Studies in Honor of Jerry Norman Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects Zev Handel University of Washington Jerry Norman is one of the world’s most widely acknowledged and respected experts on the Min dialects. Yet his crowning achievement in the field, his reconstruction of Proto-Min phonology, has not received widespread acceptance. Many scholars have argued that Norman’s reliance on the comparative method is misguided, and that in researching the history of peculiar features in Chinese dialects emphasis must be placed on uncovering the lexical layers resulting from dialect contact. Yet the opposition between a comparative-reconstructive approach and a lexical layering approach, which appears to lie at the heart of the disagreement, is not sufficient to account for the different conclusions of these scholars. I argue that a major reason lies instead in fundamental differences of opinion about data collection and interpretation, rather than in differences of methodology and analysis. Because these differences in underlying assumptions have for the most part not been explicitly addressed in the academic literature, it has not been possible to reconcile the two competing storylines about the history of the Min dialects. This study focuses in particular on Norman’s proposal for the reconstruction of a series of Proto-Min “softened” stop initials and on the data that appears to support or refute the proposal. Refutations by scholars like Hirata Shōji, Lǐ Rúlóng, and Wáng Fútáng appear to simply present alternative explanations for the same phenomena that Norman seeks to explain. In fact, however, the data sets on which these and other scholars’ work is based overlap only partially with the supporting data presented by Norman. It is only through a careful examination of how these data sets differ that the basic differences in assumptions, and the way they shape methodology, can be brought out into the open. By doing so, I hope to take a step toward providing a common framework that will allow competing hypotheses to stop “talking past each other” and instead to contribute to the development of a comprehensive natural history of Chinese dialects. Key words: Northern Min, softened initials, Tone 9, Proto-Min, historical reconstruction, dialect layers, dialect history 19 Zev Handel 1. Introduction Jerry Norman is one of the world’s most widely acknowledged and respected experts on the Min (Mǐn 閩) dialects. Yet his crowning achievement in the ield, the reconstruction of Proto-Min phonology, has not received widespread acceptance. Many scholars have argued that Norman’s reliance on the comparative method is misguided, and that in researching the history of peculiar features in Chinese dialects emphasis must be placed instead on excavating the lexical layers that have resulted from dialect contact. Yet the opposition between a comparative-reconstructive approach and a lexical layering approach, which appears to lie at the heart of the disagreement, is not suficient to account for the different conclusions of these scholars. I argue that a major reason lies instead in fundamental differences of opinion about data collection and interpretation, rather than in differences of methodology and analysis. Because these differences in underlying assumptions have for the most part not been explicitly addressed in the academic literature, it has not been possible to reconcile the two competing storylines about the history of the Min dialects. This study focuses in particular on Norman’s proposal for the reconstruction of a series of Proto-Min “softened” stop initials and on the data that appears to support or refute the proposal. Refutations by scholars like Hirata Shōji, Lǐ Rúlóng, and Wáng Fútáng appear to simply present alternative explanations for the same phenomena that Norman seeks to explain.１ In fact, however, the data sets on which these and other scholars’ work is based overlap only partially with the supporting data presented by Norman. It is only through a careful examination of how these data sets differ that the basic differences in assumptions, and the way they shape methodology, can be brought out into the open. By doing so, I hope to take a step toward providing a common framework that will allow competing hypotheses to stop “talking past each other” and instead to contribute to the development of a comprehensive natural history of Chinese dialects. 2. Background In a series of articles published over the last four decades, Jerry Norman (1973, 1974, 1981, 1986a, 1986b, 1996, 2000) has proposed and refined a reconstruction of Proto-Min, the hypothetical ancestor of the modern Min dialects, based on the application of the comparative method of historical reconstruction. His reconstruction １ 20 These explanations are by no means uniform. These and other scholars’ explanations include the layering of literary readings, of borrowings from neighboring dialects, of sound changes proceeding by lexical diffusion, and of sound changes caused by internal morphological processes, among others. Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects of the Proto-Min initial and tone systems are of particular interest because they deviate notably from what can be termed the Qièyùn System (QYS) pattern. The QYS is a reconstructed sound system based on the patterns of phonological structure and correspondence relected in medieval rhyme books (such as Qièyùn 切韻 and Guǎngyùn 廣韻) and rhyme tables (such as Yùnjìng 韻鏡 and Qīyīnlüè 七音略). ２ The QYS initial system is characterized by a three-way contrast in stop and affricate initials: voiceless unaspirated (p, t, k, etc.), voiceless aspirated (pʰ, tʰ, kʰ, etc.) and voiced (b, d, g, etc.). The system has four tones, termed píng 平, shǎng 上, qù 去, and rù 入 (the last coincident with syllables having stop codas). Broadly speaking, the initial and tone systems of nearly all modern Chinese dialects can be correlated with the QYS. By this I mean that a series of conditioned sound changes can be proposed that derive the modern pronunciations from the reconstructed sound system. In contrast to the three-series, four-tone system of QYS, Norman’s reconstructed Proto-Min has a six-way contrast of initial consonants, reconstructed as follows (using labials as cover symbols for all places of articulation): *p *ph *-p *b *bh *-b The first row of voiceless initials conditions one set of tonal developments in Min dialects, and corresponds (in cognate morphemes) to voiceless initials in the QYS such as p, pʰ. The second row of voiced initials conditions a second set of tonal developments, and corresponds to voiced initials in the QYS such as b.３ From the perspective of comparative Chinese dialectology, this reconstruction has two remarkable and surprising features. The irst is the distinction between a voiceless aspirate and a voiced aspirate (e.g. *ph vs. *bh), to account for the presence throughout modern Min dialects of aspirated initials in both upper- and lower-register tones. The second is the presence of what Norman termed “softened” (ruòhuà 弱化) initials (e.g. *-p and *-b), marked with a preceding hyphen. They are reconstructed purely on the basis of relexes in Northern Min and a handful of closely related dialects, accounting simultaneously for the presence in these dialects of voiced initials in both upper- and lower-register tones, and for correlated tone splits. ２ ３ QYS is commonly referred to as Middle Chinese. The proponents of the term QYS prefer it to Middle Chinese because they wish to stress that the reconstructed phonological system is not that of a spoken language, but rather a collection of character readings. For example, see Norman and Coblin 1995:580. These two sets of tones are traditionally referred to as “upper register” and “lower register”, respectively. See below for more detail. 