i

Modern Chinese Parts of Speech

What is the essence of a part of speech? Why is it difficult to classify parts

of speech? What are the bases and criteria for classifying them? How should
they be classified? In doing so, how should a conversional word be dealt with?
How should nominalization be treated? These are just some of the questions
answered in this book.
The classification of parts of speech in Chinese is a tough job due to the
language’s lack of morphological differences. Based on the analysis of nearly
40,000 Chinese characters, this book proposes that, essentially, a part of
speech is not of a distributional type and that its intrinsic basis is an expres-
sional function and the semantic type. Essentially, large categories such as sub-
stantive words, predicate words and modification words are classes of words
classified according to their expressional functions. Basic categories such as
nouns, verbs and adjectives are classes that combine semantic types with syn-
tactic functions. In classifying parts of speech, the book pays attention not
to identifying a single distributive characteristic that is internally universal
and externally exclusive but to clustering the grammatical functions that have
the same classification value through the “reflection-​representation” rela-
tionship among distribution, expressional function and semantic type (dis-
tribution reflects expressional function and semantic type, which are, in turn,
represented as distribution), thereby identifying the classification criteria.
It uses distributional compatibility and the correlation principle to analyze
which distributional differences represent differences in parts of speech and
which do not. In this way, grammatical functions that have equal classifica-
tion values are collected into one equivalent function cluster, each of which
represents one part of speech. The book uses four strategies to classify parts
of speech, namely the homogeneity strategy, the homomorphical strategy, the
priority homomorphical strategy and the consolidation strategy. It will be a
valuable reference for Chinese linguistic researchers and students as well as
Chinese learners.

Guo Rui is a professor in the Department of Chinese Language and Literature,

Peking University. His main research lies in Chinese grammar and semantics.
ii

Chinese Linguistics

Chinese Linguistics selects representative and frontier works in linguistic

­disciplines including lexicology, grammar, phonetics, dialectology, philology
and rhetoric. Mostly published in Chinese before, the selection has had far-​
reaching influence on China’s linguistics and offered inspiration and reference
for the world’s linguistics. The aim of this series is to reflect the general level and
latest developments of Chinese linguistics from an overall and objective view.

Titles in this series currently include:

Prosodic Syntax in Chinese
History and Changes
Feng Shengli

Modern Chinese Parts of Speech

Classification Theory
Guo Rui

Modern Chinese Parts of Speech

Systems Research
Guo Rui

The Experiential Guo in Mandarin

A Quantificational Approach
Ye Meng

Research on Functional Grammar of Chinese

Information Structure and Word Ordering Selection
Zhang Bojiang and Fang Mei

Research on Functional Grammar of Chinese

Reference and Grammatical Category
Zhang Bojiang and Fang Mei

For more information, please visit https://​www.routledge.com/​

Chinese-​Linguistics/​book-​series/​CL
 iii

Modern Chinese Parts

of Speech
Classification Theory

Guo Rui
iv

First published in English 2019

by Routledge
2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
and by Routledge
52 Vanderbilt Avenue, New York, NY 10017
Routledge is an imprint of the Taylor & Francis Group, an informa business
© 2019 Guo Rui
Translated by Zuo Liankai
The rights of Guo Rui to be identified as author of this work and Zuo Liankai
as its translator have been asserted by them in accordance with sections 77 and
78 of the Copyright, Designs and Patents Act 1988.
All rights reserved. No part of this book may be reprinted or reproduced or utilised
in any form or by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying and recording, or in any information
storage or retrieval system, without permission in writing from the publishers.
Trademark notice: Product or corporate names may be trademarks or registered trademarks,
and are used only for identification and explanation without intent to infringe.
English version by permission of The Commercial Press.
British Library Cataloguing-​in-​Publication Data
A catalogue record for this book is available from the British Library
Library of Congress Cataloging-​in-​Publication Data
Names: Guo, Rui, 1962– author.
Title: Modern Chinese parts of speech : classification theory / Guo Rui.
Other titles: Xian dai Han yu ci lei yan jiu. English
Description: London ; New York, NY : Routledge, 2019. |
Series: Chinese linguistics | Includes bibliographical references and index.
Identifiers: LCCN 2018039608 (print) | LCCN 2018046278 (ebook) |
ISBN 9781351269209 (ebook) | ISBN 9781138576711 (hardcover)
Subjects: LCSH: Chinese language–Parts of speech. |
Chinese language–Syntax. | Chinese language–Grammar.
Classification: LCC PL1231.5 (ebook) |
LCC PL1231.5 .G8613 2019 (print) | DDC 495.15–dc23
LC record available at https://lccn.loc.gov/2018039608
ISBN: 978-​1-​138-​57671-​1  (hbk)
ISBN: 978-​1-​351-​26920-​9  (ebk)
Typeset in Times New Roman
by Newgen Publishing UK
 v

Contents

List of figures  vi
List of tables  vii
Foreword to the Chinese Edition  viii

1 Introduction  1

2 Basic issues concerning classification of parts of speech  21

3 Possibilities and purposes of classifying parts of speech  41

4 Essence and expressional functions of a part of speech  54

5 Criteria for classifying parts of speech  107

6 How to classify parts of speech according to distribution  128

7 Conversional words and nominalization  157

8 Conclusions  187

Bibliography  192
Index  203
vi

Figures

1 .1 A brief description of modern Chinese part-​of-​speech system  19

4.1 The nearest neighbor linkage hierarchy  69
4.2 The furthest neighbor linkage hierarchy  70
4.3 The average neighbor linkage hierarchy  70
4.4 Clustering analysis of distributional similarity degrees
of 60 words (nearest neighbor linkage technique)  71
4.5 Clustering analysis of distributional similarity degrees
of 60 words (average linkage technique)  72
4 .6 The transformation caused by a vacant constituent  99
4.7 Syntactic constructions of reference  101
6.1 Example 1  134
6.2 Example 2  134
7.1 Properties of parts of speech  159
7.2 The approximately complete overlapping and approximately
inclusive relations between two parts of speech  165
7 .3 The dominant hierarchy of parts of speech (Guo Rui, 1997)  165
7.4 The overlapping part accounts for 40% to 80% of the
dominant part of speech (Part of Speech A)  166
7 .5 The overlapping part takes the minority (<20%)  166
7.6 The overlapping of verb property and nominal property  173
7.7 The correspondence between a part of speech and a syntactic
constituent in English and Chinese  175
7.8 The correspondence between a Chinese part of speech and a
syntactic constituent  176
 vii

Tables

2 .1 The distinctions among morphology, word formation and syntax  34

4.1 Zhu Dexi’s (1982b) criteria for distinguishing between verbs
and adjectives  61
4.2 The possible classifications obtained with the two criteria of
“~ object” and “很 (very) ~”  62
4.3 The similarity degrees among five individual words  70
4.4 Semantic bases of parts of speech  91
4.5 Croft: Correlations among syntactic category, semantic type
and pragmatic function  104
4.6 Croft: The unmarked or marked correlations in English  105
5.1 Word formation methods and the capability of verbs to take
objects (double-​syllable verbs)  125
5.2 The number of syllables and the capability of a verb to
function as an attributive  126
6.1 Quoted from Zhu Dexi’s Handouts on Grammar, p. 55  129
6.2 Compatible degrees of the main grammatical functions of
Chinese notional words  136
6.3 The correlation between compatible degree and word
frequency, and that between function and word frequency  141
6.4 The correlation between number of conversional words and
word frequency  142
6.5 The compatible degrees among specific grammatical functions
of modification words  146
6.6 The compatible degrees among specific grammatical functions
of substantive words  149
6.7 The correlation between word frequency and compatible
degree (substantive word)  150
7.1 The major overlapping parts of speech  168
viii

Foreword to the Chinese Edition

Lu Jianming

Modern Chinese Parts of Speech:  Classification Theory by Guo Rui finally

meets its audience today. The book comes up with several revisions on the
basis of his doctoral thesis and is the encouraging outcome of his devoted
exploration of modern Chinese parts of speech over more than a decade.
Chinese parts of speech have always been very difficult problems. What
are the problems? Why have these problems not been solved? Guo Rui has
summarized, analyzed and disentangled them very well in his book. He has
summed them up in the following five questions:

1. Does Chinese have parts of speech?

2. What is the nature of a part of speech?
3. What are the bases and criteria for classifying parts of speech?
4. How can parts of speech be determined (how to classify them)?
5. How can conversional words be dealt with?

These questions are occasionally discussed in academic circles but fall short
of in-​depth investigation and exploration. Guo Rui has analyzed the causes
of this in his book: first, our understanding of parts of speech comes from
Western Indo-​European linguistics. The words of Indo-​European languages
have morphological markers and changes when used in a sentence, and the
bases for classifying the parts of speech in Indo-​European languages are these
morphological markers and changes. On this basis, it is thought that parts of
speech are classified according to word forms, but Chinese words have neither
morphological markers nor changes. Therefore, Chinese words do not have
parts of speech.
Second, Indo-​ European languages essentially have a one-​ to-​
one cor-
respondence between their parts of speech and syntactic constituents.
For example, a noun functions as a subject or object; a verb functions as a
predicate; an adjective as an attributive; an adverb as an adverbial and so
on. Accordingly, it is thought that parts of speech are classified according
to a word’s capability of functioning as sentence constituents. But Chinese
essentially has a one-​to-​many correspondence between its parts of speech
and sentence constituents; a word can often function as either subject,
 ix

Foreword to the Chinese Edition ix

object, predicate, attributive or adverbial and so on. The multiple functions
of a Chinese word are a headache for some scholars in Chinese academic
circles who use sentence constituents as the basis for classifying Chinese parts
of speech. They also create another problem in that it is difficult to decide
the syntactic relations of a good many syntactic structures. For example, is
容易掌握 (easy to command) a predicate-​object relation or an “adverbial-​
headword” modifier and a modified relation? Is 便于掌握 (convenient to
command) a predicate-​object relation or an adverbial-​headword modifier and
modified relation? Is 决心干到底 (determined to fight to the end) a predicate-​
object relation or an adverbial-​headword modifier and a modified relation? It
is difficult to decide. The difficulty in deciding the syntactic relations of these
constructions may affect understanding of the parts of speech of such words
as 容易 (easy), 便于 (convenient) and 决心 (determined).
Third, words in the synchronic hierarchy of any language actually have
different historical and field hierarchies. Indo-​European languages have mor-
phological changes, and the morphological changes of words are very small
and basically not affected by historical and field hierarchies. Because Chinese
words do not have morphological changes, the use of words in different his-
torical and field hierarchies may vary greatly, undoubtedly creating many
troubles for the classification of Chinese parts of speech. However, in the
past, we always lacked an understanding of them.
Fourth, an even more fundamental reason why Chinese parts of speech
have always been a very difficult problem is that when they were discussed
in the past, only some typical examples were cited, but no one has actually
examined one by one the uses of tens of thousands of modern Chinese words.
Thus, everybody can only engage in idle theorizing.
The above analyses all show keen insights; however, I must add one more: fifth,
what is the intrinsic nature of a part of speech? Since structuralist language
views have been prevalent in the world, the intrinsic nature of a part of speech
has been distribution, that is, a part of speech belongs to a distribution type.
These views are universally accepted. Chen Wangdao expressed these views in
the discussions on grammatical innovation in the 1940s. Then, starting in the
1960s, Zhu Dexi emphasized and expounded on these views again and again
(see Zhu Dexi, Lu Jiawen & Ma Zhen 1961; Zhu Dexi 1982a; Zhu Dexi 1985a
respectively). Furthermore, these views by Zhu Dexi are considered “extremely
enlightening new views” proposed through “breaking away from the interference
and constraint of the traditional Indo-​European language views and looking at
Chinese in a plain light”. From the present perspective, this view does not have
a deep understanding of the essence of a part of speech. Therefore, the Chinese
part-​of-​speech problem has still not been solved satisfactorily.
In my opinion, the prominent contributions of Guo Rui’s Modern Chinese
Parts of Speech: Classification Theory, first and foremost, lie in the fact that
he has proposed his brand-​new and more profound views on the essence of a
part of speech: “The essence of a part of speech is not distribution. Therefore,
it is difficult to succeed in any attempt to classify Chinese parts of speech
x

x  Foreword to the Chinese Edition

through seeking the internally universal and externally exclusive distribu-
tion characteristics.” A  part of speech is the type of a word’s grammatical
meaning, which we call an expressional function, namely the meaning type
of a word in its combination, such as the main types of statement, reference
and modification, and the small types of entity, position, measurement unit,
quantity and demonstration. Therefore, “a part of speech is actually the clas-
sification of words that uses their expressional functions at lexical level as
intrinsic basis”. He figured out his views not just by racking his brains but
through conducting actual investigations and devoted analyses of more than
40,000 modern Chinese words. He figured them out through his profound
analyses of a variety of loopholes of the distribution theory, the similarity
theory and the prototype theory in the Chinese part-​of-​speech classification.
More importantly, he figured them out through his thorough examination of
this question:  the grammatical position or combinatory position has selec-
tional restriction on the words that enter into the position. The selectional
restriction definitely needs certain bases. Then what are its bases? By using his
ample perceptual knowledge gained through the actual investigation of more
than 40,000 words and long-​term and repeated research, he discovered that
the bases for the selectional restriction of a grammatical position on words
are not distribution itself but certain properties at deeper levels, namely the
expressional functions of words. The concepts of word properties such as the
so-​called substantive word and predicate word actually reflect expressional
functions such as reference and statement at the lexical level. It is a pity that
people did not realize this in the past. Zhu Dexi was the earliest person to
propose the two concepts of “reference” and “statement” (see “Self-​Reference
and Transfer Reference –​Grammatical Functions and Semantic Functions of
Chinese Nominalization Markers such as de (的), zhe (者), suo (所), zhi (之)”,
in Dialect, Issue 1, 1983). But Guo Rui’s understanding goes deeper.
First, Zhu Dexi only classified expressional functions into reference and
statement, whereas Guo Rui has classified them into four basic types:  A.
A  statement denotes an assertion; B.  A  reference denotes an object;
C. Modification restricts a statement or reference; D. An auxiliary function
denotes regulation.
Second, Guo Rui thinks that expressional functions can be classified into
two levels: intrinsic expressional function and extrinsic expressional function.
The intrinsic expressional function of a word is inherent; the external expres-
sional function is the one realized by a word in its certain grammatical pos-
ition. Under usual circumstances, the expressional functions at the two
levels are consistent, for example, 小王黄头发 (Little Wang has yellow hair),
where 小王 (Little Wang) is a reference at either level. But sometimes they
may not be consistent. For example, 黄头发 (yellow hair) is a reference as far
as intrinsic expressional function is concerned; however, it is a statement in
terms of extrinsic expressional function because it can be modified at its front
by some adverbs, for instance, 小王也黄头发／小王的确黄头发 (Little Wang
also has yellow hair; Little Wang indeed has yellow hair).
 xi

Foreword to the Chinese Edition xi

It must be pointed out that Guo Rui’s expressional function is also a
kind of grammatical meaning of a word. The intrinsic basis for classifying
words into, for example, nouns or verbs is actually the distinction between
such expressional functions as a reference and a statement. The distribution
difference and morphological difference among parts of speech are simply
the external exhibitions of differences in expressional functions. Just like
expressional functions have layers, Guo Rui correspondingly classifies parts
of speech into two layers –​the parts of speech that correspond to intrinsic
expressional functions are at the lexical level, and those corresponding to
extrinsic expressional functions are at the syntactic level. The part of speech
at the lexical level is a word’s inherent one, which is labeled in a dictionary,
while that at the syntactic level results from the use of words, which can be
controlled by syntactic rules. The part of speech of 黄头发 (yellow hair) cited
in the above example is nominal at the lexical level, but in the above example
sentence, it is a predicate at the syntactic level. One can easily understand
that “expressional functions are specified by a language’s internal compos-
ition and reflect the relationship among language signs, thereby having their
grammatical meanings. They do not reflect the relationship between language
sign and the real world, thus having no conceptual meaning; they also do not
reflect the relationship between language sign and its user, thereby having no
pragmatic meaning”.
Readers may ask, since an expressional function is also a kind of gram-
matical meaning, why shouldn’t one directly say that “parts of speech should
be classified according to a word’s grammatical meaning”? Theoretically
speaking, this is true, but grammatical meanings themselves have or can
be understood at different levels. Therefore, Guo Rui uses the expressional
function to illustrate the nature of a part of speech.
Guo Rui’s views on parts of speech are obviously a great breakthrough
in traditional views and have universal linguistic theoretical meanings.
Classifying parts of speech on the intrinsic bases of the expressional function
at the lexical level of a word or not regarding distribution but the expressional
function as the essence of a part of speech “can explain why it is comparable
with different languages in different eras,” and why words distribute differently
in different eras and languages yet belong to the same part of speech. Just as
Guo Rui illustrated, the words 看 (see) in modern Chinese and 视 (watch) in
classic Chinese distribute differently. The former can take a numeral plus a
measure word object, but cannot be modified by a numeral, such as 看三次
(see three times), but not 三看 (three sees), whereas the latter cannot take a
numeral plus a measure word object and can be modified by a numeral, such
as *视三／三视 (*watch three/​three watch). But both are verbs. The English
word “stone” can be modified by a numeral, for example “two stones”, and
can function as the object of a preposition indicating a place, for instance,
“on the stone”, whereas the Chinese word 石头 (stone) does not have these
functions. But both are nouns. The view that the essence of a part of speech
is its expressional function can produce a reasonable explanation:  because
xii

xii  Foreword to the Chinese Edition

a part of speech has the same expressional function. A  part of speech is a
category; the same part of speech must have commonality of its properties;
different parts of speech must have differences in their properties. The com-
monality and difference in property are those of expressional functions and
the bases of language comparison.
Guo Rui’s views on part of speech may not be approved of by all my
colleagues; nevertheless, I hope these colleagues will not immediately oppose
or suspect them. There is no harm in seriously pondering his new views
on parts of speech. Likewise, there is no harm in knowing the operational
procedures for his classification of Chinese parts of speech, and then deciding
whether to agree or object, to completely or partially agree, or to completely
or partially object.
The second contribution made by Guo Rui’s Modern Chinese Parts of
Speech: Classification Theory is that it proposes new operational procedures
and methods for classifying parts of speech. Namely, the book calculates the
degree of compatibility among grammatical functions to reveal the relation-
ship between grammatical function and part of speech. In what sense? As
mentioned earlier, Guo Rui considers an expressional function as the essence
of a part of speech. The essential bases for classifying parts of speech are the
expressional function of a word, which is, however, invisible and impalpable.
In reality, it is not operative to classify the parts of speech of a word in direct
accordance with its expressional function. We must find observable and spe-
cific classification criteria that can truly reflect the essence of a part of speech.
Then how should we determine classification criteria? Guo Rui points out in
his book that different grammatical functions have different values for clas-
sifying different parts of speech: some have distinctive values, whereas others
have non-​distinctive values. Therefore, the various grammatical functions
expressed by a part of speech should not be put on a par and treated equally.
We only select those distinctive grammatical functions for part-​of-​speech clas-
sification. In order to classify parts of speech effectively with those distinctive
grammatical functions, Guo Rui created the concepts of “equivalent function”
and “heterovalent function”. Some differences among the functions demon-
strate the differences in the properties of a part of speech. For example, the
pair of grammatical functions 不 (not) ~ and 〈attributive〉~ demonstrates
different properties of a part of speech and can be regarded as a heterovalent
function, whereas the grammatical functions that reflect such identical prop-
erties as functioning as subject or object can be regarded as an equivalent
function. Guo Rui points out that, theoretically speaking, if we can iden-
tify all the functions of a word, determine which functions are equivalent
and form them into clusters of equivalent functions, then we identify the dis-
tinctive grammatical functions of all the parts of speech and largely classify
all the words into different parts of speech. Therefore, the determination of
equivalent functions is the key to identifying classification criteria and deter-
mining how many parts of speech a language has. Obviously, the determin-
ation of equivalent functions is the key to classifying parts of speech. Then
 xiii

Foreword to the Chinese Edition xiii

how should we determine equivalent functions? Guo Rui suggests that the
compatibility of grammatical functions is useful in this regard. The compati-
bility is manifested in (A) whether several different grammatical functions can
be regarded as equivalent function and (B) whether the different grammat-
ical functions are valuable for classifying parts of speech. If we can calculate
the degrees of compatibility among grammatical functions, we can identify
the equivalent functions that can distinguish between parts of speech and
can use the degrees as specific classification criteria. For example, a word
that can function as a subject can also function as an object, and vice versa.
Then, functioning as a subject and object is a set of compatible grammatical
functions. For another example, if a word that can enter 很 (very) ~ can also
enter ~ 极了 (extremely) and ~ 得很 (nicely), and if a word that can enter ~
极了 (extremely) and ~ 得很 (nicely) can also enter 很 (very) ~, then 很 (very)
~, ~ 极了 (extremely) and ~ 得很 (nicely) are a set of compatible grammat-
ical functions. But the compatibility of grammatical functions between 不
(not) ~ and 〈numeral〉, and that between 〈numeral and measure word〉~
and 〈functioning as adverbial〉 is extremely small. As mentioned before, the
fundamental bases of selectional restriction of a grammatical position on a
word are its expressional functions, and their extrinsic manifestations are the
grammatical functions of the word. Obviously, the grammatical functions that
have a rather big compatibility often reflect the same selectional restriction of
the two different grammatical positions on words and also the common prop-
erties of a part of speech, thus being an equivalent grammatical function. The
two grammatical functions that have a rather small or no compatibility degree
reflect different properties of a part of speech and are generally heterovalent
functions. Hence, we can use the compatibility of grammatical functions to
determine their values for classifying parts of speech. In his book, Guo Rui
has proposed not only the concept of “compatibility between grammatical
functions of a word” but also the methods and formulae for calculating their
compatible degrees. Undoubtedly, this advances one step forward to the dem-
onstration of part-​of-​speech classification. Superficially, it seems that Guo
Rui still classifies parts of speech of a word according to its distribution and
grammatical functions, but he does it fundamentally differently from what
others did previously, and goes deeper and further, as it were. This may be
profoundly understood as long as a reader carefully reads the whole book.
As mentioned previously, Guo Rui’s new views on Chinese parts of speech
were figured out not just by racking his brains but also through undertaking
actual investigation and devoted analyses of more than 40,000 modern
Chinese words. This can be considered a prominent merit of the book. This
may also be said to be one of its features compared with other monographs of
the same kind on Chinese parts of speech. Reading through the whole book,
a reader can easily discover that it utilizes massive and detailed statistical fig-
ures to do all the following: analyze the paradoxes of the overall similarity
clustering view, the prototype theory, the partial distribution view, the overall
distribution view and so on; explain the calculation of the compatible degrees
xiv

xiv  Foreword to the Chinese Edition

among grammatical functions of parts of speech and the compatible degrees
among the major grammatical functions of Chinese; demonstrate Chinese
conversional words and homographs belonging to different parts of speech;
describe specific criteria for classifying parts of speech and specific cases;
expound the correlation between word frequency and grammatical function,
and clarify the common grammatical functions and other characteristics of
various parts of speech. You may as well look at his explanations of common
grammatical functions and other characteristics of adjectives:

1. 99.47% of adjectives can function as a predicate.

2. 98% of adjectives can be modified by absolute degree adverbs represented
by 很 (very).
3. 94% of adjectives can be negated by 不 (not), but many fewer of them can
be negated by 没有 (no).
4. 97% of adjectives can be modified by other adverbials.
5. 83% of adjectives can take complements.
6. 69% of adjectives can function as complements. 67% of adjectives can
function as compound complements; 8% of adjectives can function as
conglutinate complements.
7. Only ten adjectives can take real objects but must be followed by quanti-
tative quasi-​objects that generally denote comparison and occurrence, for
example, 高他一头 (one head higher than he), 大他两岁 (two years older
than he), 熟了一个 (ripe one piece).
8. 72% of adjectives can take auxiliary words such as le (了), zhuo (着) or
guo (过), but few of them can take zhuo (着).
9. 93% of adjectives can function as a subject or object. Similar to the verb
that functions as a subject or object, an adjective that functions as a sub-
ject or object also has the following two cases. One case is that the adjec-
tive in the position of a subject or object still retains its properties, for
example:
(16)  a. 认真不好 (Being serious is not good)
b. 不认真不好 (Not being serious is not good)
c. 太认真不好 (Being too serious is not good)
d. 办事认真不好 (Doing things seriously is not good)

The second case is that an adjective that functions as a subject or object

reflects the properties of a noun, for example:

(17)  a. 追求幸福 (Pursue happiness)

b. *追求很幸福 (*Pursue very happiness)
c. 追求自己的幸福 (Pursue one’s own happiness)
(18)  a. 保持平衡 (Keep one’s balance)
b. *保持很平衡 (*Keep very balance)
c. 保持身体的平衡 (Keep one’s bodily balance)
 xv

Foreword to the Chinese Edition xv

The adjectives in the second case that function as a subject or object are
still called adjectives, and we do not treat them as conversional words, instead
calling them “nominal adjectives”.

10. 32% of adjectives can be modified by attributives; this also has two
cases. One is that adjectives modified by attributives retain their proper-
ties; at this time, the attributive takes “de (的)”, for example:
(19) a.  形势的稳定有利于经济发展 (The stability of situation favors
economic development)
b.  形势的不稳定不利于经济发展 (The instability of situ-
ation is adverse to economic development)
c. The rapid stability of situation favors economic develop-
ment (形势的迅速稳定有利于经济发展)

Another case is that the adjective reflects only the properties of a noun
(0.43%), for example: economic difficulty (经济困难), 生命危险 (life danger).
The adjectives in the second case are all nominal adjectives.

11. Many adjectives can function as attributives, but their number is far
fewer than imagined in the past, accounting for only 29% of the total.
12. 12% of adjectives can directly function as adverbials. Although some
adjectives cannot do so, they can do so after adding “di (地)” (40% of
adjectives). Because not a high proportion of adjectives can function
as adverbials, it works to treat those adjectives that can function as
adverbials as conversional words that function as adjectives and adverbs.
13. 2.73% of adjectives are separable phrasal characters, whose adjectives
have only one form of verb-​object, for example, 吃惊 –​ 吃了一惊 (taken
aback –​taken one aback), 称心 –​称他的 心(satisfied –​satisfy his heart).
14. 15% of adjectives have their corresponding reduplicated forms.

This fully reflects his solid academic discipline.

Of course, we cannot say that Guo Rui’s theory and methods for classi-
fying parts of speech can thoroughly solve once and for all the problems of
classifying Chinese parts of speech; neither can we require him to do so. The
reasons are twofold:  on the one hand, it is not so easy to really carry out
the classification; on the other hand, just as Guo Rui himself said, “we do
not know sufficiently the status of quite special words such as onomatopoeia
in a part-​of-​speech system. We are not clear how to demonstrate it with the
methods presented in this book”. Any researcher should cherish this matter-​
of-​fact attitude toward his or her academic research.
In conclusion, Guo Rui’s Modern Chinese Parts of Speech: Classification
Theory provides readers with new spirits and thoughts on classifying Chinese
parts of speech. Of course, everybody should pay attention to his conclusions,
or his brand-​new views on parts of speech, but I think his way of thinking
and analysis strategy deserve more attention. The advancement of science
newgenprepdf

xvi

xvi  Foreword to the Chinese Edition

relies on new ideas, theories, methods and strategies constantly proposed by
researchers on the basis of their practice. The life of science consists in innov-
ation on the basis of seeking truth. This serves as my Foreword.
Lu Jianming
December 29, 2001
 1

1 
Introduction

1.1  The questions that must be answered to explore Chinese

parts of speech
The issue of parts of speech is fundamental for grammar; the study of the
grammar of a language requires the classification of its parts of speech.
However, the issue has never had a satisfactory solution; many questions
have not been answered and remain discrepant and controversial. The major
questions are as follows:

1. Does Chinese have parts of speech? The question was debated in the
1950s, and in the 1990s some scholars questioned the existence of Chinese
parts of speech.
2. What is the nature of a part of speech? Currently, it is universally accepted
that its nature is distribution, but we cannot identify the grammatical
function of any part of speech that is internally universal and externally
exclusive. Can we say that the essence of a part of speech belongs to some
other factors besides distribution?
3. What are the bases and criteria for classifying parts of speech?
4. How can parts of speech be determined at an operation level? With what
standards should we classify various Chinese parts of speech? In the past,
we relied on our language sense to identify the distribution norms for
the parts of speech existing in our mind and did not provide proof for
why we should identify these norms. We might even have skipped over
the distinctions between some parts of speech. Is there any method for
determining the distinctions between parts of speech according to the
distribution of words before they are classified into parts of speech?
5. How can we deal with conversional words? For example, is a word such as
研究 (study) a verb and concurrently noun? This book aims at answering
these questions. The author started to ponder these part-​ of-​
speech
questions in 1986. Zhu Dexi and Lu Jianming undertook the National
Seventh Five-​year (1986–​1990) Plan Key Social Science Research Project
named “Exploring Modern Chinese Parts of Speech”, in which the
author participated as a research team member. Later, he cooperated
2

2 Introduction
with the research team of the National Seventh Five-​year Plan’s Key
Natural Science Project named “Modern Chinese Grammatical
Information Database”, undertaken by the Institute of Computational
Linguistics at Peking University. The team members include Professors
Yu Shiwen and Zhu Xuefeng, and they examined the functions of over
30,000 words. During the Eighth Five-​year Plan, from 1991 to 1995, we
continued to collaborate with the Institute of Computational Linguistics
and increased the number of words and their functional examination
scope on the basis of the Seventh Five-​year Plan Research Project. The
research team members included Zhang Yunyun and Wang Hui. The
present book is based on the corpus of over 40,000 words examined,
written, revised and proofread by many scholars, including Zhu Dexi
and Lu Jianming, for more than a decade.

1.2  Difficulties in classifying Chinese parts of speech

For a long time, there have been no good solutions for the Chinese part-​
of-​
speech issue; the causes include both subjective and objective ones.
Subjectively, in the past, we talked much about theory and did little actual
work. As a result, we did not have a deep understanding of the real situation
of parts of speech. (This book is based on the examination of over 40,000
words and is supported by plentiful actual materials.) Objectively, the Chinese
language’s own characteristics make it difficult to classify its parts of speech.
These features include the following:

1. Words do not have formal tags or morphological changes. 2.  The

constituents of modern Chinese are mixed at different hierarchies
in different historical stages, making the grammatical functions and
syntactic rules of a word complicated. 3. In general, words are multi-
functional; this makes it more difficult to classify parts of speech
according to their grammatical functions when words have no mor-
phological changes. 4.  The difficulties in parsing words make us
unable to tell whether a certain usage belongs to a word or a word-​
formation. 5. It is difficult to judge the grammatical relations of a
syntactic structure. Without morphological change, it is difficult to
judge what function a word plays in its combinations; thus, it is dif-
ficult to use grammatical functions as criteria to classify parts of
speech. For example, there are different opinions on whether 决心/​
完成任务 (determined to complete a task) is a predicate-​object con-
struction or a modifier-​head construction.

To sum up, there is only one key cause for the difficulty of classifying
Chinese into parts of speech, namely Chinese has no morphological changes.
Fortunately, previous and current scholars’ work paves a solid foundation
for us. Based on past studies, the book will explore further in detail how to
classify parts of speech in Chinese that has no morphological change.
 3

Introduction 3

1.3  An overview of studies of Chinese parts of speech

1.3.1  The formation and development of the framework of parts of speech

According to Yao Xiaoping (1999), Georg von der Gabelentz, a German who
wrote Chinesische Grammatik, was the first to establish the complete Chinese
grammar system. He classified Chinese notional words into nine parts of
speech: noun, some relative (roughly equivalent to locative), numeral, adjec-
tive, verb, negative, pronoun, onomatopoeia and interjection. The Chinesische
Grammatik, however, did not have much influence on China’s academic
circle; rather, what greatly influenced the later Chinese grammar studies is
Ma’s Grammar, by Ma Jianzhong (1898). Inspired by Western grammar,
he classified the words of classical Chinese into the following nine parts of
speech: noun, pronoun, verb, static word, state word, preposition, conjunc-
tion, auxiliary word and interjection. This system provided the basic pattern
of Chinese parts of speech, and did not change greatly after that. After quite
a long time, only numerals are separated from static words; modal particles
were separated from auxiliary words; measure words and onomatopoeias
were added.
Li Jinxi’s A New Chinese Grammar (1924) was the first of its kind because
it was influential and studied systematically modern Chinese grammar.
It classified Chinese words into five major parts of speech and nine basic
classes:  entity word (noun and pronoun), narrative word (verb), distinctive
word (adjective and adverb), relative (preposition and conjunction), modal
word (auxiliary word and interjection). Its basic classes, whose names are
slightly different, are the same as those in Ma’s Grammar.
Lv Shuxiang’s Chinese Grammar Synopsis (1942–​1944) classified modal
particles (the scope is larger than the modal particles referred to generally,
including the modal particle referred to subsequently, the modal adverb
and interjection), while Wang Li’s Modern Chinese Grammar (1943, 1944)
classified words into numeral and modal particle.
Modern Chinese Lecture by Ding Shengshu et  al. (1952, 1953)  classified
measure word and onomatopoeia (including interjection), and pointed out
the special classes of time word, place word and locative.
The Temporary Chinese Teaching Grammatical System (1956) classified
Chinese words into 11 parts of speech: noun, measure word, pronoun, verb,
adjective, numeral, adverb, preposition, conjunction, auxiliary word and inter-
jection. The system was revised into The High School Teaching Grammatical
System Synopsis (1984) and onomatopoeia was added to it. This system of 12
parts of speech had a rather big influence and has been adopted by most of
the current dictionaries that tag parts of speech, except for a few dictionaries
that separate modal particles from auxiliary words.
Zhao Yuanren’s Spoken Chinese Grammar (1968) has distinctive word as
a part of speech, which includes demonstrative distinctive word (this, that),
fractional distinctive word (each, all, another), quantitatively distinctive word
(one, ten) and measurable distinctive word (integral, half, numerous).
4

4 Introduction
Chen Wangdao’s A Synopsis of Grammar (1978) classifies copula, copula-
tive and demonstrative.
Zhu Dexi’s Grammar Handouts (1982b) separated distinctive words
from adjective; place word, locative and time word from noun; and modal
particles from auxiliary words, totaling 17 parts of speech. Modern Chinese
(1993), produced by the Department of Chinese Language at Peking
University, classified state adjectives into state word, but returned place
word, locative and time word to the category of a noun, totaling 15 parts
of speech.
On the basis of the 17 parts of speech proposed by Zhu Dxi (1982b)
and the 15 parts of speech proposed by Modern Chinese, compiled by the
Department of Chinese Language at Peking University, the present book
adds numeral, measure word and demonstrative, for a total of 20 parts of
speech (see Section 1.5).

1.3.2  Two core issues for studying Chinese parts of speech

The first core issue is the relationship between part of speech and syntactic
constituent; the second core issue is the criteria for classifying parts of speech.
Almost all controversies stem from the two issues, and the reason why they
trigger controversies has something to do with the fact that Chinese has no
morphological change in the strict sense.

1.3.2.1  The relationship between part of speech and syntactic constituent

Because Chinese has no morphological change, which, therefore, cannot be
used to classify parts of speech, we can only use the grammatical functions of
a word. One of the characteristics of Chinese is that a word can function as
many syntactic constituents. For these reasons, we have the questions of the
relationship between part of speech and syntactic constituent: how does a part
of speech correspond to its syntactic constituent? Specifically, there are three
questions: (1) Does a part of speech correspond to its syntactic constituent
directly or indirectly? (2) Is the correspondence one-​to-​one or complicated?
(3) What syntactic constituents can a part of speech function as? The methods
for answering the first two questions include the double-​line system, the one-​
line system and the three-​layer system.
The double-​line system: This refers to the part of speech level and the syn-
tactic constituent level. However, because a Chinese word is generally multi-
functional, there are slots between part of speech and syntactic constituent.
To resolve the slots, scholars use various methods, including the change in
part of speech and the addition of functions.
The change in part of speech establishes the simple one-​to-​one corres-
pondence between part of speech and syntactic constituent. For example, a
noun should function as subject or object; a verb should function as predi-
cate or complement; an adjective attributive; an adverb adverbial. If a word
 5

Introduction 5
occurs at the position of different characteristics, it may be deemed to have
changed its part of speech. Most of earlier works on Chinese grammar
adopted this method. For instance, Ma’s Grammar proposed the theory that
a part of speech functions in the guise of various syntactic constituents. If
a verb occurs at the position of subject or object, then it functions in the
guise of a noun; if it occurs at the position of attributive, then it functions
in the guise of an adjective (static word). Chen Chengze’s An Initial Start
of Chinese Grammar (1922) and Jin Zhaozi’s Studies of Chinese Grammar
(1922) proposed the flexible use of parts of speech. Li Jinxi’s A New Chinese
Grammar (1924) proposed distinguishing between parts of speech according
to sentences. All these works use the change in part of speech to resolve the
slot between part of speech and syntactic constituent. The Temporary Chinese
Teaching Grammar System (1956) proposed the “nominalization theory”. In
sentences such as 他的来使大家很高兴 (His coming makes everybody happy)
and 狐狸的狡猾是很出名的 (The slyness of a fox is famous), 来 (coming) and
狡猾 (slyness) do not contain the meaning of substantial action or properties
but regard them as “things”. They lose the properties, or some properties, of
their verb and adjective forms, while acquiring some properties of a noun.
We call this the nominalization of an adjective. “Nominalization” is a vague
expression and does not clearly say that there is a change in part of speech,
but in essence it is the same as the loaned part of speech and its flexible use.
Because most Chinese words can function as many syntactic constituents
without morphological change, although the method of changing parts of
speech assigns certain syntactic constituents to a part of speech, it makes
Chinese parts of speech changeable (Zhu Dexi, 1982c). As a result, words do
not have their definite parts of speech, and we may even reach] the conclusion
that “there is no part of speech without a sentence”.
The function addition method holds that there is a complicated cor-
respondence between part of speech and syntactic constituent. A  part of
speech is multifunctional: so long as the meaning of a word does not change,
the word belongs to the same part of speech even if it occurs at different
positions. Although this method causes a word to have its definite part of
speech, it causes a part of speech to have no definite syntactic constituents.
In essence, it openly admits the slot between part of speech and syntactic
constituent instead of denying it. During the debate on grammar innovation
at the end of the 1930s and the early 1940s, Fang Guangdao proposed clas-
sifying parts of speech according to a generalized morphology, while Chen
Wangdao proposed classifying them based on integrated structural functions.
In doing so, they used the function addition method. Lv Shuxiang and Zhu
Dexi’s Talks on Grammar and Rhetoric (1951) proposed that if the meaning of
a word does not change, then its part of speech does not change. Later, what
was meant by Zhu Dexi’s saying that a part of speech is multifunctional, was
narrowed down slightly. 批评 (criticism) and 研究 (study) and so on, which
are directly modified by nouns and function as the quasi-​objects of predi-
cate verbs such as 做 (do), 进行 (carry out), 加以 (conduct) and 有 (have) are
6

6 Introduction
considered nominal verbs. In this case, they are considered to be nominal. But
Zhu Dexi still thought that 出版 (publication) in 这本书的出版 (the publica-
tion of the book) and 去 (going) in 去是应该的 (Going is obliged) have the
properties of a verb.
The one-​line system: This notion was put forward by Fu Donghua in
the 1930s. Because a Chinese word is multifunctional, the simple cor-
respondence between part of speech and syntactic constituent must use
loaned parts of speech. He held that it would be better just to presume that
it is impossible to classify words into parts of speech and said that “if a
word is not used in a sentence, there is no way to know its part of speech”.
Therefore, parts of speech should be combined with sentence constituents.
For example:

张三 (Zhang San)  作 (writes)  文 (an essay)

subject-​noun predicate object-​noun

Actually, this view is a natural development of “distinguishing between

parts of speech according to a sentence” and “there being no parts of speech
without a sentence”. During the debate on Chinese parts of speech in the
1950s, Gao Mingkai also thought that Chinese had no parts of speech mainly
because they should be classified according to the morphology of a word.
Chinese has no morphological change and therefore has no parts of speech.
Later, he changed his point of view and thought that “the morphological
changes of a word, its combination capacity and syntactic functions etc. are
all the external marks of its part-​of-​speech meaning”. Chinese has no mor-
phological change, and its parts of speech can only be classified according
to the use of a word. But because Chinese parts of speech are multifunc-
tional, “a word can be used simultaneously as noun, adjective and verb”
(meaning a notional word can appear at the positions of subject, predicate
and attributive). Therefore, every notional word belongs to several parts of
speech, implying that Chinese has no parts of speech (Gao Mingkai, 1960).
Xu Tongqiang (1994b) thought that the parts of speech in Indo-​European
languages have a one-​to-​one correspondence to sentence constituents, thus
causing the necessity to classify them. Now that one Chinese part of speech
corresponds to many sentence constituents, this implies that Chinese does
not have parts of speech at all. With regard to the theory that Chinese has
no parts of speech, Zhu Dexi (1960) thought that, although the classifica-
tion of parts of speech according to sentence constituents takes grammatical
functions into consideration, “because selectional criteria are too loose and
the method is too simple, parts of speech cannot be classified this way, instead
reaching the conclusion that a word has no definite parts of speech. The entire
error of this method lies in the fact that it assumes that there is a one-​to-​one
correspondence between part of speech and sentence constituent . . . but
as a matter of fact, the relationship between part of speech and sentence
constituent is complicated”. If the one-​line system is taken to the extreme,
 7

Introduction 7
however, it cannot solve Chinese syntactic problems and instead causes more
troubles, and few people accept this theory.
The three-​layer system: This system comes into existence by combining the
change in part of speech with the addition of functions. It adds an inter-
mediate layer between part of speech and syntactic constituent so as to close
the slot between the two, assuming that there are some changes in the prop-
erties of a word at its intermediate layer at the syntactic position of different
characteristics. This thinking not only maintains the point of view that if the
meaning of a word does not change, then its part of speech does not change,
but also explains that the same part of speech can appear at different syntactic
positions. Scholars who use the three-​level system have very different opinions
and therefore quite different effects.
Wang Li (1943, 1944) and Lv Shuxiang (1942, 1944)  borrowed Otto
Jespersen’s three-​level theory and set up the intermediate word level, as Lv
called it. A word level is classified into the first level (the words at the positions
of subject and object), the secondary level (the words at the positions of predi-
cate and attributive) and the last level (the words at the positions of adverbial
and complement). Parts of speech are classified according to the meaning of
a word and are unchangeable, and when a word is used at different syntactic
positions, its level changes. But word level is a vague concept and distinguished
according to the importance of a word in its combination. Therefore, it is more
correct to say that it classifies the syntactic constituents of a word. There is no
substantial distinction between saying that a word can function as different
word levels, and that it can function as different syntactic constituents. It is
not effective to use the three-​level theory to connect a part of speech with a
syntactic constituent in order to close the slot between the two.
Chen Aiwen (1986) assumed that there are two types of part-​of-​speech con-
cept. The first type is equivalent to the classification between part of speech
and mental structure, and has the properties of noun, verb and adjective
respectively; the second type is equivalent to the classification between part
of speech and objective world, and has the properties of a basic noun, verb
and adjective respectively. The first type of part of speech is actually classified
according to grammatical properties; the second type is actually classified
according to the meaning of a word. The first type may change with different
usages; the second type does not. For example, the word 出版 (publish) has
the properties of a verb in 同意出版 (agree to publish), but has the properties
of a noun in 图书出版 (book publication). In both cases, they are basic verbs.
Chen Aiwen proposes the two types of part-​of-​speech concept for the purpose
of understanding the correspondence between part of speech and syntactic
constituent. He is aware that adherence to the concept that the meaning of
a word does not change and then its parts of speech do not change cannot
explain that its properties may change at different positions. Therefore, it is
necessary to establish the two types of part-​of-​speech concept and form the
three-​layer system of “part of speech (X basic word) –​properties of a word –​
syntactic constituent”, taking word properties as the intermediate layer that
8

8 Introduction
connects a part of speech with a syntactic constituent. This approach actually
classifies parts of speech according to the meaning of a word and then decides
its properties, only making a clear-​cut classification between the two types of
part-​of-​speech concept but making no substantial difference from the change
in parts of speech in the double-​line system.
Xiao Guozheng (1991) and Xiang Mengbing (1991) thought that some
verbs and adjectives at the positions of subject and object have denotation,
namely using their expressional functions to denote the change in their prop-
erties at different positions.
This book proposes that the nature of a part of speech is the type of expres-
sional function in the large categories of substantive word, predicate word
and modification word. According to the two levels of expressional function,
we can classify parts of speech into the part of speech at the lexical level and
that at the syntactic level. The part of speech at the lexical level is the inherent
property of words and expressions that are labeled in a dictionary; the part
of speech at the syntactic level is generated by using words and expressions
and governed by syntactic rules. Most of the covert and flexible uses of a
part of speech pointed out by previous scholars belong to changes in parts
of speech at the syntactic level. The three-​layer system of “part of speech at
the lexical level –​part of speech at the syntactic level –​syntactic constituent”
is thus formed. There are two situations in which a word appears at different
syntactic positions: (1) a word has its properties at multiple lexical levels, for
example, 出版 (publish) has the properties of a verb in 出版两本书 (publish
two books), but has the properties of a noun in 图书出版 (book publication);
(2) there are some changes in part of speech at the syntactic level, for example,
出版 (publish) still has the properties of a verb at lexical level in 这本书的出版
(the publication of the book) but has the properties of a noun at the syn-
tactic level. The changes in part of speech at the syntactic level essentially
have covert core constituents.
Although both the three-​level theory and the viewpoints of this book on
the relationship between part of speech and syntactic constituent belong to
the three-​layer system, they are different. The intermediate layer “word class”
established between part of speech and syntactic constituent by the three-​
class theory is not identical with the intermediate layer “the properties of
a word at the syntactic level” established by this book. The word “class” is
classified into the first class (subject and object), the secondary class (predi-
cate and attributive) and the last class (adverbial and complement). The parts
of speech at the syntactic level refer to extrinsic expressional functions such
as reference, statement and modification. Subject and object generally belong
to reference; predicate belongs to statement. In this regard, the two are iden-
tical. But attributive corresponds to modification; complement corresponds
to statement. In this regard, the two are different. The root cause for the diffe-
rence between the two is their own different starting points. The three-​class
theory emphasizes the importance of a syntactic constituent, but an expres-
sional function emphasizes a word’s method of meaning expression in use.
 9

Introduction 9
The third question is to which syntactic constituent a part of speech cor-
responds. This is the most difficult question. If the question is answered, the
specific criteria for classifying parts of speech will be available. We shall dis-
cuss them in Chapter 6.

