Hakka tone sandhi: Corpus and analytical challenges
Matthew Y Chen 陈渊泉; Xiuhong Yan 严修鸿; L.H. Wee 黄良喜
The volume you have in hand is a gold mine of dialectological data. More importantly, it is an invitation and a challenge.
For nearly five years the authors have elicited, transcribed, sorted, and tabulated the dauntingly complicated tonal patterns of Changting, a Hakka dialect of Chinese. The fruits of our collective labor of love are set out in tabular form in Part Three.
What motivated us was far more than archival interest. We picked Changting Hakka because we realized from previous descriptions that this dialect exhibits tone sandhi phenomena that present a formidable analytical challenge for any theoretical model. In its bare essentials, the problem posed by Hakka is not unlike what one encounters in rudimentary arithmetic. The value of 2 + 3 x 4 depends on whether addition or multiplication comes first:
2 + (3 x 4) = 14
(2 + 3) x 4 = 20
Likewise, given a tonal sequence /ABC/, the phonetic form depends on whether the elementary sandhi rules operate first on AB, then on BC, or vice versa. The crux of the matter lies in the discovery of general principles that predict the order in which the elementary operations combine to produce the observed outputs. For descriptive purposes, we use the term ‘directionality’ to refer to this problem. Needless to say, the problem is magnified when more than two operations are involved, e.g. in longer tonal strings. With the advent of Optimality Theory (OT), rule ordering is no longer part of the theoretical vocabulary. In which case, how do we even begin to understand and describe Hakka tone sandhi in optimality theoretic terms?
The main body of the text is a narrative of our systematic attempt to render a satisfactory account of the Hakka facts, either in rule-based generative framework or in constraint-based OT terms. We came to the grudging conclusion that we have failed, despite our best efforts. Hence, we have decided to present a full range of Hakka data in Part Three, in the hope that others may find a better solution. We suspect that a solution that eventually succeeds will entail significant, perhaps even drastic, refashioning of the analytical tools presently available. Hakka severely tests the descriptive capacity of existing theoretical models. In short, our failure is your opportunity.
It was for this reason that we have decided to make this monograph accessible to two sets of potential readers. We endeavor to ensure that the language-specific facts are glossed, translated and made transparent to any generalists without special Sinological background. Hopefully, what we have said so far has piqued the curiosity of the generalist enough to want to find out not only where the analytical tools came up short, but also to have the full set of facts against which to test the adequacy of theoretical alternatives. However, we wish also to address the needs of a large number of specialists in Chinese dialectology by detailing in Chinese the theoretical significance of the corpus of data we have painstakingly assembled. Part One (in English) and Part Two (in Chinese) cover roughly the same material, but are not exact translations of each other; rather, they target different audiences, and are designed accordingly. For instance, the English narrative omits some details of interest mainly to Chinese dialectologists. By the same token, the Chinese text says nothing about Optimality Theory, the assumption being that any specialists who are attuned to OT-related issues are already familiar with the OT literature that exists almost entirely in English.
The narrative texts in Parts One (English) and Two (Chinese) are organized as follows: The introductory Chapter One provides the basic background information about Hakka, and describes the elementary sandhi rules that operate on two-tone windows. Chapter Two is devoted to the core problem of ‘directionality’. There we examine a number of ‘derivational constraints’ such as Temporal Sequence, Tonotactic Wellformedness, and Transparency etc. as potential predictors of the order in which the elementary sandhi rules must apply in order to produce the attested output. Chapter Three (English only) explores the analytical options OT has to offer, including Sympathy Theory.
Part Three begins with a quick reference chart to tritonal sandhi patterns as well as explanatory notes (in both English and Chinese) on how to read the tables. The rest of Part Three consists of 241 pages of tables, arranged according to syllable count, then in the traditional order of tonal categories (that hold remarkably stable across dialects), namely:
M – R – F – H – L
i.e. Mid, Rising, Falling, High and Low
The monograph ends with bibliographical references.
M R F H L
阴平 阳平 上声 阴去 阳去
Part 1 (in English) 第一部分 (英文版）
1.1 Background 1.2 Phonological system
1.3 Ditonal sandhi patterns
1.4 Autosegmental representation
2.2 The problématique
2.3 Directionality: Against rule ordering
2.4 Predicting Directionality
2.5 Ranking paradox
2.5.2 Conjoint constraints
2.6 Directionality in longer sequences
2.6.1 Statement of the problem
2.6.2 Predicting ‘Traffic Flow’ from ‘Overlapping Directionality’
2.6.4 Primary and Marginality
2.6.5 Neutral and underivable types
2.7 Extra-long sequences
Matthew Y Chen 陈渊泉, Xiuhong Yan 严修鸿, L.H. Wee 黄良喜
3.1 The general character of an Optimality Theoretic approach
3.2 Ditonal sandhi
3.3 Derivational challenges to OT
3.4 Improving the model with Sympathy Theory
Part 2 (in Chinese) 第二部分（中文版）
2.5 等级排序的矛盾(Ranking Paradox)