21 Zev Handel The modern voiced initials that descend from these Proto-Min softened initials are illustrated below by relexes in Jiànyáng 建陽 and Shíbēi 石陂４. The Jiànyáng relexes are typical of most Northern Min dialects. Proto-Min Jiànyáng Shíbēi (duānmǔ 端母) *-t lo3 du3 袋 dài ‘bag’ d- (dìngmǔ 定母) *-d lui6 do6 早 zǎo ‘early’ ts- (jīngmǔ 精母) *-ts lau3 字 zì ‘character’ dz- (cóngmǔ 從母) *-dz li6 dz3 dzi6 *-p ye9 ɦye9 Example word QYS initial 賭 dǔ ‘gamble’ t- 飛 fēi ‘ly (v)’ f- (fēimǔ 非母) 倍 bèi ‘-fold’ b- (bìngmǔ 並母) *-b wui5 bo5 飢 jī ‘hungry’ k- (jiànmǔ 見母) *-k ue9 gye9 跪 guì ‘kneel’ g- (qúnmǔ 群母) *-g y6 gy6 Note that the modern relexes w-, l-, zero, etc. are also sometimes referred to as “softened” initials. Norman did not specify the precise phonetic nature of the Proto-Min softened initials, but given their modern-day reflexes, he suggested that they may have been clusters. The hyphen in Norman’s notation can be taken as suggesting the presence of an unspeciied voiced element. In Norman 1986a, he drew on comparative evidence to suggest that at least some of the softened initials may have been prenasalized. Norman’s six-series, four-tone reconstruction can be recast as a three-series, eighttone reconstruction without any loss of explanatory power, as follows: Eight-tone reconstruction: *p Four-tone reconstruction: *p *b *ph *b *ph *-p *bh *-b *1 *2 *3 *4 *5 *6 *7 *8 *A *B *C *D It should be stressed that in the eight-tone reconstruction, the Proto-Min voiced initials (*b, *d, *g, etc.) do not correspond directly to voiced initials of QYS. Moreover, some of the Proto-Min voiceless initials correspond to QYS voiced initials. In other words, the typological similarity of the two initial systems does not translate in a straightforward way into regular sound correspondences. Norman’s reconstruction struck many in the field as typologically bizarre. A number of scholars objected to it, and published spirited refutations or alternative ４ 22 Data is from Norman 1996 and Norman 2000. 石陂 is sometimes romanized Shípō. Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects explanations. ５ Noting that the evidence for the softened initials was confined to Northern Min dialects, these scholars pointed to contact inluence from neighboring Wu (Wú 吳) and/or Gan (Gàn 贛) dialects as the likely origin of the correspondence patterns that led Norman to reconstruct these proto-initials, or pointed to the inluence of a nonChinese substrate. They argued that the correspondence patterns did not reflect an earlier six-series system of initials, but was an artifact of multiple lexical layers, layers relecting distinct dialect sources. I have argued (Handel 2003) that an important additional piece of evidence from the Northern Min dialects can be invoked in support of Norman’s reconstruction. The tonal values of the tone categories associated with the softened initials show a consistent pattern. Across all tones and all dialects, these pitch values are consistently lower than the corresponding tones associated with the non-softened initials. On the basis of universal features of articulatory phonetics, this tone value evidence strongly supports the reconstruction of voiced murmured initials as the origin of these tone values. Moreover, the overall pattern cannot be explained by the particular Wu and Gan dialectal pronunciations that have been pointed to as the origin of the odd Northern Min sound correspondences. Only a three-series, eight-tone reconstruction similar to Norman’s proposal (with the voiced series likely articulated with ‘aspiration’, i.e. breathiness or murmur) can provide a satisfactory account. I also argued (Handel 2009) that while Norman’s reconstruction may look typologically bizarre from the perspective of QYS and most modern Chinese dialects, it is not dissimilar to recent proposals for reconstructed Old Chinese, which have more manners of articulation that the QYS, including prenasalized obstruents. If one carefully reads Norman’s papers and those of his detractors, one is struck immediately by the very different methodologies underlying their approaches and conclusions. In many cases these methodologies are not made overt, and so academic disagreements that are in part founded on very different underlying assumptions are instead manifested in arguments that “talk past each other”. The dispute is sometimes cast as a difference between an approach based on the primacy of the comparative method (Norman) and an approach based on the primacy of dialect mixing and the resulting accumulation of lexical layers, including literary and colloquial layers (Norman’s detractors). Sometimes implicit, sometimes explicit, in the latter approach is the claim that the complexity of Chinese dialect history renders the comparative method essentially inapplicable in the Chinese linguistic context. For example, Lǐ Rúlóng (1985:148-149) says: ５ See for example Yue-Hashimoto 1976, Hirata 1988, Lǐ Rúlóng 1985, Wáng Fútáng 1999. 23 Zev Handel Is it the case that there must be a common ancestor—a “Proto-Min”—for a dialect group with such a long history, broad geographic distribution, and internal diversity? … If we say that all internal variation developed out of a uniied “Proto-Min”, then this “Proto-Min” could be nothing but an abstract, theoretical notion, difficult to conceive of as an actually existing entity. Therefore, when it comes to the internal diversity of Min, we believe that it is more appropriate and realistic to understand it in terms of the correspondence relations with Old and Middle Chinese and in terms of dialect layering, rather than reconstructing a synchronic system of “Proto-Min”.６ Yet I would argue that this apparent fundamental difference is not what really lies at the core of the dispute. Norman does not dispute the reality of lexical layering and historical convergence in the development of the Chinese dialects.７ And those who have argued against his Proto-Min reconstruction do not deny the theoretical validity of applying the comparative method to the oldest lexical layers of related languages— assuming that those layers can be reliably identiied.８ Rather, it is the presentation of different data sets that distinguishes the two approaches. A careful reading of Norman’s papers and those written in response reveals a great disparity in the data presented in support of each argument. In some cases, sets of morphemes—and the sound correspondences that they relect—adduced by one side are simply absent on the other side, and vice versa. In other cases, the same morphemes are found on both sides of the argument, but the pronunciations attested for them in the various modern dialects—and thus the sound correspondences which those morphemes support—are different. And, indeed, the data presented by Norman support the notion that the comparative method is applicable, while that presented by his detractors cast the applicability of the comparative method in doubt. Yet while each side brings up speciic evidence to serve as counter-examples to the arguments of the other side, the responses do not directly address the data that supported the original arguments. In other words, the claims and counter-claims do not simply relect different underlying methodologies, but are founded on divergent sets of raw data. This is explicitly apparent in Hirata’s ６ ７ ８ 24 像閩方言這樣歷史長、分佈廣、分歧大的方言，是否一定有一個共同的來源─「原始閩語」？…如果 說，所有的內部分歧都是從統一的「原始閩語」演化出來的，這個「原始閩語」恐怕只能是一個抽象的 理論上的概念，而很難理解為一種現實的原形。因此，我們認為，對於閩方言的內部差異，從它與上古 漢語、中古漢語的對應關係，從歷史層次的構成去理解，比起為它退出一個「原始閩語」的共時的體 系，要更切合實際些。 See for example Norman 1986b: 38. The study of lexical layers in Chinese dialects has been a very fertile area of research in recent years. For one example, see the collection of articles in Ting (ed.) 2007. In a review of that volume, Wang Feng (2009) points out the importance of both vertical transmission (i.e. inheritance from an ancestor language) and horizontal transmission (i.e. borrowing due to language contact) in the formation of a dialect features. Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects (1988:12) refutation, which acknowledges Norman’s claim that he is applying the comparative method only to the oldest dialect layers, but disputes the data on which Norman relies. Norman 1986 says: “The oldest layer of Northern Min perhaps preserves a set of initials distinct from the initials of the Qièyùn-series rhyme books.” (p. 38). Assuming that Norman’s reconstruction can stand, we would have to completely rethink the nature of the Min dialects, or revise the initial system of the pre-Qièyùn stage [of Chinese]. I believe that Norman’s reasoning process is faulty, and below will investigate the ancient Min system of initials based on actual language data.