1.3.2.2  Criteria for classifying parts of speech

The criteria for classifying parts of speech include the meaning of a word, its
morphology and its grammatical function.
In one case, parts of speech are classified purely according to the meaning
of a word. Two important grammar works, Chinese Grammar Synopsis by
Lv Shuxiang and Chinese Modern Grammar by Wang Li in the 1940s did so
without prior consultation. There are two lethal problems for the classifica-
tion of parts of speech according to the meaning of a word: (1) its meaning
does not correspond completely to its grammatical properties, and the parts
of speech classified according to the meaning are not closely related to syntax;
(2) the meaning of a word itself is not clearly observable and therefore not
operable. In another case, the parts of speech are classified according to the
meaning of a word, but its part-​of-​speech conversion is judged according to
the position of a word in a sentence, like Ma’s Grammar by Ma Jianzhong
and A New Chinese Grammar by Li Jinxi. This approach actually creates two
systems of parts of speech. One system is classified according to the meaning
of a word; the other system is classified according to syntactic constituent.
Since the second system can be classified according to syntactic constituent,
classifying the first system according to the meaning of a word is then
unnecessary, while classifying the system according to syntactic constituent
establishes a one-​to-​one correspondence between part of speech and syntactic
constituent. Thus we can reach the conclusion that a word has no definite
parts of speech.
Some Western languages have plentiful morphological changes, which can
be used to classify their parts of speech, but Chinese has no morphological
change, which, therefore, cannot be used to classify its parts of speech. In
the early 1950s, Gao Mingkai regarded morphology as the nature of a part
of speech and arrived at the conclusion that Chinese has no parts of speech.
Chen Chengze (1922) pointed out that “a part of speech cannot be decided
according to the form of a word or the word itself and can only be decided
according to the position of the word in a sentence”. Jin Zhaozi (1922)
observed that “Chinese characters do not show word properties and rely on
their positions to show them”. That is to say, parts of speech are decided
according to the grammatical functions of a word. He did not come up with
the criteria and operational procedures for classifying parts of speech. Because
he thought there was a one-​to-​one correspondence between part of speech
and syntactic constituent, he had to use the method for flexible use of parts
of speech, not fixing their use. Lu Zhiwei proposed two types of structural
relationship in his Single-​Syllable Vocabulary in Beijing Dialect (1938):  (1)
10

10 Introduction
attached relation:  red flower (attachment + thing attached); (2)  approxima-
tion relation: eat a meal (approximator + thing approximated). The two types
of structural relationship provide three kinds of basic parts of speech: a noun
that occupies the position of the thing attached or approximated, a verb
that occupies the position of approximator and an adjective that occupies
the position of attachment. This is equivalent to classifying parts of speech
according to sentence constituents. Because a word is multifunctional, if this
is strictly implemented, then certainly it will have no definite parts of speech.
For the first time, the discussion on grammar innovation explicitly proposed
classifying parts of speech according to the overall grammatical functions
of a word. Fang Guangdao proposed classifying parts of speech according
to generalized morphology, namely according to the combination capacity
of a word. Based on Ferdinand de Saussure’s syntagmatic and paradigmatic
relations, Chen Wangdao proposed classifying parts of speech of a word
according to its syntagmatic relation. But he just discussed this theoretically
and did not propose a set of operating criteria. In his Elementary Chinese
(1948), for the first time, Zhao Yuanren made a systematic use of grammatical
functions to classify parts of speech and proposed a set of operating criteria.
For example, a noun can be modified by a quantitative phrase; a verb can be
modified by “不 (not)” and followed by “le (了)”. In 1968, he performed a
systematic classification of Chinese parts of speech more entirely according
to grammatical functions. Zhu Dexi (1960, 1982b, 1985a) emphasized several
times that the essential basis for classifying parts of speech is the distribution
of a word, and that this is true not only to Chinese but also to other languages.
Due to the lack of morphological change, Chinese can only use the grammat-
ical functions of a word to classify its parts of speech. This came to be agreed
upon after discussions on Chinese parts of speech in the 1950s. However,
because of the complicated relationship between part of speech and gram-
matical function, how to identify the part of speech of a word according to its
grammatical functions becomes a prominent issue. The question, in essence,
is still the relationship between part of speech and syntactic constituent: Is
the correspondence between part of speech and syntactic constituent one-​
to-​one or complicated? If it is one-​to-​one, then nearly all the members of
noun, verb and adjective are conversional words. If it is complicated, then
how complicated? One point of view is that the same generalization word
that occurs at any position belongs to the same part of speech. In this way,
there is no conversional word whose meaning is the same. The two extreme
points of view are now scarcely acceptable. Most people think something in
between: there is no one-​to-​one correspondence between part of speech and
syntactic constituent; the same generalization word that occurs at any pos-
ition may have different parts of speech.
The subsequent question is how to determine the correspondence between
part of speech and grammatical function. Once the question is figured out,
the criteria for classifying parts of speech will be worked out, because the
question is the key to the current studies of Chinese parts of speech. However,
 11

Introduction 11
in most of the studies of Chinese parts of speech since the 1980s, parts of
speech have been classified subjectively as a first step, and the criteria for clas-
sifying them have been found as a next step. Due to the lack of clear dem-
onstration, it is difficult to achieve agreement among various scholars on
classifying, grouping and annexing the parts of speech of a word.
To work out the relationship between part of speech and grammatical
function, the Chinese grammar circle has tried various approaches. One
approach is to regard a part of speech as a prototype category or a fuzzy cat-
egory. Mo Pengling and Wang Zhidong (1988) proposed the fuzzy clustering
analysis method. Shi Youwei (1994) proposed the flexible approach to Chinese
parts of speech and tried to calculate the membership degrees of words to
deal with the relationship between part of speech and grammatical function.
Yuan Yulin (1995, 1998) used the prototype theory to deal with the relation-
ship between distribution and part of speech. Lu Yingshun (1998) also tried
to calculate the approximation degrees of words to determine the attribu-
tion of the parts of speech he analyzed. The problem with these methods is
that there is no way to demonstrate the determination of a prototype, thereby
being difficult to classify parts of speech.

1.3.3  The question of conversional words

There is a heated discussion on this question. The following three factors are
involved in whether a word is a conversional word: (1) the identity of a word;
(2) whether a word has the properties of several parts of speech or its part of
speech is multifunctional; (3) what strategy is used to classify parts of speech.
The general view on the relationship between word identity and a
conversional word is that the different senses of a polyseme or the different
usages of the same generalization word belong to different parts of speech and
are reckoned as a conversional word, whereas the homonyms whose meanings
are not associated belong to different parts of speech, are not reckoned as
conversional words, and are treated as two independent words. Xu Shu (1991)
and Lu Jianming (1994) thought that only the same generalization word that
belongs to different parts of speech can be called a conversional word, whereas
different generalization words that belong to different parts of speech are two
words belonging to utterly different parts of speech. As a matter of fact, it is
only a matter of name to regard the polyseme that belongs to different parts
of speech as a conversional word or only to regard the same generalization
word as a conversional word. The key question here is whether these words
have properties of only one or several parts of speech. In our view, the term
“conversional word” should be used in its broad sense. To distinguish two cases,
the conversional word whose meaning is identical is called a homonymous
conversional word, and a polyseme that belongs to different parts of speech is
called heteromorphic conversional word. The conversional word in its narrow
sense refers to homonymous conversional word; the generalized conversional
word includes both homonymous and heteromorphic conversional words.
12

12 Introduction
The difficulty lies in determining the identity of a word. If different gener-
alization words have the properties of different parts of speech, they must
be regarded as different parts of speech. The same generalization word that
has the properties of several parts of speech need not necessarily be treated
as a conversional word. Whether a word is treated as a conversional word
has something to do with what strategy is used to classify parts of speech.
Because of the discrepancy in judging a word’s identity, there is a difference in
the range of heteromorphic conversional words. For example, if 研究 (study)
is not identical in 研究问题 (study a problem) and 社会研究 (social study),
then it should be treated as two parts of speech, namely a verb and concur-
rently a noun. Likewise, if 木头 (wood) is not identical in 木头断了 (the wood
is broken) and 木头桌子 (wooden desk), then it should be treated as a noun
and concurrently a distinctive word (see Yang Chengkai, 1991). There is not
much discussion on how to determine word identity. Currently, we mainly rely
on our language sense to do this and have no strict operational procedures.
Although in most cases we can reach agreement, rather big discrepancies on
some difficult questions still remain.
Another question is whether a word at different positions has the prop-
erties of several parts of speech or whether its part of speech is multifunc-
tional. The question is related to the correspondence between a part of speech
and a grammatical function. Zhu Dexi (1982b) thought that 调查 (investi-
gation) in 社会调查 (social investigation) and 进行调查 (conduct investiga-
tion) is different in property from 调查 (investigate) in 调查问题 (investigate
a problem). The latter can take an adverbial and a complement and so forth,
having the properties of a verb, whereas the former can only take an attribu-
tive, having the properties of a noun. Hu Mingyang (1996b) also performed
a detailed examination of similar issues. This book suggests that the com-
patibility of grammatical functions can be used to determine the relationship
between grammatical function and a part of speech, namely in what case a
word has the properties of several parts of speech and in what case a part of
speech is multifunctional.
Zhu Dexi (1982b, 1985a) and Lu Jianming (1984) discussed what strategy
should be used to classify parts of speech. We shall discuss this in detail in
Chapter 7.

1.3.4  Operational procedures for classifying parts of speech

The study of Chinese parts of speech before the 1980s rarely discussed the
concrete operational procedures for classifying parts of speech. Since the
1980s, a good many scholars have paid attention to these processes. In his
Re-​investigation of Chinese Part-​of-​speech Classification Issues (1980), Shi
Anshi tried to classify parts of speech layer by layer, using only one cri-
terion one time. The upper layer of criteria is a bit broader than the lower
layer. This method offers progress in making criteria for classifying parts of
speech stricter and enhancing the operability of classifying them. Lu Jiawen
 13

Introduction 13
(1982), Lu Chuan (1991), Lu Jianming (1994) and Gao Gengsheng (1995)
also proposed operational procedures for classifying parts of speech layer by
layer. The layer-​by-​layer classification reveals the hierarchical nature of parts
of speech in more detail than the past major classification into only a notional
word and a functional word. The present book not only uses the hierarchical
classification method but also identifies the criteria for classifying each part
of speech layer by layer.
Xing Fuyi (1981) concentrated his discussion on the grouping of words and
proposed grouping methods such as direct decision, exclusion and analogy.

1.3.5  Other studies

In addition to the above, scholars also studied Chinese parts of speech from
various angles. Mo Pengling and Shan Qing (1995) did a statistical calcula-
tion of the grammatical functions of noun, verb and adjective in their actual
corpus and provided valuable statistical data. Chen Ningping (1987) used a
continuous statistical model to analyze the drift of verb to noun. Li Yuming
(1996) and Zhang Bojiang (1994) tried to explain the properties of distinctive
words and the flexible use of parts of speech, starting from the spatiality of a
noun and the temporality of a verb. Shen Jiaxuan (1997) used Croft’s tagging
theory to discuss the major functions of Chinese adjectives. These studies that
apply new theories are very heuristic.

1.4  Outline of chapters

The book mainly covers two parts.
Part 1 (Chapter 2) discusses the prerequisites for classifying parts of speech,
including word parsing, word identity and the internal hierarchies of modern
Chinese. To parse words, we can use their two basic characteristics –​standby
application and independent application –​to distinguish between word and
phrase. In terms of standby application, words should have finite length and
number; as an independent application unit, the rules for combining words
are different from those for combining morphemes, and the difference should
be used to parse words.
Modern Chinese is an inhomogeneous system mixed with different hier-
archical constituents. Analyze its internal hierarchies, and we can deal with
the system well. It mainly has the historical hierarchy and the field hier-
archy. A  language may have different grammatical systems at different his-
torical and field hierarchies; the grammatical functions of its words may also
be different, and the criteria for classifying their parts of speech should be
different too. The classification of modern Chinese parts of speech should
consider these different hierarchies and treat the constituents at different hier-
archies differently.
The determination of word identity involves whether a word has a certain
grammatical function, thereby influencing its classification results. Therefore,
14

14 Introduction
it is one of the issues to be solved first in order to classify parts of speech.
To determine the word identity, we need to distinguish between constituent
meaning and structural meaning, lexical transfer reference and syntactic
transfer reference, lexical transfer reference and lexical self-​reference, word-​
formation, morphology and syntax.
Part 2 (Chapters 3 through 7) discusses the theoretical issues for classifying
parts of speech. Chapter 3 discusses the possibility and purpose of classifying
parts of speech and points out that words are combined not randomly but in a
certain order. The orderliness indicates that a grammatical position has selec-
tional restrictions on words, and that different grammatical positions allow
different words to enter. This shows that we can classify words into different
parts of speech according to their different properties. From the perspective of
natural taxonomy, parts of speech are generalized from syntactic structures,
not artificially postulated beforehand for the sake of grammar. The purpose
of classifying parts of speech is to reveal the properties of a word and estab-
lish its general reference system. In other words, we think that language is not
desultory but systematically organized, and has its natural order independent
of linguists. Words have their own positions in such a natural order.
Chapter 4 discusses the nature of parts of speech. Since US descriptive lin-
guistics came into being, it has been generally believed that a part of speech
is of the distribution type. Words that have the same distribution form a part
of speech. The “distribution” here has three possibilities: (1) individual dis-
tribution, (2) population distribution, (3) fractional distribution. No matter
in what sense the distribution is, the viewpoint that a part of speech is of
the distribution type is not justifiable. The prominent issue is that we cannot
use the distribution characteristics themselves to answer why we select these
distribution characteristics instead of others as the criteria for classifying
parts of speech. If factors other than distribution characteristics are not
considered, we cannot classify parts of speech purely according to distribu-
tion characteristics. In our opinion, the nature of a part of speech is not of the
distribution type. The nature of classifying parts of speech according to dis-
tribution is to classify them according to the selectional restriction of gram-
matical positions on words. There must be some bases used as conditions
for selectional restrictions. Then, what are the bases for such selectional
restrictions? We think that the expressional function and the semantic type
of a word are the intrinsic causes for constraining its distributions. Like mor-
phological change, distribution is only the extrinsic exhibition of the proper-
ties of a part of speech. The intrinsic basis for distinction among words, for
example, the large categories of parts of speech such as a substantive word,
a predicate word, a substantive modification word, a predicate modification
word, is actually the distinction between such expressional functions as ref-
erence, statement, substantive modification and predicate modification. The
distributional difference and morphological difference among parts of speech
are simply the external exhibitions of differences in expressional functions.
Expressional functions can be classified into intrinsic expressional function
 15

Introduction 15
and external expressional function. We can also classify parts of speech into
the parts of speech at the lexical level and the syntactic level. Syntactic com-
bination is a series of expressional functions rather than a series of parts of
speech. The selectional restriction of a grammatical position on a word is
based on its expressional function, which is the nature of parts of speech such
as a substantive word, a predicate word and a modification word.
The essence of basic parts of speech such as nouns, adjectives, state words,
measure words, locatives, distinctive words, numerals and measure words is
the semantic type that combines with grammatical function. Only when the
semantic type of a word is in agreement with its difference in syntactic distri-
bution is it necessary to classify its parts of speech. Therefore, at the level of
basic parts of speech, the essence of a part of speech is the type that combines
the semantic type with the grammatical function.
The expressional function is a word’s pattern of meaning expression; the
semantic type is the categorized semantic meaning. The two belong to gram-
matical meaning; thus, we can say that grammatical meaning is the intrinsic
cause for restricting the distribution of a word.
Chapter 5 discusses criteria for classifying parts of speech, which must sat-
isfy the three conditions: reflecting the properties of a part of speech, being
observable and being comprehensive. Theoretically speaking, a word’s mor-
phological change, expressional function, grammatical meaning or intrinsic
expressional function all can be used as criteria for classifying parts of speech.
But in Chinese, only the grammatical function (a word’s distribution) can sat-
isfy these conditions and be used as criteria for classifying parts of speech.
The grammatical position occupied by a word is called its distribution. The
capability of a word to occupy a certain grammatical position is the word’s
grammatical function. Two factors define a constituent’s distribution: (1) the
grammatical relationship between the immediate constituent and its gram-
matical role; (2) the larger environment in which the structural whole made
up of the two immediate constituents can appear. The grammatical function
has different generalization levels. A word’s capability of occupying the gram-
matical position defined in the environment of evaluating it and its parts of
speech is a rather specific grammatical function; the capability of functioning
as a syntactic constituent is a rather abstract grammatical function. We can
use a word’s distribution characteristics to infer the properties of its parts of
speech. The grammatical meaning of a word is the main intrinsic cause for
restricting its distribution and basically determines its grammatical distribu-
tion. Although distribution is not the nature of a part of speech, the words
that belong to the same part of speech have a roughly similar distribution.
Like morphological change, distribution is only an extrinsic exhibition of
grammatical meaning. That is to say, there is a “reflection-​exhibition” rela-
tionship between distribution of a word and its grammatical meaning: distri-
bution reflects the grammatical meaning, which in turn exhibits distribution.
We can use the properties of parts of speech reflected by a word’s distribution
to classify its parts of speech.
16

16 Introduction
However, the relationship between distribution and a part of speech is
complicated, as exhibited in the following: (1) what determines the properties
of a part of speech is not just the distribution of a word; its lexical and prag-
matic meanings, word-​formation, prosodic characteristics, and other factors
may also influence distribution; (2)  some grammatical positions reflect the
same properties, and then the above-​ mentioned distributional difference
cannot reflect the difference in properties; for example, 很 (very) ~ and 极
(extremely) ~ have the same requirements for words to enter into their gram-
matical positions, but the difference in the two functions of the two words
does not reflect different properties of their parts of speech; (3) it is possible
that some grammatical positions allow several parts of speech to enter. For
example, either a noun, a verb, an adjective or a distinctive word and so on
can appear at the position of a subject; either a verb, an adjective, a state
word or a noun and so on can appear at the position of a predicate. Therefore,
the properties of a part of speech cannot be inferred entirely according to
word distribution. We should not only rely on word distribution to infer the
properties of a part of speech but also use some other means to eliminate
the factors that influence the nongrammatical meanings of word distribu-
tion and the interference caused by an incomplete correspondence between
properties and grammatical position. The correspondence between a part of
speech and word distribution should be sought from the complicated rela-
tionship between the properties of a part of speech and word distribution,
thereby reasonably and justifiably classifying parts of speech according to
the distribution.
Chapter 6 discusses how to classify parts of speech according to distribu-
tion. Not all differences in distribution reflect the difference in the proper-
ties of a part of speech; different functions of different parts of speech have
different values for classifying them: some functions are distinctive; some are
not. Therefore, we cannot treat equally all the functions of a part of speech
and only select those distinctive functions to classify parts of speech. We can
use the compatibility of functions and the relevant rules to determine the
values of functions for classifying parts of speech, thereby identifying the
functions that distinguish between word properties and selecting among them
the criteria for classifying parts of speech. The compatibility of functions
refers to the property that the same group of words shares two or more
grammatical functions, for example, the word that can function as a subject
can also function as an object and vice versa. But the compatibility of some
other functions is extremely small, for instance, 不 (not) ~ and (numeral) ~.
Functions that have a rather big compatibility often reflect the same selec-
tional restriction of the two different grammatical positions on words and
also the common properties of a part of speech; thus, they are equivalent-​
value functions. The functions that have a rather small or no compatibility
reflect different properties of a part of speech and are generally hetero-​value
functions. An equivalent-​value function is transferable. If the equivalent-​
value functions of a language are gathered into clusters, then we come up with
 17

Introduction 17
the correspondence between a part of speech and a grammatical function; an
equivalent-​value function cluster represents a part of speech.
Chapter 7 discusses conversional words and nominalization. The same
generalization word that has the properties of several parts of speech is not
necessarily treated as a conversional word in its narrow sense. Whether it
is treated as a conversional word has something to do with the strategy for
classifying parts of speech. To classify Chinese parts of speech, we mainly
use the two strategies: (1) the homogeneous strategy; namely, starting from
the properties of a part of speech, all the properties of the part of speech
of a generalization word are treated equally to create the one-​to-​one cor-
respondence between a part of speech thus classified and its properties. If
a generalization word has the properties of several parts of speech, then
we treat it as conversional word. (2) The priority homomorphic strategy;
namely, starting from a generalization word, we give priority to the prop-
erties of certain parts of speech and do not treat as a conversional word
the word that has the properties of several parts of speech. Instead, we
treat it as having a priority part of speech, thus creating a correspondence
between a part of speech and the properties of a priority part of speech. In
fact, different strategies for classifying parts of speech cause a good many
controversies in Chinese parts of speech. Strategies are not right or wrong
but only good or bad. Which strategy for classifying parts of speech is the
best depends on specific circumstances. To select a strategy for classifying
parts of speech, the following two factors should be taken into consider-
ation:  (1) the total numbers of parts of speech, and conversional words
in their narrow sense should be as few as possible; (2) the fewest possible
syntactic rules:  there should be as few as possible grammatical functions
of different words in the same part of speech. The two factors are exactly
contradictory: the fewest number of parts of speech is in conflict with the
fewest number of syntactic rules and vice versa. Therefore, in selecting a
strategy for classifying parts of speech, we should consider both factors,
reducing the numbers of both parts of speech and syntactic rules to the
lowest degree as possible, specifically:

1. If there are large numbers of words that have two or more parts of speech
or can be controlled by rules, then we use the priority homomorphic
strategy; if not, we use the homogeneous strategy. If the homogeneous
strategy is used for the large number of words that have two or more parts
of speech, a massive number of conversional words may appear. Although
there are a few syntactic rules, there are too many parts of speech, and a
massive number of conversional words have to be identified before syn-
tactic processing. If something goes wrong with the conversional word
identification, something may also go wrong with syntactic processing.
2. If the priority homomorphic strategy is used for a few words that have
two or more parts of speech, there will be too many syntactic rules,
although there are not too many parts of speech. This is because more
18

18 Introduction
syntactic rules are needed to deal with the use of a conversional word.
The use of the homogeneous strategy may reduce the numbers of not
only syntactic rules but also parts of speech. Therefore, in general, we
should use the homogeneous strategy. However, if words that have two or
more parts of speech can be controlled by rules, then the priority homo-
morphic strategy is desirable.

Based on this justification, according to the priority homomorphic strategy,

Chinese words that have the properties of both a verb and a noun should be
treated as a verb instead of a conversional word.
There are two types of nominalization: (1) the reference of external expres-
sional function and the nominalization at the syntactic level; nominalization still
has the general characteristics of a verb or an adjective, and does not change its
properties at the lexical level; (2) some words that have the properties of a verb
(or an adjective) and a noun embody the properties of a noun at the position
of subject or object. They no longer have the general characteristics of a verb
when functioning as subject or object but those of a noun. We can reckon that
their intrinsic expressional functions have changed; that is to say, the properties
of a part of speech at the lexical level have changed, but we do not treat them as
conversional words according to the priority homomorphic strategy.

1.5  The methods and main conclusions in the book

1.5.1  Methods and main viewpoints in the book

In our opinion, the nature of a part of speech is not distribution. Therefore,
it is difficult to succeed in any attempt to classify Chinese parts of speech
through seeking internally universal and externally exclusive distribution
characteristics. The large category of a part of speech (substantive words,
predicate words and modification words) belongs to the type of expressional
function; the basic category of a part of speech (nouns, measure words, verbs,
adjectives, state words, numerals and measure words …) is the type that
combines the semantic type of a word with its grammatical function.
To operate part-​of-​speech classification, the book calculates the compati-
bility degree among grammatical functions to reveal the relationship between
a grammatical function and a part of speech. The book suggests that there is
compatibility among some grammatical functions. For example, a word that
can function as a subject can also function as an object and vice versa. Different
grammatical functions that have a rather large degree of compatibility reflect
the same properties of a part of speech, have equivalent correspondence to it
and can be clustered together. A cluster of equivalent grammatical functions
is actually the distinctive function of a part of speech. The identification of
distinctive functions of all parts of speech is, in effect, the classification of all
the words into different parts of speech. This method, in our opinion, moves
one step forward toward the demonstration of part-​of-​speech classification.
 19

Introduction 19

1.5.2  Overview of the part-​of-​speech system in the book

Word

combinational word independent word

content word functional word

kernal word modificational word

predicate word substantive word determinative

locative

1. 2. 3. 4. 5. 6. 7. 8. 9. 10.11.12. 13. 14. 15.16.17.18. 19.

v a s n m p t 1 d n q d a i p c d a i
d t e o i o i u u e d n r o i u n
i a a sm s s ma m v t e n s x t
t s e a t n o e p j / e
e t n r m r
s j a j
r
k
20. pronoun

Figure 1.1 A brief description of modern Chinese part-​of-​speech system

Chinese parts of speech are classified according to their hierarchies. There are
20 basic parts of speech, 19 of which are classified according to their gram-
matical functions. A pronoun is special and is not classified according to its
grammatical functions, which are equivalent to a verb, a noun, a time word,
a place word, a numeral, a measure word and an adverb. The part-​of-​speech
examples are given as follows:

1. Verbs: 吃 (eat), 洗 (wash), 跑 (run), 休息 (rest), 想 (think), 病 (fall ill), 坐

(sit), 有 (have), 是 (be), 来 (come), 能 (can), 可以 (may), 完成 (accom-
plish), 思考 (ponder)…
2. Adjectives: 高 (high), 短 (short), 大 (large), 晚 (late), 干净 (clean), 认真
(serious), 伟大 (great), 不幸 (unlucky), 有趣 (interesting), 可笑 (laugh-
able), 结实 (sturdy), 相同 (identical)…
3. State words: 雪白 (snow-​white), 甜丝丝 (happy), 黑咕隆咚 (pitch-​dark),
轰轰烈烈 (vigorous), 优良 (good), 酷热 (extremely-​hot), 瘦高 (thin and
tall), 旖旎 (graceful)…
4. Nouns: 石头 (stone), 文化 (culture), 国家 (country), 人民 (people), 船只
(boat), 时间 (time), 钟头 (hour), 长江 (Changjiang), 泰山 (Mount Tai),
孔子 (Confucius), 学校 (school)…
20

20 Introduction
5. Measure words:  个 (piece), 张 (sheet), 斤 (jin), 次 (time), 天 (day), 年
(year), 分钟 (minute), 点 (point), 些 (some), 种 (type), 团 (regiment), 滴
(drop), 杯 (cup), 瓶 (bottle), 批 (batch), 套 (set) …
6. Locatives:  前 (front), 上 (above), 里 (inside), 左 (left-​side), 南 (south),
下面 (below), 后头 (back-​end), 以前 (before), 周围 (ambient), 旁边
(aside), 附近 (near), 对面 (opposite-​side) …
7. Time words: 今天 (today), 去年 (yesteryear), 上午 (morning), 刚才 (just),
过去 (past), 春节 (Spring Festival), 正月 (first month of a lunar year),
最近 (recently), 拂晓 (daybreak), 星期天 (Sunday) …
8. Place words:  当地 (local), 街头 (street), 门口 (doorway), 野外 (field),
一旁 (on one side), 乡下 (countryside), 民间 (folk), 原处 (original place),
远处 (distant), 基层 (grass-​root level), 头里 (front) …
9. Distinctive words:  高等 (high-​class), 公共 (public), 亲爱 (dear), 民用
(civil), 日常 (routine), 随机 (random), 袖珍 (miniature), 现行 (current),
野生 (wild), 业余 (amateur), 男 (male) …
10. Numerals: 一 (one), 二 (two), 三 (three), 五 (five), 十 (ten), 百 (hundred),
千 (thousand), 万 (ten thousand), 亿 (hundred million), 半 (half), 几 (sev-
eral), 数 (several), 多 (multiple), 诸 (all) …
11. Measure words: 一切 (each and all), 大量 (bulk), 不少 (good many), 所有
(all), 大批 (large batch), 部分 (some), 个把 (one or two), 少许 (a little), 俩
(both), 片刻 (a while), 许久 (long time) …
12. Demonstratives: 这 (this), 那 (that), 每 (each), 其他 (other), 任何 (any),
另 (another), 唯一 (only), 上 (above), 下 (below), 前 (previous), 后 (rear),
头 (first) …
13. Adverbs: 很 (very), 都 (totally), 只 (only), 也 (also), 又 (again), 就 (just),
不 (not), 赶紧 (speedily), 常常 (often), 正在 (be doing), 亲自 (personally),
难道 (surely), 究竟 (on earth) …
14. Onomatopoeia: 啪 (bang), 叮当 (dingdong), 哗啦 (crash) …
15. Prepositions: 被 (by), 从 (from), 对 (versus), 在 (at), 按照 (according to),
比 (than), 跟 (with), 向 (to), 由 (via), 凭 (on) …
16. Conjunctions:  和 (and), 或者 (or), 而 (but), 不但 (not only), 而且
(but also), 虽然 (although), 但是 (however), 即使 (even though), 况且
(besides) …
17. Modal words: 吗 (ah), 呢 (well), 来着 (only), 罢了 (just) …
18. Auxiliaries: 了(le), 着 (zhuo), 过 (guo), 的 (de), 地 (di), 得 (de), 所 (suo),
等 (etc.), 似的 (like) …
19. Interjections:  啊 (ah), 唉 (alas), 喂 (hello), 哎哟 (oh), 哼 (hum), 哎呀
(gosh) …
20. Pronouns:  我 (I), 你 (you), 他们 (they), 谁 (who), 什么 (what), 哪里
(where), 几 (several), 多少 (how many), 怎样 (how), 怎么样 (what about),
这儿 (here) …
 21

2 
Basic issues concerning classification
of parts of speech

2.1  Parsing of words

The parsing of words is not discussed in detail here.
It is difficult to divide the boundary line between word and phrase in
Chinese; hence, a good many scholars have raised the question of continuity1
between word and phrase (see Wang Hongjun, 1994; Shen Yang, 1996).
Perhaps we should not emphasize the continuity between word and phrase
but instead emphasize that there is indeed a boundary line between them,
although it is difficult to divide the line.
The parsing of words requires an understanding of their properties.
Functionally speaking, a word is the smallest sentence-​making unit. From the
perspective of properties, it is the smallest applicable linguistic constituent.
A word has the following three properties:

1. Integrity: A word is a whole or integral, and is not separable.

2. Independent application: As a sentence-​making unit, a word may be used
independently or combined with other linguistic constituents temporarily
but not fixedly.
3. Stand-​by application:  A word is a standby linguistic unit for language
application, which is kept in stock in a word bank rather than generated
temporarily.

Taking into consideration the properties of a word, we can propose the

following three principles to determine its boundary line.

2.1.1  The integrity principle

A word has its integrity, and as a whole it is not separable or incomplete. The
integrity principle can be described with the following two criteria:

2.1.1.1  Expansion criteria

A word usually cannot be expanded. In principle, an expandable segment is a
phrase; a segment that cannot be expanded is a word.
22

22  Basic issues
Expandable: 手疼 (hand pain) → 手很疼 (The hand is very painful); 手上
(on hand) → 我的手/​上(on my hand)
Not expandable (word): 肉麻 (disgusting) →*肉很麻 (*The flesh tingles);
身上 (body) → 我的/​身上 (on my body)
手很疼 (The hand is very painful) and 手上 (on hand) are expandable and
therefore are phrases; 肉麻 (disgusting) and 身上 (body) are not expandable
and therefore are words.

2.1.1.2  Integrity criteria

A word usually must be integral. In principle, a segment without integrity is a
phrase; an integral segment is a word.
Not integral:  急性肠炎和慢性肠炎 (acute enteritis and chronic enteritis)
→ 急性和慢性肠炎 (acute and chronic enteritis); 火车上汽车上 (in train and
in automobile) → 火车汽车上 (in train and automobile)
Integral: train and automobile (火车和汽车) →*火汽车
桌子椅子 (desk and chair) →*桌椅子
地上身上 (on the ground and on one’s body) →*地身上
急性肠炎 (acute enteritis) and 火车上 (in train) can be separable and there-
fore are phrases; 火车 (train), 桌子 (desk) and 身上 (body) are not separable
and therefore are words.

2.1.2  The transparency principle

As a temporary combination unit, a word has its structural and func-
tional transparency and transparent meaning when combined with other
constituents, whereas the morphemes inside a word are often not transparent.
The transparency principle has the following three criteria:

2.1.2.1  Structural transparency criteria

A word has a transparent structure when combined with other words or
phrases, whereas the structure inside a word may not be transparent. The
structural transparency is manifested as follows:

(1) Whether a structural relation is apparent or not. A  segment whose

internal structural relation is not apparent is usually a word.

The structural relation is apparent: 新书 (new book), 桌子椅子 (desk and

chair), 吃完 (eat up), 白菜 (celery cabbage), 提高 (raise high) (phrase).
The structural relation is not apparent:  天真 (naive), 个别 (individual),
若干 (some), 冬烘 (pedant) (word).

(2) Whether the type of structural relation exists or is in agreement with a

typical phrase or not.
 23

Basic issues 23
Not existing in a typical phrase: 纸张 (paper), 船只 (boat), 马匹 (horse),
人口 (population) (word).
Not in agreement with the structural relational type of a typical phrase: 枣红
(claret) (noun + adjective = attributive and headword construction consisting
of modifier and headword) (word).
The above two cases should be regarded as words.

(3) Whether there is structural reanalysis or not:

The segment that causes structural reanalysis should be regarded as a

word:  等于 (equal), 善于 (skillful), 所以 (therefore), 不仅 (not only), 毫不
(no), 之一 (among), 之前 (before), 加以 (conduct), 有所 (somewhat).

2.1.2.2  Functional transparency criteria

Functional transparency means that the function of the segment combined in
a certain mode is the same as that of the typical phrase combined in the same
mode. The segment whose function is not transparent should be regarded
as word.
Transparent function:  手疼 (hand pain) →*很手疼 (*very hand pain);
很新 (very new) →*很新衣服; 加水 (add water) →* add water cup (加水杯子)
Nontransparent function: 头疼 (headache) → 很头疼 (terrible headache);
很多 (many) → 很多衣服 (many clothes); 加工 (process) → 加工木材 (process
timber)
The functions of 手疼 (hand pain) and 加水 (add water) are trans-
parent, and therefore they are phrases, whereas those of 头疼 (headache),
很多 (many), and 加工 (process) are not transparent, and therefore they
are words.

2.1.2.3  Transparent meaning criteria

Transparent meaning denotes that the meaning of the segment as a whole can
be inferred from its component constituent. The segment whose meaning is
not transparent should be regarded as word.
Transparent meaning: 白鞋 (white shoe), 白云 (white cloud), 白发 (white
hair) (phrase or word).
Non-​transparent meaning:  白菜 (cabbage), 痛快 (delighted), 丈夫 (hus-
band) (word).

2.1.3  The finite principle

As a standby unit, a word has a finite number and length.
Finite criteria: the segment whose length and number of constituents of
the same kind should be regarded as a phrase.
24

24  Basic issues
The length of segments such as “3245” can be expanded infinitely, and the
number of constituents of their same kind is infinite; therefore, they should
be regarded as phrases.

