iu—in these finals u is to be taken as the main vowel.

(11) .A lower mid to low unrounded central vowel, []: ii, iwi, im, iwm,ip, iwp, in, iwn, it, iwt, ng, wng, ing, iwng, k, wk, ik.

(12) u. A high back rounded vowel, [u]: iu, ung, iung, uk, iuk.

(13) o. An upper mid, back rounded vowel, [o]: uo, iwo, uong, iwong, uok,iwok.

(14) . A lower mid, back rounded vowel, []: ng, k.

(15) . A low back, unrounded vowel, []: , u, i, iw, i, ui, u, m, p, n,un, t, ut, ng, wng, k, wk.

(16) . A shorter variety of vowel (15): i, ui, m, p.

The four medials may be described as follows:

(1) i.A short, consonantal palatal onglide, [j]. F. K. Li uses the IPA symbol given here for this medial, a practice which we shall also follow hereafter.

(2) i. A palatal glide, longer and more vocalic than medial (1). It also occurs as an offglide.

(3) w. A short back rounded medial, [w].

(4) u. A back rounded medial, longer and more vocalic than medial (3).

Karlgrens reconstruction of the Middle Chinese finals has been criticized by a number of scholars. Y. R. Chao (1941) examined his system with a view to determining its contrastive elements; in the process he proposed several revisions in he system, some of which were subsequently accepted by Karlgren. Samuel Martin (1953) made a thoroughgoing phonemic analysis of Karlgrens reconstruction, reducing the number of vocalic contrasts to six. Martin treated Karlgrens reconstruction, along with Chaos revisions of it, as a valid phonetic reconstruction of Middle Chinese; his resulting phonemicization, while interesting from the point of view of phonological analysis, did little to advance our knowledge of Middle Chinese itself, Lǐ Róng (1952) and Dǒng Tónghé (1954) subjected Karlgrens reconstruction to a more searching examination, producing in the process fairly drastic revisions of it. Tōdō Akiyasu (1957), taking into account the work of the abovementioned scholars as well as that of several Japanese predecessors, proposed an even more drastically revised version of Middle Chinese. In 1962 Edwin Pulleyblank, as a preliminary step toward reexamining the probblem of Old Chinese, proposed his own revised reconstruction of Middle Chinese (Pulleyblank 1962a; 1962b). Although all the aforementioned revisions differ from Karlgrens original reconstruction in many important ways, both in interpretation and in substance, from a methodological point of view these revisions depart little (if at all) from Karlgrens views. The differences come about because of a refinement of technique: a closer examination of the sources reveals more distinctions than Karlgren posited; more and better information relevant to reconstructing some of the more difficult contrasts has come to light; the Chinese transcriptions of Buddhist names and terms and Chinese words transcribed in alphabetic scripts (Tibetan, Brahmi, Uygur) have also thrown some light on problems of reconstruction. Such material can be employed, however, without questioning Karlgrens basic methodology. All the reconstructions and revisions of Middle Chinese published up until the present can and should be viewed as products of a single methodological tradition going back to the original work of Bernhard Karlgren and his predecessors.

Pulleyblank (1970) and Miller (1975) have both raised serious questions concerning this traditional methodology. Pulleyblank for the first time rejects the direct relevance of the rhyme table categories to the reconstruction of the Qièyùn language; this is because the earliest rhyme tables are based on another dialect, two centuries later in time and based on different geographic standards. In Pulleyblanks view this later form of Chinese should be first reconstructed independently, and only then be used along with other data to reconstruct the categories of the Qièyùn language. Millers objections to the methodology outlined here are much more radical; he describes it as a “highly eclectic, pickandchoose system” that brings together elements of the traditional historical comparative method with purely intuitive and, at times, arbitrary processes. He finds traditional Chinese linguistic terminology rife with terms chosen for their exemplary function, and even tainted in many cases by philosophical or mystical speculation. As a result, Miller sees little value in most of the reconstructions produced before 1975. While I do not subscribe to this criticism in all its details, I think that Millers criticism can be accounted useful if it leads to a sober reexamination of the whole question of historical reconstruction in Chinese. It indeed seems that Karlgrens sapproach to this topic has pretty much run its course; pursued further, it can only lead to an unending process of juggling and rejuggling of the same old elements, without any really new insights into the historical process. If Chinese historical linguistics is to be rescued from scholasticism, a thoroughgoing reevaluation of basic assumptions about methodology is essential. Although it is well beyond the scope of a book of this type to carry out such a reevaluation, I will offer a few preliminary thoughts on the subject. The Qièyùn itself must be viewed as the primary source for Middle Chinese, the stable core to which other bodies of information are referred; despite disclaimers to the contrary, Karlgrens reconstruction of “Ancient” Chinese is in its essence a reconstruction of the rhyme table categories. The Qièyùn is the basic datum, and the rhyme tables are interpretations of this datum based on later and geographically disparate dialects. We are under no compulsion to accept such interpretations; they are no more than one other type of evidence, to be judged along with other equally valuable types such as modern dialect forms, ancient transcriptions and Sinoxenic readings. The value attributed them because of their early date is offset by the fact that their meaning is anything but clear, and consequently any interpretation of them is of necessity speculative. The native Chinese dialects should take precedence over the Sinoxenic materials: the latter are loanwords taken into a foreign medium, and subject to the internal historical processes of that medium. The Sinoxenic material needs to be studied much more thoroughly before being used to interpret the earlier stages of Chinese. The Chinese dialects are the organic, autochthonous descendants of Middle Chinese, and clearly should be the primary data on which any reconstruction of earlier stages of the language is based. More account needs to be taken of the rich stock of Chinese loanwords in neighboring languages such as Vietnamese, the various Tai languages, MiaoYao and perhaps others as well. Ancient transcriptional evidence, Buddhist as well as other types, needs to be studied more systematically. Finally there should be a more conscious and rigorous approach to methodology; we should know precisely what we are doing at every stage of the reconstructive process.