９ The purpose of this short study is to try to bridge the gap between the two sides by bringing the discrepancies in both methodological approach and underlying data out into the open. Is there a sense in which both sides are “right”? If not, can the counterevidence raised by one side be directly explained, rather than sidestepped? It is my belief that the ield of Chinese historical dialectology cannot be advanced unless and until a mutual understanding about different methodological approaches to data, and the signiicance of the conclusions that they lead to, is arrived at. 3. Data differences A simple example will serve to illustrate the differences in data that are found in the literature. According to Norman (1974, 1986b, 1996), Northern Min dialects show very speciic and regular correlations between initial and tone. Consider those Northern Min morphemes whose cognates have initial consonant d- (dìngmǔ 定母) and píng tone in the QYS. In most modern Chinese dialects, these words have developed into the lower register relex of the píng tone (conditioned by the voiced quality of the initial), called yángpíng 陽平, and the initial has devoiced to t- or tʰ-, depending on the development rule of the speciic dialect. In some dialect groups (notably Wu and Xiāng 湘), the voiced quality of the QYS initial is preserved in the modern pronunciation.１０ １１ For example: ９ 羅杰瑞1986說：「閩北方言最古老的層次，可能保留有一套和《切韻》系統韻書不同的聲母」（38 頁）。假如羅杰瑞氏的構擬能夠成立，我們不得不重新考慮閩語的性質，或者修改前《切韻》階段的聲 母系統。筆者懷疑羅氏的推論過程有問題，下面根據語言事實檢查古閩語聲母系統。 １０ For simplicity, I will speak of “historical developments” from QYS to modern non-Min dialects. In fact, the modern dialects are not directly descended from QYS, which cannot be considered a language in the conventional sense. The QYS can, however, serve as a reasonable proxy for a set of phonological features that must have been present in the common ancestor or ancestors of most of the modern dialects, and it is in this role that correspondence patters are conveniently cast as historical developments. １１ In many Northern Wu dialects these initials are not actually voiced, but consist of a voiceless stop followed by breathy release, which may persist through the articulation of the following vowel. 25 Zev Handel Sūzhōu Shuāngfēng Nánchāng Méixiàn Guǎngzhōu Běijīng 北京 蘇州 雙峰 南昌 梅縣 廣州 (Mandarin) (Wu) (Xiang) (Gan) (Hakka) (Yue) ‘same’ (同) ‘tube’ (筒) ‘sugar’ (糖) ‘copper’ (銅) tʰuŋ (tóng) tʰuŋ (tǒng) tʰuŋ (tóng) tʰuŋ (tóng) doŋ doŋ doŋ doŋ dan tʰuŋ tʰuŋ dan tʰuŋ tʰuŋ dan tʰuŋ tʰuŋ dan tʰuŋ tʰuŋ tʰUŋ tʰUŋ tʰUŋ tʰUŋ All four of the above words have QYS voiced initial d-.１２ In Northern Min, three distinct sets of initial correspondences correlate with this single QYS initial. These sets, along with Norman’s Proto-Min reconstruction, are as follows: d1) d2) d3) Zhènqián 鎮前 ttht- Jiàn’ōu 建甌 ttht- Jiànyáng 建陽 thl- Wǔfū 五夫 thl- Shíbēi 石陂 tthd- Proto-Min *d*dh*-d- These three correspondence patterns—which I’ve labeled d1, d2, and d3 respectively—are illustrated by the following three cognate sets, which are cognate to Mandarin tǒng 筒 ‘tube’, táng 糖 ‘sugar’, and tóng 銅 ‘copper’, respectively. d1) d2) d3) ‘tube’ ‘sugar’ ‘copper’ Zhènqián toŋ2 thauŋ2 toŋ9 Jiàn’ōu toŋ5 thoŋ5 toŋ3 Jiànyáng toŋ2 hoŋ2 loŋ9 Wǔfū tuŋ2 hɔuŋ2 luŋ9 Shíbēi toŋ5 thɔŋ5 doŋ9 These three sets of initials participate in two distinct patterns of tonal correspondence. For words with the QYS píng tone, the two patterns are: Zhènqián 2 9 A1) A2) Jiàn’ōu 5 3 Jiànyáng 2 9 Wǔfū 2 9 Shíbēi 5 9 The tonal notation is based on a conventional numbering of an idealized tworegister relex of the QYS four-tone system, as follows: 1 yīnpíng 陰平 3 yīnshǎng 陰上 5 yīnqù 陰去 7 yīnrù 陰入 2 yángpíng 陽平 4 yángshǎng 陽上 6 yángqù 陽去 8 yángrù 陽入 １２ They are also all píng tone words. From the perspective of the QYS, the Mandarin tonal reflex of ‘tube’ 筒 is irregular; we would expect second-tone tóng rather than third-tone tǒng. In fact, ‘same’ 同, ‘tube’ 筒, and ‘copper’ 銅 are all homophonous in QYS, as well as in most modern Chinese dialects. 26 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects The odd-numbered or “upper-register” (yīn) tones are those associated with QYS voiceless initials; the even-numbered or “lower-register” (yáng) tones are those associated with QYS voiced obstruents. Where tone categories in Northern Min dialects mostly correspond to these QYS categories, they are assigned the equivalent number. The number 9 labels a tone category that does not correlate to the QYS system. For the purposes of this example, however, the numerical labels of the tones may be thought of as arbitrary; what is significant is the fact that there is a distinct pattern of tonal correspondence (labeled A2) that correlates with a distinct pattern of initial correspondences (labeled d3). (This same A2 pattern also occurs with all the equivalent initial correspondences at other places of articulation.) Norman’s reconstruction of the three Proto-Min initial types *d, *dh, *-d is predicated as much on the regularity of the tonal correspondence as it is on the regularity of the initial correspondence.１３ Indeed, in some Northern Min dialects, such as Jiàn’ōu, the tone difference alone relects the distinction between Norman’s protoinitials *d and *-d; both initials have Jiàn’ōu relex t-. According to Norman’s reconstruction, a Jiànyáng syllable of the shape loŋ² should not exist as a relex of a Proto-Min morpheme whose QYS cognate has initial *d- and píng tone. The initial l- could only derive from Proto-Min *-d; that initial should in turn condition tonal relex 9. Put another way, toŋ2, hoŋ2 and loŋ9 should be the only possible Jiànyáng forms with rhyme -oŋ corresponding to a QYS syllable with initial dand píng tone, as seen in the words for ‘tube’, ‘sugar’, and ‘copper’ above. Yet Hirata (1988:16) lists four morphemes in Jiànyáng pronounced loŋ2. Their very existence would seem to cast doubt on Norman’s entire reconstruction. They are cognate to the standard Chinese morphemes conventionally written 同 ‘same’, 桐 ‘camphor’, 童 ‘youth, son’, and 瞳 ‘pupil (of eye)’, all pronounced tóng in modern standard Mandarin. None of these four morphemes is found in any of Norman’s published articles on his Proto-Min reconstruction; nor, so far as I can tell, does Norman attempt to provide an explanation for their pronunciation in any of his publications that appeared after Hirata 1988. Most surprising of all, Hirata’s source for these Jiànyáng pronunciations is Norman’s own 1971 monograph on Jiànyáng. Hirata not only argues that the pronunciations of these Jiànyáng words and many others refute the sound correspondences on which Norman’s Proto-Min reconstruction are based, he also goes on to argue that these Northern Min syllables can be explained １３ Analogous correlations are found not just with the particular initials and tones given in these examples, but throughout the phonological system. The significance of the overall pattern of these correlations, both phonological and phonetic, is reiterated in Handel 2003 and 2009. 27 Zev Handel by the inluence of Wu dialects (1988:22). What are we to make of this? One might easily criticize Norman for turning a blind eye to evidence that contradicts his hypothesis, especially evidence that he himself has published; one might similarly criticize Hirata for failing to accept the validity of the comparative method and for failing to recognize the signiicance of the correspondence patterns uncovered by Norman. 