2.2  Internal hierarchies of modern Chinese

To divide parts of speech, the first thing to do is distinguish whether a con-
stituent is a word or not. But sometimes it is hard to tell whether a constituent
is a word or not. Lv Shuxiang (1979) pointed out that it is easy to tell whether
some constituents are words or not, but that it is complicated to tell some
others. For example, 叶 (leaf) does not form a word in most cases but can
do so in botany; 时 (time) does not form a word in speech but can do so in
writing. If 叶 (leaf) and 时 (time) are recognized as words, then an important
fact is obliterated:  these morphemes cannot form words in most cases.
Furthermore, which part of speech a word belongs to sometimes depends on
different occasions. For example, 杯 (cup) should be regarded as a measure
word in spoken Chinese but can be used as noun in writing, for example,
杯中 (in a cup). 金 (gold) is a distinctive word in everyday spoken Chinese
but a noun in scientific and technical Chinese. In the past, modern Chinese
was considered to be homogeneous, thereby having no way to deal with this
phenomenon. In reality, modern Chinese is an inhomogeneous system mixed
with different hierarchical constituents. By analyzing its internal hierarchies,
we can deal with the phenomenon well.
Modern Chinese mainly has its historical hierarchy and field hierarchy
(in other words, a two-​hierarchy system). Its writing is actually a mixture of
different historical hierarchies and can be largely classified into two hierarchies
(systems): modern colloquial Chinese hierarchy and classical Chinese hierarchy.
Written modern Chinese is based on modern colloquial Chinese and mingled
with some expressions of classical Chinese. In written modern Chinese, the clas-
sical Chinese hierarchy mainly manifests itself in the following two forms:

1. The classical Chinese words that are mingled with the modern collo-
quial Chinese. For example, 国 (country), 时 (when, e.g., 来时 (when
coming)), 战 (war, e.g., 站前 (before war)), 该 (the, e.g., 该同志 (the com-
rade)), 本 (this, e.g., 本校 (this school)), 虽 (though, e.g., 虽难 (though
difficult)), etc.
2. The classical Chinese usage mingled with modern colloquial Chinese.
Although some words exist in modern colloquial Chinese, they can still
be used in classical Chinese, for example:
(31) 在一个“关于儿童对物体运动速度”的认知研究中,实验人员让儿
童比较两个小汽车运动的速度是否一致(事实上两车运动的速
度是一致的),其目的是要看看儿童是怎样理解求速度的公式的.
(《儿童的心理世界》,方富熹、方格主编,北京大学出版社) (To
study children’s cognition of an object’s motion speed, the experi-
mental personnel let them compare whether the motion speeds of
 25

Basic issues 25
two vehicles are the same or not (in fact, their speeds are the same),
so as to understand how children solve formulae for speed. (The
Psychological World of Children, compiled by Fang Fuxi and Fang
Ge, Peking University Press)
(32)
一些人纪律和全局意识淡化,搞上有政策,下有对策,有令不听,我行
我素,导致小团体主义和无政府主义滋长. (《人民日报》1995年6
月27日) (Some people have a weak discipline and weak global
awareness; the inferior have their own countermeasures when the
superior have their policies. They do not obey their superior and
act according to their own wishes, thus giving rise to the spread of
factionalism and anarchism.) (The People’s Daily, June 27, 1995)
(33)
于1993年8月8日举办新兴公司成立两周年的盛大庆典,宴开137桌
,发给与会千余来宾每人288元的”红包“。 (《人民日报》1995年1
2月4日) (The second grand anniversary celebration ceremony for
the Xinxing Company foundation will be held on August 8, 1993;
the banquet consists of 137 tables; each of around thousands of
participants will receive their own “red envelope” of 288 yuan.)
(The People’s Daily, December 4, 1995)

The word 车 (vehicle) in Example (31) is in the modern colloquial Chinese

hierarchy, but is directly modified by a numeral without a measure word, thus
being used as classical Chinese. The words 上 (superior) and 下 (inferior) in
Example (32) are also in the modern colloquial Chinese hierarchy, but here
they are used as subjects independently, just like classical Chinese words.
Before the numeral 千 (thousand) in Example (33), there is no measure word,
which is used as a classical Chinese word.
The constituents in the classical Chinese hierarchy primarily occur in verse,
title, four-​word phrase, parallel format, and in front of monosyllable locatives
such as 前 (front), 后 (rear), 内(in), 外 (outside), 中 (middle), military lan-
guage, journalistic language, official document language, and scientific and
technical language.
The constituents in different historical hierarchies have three possible
grammatical differences:

1. Whether being a word or not: For example, 国 (state), 校 (school), 敌

(enemy), 鸭 (duck), 应 (should), 该 (this), 置 (put), 共 (total) and 俩
(both) do not form words in the colloquial modern Chinese hierarchy but
do so in the classic Chinese hierarchy. For example:
(34) 冷战结束后,西方国家在对外政策中突出了经济因素,在处理国与
国关系时,对本国经济利益的考虑往往多于对政治的考虑. (《人民
日报》1995年1月1日) (After the end of the Cold War, the foreign
policies of the Western countries emphasized economic factors;
in handling the relationship between their countries and other
countries,2 they often considered their economic interests more
important than their politics.) (The People’s Daily, January 1, 1995)
26

26  Basic issues
(35) 十多年来,该校共培养了2.7万名干部,超过了同期该校普通高等
教育培养的人数. (《人民日报》1995年1月4日) (For more than
a decade, the school has trained a total of 27,000 cadres, which
exceeds the number of people who received higher education by the
school in the same period.) (The People’s Daily, January 4, 1995)
(36) 左手一只鸡,右手一只鸭. (歌词) (A chicken in the left hand; a duck
in the right hand.) (a song)
(37) 运 用 观 察 法 时 ,应 使 被 观 察 者 始 终 处 于 日 常
教 学 或 生 活 的 情 景 之 中 ,尽 量 避 免 外 来 干 扰 .
(《儿童的心理世界》,方富熹,方格主编,北京大学出版社) (In
applying the observation method, the observed should always
be in an everyday teaching or living context; every effort should
be made to avoid extraneous interference.) (The Psychological
World of Children, compiled by Fang Fuxi and Fang Ge, Peking
University Press)
(38) “抓两头”,就不能一头热,一头冷,更不能”嫌贫爱富“,置贫困村于
不顾. (《人民日报》1995年1月4日) (“To take charge of two ends”,
you cannot just emphasize one end but ignore another, not to
mention that “you like the rich but dislike the poor”, ignoring the
poor villages.) (The People’s Daily, January 4, 1995)
(39) 他先以一部兵力围攻徐水,伺机诱歼来援之敌. (《人民日报》1995
年4月28日) (He first uses part of his military force to lay siege to
Xushui County and then waits for his chance to seduce and annihi-
late the enemy reinforcements.) (The People’s Daily, April 28, 1995)
2. Grammatical functions: Namely the differences between the classical
Chinese usage and the general usage that are mingled in modern collo-
quial Chinese, as mentioned above.
3. Parts of speech: For example, the words like 杯 (cup), 盘 (plate), 拳 (fist)
and 口 (mouth) belong to measure words in the modern colloquial Chinese
hierarchy like 一杯酒 (a cup of wine), 一盘菜 (a plate of vegetable), 打一拳
(hit a fist) and 咬一口 (give a bite) but belong to nouns in the classical
Chinese hierarchy like 杯中 (in a cup), 盘中 (on a plate), 挥拳 (shake one’s
fist), and 口中 (inside a mouth). The word 依然 (still) belongs to an adverb
in modern colloquial Chinese, for example, 国际形势依然不稳定 (The
international situation is still unstable). But it belongs to state words in the
classical Chinese hierarchy, like 风采依然 (The charm is still the same).

The field hierarchy is mainly classified into everyday language hierarchy

and specialized language (including scientific and technical language) hier-
archy. For example, 叶 (leaf) and 鳄 (crocodile) are not words in the everyday
Chinese hierarchy, but are words in biological Chinese. 胸 (chest) and 肩
(shoulder)3 are not words in the everyday Chinese hierarchy but are words in
the physiological and medical Chinese hierarchies. For another example, the
word 金 (gold) is a distinctive word in the everyday Chinese hierarchy, but is
a noun in the scientific and technical Chinese hierarchy. The word 热 (hot) is
 27

Basic issues 27
an adjective or a verb in the everyday Chinese hierarchy, but can be a noun in
the scientific and technical Chinese hierarchy.
Some dictionaries only recognize the universal hierarchies of the
words that form words in different hierarchies and have different parts
of speech (modern colloquial Chinese and everyday Chinese hierarchies).
Those that do not form words in the universal hierarchy are treated as
morphemes and labeled into parts of speech according to the universal
hierarchy.4 Other dictionaries recognize both the universal hierarchy and
the specialized hierarchy, but do not divide their parts of speech.5 However,
the two types of dictionaries are too simple; it is better to recognize the
constituents in various hierarchies and label their hierarchies. For example,
the constituents in modern colloquial Chinese and everyday Chinese hier-
archies are not labeled, but the constituents in the classical Chinese and
specialized Chinese hierarchies are labeled as follows:  国 (country:  [clas-
sical Chinese], noun; 口 (mouth): measure word, [classical Chinese], noun;
金 (gold): distinctive word; [chemistry], noun.
A language may have different grammatical systems at different histor-
ical and field hierarchies; the grammatical functions of its words may also
be different, and the criteria for classifying their parts of speech should be
different too. The classification of modern Chinese parts of speech should
consider these different historical hierarchies and treat the constituents in
different hierarchies differently. The ideal way of dealing with this is to sep-
arate the two hierarchies completely and establish a set of criteria for classi-
fying the parts of speech for them separately. However, there is no clear-​cut
boundary between modern colloquial Chinese and classical Chinese hier-
archies, and it is difficult to completely divide the two hierarchies. Moreover,
there are more grammatical similarities between the two hierarchies than
differences. Since the purpose of our study is to parse all the words in written
and spoken modern Chinese, we consider the colloquial modern Chinese and
classical Chinese hierarchies jointly together and establish the same criteria
for classifying their parts of speech. For example, only a word that meets the
criterion of “numeral ~Λ* subject or object” is a measure word.

2.3  The identity of words

2.3.1  Individual words and generalization words
Each appearance of a word is a token (individual word); several individual
words that have the same sound form and meaning can be grouped into the
same type (generalization word). The determination of a generalization word
is to determine whether the word that appears in different places is the same
word that appears several times or a different word, namely determining its
identity. The object of classification is type (generalization word). Therefore,
words must be grouped before classification, and grouping is a precondition
for classification (see Zhu Dexi, Lu Jiawen & Ma Zhen, 1961).
28

28  Basic issues
What we mean by classifying parts of speech is to take a generalization
word as the unit to divide its parts of speech. For this reason, if the individual
words that appear at different grammatical positions are identical, then all
their functions should be regarded as different functions of the same word;
however, if not, they should be regarded as different generalization words.
The determination of word identity involves whether a word has certain
grammatical functions and thereby may influence grouping results. Besides, it
may influence whether a word should be included in a dictionary. Therefore,
it is one of the issues to be solved first in order to divide parts of speech. For
example, if 木头 (wood) in 木头桌子 (wooden desk) is regarded as not iden-
tical with 木头 (wood) in 买木头 (buy wood), then it should be treated as a
noun and concurrently as an adjective (non-​predicate).
2.3.2  Case studies
The principles for word identity determination are homophony and syn-
onymy, but how to use the principles is a tricky question. In the following, we
will explain in some detail several cases of word identity determination with
regard to the classification of parts of speech.

2.3.2.1  Constituent meaning and structural meaning

In determining the identity of meaning, we should distinguish between the
meaning possessed by a word itself (constituent meaning) and the struc-
tural meaning. Some grammar books hold that the meaning of 死 (die) in
他死了 (he died) and 他死了父亲 (he died his father) (in Chinese grammar) is
different and regard them as two generalization words. The former is intransi-
tive, while the latter is transitive. In our view, the difference in meaning
between 他死了(he died) and 他死了父亲 (he died his father) is not caused
by 死 (die) itself but by its grammatical structure. They belong to the same
generalization word.
Then, how should we tell whether a certain meaning belongs to a word’s
own meaning or its structural meaning? Our answer is that the systematic
and analogical meaning is the structural meaning and, if not, the constituent
meaning. For example, 死 (die) is a verb with univalent change. We can place
its only argument element at the position of subject or object, taking the
above 他死了 (he died) and 他死了父亲 (he died his father) for examples. The
majority of or almost all verbs and state verbs with one syllable and univalent
change can behave like this. Other examples are the following:

(1) a. 腿断了 (A leg breaks). b. 桌

 子断了条腿 (The desk
breaks one of its legs).
(2) a. 眼睛瞎了 (The eye is blind). b. 他
 瞎了一只眼睛 (He blinds
one of his eyes).
(3) a. 钱丢了 (Money is lost). b. 我丢了钱 (I lose money).
(4) a. 第一段漏了 (The first b. 漏
 了第一段 (the first para-
paragraph leaves out).     graph is left out)
 29

Basic issues 29
(5) a. 电线杆倒了 (The electric b. 倒
 了一根电线杆 (fell an
pole falls).           electric pole).

The verbs in the above five examples should be regarded as the identical
generalization words in Groups a and b respectively. Other examples are the
following:

(6) a. 他很高 (He is very tall). b. 他

 高我一头 (He is one head taller
than me).
(7) a. 他很小 (He is young). b. 他  小我两岁 (He is two years
younger than me).
(8) a. 他很大 (He is older). b. 他
 大我半岁 (He is half a year older
than me).
(9) a.他很矮 (He is short). b. 他矮我一寸 (He is one inch shorter
than me).

In these examples, the adjectives in Groups a and b are systematically

different. The comparative meanings in Group b should be understood as
structural meanings and not possessed by the adjectives themselves. The
adjectives in both groups should be regarded as generalization words.
Let us analyze the following groups of examples:

(10) a. 文  化生活丰富 (The cultural b. 丰

 富文化生活 (enrich
life is rich).              cultural life)
(11) a. 思想统一 (Thoughts are unitary). b. 统一思想 (unify
thoughts)
(12) a. 制度完善 (The system is complete). b. 完善制度 (complete the
system)
(13) a. 关系密切 (The relation is intimate). b. 密切与群众的关系
(establish close relations
with the masses)
(14) a. 队伍壮大 (The army is strong). b. 壮大队伍 (strengthen
the army)
(15) a. 生活充实 (Life is rich). b. 充实生活 (enrich life)

The above examples show that the “make-​do” meaning in the type of word
in 丰富文化生活 (enrich cultural life) is systematic and analogical and should
be regarded as structural meaning. That is to say, this type of word has the
“make-​do” meaning when it is used in the “~ (object)” environment. The
“make-​do” meaning is grouped into the structural meaning brought about
by the “adjective + object” construction. But so long as we enlarge the scope
of examination, we may find three facts: (1) there is only a limited number
of adjectives that have the so-​called “make-​do” usage; (2)  many adjectives
that are synonymous or near-​synonymous with the type of words such as
丰富 (rich) and 统一 (unitary) have no “make-​do” usage. For example, 丰盛
30

30  Basic issues
(rich), 一致 (consistent), 完备 (complete), 紧密 (close), 强大 (powerful) and
实在 (actual) are nearly synonymous or synonymous with 丰富 (rich), 统一
(unify), 完善 (complete), 密切 (close), 壮大 (strengthen) and 充实 (enrich) but
have no “make-​do” usage. (3) The so-​called “make-​do” meaning appears not
only when adjectives have objects but also on other occasions, for example,
基础得到巩固 (the foundation is consolidated), 对制度加以完善 (perfect a
system), 使《纲要》提出的原则不断得到丰富和发展 (continuously enrich
and develop the principles raised in the program), 意见不能统一 (opinions
cannot be unified). In these examples, the “make-​do” meanings of 巩固 (con-
solidate), 完善 (perfect) and 统一 (unify) can only be regarded as the meanings
possessed by these words themselves, not their structural meanings. Hence, we
have to admit that the “make-​do” meaning of the type of words such as “rich”
is possessed by the words themselves.
We have another reason to regard the “make-​do” meaning as the words’
own meanings instead of their structural meanings. Namely, if we regard the
“make-​do” meaning as the structural meaning, then we have to admit that it
is a syntactic means in modern Chinese. As a result, we can only group the
type of words such as 丰富 (rich) and 完善 (perfect) into adjectives instead of
conversional words that function as adjective and verb, thus making syntax
unnecessarily complicated. In fact, on the whole, no one thinks that modern
Chinese has the “make-​do” syntactic means. In dealing with word identity, we
should coordinate words with syntax. Since we think that there is no “make-​
do” usage in Chinese syntax, we can only group the “make-​do” meaning of
the type of word 丰富 (rich) into its own meaning. Therefore, words such as
丰富 (rich), 统一 (unitary), 完善 (perfect), 密切 (intimate), 壮大 (strong) and
巩固 (consolidated) should belong to identical generalization words.

2.3.2.2  Lexical transfer reference and syntactic transfer reference

The type of words such as 领导 (leader), 导演 (director), 锁 (lock), 铲 (shovel),
练习 (exercise), 报告 (report), 典型 (representation) and 内行 (expert) has the
two uses of predicate and noun. The two uses of 领导 (leader) and 导演 (dir-
ector) have the relationship between action and its doer; the two uses of 锁
(lock) and 铲 (shovel) have the relationship between action and instrument;
the two uses of 练习 (exercise) and 报告 (report) have the relationship between
action and content; the two uses of 典型 (representation) and 内行 (expert)
have the relationship between attributive and subject. In these examples,
when used as noun, these words have a reference meanings; what they refer
to are not action, behavior and attribute themselves but other objects related
to them. This reference is what we call transfer reference. The transfer refer-
ence of the above words is a kind of fixed usage. A dictionary should take
them as independent senses, and we call this type of transfer reference lexical
transfer reference. Therefore, the lexical transfer reference should be regarded
as an independent generalization word and should differ from a generaliza-
tion word’s predicate usage.
 31

Basic issues 31
Opposite to lexical transfer reference is syntactic transfer reference,
which refers to the temporary transfer reference when some words appear at
the position of subject or object. Because syntactic transfer reference has no
fixed meaning, we regard syntactic transfer reference and its corresponding
non-​transfer reference as the same generalization word, for example:

(16) a. 很大/​很小 (very big/​very small) b. 有大有小 (have a big or small

thing)
(17) a. 很肥/​很瘦 (very fat/​very thin) b. 有肥有瘦 (have fat and thin)

The 大 (big) and 小 (small) in Example (16 b) refer to a big or small thing;
肥 (fat) and 瘦 (thin) in Example (17 b) refer to fat meat or thin meat respect-
ively. In both cases, there is a syntactic transfer reference. We group all of
them in both groups into the same generalization word respectively.
The syntactic transfer reference of a predicate is very rare in modern
Chinese but very common in classical Chinese, for example:

(18) 见贤思齐,见不贤而内自省(《论语·里仁》) (When we see a man of

virtue and talent, we should think of equaling them; when we see a
man of a contrary character, we should turn inward and examine
ourselves.) (Confucian Analects)
(19) 然 则 小 故 不 可 以 敌 大 , 寡 故 不 可 以 敌 众 , 弱 故 不 可 以 敌 强
(《孟子·梁惠王上》) (However, the small naturally cannot rival the
big; the few, of course, are no match for the many; the weak cannot
rival the strong.) (Book One on Lianghuiwang, Mencius)
(20) 失所长则国家无功,守所短则民不乐生.以无功御不乐生,不可行于
齐民(《韩非子·安危》) (With lack of the meritorious, the state
is powerless; if keeping the unmeritorious, the people do not enjoy
their life. Under these circumstances, it is impossible for the Qi
people to live without merit.) (Chapter on Safety and Danger in Han
Fei Zi)
(21) 将 军 身 披 坚 执 锐 ,伐 无 道 ,诛 暴 秦 ,复 立 楚 国 之 社 稷 ,功 宜 为 王
(《史记·陈涉世家》) (The general is well armed to the teeth,
combats the brutal and the tyrannical Qin state and reestablishes the
Chu state. His merit makes him deserve to be a king.) (Chen She’s
Story in Historical Records)
(22) 今梁赵相攻,轻兵锐卒心竭于外,老弱罢于内 (《史记·孙子吴起
列传》) (Now the Liang and Zhao states attack each other; the
light-​armed and well-​protected soldiers aspire to fight outside, and
the old and weak rest inside.) (The Biographies of Sun Zi and Wu Qi
in Historical Records)

The syntactic transfer reference of a modification word is common in

modern Chinese, for example:
32

32  Basic issues
(23) a. 
急性肠炎好治,慢性肠炎不好治 (Acute enteritis is easy to cure,
but chronic enteritis is not easy).
b. 
急性好治,慢性不好治 (The acute are easy to cure, but the
chronic are not easy).
(24) a. 打一个长途电话 (make a long-​distance phone call)
b. 打一个长途 (call a long distance)
(25) a. 
许多苹果都坏了 (Many apples get rotten).  b. 许  多都坏了(Many
get rotten).
(26) a. 没收一切财产 (confiscate all property) b. 没收一切 (confis-
cate all)

The distinctive words such as 急性 (acute), 慢性 (chronic) and 长途 (long-​

distance) in Group b of Examples (23) and (24), and the measure words 许多
(many) and 一切 (all) in Group b of Examples (25) and (26) are all syntactic
transfer references and grouped into the same generalization words with those
in Group a respectively.

2.3.2.3  Lexical transfer reference and lexical self-​reference

Some predicates can be used for a reference but have no transfer refer-
ence meaning except self-​reference, namely referring to action, behavior or
attribute itself. For example:

(27) a. 研究问题 (study a problem) b. 进行研究 (carry out study)

(28) a. 答复对方 (answer the other party) b. 给 对方一个答复 (give the
other party an answer)
(29) a. 收支不平衡 (Income and b. 保持平衡 (keep one’s
payment are not balanced). balance)
(30) a. 很不幸 (very unfortunate) b. 遭受不幸 (suffer from
misfortune)

研究 (study), 答复 (answer), 平衡 (balanced) and 不幸 (unfortunate) in

Group a of Examples (27) through (30) are used as predicates, but in Group
b they are used as nouns and have self-​reference meanings. The self-​reference
has its fixed meaning and is called lexical self-​reference. We regard the lexical
self-​reference and its corresponding predicate use as the same generalization
word. For example, the above words belong to the same generalization words
in Groups a and b respectively.
Although there is an evident distinction between typical lexical self-​
reference and typical lexical transfer reference, the boundary line between
lexical self-​reference and transfer reference is still rather indistinct. We will
specify the distinction between the two in the following.
The relationship between action and reference is as follows:

1. The relationship between action and its doer, for example, 代表 (represen-
tative), 领导 (leader), 编辑 (editor), 导演 (director), 指挥 (commander).
 33

Basic issues 33
2. The relationship between action and tool, for example, 锁 (lock), 锯
(saw), 铲 (shovel).
3. The relationship between action and its object, for example, 摆设 (fur-
nish), 储蓄 (save), 花费 (expenditure), 建筑 (building), 武装 (armament).
4. The relationship between action and content. The content means the
things existing in the form of substance or abstract things, for example,
通知 (information), 报告 (report), 补助 (subsidy), 练习 (exercise), 要求
(requirement), 计划 (plan), 忠告 (advice), 根据 (basis), 感觉 (perception),
命令 (order), 建议 (suggestion), 定义 (definition), 主张 (claim).
5. The relationship between action and source or credential, for example,
区别 (difference), 趋向 (tendency), 仇恨 (hatred), 依据 (basis), 病 (dis-
ease), 梦 (dream).
6. The use as a noun purely means the action itself, for example, 研究
(research), 调查 (investigation), 生产 (production), 解决 (solution), 改革
(reform).

Among the above six cases, we regard the first five cases as transfer refer-
ence and treat them as verb and concurrently noun (see Chapter 7); the sixth
case is regarded as self-​reference.

2.3.2.4  Word-​formation, morphology and syntax

In the above, we discussed the word identity caused by meaning. In the
following, we will discuss the identity caused by changes in form.
Morphology and word-​ formation come from the Indo-​ European
languages, where the changes in a word’s form bring about two results, taking
English for example:

a. Changes in lexical meaning:  work -​ worker (addition), sing -​ song

(inflection)
b. Changes in grammatical meaning (no change in lexical meaning): work
-​worked (addition), sing -​sang (inflection)

Group a belongs to word-​formation, namely the change in a word form

creates a new word (derivative); Group b belongs to morphological change.
That is to say, the morphological change does not create a new word, only a
different form of the same word. In other words, the constituents represented
by the two forms in Group a are not identical and should be regarded as two
words, but the constituents represented by the forms in Group b are identical
and should be regarded as one word.
It is generally held that Chinese has no morphological changes like those in
Indo-​European languages but has its morphology, which is achieved by redupli-
cation and addition of -​zi (子), -​er (儿), -​tou (头) and -​hua (化). But they are
very likely to belong to neither morphology nor word-​formation but to syntax.
The so-​called syntax means that the changes in form can be regarded as a kind
of word combination. For example, “走 (going) + le (了)” does not belong to
34

34  Basic issues
word-​formation or morphology but belongs to a syntactic combination. If the
changes in form are syntactic, then the original form and the varied form belong
to the same generalization word. We will distinguish between word-​formation,
morphology and syntax based on the following six criteria:

A. System: it can be analogized in a certain scope. A system does not neces-

sarily mean a big quantity; under certain conditions, most changes in
form are systematic.
B. Compulsion: whether it must be so as long as conditions are met.
C. Functional identity: whether the grammatical functions of a varied form
are the same as the original form or still have the basic functions of the
original form.
D. Whether there is a systematic difference in meaning between varied form
and original form.
E. Whether the original form is a word: English has the word “porter”, but
its original form “port” is not the word from which “porter” is derived.
F. Whether the root part and the part to be used to change form are words.

In the following, we will analyze reduplication. There are three types of

reduplication:  morphological reduplication, word-​ formation reduplication
and syntactic reduplication.
Zhang Shoukang (1985) listed the word-​ formation and morphology
of noun, verb, adjective, measure word, numeral, pronoun and so on, and
thought that the reduplication of verb, adjective and measure word belongs
to morphology. We will analyze them one by one according to the six criteria.

1.  VERB REDUPLICATION

There are two forms of verb reduplication: ABAB and AABB.

The ABAB reduplication form:  商量商量 (consult), 讨论讨论 (discuss).
They are systematic; their grammatical functions are basically the same and
have no compulsion; both parts are words. Therefore, they should be regarded
as not morphological reduplication but syntactic reduplication.

Table 2.1 The distinctions among morphology, word formation and syntax

System Com­pulsion Functional Systematic A word in Two Systematic

identity difference its original ­independent difference
in meaning form words in meaning

Morphology + + + + + − −/​+
Word ? − ? ? ? − −/​+
formation
Syntax + − + + + + +

Note: + denotes the existence of the characteristic; -​denotes the non-​existence of the character-
istic; ? denotes no requirement for the characteristic, which is optional.
 35

Basic issues 35
The AABB reduplication form: a. 比比划划 (gesticulate), 拉拉扯扯 (pull
and push):  they are not systematic, have different grammatical functions
and belong to word-​formation reduplication. b.  走走停停 (walk and stop),
说说笑笑 (josh): their order can be changed into 停停走走 (stop and walk),
denoting that the two actions alter in turn (or both actions happen). They
should be regarded as the reduplication of the two verbs 走 (walk) and 停
(stop). This is syntactic reduplication.

2.  ADJECTIVE REDUPLICATION

In the past, the reduplicated form of an adjective was called the vivid form
and regarded as identical.
The adjective reduplication is systematic (a good many adjectives can be
reduplicated). But the original form that is not a word has no functional
identity:

a. The original word is not a word:  堂堂正正 (dignified and imposing),

风风火火 (rash and too much in haste), 密密麻麻 (close and numerous,
very dense); the reduplicated form is not a word: 胖胖 (fat), 红红 (red),
圆圆 (round)
b. The original form and the reduplicated form have different functions in
the following ways:

Compare:

Not ~ very ~ combined complement ~ complement

干净 (clean) 十 十 + +
干干净净 (very clean) -​ -​ -​ -​

A reduplicated adjective no longer has the basic functions of an adjective

很 (very) ~; it has the properties of a state word.
On the other hand, some original forms belong to other parts of speech.

The original form is a verb: 比划 (gesticulate), 勾搭 (gang up with), 踉跄

(stagger), 商量 (consult), 溜达 (stroll), 哆嗦 (tremble)
The original form is a noun: 疙瘩 (knot), 枝节 (branch and knot)6
The original form is an adverb: 切实 (earnestly)
The original form is a numeral and measure word: 半拉 (half)

Thus we can see that the AABB form that has the properties of a state
word belongs to word-​formation reduplication.
Furthermore, some adjective reduplications mean “either A  or B”, for
example, 好好坏坏 (either good or bad), 大大小小 (either big or small). This
belongs to syntactic reduplication.
36

36  Basic issues
3.  STATE WORD REDUPLICATION

For example: 雪白雪白 (snow-​white), 笔直笔直 (very straight), 通红通红 (red

through), 乌黑乌黑 (pitch dark), 喷香喷香 (extremely delicious-​smelling).
They are systematic, have the same functions (all have the properties of a state
word) but no compulsion; both parts are words. They should be regarded as
syntactic reduplication.

4.  MEASURE WORD REDUPLICATION

For example, 个个 (piece by piece), 条条 (sections after sections of), 天天 (day

by day), 次次 (time and time again)
The reduplication is systematic; the grammatical function of a measure
word changes. The reduplicated form has the functions of both a numeral and
a measure word, for example:

Measure word: 一条条道路 (sections after sections of a road)

Numeral and measure word: 九条道路 (nine sections of a road)
Measure word: 一次次摔倒 (fall over time and time again)
Numeral and measure word: 九次摔倒 (fall over for nine times)

However, because the basic functions of a measure word are kept (being
able to be modified by numerals, and the functions of increased numerals
and measure words are incomplete, the reduplicated measure words are not
regarded as word-​formation reduplication. Because there is no enforceability,
both reduplicating and reduplicated parts are words, and they should be
regarded as syntactic reduplication.

5.  NUMERAL REDUPLICATION

For example, 三三两两 (in twos and threes), 千千 (thousands after

thousands), 万万 (ten thousands after ten thousands), 许许多多 (many
many)
The reduplication of a numeral is not systematic and has some changes in
function. For example, the reduplicated numeral cannot modify a measure
word, and there is no systematic difference in meaning, thus being ought to be
regarded as word-​formation reduplication.

6.  NOUN REDUPLICATION

For example, 风风雨雨 (hardships), 方方面面 (respects after respects),

恩恩怨怨 (grievances after grievances), 条条框框 (regulations and
restrictions), 瓶瓶罐罐 (bottles and cans). Their reduplication is not sys-
tematic; some nouns before reduplication have no word form and should be
regarded as word-​formation reduplication.
 37

Basic issues 37
7.  LOCATIVE REDUPLICATION

For example, 上上下下 (up and down), 前前后后 (front and rear), 里里外外
(inner and outer), 左左右右 (left and right). Their reduplication is system-
atic; their function does not change fundamentally but has no enforceability;
both reduplicating and reduplicated parts are words, belonging to syntactic
reduplication.

8.  ADVERB REDUPLICATION

For example, Type A: 刚刚 (just), 常常 (very often), 白白 (fruitlessly), 早早

(very early). 渐渐 (gradually), 悄悄 (silently), 偏偏 (unexpectedly), 隐隐约约
(faintly), 的的确确 (very indeed), 陆陆续续 (successively)
Type B:  最最 (the very most), 永远永远 (forever and forever), 非常非常
(very much and very much)
Type A cannot be reasoned by analogy and has no uniform grammatical
meaning; some adverbs before reduplication are not words. This type can
only be regarded as word-​formation reduplication. In fact, Type B does not
belong to reduplication but to repetition: it can be repeated for more than two
times: very very good, ignore you forever forever forever. Repetition is neither
word-​formation nor morphological formation but the use of words.
Yang Chengkai (1991) thought that there is no need to group the
reduplicated form of a word as a part of speech and to distinguish between
word-​formation reduplication and syntactic reduplication. Indeed, there is
no need to classify syntactic reduplication as a word into a part of speech,
but word-​ formation reduplication should be stored in a word bank as
independent words.
Chinese essentially has no attached word-​formation. “X + zi (子), er (儿),
tou (头), hua (化)” are used to form words, such as 竹子 (bamboo), 尖子 (tip),
兔儿 (rabbit), 石头 (stone), 看头 (worth seeing), 液化 (liquefy), 淡化 (play
down). le (了), zhuo (着), guo (过) and men (们)7 form phrases (see Chen
Baoya, 2000) and are syntactic. But their functions are like word-​formation,
and the words thus formed should be regarded as identical units. For example,
来了一个人 (a person comes) should be regarded as 来 (come) having its
object 一个人 (a person).

2.3.2.5 With de (的)/​di (地) or without de (的)/​di (地)

In examining the grammatical functions of a word, a good many studies that
discuss parts of speech suggest that a word that takes de (的) or di (地) has the
grammatical function of the word itself. We do not agree with this.
Grammatically speaking, de (的) or di (地) can be analyzed into the suffix
of adverb de1 (的), the suffix of state word de2 (的) and the attached con-
stituent of a noun-​like phrase de3 (的) (see Zhu Dexi, 1961). A  constituent
that takes de1 (的) or de3 (的) has quite different grammatical functions.
For example, 逻辑 (logic) and 历史 (history) cannot function as adverbial
38

38  Basic issues
originally, but can function as adverbial after taking de1 (的). Moreover, they
can no longer be used as subject or object. 急躁 (irritable), 狡猾 (sly), 谦虚
(modest) cannot function as adverbial originally, but can do so as adverbial
after taking de1 (的). However, the addition of de3 (的) to 吃 (eat), 买 (buy),
做 (do) brings about changes in not only the grammatical function but also
meaning. Therefore, in our opinion, words that take de1 (的) and de3 (的) or
not, are not identical at all, and we cannot reckon that these words have the
grammatical function of de (的).
de2 (的) is rather special. Under usual circumstances, the functions of the
constituent before de2 (的) are basically the same, both having the properties
of a state word, for example, 干干净净 (very clean), 干干净净的 (very clean
plus de (的)). Only a constituent that is the reduplicated form of a monosyl-
lable adjective has different properties from de2 (的). There are two cases: (a)
the monosyllable adjectives such as 胖胖 (fat fat), 红红 (red red), 长长 (long
long) do not form words, but the constituent with de2 (的) thus formed has
the properties of a state word; (b)  the reduplicated forms of monosyllable
adjectives such as 好好 (good good), 慢慢 (slow slow), 大大 (big big) have the
properties of an adverb, but the constituent with de2 (的) thus formed has the
properties of a state word. We regard “胖胖 (fat fat) plus de (的)” and “大大
(big big) plus de (的)” wholly as one word; the reduplicated forms of mono-
syllable adjectives such as 胖胖 (fat fat) are regarded as not being a word;
the reduplicated forms of monosyllable adjectives such as 大大 (big big) are
regarded as adverbs and different generalization words of “大大 (big big) plus
de (的)”. Other constituents with de2 (的) and the constituents before de2 (的)
are regarded as one generalization word.

2.3.3  Brief summary

In classifying parts of speech, why should we discuss the identity of words?
The fundamental reason is that part-​of-​speech classification aims at general-
ization words, as shown in the following two respects:

1. All the functions of the same generalization word belong to its own and
cannot be classified into different units.
2. Different generalization words should be classified into different units,
their grammatical functions should be examined respectively, and they
should not be regarded as one combined unit.

2.4  Investigating the functions of a word

2.4.1  The materials used to investigate the functions

In investigating the functions of a word, we mainly rely on our language sense,
the written materials after the foundation of New China and contemporary
audio language materials such as radio broadcasting, movies and television.
 39

Basic issues 39
For a few words that rarely occur in materials after the New China founda-
tion, we use the materials that appeared after 1919. The materials before 1919
are not used to investigate the functions of a word.
Language is constantly changing; new usages and coinages may arise at
any time. If new usages and coinages are extensively used, strongly acceptable
and rather stable, then we use the new usages and some coinages to investigate
their functions. Otherwise, new usages and coinages are not used to investigate
their functions. For example, 投入 (input) is fundamentally a verb and cannot
be modified by 很 (very), but recently there is the expression 很投入 (very
input), whose meaning is different from the verb 投入 (input). The expres-
sion is extensively used and stable. Therefore, we determine that 投入 (input)
should be a verb and adjective concurrently. 专业 (profession) can only be
used as attributive and take de (的), thus being a distinctive word. However,
recently the expression 很专业 (very profession) has arisen, but it is not exten-
sively used and is not stable. This is only a flexible use, and we do not use it to
investigate its function. 太业余了 (too amateur) is not a fixed expression and
cannot be used to investigate its function.

2.4.2  Dealing with special uses

Some words have their special uses in addition to their ordinary uses. We
divide their parts of speech according to their ordinary uses, not according to
their special uses. The special uses include the following cases:

(1) Syntactic transfer reference as discussed previously. It is a temporary use

and thus is not used to investigate the functions of a word.
(2) An obvious elliptical use. For example, sometimes an adverb can function
as a predicate, which, however, obviously omits the head word. We thus
do not reckon that the use of an adverb as a predicate is its function. For
example, 我不 (I do not), 你赶快 (You hurry up!)
(3) A word is used as quotation. Any word if used as quotation may change its
function and behave like a noun, therefore not being usable to investigate
its function. For example, the auxiliary word de (的) cannot be used as
subject; however, when used as quotation, it can function as subject, for
example, “de (的)” is an auxiliary.
(4) A word is used to denote numerical conversion and calculation. Conversion
and calculation are special expressions and have their own methods of
expression, which are different from those of everyday Chinese, thus not
being usable to investigate their functions. For example, 两个五是十 (two
times five is ten); 把斤换算成克 (convert jin into gram).
(5) A word is used as a fixed expression, and then we do not use it to
investigate its function. For example, the fixed expression 不男不女 (not
man not woman) cannot be used to claim that the words 男 (man) and
女 (woman) can be modified by 不 (not). The fixed expression 硬着头皮
(brave all rebuff) also cannot be used to treat 硬着 (brave) as verb.
40

40  Basic issues
(6) A few exceptions. Some words are historical legacies; some have no clear
source. For example, 忽然之间 (suddenly), 短期内 (within a brief period)
and 很 (very) are used as complements.

Notes
1 In fact, it is difficult to divide the boundary line between the compound word
formed by combining root words and phrases not only in Chinese but also in other
languages such as English. For example, “airmail” and “air mail”. In the same
dictionary such as Webster’s Ninth New Collegiate Dictionary, the following pairs
are correct: full moon/​half-​moon, back stroke/​breaststroke, morning coat/​tailcoat,
water buffalo/​watermelon.
2 Here, 国 (country) in classical Chinese collocates 与 (and) in classical Chinese; if 和
(and) in modern Chinese is used to replace 与 in classical Chinese, then the sentence
will be ungrammatical.
3 The technical language hierarchy often uses classical Chinese elements; 叶 (leaf),
鳄 (crocodile), 胸 (chest), 肩 (shoulder) are also words in the classical Chinese
­hierarchy used in the everyday Chinese.
4 See The Modern Chinese Learning Dictionary.
5 See The Modern Chinese Usage Dictionary.
6 It means 支支吾吾 (hem and haw). For example, a line by Chen Jinfu: 枝枝节节,
meaning somewhat hesitant to say something (Hong Shen’s drama: Wukui Bridge).
It also means fragmentary, for example: What I love about Peking is not something
枝枝节节 (fragmentary) but a whole portion of history that unites my soul (Missing
Peking by Lao She).
7 A few combinations such as 为了 (for), 除了 (except), 人们 (people) and 他们 (they)
are words; except for these, others should be regarded as phrases.
 41

3 
Possibilities and purposes of
classifying parts of speech

3.1  Possibilities of classifying parts of speech

3.1.1  Controversy on the existence of Chinese parts of speech

Traditional Chinese linguistics does not deal with grammar and, of course,
not parts of speech. Though traditional Chinese linguistics makes distinctions
between notional words and functional words, motional words and still
words, dead words and living words, these are only semantic distinctions and
are not the strictly grammatical classification of parts of speech. The real
classification of Chinese parts of speech began with Ma’s Grammar (1898) by
Ma Jianzhong. Drawing on the Western grammars, he classified the ancient
Chinese parts of speech into nine parts of speech as follows: nouns, pronouns,
motional words, still words (adjectives), descriptive words (adverbs),
prepositions, conjunctions, auxiliary words, interjections.
But in the 1950s, Gao Mingkai (1953) put forward his opinion that there
is no classification of parts of speech in Chinese; thus, a debate was triggered
by On Chinese Language (1952), written by a former Soviet Union scholar
named H.  И.KoHpaд (see Gong Qianyan, 1997). KoHpaд maintained that
it was wrong that H. я. Mapp of the Soviet Union and the Western linguists
N.  Maspero and B.  C. Karlgren believed that Chinese was a primitive lan-
guage. He pointed out that Chinese has a rich vocabulary, is expressive and
is one of the most developed languages in the world. Maspero maintained
that Chinese has no grammatical categories and parts of speech; his main
evidence was that Chinese has no morphology. Other Soviet Union scholars
and Karlgren had similar viewpoints. However, KoHpaд refuted them from
the starting point that it is proved that Chinese has a rich morphology and,
hence, has its parts of speech.