4. Methodological differences Upon deeper analysis, however, both criticisms are unfounded. Each scholar has a fundamentally different understanding of what it means to describe and explain the history of a dialect; these understandings in turn mandate different methodological approaches and a different understanding of the applicability of dialect data. They have also prevented each scholar from making an argument that can convince the other. Is it possible to recognize validity in each approach? If so, is there a broader, more comprehensive perspective from which we can look at each approach, and attempt to make a more objective analysis? By posing this question I do not mean to imply that the problem is an inherently subjective one, or that there are no right and wrong answers to questions about dialect history. On the contrary, it is by posing this question that I hope to be able to make a more meaningful critique and evaluation of competing hypotheses. The most recent articulation of Norman’s approach to dialectology is found in his 2007 article published in the Journal of Peking University. The main point of this minimanifesto can be found in the inal sentence of the abstract: “In the earlier period dialect fieldwork was based on character readings; later it was recognized that it is popular words that are the key to investigating dialect history” (2007:91).１４ Norman argues that the conventional practice of carrying out dialect fieldwork by eliciting readings of written characters arranged by QYS phonological categories is fundamentally lawed. Instead of using a list of characters in homophone groups (tóngyīn zìbiǎo 同 音字表), one should make use of a word list (cíhuìbiǎo 辭彙表) comprised primarily of vocabulary related to everyday life. Instead of thinking about dialect vocabulary as made up of literary (wén 文) and colloquial (bái 白) character readings, one should think of it as being made up of learned words (shūmiàn cíyǔ 書面詞語) and colloquial words (kǒuyǔcí 口語詞). Learned words are those more formal words that are used only by educated speakers; colloquial words are found in the speech of all levels of society. １４ 早期的方言田野調查以字音為主，後來認識到俗傳詞才是探索方言歷史的關鍵。 28 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects From a historical perspective, however, these two functional categories are crosscut by two other categories: words of literary origin (wénchuáncí 文傳詞) and words of popular origin (súchuáncí 俗傳詞). Literary words entered the spoken vocabulary at some point in the dialect’s history from a literary source; popular words have been passed down orally from generation to generation. “Colloquial words can be literary or popular in origin, but learned words are not used in ordinary speech, so they are not colloquial words. Popular words do not all have written character forms” (2007:92).１５ To understand the historical origins of a dialect, the ieldworker should concentrate on popular words, which are transmitted with minimal interference from literate inluences and so are more likely to directly relect the operation of regular rules of sound change on the earliest stages of the dialect. In contrast, the traditional “concept of wénbáidú 文白讀 specifically refers to reading pronunciations of Chinese characters, and so naturally does not take into consideration the conditions of ordinary people’s speech” (2007:93).１６ Although Norman’s early fieldwork made use of character lists, as he himself points out (2007:92), his work from the 1980s on seems to be based almost entirely on wordlists. If one examines the published dialect reports of most Chinese ieldworkers, one discovers that the concept of wénbáidú is indeed strongly embedded in both their elicitation process and in the method of presentation of data. But it is also true that these reports contain vocabulary lists that are independent of character readings, and are more similar to the materials that Norman employs in his fieldwork. Consider, for example, Chén and Lǐ 1991. The chapter on Northern Min dialects irst illustrates the relationship between QYS phonological categories and the sound systems of eight Northern Min dialects by listing the initials, inals, and tones of character readings (pp. 140-156). It is followed, however, by a vocabulary list which includes morphemes for which no standard character has been identiied (pp. 157-181). When we encounter discrepancies in the data presented by Norman on the one hand and by other researchers, such as Lǐ Rúlóng (who authored the Northern Min chapter in Chén and Lǐ 1991)１７, it is difficult to know how to interpret these discrepancies. For example, consider the following data from Norman 2000:279 and Chén and Lǐ 1991:152.１８ １５ 口語詞可能是文傳或俗傳的，但是書面語詞不是普通口語裡用的，所以不是口語詞。俗傳詞不都有漢字 可寫。 １６ 文白讀這個概念本來專指漢字的讀法，自然就沒有太考慮老百姓語言裡的情況怎麼樣。 １７ This chapter appears separately in the bibliography as Lǐ Rúlóng 1991. １８ I have substituted tone category numbers 1-8 for Chén and Lǐ’s tonal notations in order to facilitate comparison. 29 Zev Handel Norman 2000:279 字 ‘graph’ Shíbēi Jiànyáng Chóng’ān 崇安 Jiàn’ōu dzi6 lɔi6 lei6 tsi6 Shíbēi Jiànyáng Chóng’ān Jiàn’ōu tsi6 lɔi6 lei6 tsu6 Chén and Lǐ 1991:152 字 I have underlined the two points of discrepancy: the initial of the Shíbēi form and the inal of the Jiàn’ōu form. There are many explanations that one might posit to account for these discrepancies. Some possibilities that come to mind are: 1) One of the forms might be in error, either because the researcher misheard or misrecorded the form, or because of a typographical mistake in printing. 2) The native speakers consulted by each researcher may have had different dialectal or ideolectal pronunciations. 3) Norman’s form might be a popular vocabulary word elicited independently of the written form, while Lǐ’s form might be a character reading (whether ‘literary’ or ‘colloquial’). Before we look at the discrepancies in light of these three possibilities, an important point must be noted: both Norman and Lǐ have listed a Chinese character (in this case「字」) followed by attested forms in several Northern Min dialects.１９ But this superficial similarity is quite misleading. Norman has elicited the popular word for ‘graph; Chinese character’ in these dialects, i.e. the word that is used in ordinary speech. He has then determined that it is cognate with the words in other dialects and in older texts that are ordinarily written with the character「字」. In contrast, Lǐ is presenting a set of pronunciations of the character「字」according to educated, literate speakers of the dialects in question. Although the data is presented in a similar fashion, the underlying methods, and thus the signiicance of the Chinese character that labels １９ It is common practice in publications on Chinese dialectology to use a Chinese character as a convenient label for a cognate set, although this practice is seldom explicitly discussed. There is no doubt that these characters provide an efficient shorthand for referencing entire cognate sets. This is a uniquely Chinese convenience, as no analogous usage is possible for other languages that are the subject of dialectological inquiry. However, the convenience may be outweighed by the potential for misunderstanding. The problem is even more vexing in Chinese-language publications, because one alternative available in non-Chinese-language publications—labeling cognate sets with a gloss— is still potentially confusing in Chinese, where the gloss may itself be a single Chinese character that happens to be identical to the character associated with the cognate set. 30 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects the data set, is quite different. In the case of the Jiàn’ōu inals, one suspects either a typographic error or some sort of hypercorrected or literary pronunciation in Lǐ’s data. This is because Lǐ’s chart of final correspondences lists -i as the only Jiàn’ōu reflex of Zhǐ 止攝 rhyme group words like 字 (p. 142). The discrepancy in the Shíbēi initial is more interesting. According to Norman’s data and his Proto-Min reconstruction, we should always find a voiced Shíbēi initial corresponding to Jiànyáng l- when other dialects, like Jiàn’ōu, have a voiceless obstruent initial.２０ This correspondence pattern is completely consistent in the data presented in Norman 2000. The other cognate sets in Norman 2000 that reflect this pattern are ‘ramie’ (苧), ‘long’ (長), ‘bag’ (袋), ‘worth’ (值), and ‘far’ (for which no character is listed). Yet in Lǐ’s data we find two distinct correspondence patterns. One matches Norman’s, and is found in the pronunciations of the characters 雜, 達, 池, 寨, 銅, 讀, 除, and 毒. In the second pattern it is a Shíbēi voiceless 齊, 情, 斜, 徐, 謝, 崇, 苧, 道, initial that corresponds to Jiànyáng l-. Besides 字, other characters whose readings exemplify this second pattern are 重, 隊, 沉, 澤, 助. Because the second correspondence pattern occurs in a number of other words besides 字, we can be fairly conident that the Shíbēi pronunciation of 字 listed by Lǐ is not an error of transcription or typography. Most likely it reflects a difference in methodological approach to dialect ieldwork. Among all these words, only three are found in both data sets: ‘graph (字), ‘ramie’ (苧), and ‘far’.２１ ‘Heavy’ (重) also occurs in Norman’s data, but he gives initial t- for both the Shíbēi and Jiànyáng forms. This comparison of Norman’s and Lǐ’s data demonstrates that different approaches to gathering dialect data can lead to significantly different results. Perhaps more importantly, it also demonstrates that the manner of presenting that data can obscure important underlying differences. This is because Chinese characters commonly serve two distinct functions in Chinese dialectology: they are used on the one hand as a convenient representation of the standard Chinese cognate (whether that cognate is thought of as a Mandarin morpheme written with that character, as a lexical item in the QYS associated with that character, or as a word underlying the Literary Chinese use ２０ This sub-pattern is part of a broader pattern in which the ‘softened’ initials of Jiàn’ōu correspond to the voiced initials of Shíbēi. ２１ The forms Lǐ lists under the character are apparently the same cognate set as Norman’s ‘far’. See note 9 in Chén and Lǐ 1992: 190 on the identification of this character. 31 Zev Handel of the character), and on the other hand as a written form that has a particular dialectal pronunciation, or “reading”. Yet these distinct functions are seldom made explicit, can sometimes overlap, and may not even be consciously recognized by some scholars in the ield. But these discrepancies have further, crucial ramifications. Norman presents us with a set of colloquial words showing regular sound correspondences. It is entirely natural to see these as cognate sets derived by regular sound change from a common ancestor, and therefore to attempt to reconstruct the proto-forms according to the comparative method. Lǐ, on the other hand, presents us with data showing irregular sound correspondences (though with some underlying sub-patterns), which suggests the effects of dialect mixture and the presence of lexical layers. It is natural, therefore, to look to neighboring dialects and/or literary readings for the source of these phonological properties. It is also obvious why neither side of the debate is able to effectively argue against the other. Norman’s publications make no explicit mention of the correspondence pattern exemplified by Li’s data for 字, 重, 隊, 沉, 澤 and 助, which appears to constitute counter-evidence against Norman’s Proto-Min reconstruction. Because Norman fails to explain this data, scholars that have collected and presented it are reluctant to accept his conclusions. At the same time, Norman is suspicious of the data found in publications like Lǐ’s, because it is founded on the “wénbáidú concept” that he sees as largely irrelevant to an understanding of dialect history. Indeed, it is entirely possible that Norman has elicited colloquial forms that are similar to Lǐ’s, but has determined that they are not true popular forms, and has therefore omitted them from his publications. On what basis might Norman have determined that they are not popular forms? In some cases it may be because they are not commonly used words, and thus are likely of literary origin. But in other cases—and here I can do no more than speculate—it may be because they do not it the correspondence patterns that the bulk of popular words exemplify. This may appear to be a dangerously circular line of reasoning, but it is in fact a fundamental and important component of the comparative method. It is by identifying the regular correspondence patterns exemplified by the majority of basic-vocabulary cognate sets that one determines the standard of regularity against which irregular and exceptional developments are identiied and judged. Admittedly, some of my discussion above is based on speculation concerning the methods used by these scholars in collecting and presenting their data. But it is precisely the fact that speculation on these points is possible, indeed unavoidable, that demonstrates the acute need for a more explicit description of those methods on 32 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects the part of researchers who carry them out and publish the results. Without such a description the validity and signiicance of their data cannot be properly evaluated, and a meaningful academic exchange of ideas on the interpretation of the data cannot be undertaken. 5. Bridging the gap In two earlier publications (Handel 2003, 2009) I have argued in favor of the basic features of Norman’s Proto-Min reconstruction. At the same time, I have pointed out that the reconstruction is not in and of itself incompatible with the hypothesis that the Proto-Min phonological system is itself a stratified result of dialect mixture (Handel 2009:11-13). However, I have also noted that as of yet no convincing external source for the uniquely Min dialect features has been identified. Therefore, dialect mixture is not so much an explanation for those features as it is an as yet unsubstantiated hypothesis.２２ My support of Norman’s hypothesis is based on the acceptance of his data. To my mind, Norman’s Proto-Min reconstruction is the best, and at the moment the only convincing, explanation available for the set of popular words showing regular sound correspondences. But what about the dialect data beyond the set used by Norman in his reconstruction? Colloquial words of literary, rather than popular, origin are part of a dialect’s lexicon, and any comprehensive history must take them into account, and provide an explanation for their pronunciations. Words and pronunciations borrowed from other dialects are also a part of that history. These sets of words—borrowings, “artiicial” readings, literary forms, etc.—cannot be dismissed as autonomous subsets of the lexicon, independent of and irrelevant to the development of the oldest layers of popular vocabulary. Indeed, such words can affect the overall development of a dialect’s phonology, and thus have a signiicant impact on the pronunciation of popular words as well.２３ This leads us to ask a rather straightforward question: How can we explain the numerous apparent exceptions to Norman’s correspondences? Can the explanation shed ２２ I have argued that claims by Hirata, Wáng Fútáng, Zhèng-Zhāng Shàngfāng and others that neighboring Wu dialects are the main source of these features are unable to account for the Min dialect features in a comprehensive way. See Handel 2003: 69 footnote 28 and Handel 2009: 5. ２３ For a simple example, consider the way that the important of large amount of French vocabulary led to the phonologization of the previously allophonic distinction between voiceless and voiced fricatives in English. This had a major impact on the phonology and morphophonology of native English vocabulary. The history of the popular layer of the English lexicon cannot be fully understand without taking into account the influence of the borrowed layer. 33 Zev Handel further light on internal developments involving Min popular words? As a irst step, let us consider the set of Jiànyáng words that Hirata 1988 lists as exceptions to Norman’s rules of correspondence—words that are taken from Norman’s own early fieldwork on the dialect. Taking Norman’s Proto-Min reconstruction as a given, what range of patterns do we ind holding between the “regular” correspondences and these “irregular” exceptions? Finally, what conclusions can we draw about the overall history of Jiànyáng and the Northern Min dialects through a careful comparison of the regular and the irregular? 