1. Affixation:  prefixes or suffixes:  -​zi (子), -​er (儿), -​tou (头), -​jia (家), for
example, 资本家 (capitalist); di-​(第-​), wu-​ (无-​), for example, 无条件地
(unconditionally); zong-​(总), for example, 总领事 (councilor general);
fan-​ (反-​), for example, 反革命 (counterrevolutionary).
42

42  Possibilities and purposes

2. Stress: when the latter syllable of a double-​syllable word is stressed, it is
a verb; if the former syllable is stressed, it is a noun, for example, 写字
(write), 道路 (road).
3. Change in tone:  for example, 好 (hǎo, adjective, meaning “good”), 好
(hào, verb, meaning “like, love”).
4. Lexicalization of grammatical forms:  The analytical form of possible
voices: 笑 (laugh), 可笑 (laughable), 可笑的 (laughable).
The lexicalization of passive voice: 拘留 (detain), 被拘留 (be detained),
被拘留的 (detainable).
The morphology of cases of a noun: 桥 (bridge, MOCT), 桥的 (of the
bridge, MOCTa)
臂 (arm, KИCTb), 用臂 (with the arm, KИCTbю).
他 (he, OH), 给他 (give him, eMy, dative).
The morphology of tenses of a verb:  来 (come), 来了 (has come); 来
(come), 要来 (will come).

In addition to the above morphological evidence, a syntactic marker also

serves as evidence that Chinese has parts of speech because it signifies the cap-
ability of a word to occupy a certain syntactic position in a sentence.
Therefore, KoHpaд concluded the following

1. The modern Chinese language, the language of the Han nationality, is

famous for its incomparably rich vocabulary.
2. Chinese vocabulary consists of words formed by morphology, whereas
the word formation methods are diverse.
3. Chinese words can be classified into various parts of speech and contain
all the concepts needed by a language. Chinese has its own special word
formation methods.
4. When words are combined into speech, their grammatical forms come
into being and have rich and diversified characteristics and meanings.
5. Chinese syntactic structure has its own precise rules, which reveal all the
relations among words and phrases in a sentence.
6. Such highly developed vocabulary and grammatical structures enable
Chinese to express any ideas of the people at the highest cultural level.

Gao Mingkai supported Maspero’s view that “Chinese has no parts of

speech” and objected to KoHpaд’s view. He proved his idea in the following way:

Major premise:  the classification of parts of speech is based on the

morphology of a word.
Minor premise:  a Chinese notional word has no morphology (or its
morphology is not enough to classify parts of speech).
Conclusion: therefore, there is no way to classify the parts of speech of a
Chinese notional word.
Gao Mingkai proved that Chinese has no morphology in the following  way:
 43

Possibilities and purposes 43

1. Affixation is no morphology. Indo-​European languages use word
roots to express their forms. For example, lego (I’m reading now),
legis (You’re reading now) and legit (He’s reading now) in Latin. In
these examples, word roots and morphology are combined; the word
root leg-​cannot be independent. Latin has its genitive case:  liber
(book, subjective case), -​ libri (of the book/​ the book’s, genitive
case), but French uses the preposition “de”, for example, de le livre,
meaning “of the book”, to express the same meaning. However, “de”
is not in the genitive case, for it is separable from “livre”. The affix-
ation elements in Chinese like zhe (者), men (们), de (的) also belong
to this type of constituent.
2. Changes in tone do not belong to morphology because they can
only differentiate varied meanings rather than parts of speech.
For example, 好 (hǎo, adjective, meaning “good”), 好 (hào, verb,
meaning “like or love”), 钉 (ding, noun, meaning “nail”), 钉 (dìng,
verb, meaning “put a nail into”), 背 (bēi, verb, meaning “carry on the
back”), 背 (bèi, noun, meaning “back”; verb, meaning “recite”).

Gao Mingkai summarized that for four reasons, Chinese has parts of
speech; however, none of them is valid.
First, if Chinese had no parts of speech, then Chinese would be regarded as
a kind of low-​level language. However, whether there is classification of parts
of speech or not is not an adequate basis for judging the level of a language.
Though Chinese has no classification of parts of speech, its speakers have the
notions of “noun, verb, adjective”. The only thing is that these concepts are
expressed by words rather than by grammar. Language development facts
show that the morphology of Indo-​European languages tends to be simpli-
fied, but we cannot say that they are low-​level languages.
Second, the statement that Chinese has parts of speech is based on the
meaning of a word instead of its form. But Chinese has no special forms to
show its parts of speech, so the conclusion is that it surely has no parts of
speech.
Third, when somebody believes that Chinese has parts of speech, their
belief is based on the evidence that Chinese has a morphology. However, they
are not clear about what morphology is, thinking that whatever constituent
that follows or precedes a word root is morphology. As a matter of fact, zhe
(者), de (的), le (了) and so forth in Chinese are just grammatical means (syn-
tactic means); they are merely functional words rather than morphology.
Fourth, when somebody believes that Chinese has its parts of speech, their
belief is based on the fact that it has changes in tone. But the changes in tone
represent different meanings, and the reason why they think there is a diffe-
rence in parts of speech after changes in tone is still based on the meaning of
a word.
KoHpaд’s belief that Chinese has a morphology actually reveals that he
looked at Chinese from the perspective of Indo-​European languages. For
44

44  Possibilities and purposes

example, he regarded 给他 (give him) as the dative form of a noun, and
用臂 (with the arm) as the instrumental case of a noun. In Gao Mingkai’s
opinion, most of the reasons why Chinese has no morphology are valid.
But what a morphology is, is still a question that is difficult to answer.
Judging from the criteria that a word root is part of a morphological form,
“books”, and “looked” cannot be regarded as a morphology, either. The
difference between Gao Mingkai and KoHpaд lies in whether Chinese has
a morphology or not, and the common point is that both agreed that the
nature of a part of speech is morphology. Because the nature of a part of
speech is attributed to morphology and there are no definite criteria to
judge what morphology is, it is hard to decide whether Chinese has parts
of speech or not.
The publication of Gao Mingkai’s essay resulted in a series of criticisms.
His critics maintained the following views:

A. They agreed that the basis for classifying parts of speech is morphology
(in other words, they agreed with the major premise), but they believed
that Chinese has a morphology. Therefore, it has parts of speech. Such
critics are represented by Lu Zongda (1955) and Yu Min (1955). Yu Min’s
Chinese morphology is shown in the following table:

Original Reduplication Change Number Meaning Part of

word pattern of words produced by speech
redupli­cation

人 (rén, 人人 (rén′rén, Stress is shifted 1 every, each noun

meaning meaning and a non-​
person) each person) syllabic “r”
is suffixed
to noun and
sometimes
verb
好 (hǎo, 好好儿 Stress is shifted 1 very adjective
meaning (hǎohâo, and a non-​
good) meaning very syllabic “r”
good) is suffixed
to noun and
sometimes
verb, and
there is
change in
tone
飞 (fêi, 飞飞 (fêifei, weakened tone 1 one time verb
meaning meaning fly
fly) one time)
三 (sân, 三三 (sânsân, zero 2 multiple numeral
meaning meaning
three) many
 45

Possibilities and purposes 45

According to Yu Min and Lu Zonda, the table shows that nouns, adjectives
and verbs have their respective rules of change. Therefore, they used these
rules to classify the parts of speech of the three types of words:

B. They agree that the basis for classifying parts of speech is morphology in
its narrow and broad senses, and that it can be used to classify parts of
speech accordingly. Such critics are represented by Wen Lian and Hu Fu
(1954), Б. Г. Myдpob (1954).
Morphology in its narrow sense: affix. For example, -​zi (子), -​er (儿), -​tou
(头) and reduplication.
Morphology in its broad sense: the capability of words to combine with
each other. For example, 人 (person), 一个人 (one person), 笔 (pen),
三只笔 (three pens), 笑 (smile), 不笑 (not smile), 会笑 (can smile), 快
(fast), 很快 (very fast).
C. The function of a word is the basis for classifying parts of speech, and its
meaning should be taken into account at the same time. This is represented
by Cao Bohan (1955), who pointed out that “we feel that the criteria for
classifying parts of speech must be decided by the function of a word in a
sentence, and that the classification must be based on its meaning”. Why
must it be done in this way? The main reason is that parts of speech and
the constituents of a sentence do not fit well with each other (namely, they
do not correspond well with each other). In such a case, word meaning is
needed to help classify parts of speech. For example, it is unnecessary to
classify words that function as subject and object into two parts of speech
because nouns functioning as subject and object all mean the names of
things, whose attributes are the same.

Moreover, some scholars like Wang Li (1955) held that meaning, morph-
ology and syntax should be jointly used to classify parts of speech.
Gao Mingkai (1954, 1955) refuted this as follows:

1. The function of a word and its capability to combine with other words
cannot be used to classify parts of speech. Grammar can be classified
into morphology and syntax, but parts of speech belong to morphology.
The so-​called parts of speech are the classification of words functioning
as the building block of a language (in a word-​bank, words are not
combined into sentences); they do not refer to the position or function of
a word in a sentence. If we classify parts of speech according to morph-
ology in its broad sense, then we may commit the following mistakes:
a. The borderline between morphology and syntax is not clear.

Morphology in its broad sense and the function of a word fall into
the category of syntax rather than morphology. The part of speech
of a word and its function in a sentence are two different concepts.
For instance, in the past, wood was often used as the beam of a
house, but now steel bars are used instead; thus, steel functions as
46

46  Possibilities and purposes

wood, but steel is not wood. A part of speech is similar to the distinc-
tion between wood and steel, which has a static nature, but a subject
and a predicate are similar to the distinction among a pillar, a beam
and a wall, and have a dynamic nature. A part of speech refers to the
attribute type of a word itself rather than its functional type. Wood,
brick and steel are different from each other when they are stored in
a warehouse and before they are put into use. They do not need to
come into use for their distinction. If that is the case, it is just like
what Li Jinxi observed, “The part of speech of a word should be
judged by sentence, and without sentence, there is no part of speech”.
b. Because parts of speech of a word do not correspond to its functions,
it is impossible to classify them according to the functions. For
example, in English, a noun, the present participle of a verb and
an infinitive can serve as the subject of a sentence. In Chinese, the
problem is even more serious, not only the so-​called noun, verb and
adjective can occur at the position where a subject can do, but also
the so-​called morphology in its narrow sense cannot work effect-
ively. For instance, 不去 (not go) is a so-​called verb, while 不红 (not
popular) is a so-​called adjective.
2. Morphology in its narrow sense cannot be used to classify parts of speech.
The existence of a morphology in Chinese should be recognized, but it is
not enough to classify parts of speech. Morphology can be classified into
two types: (1) word formation type (new words are formed by inflections);
(2) inflectional type (changes in a word form are special signs for changes
in parts of speech or express a particular grammatical category). Only
the inflectional type can distinguish between parts of speech. But Chinese
morphology belongs to the word-​formation type: 者 (zhe), 子 (zi), 儿 (er)
are all word-​formation affixations, which have nothing to do with the
classification of parts of speech. Words like 了 (le) and 们 (men) are func-
tional words rather than a morphology.

Gao Mingkai stressed that the statement that Chinese has no parts of
speech is based on grammar; however, Chinese speakers use concepts to tell
difference between things, actions and properties. The old Slavic language
has the category of even numbers, but Russian does not. However, Russian
speakers also have the concept of even numbers.
The apparent reason why Gao Mingkai held that Chinese has no parts of
speech is that he looked at the issue from the perspective of Indo-​European
languages, believing that parts of speech can be classified only according to a
morphology. But the fundamental reason is that he failed to realize the nature
of parts of speech. Their nature is not the morphological type, just like male
and female are not distinguished by their physical features or their clothing.
The critics of Gao Mingkai also failed to realize the nature of parts of speech,
either continuing to regard morphology as their nature, insisting on finding
some examples of Chinese morphology (Lu Zongda, Yu Min), or regarding
 47

Possibilities and purposes 47

the essence of a word as the function of a word (including the combination of
words and syntactic constituents) (Wen Lian, Hu Fu, Cao Bohan). Actually,
function (relations and positions in combination) is not the nature of parts
of speech, either. Judging from this point of view, Gao Mingkai’s viewpoint
is correct:  the part of speech and the function of a word in a sentence are
different concepts; a grammatical function is not the nature of a part of
speech. It is exactly for this reason that he was not convinced.
As Lv Shuxiang (1954) summed up, “if one thing or several things can be
used to classify parts of speech, what does it matter if morphology cannot be
used? In other words, we can put Gao Mingkai’s minor premise aside for a
while and change his major premise to do experiments”.
Later, Gao Mingkai (1960) changed his viewpoint in a certain way and held
that a “part of speech is, in essence, a matter of grammatical meaning; it is the
grammatical meaning expressed by form” (p. 37). “Morphology and the com-
bination capability of a word (including the syntactic combination capability,
namely the syntactic function) both express grammatical meanings through
some forms” (p. 37). “In the past, I just regarded morphology as the form to
express the meaning of a part of speech. This view is too narrow, and now
I should say that the morphological change, combination capability and syn-
tactic function of a word and so on are all the exterior signs for the meaning
of a part of speech. The problem is that no matter from what perspective,
Chinese notional words show their multiple types of meaning, including the
part-​of-​speech meaning without displaying the fixed characteristics of a part
of speech” (p. 38). “A word having multiple parts of speech means that it has
no parts of speech” (p. 38). Gao Mingkai admitted that “Chinese has the cat-
egory of a part of speech” and that a grammatical function may reflect the
properties of a part of speech. But because a Chinese part of speech has mul-
tiple functions, “a word can be simultaneously used as noun, adjective and
verb” (meaning that a notional word can occur at the position where subject,
predicate or attributive can). Hence, every notional word has multiple parts of
speech, and this is the same as saying that there is no part of speech.
Gao Mingkai’s arguments can be summed up as follows:

1. A part of speech is, in essence, the type of grammatical meaning a word

expresses.
2. The morphology and grammatical function of a word express its gram-
matical meaning in a certain form.
3. Both the morphology and the grammatical function of a word express
the exterior form of a part of speech.
4. Both the morphological and the grammatical functions of Chinese
notional words demonstrate multiple types of part-​of-​speech meaning
without displaying the fixed characteristics of a part of speech.
5. A word having multiple parts of speech means that there is no part of
speech.
6. Therefore, a Chinese notional word does not have its part of speech.
48

48  Possibilities and purposes

Zhu Dexi (1960) pointed out that “the classification of parts of speech
according to morphology is just a method or a means, which is possible
because it is still based on the syntactic function of a word. Morphology is
merely the indication of a function”. Although the classification of parts
of speech according to syntactic constituents is done from the perspective
of function, “because the criterion is chosen too carelessly and the method
too simple, it not only fails to classify parts of speech but also reaches the
conclusion that a word does not have its fixed part of speech. This method
has the fundamental mistake that it assumes that there is a close correspond-
ence between sentence constituent and part of speech . . . but as a matter
of fact, the relations between a sentence constituent and a part of speech
are complicated” (p. 42). The classification of parts of speech according to
sentence constituents concludes that a word has no fixed parts of speech,
there being no part of speech in essence. This is one of the two important
pieces of evidence for Gao Mingkai’s insistence that Chinese has no parts of
speech. Zhu Dexi pointed out that a word occurring at the position of subject
is not necessarily a noun. For example, the adverb 不 (not) in the sentence
哭是不好的 (crying is not good) can be added before 哭 (cry); this proves that
哭 (cry) is a verb (p. 43). Therefore, it is wrong to believe that Chinese has no
parts of speech simply because a sentence constituent is not enough to classify
parts of speech.
The discussions on parts of speech in the 1950s mainly centered on
whether Chinese has parts of speech or not. Earlier, Gao Mingkai, Liu
Zhengtan and Li Xingjian held that Chinese does not have parts of speech,
whereas others held the basic conclusion that Chinese has parts of speech,
which, however, should be classified according to the function of a word.
This raises two critical questions:  (1) What is the criterion for classifica-
tion? Because a Chinese part of speech does not correspond with a sen-
tence constituent, the question is how to choose the classification criterion;
(2) What is the essence of a part of speech? Because these two questions
have not been answered, there are underlying disagreements. Moreover,
the affirmative part of Gao Mingkai’s arguments (for example, Arguments
1 and 3)  failed to be accepted; instead, it was rejected together with the
mistaken part.
Zhu Dexi (1985a) maintained that Chinese has two characteristics:  (1)
a part of speech does not correspond with a syntax constituent; (2)  the
principles for constructing a sentence and a phrase are the same. Xu
Tongqiang (1994a, b) held that the discussions on parts of speech in the
1950s did criticize the approach of looking at Chinese from the perspective
of Indo-​European languages (a Chinese notional word does not have parts of
speech because it has no morphology). This, however, opened up a new road
for the popularity and development of another Indo-​European language per-
spective. Parts of speech in Indo-​European languages correspond well with
sentence constituents; hence, it is necessary to classify their parts of speech.
Since Chinese parts of speech correspond with multiple sentence constituents,
 49

Possibilities and purposes 49

Chinese has no parts of speech. One sentence construction of Indo-​European
languages is “subject + predicate”; hence, their parts of speech correspond
with sentence constituents. One Chinese sentence construction is “topic +
description”, which cannot be analyzed as “subject + predicate”; hence, it is
impossible to classify Chinese parts of speech. The classification of Chinese
parts of speech looks at Chinese from the perspective of Indo-​European
languages.
As a word has multiple functions and corresponds with multiple sentence
constituents, it is impossible to classify its part of speech. Such a viewpoint is
essentially identical to that of Gao Mingkai (1960). Namely, he assumed that
there is a strict correspondence between a sentence constituent and a part of
speech. If there is no such correspondence, then there is no classification of
parts of speech. Zhu Dexi (1960) criticized this viewpoint. Its root cause is the
failure to understand the nature of a part of speech, which is viewed as the
capability of a word to function as sentence constituents.
Lv Shuxiang’s criticism of Gao Mingkai’s morphological classifica-
tion of parts of speech also applies to classification according to sentence
constituents. If one thing or several things can be used to classify parts of
speech, what does it matter if morphology cannot be used? We can see that
the viewpoint that Chinese has no part of speech mainly appears in two forms.
In the 1950s, morphology was seen as the sole criterion for classifying parts of
speech, and since Chinese has no morphology, it has no parts of speech. After
the 1960s, the correspondence of a sentence constituent with a part of speech
was established. Thus, since Chinese parts of speech do not correspond with
sentence constituents, there is no way to classify them.

3.1.2  The evidence that Chinese has parts of speech: selectional

restriction of grammatical positions on words
Grammatical positions have selectional restriction on words, as shown by the
following example:

˄˅ 䘉 а ᵜ Җ
DE
FG
H I
The hierarchical analysis shows that the example has a total of six gram-
matical positions, the five positions of b, c, d, e, f can be substituted by the
following words:

Position b:  纸 (paper), 鱼 (fish), 石头 (stone), 花生 (peanut), 苹果

(apple) . . .
Position c: 那 (that), 每 (each), 任何 (any), 某 (certain), 另 (another), 惟一
(only) . . .
50

50  Possibilities and purposes

Position d: 不少 (not a few), 一切 (all), 许多 (many), 俩 (both) . . .
Position e:  两 (two), 三 (three), 四 (four), 十 (ten), 半 (half), 几
(several) . . .
Position f: 张 (sheet), 条 (bar), 块 (piece), 粒 (grain) . . .

Such substitutions have restrictions. For example, words in Position b

cannot substitute words in Position c, and words in Position c cannot sub-
stitute words in Position f, and vice versa. That is to say, when words are
combined, they are arranged not randomly but in an orderly way. Such order-
liness is shown by the selectional restriction of grammatical positions on
words. Different grammatical positions allow different words to enter. This
reveals that a word itself has different properties, which can be used to classify
words into different parts of speech.
Gao Mingkai and Xu Tongqiang mainly held that it is difficult to classify
Chinese words into nouns, verbs and adjectives. With the above method, we
can do so:

no/​not ~ very ~ very ~ <object> subject/​object

Verb + −/​+ −/​+ −/​+

Adjective + + − −/​+
Noun − − − +

Gao Mingkai maintained that the syntactic function of a word cannot

be used to classify parts of speech. He admitted that words have different
functions in different sentences, but that a part of speech belongs to the cat-
egory of morphology, whereas a syntactic function belongs to the category
of syntax. A part of speech belongs to a type of a back-​up unit rather than
the type in use. Therefore, the syntactic function cannot be used to classify
parts of speech. Since a word has different syntactic functions, we have to ask
why it has such differences. The reason still lies in the fact that a word itself
has different properties. Though a syntactic function is not a part of speech,
it reflects differences in the latter. This can be compared to wood and steel.
Though their function is not identical, they both can be used as a beam, but
wood can be used to make fire, while steel cannot. In contrast, steel can be used
to make artilleries, while wood cannot. Their differences in function still result
from their differences in properties. If we cannot directly tell them apart, we
can distinguish them by their functions. That is, we should distinguish between
the nature of things and the means to distinguish them. To distinguish them,
we tend to adopt the exterior form or function as the means rather than dir-
ectly recognize their essential properties. For example, we recognize something
as wood by its appearance (wood texture) or by the feeling that it is heavy.
The use of grammatical functions as the criteria for classifying parts of
speech does not mean that every difference in grammatical function can
 51

Possibilities and purposes 51

distinguish parts of speech. We just choose some of the grammatical functions
to serve as the classification criteria. We will discuss this in detail later.
Judging from the universal characteristics of languages in the world, clas-
sifying parts of speech, especially classifying words into nouns and verbs, is
common. Some scholars believe that some languages, for example, Nootka,
do not have the classification of nouns and verbs (see Schachter, 1985):

(2) a. Mamu•k-​ ma   qu•?as-​?i
   Work (present tense)  man this
    This man is working.
b. Qu•?as-​
ma    mamu•k-​?i (mamu•k: statement→reference)
    Man (present tense) work  this
    The one who is working is a man.

Tagalog:

(3) a. Nagtatrabaho ang  lalaki
   is working (topic)  man
    Man is working.
b. Lalaki ang  nagtatrabaho (nagtatrabaho: statement → reference)
    Man  (topic) is working
    The one who is working is a man.

Actually, this is just the transfer reference of a verb. A noun serves as the
predicate of a judgment sentence. This is not enough to explain that there is
no classification of nouns and verbs, however. For example, when a verb has
a transfer reference and serves as a subject or object, the definite determiner
“?i” is required (similar to the transfer reference of an adjective in English).
Hence, Example (4) in the following is invalid. Furthermore, the meaning of
a transfer reference is produced only in a special context.

(4) * Qu•?as-​ma mamu• k

  man (present tense) work
  One is working is a man.

3.2  Purpose for classifying parts of speech

The purpose for classifying parts of speech is actually concerned with the
attitude toward language. Householder (1952) classified the then structural
linguists into two schools: the Hocuspocus group and the God’s Truth School.
The attitude of the Hocuspocus group to language structure is that language
is merely a pile of messy materials, and that a linguist’s task is just to arrange
and combine them together so as to work out a structure; hence, structure
relies on man’s arrangement to some extent.
52

52  Possibilities and purposes

The attitude of the God’s Truth School to language structure is that the
structure of a language is inherent and lies in its materials, and that the task
of a linguist is to find out such a structure and describe it as clearly, econom-
ically and precisely as possible.
Z.S. Harris and J.R. Firth belong to the Hocuspocus group, and K. L. Pike
belongs to the God’s Truth School.
Today’s argument on the purpose for classifying Chinese parts of speech
and whether they are objective is similar to the attitude of the above two
schools toward language.
A part of speech is based on the properties of a word, which exist object-
ively in language. To classify parts of speech, one should first determine
the properties of a word in that language and how many properties a word
contains, and then one can consider how to classify it into parts of speech
according to its properties.
The classification has two types:  natural classification and artificial clas-
sification. Classification according to the basic characteristics of an object is
called natural classification, while artificial classification is opposite to natural
classification. Artificial classification tends to serve a particular practical pur-
pose, whereas natural classification mainly reveals natural laws but does not
serve a practical purpose.
From the perspective of natural classification, parts of speech are summed
up from syntactic structures rather than artificially determined in advance for
the convenience of grammatical analysis. Judging from the processes of clas-
sifying parts of speech, syntactic rules exist first and then the classification.
So, the convenience of grammatical analysis is not the purpose for, but the
result of, classifying parts of speech. The purpose is to reveal the nature of
a word itself and to build a general reference system. In other words, a lan-
guage is not chaotic but well structured and has its natural order independent
of a linguist. A word has its own grammatical positions in this natural order.
Our views on language belong to the God’s Truth School. We do not object
to other scholars for their adopting the language views of the Hocuspocus
group, but this book adopts those of the God’s Truth School.
Therefore, from the perspective of natural classification, we should not con-
sider how to make part-​of-​speech classification helpful for syntactic analysis;
instead, we should do so as it should be done and, as a matter of fact, not take
the convenience of syntactic analysis into consideration. For example, “the
old” in English can serve as a subject; for the convenience of syntactic analysis,
“old” can be classified into noun, as was done in the past. This is indeed con-
venient for syntactic analysis: noun + verb = sentence. But this analysis is not
true to reality; in fact, the “old” here still has the properties of an adjective:

The (extremely) old need a great deal of attention.

The (very) best are yet to come.
He is acceptable to both (the) young and (the) old.
The number of jobless is rising.
 53

Possibilities and purposes 53

From the perspective of natural classification, part-​of-​speech classification
for the purpose of syntactic analysis may easily lead to wrong parts of speech,
and the so-​called convenience thus obtained may be false. For example, Li
Jinxi’s loaned parts of speech and his statement (1924) that the part of speech
of a word is based on a sentence may be very convenient for syntactic analysis.
A verb becomes a noun when it serves as a subject; a noun becomes an adjec-
tive when it serves as an attributive. Such treatment can simplify syntactic
rules, yet this is convenient but not true to fact. At least a large number of
verbs in the position of a subject still have the properties of verbs. Actually,
this is not convenient at all because such treatment changes the question into
judging whether a word is a noun or a verb. This is still as difficult as before.
Only when we deal with a word that belongs to several parts of speech sim-
ultaneously should we consider the convenience of grammatical analysis (see
Chapter 7).
Classification not only discovers an objective natural order but also
establishes a general reference system, which refers to such a system that
expresses multiple relationships among things or phenomena. It has the char-
acteristic that, though it may be established on the basis of one characteristic,
it can explain many other relationships (Forey, 1983:  152–​153). Biological
classification is such a system (which reveals not only the evolutionary rela-
tionship of living things but also their habitus and physical characteristics).
The classification for the Periodic Table of Elements also falls into such a
system (which reveals both the chemical and physical properties of an
element). The classification of parts of speech also belongs to such a general
reference system. Of course, in specific applications, there are some changes
in how elaborately detailed the classification is. The strategies for classifying
conversional words and some minor parts of speech may be ignored, but the
overall pattern should be kept.
Many people do not think about universality and believe that the purpose
for classifying parts of speech is for the convenience of grammatical analysis.
Therefore, they think that different types of grammar (grammar for experts,
for computing and for foreigners to learn Chinese) should have different
systems of parts of speech. In fact, behind these different systems, there exists
a fixed and universal system, and the changed system is just the temporary
adaptation of the universal system to a circumstance.
54

4 
Essence and expressional functions
of a part of speech

4.1  The paradoxes of distribution nature theory

4.1.1  Viewing distribution as the essence of a part of speech

The essence of a part of speech is concerned with what it is.
One view holds that a part of speech is concerned with distribution, and
that its essence is distribution. This has been a common view since structural
linguistics was born and is mainly explained in two ways: (1) directly explained
with the distribution view; (2)  explained with Saussure’s syntagmatic and
paradigmatic relations in addition to the distribution analysis according to
US descriptive linguistics.
US descriptive linguistics links distribution with word classes. Bloomfield
(1926) held that the position occupied by a form is its function, and that all
the forms that share the same function make up a word class. The largest word
class in a language is the part of speech in the language.
Saussure’s Course of General Linguistics points out that language is a form
but not an entity. It is so called because the value of a linguistic constituent
is not decided by the constituent itself but by relations among constituents.
There are two types of relations among linguistic constituents: syntagmatic
relation and associative relation. The syntagmatic relation means that lin-
guistic constituents combine one with another, for example, in French:

Dieu est bon (God is kind).

Re-​lire (reread).

The associative relation means that the constituents that share common
things cluster into classes through association in a human’s memory. For
example, the French word enseignement (education, noun) makes one asso-
ciate enseigner (educate, verb) because they are cognates. Changement
(change, noun, the suffix -​ment is the same as -​ment in enseignement), educa-
tion (education, noun, synonymous). Enseignement has associative relations
with these words.
 55

Essence and expressional functions 55

Saussure’s associative relation is not the relation in the sense of distribu-
tion; it mainly refers to a cognate relation, a similar word-​formation relation,
a synonymous relation and so on.
To avoid the notion that the term “associative relation” may have a psy-
chological meaning, later scholars use the term “paradigmatic relation” (see
Cheng Zenghou, 1988).
Furthermore, Chen Wangdao (1941, 1942, 1978) combined paradigmatic
relation with distribution analysis: constituents make up larger combination
units; the constituents located in the same grammatical positions in a com-
bination unit form paradigmatic classes (in terms of words’ paradigmatic
relations, a paradigmatic class is a part of speech); however, combination
selects members from a paradigmatic class. That is to say, a combination unit
is a sequence of parts of speech. For example:

Combination relation

p │ 我   读   书 (I read books)
a │ 他   看   报 (He reads newspaper)
r │王同志  写  文章 (Comrade Wang writes articles)
a
d a    b    c
i
g
m
a
t
i
c  noun  + verb + noun

There are three possible scenarios of distribution in the idea that words
having the same distribution form a part of speech:

1. Single-​item distribution (for the convenience of explanation, distribution

here refers to the grammatical position a constituent occupies, differing
from the definition of Harris). The definition of a part of speech in terms
of paradigmatic relation belongs to this scenario; that is to say, words
having the same function belong to the same part of speech.
2. Overall distribution. It means that all the distributions are the same,
which is basically the view held by Bloomfield.
3. Partial distribution. Words having in common partial distribution
characteristics are clustered into one part of speech.
56

56  Essence and expressional functions

No matter in what sense the distribution is, the viewpoint that a part of
speech is of distribution type cannot be upheld. The viewpoint that a part of
speech is of a distribution type is not self-​contained. In the following, we will
prove this.

4.1.2  Paradoxes of the single-​item distribution view

The explanation for this is threefold.

1. Words having the same single-​item distribution do not necessarily belong

to the same part of speech, for example:
(1) Adverb-​headword: 很大 (very big); 拳头大 (the fist is big)
Subject-​predicate:  今天晴 (today is fine); 今天晴天 (today is a
fine day)
Subject-​predicate: 人好 (the person is good); 去好 (to go is good)
Predicate-​object: 喜欢吃 (like to eat); 喜欢酒 (like wine)
For another example, words that can function as subject include
the words that belong to different parts of speech, such as
电话 (telephone), 认真 (serious), 休息 (rest), 许多 (many),
慢性 (chronic) and 十 (ten).
2. Words falling into the category of the same part of speech do not neces-
sarily have the same single-​item distribution. For example, 年事 (person’s
age) and 现年 (current year) can only function as subjects; 剧毒 (high tox-
icity) and 泡影 (zilch) can only function as an object. They do not have
the same single-​item distribution, but all are nouns.
3. There are quite a huge number of grammatical positions. It is hard to
tell clearly how many grammatical positions there are exactly; however,
words that can enter into each and every grammatical position are quite
different. If a single-​item distribution represents one part of speech, then
there may also be a huge number of parts of speech, many of whose
members are conversional. The classification of words into their parts
of speech according to single-​item distribution makes a part of speech
have one-​to-​one correspondence with a syntactic constituent; thus, nearly
every word belongs to several parts of speech. Therefore, such a classifica-
tion makes no sense.

Hence, the view that words having the same single-​item distribution fall
into the same part of speech is not valid.

4.1.3  Paradoxes of overall distribution

1. The more grammatical positions for classifying parts of speech there are,
the larger numbers of parts of speech may exist. Even if the grammatical
position for classifying parts of speech is a little different, the parts of
 57

Essence and expressional functions 57

speech thus classified may be quite different. Therefore, the classification
of words into their parts of speech according to overall distribution has
almost no possibility of definite parts of speech.
2. Even if we can discover all the grammatical positions in a language to
classify parts of speech, we may find that words whose distributions are
completely identical barely exist. If we firmly believe that words having
the same distributions belong to one part of speech, then nearly every
word belongs to a part of speech, which is almost like saying that there is
no part of speech.

To demonstrate this point of view, we select 60 words and examine their

distributions to 36 grammatical positions, with the examination results
given in Table  4.1-​1 (see appendices). The results show that among the 60
words, only four pairs, such as 人 (person)  –​ 桶 (barrel), 岁数 (age)  –​ 举动
(action),慢性 (chronic) –​ 私人 (private) and 究竟 (on earth) –​亲自 (person-
ally) have completely identical distributions. If parts of speech are classified
according to overall distribution, then the 60 words should be classified into
56 parts of speech. Chen Xiaohe (1999) tried to classify parts of speech with
the overall distribution of words having the capability to function as syntactic
constituents, resulting in more than 1,000 parts of speech. Such a classification
of parts of speech follows a rather strict distribution. The word classes thus
classified belong to the “distribution classes of words” rather than “parts of
speech”. Obviously, a “part of speech” is a class in its specific sense, and not
all “classifications of word classes” belong to a part of speech. The classifica-
tion of parts of speech is based on categorization. That is to say, the classified
parts of speech are different in their categories and patterns. The classification
of words solely according to the same or different distribution and the classifi-
cation of words into monosyllable/​double-​syllable words or simple/​compound
words do not constitute parts of speech. In fact, the way most scholars classify
parts of speech is to select some of distribution characteristics, while neglecting
some other distribution characteristics. This way of classifying parts of speech
is in conflict with the view that a part of speech has the nature of distribution.
If we accept that this is reasonable, we should negate it. Furthermore, what
distributions should be selected for classifying parts of speech is not decided
by distributions themselves but relies on other factors.

4.1.4  Paradoxes of the partial distribution view

Some people have held the first two distribution views, but there are only a
few practitioners. Most scholars select partial distribution characteristics to
classify parts of speech, while neglecting other distribution characteristics.
If parts of speech are viewed as distribution classes, in theory, what problem
does this view bring about?
58

58  Essence and expressional functions

1. If distribution is viewed as the nature of a part of speech and partial dis-
tribution characteristics can be selected among the distribution of words
as the criteria for classifying parts of speech, then a certain class of words
should have the distributions of internal universality and external exclu-
sivity. But our examination results show that such distributions cannot be
found actually.
This phenomenon will be illustrated by the main grammatical functions
of the following major parts of speech. First, let us look at the distribu-
tion that has no internal universality (for specific data on noun, verb and
adjective, see the statistical results in Chapter 2.1 in Volume 2):
(1) Nouns
• * Subject: 作为 (action), 地步 (extent), 新生 (rebirth), 剧毒 (high
toxicity), 来由 (cause), 泡影 (zilch), 一体 (oneness), 着落 (where-
abouts), 国际 (internationality), 乐子 (fun), 鬼胎 (evil plot)
• * Object:  年事 (person’s age), 谈锋 (eloquence), 现年 (current
year), 浑身 (whole body), 爱憎 (love and hatred), 常言 (proverb)
• * Quantity ~: 私人 (privacy), 人类 (mankind), 总和 (sum), 航运
(navigation), 列强 (power), 内心 (heart), 手工 (handwork),
哥 (brother), 军事 (military affairs), 利弊 (advantages and
disadvantages)
• *Attributive ~: 私人 (privacy), 外界 (exteriors), 出手 (disposal),
现年 (current year), 敌我 (enemy and ourselves), 中外 (China
and foreign), 国际 (international)
• *Attributive:  把戏 (trick), 称号 (title), 大局 (overall situation),
法子 (method), 方针 (policy), 害处 (harm), 举动 (action), 巨响
(blare), 计策 (stratagem), 理由 (reason), 勾当 (deal), 胆子 (gut),
跟头 (fall), 差错 (error), 措施 (measure)
(2) Verbs
• * 不 (not) /​没 (not) ~: 致使 (cause), 看待 (regard), 注定 (doom),
查收 (check), 胆敢 (dare), 活像 (look exactly like), 有关 (relate
to), 总计 (total), 连绵 (stretch), 出落 (grow), 有待 (await), 捉摸
(fathom), 地处 (locate), 肄业 (attend college)
• *Predicate: 住 (able to, for example, 抓不住 (unable to grasp)),
着 (used for emphasis, for example, 打不着 (cannot beat)), 透顶
(thoroughly), 绝顶 (extremely)
• *Adverbial:  备用 (standby), 参半 (half), 留念 (souvenir), 不息
(incessantly), 不等 (variously), 同上 (ibid.), 扑鼻 (assail nostrils)
• *~ real object:  工作 (work), 休息 (rest), 飞跃 (leap), 搏斗 (fight),
劳动 (labor), 成长 (grow), 到来 (arrive), 倒退 (reverse), 着想 (con-
sider), 发育 (growth), 浮动 (float), 交际 (social), 考试 (examination),
崩溃 (collapse), 相反 (opposite), 颤 (quivering), 颠倒 (reversed)
(3) Adjectives
• *Not ~:  荣幸 (honored), 异常 (abnormal), 微小 (trivial), 有趣
(interesting), 难免 (unavoidable), 无情 (ruthless), 不安 (uneasy),
不错 (alright)
 59

Essence and expressional functions 59

• *very ~1:  平衡 (balanced), 相同 (same), 耐烦 (patient), 景气
(prosperous), 道德 (moral), 像话 (right), 起眼 (eye-​catching),
要脸 (face-​saving), 寻常 (ordinary)
• *Predicate: 个别 (individual), 巧合 (coincidental), 停当 (ready),
停妥 (orderly)
• *Complement: 荣幸 (honored), 异常 (abnormal), 辛勤 (diligent),
间接 (indirect), 静 (quiet), 固执 (stubborn), 抱歉 (sorry), 个别
(individual)
• *Attributive:  不错 (good), 高兴 (glad), 荣幸 (honored), 齐全
(complete), 停当 (ready), 糟糕 (bad), 一样 (same), 对 (right),
多 (numerous), 挤 (crowded), 久 (long time), 痒 (aching), 痛
(painful)
(4) State  words
• *Predicate: 飞快 (fast), 崭新 (brand-​new), 火热 (fervent)
• *Complement: 羞答答 (bashful), 皑皑 (snow-​white), 好端端 (in
perfectly good condition), 雄赳赳 (valiant), 指指点点 (gossiping-​
about), 磨磨蹭蹭 (dawdling), 旖旎 (graceful), 闪闪 (sparkling),
重重 (numerous)
• *Followed by de2:  皑皑 (snow-​white), 闪闪 (glittering), 旖旎
(graceful), 优良 (excellent), 崭新 (brand-​new), 金黄 (golden-​
yellow), 碧绿 (green)
( 5) Distinctive  words
• *Attributive: 亲爱 (dear), 心爱 (lovely)
• *Followed by ~ de (的):  公共 (public), 机要 (confidential),
日用 (daily), 聋哑 (deaf and dumb), 接力 (relaying), 集约
(intensive)

Distributions that genuinely have internal universality presumably

include “~ 量 (quantity)” of a numeral, “数 (several) ~” of a measure word
and an adverb’s function as adverbial, but all of them have no external
exclusivity.
Let us look at the distributions that have no external exclusivity in the
following:2

(1) Subjects, objects: Nearly all the parts of speech of notional words, except
adverbs, can function as subjects and objects; for example, nouns, verbs,
adjectives, locatives, time words, place words and numerals can do so.
(2) Attributives: Nearly all of the parts of speech of notional words except
adverbs and measure words can function as attributives; for example,
nouns, verbs, adjectives, distinctive words and numerals can do so.
(3) Attributives ~: Nearly all of the parts of speech except distinctive words,
adverbs and numerals can be modified by attributives; for example,
nouns, time words, measure words, verbs and adjectives can do so.
(4) Predicates:  Nearly all of the parts of speech except distinctive words,
adverbs and numerals can function as predicates. Generally, verbs,
60

60  Essence and expressional functions

adjectives and state words can function as predicates. Under certain
conditions, nouns, time words, numerals and adverbs can also function as
predicates.
(5) Adverbials:  adverbs, adjectives, state words, numerals and measure
words, time words, locatives and place words can function as adverbials.
(6) Verbs and adjectives have the grammatical function that they can be
placed after 不 (not).
(7) Verbs and adjectives have the grammatical function that they can be
placed after 很 (very) or 很不 (not at all).
(8) Verbs, adjectives and state words can function as complements.
(9) Verbs and adjectives can have their complements.
(10) Verbs and adjectives can function as objects.
(11) Verbs, adjectives, state words, nouns, time words and numerals can be
modified by adverbials.
(12) Link-​verb predicates: verbs, adjectives and state words can be the imme-
diate constituents of a link-​verb predicate construction.
(13) Verbs, adjectives, nouns and time words can be followed by le (了).
(14) Verbs and adjectives can be followed by zhuo (着).
(15) Verbs and adjectives can be followed by guo (过).
(16) Nouns, verbs, adjectives and time words can be modified by numerals
and measure word phrases.
(17) Measure words and nouns can be modified by numerals.
(18) Locatives, place words and time words can function as the objects of 在
(in), 到 (to), 往 (toward).
(19) Numerals, adjectives and demonstratives have the grammatical function
of being followed by measure words.