6. The data Hirata 1988:16 lists seventy-eight characters whose Jiànyáng pronunciations do not conform with Norman’s regular sound correspondences. The pronunciations are taken from Norman 1971. All of these characters have softened initials, and therefore should be expected to have tone values that correlate with those softened initials. Hirata (1988:15), following Norman, listed those correlations as follows: 平 上 去 入 QYS voiceless initials *p *ph *-p 1 1 9 3 3 3 5 5 9 7 7 3 QYS voiced initials *b *bh *-b 2 2 9 5 5 5 6 6 6 8 8 8 The irst three columns, with voiceless initial reconstructions, correspond to QYS voiceless initials and upper-register tones. The second three columns, with voiced initial reconstructions, correspond to QYS voiced initials and lower-register tones. The tonal relexes in Jiànyáng indicate that in four of the eight tones of the QYS system, namely yīnpíng, yángpíng, yīnqù, and yīnrù, the Proto-Min softened initials *-p and *-b have conditioned a tone split (as well as developing into uniquely ‘softened’ initial relexes). I have bolded the split tones to highlight the distinctions. We have already seen one example of this split above, in the tonal relexes of ‘tube’ (筒) and ‘sugar’ (糖) on the one hand and of ‘copper’ (銅) on the other. All three of these words fall into the QYS yángpíng tone category. The third has developed a softened initial l- and a distinct tonal relex. Hirata’s exceptions fall into 46 homophone groups, according to their QYS pronunciations. I have listed these 46 groups below. Following Hirata, I also list QYS homophones that do conform to Norman’s sound correspondences. In all cases, the exceptions involve a tone 2 pronunciation of a word with a softened initial that should, 34 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects according to Norman’s correspondences, have a tone 9 pronunciation. For example, in homophone group 7 the chart lists the two characters 培 and 陪, both pronounced vui². This is irregular according to Norman’s listed correspondences, since we would expect a tone 9 relex to correlate with the softened initial v-. The reading of 裴, listed to the right, is vui9, which shows the expected regular correlation. I have replaced Hirata’s simplified characters with the corresponding traditional forms. 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12) 13) 14) 15) 16) 17) 18) 19) 20) 21) 22) 23) 24) 25) 26) 27) “Irregular” characters lɔ² 駝 陀24 va² 琵杷 lo² 徒途圖 ly² 除儲 lai² 台苔抬 loi² 提 vui² 培陪 loi² 池馳 ki² ~ i² 奇 25 ki² ~ i² 岐祇 loi² 遲 26 ki² ~ i² 祁 loi² 持 ki² ~ i² 其麒 ki² ~ i² 祈 y² 葵 vau² 袍 lau² 逃萄 lio² 朝潮 kiu² ~ iu²; iu² 球; 裘 laŋ² 潭譚 laŋ² 談痰 loiŋ² 沉 kiŋ² ~ iŋ² 琴禽擒 lueŋ² 檀壇 lieŋ² 田 lueŋ² 團 馱 “Regular” characters lɔ⁹ 屠涂 lo⁹ 題 裴 loi⁹ vui⁹ 騎 i⁹ 棋期旗 ki⁹ ~ i⁹ 淘 lau⁹ ２４ This characters appears as 駝 (identical to the preceding character) in Hirata 1988. I have corrected it according to what appears in Norman 1971. ２５ This character appears as 祁 in Hirata 1988. I have corrected it according to what appears in Norman 1971. 35 Zev Handel 28) 29) 30) 31) 32) 33) 34) 35) 36) 37) 38) 39) 40) 41) 42) 43) 44) 45) 46) 傳 權 貧 頻 勤懃 群 旁 堂棠螳唐塘 強 狂 朋 籐騰滕 曾 橙 平 亭; 停 庭 蜓 瓊 同桐童瞳 窮 lyeŋ² kyeŋ² ~ yeŋ² voiŋ² voi² 27 keŋ² ~ eŋ² keŋ² ~ eŋ² vɔŋ² lɔŋ² kioŋ² ~ ioŋ² kuoŋ² ~ uoŋ² vaiŋ² laiŋ² laiŋ² 層 laiŋ² voiŋ² 評 laiŋ²; loiŋ² keŋ² ~ eŋ² loŋ² 銅 keŋ² ~ eŋ² laiŋ⁹ voiŋ⁹ loŋ⁹ It is important to note that the data in Norman 1971 was collected using character lists, rather than word lists.２８ At that time Norman had not yet fully developed his approach of using only popular words for comparative purposes. Hirata notes that among the seventy-eight exceptional character readings quite a few are for commonly occurring characters. Although Hirata doesn’t specify which words he is referring to, we might identify 「田」, 「平」, 「同」, 「朋」, and 「談」 as examples of commonly occurring characters. By this statement Hirata presumably means to refute Norman’s claim that his regular correspondences are based only on popular forms. However, we must take careful note of the fact that educated readings of even commonly occurring characters may not have a direct relationship to popular vocabulary and pronunciations. To raise one simple example, 「田」 is an extremely common character in the written Chinese tradition, representing the common Chinese word tián ‘ield’. But one cannot therefore assume that the character reading must be ２６ This character appears as 祈 in Hirata 1988. I have corrected it according to what appears in Norman 1971. ２７ The lack of -ŋ ending on this syllable is surprising, but this is indeed how the form appears in Norman 1971. ２８ Norman 1971: 24/29, in the introduction to section 2 “Lexicon”, indicates that the list was compiled by asking the informant to give “pronunciations for a character”. 36 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects in the popular layer of the language. As seen in both Norman 1996:31 and Chen and Li 1991:158, the regular word for ‘(rice) ield’ in Jiànyáng is thaiŋ², and cognates of this word occur in all the other Northern Min dialects.２９ In my own ieldwork on the Northern Min dialect of Chéngcūn 城村, I elicited the pronunciation liŋ² as the second syllable of the place name 興田, located near Chéngcūn.３０ This is a common, but in Norman’s terminology not a “popular”, word. In some cases, pairs of readings suggest a clear “literary” versus “colloquial” distinction. For example, in set 14, 其 ‘its’ and 麒 ‘part of word for kirin, a mythical animal’ are not morphemes in everyday use, while 棋 ‘chess’, 期 ‘time period’, and 旗 ‘lag’ are. This is in fact consistent with what Hirata (p. 14) notes about characters that Norman 1971 lists with both a literary and a colloquial reading, namely that in such cases the tone 9 reading is found in the colloquial reading. We can raise several questions about all of these “irregular” character readings from Norman 1971, especially those that appear to be common, basic vocabulary. Which of these character readings correspond to spoken morphemes? And which of those spoken morphemes are in ‘popular words’, words that have a history of continuous use throughout the history of the Min dialects, as opposed to words that entered the dialect from other sources? To the extent that any of these character readings are not associated with popular pronunciations, they could be considered irrelevant to Norman’s Proto-Min reconstruction. But their presence still requires an explanation: where did they come from? To give a concrete example, if the pronunciation lieŋ² for 「田」 is not from the popular Min layer, then where did the uniquely Northern Min softened initial l- come from? An examination of data from Norman’s later publications indicates that for some of the characters that in Norman 1971 are given only a single reading, and that appear in Hirata’s chart, Norman has later elicited popular pronunciations. For example, Norman 1996:34 has iu⁹ for Jiànyáng ‘ball’ (球) (compare Norman 1971’s kiu² ~ iu²), and Norman 1996:35 has loŋ⁹ for Jiànyáng ‘pond’ (塘) (compare Norman 1971’s lɔŋ²)３１. These forms—elicited by Norman as words, rather than as character readings—support his argument that the tone 9 pronunciations relect the popular layer, directly inherited ２９ Norman suggests that the word may be related to céng 層 ‘layer’, while Chén and Lǐ list the character chéng ‘raised path between fields’. ３０ The regular word for ‘field’ in Chéngcūn is thaiŋ². ３１ The discrepancy in finals between Norman 1971 forms with -ɔŋ and Norman 1996 forms with -oŋ, as in ‘pond’, is due to a difference in pronunciation between the speakers that provided Norman with the data. The historical distinction between -ɔŋ and -oŋ had merged to -oŋ in the dialect of the 1996 speaker. (Jerry Norman, personal communication, June 29 2009.) The vowel discrepancy is not relevant to the issue under discussion here. 37 Zev Handel from the earliest stages of Min. The forms would probably be considered “colloquial” character readings by Hirata, but conlating a colloquial character reading (báidú 白讀) with a popular spoken form obscures an important distinction. One approach to reconciling these various interpretations and presentations of dialect data is to think in more concrete terms about the history of particular dialects, the ways that reading traditions of characters develop, and the various mechanisms by which literary and popular forms inluence each other. Jiànyáng was a major commercial and publishing center from the 11th through the 17th centuries; in fact, it was one of the major centers of the Chinese publishing industry for much of that period.３２ As a thriving cultural and literary mecca, Jiànyáng was home to numerous schools and attracted a large population of literate and erudite Chinese, many of whom were studying for the imperial examinations and aspired to national prominence.３３ In this kind of environment, it would have been inevitable for one or more strong reading traditions to develop, based on Northern Chinese prestige readings of characters iltered through the local Northern Min phonology. As we see in many southern dialects with literary layers, these readings would have shown a high degree of correlation with the QYS tone and initial systems. Within the local speech community, they also would have served as high-prestige forms. Words and idioms found in Classical Chinese texts, as well as words (both mono- and bi-morphemic) from contemporary Northern Chinese would have been employed in the speech of the upper strata of society, and because of their prestige many of these words and forms would have iltered into the speech of the local community as a whole. At the same time, we might expect that attempts by the less-educated populace to sound more sophisticated might have led to the incorporation of salient features of the literary character readings into the pronunciations of cognate popular forms. These would in turn lead to the creation of hyper-corrected hybrid forms, in which literary features are grafted onto popular pronunciations. For example, one of the salient features of literary pronunciation would be the yángpíng tone pronunciation of many morphemes with tone 9 pronunciations in ordinary speech. It is not hard to imagine some speakers ３２ For a detailed description of the commercial and cultural role of Jiànyáng in Chinese history, see Chia 2002. ３３ According to Chia (2002:73): “Northern Fujian … was something of a cultural and political backwater until the Northern Song, when men from the region succeeded spectacularly in the government examinations: in numbers of jinshi [進士], Jianzhou ranked first in the empire with 809 during the Northern Song and continued to do well in the Southern Song… By around 1080, every county in Jianzhou had at least one government school, and many families had established schools for their members, with some wealthy families boasting more than one. By the time Zhu Xi [Zhū Xǐ 朱 熹 (1130-1200), the famous neo-Confucian scholar] and his students made the area a Daoxue [道學] stronghold in the late twelfth century, the Jianyang area had reached a cultural (and economic) level it has never again attained.” 38 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects deliberately pronouncing such morphemes with yángpíng (tone 2) in order to sound more learned. In some cases such a “hyper-corrected” pronunciation would end up becoming the formal reading pronunciation of a Chinese character; in other cases it would spread into the vernacular, co-existing with tone 9 pronunciations; and in cases where the vocabulary item was less basic, and thus less irmly embedded in the everyday speech of the community, the hyper-corrected form might completely displace the original tone 9 pronunciation. To take a concrete example, consider the word ‘pond’ (塘). The modern Jiànyáng pronunciation found in Norman 1996, loŋ⁹, relects the regular process of sound change from an earlier stage of the language. In contrast, the pronunciation lɔŋ², as found in Norman 1971 and listed by Hirata, reflects a hyper-correction with tone modified to conform to regular patterns of literary character pronunciation. The two exist side-byside in the modern language.３４ By way of comparison, I am not aware of the morpheme for ‘hall’ (堂) having any pronunciation other than lɔŋ², as found in Norman 1971 and listed by Hirata. In this case, Jiànyáng presumably had a pronunciation that would have been ancestral to modern loŋ⁹ or lɔŋ⁹, but the hyper-corrected form lɔŋ² completely displaced it. The pertinent question to ask then about such words is not “Where did the l- initial come from?” It was, in a sense, there all along. The question instead is “Where did the tone 2 pronunciation come from?” The answer I am proposing is that it was inluenced by prestigious reading pronunciations.３５ If the conjecture outlined above is correct, than we would expect to find signiicantly more “hyper-corrected” forms in Jiànyáng, a center of learning and culture, than in smaller and more remote locations. Away from the culture center, a much smaller percentage of the population would be familiar with literary pronunciations, and their inluence on the common language—in terms of both numbers and prestige— would be much smaller. And, in fact, this is precisely what we ind. Hirata (1988:16) observes that many characters pronounced in Jiànyáng with yángpíng (tone 2) pronunciations have tone 9 pronunciations in Sōngxī 松溪. Among such characters listed in his Table 5 we ind, for example, 堂 ‘hall’. Hirata takes this as evidence of a problem with Norman’s basic ３４ It is entirely possible that the “literary” pronunciation of the morpheme, elicited as a character reading, also occurs in everyday speech as part of learned compound words. ３５ The presumption that one part of a morpheme—just the initial, the final, or the tone—may undergo shifts due to dialect contact is not by any means a new proposal. See, for example, Ho 1988 or indeed the conclusion of Hirata 1988: 23. Similarly, hyper-correction, or the imitation of prestige pronunciations, may also affect just one part of a syllable. 39 Zev Handel claim that Tone 9 is a relection of Proto-Min distinctions, since it belies Norman’s claim of regular relexes and sound correspondences. But one should perhaps instead take it as evidence that in Jiànyáng, literary readings and the particular features associated with them had a greater overall inluence on the development of the vernacular. 7. Conclusion The “story” I outlined above, though plausible, is nothing more than a conjecture. The nature of the published data available on Jiànyáng and other Northern Min dialects makes it impossible to pursue or test the conjecture further. One would need a great deal more information on exactly how the data was collected and how it was collated, and in all likelihood one would need to collect additional data using new methods that carefully distinguish variant pronunciations of cognate morphemes within and across dialects according to their usage and history.３６ What all this suggests is the need in Chinese dialectology for a more integrated approach to dialect history, one that makes the methodology of both data collection and data analysis transparent, allowing for judicious application of three distinct approaches: the historical-comparative method, the study of dialect contact and lexical layering, and the investigation of reading practices and the formation of literate character pronunciations. My impression is that all the scholars whose work has been referenced here would agree to the same basic axioms of historical dialectology. First, that most dialects (and certainly the Min dialects) have lexicons composed of multiple strata as a result of dialect contact. Second, that the dissemination of character readings into the spoken language has created additional lexical layers. Third, that the oldest layers relect direct inheritance from the earliest historical stages of the dialects. Based on these axioms, a comprehensive and integrated approach to reconstructing the history of an individual dialect and its dialect group should involve all of the following steps: 1) Elicitation of as many forms as possible, including words used in everyday speech and learned pronunciations of Chinese characters. In all cases, the ３６ It would be interesting and useful, for example, to look anew at QYS yángpíng words in Jiàn’ōu in light of the proposals I have put forth here. In Jiàn’ōu, as in nearby Jiànyáng, we would expect to see a fairly strong influence from Northern literary pronunciations. Are there in fact hybrid forms in Jiàn’ōu similar to those in Jiànyáng, and do they occur in greater numbers than in Sōngxī? To what degree do we find hybrid forms in Jiàn’ōu and Jiànyáng correlated with particular lexical items, and to what degree are such developments independent? 