The general grammatical functions3 that can be genuinely deemed to have

external exclusivity presumably include only “suo (所) plus verb”, “state
word followed by de (的)” and the “possessive attributive plus numeral and
measure word plus noun” of demonstratives, but all of them have no internal
universality.

2. Some words have quite different distributions but belong to the same part
of speech. 年事 (person’s age) and 现年 (current year), as cited above, can
only function as subjects; 剧毒 (high toxicity) and 泡影 (zilch) can only
function as objects. They have quite different distributions, but all are
nouns. 活像 (look exactly like) and 企图 (attempt) can only appear at the
central position of a predicate and cannot be modified by 不 (not); zhu
(住) and zhuo (着) can only appear at the position of a complement but
belong to verbs.
3. Even if we can select some distribution characteristics to classify parts of
speech, we find that we cannot actually use the distribution characteristics
themselves to answer why we select these distribution characteristics
instead of others as the criteria for classifying parts of speech. Different
 61

Essence and expressional functions 61

parts of speech are obtained with the selection of different criteria for
classifying them. For example, we classify into adverbs the words that can
only function as adverbials, but why cannot we classify the words that can
only function as complements into only-​complement words and the words
that can only function as objects into only-​object words? In other words, if
we have no classification of parts of speech in mind beforehand, if factors
other than distribution characteristics are not considered, we cannot clas-
sify parts of speech purely according to distribution characteristics.

In the following, the distinctions between verb and adjective are taken as
examples to show that they cannot be determined only by distribution.
Table  4.1 gives the distinctions between verb and adjective proposed by
Zhu Dexi (1982b):
Theoretically speaking, there are, indeed, at least the following methods for
classifying parts of speech of the language facts reflected by these two gram-
matical functions:

1. Classifying the language facts into two types according to whether they
can take their real objects or not: Type A (transitive predicate words: 想
(think), 唱 (sing)) and Type B (intransitive predicate words: 醒 (wake), 大
(great)).
2. Classifying the language facts into two types according to whether they
can be modified by 很 (very) or not: Type A (degree predicate words: 想
(think), 大 (big), Type B (non-​ degree predicate words:  唱 (sing), 醒
(wake)).
3. Give conjunctive relationship to words that can be modified by 很 (very)
and take objects, and they are classified into four types: Type A (+ very ~
∧ + ~ object: 想 (think)), Type B (-​very ~ ∧ + ~ object: 唱 (sing)), Type
C (* very ~ ∧ –​~ object: 醒 (wake), Type D (+ very ~ ∧ * ~ object: 大
(big)).
4. Give conjunctive or disjunctive relationship to words that can be modi-
fied by 很 (very) and take objects, and classify them into two types: Type

Table 4.1 Zhu Dexi’s (1982b) criteria for distinguishing between verbs and adjectives

Taking 很 Taking Example

(very) objects

1 + + 想 (think), 怕 (fear), 爱 (love), 喜欢(like), 关心

(concern), 相信 (believe)
2 —​ + 唱 (sing), 看 (watch), 切 (cut), 有 (have), 讨论
(discuss), 分析 (analyze)
3 —​ —​ 醒 (wake), 锈 (rust), 肿 (swell), 咳嗽(cough), 游行
(parade), 休息 (rest), 死(die)
4 + —​ 大 (big), 红 (red), 远 (far), 累 (tired), 饱(full), 结实
(strong), 干净 (clean)
62

62  Essence and expressional functions

A (* very ~ ∨ + ~ object: 想 (think), 唱 (sing), 醒 (wake)), Type B (+ very
~ ∧ * ~ object: 大 (big)).
5. Give conjunctive or disjunctive relationship to words that can be modi-
fied by 很 (very) and take objects, and classify them into two types: Type
A (+ very ~ ∨* ~ object: 想 (think), 醒 (wake), 大 (big)), Type B (* very
~ ∧ + object: 唱 (sing)).
6. Give conjunctive or disjunctive relationship to words that can be modi-
fied by 很 (very) and take objects, and classify them into two types: Type
A (+ very ~ ∨ + ~ object: 想 (think), 唱 (sing), 大 (big)), Type B (* very ~
∧ * ~ object: 醒 (wake)).
7. Give conjunctive or disjunctive relationship to words that can be modi-
fied by 很 (very) and take objects, and classify them into two types: Type
A (* very ~ ∨ * ~ object: 想 (think), 唱 (sing), 醒 (wake), 大 (big)), Type
B (+ very ~ ∧ + ~ object: 想 (think)).
8. Give conjunctive or disjunctive relationship to words that can be modi-
fied by 很 (very) and take objects, and classify them into two types: Type
A  ((+ very ~ ∨ + ~ object ∨ (* very ~ ∧ * ~ object)):  想 (think), 醒
(wake)), Type B ((* very ~ ∧ + ~ object)∨(very ~ ∧ * ~ object)):  唱
(sing), 大 (big).

In addition, there are some other classification methods and criteria.

Table 4.2 gives all the possible classifications:
The table shows that both “很 (very) ~” and “~ object” are not the
grammatical characteristics of a verb or an adjective. Only when the two
functions are given a conjunctive or disjunctive relationship can grammat-
ical characteristics be obtained. For example, “*很 (very) ~ ∨ + ~ object” is
the grammatical characteristic of a verb. “+ 很 (very) ~ ∧ * ~ object” is the
grammatical characteristic of an adjective. But the problem is that, as shown

Table 4.2 The possible classifications obtained with the two criteria of “~ object” and
“很 (very) ~”

a b c d a b c d

1 想 (think), 醒(wake), 8 想(think), 唱(sing),

唱 (sing) 大 (big) 醒(wake) 大(big)
2 think, big sing, wake 9 think sing wake big
3 think sing wake big 10 think wake sing big
4 think, sing, big 11 think sing wake
wake big
5 think, wake, sing 12 sing, wake think big
big
6 think, sing, wake 13 sing, big think wake
big
7 sing, wake, think 14 wake, big think sing
big
 63

Essence and expressional functions 63

in the above table, when the two functions are joined together with a conjunc-
tive or disjunctive relationship, there can be numerous classifications. If only
distribution is taken into account, it is hard to explain the reason why only
Method Four instead of other methods should be selected. The reason why
we select Method Four is that its syntactic function is in agreement with its
semantic type. The parts of speech classified according to this syntactic cri-
terion exactly exhibit the differences in semantic type: action and properties.

4.1.5  Paradoxes of the similarity theory

Because it is difficult to identify the distribution characteristics of a part of
speech that have internal universality and external exclusivity, the view that
distribution characteristics are its nature can be questioned. Scholars attempt
to use the similarity theory to explain the distribution nature of a part of
speech, thinking that it is collected according to the distribution similarity of
words. According to whether or not words are collected with their prototype,
there are the prototype theory of a part of speech and its overall similarity
theory.

4.1.5.1  Paradoxes of the prototype theory

Shi Youwei (1994) held that Chinese parts of speech should be treated flexibly.
Yuan Yulin (1995, 1998) used the prototype theory or the family similarity
theory to deal with the relationship between distribution and a part of speech.
He (1995, 1998, 2001) thought that a part of speech belongs to the prototype
category, and that its typical members share some distribution characteristics
that other parts of speech do not have. The distribution characteristics of atyp-
ical members are incomplete, but we can put them into one class according
to their similarity with typical members. For example, 耐烦 (patience) and
景气 (prosperity) cannot be modified by 很 (very), but typical adjectives can
not only be modified by 很 (very) but also be followed by 不很 (not very).
The two words in the above example can also be followed by 不很 (not very).
Therefore, according to the family similarity principle of distribution, we can
assert that 耐烦 (patient) and 景气 (prosperous) belong to adjectives. For
another example, words such as 极 (extremely) can be used as adverbials or
complements. In the light of the typical adverb 太 (too), which can only be
used as an adverbial, we can classify 极 (extremely) into an adverb. Then, in
light of the atypical adverb 极 (extremely), we can classify into an adverb the
word 透顶 (thoroughly), which can only be used as a complement.
This brings about the following problems:

1. According to the family similarity principles of distribution, we can place

almost all notional words into one class. For example, typical adjectives
can function as attributives; hence, we may think that nouns such as 相同
(sameness), 大型 (large-​scale), 彩色 (color), 木头 (wood), 社会 (society)
64

64  Essence and expressional functions

that can function as attributives belong to atypical members of adjectives,
thereby classifying them as adjectives. Typical adjectives can function as
adverbials; hence, we may classify into adjectives words such as 亲自
(personal) and 全力 (all-​out effort) that can function as adverbials. They
may function as complements; hence, we may classify into adjectives
words 雪白 (snow-​white), 来 (come), 着 (zháo) that can function as
complements. They can function as predicates; hence, we may classify
into adjectives words such as 吃 (eat), 洗 (wash) and 休息 (rest). You may
presumably say that 颜色 (color), 木头 (wood) and 雪白 (snow-​white)
cannot be modified by 很 (very); therefore, they are not adjectives. But
the question is: how do you know that being preceded by 很 (very) is a
necessary grammatical function of an adjective, whereas “functioning as
an attributive is its sufficient grammatical function”? The family simi-
larity principle itself cannot answer this question. Because of this, there
is no way to operate in reality. For example, because typical adjectives can
be preceded by 完全 (completely) ~, we may classify into adjectives not
only words such as 相同 (same) and 相反 (opposite) that behave like this
but also words such as 属实(verify), 腐烂 (perish), 静止 (standstill), 停顿
(pause) and 融化 (melt) that usually belong to verbs.
2. A word may be classified into Class A  or B according to Distribution
Characteristics x or y. Then, the family similarity principle works by
relying on perception. For example, it is possible to classify into distinctive
words, words such as 主要 (primary), 次要 (secondary), 新型 (new-​type),
亲爱 (dear) and 基本 (basic) that can be used as attributives but not
predicates. It is also possible to classify them into adjectives because they
are typical adjectives and thus can be modified by 最 (the most). In other
words, when a word has the distributions of either Class A or B words, the
family similarity principle cannot be used to classify their parts of speech.
3. In reality, there is no way to implement this method.

Shi Youwei (1994) first determined the distribution characteristics of typ-

ical members of a part of speech and then determined which part of
speech the atypical members belong to according to the correlation values
(membership degree) of the distribution and the word-​formation of atyp-
ical and typical members. For example, compared with 冷 (cold), a typical
member of adjectives, the correlation value of 温 (warm) is only 1.5; its
value of correlation with 彩色 (colorful), a typical member of distinctive
words, is 6. Because 温 (warm) has a higher correlation value than a typ-
ical member of distinctive words, it is classified into distinctive words.
But the fact that it can be preceded by 有点~了 (a bit ~) can contra-
dict that it belongs to distinctive words; instead, it should be classified
into adjectives. Hence, correlation values do not truly reflect the parts of
speech of a word.
Lu Yingshun (1998) proposed a formula for calculating the similarity
degree: S=N/​M, where M refers to the number of standards of a certain part
 65

Essence and expressional functions 65

of speech, and N refers to the number of words that satisfy such standards.
If S ≥  =  0.5, then words basically belong to the part of speech. He gave
the five standards of a verb: (1) A verb can be modified by 不 (not) and can
have the question form of “X or not X”. (2) It can be followed by dynamic
auxiliary words such as le (了) and qilai (起来) and so on to denote perfec-
tion, commencement and so on, or be followed by a zero label to denote a
habitual action or attribute. (3)  A  good many verbs can be followed by an
object. (4) A verb can occur after a noun to form a subject-​predicate relation.
(5) It can be followed by numeral and measure-​word objects and go together
with other constituents to form the predicate and complement constructions.
According to his calculation, 吃 (eat) satisfies all these standards, and S=1.
跑步 (run) satisfies Standards 1, 2 and 4, and S=0.6; thus, they can be classified
into verbs. 课文 (text) satisfies no standards, and S=0. 春天 (spring) satisfies
Standards 2 and 4, and S=0.4; thus, they cannot be classified into verbs.
The standard that S≥0.5 is too broad and includes ordinary adjectives. But
Lu Yingshun thought that verbs and adjectives should belong to one part of
speech. For the time being, we will not discuss this issue here. We should look
at whether or not they have internal universality. Now we use this method to
test the following words:

担待 (undertake) satisfies Standards 4 and 5 but does not satisfy 1, 2 and

3. S=0.4;
雷动 (thunderous) satisfies Standard 4 but does not satisfy Standards 1,
2, 3 and 5. S=0.2;
媲美 (rival) satisfies Standards 3 and 4 but does not satisfy Standards 1,
2 and 5. S=0.4.

All these words belong to verbs; if S ≥ 0.5 is followed, then we can only
exclude them from verbs; hence, this method still does not work. Yuan Yulin
(1995) thought that “because a part of speech is a prototype category, all the
members of a part of speech often do not share the distribution characteristics
that the members of another part of speech do not have. Therefore, there is
no way to use the conjunction/​disjunction relations among several distribu-
tion characteristics to strictly determine the part of speech of a word. We can
only use the advantageous distribution probability of a certain class of words
to broadly determine their parts of speech. But the broad determination is
too fuzzy and not satisfactory”. “We can use the distribution characteristics
unique to typical members to determine their parts of speech comparatively
strictly”.
Let us look at whether or not we can use the prototype theory as the cri-
terion for classifying parts of speech.
The strict definition of noun: a noun is a class of words whose members
can be modified by numerals and measure words but not adverbs.
This definition is very strict indeed. Only 78% of nouns can be modi-
fied by numerals and measure words, and only a small number of nouns
66

66  Essence and expressional functions

can be modified by ordinary adverbs (2%). If scope adverbs are also taken
into account, then many more nouns can be modified by adverbs, for
example: 光馒头就吃了三个 (Three pieces of steamed bread alone have been
taken). Thus, we do not know which nouns are typical nouns.
The broad definition of a noun: a noun is a class of words that is often
used as a typical subject (doer of an action) and a typical object (object of
an action and resultant object), and generally cannot be modified by adverbs.
Yuan Yulin treated 野外 (out in the field) and 下面 (underneath) as nouns,
but following the prototype theory, it is difficult to classify them into nouns.
If nouns generally cannot be modified by adverbs, then how should nouns
that can be modified by adverbs be handled? Should they be regarded as gen-
erally not being modified by adverbs or generally being able to be modified
by adverbs? Should words such as 年事 (person’s age) that can only be used
as subjects, and words such as 地步 (extent) that can only be used as objects
still be classified into nouns? Should words such as 私人 (private) that are not
often used as subjects still be classified into nouns?
Adverbs are defined as words that can only be used mainly as adverbial.
Can 很 (very) and 极 (extremely) be considered to be used only as adverbials
by and large? Yuan Yulin classified 透顶 (thoroughly) and 透 (thorough) into
adverbs, but these two words really fall short of the criterion that they can
only be used as adverbials by and large.
The broad definition of state words: a state word is often used as a predi-
cate, a complement and an adverbial, and cannot be modified by adverbs such
as 很 (very) and 不 (not) and so on. Only 86% of state words can be used as a
predicate; only 50% of them can be used as a complement; only 15% of them
can be used as an adverbial. It is not clear whether there are conjunctive or
disjunctive relations among the three grammatical functions. If there is a con-
junctive relation, then only 6% of words can be classified into state words; if
there is a disjunctive relation, then adverbs are also classified into state words.
All in all, (1)  we should not only look at the grammatical functions of
a word but also their frequencies, but it is almost impossible to do such
complicated calculations. (2) There are no clear criteria for judging “often”.
(3)  Almost every part of speech has an imbalance among its grammatical
functions; therefore, it is unsafe to classify parts of speech according to their
major grammatical functions (see Chapter 2.4.7 in Volume 2).

4. The prototype theory does not give a method for determining the proto-
type of a part of speech. In fact, it is impossible to determine its proto-
type according to distribution itself.
5. The root problem is that the prototype theory argues in a logical circular
way. The determination of the prototype of a part of speech can only
be achieved under the condition that, first, categories are classified and
their typical distribution characteristics are known. In other words, when
you say that words such as 桌子 (desk), 石头 (stone), 人 (person) and 马
(horse) are typical members of nouns, you already know that a noun is
 67

Essence and expressional functions 67

a part of speech and that a typical noun can be modified by a numeral
and a measure word. If you do not know these facts, you cannot say so.
Therefore, this actually causes circular arguments: on the one hand, the
prototype of a part of speech and its distribution characteristics must be
determined after categories are classified; on the other hand, the category
classification depends on the determination of the prototype of a part of
speech.

We do not prefer to classify parts of speech based on the prototype theory;

this does not mean that we altogether deny their prototype nature. In our
opinion, although parts of speech do not belong to the prototype category
and the prototype theory is not operable, they do have a prototype nature in
terms of apparent characteristics (rather than intrinsic characteristics). The
prototype nature expresses itself mainly in the following four respects:

1. The prototype nature of lexical meaning. A part of speech has its proto-
type lexical meaning. For example, nouns denote things; verbs denote
action; typical nouns denote three-​ dimensional material objects (see
Taylor, 1989).
2. The interconnection among parts of speech, lexical meanings and syn-
tactic constituents has a prototype nature. For example, the interconnec-
tion among actions, verbs and predicates is prototypical and unmarked,
whereas the interconnection among actions, verbs and subjects is non-​
prototypical and marked (see Croft, 1991).
3. Humans’ perception of parts of speech is prototypical. Namely, they rely
on a prototype to identify them.
4. A certain class of words has a prototypical nature in terms of distribution.

But these are not of the intrinsic prototypical nature of parts of speech.
What is similar to this phenomenon is sex. Sex is not a prototypical class. What
determines sex is the XY sex chromosome or the XX sex chromosome. Male
and female can be clearly classified according to whether there is a Y chromo-
some or not, but from the perspective of the apparent characteristics of sex,
it has its prototype. These apparent characteristics include (taking human
beings as an example) physiological characteristics, physical appearance
characteristics, clothing characteristics, behavior characteristics, disposition
characteristics, occupational characteristics and so on. Sometimes things of
the same type are so different in apparent characteristics that it is difficult to
establish the interconnection among them only according to these apparent
characteristics. Only through their intrinsic characteristics can we establish
their interconnection, for example, between coal and diamonds.
The prototype of a class expresses itself at different levels. We can neither
regard the prototypical nature of apparent characteristics as that of intrinsic
characteristics nor classify objects into a class solely according to the similarity
of apparent characteristics. We cannot think that a part of speech belongs
68

68  Essence and expressional functions

to a prototype category because it has a distributional prototype, just as we
cannot think that sex belongs to a prototype category because it is prototyp-
ical in terms of clothing characteristics and disposition characteristics.
Because the prototype theory does not hold and is not operable, we do
not use it; instead, we use the characteristic theory. Fundamentally speaking,
the reason why the prototype theory does not hold is that distribution is not
an intrinsic characteristic of a part of speech; its intrinsic characteristic is an
expressional function. Distribution is but the apparent characteristics of a
part of speech. Therefore, we cannot successfully classify parts of speech only
according to distribution.

4.1.5.2  The overall similarity clustering view

To avoid a circular argument, we can use another method, which is not to
determine the prototype of a category in advance but to cluster words
into their classes solely according to their distributional overall similarity.
Following the prototype theory, a part of speech is a class of words classified
and clustered according to their distributional similarity. To verify the feasi-
bility of the method, we carried out the clustering analysis of the distribu-
tional similarity degree of the 60 words listed in Table 4.1-​1 (see appendices
in Volume 2). Assuming 0 ≤ S ≤ =100, the formula for calculating the distribu-
tional similarity degree (S) between two words is as follows:

S = 100× I/​(P –​  I),

where I is the number of identical grammatical positions at which two words
appear, and P is the sum of the numbers of grammatical positions. For
example, there are 19 grammatical positions where 干净 (clean) may appear,
and there are nine grammatical positions where 附近 (nearby) may appear.
Five grammatical positions are shared by the two words, whose similarity
degrees are:

S = 100×5/​(285)=22.

Table  4.1-​2 (see appendices in Volume 2)  gives the mutual distributional
similarity degrees of the 60 words calculated according to Table 4.1-​1. The
sequencing of the mutual similarity degrees of the 60 words according to their
size produces Table 4.1-​3 (see appendices in Volume 2).
The tables show an interesting fact:  the word that is the most similar to
another word in terms of distribution does not necessarily belong to the same
part of speech. For example, the ten words that have the highest degree of
distributional similarity to the state word 花白 (grizzled) include 日常 (daily),
临时 (temporary), 野生 (wild), 慢性 (chronic, distinctive word), 私人 (pri-
vate, noun), 众多 (multitude, numeral), 相同 (same, adjective), 注定 (doom,
verb), 个别 (a few, numeral) and 钢笔 (fountain-​pen, noun). However, the ten
 69

Essence and expressional functions 69

words that have the highest degree of distributional similarity to the verb 着想
(consider) include 荣幸 (honored, adjective), 酷热 (swelter, state word), 注定
(doom), 休息 (rest, verb), 野生 (wild, distinctive word), 相同 (same, adjec-
tive), 洗 (wash, verb), 花白 (grizzled, state word), 慢性 (chronic, distinctive
word) and 亲爱 (dear, distinctive word).
Of course, to cluster individual words into their classes strictly according
to their overall similarity, we need to do so with clustering analysis. Clustering
analysis uses a mathematical technique to gather individual words of a close
distance (large similarity degree) into clusters and further gather them into
larger clusters and eventually into one large cluster. There are different
methods for calculating the distance among individuals or clusters, thereby
forming different clustering analysis methods. The most commonly used
clustering analysis methods include the nearest neighbor linkage technique,
the neighbor linkage technique and the average linkage technique. The nearest
neighbor linkage technique treats the nearest distance (the largest similarity
degree) among individuals in two clusters as the distance between them. For
example, the similarity degrees among five individual words such as a, b, c, d,
e respectively are:
The similarity degree between a and b is the largest (10), and we gather
them first and foremost into one cluster (A). Among the remaining individ-
uals, the similarity degree between d and e is the largest (9). We then gather
them into one cluster (B). The similarity degree between the individual a in
Cluster A  and the individual e in Cluster B is 8, larger than the remaining
similarity degrees. Therefore, Clusters A and B are gathered into the larger
Cluster C. The similarity degree between the individuals c and a in Cluster C
is 7. Thus Cluster C is gathered into the larger Cluster D. The above clustering
operation can be expressed in a tree-​hierarchy diagram (Figure 4.1):
The furthest neighbor linkage technique treats the longest distance (the
smallest similarity degree) among clusters as the distance between them.
Taking the scenario in Table 4.3 as an example, the clustering results with the
furthest neighbor linkage technique are represented as Figure 4.2.
The average linkage technique treats the average distance among individ-
uals in two clusters as the distance between two clusters. Taking the scen-
ario in Table 4.3 as example, the clustering results with the furthest neighbor
linkage technique are represented as Figure 4.3 (see p. 70):

D
C
7

A 8 B

10 9

a b d e c

Figure 4.1 The nearest neighbor linkage hierarchy

70

70  Essence and expressional functions

Table 4.3 The similarity degrees among five individual words

a b C d

b 10
c 7 4
d 3 5 5
e 5 8 5 9

We carried out a clustering analysis of the distributional similarity

degrees among the 60 words listed in Table  4.1-​ 3 (see appendices in
Volume  2). Because the furthest neighbor linkage technique is rather
extreme, the book uses only the nearest neighbor linkage technique and the
average linkage technique to perform the clustering analysis, whose results
are shown in Figures 4.4 and 4.5.
As we can see, the results clustered with either the nearest neighbor linkage
technique or the average linkage technique are quite different from the parts

10 9

a b c d e

Figure 4.2 The furthest neighbor linkage hierarchy

5.2

5.5

10 9

a b c d e

Figure 4.3 The average neighbor linkage hierarchy

 71

honored
same
rest 61
build 80
wash
like 81 67
drunk 
small
clean 86
serious 82
extremely hot 75
act
years 100 88
person
barrel 100 89 80
sisters
people
pen
phone 80 75
sunny
playground 90
China 82
afternoon 92
today
temporary 70
everyday 80
chronic 50
private 100 83
wild 67
field
before 78
nearby
any
individual
lots of 80
gray 64
moment 45
recently 56
numerous
two 80
ten
many
thousand 43-44
standing
dear
all
this 40
snow white
think about
destined
every 33
hour 67
piece 26
extent
torrential 20
diachronic
ready
on earth 100 17
personally
age

Figure 4.4 Clustering analysis of distributional similarity degrees of 60 words (nearest

neighbor linkage technique)
72

build 72
like 81
wash 77
drunk 59
serious 86
clean 80 55
small
rest 61 34
same
destined 38
think about 45
honored
extremely hot
act 100
year 85
person 100
barrel 89 68 61
sister
people
sunny 90
playground 86 67
China 74 25
afternoon 92 55
today 69
phone 80
pen 46
field 69
nearby 78
before 43
many 55 15
ten 80
two
recently
everyday 80
temporary 66 33
private 100 83
chronic 56
wild
gray
numerous 44 28
lots of 80 60
individual 74 50 39 12
any
moment
dear
snow-white
this 43 11
every
standing 33
all
thousand 37 8
hour 67 25
piece
extent 6
diachronic 33 4
torrential
age
ready 3
on earth 100
personally

Figure 4.5 Clustering analysis of distributional similarity degrees of 60 words (average

linkage technique)
 73

Essence and expressional functions 73

of speech classified in the usual ways. For example, if the average linkage
technique is used, the state word “grizzled” first and foremost clusters into
one class the words that are usually classified into distinctive words. The verb
着想 (consider) gathers the adjective 荣幸 (honored) first and then the tran-
sitive verb 注定 (doom). The adjective 相同 (same) first gathers the verb 休息
(rest). The distinctive word 慢性 (chronic) first gathers the noun 私人 (pri-
vate). Classifying parts of speech in this way is perhaps meaningful in the
engineering sense but obviously has no linguistic meaning. These are not the
parts of speech in a linguistic sense.
In 1986, when we had just begun the study of Chinese parts of speech,
we held the distribution view and tried to classify them according to the
overall distribution similarity degrees of words. But small-​scale experiments
indicated that parts of speech thus classified show no resemblance to those
in a linguistic sense. Therefore, we abandoned our view of their distribution
nature and explored their further level of distinction.

4.1.6  Brief summary

It is impossible for us to classify parts of speech solely according to the distri-
bution of words. We actually neglect some of their distributional differences
and select some others in order to do so. Which distributional characteristics
are to be selected are not determined by distribution itself; some other factors
must be taken into account. In fact, we are seeking valuable grammatical
meanings of the parts of speech thus classified. In the following sections, we
shall mention that we classify parts of speech by selecting those characteristics
that can separate parts of speech in terms of their grammatical meanings.
Hence, the nature of a part of speech is not of the distribution type.
The nature of classifying parts of speech according to distribution is to
classify them according to the selectional restriction of the syntactic positions
on words. This selectional restriction must have certain bases to be used as the
conditions for selectional restriction; otherwise, the selection may be unre-
stricted. Then, what are the bases for the selectional restriction? If we think
that parts of speech are of the distribution type and that the basis for selec-
tional restriction is distribution, then we can only say that the reason why
a word can appear at a certain grammatical position is that the word can
appear at that position. This is actually a tautology. In other words, some
words belong to a certain part of speech is because they have the same distri-
bution. The reason why they have the same distribution is that they belong to
the same part of speech. This becomes a circular argument. But in the view of
conjuring structuralists, this is the final solution to the problem; they refuse
to talk about meaning or justification. At this moment of linguistic devel-
opment, we are already not satisfied with such a solution and still desire to
understand the reason behind it. In our opinion, the reason why words have
identical or different grammatical distributions is that they have identical or
74

74  Essence and expressional functions

different expressional functions and semantic types, which are the intrinsic
nature of a part of speech.

4.2  Bases of selectional restriction of a grammatical position

on words
What is the intrinsic nature of part of speech? We discuss this question on the
basis of a grammatical position’s selectional restriction on words.
As mentioned before, a grammatical position has selectional restrictions
on words that may enter into it. Here we explain that this selectional
restriction requires bases. In the following, we shall cite examples of how
a syntagmatic position selects words, to explain the basis for selectional
restriction.

(1) Among words such as 送 (send), 买 (buy), 卖 (sell), 借 (borrow), 抢

(rob), 炒 (fry), 还 (return), 赔 (compensate), 偷 (steal), 沏 (infuse),
做 (make) and 织 (weave), some can enter Position I, and some can
enter Position II, while others can enter Position III:
(2) Words such as 送 (send), 买 (buy), 卖 (sell), 借 (borrow), 还 (return)
and 赔 (compensate) can have two objects, for example:
I. ~ + 给 (give) + A  + B.  For example, 送给他一本书 (give him
a book):  送 (send), 卖 (sell), 借 (borrow), 还 (return), 赔
(compensate)
II. 给 (give) + A + ~ + B. For example, 给他买一本书 (buy a book
for him): 买 (buy), 借 (borrow), 偷 (steal), 抢 (rob), 炒 (fry), 做
(make), 沏 (infuse), 织 (weave)
III. ~ + A  + B.  For example, 送他一本书 (send him a book):  送
(send), 卖 (sell), 借 (borrow), 还 (return), 赔 (compensate), 买
(buy), 偷 (steal), 抢 (rob)

According to the different positions in which they appear, we can classify

them into the following three classes (see Zhu Dexi, 1979):

a. Send, sell, borrow, return, compensate (appearing at Positions I and III)

b. Buy, borrow, steal, rob (appearing at Positions II and III)
c. Fry, infuse, weave (appearing at Position II)

Although the parts of speech are classified according to distribution, they

are scarcely regarded as a distributional type. What factors influence a word’s
appearance position? We may not hesitate to give the answer that it is the
semantic meanings themselves:  Class A  has the “giving” meaning; Class B
has the “obtaining” meaning; Class C has the “making” meaning. In other
words, the selectional restriction of the above three positions is based on
the semantic meanings of the words themselves; words of different semantic
 75

Essence and expressional functions 75

meanings enter into different positions. Therefore, speaking intrinsically, the
three classes belong to semantic types.
As a matter of fact, there must be bases for the syntagmatic position’s selec-
tional restriction on words. Unless a syntagmatic position has no selectional
restriction on words, then any words can enter into the position optionally.
The above examples are based on the semantic meanings; similarly, the more
abstract grammatical positions that exclude semantic meanings are also based
on the selectional restriction on words. Let us consider the following English
example:

(3) (The) ~ are/​is acceptable.

At the position given in Example (3), the following can appear: children,

adults, students, teachers, young, old, smoking, teaching and son on. But the
appearance of “grow, born, study, teach, smoke” is not allowed. Then what
are their selectional bases? Apparently it is not their semantic meanings. We
should not think that the basis for selectional restriction is a part of speech
because this leads to a circular argument. Not all words that can appear at
this position are nouns. “Young” and “old” can take adverbials of degree,
for example:  “The very young are acceptable”. They can have comparative
degrees and superlative degrees. “Smoking” is not a noun because it can take
its object, for example:  “Smoking cigarettes is acceptable”. Analysis of the
similarity and difference of the words that can or cannot appear at a certain
position shows that the similarity of the words that can appear at Position 3
is that they have the function of reference, while those that cannot have the
function of statement. The basis for the positional selectional restriction is
expressional function.
Let us consider the following Chinese examples:

(4) X de (的) ~   (5) ~ 好治 (It is easy to treat ~)

The words that can appear at the position in Example (4)  include 书
(book), 桌子 (desk), 愿望 (desire), 感觉 (perception), 出版 (publication),
研究 (study), 到来 (arrival), 依赖 (dependence), 美丽 (beauty), 开明 (enlight-
enment) and 邪恶 (evil), and all having referential meanings. Words that
cannot appear at this position include 是 (be), 有 (have), 知道 (know), 觉得
(feel), 舍得 (willing to part with), 看见 (see) and 愿意 (willing), and all having
statement meanings. Words that can appear at the position in Example
(5) include 肠炎 (enteritis), 红眼病 (acute conjunctivitis), 感冒 (cold), 拉肚子
(diarrhea), 流鼻血 (nosebleed), 急性 (acute), 慢性 (chronic) and so on. What
do these words have in common? They are all referential words. The basis of
the two grammatical positions’ selectional restriction on words can also be
regarded as their expressional function.
In our opinion, the reason why words have different distributions is
that they have different properties, which are, therefore, their expressional
functions and semantic types.
76

76  Essence and expressional functions

4.3  Types and hierarchies of the expressional function

4.3.1  What is the expressional function?

In our opinion, the intrinsic nature of the large category of parts of speech
such as substantive words, predicate words and modification words is the
expressional function. Therefore, it is necessary to make a special effort to
discuss it.
As mentioned before, a grammatical position has a selectional restriction
on words, whose root basis is the expressional function of a word.
Zhu Dexi (1982b) was the first to propose the concepts of reference and
statement, which we call the expressional function.
The main reason for proposing the expressional function is to describe the
distinction between a and b in the following examples:

(6)
a. 想打球 (want to play basketball) (object of statement) –​want what
b. 看打球 (watch playing basketball) (object of reference) –​watch
what
(7) a. 子贡贤于仲尼(《论语·子张》) (Zigong is more virtuous
than Zhongni) (Zizhang in Analects of Confucius)
b. 见贤思齐焉,见不贤而内自省焉(《论语·里仁》) (When we see
a man of virtue and talent, we should think of equaling him;
when we see a man of a contrary character, we should turn
inward and examine ourselves.) (Liren in Analects of Confucius)
(8) a. 失所长则国家无功,守所短则民不乐生. (With a lack of the meri-
torious, the state will be powerless; if keeping the unmeritorious,
the people will not enjoy their life.)
b. 以无功御不乐生,不可行于齐民. (《韩非子·安危》) (Without
power, it is not feasible for the Qi people to enjoy their life.)
(Chapter on Safety and Danger in the book Han Fei Zi)
(9) a. 急性肠炎好治,慢性肠炎不好治. (Acute enteritis is easy to treat,
but chronic enteritis is not easy to treat) (modification).
b. 急性好治,慢性不好治. (The acute are easy to treat, but the
chronic are not easy to treat.) (reference)
(10) a. 我们研究问题 (We study problems), (embodied as an assertion)
b. 研究很成功 (The study is highly successful), (embodied as an
object)
(11) a. 这个苹果大 (The apple is large)
b. 有大有小 (Some are large; some are small)
(12) a. 一切财产 (all property)
b. 放弃一切 (abandon all)
(13) a. I walk every day (embodied as an assertion)
b. I take walk every day (embodied as an object)

Although the meaning of a word is identical on different occasions, it can be

expressed by different patterns. For example, 研究 (study), 大 (large), 一切
 77

Essence and expressional functions 77

(all) and “walk” in Examples (10) a and b through (13) have identical lexical
meanings, their meaning expression patterns are different: a is expressed as a
statement, while b is expressed as a reference. An expressional function refers
to the pattern by which words express their semantic meanings.
The expressional function is different from a syntactic constituent. On the
one hand, an identical expressional function can be used as different syntactic
constituents, for example, 研究 (study) in Example  10 (b)  and 一切 (all) in
Example 12 (b) both have reference meanings, but the former is a subject, while
the latter is an object. On the other hand, a different expressional function can
be used as identical syntactic constituents, for example, 保持安静 (keep quiet)
and 觉得安静 (feel quiet). The first 安静 (quiet) has a reference meaning, and
the second has a statement meaning, but both are used as objects. A syntactic
constituent starts from the relationship among immediate constituents, while
an expressional function is based on a word’s own properties. In other words,
we should regard the expressional function as a word’s own properties rather
than the properties of its grammatical environment.
An expressional function has four basic types: a statement denotes assertion
and can be modified by adverbials; a reference denotes object and can be
modified by attributives. A modification modifies or restricts a statement or
reference, which depends on whether a word modifies a statement or a refer-
ence; it can be further classified into a predicate modification and a substan-
tive modification. An auxiliary has a regulatory effect.

4.3.2  Types of expressional function

4.3.2.1  Statement and reference

The two most fundamental expressional functions of a language are a
statement and a reference. Zhu Dexi (1982b) distinguished between a
statement and a reference, pointing out that a statement is used to answer
the question “in what manner?” and that a reference is used to answer the
question of “what?” An object and a subject both can be a statement or a
reference. For example, Example (6)  a is a statement, and b is a reference.
He further pointed out that a reference means that there is a referent, and
that a statement means that something is signified. Further explanation is as
follows:

A statement denotes an assertion, points to another constituent and is

generally used to answer the question of “in what manner”.

A reference denotes an object, has an inward meaning and is generally used

to answer the question “what?”
The distinction between de (的) and di (地) in a modifier position some-
what reflects the mental distinction between a reference and a statement.
A statement can be modified by an adverbial; a reference denotes an object
and can be modified by an attributive.
78

78  Essence and expressional functions

There are exceptions for the above formal characteristics, for example:

(14)
公认他是好人 (It is generally recognized that he is a good person) –​
*recognize what; *recognize in what manner
(15)
彩色电视 (color television) –​what television (but 彩色 (color) is not
a reference)
(16)
悄悄对他说 (talk to him in whispers) –​talk to him in what manner
(but 悄悄 (in whispers) is not a statement)
The opposition between a statement and a reference is the most fundamental
in language, but it still has not been completely settled how to use formal
characteristics to distinguish between the two.

4.3.2.2 Modification
Some constituents are neither references nor statements. The following three
scenarios show that the constituent at a modifier position should be regarded
as the third type of expressional function –​a modification.

1. When distinctive words, adverbs, numerals and measure words are used
as modifiers, they are neither subjects nor objects. Only at the position
of subject or object are they subjects or objects; only at the position of
predicate are they statements. Compare the following:
(17) a. 急性肠炎好治，慢性肠炎不好治 (Acute enteritis is easy to
treat, but chronic enteritis is not easy to treat).
b. 急性好治,慢性不好治 (The acute are easy to treat, but the
chronic are not easy to treat).
(18) a. 有男生,有女生 (There are schoolboys and schoolgirls).
b. 有男有女 (There are men and women).
(19) a. 许多题都不会 (I cannot answer many questions).
b. 许多都不会 (Many I cannot answer).
(20) a. 我不去 (I do not go).
b. 我不 (I do not).
2. An adjective has different properties at a predicate position or a modifier
position:
(21) a. 衣服干净 (The clothes are clean). 衣服不/​很干净 (The clothes
are not/​very clean).
b. 干净衣服 (the clean clothes)   *不/​很干净衣服 (the not/​very
clean clothes)
(22) a. 学习认真 (study seriously)   学习不／很认真 (The study is
not/​very serious)
b. 认真学习 (serious study)   *不/​很认真学习 (study not/​very
seriously)
 79

Essence and expressional functions 79

干净 (clean) and 认真 (serious) can be modified by adverbials at the predi-
cate position but cannot be modified at the modifier position, indicating that
their properties have changed.
A noun has different properties at the subject or object position, or at the
modifier position:

(23) a. 买木头 (buy wood)   买十根木头 (buy ten pieces of wood)

b. 木头房子 (wooden house)   *+根木头房子 (a house of ten
pieces of wood)

木头 (wood) can be modified by numerals and measure words at the sub-

ject or object position, but cannot be modified at the modifier position, indi-
cating that its properties have changed. But with de (的) or di (地) added, the
original ungrammatical sentences are now grammatical:

(24)
不/​很干净的衣服 (not/​very clean clothes)
(25)
不/​很认真地学习 (study not/​very seriously)
(26)
十根木头的房子 (a house of ten pieces of wood)
Therefore, the constituent at a modifier position should be regarded as the
third expressional function: a modification.
The characteristics of a modification are as follows: it modifies and restricts
a statement or reference; its meaning is outward oriented, dependent by itself
and dependent on a statement or a reference, and cannot be modified by
attributives and adverbials.
There are two types of modifications: substantive modifications and predi-
cate modifications. A  substantive modification modifies a reference and
functions as an attributive; a predicate modification modifies a statement and
functions as an adverbial.
The direct use of 干净 (clean), 认真 (serious) and 木头 (wood) as
attributives or adverbials has different properties from the above words
followed by de (的)/​di (地) to be used as attributives or adverbials. When they
are directly used as attributives or adverbials, their expressional function
is a modification, and thus they cannot be modified by other modifiers.
But when they are followed by de (的)/​di (地) and used as attributives or
adverbials, they still function as statements or references, and still have the
characteristics of a statement or a reference. Namely, they can be modified
by other modifiers.