40 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects precise method of elicitation should remain attached to the elicited forms. 2) Comparison of cognate forms across related dialects to establish patterns of regular sound correspondence. Multiple sets of correspondences will reveal and relect multiple lexical layers. 3) The most common correspondence patterns found in basic vocabulary elicited from everyday speech, across a wide variety of related dialects, can be assumed to relect the oldest layer. These cognate sets and correspondence patterns can be used to reconstruct the vocabulary and phonological system of the proto-language according to the traditional comparative method. 4) Other layers need to be compared to neighboring dialects and/or with QYS categories to determine their origin. Included in this comparison must be a determination of the degree of likely literary influence on local speech habits. The simplistic bipartite division of character readings into “literary” and “colloquial” should be abandoned in favor of a more sophisticated, openended view. Most modern dialectologists have worked on one or more of these four approaches. But they have not always been explicit about their methods or the nature of their data when doing so. More problematic, when debating the nature and historical origin of certain dialect features, they are often talking at cross-purposes precisely because of the lack of clarity about data and methods. This has had the unfortunate effect of creating the appearance of disagreement when there should not be any, and of diminishing the degree of persuasiveness of many important arguments in the ield. It is my belief that the application of the overall method I have just described will continue to support Jerry Norman’s basic conclusions regarding the initial and tone system of Proto-Min. However, it is not until all discussions about the problem are made more explicit and comprehensive that the ield will be able to move forward and reach a inal consensus. As a irst step, we might begin with a simple proposal. When presenting cognate set data like the following: 字 Shíbēi Jiànyáng Chóng’ān Jiàn’ōu tsi⁶ lɔi⁶ lei⁶ tsu⁶ we should strive to clearly indicate whether these pronunciations are (1) elicited morphemes which have been determined to be cognate to the Chinese morpheme conventionally written with the character labeling them or (2) pronunciations that have 41 Zev Handel been elicited as readings of the written character. In the case of (1), it should also be noted whether the morpheme has been elicited as part of a larger word, and if so, what the word was. Admittedly, the unambiguous presentation of Chinese dialect data is not always as easy to achieve as one might hope. Throughout the composition of this article, I have struggled with the question of how best to make use of Chinese characters in the presentation of cognate sets drawn from published sources in a way that is not potentially misleading. Overcoming difficulties like this will require a concentrated effort among researchers in the field to develop new, transparent conventions of typography and layout. I would argue that such an effort is a necessity for the field. Unless the presentation of dialect data is accompanied by an explicit description of the method of collection of that data, the proper interpretation of that data, and conclusions about dialect history based on that interpretation, cannot be considered reliable. Under such circumstances, progress in our understanding of the history of all forms of Chinese will be slow in coming. 42 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects Bibliography Chia, Lucille. 2002. Printing for Profit: the Commercial Publishers of Jianyang, Fujian (11th-17th Centuries). Cambridge: Harvard University Press. Chén, Zhāngtài (陳章太) and Lǐ Rúlóng (李如龍). 1991. 《閩語研究》。北京：語文出版 社。 Handel, Zev (韓哲夫). 2003. Northern Min tone values and the reconstruction of “softened initials”. Language and Linguistics 《語言暨語言學》4.1:47-84. 中譯本為《方言》 2004.1:54-75 中的〈閩北方言的調值與「弱化聲母」的擬測〉。 Handel, Zev (韓哲夫). 2009. 〈論閩北方言弱化聲母的歷史來源〉,《語言暨語言學》 10.1:1-16。 Hirata, Shōji (平田昌司). 1988. 〈閩北方言「第九調」的性質〉,《方言》 1: 12-24。 Ho, Dah-An (何大安). 1988.《規律與方向：變遷中的音韻結構》,中央研究院歷史語言研 究所專刊之90。台北：中央研究院歷史語言研究所。 Ho, Dah-An (何大安). 2007. 〈語言史研究中的層次問題〉，丁邦新編《歷史層次與方言 研究》，11-21。上海：上海教育出版社。 Ho, Dah-An. 1996. Stages and strata in dialectal history—case studies of Heng County, Da County and Shipo. New Horizons in Chinese Linguistics, ed. by James C.-T. Huang and Audrey Y.-H. Li, 215-234. Studies in Natural Language and Linguistic Theory 36. Dordrecht, Boston, London: Kluwer. Lǐ, Rúlóng (李如龍). 1985. 〈中古全濁聲母閩方言今讀的分析〉，《語言研究》 1:139149。 Lǐ, Rúlóng (李如龍). 1991. 〈閩北方言〉. 陳章太、李如龍著《閩語研究》，139-190. 北 京：語文出版社。 Lǐ, Rúlóng (李如龍) and Dèng Xiǎngzhāng (鄧享璋). 2007. 〈中古全濁聲母字閩方言今讀 得歷史層次〉。丁邦新編《歷史層次與方言研究》，227-246。上海：上海教育出 版社。 Norman, Jerry. 1971. A Guide to the Chien-yang Dialect. Washington, D.C.: Education Resources Information Center. Norman, Jerry. 1973. Tonal development in Min. Journal of Chinese Linguistics 1.2:222-238. Norman, Jerry. 1974. The initials of Proto-Min. Journal of Chinese Linguistics 2.1:27-36. Norman, Jerry. 1981. The Proto-Min finals. Proceedings of the International Conference on Sinology August 15-17, 1980 (Section on Linguistics and Paleography), 35-73. Taipei: Academia Sinica. 43 Zev Handel Norman, Jerry. 1986a. The origins of the Proto-Min softened stops. Contributions to SinoTibetan Studies, ed. by John McCoy and Timothy Light, 375-384. Cornell Linguistic Contributions 5. Leiden: E. J. Brill. Norman, Jerry (羅杰瑞). 1986b. 〈閩北方言的第三套清塞音和清塞擦音〉，《中國語 文》. 1:38-41。 Norman, Jerry. 1996. Tonal development in the Jennchyan dialect. Yuen Ren Society Treasury of Chinese Dialect Data 2:7-41. Norman, Jerry. 2000. Voiced initials in Shyrbei. In Memory of Professor Li Fang-kuei: Essays on Linguistic Change and the Chinese Dialects, ed. by Anne Yue-Hashimoto and Pang-Hsin Ting, 271-280. Taiwan: Academia Sinica Institute of Linguistics Preparatory Ofice. Norman, Jerry (羅杰瑞). 2007. 〈漢語方言田野調查與音韻學〉，《北京大學學報（哲學 社會科學版）》44.2:91-94。 Norman, Jerry and W. South Coblin, 1995. “A New Approach to Chinese Historical Linguistics.” Journal of the American Oriental Society 115.4:576-584. Ting, Pang-Hsin (丁邦新), ed. 2007. 《歷史層次與方言研究》。上海：上海教育出版社。 Wang, Feng. 2009. Review of Chinese Dialects and Historical Strata edited by Ting PangHsin. Journal of Chinese Linguistics 37.1:163-183. Wáng, Fútáng (王福堂). 1999. 《漢語方言語音的演變和層次》，北京：語文出版社。 Wáng, Fútáng (王福堂). 2004.〈原始閩語中的清弱化聲母和相關的「第九調」〉，《中 國語文》. 2: 135-144。 Wáng, Fútáng (王福堂). 2005.〈原始閩語構擬問題的研究過程〉，《語言暨語言學》6.3: 473-481。 Yue-Hashimoto, Anne (余靄芹). 1976. 〈論古漢語聲母的音韻對立〉，《中國語學》223: 2-5。 Zhèng-Zhāng Shàngfāng (鄭張尚芳). 1985. 〈浦城方言的南北區分〉，《方言》. 1:3945。 44 Competing Methodologies of Chinese Dialect Fieldwork, and their Implications for the Study of the History of the Northern Min Dialects 漢語方言調查不同的方法對研究 閩北方言歷史的含義 韓哲夫 西雅圖華盛頓大學 羅杰瑞先生所構擬的原始閩語已發表三十餘年，到目前為止尚未得到學界的 廣泛接受。有學者提出，羅氏構擬所依據的歷史比較法不適用於閩方言的特殊歷史 情況，主張需要先探索閩語層次才能了解閩語特殊的歷史來源。從表面上看，爭論 的關鍵似乎在於歷史比較法與層次分析法的對立。其實，兩种意見分歧的主要原因 是在於方言資料的收集與闡釋上的不同，而不是方法與分析上的不同。就是因為學 者們很少清楚地揭示這些基本設想上的差異，兩種對閩方言歷史的不同解釋的對立 至今都無法得到協調。 本文從另外一個角度重新探討羅氏所構擬的原始閩語「弱化聲母」，不僅考 察反對意見，也將他們所根據的資料和羅氏所根據的資料進行了比較。本人認為只 有透徹揭示出這些資料反應的基本設想上的差異，才能朝建立一個構擬閩方言（或 任何歷史複雜的方言）的共同方法的發展方向邁進。 關鍵詞： 閩北方言，弱化聲母，第九調，原始閩語，歷史比較法，方言層次，方 言歷史 45 Zev Handel 46