4.3.2.3 Auxiliary
Auxiliaries refer to the expressional functions of functional words such as
prepositions, conjunctions, modal particles, auxiliary words and so on. They
can be regarded as neither statements nor references nor modifications, and
are added to functional words to play the following auxiliary roles:
80

80  Essence and expressional functions

1. Change the expressional function of a supplementary constituent. For
example, de (的) denotes a modification; zhe (者) a reference; zhi (之) a
modification.
2. Denote a certain supplementary meaning. For example, a modal par-
ticle denotes a mood; auxiliary words like le (了), zhuo (着) and guo (过)
denote tense or aspect.
3. Play a conjunctive role like a conjunction.

In addition, the expressional function of an interjection is neither a

statement, a reference, a modification nor an auxiliary. An interjection
denotes a call, answer or exclamation. In fact, it is a constituent beyond the
language system because it has a special phonological system without tone
and is always used independently.

4.3.3  The expressional patterns and performance mechanisms of a language

References and statements form the basic building blocks of expressional
patterns of a language. A  modification cannot be used independently but
modifies or restricts referential and statement constituents by being attached
to them. An auxiliary word is attached to referential, statement or modifi-
cation constituents to play a regulatory or transformational role (see Guo
Rui, 1997). The relationships among expressional functions are shown in the
following diagram:

reference———statement (primary opposites)

│ │

Substantive modification———predicate modification (secondary opposites)

Auxiliary constituents are attached to references, statements or

modifications and are not shown in the diagram. The diagram indicates that
a complete meaning can be expressed with the opposite of a reference and
a statement. A  modification does not change the primary opposition but
restricts it. In other words, a modification is attached to a reference, and its
combined form is still a reference, for example, 彩色电视 (color television).
A  modification is attached to a statement, and its combined form is still a
statement, for example, 认真学习 (study seriously). Namely, a modification
has no independence. An auxiliary word can be attached to any position of
the two opposites and play a regulatory role.
We notice that the combination of a statement with a reference forms
a larger unit whose expressional function is still a statement, for example,
both 看+书 (read a book) and 他+看 (he reads) are still statements. Thus,
we can say that a statement is more independent than a reference. Therefore,
we can conclude that a statement is the most primary and important expres-
sional function in a language. A  modification depends on a statement or a
 81

Essence and expressional functions 81

reference; the combination of a statement with a modification still produces a
statement on the whole. The combination of a reference with a modification
still produces a reference on the whole. Therefore, the status of a modification
is the lowest. Its independence, importance and size can be expressed with the
following inequation:

Statement > reference > modification > auxiliary

4.3.4  Hierarchies of expressional functions

(27) a. 小王黄头发 (Little Wang has yellow hair).
b. 小王也黄头发 (Little Wang has yellow hair, too).
c. 小王一头黄头发 (Little Wang has a head of yellow hair).

In Example (27) a, 黄头发 (yellow hair) can be used to answer the question
“what is it like?” and can be added with an adverbial, thus being a statement.
However, it can also be added with an attributive, and in this regard, it should
be considered as reference. This is contradictory. But if an adverbial and an
attributive appear simultaneously, the adverbial is always preceded by the
attributive. Namely, the adverbial is in the outer layer; the attributive is in the
inner layer.

(28) a. 小王也一头黄头发 (Little Wang has a head of yellow hair, too).

b. 小王一头也黄头发 (The head of Little Wang has yellow
hair, too).

This indicates that the expressional function of 黄头发 (yellow hair) has
two layers: the inner layer is a reference; the outer layer is a statement. For
another example:

(29) a. 这本书的出版 (the book’s publication)

b. 这本书的及时出版 (The book’s punctual publication)
c. *及时这本书的出版 (*Punctual the book’s publication)

In Example (29) a, 出版 (publication) is modified by an attributive and

is thus a reference. But it can be added with an adverbial, which can only
be followed by attributive b but not preceded by attributive c. Therefore,
出版 (publication) is a statement in its inner layer but a reference in its
outer layer.4
In the above, we classified expressional functions into two combination
layers. From the perspective of their own properties, we can classify them
into intrinsic and external expressional functions. The intrinsic expressional
function of a word is inherent; the external expressional function is the
one ultimately realized by a word in its certain grammatical position. The
expressional functions in the two layers are in agreement. 小王 (Little Wang)
82

82  Essence and expressional functions

in Example  25 is a reference either in the inner layer or in the outer layer.
Sometimes they are not in agreement, for example, 黄头发 (yellow hair) in
Example (28) a and 出版 (publication) in Example (29) a. In this case, the
external expressional function of a word works temporarily.
The distinction between a statement and a reference by using “what is it
like?” and “what” is in the outer layer. Example (28) a can only answer the
question “what is it like”? instead of “what”? Example (29) a can only answer
the question “what”? instead of “what is it like”?

4.3.5  Expressional functions and parts of speech

In essence, expressional functions are the bases for distinguishing words in the
large category of parts of speech such as substantive words, predicate words
and modification words. Please look at the following examples (Lu Jianming,
1991a):

(30) 我们厂只做~,不做~ (Our factory only makes ~, does not make ~).

The words like 柜子 (cabinet), 桌子 (desk), 沙发 (sofa), 板式 (plate type) and

框式 (frame type) can appear at the two positions in Example (30), having the
construction “make ~”. Some of them are substantive words and others are
non-​substantive words like 板式 (plate type) and 框式 (frame type), but 做
(make) is usually regarded as a verb that can take only a substantive object. In
fact, the so-​called substantive object is a reference object.
Furthermore, 应该 (should) in Chinese can be followed by either 去 (go),
看 (see) or 三个人 (three persons), 星期一 (Monday), 阴天 (overcast) and
others. It is regarded as a verb that can take a predicate object.
Words like 赏 (reward), 罚 (punish) and 诛 (kill) in classical Chinese are
regarded as verbs that take substantive objects, but we can find examples such
as 赏有功 (reward the meritorious), 罚有罪 (punish the guilty) and 诛不义
(kill the unrighteous). In fact, the so-​called substantive object is a reference
object.
In the above examples, the concepts of word properties such as the so-​
called substantive word and predicate word actually involve the expres-
sional functions such as reference and statement. It is a pity that we did not
realize this in the past. Here we can identify word properties with expres-
sional functions. But how should we explain the contradictions that a verb
functions as a substantive object (reference object) such as 有功 (meritorious)
in the phrase 赏有功 (reward the meritorious), and that a substantive word
functions as a predicate object (statement object) such as 阴天 (overcast) in
the phrase 应该阴天 (should be overcast)? This has something to do with the
hierarchies of parts of speech. Just like an expressional function is classified
into two layers, so is a part of speech.
 83

Essence and expressional functions 83

The intrinsic basis for the distinction among parts of speech such as a noun
and a verb is actually the distinction between expressional functions such as
reference and statement. The distributional and morphological distinctions
between parts of speech are simply the extrinsic manifestations of distinctions
in expressional function. Just like expressional functions have layers, so should
parts of speech be classified into two levels. The parts of speech that corres-
pond to intrinsic expressional function are called parts of speech at the lexical
level; the parts of speech that correspond to extrinsic expressional function
are called parts of speech at the syntactic level. The part of speech at the
lexical level is a word’s inherent one, which is that labeled in a dictionary,
while that at the syntactic level results from the use of a word, which needs
to be controlled by syntactic rules. The parts of speech at the two levels are
generally the same, but are different in a few cases. The 出版 (publication)
in Example (29) a is a verb at the lexical level but a noun at the syntactic
level. Corresponding to the intrinsic levels of three expressional functions, the
parts of speech of substantive words can also be classified into three: predi-
cate word (statement), substantive word (reference) and modification word
(modification). If parts of speech are not classified into levels, it may be dif-
ficult to describe Chinese sentence structures with categories such as NP and
VP. For example, Example (10) b is represented as “VP+VP” and Example
(27) a as “N+NP”. Such representations hardly reveal their phrasal structures.
How should we understand the difference in the two layers of expressional
function? We believe that there exists a covert marker of expressional function
transformation, as shown in the following analysis:

(31) 小王也一头黄头发 (Little Wang has a head of yellow hair, too).

IP
╱╲
NP I’
│ ╱╲
小王 I vP
╱╲
AdP v’
│ ╱╲
Ad v NP
│ │ ╱╲
也 (是Ø) QP N’

一头 黄头发
84

84  Essence and expressional functions

(32) 学习很重要 (Study is very important).5

IP
╱╲
NP I’
╱╲ ╱╲
NMk VP I VP
│ │ ╱╲
Ø V AdP V
│ │
学习 很 重要

(33) 这本书的出版 (the publication of the book)

NP
╱╲
AP N’
╱╲ ╱╲
NP AMk NMk VP
│ │
这本书 的 Ø 出版

黄头发 (yellow hair) in Example (31) still has the properties of a noun
because it can take its attributive 一头 (a head of). Because the functioning
of a substantive phrase as a predicate denotes judgment and requires the
appearance of the overt 是 (is), for example, 小王不是黄头发 (Little Wang
is not of yellow hair.), it can be treated in the way that its syntax requires
that it should have a covert link verb. Because the link verb 是 (is) is silent,
it seems that the pronounced 一头黄头发 (a head of yellow hair) undertakes
the function of 是一头黄头发 (is of a head of yellow hair), thus having the
properties of a predicate word on the whole. In Example (32), the verb 学习
(study) does not turn into a noun because it can still take its object or adver-
bial, for example, 学习语法很重要 (studying grammar is very important);
认真学习很重要 (studying seriously is very important). We believe that the
verb 学习 (study) in the position of subject takes the covert marker (NMk)
of a substantive word, whose property of reference is created by the covert
marker. In Example (33), AP denotes a substantive modification phrase;
AMk denotes the substantive modification word de (的), namely the attribu-
tive marker de (的, meaning “of ”). The attributive 这本书的 (of this book)
modifies 出版 (publication), which has the covert marker of the properties of
a noun. Therefore, the nature of the distinction between the parts of speech at
 85

Essence and expressional functions 85

the internal and external levels is the distinction between the language classes
at the upper and lower levels: the upper level is NP; the lower level is the nom-
inalization label and VP. Because the nominalization label is covert, the upper
and lower levels have the same form.
It is difficult to do a comprehensive analysis of Chinese with formal grammar.
One of the reasons for this is the difficulty in describing phrasal structural rules,
which are caused by the multiple functions of Chinese parts of speech. It is per-
haps feasible to solve the problem by classifying parts of speech into two levels
and treating the difference in the two levels with a covert constituent.
The classification of parts of speech into two levels can better explain the
contradictions in such examples as 诛无礼 (kill the impolite) and 敌大 (rival the
large), in which verbs that should take substantive objects, however, take predi-
cate objects. The two phrases come from:  得志於诸侯,而诛无礼,曹其首也
(successful because of dukes and princes, but kill the impolite and get them
beheaded) (The 23rd Year during the Xigong’s Reign in The Book of Zuozhuan);
小固不可以敌大 (the small cannot rival the large) (Book One on Lianghuiwang
in Mencius). 无礼 (impolite) and 大 (large) are predicate words at the lexical
level but substantive words at the syntactic level, meeting the requirements
for the verbs 诛 (kill) and 敌 (rival) that take substantive objects. Therefore,
the verbs that take substantive objects and predicate objects use the parts of
speech at the syntactic level. Auxiliary verbs in modern Chinese are usually
regarded as verbs that take predicate objects, but we also find that some aux-
iliary verbs can take substantive objects, for example, 应该阴天 (should be
overcast). The contradiction that verbs that usually take predicate objects can
take substantive objects can be solved only by classifying parts of speech into
two levels. 阴天 (overcast) is a substantive word at the lexical level but a predi-
cate word at the syntactic level; therefore, it is still a predicate object.
In general, English adjectives can play the role of the headword of a noun
phrase. Similar to all noun phrases, these adjectives can function as subject,
object and prepositional complement, but the difference is that they have no
plural suffices and no changes in possessive case suffices (Quirk, Greenbaum,
Leech & Svartvik, 1985). This can be explained at the level of word proper-
ties:  adjectives have their noun properties at the syntactic level but are still
adjectives at the lexical level, thereby being able to be modified by degree
adverbs and having changes in degrees. However, the plural forms and
possessive cases of English nouns have their noun properties at the lexical
level, which cannot appear at the syntactic level.

4.4  Essence of a part of speech

At the hierarchy of large categories such as substantive words, predicate words
and modification words, a part of speech is actually the classification of words
into their categories that uses their expressional functions at the lexical level
as an intrinsic basis. The difference in a word’s part of speech comes before
86

86  Essence and expressional functions

its distributional difference; the properties of its part of speech are inherent
rather than temporarily resultant in use.
Hengeveld (1992) corresponded parts of speech with the four grammat-
ical positions in a statement phrase and a reference phrase (see Chapter 3.1
in Volume 2 for detailed discussion). The four grammatical positions out of
two core positions actually contain different specific grammatical positions
respectively:

Reference phrase core: subject, object, attributive ~ (adjective ~, numeral

and measure word ~, numeral ~)
Statement phrase core: predicate, ~ object, ~ complement, complement,
adverbial ~
Modifier of a reference phrase: attributive
Modifier of a statement phrase: adverbial

In other words, grammatical positions such as “predicate, ~ object, ~

complement, complement, adverbial ~” have the same functions for clas-
sifying parts of speech. The presence of one of the functions indicates
that a grammatical position has the properties of a predicate word. Why?
Because these functions have the same expressional function –​statement.
Grammatical positions such as “subject, object, attributive ~” represent the
properties of a substantive word; a modifier represents the properties of a
modification word.
In practice, thousands upon thousands of possible classifications of
parts of speech may be obtained purely according to distribution. The
fundamental reason why we choose the part-​of-​speech system of predi-
cate words, substantive words, predicate modification words and substan-
tive modification words is that out of the thousands upon thousands of
possible classification results, we merely choose the classification results
that are in agreement with expressional functions. The fundamental reason
why we should choose the part-​of-​speech system that is in agreement with
expressional functions is that expressional functions such as statements,
references, predicate modifications and substantive modifications reflect
the basic working mechanisms of a human language, which, in other words,
express meanings by forming sentences through combining statements with
references. The vocabulary of a human language by and large differs in
grammatical function and morphology in accordance with differences
in its expressional functions. Just for these reasons, the vast majority of
human languages distinguish among predicate words (verb), substantive
words (noun), adverbs and adjectives (distinctive word) that are dependent
on predicate words and substantive words. Therefore, the part-​of-​speech
system that is in agreement with expressional functions can most effectively
grasp the basic grammatical rules of a language.
The large category of part of speech (predicate words, substantive
words, substantive modification words and predicate modification words) is,
 87

Essence and expressional functions 87

essentially, a type of expressional function, but whether a type is classified
into different parts of speech or not depends on whether a lexical item differs
in terms of expressional function or not. For example, in some languages,
there is no difference between words that can enter the reference phrase core
and words that can enter the reference phrase modifier and the statement
phrase modifier; they basically belong to the same type of words. Then,
in spite of differences in expressional functions such as a reference and a
modification, there is no need to classify substantive words and modifica-
tion words. Only when lexical items roughly differ in terms of expressional
functions, and when the words that appear in different grammatical positions
do not basically coincide is it necessary to classify them into different large
categories.

Pattern 1: No need to classify parts of speech

Pattern 2: Need to classify parts of speech

Pattern 3:  Need to classify parts of speech, and treat the overlapping part as a
conversional word
88

88  Essence and expressional functions

Pattern 4:  Need to classify parts of speech, and treat the overlapping part as
conversional word or not as a conversional word

The classification of predicate words, substantive words and modification

words into such minor categories as nouns, verb, adjectives, distinctive words
and adverbs is not decided by distribution entirely. The minor categories inside
predicate words, substantive words and modification words (nouns, measure
words, locatives, time words, place words, verbs, adjectives, state words, dis-
tinctive words, adverbs, numerals, measure words and demonstratives) are
types that combine a semantic type with its syntactic function. A  semantic
type is the categorized semantic meaning, including entity, position, state,
method, number, numeral and measure word, demonstrative and so on. Only
when the semantic type is in agreement with its syntactic function is it neces-
sary to classify words into their independent parts of speech. In other words,
only when the difference in semantic type agrees with the difference in syn-
tactic distribution is it necessary to classify words into their parts of speech.
For example, a unit word is different from an entity word in terms of semantic
meaning, but in terms of syntactic distribution, the two are differentiated: a
unit word can enter into the “numeral ~” environment, whereas an entity word
cannot. Therefore, modern Chinese can have the different parts of speech of
measure words and nouns. However, although the Chinese during the Pre-​Qin
dynasty (221–​207 BC) differed in semantic types such as unit word and entity
word, its syntactic distribution did not differ. Both types of words can enter
the “numeral ~” environment, and as a result, there is no need to classify them
into measure words and nouns.
Action words and attribute words within a predicate word differ in
semantic type. Their syntactic distributions also differ:  an attribute word
can enter the 很 (very) ~ environment, but an action word generally cannot.
Therefore, adjectives and verbs should be classified into two parts of speech,
but this difference in distribution is not very thorough. Words such as 害怕
(fear) and 喜欢 (like) can enter the 很 (very) ~ environment but are closer to
action words. Obviously it is inappropriate to classify them into adjectives.
Accordingly, it is necessary to supplement the criterion “很 (very) ~, ~ object”
and to take 害怕 (fear) and 喜欢 (like) away from adjectives and include them
in verbs.
Attribute words and state words within the category of predicate words
are different in semantic type; their syntactic functions are also different: state
 89

Essence and expressional functions 89

words cannot enter the 很 (very) ~ environment. Therefore, they should be
classified into different parts of speech.
The difference in pure distribution cannot be taken as the reason for parts-​
of-​speech classification. For example, 年事 (person’s age) can only function as
a subject; 地步 (extent) can only function as an object. They are different in
distribution, but their semantic meanings belong to the same category (entity);
therefore, there is no need to classify them into different parts of speech. 采取
(adopt) can take object, but 休息 (rest) cannot. They are different in distribu-
tion, but their semantic type is the same (behavior); therefore, they should be
classified into verbs.
Neither is the pure identity of semantic type the reason for a word to
be in the same part of speech. For example, 睡觉 (sleep), 打仗 (fight) and
开刀 (operate on) semantically denote action. 睡眠 (sleep), 战争 (war), 手术
(operate) also semantically denote action, but their distribution differ greatly.
Therefore, they cannot be classified into the same part of speech. The former
belong to verbs, while the latter belong to nouns. 分 (minute), 天 (day), 年
(year) 月 (month) and 钟头 (o’clock) are all units for measuring time, but
their distribution is different. The former can enter the “number ~” environ-
ment, whereas the latter cannot. Therefore, the former should be classified
into measure words, and the latter should be classified into nouns.
荣幸 (honored), 痛 (painful), 饱 (full) and 大型 (large-​scale), 野生 (wild)
and 急性 (acute) all denote attributes, but their distribution is greatly
different. The former can function as predicates but cannot function as
attributives; the latter can function as attributives but cannot function as
predicates. More importantly, differences in predicates and attributives reflect
those in the expressional functions such as statements and modifications. The
two large categories of predicate words and modification words should be
classified. Therefore, it is inadvisable to classify the two categories of words
into adjectives. The classification of the former into adjectives and the latter
into distinctive words has not only a distribution basis but also a semantic
basis. Although adjectives and distinctive words have the semantic type of
properties, a Chinese adjective (predicate word) is an attribute word that has
the properties of degree and can be modified by a degree adverb, whereas a
distinctive word is an attribute word that has no properties of degree (modifi-
cation word) and cannot be modified by a degree adverb.
Therefore, the classification of words into different parts of speech requires
both a distribution basis and a semantic basis. The difference in pure distri-
bution and pure semantic meaning cannot be used as the basis for classifying
parts of speech. Comparatively speaking, the difference in semantic type may
be more important for classifying parts of speech. If there is a difference in
semantic type but no difference in distribution, then the classification should be
preferably based on semantic type. For example, 滂沱 (pour), 逶迤 (meander),
斑驳 (mottle), 旖旎 (graceful) and 交加 (mixed), 参半 (half-​and-​half), 洞开
(bright) and 参天 (towering) are predicate words, which, however, can only
function as predicates so far as their predication function is concerned. But
90

90  Essence and expressional functions

so far as their semantic type is concerned, the former denote state, whereas
the latter denote action. Therefore, words like 滂沱 (pour) should be classified
into state words, and those like 交加 (mixed) should be classified into verbs.
The study of Chinese grammar in earlier days tended to regard parts of
speech as semantic types. Since the 1950s, the distribution theory has come
into fashion in the Chinese grammar circle. During recent years, some scholars
have picked up again the view that a part of speech belongs to a semantic type
(Shi Dingxu, 2007, 2009; Lu Bingfu, 2010; Ren Ying, 2010). My view lies
in between the two extreme views: a part of speech, in essence, is neither a
pure distribution type nor a pure semantic type. Basically, the large categories
of parts of speech (substantive words, predicate words, modification words)
are classified according to the intrinsic expressional functions of a word. The
basic categories of parts of speech (nouns, measure words, locatives, place
words, time words, verbs, adjectives, state words, distinctive words, adverbs,
numerals, measure words, demonstratives and others) are types that combine
the semantic type of the word with its syntactic function. The reason why
the semantic type is important for a part of speech is that it reflects language
categorization, while the distinctions among parts of speech are exactly the
categorization results. The expressional function is the meaning-​expression
pattern of a word, and the semantic type is the categorized meaning. The two
belong to grammatical meaning; thus, we can say that grammatical meaning
is the intrinsic basis for parts of speech.
Regarding the combination of expressional function, semantic type and
distributional type rather than distribution as the nature of a part of speech
can explain why parts of speech are comparable in terms of different times and
languages. The view that distribution is the nature of a part of speech cannot
explain why words distributing differently in different times and languages
yet belong to the same part of speech. For example, the words 看 (see) in
modern Chinese and 视 (watch) in classical Chinese distribute differently. The
former can take a numeral and a measure-​word object but cannot be modi-
fied by a numeral, such as 看三次 (see three times) but not *三看 (three sees),
while the latter cannot take a numeral and a measure-​word object and can be
modified by a numeral, such as 三视 (watch three times) but not *视三 (watch
three). But both are verbs. The English word 石头 (stone) can be modified by
a numeral, for example, two stones, and can function as the object of a place
preposition, for instance, on the stone, whereas the Chinese word 石头 (stone)
does not have these functions. But both are nouns. The view that the nature
of a part of speech is the combination of its expressional function, semantic
type and syntactic function can produce a reasonable explanation:  because
they have the same expressional functions and semantic types.
A part of speech is a category. That is to say, different parts of speech
must have qualitative distinctions, which are the basis for comparison among
different languages.
We regard the basic category of parts of speech as the combination of the
distribution with the semantic type. Just because of this, different parts of
 91

Essence and expressional functions 91

Table 4.4 Semantic bases of parts of speech

Part of speech Expressional Semantic type

function

Verb Statement Behavior and action

Adjective Statement Property or degree
State word Statement State description
Noun Reference Entity
Measure word Reference Measurement unit
Time word Reference Absolute time position
Place word Reference Absolute space position
Locative Reference Relative position
Numeral Modification Numerical value
Numeral and measure word Modification Quantity
Distinctive word Modification Property of non-​degree
Demonstrative Modification Demonstration
Onomatopoeia Modification Sound imitation
Functional word Supplementation Supplementary meaning
Interjection word Interjection Independent interjection and echo

speech have different semantic bases, and we can use the lexical meaning to
help judge them. But definition and explanation do not mean criteria; expres-
sional functions or meanings can be used to give definitions and explain the
nature of parts of speech, but cannot be used as criteria for classifying parts
of speech simply because they cannot be directly observed. Therefore, we still
should find directly observable criteria for classifying parts of speech, such as
distribution and morphology.
In fact, it is not this book that originally created the view that distribution
is not the essence of a part of speech. Many other scholars hold a similar view-
point, which can even be regarded as returning to old ways. In the following,
we present other views that hold that distribution is not the essence of a part
of speech:

1. Discourse function is the nature of a part of speech. Hopper and Thompson

(1984) pointed out that the aim of discourse is to report events that
happen to a participant. There are two basic discourse functions: eliciting
participants and reporting events, which are lexicalized exactly by nouns
and verbs. Against a deeper background, Hopper and Thompson thought
that discourse function is the root factor for restricting grammar, and that
grammatical forms simply make it permanent. Speaking in the extreme,
there is only discourse function and no grammar at all. From a histor-
ical genesis point of view, language forms lack categories in principle.
Namely, there is no categorical opposition between parts of speech per
se; rather, there are only different discourse requirements. It is considered
that a large number of words have such requirements, and an opposition
between a noun and a verb is thus formed.
92

92  Essence and expressional functions

Our viewpoint is different; we use the language expression pattern to
observe the opposition between parts of speech in terms of category,
holding that the opposition between parts of speech is fundamentally the
different classification of labor of lexical forms in a language expression
pattern. In other words, there exists first and foremost the opposition
between expressional functions, which simply lexicalize parts of speech
or make them permanent in a lexical form. A discourse function is quite
exterior and based on an expressional function. Its motive tends to be
grammatical and semantic features, which are fundamental. The dis-
course function is simply the manifestation of grammar and semantics.
In other words, it must rely on a certain grammatical or semantic feature.
For example, the basis of “eliciting participants” is reference, and that of
“reporting events” is statement or assertion. The expressional function
comes before discourse function and is therefore the nature of a part of
speech.
2. Meaning is the nature of a part of speech. This view’s different versions
are briefly described below:
(1) Traditional grammar:  nouns denote names of things; verbs denote
actions; adjectives (attached to nouns) denote properties. This starts
from a conceptual meaning, which obviously cannot give a clear
explanation. For example, 手术 (operation on) denotes an action but
it is a noun.
(2) The theory of four categories by Magnusson (1994) borrows Aristotle’s
theory of category to explain the nature of a part of speech. Aristotle
classified solitary words into ten categories: material object: 人 (person)
or 马 (horse), numeral and quantity, for example: 三尺长 (three feet
long); property：白的 (white), 懂语法的 (grammar-​knowing)); rela-
tion:  一倍 (one time), 大于 (larger than); place:  在市场上 (in the
market); time:  昨天 (yesterday), 去年 (last year); posture:  躺着 (lie
down), 坐着 (sit); state:  穿鞋的 (show-​wearing, 武装的 (armed));
activity: 切割 (cut), 烧灼 (burn); passive meaning: 被刺 (assassinated),
被烧灼 (get burned) (Aristotle, Categories). He mainly discussed
the categories from two perspectives. One is the ontological per-
spective, which considers category as an existential class and holds
that a material object can exist independently and be self-​explained.
Other categories are generally called attributes (Aristotle, Posterior
Analytics) and cannot exist by themselves but exist in their subjects,
thus being substantially concomitant (Aristotle, Metaphysics) and
forming an opposition between substance and property. Magnusson
extracted the four most important categories of substance, quantity,
quality and relation to explain the essence of a part of speech:  an
adjective denotes attribute; a noun denotes substance and property
instead of relation; a pronoun denotes substance instead of property;
a preposition, adverb and conjunction denote relations; an article
denotes quantity; a verb denotes substance, property and relation.
 93

Essence and expressional functions 93

Fundamentally speaking, this is still a part of speech from the perspec-
tive of conceptual meaning.
(3) The universal and rational grammar thinks that a word denoting
substance is a noun, and that a word denoting accident is an adjec-
tive. Both noun and adjective denote the object of thought. A verb
denotes a way of thinking; in other words, it denotes the judgment
of a thing and makes an assertion (see Antoine Arnaud, Claude
Lancelot 1660). We agree with this viewpoint and think that the view
that an expressional function is the nature of part of speech returns
to the old view of universal and rational grammar in this regard.
The universal and rational grammar simply did not say explicitly the
expressional function, but the essential idea is the same. That is to say,
both start from the perspective of a language expressional pattern
or organization (the universal and rational grammar called this a
form of thinking and held that speech is its expression) to explain
the opposition between parts of speech. Of course, the universal and
rational grammar only rather carelessly distinguishes between two
types of functions:  object of thinking (nouns, articles, participles,
prepositions, adverbs) and way of thinking (verbs, conjunctions,
interjections).
(4) Givón (1984) uses time stability to explain the meanings of a part of
speech. Language tends to encode what is relatively stable in time as
a noun and what is rapidly changing as a verb. The attribute meaning
is intermediate in time stability; therefore, some languages encode it
as a noun, and others encode it as a verb.
Some scholars argue that it is difficult to control time stability. For
example, the time stability of “fire” and “flicker” is weak, but they
are nouns, whereas that of “to tower” is strong, but it is a verb (see
Whaley, 1996).
(5) Langacker (1987a, b) started from cognitive grammar to explain
the meanings of a part of speech. Nouns denote things; verbs
denote processes; adjectives/​adverbs denote a temporal relation.
Fundamentally speaking, this approach still uses conceptual meaning
and is perhaps right, but its operation and falsifiability are too weak.

4.5  Nature of classifying parts of speech

The aim of part-​of-​speech classification is actually to deduce the intrinsic
grammatical properties (expressional functions) of a word from observable
extrinsic characteristics. Strictly speaking, the nature of part-​of-​speech clas-
sification is thus to discover and identify classes. The criteria for classification
are not necessarily the intrinsic characteristics of a class. A form is not the
nature of a part of speech, but we can use it as criteria for classifying parts of
speech. Similarly, distribution also is not the nature of a part of speech, but
we simply use it as criteria for classifying parts of speech. Of course, some
94

94  Essence and expressional functions

of the extrinsic characteristics we use to classify parts of speech are closer
to intrinsic characteristics; others are further away. For the purpose of clas-
sifying parts of speech, form is a relatively extrinsic characteristic, whereas
distribution is a rather intrinsic characteristic. Generally speaking, intrinsic
characteristics are more reliable than extrinsic characteristics but are not
easily observable, whereas extrinsic characteristics are easily observable but
not so reliable.
Then, can we classify parts of speech directly according to intrinsic
characteristics instead of deducing them indirectly according to extrinsic
characteristics? Theoretically yes. The key is that we can directly observe
intrinsic characteristics. Whether they are observable or not sometimes
depends on our analysis instrument. As progress in analysis instruments is
made, what we cannot observe today will perhaps be observable in the future.
The reason why today we do not use the intrinsic characteristics (expressional
functions) of parts of speech as the criteria for classifying them is that the
characteristics are not directly observable. If some day progress in analysis
instruments is made and we can observe expressional functions, we may well
use them as criteria for classifying parts of speech.
After the intrinsic characteristics for classifying parts of speech are well
understood, we can answer the following question raised by Gao Mingkai: since
the properties of parts of speech are static and inherent, why can they be
classified according to grammatical functions (dynamic uses)? This simply
deduces the inherent properties of a word according to its extrinsic form. The
functions in use are not used as the nature of a part of speech.
Just because distribution is not the nature of a part of speech but simply
its extrinsic characteristics, it is not the only criterion for classifying parts of
speech.
This book suggests that the properties of a word exist before a linguist’s
classification of parts of speech and are part of a language’s constituent
structures. The part of speech based on the properties of a word is not of the
distributional class; distribution is only their extrinsic manifestation. Parts
of speech have no entire correspondence with distribution. It is impossible
to classify parts of speech purely according to the similarity or difference in
distribution even if the prototypal model is used. The root cause of the selec-
tional restriction of a grammatical position on words is that words themselves
have differences in their properties. To classify parts of speech, we use distri-
bution to infer the differences in a word’s own properties that cause different
distributions. These properties are the nature of a part of speech.
Therefore, we entirely agree with Gao Mingkai’s (1960) view that a word’s
morphological changes, combinational capability, syntactic functions and
so on are all the extrinsic manifestations of its part-​of-​speech meaning. Our
difference from him is that Gao Mingkai thought that Chinese notional words
are versatile and that a word has no parts of speech. But we do not think so
and believe that the selectional restriction of grammatical position on a word
exists. Although the selectional restriction of the rather abstract grammatical
 95

Essence and expressional functions 95

position (syntactic constituent) on Chinese words is not as strict as that on
Indo-​European languages, the selectional restriction is quite strict at rather
specific grammatical positions such as 很 (very) ~ or 不 (not) ~. Hence, we can
use the rather specific distributional characteristics of a word to classify its
part of speech. Moreover, at the position of a syntactic constituent, the dom-
inant selection is very obvious. For example, more than 88% of subjects are
held by substantive constituents; over 99% of predicates are held by predicate
constituents.

4.6  Bases and criteria for classifying parts of speech

We accept Wen Lian’s (1995) view that there should be a distinction between
the basis for classifying parts of speech and the criteria for doing so. The
basis for the classification refers to the intrinsic characteristics of a part of
speech; the criteria refer to the conditions required to pinpoint which class a
word belongs to. The basis may be what cannot be observed directly, but the
criteria must be what can be observed. The two may be in conformity or not,
but the criteria must be able to reflect the basis for classification. The basis is
the intrinsic expressional function or grammatical meaning of a word, but
because they cannot be observed directly, we use its observable morphology
or grammatical function as the criteria. They can reflect the intrinsic expres-
sional function or grammatical meaning.
The use of different things as the intrinsic basis for classifying parts of
speech may produce very different results. The use of morphology as the
intrinsic basis may produce Varro’s classification. Not supported by expres-
sional function or grammatical meaning, this kind of classification is similar
to Linnaeus’ bio-​taxonomy. The use of distribution as the intrinsic basis
may produce Chen Xiaohe’s (1998) classification, which is also similar to
Linnaeus’ bio-​taxonomy, and in which the classes do not reflect categories.
But the use of expressional function as the intrinsic basis is different from
the above-​mentioned two kinds of classification because the difference in
distribution, in our opinion, is only the extrinsic manifestation of expres-
sional function. We should obtain different or similar expressional functions
from different or similar distributions. In this way we do not classify parts
of speech simply according to distribution but neglect some differences in
distribution, for example, whether a word can be modified by 没 (no) or 不
(not). We also neglect some commonality in distribution (for example, nouns,
verbs and adjectives all can function as subjects). Through analyzing distri-
bution compatibility, we find the expressional functions that restrict distribu-
tion and connect certain distributions with certain expressional functions (see
Chapter 6).
Therefore, classifying parts of speech is to identify the existing properties
of a word and then do the classification according to certain classification
strategies (see Chapter 7). The properties of a word exist first, and then cri-
teria are identified.
96

96  Essence and expressional functions

4.7  Transformation of expressional function and part of speech

4.7.1  What is the transformation of expressional function?

Zhu Dexi (1983) pointed out that a statement can be transformed into a ref-
erence. “The addition of de (的) to the end of VP transforms the VP that
originally indicates a statement into the ‘VP de (的)’ that indicates a refer-
ence”. Although we do not agree that the function of de (的) is a reference, we
agree that an expressional function can be transformed. Lu Jianming (1991)
discussed the transfer reference in modern Chinese. We will further discuss the
transformation of expressional functions as follows:
急性 (acute) and 慢性 (chronic) in Example (9)  b in Chapter  4.3 can be
regarded as the transformation of a modification into a reference; 黄头发
(yellow hair) in Example (27) a is the transformation of a reference into a
statement; 出版 (publication) in Example (29) a is the transformation of a
statement into a reference. Classical Chinese also has the transformation of
expressional functions, for example, 贤 (virtuous), 不贤 (a person of no virtue
and ability), 无功 (not meritorious) and 不乐生 (not enjoy life) are the trans-
formations of a statement into a reference.
As a universal phenomenon, other languages also have the transformation
of expressional functions, for example:

In English:
(34) The extremely old need a great deal of attention. (extremely old: a
modification transformed into a reference)
(35) The number of jobless is rising. (jobless: a modification transformed
into a reference)
(36) Mary’s was the prettiest dress. (Mary’s: a modification transformed
into a reference)
(37) We’ll meet at Bill’s. (Bill’s:  a modification transformed into a
reference)

In Spanish:
(38) a. Saludemos a los valientes combatientes.
Salute first person to fixed reference.
b. Siempre muestra gran respeto a los valientes. (valientes: modi-
fication→ reference)

In Hungarian:
(39) a. szép ház      b. A ház szép.
c. A szép kevés, a rossz sok. (szép, rossz: a modification or a
statement → a reference)
 97

Essence and expressional functions 97

In the Amis language:
(40) a. ta ajaj kuni a kuar.
b. nilkaj tuni a taʔaajaj kaku.(taʔarajaj:  a statement → a
reference)
(41) a. maumah-​aj ku tʃiwama
b. ninukaj-​ aj tu ku maumah-​ aj a maəmin. (maumah-​
aj: statement → a reference)
In addition, verbs or adjectives in a good many languages in the Altaic
language family also have transfer reference, for example, the Uiguric lan-
guage, Dongxiang language, Yugur language, Bonan language, Hezhen lan-
guage and so on. Adjectives in the Qiang, Monba, Yao and Bisu languages
and verbs in Vietnamese also have transfer references.

4.7.2  Types of expressional function transformation

First, the expressional function transformation can be classified into lexical
and syntactic ones.
The lexical transformation refers to the expressional function transform-
ation through word formation, taking English for example, from “see” to
“worth seeing”, from “read” to “reader”, from “work” as a verb to “worker”,
from “happy” to “happiness”. These are marked. “Lead” from verb to noun;
“define” from noun to verb, “water” from noun to verb, “cook” from verb to
noun. These have no markers.
The lexical transformation occurs at the level of intrinsic expressional
function. The marked lexical transformation actually derives a new word; the
part of speech of the word with no-​marker transformation at its lexical level
changes, thus being actually a conversional word. The lexical transformation
is beyond the detailed discussion in this book.
The syntactic transformation refers to the change in expressional function
through syntactic means, also having no-​marker and marked transformations.
The no-​marker transformation has no marker, but there is a temporary trans-
formation of the expressional function of a word at a certain syntactic position,
namely changing its part of speech at the syntactic level. The marked trans-
formation refers to the change in the expressional function of a word by adding
a functional word. Examples for the marked transformation are as follows:

(42)
小王黄头发 (Little Wang has yellow hair). (from reference to
statement)
(43)
这本书的出版 (the publication of this book) (from statement to
reference)
(44)
急性好治 (The acute are easy to treat). (from a modification to a
reference)
(45)
夫 尚 贤 使 能 ， 赏 有 功 ， 罚 有 罪 ， 非 独 一 人 为 之 也 ……
(《荀子·强国》(If the virtuous are esteemed and the able are
motivated, then the meritorious are rewarded and the guilty
98

98  Essence and expressional functions

are punished; these cannot be done by one person (Empower the
Country in Xunzi)). (from statement → reference)

The no-​marker syntactic transformation does so actually from intrinsic to

extrinsic expressional functions, namely transforming parts of speech from
lexical to syntactic levels. The transformation that involves various intrinsic
expressional functions is not a grammatical one but a lexical one, for example,
研究 (study), 讨论 (discussion), 危险 (danger) and 困难 (difficulty) have the
intrinsic expressional functions of statement and reference. Examples for the
marked syntactic transformation are as follows:

(46) a. 他看书 (He reads a book).

b. 看书的人 (the person who reads a book) (transformed from
a statement to a modification with the marker de (的)
(47) a. 生产有计划 (The production has a plan).
b. 有计划地生产 (produce in a planned way) (transformed from
a statement to a modification with the marker di (地)

The word zhi (之) in classical Chinese can be regarded as a marker for a
modification; zhe (者) is a marker for reference. The word “to”, a marker for
an English infinitive, can be regarded as a marker for reference; the possessive
case “-​’s” attached to a noun is a marker for a modification. The marked syn-
tactic transformation can be analyzed as follows (in the diagram, Mk indicates
transformation marker; AMk indicates the marker for distinctive word;
AdMk indicates the marker for adverb; NMk indicates the marker for noun):

(48)
有计划地生产 (produce in a planned way)
VP

AdP
V’
│
VP AdMk V
│ │
有计划 地 生产

(49)
一箪食,一豆羹,得之则生,弗得则死,呼尔而与之,行道之人弗受. (If
you get a basket of rice or a bowl of thick soup, you can live on;
if not, you may die. If you are asked to give, do not give it to the
person who preaches.) (Volume 1 on Giving You Advice in Mencius)
NP

AP N’
│
VP AMk N

│ │
行道 之 人
 99

Essence and expressional functions 99

The no-​marker syntactic transformation is caused by core constituent
vacancy, including modification core constituent vacancy and marked core
constituent vacancy, as analyzed previously. It is caused by different reasons.
For example:  小王一头黄头发 (Little Wang has a head of yellow hair).
学习很重要 (Study is very important). 这本书的出版 (the publication of this
book). The reasons are shown in Figure 4.6. The vacancy of Core Constituent
X causes Constituent Y that collocates with Core Constituent X independ-
ently functions as the mother node XP of X and Y, thus changing Y’s extrinsic
expressional function and part of speech at the syntactic level. For another
example, the VP modified by 不 (not) in the following Example (50) is actu-
ally vacant, and 不 (not), which was originally a modifier, now functions as a
statement. 急性 (acute) in Example (51), which was originally a modifier, now
functions as NP because the noun modified by 急性 (acute) is vacant. The
modifier 干净的 (clean) in the sentence 衣服干净 (The clothes (are) clean),
which originally formed a statement together with the link verb 是 (are), now
independently functions as a statement because the link verb is vacant. In the
following tree diagrams, Ø denotes a vacant constituent.

(50)
我不 (I do not).       (51) 急性好治 (The acute is easy to treat).

IP IP

NP I’ NP I’

A N I VP
ᡁ I VP
│ │
AdP V’ 急性 Ø 好治

н Ø       
From this perspective, the no-​marker syntactic transformation of expres-
sional functions is caused by two types of constituent vacancy: (1) the vacancy
of an expressional function transformation marker, for example, 学习很重要
(Study is very important); 这本书的出版 (the publication of this book); (2) the
vacancy of a core constituent, for example, 小王一头黄头发 (Little Wang has
a head of yellow hair); 我不 (I not). The expressional function transformation
marker is also a core constituent; therefore, the no-​marker transformation is
actually caused by the core constituent vacancy.

XP

Y (XØ)

Figure 4.6 The transformation caused by a vacant constituent

100

100  Essence and expressional functions

4.7.3  Reference and nominalization

Reference and nominalization happen at two levels respectively. The refer-
ence of an intrinsic expressional function corresponds to the nominalization
at the lexical level, and the reference of an extrinsic expressional function
corresponds to the nominalization at the syntactic level. The nominalization
at the two levels must be distinguished; the one at the syntactic level cannot
be identical with the one at the lexical level. The following examples show the
nominalization at the syntactic level. The parts of speech at the lexical level
do not change:

(52)
学习很重要 (Study is very important). (学习 (study) is a verb)
(53)
急性好治 (The acute are easy to treat). (急性 (acute) is a
distinctive word)
(54)
有大有小 (Some are large; some are small). (大 (large) and 小
(small) are adjectives)
(55)
赏有功，罚有罪 (reward the meritorious; punish the guilty). (有功
(meritorious) and 有罪 (guilty) are verb phrases)
(56)
贤者以其昭昭使人昭昭 (The wise use their clarity to make others
clear). (With All Your Heart in Mencius). Here 昭昭 (clear) is a
state word.

4.7.4  Syntactic constructions of self-​reference and transfer reference

Reference refers to the transformation from a non-​referential into a referen-
tial constituent. It has the following three constructions (Y indicates a non-​
noun constituent; NMk indicates the marker for a noun):
Types A and B are referenced through a reference marker. The reference
marker of Type A  is overt; the reference marker of Type B is covert. Type
C does not take a reference marker but has a covert noun headword. The
three constructions have the following in common: a non-​noun constituent
functions as the mother node of a noun independently or together with a
marker word.
Reference has self-​reference and transfer reference. From the perspec-
tive of formal grammar, the difference between the designations such as
self-​reference and transfer reference lies in the referential relations between
the mother nodes NP and Y. If the mother node NP refers to Y itself, then
the self-​ reference is designated, for example, 学习很重要 (Study is very
important), 这本书的出版 (the publication of this book). If the mother node
NP does not refer to Y itself but refers to the same thing as another noun
phrase at a certain node dominated by the mother node NP, then the transfer
reference is designated. This noun phrase is usually vacant, for example, in
Type C: 急性好治 (The acute are easy to treat), 看书的 (A person who reads
books). Let us look at the following examples of Type C:
 101

Essence and expressional functions 101

(Type A) NP

͔͕

Y NMk

Ί Ί

䘍⌅(violate law) 㘵(zhe)

(Type B) NP

͔͕

NMk Y

Ί Ί

᱕ཙ (spring)Ø ࡠᶕ(arrival)

(Type C) NP

͔͕

Y N

Ί Ί

ᙕᙗ (acute) Øྭ⋫(easy to treat)

Figure 4.7 Syntactic constructions of reference

(57)
不备不虞,不可以师. (《左传·隐公五年》) (If not prepared and
not dangerous, then no fighting) (The Fifth Year of the Yin Prince in
Zuozhuan)

VP
͔͕
V’ NP
͔͕
н༷ AP NP
͔͕ │
VP AMk N

н㲎 Ø Ø
102

102  Essence and expressional functions

(58)
王亲受而劳之,所以惩不敬,劝有功也. (《左传·成公二年》) (The
king himself receives and rewards him to punish those who are not
respectful and to persuade the meritorious.) (The Second Year of the
Cheng Prince in Zuozhuan)

VP
╱╲
V’ NP
╱╲
劝 AP N’
╱╲ │
VP AMk N
│ │
Meritorious Ø Ø

(59)
他有两个哥哥,一个高,一个矮 (He has two brothers; one is tall, and
the other is short).

IP
╱╲
NP I’
╱╲ ╱╲
QP N’ I VP

一个 Ø 高

(60)
看书的比买书的多 (there are more book readers than buyers).

NP
╱╲
AP N’
╱╲ │
VP AMk N
│ │
看书 的 Ø

(61)
The extremely old need a great deal of attention.

DP
╱╲
Det NP
│ ╱╲
the AP N’
╱╲ │
Adv A N
│ │ │
extremely old Ø
 103

Essence and expressional functions 103

(62)
Mary’s was the prettiest dress.

NP
╱╲
AP N’
╱╲ │
NP AMk N
│ │
Mary ’s Ø

We call the constituent referred to by the mother node NP a reference

source and borrow the concept of “extraction” proposed by Zhu Dexi (1983).
The process in which a reference constituent refers to a reference source
is called extraction. Thus, we can say that self-​reference extracts Y itself,
and that transfer reference extracts an NP below the mother node NP. For
example, the transfer reference of 急性 (acute) extracts the headword modi-
fied by 急性 (acute). 有功 (meritorious) is an extracted noun phrase modified
by 有功 (meritorious).6
The above examples show that the transfer reference of a modification
constituent in either Chinese or English usually extracts the modified vacant
NP. The transfer reference of the predicate constituent in modern Chinese
basically disappears.7 In classical Chinese, under usual circumstances, the
subject of a predicate constituent is extracted. For example (58), generally,
the marker suo (所) should be added to the object to be extracted. For
another example, 病而乞盟,所丧多矣 (《左传·僖公五年》) (When you
are sick and ask for union, what you lose will be plentiful.) (In the Fifth
Year of Xigong Prince in Zuozhuan). 寡人所好者,音也 (《韩非子·十过》)
(What I  like is music) (Chapter on Ten Faults in Han Fei Zi). There are
seldom if ever objects of transfer reference without adding suo (所),8 for
example in (57). The predicate constituents in the Nootka (see example
(2) in Chapter 3) and Tagalog (see example (3) in Chapter 3) languages also
extract subjects.

4.7.5  Functions of de (的) and di (地) in Chinese syntax

The function of de3 (的3) is actually related to the transformation of expres-
sional functions and parts of speech at the syntactic level. It is a marker for a
substantive modification word and can transform a predicate constituent or
substantive constituent into a substantive modification constituent. The “X de
(的)” that independently functions as a subject or an object is the no-​marker
transfer reference of a predicate modification constituent, namely the nom-
inalization at the syntactic level. The function of di1 (地) also transforms a
predicate or substantive constituent into a predicate modification constituent.
104

104  Essence and expressional functions

看书的 (a person who reads a book) transforms a predicate constituent into
a modification word. 木头 (wood) + de transforms a substantive constituent
into a substantive modification constituent. 有计划地 (生产) (produce in a
planned way) transforms a predicate constituent into a predicate modification
word; 历史地 (看问题) ((look at an issue) historically) transforms a substan-
tive word into a predicate modification word.
干净 (clean) in 干净的衣服 (clean clothes) is still a predicate word and
equivalent to a clause. It still has the general characteristics of a predicate
word:  不干净的衣服 (unclean clothes), 很干净的衣服 (very clean clothes).
But 干净 (clean) in 干净的衣服 (clean clothes) is actually a modification
word; therefore, it cannot be modified by 不 (not) and 很 (very): *不干净衣服
(* not clean clothes), *很干净衣服 (*very clean clothes).
It is crucially important to point out that de3 (的) and di (地) function
as a transformation marker in Chinese syntax and cause constituents with
different parts of speech to convert to each other. Therefore, the differences
that a word can directly function as attributive and that only by adding de3
(的) can it do so reflect its different parts of speech. A word that can directly
function as an attributive has a modifier property; that which can do so only
by adding de3 (的) has no modifier property and is still a substantive con-
stituent or a predicate constituent. Because of the special function of de3 (的)
and di (地) in Chinese syntax, we strictly distinguish between the function a
word has by adding de3 (的) and di (地) and the word’s own function.

4.8  Correlations among part of speech, expressional function

and syntactic constituent
Croft (1991) used typology to observe the universal parts of speech in the
world languages. He links part of speech with semantic meaning and prag-
matic function and believes that the three have correlations:
These are unmarked prototype correlations, whereas others are marked.
For example, the noun “vehicle” used as reference is unmarked, but we can
add a marker “ ’s” or use the derived adjective “vehicle’s” to indicate a modifi-
cation, or add the marker “be” to indicate a statement. The adjective “white”
used as a modification is unmarked, becomes the noun “whiteness” by adding
the marker “-​ness” to indicate reference and becomes “be white” by adding the
marker “be” to indicate a statement. The verb “destroy” used as a statement is

Table 4.5 Croft: Correlations among syntactic category, semantic type and pragmatic

function

Syntactic category Noun Adjective Verb

Semantic type Object Property Action
Pragmatic function Reference Modification Predication
 105

Essence and expressional functions 105

Table 4.6 Croft: The unmarked or marked correlations in English

Reference Modification Statement

Object vehicle Vehicle’s, vehicular, be a/​the vehicle

of vehicle, in vehicle
Property whiteness white be white
Action Destruction, to Destroying, destroyed destroy
destroy

Note: Bold letters indicate marked constituents.

unmarked, but we add the marker to or -​tion to indicate a reference and add
the marker -​ing or -​ed to indicate a modification.
According to Croft’s marker theory, the correlation among things,
references and nouns, and that among actions, statements and verbs are all
valid. But the correlation among attributives, modifications and adjectives is
not universal. In Chinese and most of the Sino-​Tibetan languages, an attribu-
tive mainly corresponds to a statement.9 The correspondence between part
of speech and semantic type is actually a matter of word formation but not
a matter of syntax. Some markers such as “-​ness” are also a matter of word
formation but not a syntactic one. If we only take into account the syntactic
matter, we can observe the correlation among part of speech, expressional
function and syntactic constituent. The correlation among the three is as
follows:

Part of speech Expressional function Syntactic constituent

Predicate word Statement Predication constituents

(predicate, complement, object
of real predicate-​object verb)
Substantive word Reference Subject, objects of verbs with
substantive object and quasi-​
predicate object, attributive
Modifier Modification Attributive, adverbial

The above is a prototypical correlation. Other correlations regarded as the

multiple functions of a part of speech are non-​prototypical: there is an expres-
sional function transformation (marked or extrinsic expressional function
transformation). For example, de (的)/​di (地) is added in order for a statement
or a reference to function as a modifier. To function as a predicate, a reference
should be transformed into a statement; to function as a subject or object, a
statement should be transformed into a reference; to function as a subject or
object, a modification should also be transformed into a reference.
106

106  Essence and expressional functions

There is a loose correspondence between the position of a Chinese syn-
tactic constituent and a part of speech, but at the combinatory positions of
notional and functional words, there are rather strict requirements for parts
of speech. For example, the positions followed by 不 (not) or 很 (very) permit
the entry of only predicate constituents. That is the reason why we have to use
specific distributions as the criteria for classifying parts of speech.

Notes
1 很 (very) represents an absolute degree adverb.
2 Because there are a great many grammatical functions, we discuss here only the
greatly universal grammatical functions. As mentioned before, nearly all of the
grammatical functions have no internal universality. For the convenience of the dis-
cussion, in the following, when we discuss the external exclusivity of a grammatical
function, the grammatical function of a certain part of speech means that at least
some of the words in the part of speech have a grammatical function.
3 A grammatical function has a distinction between general and concrete. For
example, “combined with other constituents” is more general than “functioning as
a syntactic constituent”. “Functioning as a syntactic constituent” is more general
than “functioning as a head word”. “Functioning as a head word” is more general
than “being modified by an adverbial”. “Being modified by an adverbial” is more
general than “being modified by 很 (very)”.
4 See Xiao Guozheng (1991) for the inner and outer distinctions between references
of predicate constituents at an object’s position.
5 Chinese adjectives such as 大 (large), 红 (red) and 重要 (important) are predicate
words and belong to VP syntactically, while English adjectives belong to AP (a
modifier) syntactically. Chinese distinctive words belong to AP syntactically and
are equivalent to English adjectives. Chinese adjectives that can directly function as
attributives can be regarded as adjectives and distinctive words concurrently.
6 Lu Jianming (1991) discussed in some detail the transfer reference of a modifier.
7 大 (big) and 小 (small) in 那商店卖的盆儿有大有小 (some basins on sale in that
store are big; others are small) are transfer references, but it seems that they should
be analyzed into “NP→AP + Pro”. Their constructions are the same as those for
the transfer reference of a distinctive word, but different from the construction “NP
→ Pro + VP” for the transfer reference of an adjective in classical Chinese.
8 The object of a transfer reference without adding suo (所) usually has conditions: the
predicate word takes a negative word or a modal verb 可 (can).
9 Shen Jiaxuan (1997) believed that the major function of a Chinese adjective is
attributive. In the opinion of the author of this book, it is predication, and an
adjective belongs to a predicate word; its direct function as an attributive is actually
because some of adjectives concurrently have the property of a distinctive word.
 107

5 
Criteria for classifying parts of speech

5.1  Conditions for criteria for classifying parts of speech

Chapter 4.6 presented the idea that the bases for classifying parts of speech
and the criteria are different. The two may be in agreement or not, but the cri-
teria must be able to reflect the bases for classification.
Any factor that can be used as the criteria for classifying parts of speech
must meet the following three conditions:

A. It can reflect the nature of a part of speech, namely the intrinsic expres-
sional function. Different parts of speech show different expressional
functions of a word. Therefore, only those factors that can reflect its
intrinsic expressional functions can be used as the criteria for classifying
parts of speech.
B. Observation. This means that the factor has an obvious extrinsic form or
is itself a certain extrinsic form, thereby being definitely comprehensible.
This is important because only through using observable things as the
criteria for classifying parts of speech can discrepant classifications be
avoided, thus making definite, reliable and operable classifications.
C. Comprehensiveness. This means that the factors used as the criteria for
classifying parts of speech are applicable to all or most of the words. Only
comprehensive factors can be used as the primary criteria for classifying
parts of speech, whereas incomprehensive factors can at most be used as
subsidiary criteria to supplement the primary criteria and can be used to
classify the parts of speech of some other words when the primary cri-
teria cannot be used to classify their parts of speech.

The previously mentioned classification criteria mainly include the

following three: (1) the morphology of a word; (2) its meaning; (3) its gram-
matical function. In the present study, we use the grammatical function as
classification criteria. Why? Our analysis is as follows:
108

108  Criteria for classifying parts of speech

5.2  A word’s morphology, meaning and grammatical function

5.2.1  Whether the morphology can be used as classification criteria

Morphology means a word form and its changes. Morphology displays part-​
of-​speech differences in two scenarios. First, the word form itself has system-
atic differences that reveal parts of speech. For example, Italian:

Verbs: -​are/​-​ire/​-​ere: benefiare, leggere, finire, amare

Nouns:  -​o (masculine)/​-​a (feminine)/​-​ e (neutral):  beneficio, leggenda,
fine, amore

Second, a word has its systematic changes that reveal differences in parts
of speech. On Latin Language, written by W. T. Varro, a Roman who lived
from 116 to 27 BC, classified words into four classes according to changes in
word form:

Noun: a word that has changes in case.

Verb: a word that has changes in tense.
Participle: a word that has changes in case and tense.
Functional word: a word that has no changes in case and tense.

Only when the difference or change in word form that shows differences
in parts of speech is systematic can it be sufficiently used to classify parts of
speech. Intrinsically speaking, a part of speech is the class of intrinsic expres-
sional functions, whereas a word form is not the nature of a part of speech
but only the most extrinsic manifestation of a word’s expressional function.
Therefore, we should not think that a language has no parts of speech just
because its words have no systematic differences or changes in word form.
But the word form of a formal language often has a rather strict corres-
pondence with the expressional function of a word. For example, in English,
a word that has changes in plural form can function as a subject or object,
and has a reference meaning. A word that has changes in tense and aspect
can function as a predicate and has a statement meaning. Therefore, we can
assume that the form of a word can indirectly reflect its expressional function
and meet the first condition of the criteria for classifying parts of speech.
The reason why we can use a word form to classify the parts of speech of
a language that is rich in form is that there is a rather strict correspondence
between its word form and the expressional function of its words. In other
words, the form of a word can be regarded as a symbol of its intrinsic expres-
sional function.
A word form, of course, is observable and also meets the second condition
for classifying parts of speech.
But Chinese lacks word form symbols and morphological changes in their
strict sense; its word form is incomprehensive. Only grammatical function can
 109

Criteria for classifying parts of speech 109

be used to classify the parts of speech of the vast majority of words that have
no morphological change. Thus, the Chinese word form can at most only be
used as the reference criteria for classifying parts of speech. For example,
words such as 轻松 (relaxed), 凉快 (nice and cool) and 暖和 (warm) can only
be classified into adjectives according to their grammatical functions, but they
have their ABAB reduplication forms. If we regard the ABAB reduplication
forms as a kind of morphological change of verbs, then we can use this as
the basis to regard 轻松 (relaxed), 凉快 (nice and cool) and 暖和 (warm) as
conversional words between adjective and verb. Nevertheless, as mentioned
in Chapter  2.2, strictly speaking, Chinese has no genuine morphological
reduplication but only word formation reduplication and syntactic reduplica-
tion. The ABAB reduplication form is a syntactic phenomenon and should be
looked upon as a grammatical function.

5.2.2  Can word meaning be used as the criteria for classifying parts of speech?
Ma Jianzhong (1898) and Wang Li (China’s Modern Grammar, 1943, China’s
Grammatical Theory, 1944) both used meaning as the criteria for classifying
parts of speech.
There are two types of meanings: one is lexical meaning, and the other is
grammatical meaning or category meaning. These two aspects must be taken
into account when discussing whether meaning can be used as criteria for
classifying parts of speech.
We begin with the above-​mentioned three conditions for classifying parts
of speech to examine whether the lexical meaning of a word can be used as
criteria for classifying its parts of speech. First, the lexical meaning cannot
reflect the expressional function of a word. For example, the following words
in their groups have the same or about the same lexical meanings, but their
expressional functions are quite different:

a) 不 (not) ~ 很 (very) ~ complement predicate adverbial

突然 (sudden) + + + + +
忽然 (unexpectedly) 一 一 一 一 +

b) 不 (not) ~ 很 (very) ~ complement predicate

白 (white) + + + +
白雪 (snow-​white) 一 一 + 一
白色 (white color) -​ -​ -​ -​

c) 没 (not yet) ~ 很 (very) ~

胜利 (victory) + 一
成功 (successful) + +
110

110  Criteria for classifying parts of speech

d) subject 在 (in) ~ 三天 (three days) ~ 春节 (spring festival) ~

以前 (ago) + + + +
过去 (past) + + 一 一

Second, lexical meaning is not observable; it can only be perceived but

cannot be definitely grasped. Therefore, although lexical meaning meets the
third condition for the criteria, it cannot be used as the criteria for classi-
fying parts of speech because it does not meet the first and second conditions.
Can we use the grammatical meaning (or category meaning) of a word to
classify its parts of speech? Wang Li (1943) took the category meaning as a
criterion for classifying parts of speech; he was against using a word form or
a grammatical function to classify them. “When we say that parts of speech
can be indicated in a dictionary, we mean that a word itself can be grouped
into a class, and that there is no need to wait for it to come into a sentence to
decide its parts of speech. If a word is classified into classes according to its
function in a sentence, then word properties are obtained” (p. 19). “Words can
be classified into two major classes: any word itself that can indicate a concept
is called a notional word; any word itself that cannot indicate a concept but is
the tool for constructing a language is called a functional word. The classifica-
tion of a notional word into parts of speech should be based on the type of its
concept; the classification of a functional word into its parts of speech should
be based on its function in a sentence”. The following are Wang Li’s criteria
for classifying some notional words into their parts of speech (pp. 21–​29):

Noun: all names of objects are called nouns.

Numeral: all words that indicate the numbers of objects are called numerals.
Adjective: all words that indicate the properties of objects are called adjectives.
Verb: all words that indicate actions or events are called action words or verbs.
Adverb: all words that can only indicate degree, scope, time, possibility, neg-
ation and so on but cannot separately indicate objects, truths or facts are
called adverbs.

In Wang Li’s opinion, because Chinese has no word forms, its classifica-
tion into parts of speech is easier than Western languages. Although Western
languages have word forms, they are incomplete, so the classification into
parts of speech still relies on a grammatical function, which, however, does
not completely correspond to a part of speech, ultimately relying on concepts
to classify parts of speech. For example, in French, “Je suis fort (I am strong)”
and “Je suis roi (I am a king)”, “fort” and “roi” have no morphological diffe-
rence, and we can only make conceptual distinctions that “fort” is an adjective
and that “roi” is a noun. But Chinese has no part-​of-​speech label at all, and
this happens to let us classify parts of speech purely according to concept, not
being restricted by any form. As a result, it is easier to classify Chinese parts
of speech than those of Western languages (p. 28).
 111

Criteria for classifying parts of speech 111

The category meaning of a word has an approximate correspondence with
its expressional function. For example, we say that objects correspond to
reference, and that actions correspond to statement, but category meaning
is not directly observable and cannot directly and definitely be grasped.
Furthermore, it is extremely complicated: What do we mean by action, object
or property? Where are their boundaries? How many types of category
meaning in all are there? These can create inexhaustible controversies that
may never be solved (Zhu Dexi, 1985a). Therefore, the category meaning does
not meet the second condition for the criteria of classifying parts of speech.
The statement that it is easy to classify Chinese parts of speech as Wang Li
put it is a false appearance. The category meaning also does not completely
correspond to an expressional function. For example, an action can corres-
pond to a reference, such as “walk” in “to have a walk”. Attribute can corres-
pond to either a modification or a statement. Therefore, the category meaning
does not meet the first condition for the criteria of classifying parts of speech.
In our opinion, the grammatical meaning of a word is not the experiential
meaning such as an object, action or properties but the expressional function
based on the relations among the internal constituents of a language, such as
a reference, statement and modification. As mentioned before, because the
expressional function is not directly observable, it also cannot be directly used
as the criteria for classifying parts of speech.
A good many scholars recognize that the main basis for classifying parts of
speech is the grammatical function of a word; however, they also emphasize
that grammatical meaning is an important reference criterion. Some books say
that equal attention should be paid to both function and meaning (Zhu Dexi,
1985a). But the reality is often that when the uses of a grammatical function as
the criterion for classifying parts of speech are smooth, this is done. Once this
is troublesome, or the parts of speech classified according to a grammatical
function do not agree with the human intuition of the meaning, the grammat-
ical function is just abandoned, and instead a grammatical meaning is used.
Sometimes a grammatical function and sometimes an indefinite grammat-
ical meaning are used as the criteria, thus causing chaos for the classification.
For example, in the past, a good many scholars maintained that words that
express properties are adjectives, and then they classified into adjectives the
words such as 高等 (high-​class) and 大型 (large-​scale). Indeed, the differences
in the grammatical function between 高等 (high-​class) and 大型 (large-​scale)
and typical adjectives such as 高级 (advanced) and 大 (large) lie in the fact
that 高级 (advanced) and 大 (large) can be modified by 不 (not) or 很 (very),
and can function as predicates, while 高等 (high-​class) and 大型 (large-​scale)
do not have these functions. But because they are deemed to indicate the same
grammatical meaning, scholars would rather call words such as 高等 (high-​
class) and 大型 (large-​scale) non-​predicate adjectives than eliminate them from
adjectives, consequently making adjectives become a mélange where the gram-
matical functions of their members have large discrepancies. In the last ana-
lysis, only the use of a grammatical function can help classify parts of speech.
112

112  Criteria for classifying parts of speech

5.2.3  Classifying parts of speech according to a grammatical function

The following three methods are used:

I. Classifying parts of speech according to a syntactic constituent. With the

one-​to-​one correspondence between a syntactic constituent and a part of
speech, we use the implemented functions as the criteria for classifying
parts of speech. This is inconsistent with Chinese facts, easily leading to
the conclusion that words have no definite parts of speech or even that
Chinese has no parts of speech at all, as Li Jinxi (1994) held.
II. Classifying parts of speech according to distribution in its narrow sense,
namely using the combinatory environment of a word in relation to
another word or phrase as the criteria for classifying parts of speech. For
example:
Numeral~[measure word: 个 (piece), 支 (count), 斤 (jin)]
Numeral and measure word~[noun:  苹果 (apple), 山 (mountain), 纸
(paper)]
The so-​called identification word or test slot and the generalized word
form (Fang Guangdao, 1939) use this method. English grammar com-
monly uses this method to classify parts of speech. The following are
the criteria that J. M. Y Simpson (1979) used to classify English parts of
speech:
1. ~ hat is on the table.[Det.: a, the, that …]
2. (Det.) ~ is/​are good. [Noun: man, oats, John…]
3. (Det.) Noun ~ (Det.) Noun.[Verb: is, sings, smokes…]
4. (Det.) ~ noun.[adj.: big, green, brackish…]
5. Det. noun is ~ adj.[adv.: too, very, badly…]
6. ~ Verb (Det.) noun.[pron.: I, you, he…]
7. Noun Verb Pron. ~ (υ ≠adv.)[particle: ~ (up, over, out…)]
8. Det. noun verb ~ Det. noun (~ ≠Adv.) [prep.: into, under, up…]
The problem of this method is that only when the part of speech that
enters into a certain position is the sole one can distribution effectively
classify parts of speech. This method is problematic even in English. For
example, it is hard to distinguish between noun and pronoun, and also
between noun and adjective.
In Chinese, because one grammatical position is commonly occu-
pied by various parts of speech, a word combined with another word
in different parts of speech often has more than one grammatical rela-
tion. Therefore, the use of this method for classifying parts of speech
must attach some conditions, but too many conditions make the criteria
complicated and not easy to control.
III. Classifying parts of speech according to syntactic constituent and distri-
bution in its narrow sense. This has the following two aspects: (1) the cap-
ability of a word functioning as syntactic constituents; (2) the capability
of one word to be combined with another.
 113

Criteria for classifying parts of speech 113

The two capabilities combined make up for the shortcomings of the classi-
fication with distribution in its narrow sense, which is a rather effective way to
classify Chinese parts of speech.
This book uses the third method. The first method has the problem that
it is difficult to determine syntactic constituents, and hence only the second
method is used if possible. Only when the second method cannot effectively
classify parts of speech, or when the criteria are too complicated, is the first
method used to classify them. Both the first and the second methods use
grammatical function and distribution. In the following, we shall explain the
unity of the two.
Why can grammatical function be used to classify parts of speech?
Previously, we mentioned that a grammatical position has selectional restric-
tion on parts of speech. The criteria for the selectional restriction to select
words are the words’ own grammatical attributes. In other words, because the
grammatical attributes are different, the grammatical positions a word can
occupy are also different. The selectional restriction of grammatical position
on words, namely distribution, can thus be used to classify parts of speech.
A word is capable of occupying grammatical positions (namely grammat-
ical function reflects its intrinsic expressional function). The parts of speech
classified according to the grammatical functions of a word are indeed gram-
matical, thereby meeting the first condition of the classification criteria.
Grammatical functions also meet the second condition of the classification
criteria. It has two aspects: (1) the capability of functioning as a syntactic con-
stituent such as a subject, object, predicate or attributive; (2) the capability of
being combined with other words or phrases, for example, being modified by
很 (very) or numerals and measure words, taking locatives and so on. The two
aspects are observable.
Grammatical function also meets the third condition of the classification
criteria. Every word has its own grammatical functions, which are therefore
applicable to the classification of every word’s parts of speech as a criterion.
Therefore, grammatical function meets all the conditions and thus can be
used as criteria for classifying parts of speech.

5.2.4  Brief summary

Theoretically speaking, a word’s grammatical function, grammatical meaning
or intrinsic expressional function all can be used as criteria for classifying
parts of speech.
The form of a word is highly observable and can reflect its expressional
function, but the Chinese word form is not comprehensive and can only be
used as supplementary criteria.
A word has its lexical meaning and category meaning. The lexical meaning
is comprehensive but does not reflect the word’s expressional functions and
is not observable, therefore being unable to be used as classification criteria.
The category meaning is comprehensive but not observable and cannot fully
114

114  Criteria for classifying parts of speech

reflect a word’s intrinsic expressional function, while the expressional function
itself also is not directly observable, thus being unable to be used as classifi-
cation criteria.
Grammatical function can reflect a word’s expressional function and is
observable and comprehensive, thus being able to be used as classification
criteria.

5.3  What is grammatical function?

The classification of parts of speech according to distribution can be traced
back to the distribution analysis proposed by US descriptive linguistics.
Bloomfield (1926) said,

29. [Definition]Every methodical unit in a construction is a position.

30. [Assumption]Every position in a construction can only be filled in by
a certain form.
32. [Def.]The position a form occupies is its function.
33. [Def.]All the forms that have the same functions compose a form class.
37. [Def.]The form class of a word is the part of speech.
38. [Def.] The largest word class in a language is its part of speech.

Harris (1946) said, every class of morphemes has its special sentence
positions and can be filled by any members of the class and only by those
members. For example:

N: Appearing before the plural form -​s or its variant or after “the” or adjec-
tive: hotel, butter, two.
V: Appearing before the past tense “-​ed” or its variant; before “-​ing”, after an
N plus “should, will, might”: go, take, do.
Adjective: Appearing between “the” and N, but never appearing before “–​
s”: young, pretty, happy.
Adverb: appearing between “the” and adjective, but not between “the” and a
noun: rather, very, now, not.

Fries (1952) said that in a single free mode of discourse, English words that
occupy the same positions must belong to the same part of speech.
What is distribution? Harris said that “the distribution of an element is the
total of all environments in which it occurs, i.e. the sum of all the (different)
positions (or occurrences) of an element relative to the occurrence of other
elements” (Harris, 1951: 15–​16).
What is the environment or position? Harris said that the environment or
position of an element in a discourse is made up of its adjacent elements.
The so-​called “adjacency” refers to the position before or after the element
or the position in which nonlinear elements such as intonation, stress and
 115

Criteria for classifying parts of speech 115

others appear simultaneously. (Harris, 1951: 15) If intonation is considered, it
includes a simultaneous statement intonation.
The definitions of Bloomfield and Harris show that the position concept
of US structural linguistics is based on a superficial and sequential position
relationship among elements:  before, after or simultaneously, not on the
grammatical relationship between hierarchy and element. Yet this concept of
distribution is too superficial. The root cause for the use of grammatical pos-
ition to classify the parts of speech of an element is its selectional restriction
on words, which is not decided by the superficial and sequential positional
relationship, but by hierarchies and grammatical relationships. For example:

(1) 给我书／给我哥 (Give me a book/​give to my brother)

The consideration of only a superficial and sequential positional relationship

may lead to the view that 书 (book) and 哥 (brother) have the same grammat-
ical positions. But actually the hierarchical structures of the two constituents
are different, thereby occupying different grammatical positions. The position
that the whole phrase 我哥 (my brother) occupies is equivalent to that occu-
pied by 我 (me) in 给我书 (Give me a book).

(2) 这三个苹果／这三个好吃 (The three apples/​the three are delicious)

Here, no consideration of a grammatical relationship may lead to the view

that 苹果 (apples) and 好吃 (delicious) appear in the same grammatical pos-
ition, but because their grammatical relationship is different, the two actually
occupy different grammatical positions.

(3) 学习文件 (study documents)

In terms of only a sequential positional relationship, 学习 (study) in the

phrase represents only one distribution, but actually it represents two types
of distribution:1 one type is equivalent to 学习 (study) in the phrase 学习汉语
(studying Chinese); another is equivalent to 学习 (study) in the phrase
学习时间 (study time).
Therefore, in our opinion, the position that is used to define distribution
contains hierarchies and grammatical relations, and is thus called the gram-
matical position. Two factors stipulate the distribution of constituents:

1. The grammatical relations among immediate constituents and the gram-

matical roles of the immediate constituents thus stipulated.
The grammatical relations include both quite abstract relations such
as subject-​predicate, predicate-​object, attribute-​headword and predicate-​
complement and quite specific relations, for example:  许多书 (many
books), 新书 (new book), 一个 (one). In terms of their general relations,
the distribution of 许多 (many), 新 (new) and 一 (one) is the same in these
116

116  Criteria for classifying parts of speech

phrases, but in terms of their specific relations, they are different. This is
also similar to 桌子腿 (desk leg) and 桌子上 (on the desk).
2. The larger environment in which the entire construction made up of two
immediate constituents can exist.
For example, both 大型 (large-​scale) and 大量 (massive) can appear in
“~ noun”, but “大型 (large-​scale) + noun” can also appear in “numeral
and measure word ~”, while “大量 (massive) + noun” cannot. Therefore,
the specific grammatical positions in which 大型 (large-​scale) and 大量
(massive) appear are different.

The above analysis produces the following three important definitions:

Definition 1:  In a syntactic structure, the position in which the imme-

diate constituents that have certain grammatical relations exist is a
grammatical position, which contains information on hierarchy and
grammatical relation.
Definition 2:  The grammatical position occupied by a word is its
distribution.
Definition 3: The capability of a word to occupy a certain grammatical
position is the word’s grammatical function.

A word’s distribution or function is often expressed as the position (or

test slot) whose environment is a syntactic constituent, identification word
or part of speech, for example: ~ object, 很 (very) ~, numeral ~. It is some-
times expressed as a syntactic constituent but is equivalent to the position
that is expressed as the environment consisting of syntactic constituents. For
example, a predicate is equivalent to a “subject ~”.
The reason why we do not mention the sequential relational positions 之前
(before) and 之后 (after) is that the essence of a grammatical position is a
grammatical relational position, namely the position decided by grammatical
relations, while the sequential position does not matter. For example, because
Latin has case markers, the sequential positions of a subject, object and predi-
cate verb are fairly flexible:

(4) Puer amat puellam.

A teenager loves a teenager girl.
Puer puellam amamat.
A teenager a teenager girl loves
Amat puer puellam
Love a teenager a teenager girl.
Amat puellam puer.
Love a teenager girl a teenager.
Puellam puer amat.
A teenager girl a teenager loves.
Puellam amat puer.
A teenager girl loves a teenager.
 117

Criteria for classifying parts of speech 117

Where puer is the subject case (the original form: puer); puellam is the object
case (the original form: puella); amat is the third person singular (the original
form:  amere). There are no changes in grammatical relations in the above
different sentences, and so they should be deemed as occupying the same
grammatical positions respectively.
Russian is similar to this (Shi Anshi & Zhan Renfeng, 1988: 110–​111):

(5) ９ａｕｏ６ｍｏｍａ. I love my mother.

５１ＭａｗＭＯＵ　Ｋｏ. I mother love.
Ｊｈｏ６ｓｍＳＩｍａ. Love my mother.
ＪＬｏ６ａｉｉｏ　ｍａＴｈ　５１. Love mother I.
Ｍａｗ９ｍｏ６ｍ. Mother I love.
Ｍａｗｍｏ６ｍｏ５　Ｉ. Mother loves me.

Because of the case markers:  ＭａＴｂ (object case, ＪＨＯＪＪＯ Ｍ

(the third person singular), all the different sequences mean the same:  “I
love mother”. ｑｊｕｏ６．ｎｔｏ. There are no changes in grammatical
relations in the above different sentences, and so they should be deemed as
occupying the same grammatical positions respectively.
Thus, we can understand why some English words appear before or after a
sequential position but are classified into adverbs:

(6) a. He always loses his pencils. b. He loved her deeply.

Although “always” and “deeply” occupy different sequential positions,

their grammatical relational positions are the same: both form the adverbial-​
headword relation with their immediate constituents and occupy the positions
in which they function as modifiers.
Then, why do we often express a word’s distribution as a sequential pos-
ition? This is because under usual circumstances, a sequential position reflects
the only grammatical relation (relation and role). For example, 很 (very) ~
usually reflects the adverbial-​headword relation. The positions stipulated by
an identification word do not have grammatical relations superficially but
actually contain them. Simply there is no mention of them because under
usual circumstances grammatical relations are unique. For example, the sen-
tence “Det. ~ is/​are good” always has the subject-​predicate relation. Zhu Dexi
(1982a) said that “syntax can be discussed to a certain degree when the part
of speech and hierarchy in Indo-​European languages are grasped because
they can control structural relations to a certain degree”. In his opinion, that
is one of the reasons why American structuralism does not stress grammat-
ical relations. Therefore, syntactic constituents and grammatical positions
stipulated by the environment consisting of an identification word and a part
of speech can be unified by the capability of a word to occupy grammatical
positions.
118

118  Criteria for classifying parts of speech

Two points need to be clarified:

A. Previously we said that distribution is determined by two aspects. But

not all distributional differences reflect differences in parts of speech.
Some distributional differences reflect differences in parts of speech,
for example, 很突然/​*很忽然 (very sudden/​*very unexpectedness), but
others do not, for example, 九个人/​*九个人们 (nine persons/​*nine folk).
Some distributional differences only reflect the differences in subclasses,
for example, 一个+名/​*一下+名 (one + noun/​*once + noun). What rela-
tion on earth do distribution and parts of speech have? How should dis-
tribution be used to classify parts of speech? We shall discuss these in
Chapter 6.
B. Although distribution in its narrow sense and the capability to function
as syntactic constituents can be unified, the two are somewhat different.
In the following we shall discuss how different they are.

5.4  Generalization level of a grammatical function

5.4.1  Specific function and general function

Previously, we mentioned the factors that decide grammatical positions and
excluded specific semantic differences. For example:

(7) a. 一 ~ 苹果(个) (a piece of apple)     b. 一 ~ 树(棵) (a stem of tree)

This kind of positional difference is purely caused by semantic factors, not

by differences in grammatical relations. Therefore, the difference in grammat-
ical positions does not tell the grammatical distributional difference of 一个
(a piece of) and 一颗 (a stem of). In other words, a grammatical position is
actually a rather abstract position in which abstract words are combined. The
degree of abstraction may be high or low. Comparatively speaking, the pos-
ition that takes specific words or a class of words as its environment is a rather
specific position, for example:

(8) numeral + measure word      measure word + noun

九 (nine)    个 (piece)      一切 (all)    人 (persons)
adjective + noun        demonstrative  + noun
坏 (bad)     人 (person)     这 (this)     人 (person)

The reason why we say that these four examples represent four pairs of
grammatical position is not that the parts of speech or specific words that
work as environmental constituents are different. The root cause is still that
their grammatical relations are different and rather specific:  the relations
between number and measurement unit, between quantity and object, between
attributive and object and between demonstrative word and its object. When
 119

Criteria for classifying parts of speech 119

these constituents co-​occur, they more obviously occupy different grammat-
ical positions.

(9) 这九个坏人 (these nine bad persons) 这一切坏人 (all these bad persons)
*这个九坏人 (*this piece nine bad persons) *这坏一切人 (*this bad all persons)
*人坏九这个 (*persons bad nine this piece) *人坏一切该 (*persons bad all these)

If these attributives have strict substitution relations, then they cannot co-​
occur in the same noun phrase. But actually they can co-​occur and hence
are not real substitutions. Different attributives do not necessarily occupy the
same grammatical position.
Because it is not easy or very trivial to express the grammatical relations
in a specific grammatical position, we just directly use the environment
consisting of specific words or parts of speech instead of using concrete syn-
tactic relations. Clearly, this actually contains specific grammatical relations.
If we generalize these grammatical positions and abandon the differences
in specific grammatical relations, only considering the more abstract gram-
matical relations such as subject-​predicate, predicate-​object and modifier-​
modified, then these grammatical positions can be generalized as two more
abstract grammatical positions such as “attributive and headword”.
Thus, we call the grammatical position based on concrete syntactic
relations the specific grammatical position, and the grammatical position
based on abstract syntactic relations the abstract grammatical position.
The capability to occupy concrete syntactic positions is called specific dis-
tribution, and the capability to occupy abstract syntactic positions is called
generalized distribution. This is the difference between the distribution in its
narrow sense and the capability to occupy syntactic constituents:  different
degrees of abstraction.
Now, we can sum up the conclusion that a grammatical function includes
two respects:

(1) The capability to combine one word or phrase with another (specific
distribution).
(2) The capability to function as syntactic constituents (generalized

distribution).

5.4.2  Whether we can classify parts of speech only on the basis of

generalized or specific distribution
Can we classify parts of speech only on the basis of generalized distribu-
tion? This question is raised by Li Jinxi and Lu Zhiwei. Although Li Jinxi
(1924) defined parts of speech from the perspective of meaning, he identi-
fied this term according to sentence constituents. “There is no way to tell
differences among Chinese parts of speech from words themselves or their
forms; which part of speech a word belongs to is set only by its position
120

120  Criteria for classifying parts of speech

or function in a sentence”. “Parts of speech are mostly differentiated from
sentence constituents”. He classified syntactic constituents into subjects,
predicates, objects, complements, attributives (supplements of an adjective)
and adverbials (supplements of an adverb) and established correspondence
between parts of speech of notional words and their syntactic constituents:

Subject, object –​noun, pronoun

Predicate –​  verb
Attributive –​ adjective
Adverbial –​  adverb

But because Chinese parts of speech have no simple correspondence with

syntactic constituents, Li Jinxi proposed that “the part of speech of a word
relies on sentence”: “The nine Chinese parts of speech vary as the position or
function of a word varies in a sentence, and there is no strict difference”. For
example:

铁桥 (iron bridge): 铁 (iron) changes from a noun into an adjective

律师的辩护 (a lawyer’s defense):  辩护 (defense) changes from a verb
into a noun
飞鸟 (flying bird): 飞 (flying) changes from a verb into an adjective

As a result, words have no definite parts of speech. Then he reached the

conclusion that “no part of speech without a sentence”. This actually means
that there is no part of speech because a part of speech is the class of words
themselves, implying that words have no parts of speech of their own.
Lu Zhiwei (1938) proposed the following two types of structural
relationship:

(1) Supplementation: 红花 (red flower), 大海 (high sea), 好人 (good person)

(supplement word + person or thing supplemented)
(2) Approximation:  吃饭 (have meal), 在家 (stay home), 指着他 (point to
him) (approximator + person or thing approximated)

The two types of structural relationship stipulate the three kinds of basic
parts of speech:

Noun: occupying the positions of the person or thing supplemented or

approximated
Verb: occupying the position of an approximator
Adjective: occupying the position of a supplementer

The distribution proposed by Lu Zhiwei for classifying parts of speech is

roughly equivalent to such generalizations as subjects, objects, predicates and
attributives. Therefore, it is not substantially different from the classification
 121

Criteria for classifying parts of speech 121

method by Li Jinxi. If the Lu method is strictly implemented, then words still
have no definite parts of speech, but it is not strictly implemented:

布鞋 (cloth shoe), 狗 (dog), 尾巴 (tail): It seems that 布 (cloth) becomes

an adjective, but with the phrase 织布 (weave cloth), 布 (cloth) should
be determined as a noun, which is used at the supplementation pos-
ition, but still is a noun.
风大 (wind big), 墨臭 (ink odor):  It seems that 大 (big) and 臭 (odor)
become verbs. But with the phrases 大风 (big wind), 臭墨 (odorous
ink), 大 (big) and 臭 (odor) are still adjectives.
死鱼 (dead fish), 断桥 (broken bridge): 死 (dead) and 断 (broken) are still
verbs. (The examples cited by Lu Zhiweu are 飞船 (flying ship), 包车
(chartered vehicle), which are multisyllable words; if they could be
broken down, they would still be “verb + noun”. These two examples
are imitated by the author according to Lu’s statement.)

In Lu Zhiwei’s mind, a part of speech does not change; therefore, once a word
is identified as an adjective, it will be an adjective at any position. The initial iden-
tification is based on the four positions of two types of relations. But if it can
occupy two positions, which position should be used? For example, we can well
say that because there is the phrase 风大 (wind big), 大 (big) in the phrase 大风
(big wind) is also a verb. Lu does not explain this, but as we can see, he reverts to
determining parts of speech according to meaning. In other words, although dis-
tribution is superficially used as the criteria, the classification is carried out not
strictly according to the uniform distribution criteria but as needed according to
the meaning of a word. The distribution criterion becomes only nominal.
Hence, Chinese parts of speech should not be classified according to
generalized distribution.
Can we classify parts of speech only according to specific distribution?
Hu Mingyang (1996a) thought that “it seems that the simple classification of
parts of speech with identification words or identification formats is objective
and scientific, but actually it tends to fall into a circular argument because the
initial criteria are subjectively selected and not proved”. But in fact the classi-
fication of parts of speech with identification words is reasonable, merely not
proved with a specific method. In the next chapter, we shall mention that the
classification of parts of speech according to specific grammatical functions
(including the specific functions stipulated by identification words) is demon-
strable. That is to say, the great compatibility of some grammatical functions
reflects the properties of the same part of speech and their equivalence. A series
of equivalent functions represents the distinctive functions of a part of speech,
and the criteria for classifying parts of speech (including the functions stipulated
by identification words) are selected from the equivalent functions.
In contrast with the views of Hu Mingyang (1996a), Chen Baoya (1999)
proposed that, due to the complicated correspondence between a part of
122

122  Criteria for classifying parts of speech

speech and a syntactic constituent, the classification of parts of speech with
syntactic constituents as criteria cannot resolve the criteria’s external exclu-
sivity. It is difficult to decide grammatical relations. Therefore, parts of speech
should be classified only according to specific distribution (identification
words). This method brings about the following problems:

1. It still cannot completely resolve the criteria’s external exclusivity; there is

scarcely any distribution in its narrow sense for only one class of words.
For example:
Verbs and adjectives have the grammatical function that they can be
placed after 不 (not).
Verbs and adjectives have the grammatical function that they can be
placed after 很 (very) or 最不 (least).
Verbs, adjectives, nouns and time words can be followed by ~了 (le).
Verbs and adjectives can be followed by 着 (zhao).
Verbs and adjectives can be followed by 过 (guo).
Nouns, verbs, adjectives and time words can be modified by numeral
and measure words.
Measure words and nouns can be modified by numerals.
Locatives, place words and nouns can function as the objects of 在
(in), 到 (to), 往 (toward).
2. The criteria’s internal universality is poorer (see Chapter 2.3.4.1).
Consequently, plentiful disjunctive criteria have to be employed,
making them too complicated and not operational. For example, it is dif-
ficult to use specific distribution to classify words into distinctive words,
adverbs and nouns, not to mention verbs and adjectives.
3. Parts of speech are not systematic. We shall discuss the relationship
between systematicity and the criteria for classifying parts of speech in
Chapter 1.2 in Volume 2.
Therefore, only the combination of specific distribution with
generalized distribution can effectively classify Chinese parts of speech.
But sometimes it is really difficult for generalized distribution to decide
grammatical relations; therefore, our method is to use specific distribu-
tion as much as possible, and only when specific distribution does not
work or is too complicated do we use generalized distribution.
4. In fact, an identification word and syntactic constituent jointly indicate
the grammatical position occupation capability. The identification word
contains grammatical relations but clearly does not mention them. For
example, the “numeral and measure word ~” is used to identify a noun,
but 好 (good), 坏了 (bad) and other predicate words may also appear
in this position and must be eliminated according to different grammat-
ical relations. In other words, an identification word only has a specific
function, and its difference from a syntactic constituent lies only in the
generalization level.
 123

Criteria for classifying parts of speech 123

5.5  Why can grammatical function be used to classify

parts of speech?
Every position in a construction can only be filled in by a certain form
(Bloomfield, 1926). Words are combined not randomly but in a certain order.
This orderliness indicates that a syntactic position has selectional restrictions
on words, and different syntactic positions allow different words to enter.
There must be bases for a syntagmatic position’s selectional restriction
on words. Unless the syntagmatic position has no selectional restriction on
words, then any words can enter into the position optionally. There are bases
for a grammatical position’s selectional restriction on words. For example,
the word that may appear in “不 (not) ~” indicates a statement; the word
that can appear in “< quantity > ~” usually is a constituent with a reference
meaning; the word that may appear in “在 (in) ~” is usually a locative word;
the word that may appear in “numeral ~” is usually a measurement unit. The
main basis of the grammatical position’s selectional restriction on words is
the grammatical meaning of a word. Therefore, the parts of speech classified
according to distribution, in essence, have their grammatical meanings. The
type of grammatical meaning of a word is its part of speech. We can use the
grammatical position’s selectional restriction on a word (namely the parts of
speech reflected by a word’s distribution) to classify its parts of speech.
The classification of parts of speech is essentially to infer the part of speech
of a word according to its distributional characteristics. The reason why some
words may appear in the same grammatical position is that they have the
same grammatical meaning; the reason why some words cannot appear in the
same grammatical position may be that they do not have the same grammat-
ical meaning. In other words, the grammatical meaning of a word is the main
intrinsic cause for constraining its distribution and basically determines its
grammatical distribution. This is also the reason why although distribution is
not the nature of a part of speech, words that belong to the same part of speech
have a roughly similar distribution. Like a word form, distribution is only the
extrinsic exhibition of the grammatical meaning of a word. That is to say, there
is a “reflection-​exhibition” relationship between the distribution of a word and
its grammatical meaning: distribution reflects grammatical meaning, which in
turn exhibits distribution. Therefore, although we think that the intrinsic basis
of difference in parts of speech is the expressional function of a word or its
grammatical meaning, and that a part of speech is not essentially of a distribu-
tional type, a word’s distribution can be used as formal criteria for classifying
its parts of speech because it reflects grammatical meaning.

5.6  How effective for classifying parts of speech is distribution?

Although a word’s distribution can reflect its parts of speech, the relationship
between distribution and a part of speech is complicated and displays the
following:
124

124  Criteria for classifying parts of speech

I. Only the part of speech of a word decides its distribution; its lexical meaning,
pragmatic factors, word formation method, rhythmical characteristics and
so on may also influence its distribution. This is illustrated as follows:
1. The lexical meaning of a word influences distribution. For example,
认识 (recognize) does not take 过 (guo) because the process of rec-
ognition does not end; 完成 (finish) and 毕业 (graduate) cannot take
着 (zhuo) because their actions do not continue (see Guo Rui, 1993);
切记 (caution) does not take 了 (le), 着 (zhuo) and 过 (guo) because
its lexical meaning is imperative, while the predicate verb in an
imperative sentence in general cannot take tense or aspect elements.
2. Pragmatic factors influence a word’s distribution. For example,

耐烦 (patience), 像话 (proper), 好意思 (have the cheek), 要脸 (have
a sense of shame) and 起眼 (attraction) cannot be modified by 很
(very). Words like 耐烦 (patience) indicate base quantity, namely
the minimal quantity in a sense, which only appears in a negative
sentence (see Shi Yuzhi, 1992) but cannot appear after the affirma-
tive 很 (very). They can be modified again by 很 (very) only after
being negated by 不 (not), for example, 很不耐烦 (very impatient),
很不像话 (very improper). This is decided by the pragmatic principle
of sufficient-​quantity expression. Namely human language usually
requires that a quantity should be sufficiently expressed. But words
like 耐烦 (patience) express a minimal quantity and a lower limit in a
certain sense; its use as affirmative expression cannot meet this prin-
ciple. The negation of a lower limit is the negation of a whole sense
domain, meeting the sufficient-​ quantity principle. Therefore, the
minimal quantity is commonly used to express a complete negation.
3. Word formation factors influence a word’s distribution. A  good

many compound words in modern Chinese are formed by combining
phrases; syntactic rules that restrict phrases still act on the word-​form
morphemes to a certain degree, thus influencing the function of a word.

The first prominent example in this respect is that when the first root of a verb
or adjective is 不 (not), 无 (without) or 有 (have), the verb or adjective cannot
be modified by 不 (not), for instance:

(10) 不顾 (disregard), 不容 (not allow), 不如 (inferior, verb); 不安

(uneasy), 不幸 (unfortunate), 不利 (disadvantageous, adjective)

At a syntactic level, verbs such as 有 (have), 无 (without) and 没有 (have no)

cannot be modified by 不 (not) and 没有 (have not); this kind of restriction
also remains at the morphological level.2

(11) 有喜 (expect), 有助 (conduce), 有请 (please, verb); 有趣

(interesting), 有名 (famous), 有害 (harmful, adjective)
(12) 无关 (not matter), 无意 (not intend), 无视 (disregard, verb); 无情
(ruthless), 无聊 (dull), 无知 (ignorant, adjective)
 125

Criteria for classifying parts of speech 125

Table 5.1 Word formation methods and the capability of verbs to take objects
(double-​syllable  verbs)

Condition VO case Non-​VO case

taking real object 194 4709

proportion 7.2% 75.4%
not taking real object 2495 1534
proportion 92.8% 24.6%
total 2689 6243
Z1/​prominent level –​92.331 ++

Note: 1. Z refers to the standard score of normal distribution, see Chapter 2.2 in Volume 2 for
details.

The second example is that a verb in the VO case usually cannot take
a real object. At the syntactic level, except for verbs that can take double
objects such as 给 (give), 送 (send) and 借 (lend), an ordinary verb that has
taken a real object cannot take another real object. This kind of restric-
tion still basically remains at the morphological level. The following verbs
in their VO case cannot take real objects: 办公 (work), 闭幕 (close), 毕业
(graduate), 打仗 (fight), 发言 (speak), 见面 (meet), 就业 (employ), 失效
(fail), 破产 (bankrupt), 缺席 (absent), 上班 (go to work), 照相 (take photo).
But this is just a tendency. Some verbs in their VO case can take real objects,
for example, 进口 (import), 出土 (unearth), 担心 (concern) and 加工 (pro-
cess). But statistical analysis shows that there is a significant difference in
the capability of verbs in their VO and non-​VO cases to take real objects
(see Table 5.1).

(5) Rhythmical factors influence a word’s distribution. This is a tendency;

(6) for example, verbs that function as attributives are usually polysyllabic
ones (Table 5.2); the capability of a monosyllable adjective to function
as a complement is much greater than that of a double-​ syllabic
adjective.
II. Some grammatical positions reflect the same properties of a part of
speech, and the distributional difference thus caused cannot reflect the
difference in its properties. For example, “很 (very) ~” and “极 (extremely)
~” have the same requirements for words to enter into grammatical
positions, but the difference in the two functions of the two words does
not reflect different properties of their parts of speech.
III. It is possible that some grammatical positions allow several parts of speech
to enter. For example, a noun, a verb, an adjective or a distinctive word can
appear at the subject position, for example: 去是应该的 (Going is obliga-
tory), 骄傲使人落后 (Being proud makes one lag behind), 急性好治 (The
acute are easy to treat). Either a verb, an adjective, a state word or a noun
can appear at the predicate position, for example:  今天阴天 (Today is
overcast).
126

126  Criteria for classifying parts of speech

Table 5.2 The number of syllables and the capability of a verb to function as an
attributive

Condition Monosyllable Double-​syllable

as attributive 9 3147
proportion 1.0% 35.2%
not as attributive 872 5785
proportion 99.0% 64.8%
total 881 8932
Z1/​prominent level –​56.224 ++

Note: 1. See Chapter 6.

Therefore, the difference in a part of speech is not simply manifested in the

difference in distribution. “All the forms that have the same function compose
a form class”. The viewpoint of Bloomfield (1926) is not valid. It is impossible
to classify parts of speech purely according to the similarity with a word’s dis-
tribution (see Chapter 4).
In other words, the properties of a part of speech cannot be inferred
entirely according to word distribution. Not all differences in grammatical
function reflect differences in the properties of a part of speech. Thus, it is
not difficult to understand why we shall give so many disjunctive classifica-
tion criteria in the coming chapters. Even so, the classification of parts of
speech entirely according to grammatical function may still be problematic.
For example, distinctive words like 男 (male), 女 (female), 急性 (acute) and
慢性 (chronic) may function as subjects or objects and be classified into nouns
by altogether using grammatical functions. For another example, the use of
our criteria may classify most adjectives and verbs, but it is still hard to clas-
sify a few other adjectives and verbs. For example, 温 (warm) and 紫 (purple)
may function as predicates and be modified by adverbs of an absolute degree
but cannot take real objects, are not distinctive words and thus should be
classified into adjectives. If 有点 (somewhat) is regarded as an absolute-​degree
adverb like 很 (very), then 温 (warm) and 紫 (purple) may be classified into
adjectives, but intransitive verbs like 感冒 (catch cold), 咳嗽 (cough), 下雨
(rain), 哆嗦 (shiver) and 发慌 (panic) can also be modified by 有点 (some-
what); the use of this criterion may classify these verbs into adjectives. If 有点
(somewhat) is excluded from absolute-​degree adverbs, then 温 (warm) and 紫
(purple) can only be classified into verbs.
In addition, some predicate words can only function as predicates, and it is
difficult to determine whether they are verbs or state words, for example: 奇缺
(rare), 洞开 (bright), 参半 (half), 无双 (unrivaled), 交加 (occur simultan-
eously), 斑驳 (mottled), 依然 (still, classical Chinese).
Because there are limitations to the use of grammatical function as the cri-
terion, we need other approaches to overcome this difficulty. First, we need to
limit grammatical functions, namely excluding special uses. Second, we need
 127

Criteria for classifying parts of speech 127

to recognize exceptions. That is to say, the criteria of grammatical function
cannot be used to classify all words. Before effective criteria are identified, 温
(warm) and 紫 (purple) can only be regarded as exceptions.
The limitations of distribution criteria are mainly caused by three
factors: (1) there is no one-​to-​one correspondence between a part of speech
and distribution; (2) the temporary transformation of expressional functions;
for example, the functioning of a distinctive word as subject or object is the
result of reference (transfer reference); (3) the difference in distribution caused
by factors other than grammatical factors. The fundamental reason why dis-
tribution criteria have limitations is that parts of speech, in essence, are not of
the distribution type.
But after all, the properties of a word’s part of speech are the major factors
that condition its distribution, whereas its grammatical meaning is not directly
observable. Distribution is still the fundamental basis for classifying parts of
speech. We should not only rely on word distribution to infer the properties of
a part of speech but also use some means to eliminate the factors that influ-
ence the non-​grammatical meanings of word distribution and the interference
caused by incomplete correspondence between properties and grammatical
positions. The correspondence between a part of speech and word distribu-
tion should be sought from the complicated relationship between properties
of a part of speech and word distribution, thereby reasonably and justifiably
classifying parts of speech according to distribution.

Notes
1 The distribution as Harris meant it refers to the total number of positions in which
an element appears. For the convenience of description in this book, it calls a pos-
ition in which the element appears a distribution of the element.
2 不无 (not without) is an exception.
128

6 
How to classify parts of speech
according to distribution

6.1  Other scholars’ studies

After many years’ discussion, scholars have reached the consensus that Chinese
part-​of-​speech classification should be based on the distribution of words and
proposed their specific classification criteria. But fundamentally they first have
parts of speech in mind and then look for criteria, failing to prove them effect-
ively. For example, why should we follow the criteria that a word that functions
only as an attributive is a distinctive word. And that a word that functions
only as an adverbial is an adverb, instead of classifying them into adjectives?
Why don’t we classify into an independent class words that can only function
as subjects and classify into another independent class words that can only
function as objects? Why do we classify them into nouns instead? We do not
know the reason for using these distributional characteristics to have these
parts of speech instead of using some other distributional characteristics to
have other parts of speech. In the following, we will preliminarily prove our
system of parts of speech and explain the reasons why these distributional
characteristics should be used to classify parts of speech.
How should distributional characteristics be used to classify parts of
speech? The common way is to choose some distributional characteristics to
do the classification without explaining why these distributional characteristics
are chosen. Some scholars have attempted to demonstrate their point of view:

6.1.1  Identification of classification criteria from grammatical characteristics

Lv Shuxiang (1979) said that the ideal classification criteria should have
internal universality and external exclusivity. Zhu Dexi (1985a) thought that
classification criteria should come from grammatical characteristics and gave
an inequation that describes the relationship between classification criteria
and grammatical characteristics: U>V>W, in which U denotes all the gram-
matical properties of a certain class of words; V denotes all the grammatical
characteristics that only this class of words has but other classes do not have
and meanwhile all the members of the class have; and W denotes the criteria
for classifying parts of speech.
But in reality, U is not available.
 129

How to classify parts of speech 129

Table 6.1 Quoted from Zhu Dexi’s Handouts on Grammar, p. 55

很 (very) ~ ~ object Example

A + + 想 (think), 怕 (fear), 喜欢 (like)

B —​ + 唱 (sing), 看 (watch), 讨论 (discuss)
C —​ —​ 醒 (wake), 锈 (rust), 休息 (rest)
D + —​ 大 (big), 好 (good), 干净 (clean)

To identify a grammatical function that has external exclusivity, we can

endow some grammatical functions with conjunctive relations and regard
them as a whole. For example, if the grammatical functions “很 (very) ~” and
“~〈object〉” are endowed with a conjunctive relation, then words can be
classified into four groups:
In other words, although the individual “很 (very) ~” and “~ 〈object〉”
in Groups A, B, C and D have no externally exclusive grammatical functions,
if the two are regarded as a whole, then [+very ~, + ~ 〈object〉] in Group
A, [-​very ~, + ~〈object〉] in Group B, [-​very ~, –​~ 〈object〉] in Group C
and[+ very ~, –​~〈object〉]in Group D all have external exclusivity.
However, as mentioned previously, the vast majority of grammatical
functions do not have internal universality. The endowment of some grammat-
ical functions with a conjunctive relation can help identify externally exclusive
grammatical functions, but their internal universality is even poorer. Therefore,
Groups A, B and C must be combined with one part of speech, namely a
verb. This actually creates a self-​contradiction: none of the three criteria for
classifying into verbs are grammatical characteristics. If we strictly follow
the principle that classification criteria must be identified from grammatical
characteristics, then only three independent parts of speech are available. We
actually endow some grammatical functions with a disjunctive relation and
collectively classify words into one part of speech. For example, according to
Table 6.1, the criteria for verbs are: ~1〈object〉(│* ~〈object〉Λ * very ~).
Put another way, the predicate words that can take objects or can neither take
objects nor be modified by 很 (very) are verbs. But on what basis do we com-
bine Groups A, B and C with one part of speech?
The most fundamental issue is that the identification of classification cri-
teria from grammatical characteristics may actually fall into a circular argu-
ment: only after parts of speech have already been classified can it be decided
what grammatical characteristics exist; however, the classification itself
depends on grammatical characteristics.

6.1.2  The use of major functions as classification criteria

Wang Li (1960: 13) said that “the determination whether a word is a noun
or not depends on whether it often functions as subject or object. The
determination whether a word is a verb or not depends on whether it often
130

130  How to classify parts of speech

functions as a descriptor (the predicate headword in a declarative sentence).
The determination whether a word is an adjective or not depends on whether
it often functions as an attributive”. After that, Chen Wangdao (1978:  48)
also proposed that “the part-​of-​speech classification should be based on
major functions”. At first, the determination of parts of speech based on
major functions meant that the determination should be based on the major
functions (or regular function, advantageous distribution) of a word, but later
the determination was changed to the determination of part-​of-​speech clas-
sification criteria based on major functions of a part of speech, for example,
Mo Pengling and Shan Qing (1985).
This has the two problems:

(1) The overall functional frequency of a part of speech may differ greatly
from the functional frequency of some words in the part of speech.
Namely, the functions are out of balance. This is reflected in two
respects: first, there is a great difference in functional frequency among
the minor classes of words of the same major class; second, there may
also be a great difference in functional frequency among various members
of a class. Because of this unbalanced function, classifying parts of
speech according to major functions certainly brings about chaos in the
classification.
(2) The most fundamental problem is that the use of major functions to
classify parts of speech has the logical fault of a circular argument: on
the one hand, first parts of speech are classified and then their major
functions are determined; on the other hand, the classification must be
based on major functions. The classification of parts of speech according
to the major functions of a word actually means that parts of speech have
already been classified, thus leaving the classification not demonstrated.2

6.1.3  Classification based on the prototype theory

Please refer to Chapter 4 for the discussion on part-​of-​speech classification
based on the prototype theory.

6.1.4  Classifying parts of speech purely according to general distribution

Please refer to Chapter 4 for the discussion on part-​of-​speech classification
according to general distribution.
The common problem with these methods is a circular argument: on the one
hand, first parts of speech are classified and then grammatical characteristics,
major functions and typical members are determined; on the other hand, the
classification must depend on the determination of the latter. Therefore, these
methods are not effective.
In our opinion, the fundamental reasons why the previously mentioned
methods are not successful are that parts of speech are treated as a distribution
 131

How to classify parts of speech 131

type, and that their classification totally depends on distributional simi-
larity. In fact, the relationship between distribution and a part of speech
is complicated. The difference in a part of speech is not simply manifested
in the difference in distribution. A  part of speech is essentially of a type
that combines expressional function with a semantic type and grammatical
function rather than a distributional type. Therefore, our methods should be
changed from classifying the parts of speech of a word according to its dis-
tributional similarity to deducing them from its distribution. As mentioned
previously, the grammatical meaning (namely the expressional function and
the semantic type) of a word is the fundamental cause for constraining its
distribution and basically determines its grammatical distribution. Like a
word form, distribution is only the extrinsic exhibition of the grammatical
meaning of a word. It reflects the grammatical meaning of a word, which is
represented as distribution. Therefore, the basic task of classifying parts of
speech is to determine the correspondence between grammatical function
and a part of speech. Against this theoretical background, we will demon-
strate the Chinese part-​of-​speech classification as follows:

6.2  Functional compatibility and selection of classification criteria

6.2.1  Grammatical function’s values for classification

The reason why part-​of-​speech classification should be based on distribu-
tional characteristics is that the difference among some grammatical functions
indicates the difference in parts of speech. For example, Word A  has the
function of “不 (not) ~” but does not have the function of “〈attributive〉~”,
while the opposite is true for Word B.  We can then deduce that Word A  is
a predicate word but that Word B is a notional word. Such a pair of gram-
matical functions can be called a heterovalent function. But the difference
in some functions does not indicate the difference in parts of speech. For
example, Word C has only the function of〈subject〉, and Word D has only
the function of〈object〉, but both words are nouns. Such a function can be
called equivalent function, namely reflecting several grammatical functions of
the same part of speech. Because equivalent function cannot be used to dis-
tinguish one part of speech from another, it can be used as disjunctive criteria
to classify the same part of speech. For example, the disjunctive criterion〈s
ubject〉│〈object〉│〈attributive〉~ can be used as the criterion for classi-
fying notional words. The presence of an equivalent function shows that not
all differences in distribution reflect differences in parts of speech.
Equivalent functions are transferable and form a cluster of functions, which
are actually the distinctive functions of a part of speech and from which classifi-
cation criteria are selected. A distinctive function can reflect the properties of the
part of speech of a certain class of words. For example, the equivalent function
clusters “不 (not) ~”, 〈predicate〉, ~〈object〉and ~〈complement〉reflect
the properties of a verb; 〈subject〉~, 〈object〉~ and〈attributive〉~ reflect
132

132  How to classify parts of speech

the properties of a noun; and “很 (very) ~”, “很不 (quite not) ~” and * very
~ 〈object〉reflect the properties of an adjective. Different parts of speech
have different distinctive functions. For example,〈subject〉is not the dis-
tinctive function of a verb or adjective. Strictly speaking, a distinctive function
reflects the nature (expressional function or semantic type) of a certain class of
words. For example, the properties of a noun are reference; then〈subject〉~,
〈object〉~ and〈attributive〉~ reflect the properties of reference. The prop-
erty of a verb is a statement; then 不 (not) ~, 〈predicate〉~ and ~〈object〉
reflect the properties of a statement. The nature of an adjective is the statement
of degree property, which is reflected by “很 (very) ~”, “~ degree complement”
and “more than X ~”.
In other words, different functions have different values for classifying
different parts of speech: some have distinctive values; others do not have dis-
tinctive values. Therefore, we cannot treat equally all the functions of a part of
speech and only select those distinctive functions to classify parts of speech.
This is the reason why we select only some distributional characteristics as the
criteria for classifying parts of speech. Based on the selectional restriction of
the grammatical position on words, the fundamental reason why words have
their different distributions is that they themselves have differences in gram-
matical meaning. Therefore, the essence of a part of speech is the category
of the expressional function or the semantic type of a word; distribution is
only the extrinsic manifestation of the expressional function or semantic type.
But some differences in distributional characteristics are distinctive, whereas
others are not distinctive. Therefore, it is impossible to classify parts of speech
purely according to words’ distributional similarity. We select only some dis-
tributional characteristics, namely distinctive distributional characteristics, as
the criteria for classifying parts of speech. In fact, the distinctive distributional
characteristics refer to the grammatical functions that reflect the expressional
function and semantic type of a word.
Theoretically speaking, the identification of all the functions of a language
and the determination of their equivalence gather grammatical functions into
clusters of equivalent functions. In this way the identification of distinctive
functions of all parts of speech is, in effect, the classification of all the words
into different parts of speech. Therefore, the determination of equivalent
functions is the key to identifying classification criteria and determining how
many parts of speech a language has.

6.2.2  The compatibility of grammatical functions and the determination

of equivalent function

6.2.2.1  What information can the compatibility of functions provide?

Now that classification criteria must be selected from distinctive functions, the
determination of equivalent functions becomes the key to classifying parts
of speech. How should we determine equivalent function? We can use the
 133

How to classify parts of speech 133

compatibility of functions to determine their values so as to classify parts of
speech, thereby identifying the functions that distinguish one part of speech
from another and selecting among them the criteria for classifying parts of
speech.
The compatibility of functions refers to the property that the same batch
of words shares two or more grammatical functions. For example, words
that can function as subjects can also function as objects, and vice versa.
Words that can enter “很 (very) ~” can also enter “~ 极了 (extremely)”
and “~ 得很 (nicely)”; words that can enter “~ 极了 (extremely)” and “~
得很 (nicely)” can also enter “很 (very) ~”. But the compatibility of some
other functions is extremely small, for instance, “不 (not) ~”, “numeral
~”, “〈numeral and measure word〉~” and 〈adverbial〉. What does the
large compatibility between two functions mean? What does the extremely
small compatibility between two functions mean? As mentioned before,
the fundamental basis for the selectional restriction of grammatical pos-
ition on words is its grammatical meaning. Therefore, functions that have a
rather big compatibility often reflect the same selectional restriction of two
different grammatical positions on words and also the common properties
of parts of speech, thus their functions being equivalent. Functions that
have a rather small or no compatibility reflect different properties of a part
of speech and are generally heterovalent functions. Hence, we can use the
compatibility of functions to determine whether a grammatical function is
equivalent or heterovalent.

6.2.2.2  Compatibility calculation and compatibility among major

functions of Chinese
The size of compatibility among functions is different. The use of functional
compatibility to determine the values for parts-​of-​speech classification should
calculate the compatibility (c) between two functions. The formulae for cal-
culating compatible degrees are given as follows: assuming C≤1, the unidir-
ectional compatible degree of Function x relative to Function Y (cx-​y), the
unidirectional compatible degree of Function Y relative to Function x (cy-​x)
and the total compatible degree between Functions X and Y (cx*y):

(1) Cx-​y=xy number of overlapping words/​number of x words

(2) Cy-​x=xy number of overlapping words/​number of Y words
(3) Cx*y=xy number of overlapping words/​(number of x words +
number of Y words –​xy numbers of overlapping words)

For example:
In Figure 6.1, 100 words have Function X and 40 words have Function Y,
among which 80 words have only Function X, 20 words have only Function Y
and 20 words have concurrently X and Y functions. The compatible degrees
are as follows:
134

134  How to classify parts of speech

Cx-​y=20/​100=0.2
Cy-​x=20/​40=0.5

The total compatible degree is not significant. As Figure  6.2 shows,

Cx*y=20/​(100+20-​20)=20/​100=0.2. Although the total compatible degree can
reflect the compatible degree of X in relation to Y, it cannot reflect the com-
patible degree of Y in relation to x: Cy-​x=20/​20=1.
We use the above-​mentioned method to calculate the compatible degrees
among the major functions of Chinese notional words (see Table 6.2).

6.2.2.3  Methods for determining equivalent functions

Equivalent functions can be fundamentally determined by distributional
compatible degrees. Although the distribution of a word is mainly decided
by its part of speech, it may also be influenced by other factors such as
semantic and pragmatic meanings, rhyme and word formation methods.
Moreover, some grammatical positions may allow several parts of speech to
enter; some words may have several parts of speech. Therefore, we cannot
totally rely on functional compatible degrees to determine a function’s
values for classifying parts of speech; other factors must be considered
and the interference on functional compatibility caused by them must be
eliminated.

x
80
20 y
20

Figure 6.1 Example 1

80

20

Figure 6.2 Example 2
 135

How to classify parts of speech 135

We use the following rules to decide the equivalent relations among gram-
matical functions.
Rule 1: if the unidirectional compatible degree of two functions is less than
0.5 (c<0.5), then we can infer that they reflect different parts of speech and
belong to heterovalent functions.3 If a word has concurrently two functions,
then it may be deemed to have concurrently two parts of speech. But if it is
found and properly reasoned that a word meets the conditions for a certain
function, then the word cannot be deemed to have several parts of speech;
instead, it is because its grammatical position allows several parts of speech to
enter. If any unidirectional compatible degree of two functions is larger than
0.5 (c>0.5), then it must be tested with Rules 2 through 4.
For example, the unidirectional compatible degree of 不 (not) ~
&〈attribute〉, 不 (not) ~ &〈adverbial〉, 很 (very) ~ &〈attributive〉,
很 (very) ~ &〈adverbial〉, 不 (not) ~ &〈noun〉~〈attributive and head-
word〉, 〈attributive〉&〈adverbial〉 respectively is less than 0.5 and can
be deemed to have a heterovalent function. The 〈numeral and measure
word〉~ &〈adverbial〉 has a unidirectional compatible degree of less than
0.1, but can use as an adverbial a word that can be modified by numerals and
measure words. For the word, conditions can be found and properly reasoned.
There are mainly three kinds of conditions:  (1) metaphors such as 拳头大
(as big as fist) and 碗口粗 (as wide as bowl rim);4 (2) methods and tools such
as 电话采访 (telephone interview), 公款请客 (meal treatment with public
money) and 掌声鼓励 (encourage with applause); (3) places such as 操场去
(go to sports ground), 学校见 (see you at school) and 主场迎战对手 (play
with an opponent head-​on at home ground). The function whose conditions
can be found and properly reasoned is a syntactic function but not a lexical
one. Therefore, these words or phrases should not be considered as adverbs.
In other words, the adverbial position also allows some nouns to enter.
Rule 2:  if a word having one of the functions and a word not having the
function exhibit great difference at other grammatical positions, then the
functional compatibility is because the grammatical position allows several
parts of speech to enter and because a word has several parts of speech. They
should not be deemed to have equivalent functions. If the difference is small
at other grammatical positions, then they are equivalent functions.
For example, the unidirectional compatible degrees of 不 (not) ~│no ~
&〈subject〉│〈object〉 are 0.58 and 0.2 respectively. But at the positions
of〈numeral and measure word〉~, predicate and complement, they
exhibit great differences:  the words that can enter 不 (not) ~ usually can
also enter〈predicate〉and ~〈complement〉but cannot enter〈numeral
and measure word〉~. But most of the words that can enter 不 (not) ~ can
enter〈numeral and measure word〉~ but cannot enter〈predicate〉and
~〈complement〉. Therefore, we can conclude that compatibility is because
the position〈subject〉│〈object〉 allows several parts of speech to enter.
Furthermore, the unidirectional compatible degree of 〈predicate〉 &
 newgenrtpdf
136
Table 6.2 Compatible degrees of the main grammatical functions of Chinese notional words

Grammatical function Number of words Compatibility Equal

value

x y x&y Cx-​y grade Cy-​x grade Cx﹡y grade

Not ~& no ~ 11020 10790 9918 0.90 ++ 0.92 ++ 0.83 + +

Not~& predicate 11020 13261 11000 1.00 ++ 0.83 + 0.83 + +
No ~& predicate 10790 13261 10754 1.00 –​ + 0.81 + 0.81 + +
Not: no ~ &predicate 11809 13261 11788 1.00 ++ 0.89 + 0.89 + +
Not|no ~& bound 11809 506 497 0.04 –​ –​ 0.98 ++ 0.04 –​ –​ +
complement
Not|no ~& group 11809 6484 6090 0.52 + 0.94 ++ 0.50 + +
complement
Not|no ~& ~ bound 11809 6820 6748 0.57 + 0.99 ++ 0.57 + +
complement
Not|no ~& ~ real object 11809 6163 5783 0.49 –​ 0.94 ++ 0.47 –​ –​
Not|no ~& ~ 11809 7467 6755 0.57 + 0.90 ++ 0.54 + +
semi-​ object
Not|no ~ & ~ zhuo, le, 11809 10459 10240 0.87 + 0.98 ++ 0.85 –​ +
guo
Not|no ~ & adverbial~ 11809 13345 11837 1.00 ++ 0.89 + 0.89 + +
Not|no ~ & very ~ 11809 2552 2509 0.21 –​ 0.98 ++ 0.21 –​ +
Not|no ~ & attributive 11809 23544 3764 0.32 –​ 0.16 –​ 0.12 –​ –​
Not|no ~ & adverbial 11809 1592 462 0.04 –​ –​ 0.29 –​ 0.04 + –​ –​
Subject & object 31394 33989 30351 0.97 ++ 0.89 + 0.87 + +
Not|no ~& 11809 34796 6842 0.58 + 0.20 –​ 0.17 –​ –​?
subject|object
Predicate & 13261 34796 7481 0.56 + 0.22 –​ 0.18 –​ –​?
subject|object
 newgenrtpdf
 137
Grammatical function Number of words Compatibility Equal
value

x y x&y Cx-​y grade Cy-​x grade Cx﹡y grade

Not|no ~ & noun 11790 14298 630 0.05 –​ –​ 0.04 –​ –​ 0.02 –​ –​ –​
(attributive)~
Predicate & adverbial~ 13261 13477 13122 0.99 ++ 0.87 ++ 0.96 ++ +
Predicate & bound 1573 506 53 0.03 –​ –​ 0.10 –​ 0.03 –​ –​ –​
complement
Very ~&~ 得很 (nicely) 2552 1607 1607 0.63 –​ 1.00 ++ 0.63 –​ +
Very ~&~ 极了 2552 2012 2012 0.79 + 1.00 ++ 0.79 + +
(extremely)
Very ~& 很不 (very 2552 1008 985 0.39 –​ 0.98 ++ 0.38 –​ +
not) ~
Very ~ & bound 2552 506 192 0.08 –​ –​ 0.38 –​ 0.07 –​ –​ +?
complement
Very ~& group 2552 6494 1558 0.61 + 0.24 –​ 0.21 –​ +
complement
Very ~& attributive 2552 23544 709 0.28 –​ 0.03 –​ –​ 0.03 –​ –​ –​
Very ~& adverbial 2552 1592 273 0.11 –​ 0.17 –​ 0.07 –​ –​ –​
Very ~&~ zhuo, le, guo 2552 10459 1888 0.74 + 0.18 –​ 0.17 –​ +
Very ~& predicate 2552 13261 2533 0.99 ++ 0.19 –​ 0.19 –​ +
Very ~&~ real object 2552 6163 203 0.08 –​ –​ 0.03 –​ –​ 0.02 –​ –​ +?
attributive & adverbial 23544 1592 296 0.01 –​ –​ 0.19 –​ 0.01 –​ –​ –​
adverbial & group 1550 6494 240 0.15 –​ 0.04 –​ –​ 0.03 –​ –​
complement
number ~& attributive 24314 23544 17731 0.73 + 0.75 + 0.59 + –​?
number ~ & adverbial 24314 1592 112 0.00 –​ –​ 0.07 –​ –​ 0.00 –​ –​ –​

(continued)
 newgenrtpdf
138
Table 6.2  (Cont.)

Grammatical function Number of words Compatibility Equal

value

x y x&y Cx-​y grade Cy-​x grade Cx﹡y grade

subject| object & 34796 24314 22887 0.66 + 0.94 ++ 0.63 + +

number ~
subject| object & 34796 23544 22105 0.64 + 0.94 ++ 0.61 + –​?
attributive
subject| object & 34796 1592 504 0.01 –​ –​ 0.32 –​ 0.01 –​ –​ –​
adverbial
subject| object & 34796 33706 30881 0.89 + 0.92 ++ 0.82 + +
attributive ~
subject| object & noun 34796 14538 14124 0.41 –​ 0.97 ++ 0.40 –​ +
(attributive) ~
subject| object & 34796 13472 3272 0.09 –​ –​ 0.24 –​ 0.07 –​ –​
adverbial ~

Notes: 1. The number of words that have the ※ function is calculated from the statistical sampling proportion. 2. The critical value of compatible degrees: high
compatible degree: (+)c>=0.5, including extremely high compatible degree (++): c>=0.9; low compatible degree (-​):c<0.5, including extremely low compatible
degree (–​): c<0.1. 3. The “x & y” denotes the number of words that have Functions X and Y. 4. In the “equivalent” column, “+” denotes equivalence; “-​” denotes
unequivalence; “-​?” indicates that the compatible degree is higher than 0.5, but according to relevant rules, 4j should be unequivalent; “+?” indicates that the
compatible degree is less than 0.5 but should be applied according to relevant rules.
 139

How to classify parts of speech 139

〈subject or object〉 is at most 0.56. Words that can function as predicates
can also enter 不 (not) ~ but usually cannot enter 〈numeral and measure
word〉~, while those that cannot function as predicates usually can enter
〈numeral and measure word〉~ but cannot enter 不 (not) ~. The compatible
degree of the two functions is because the position〈subject or object〉allows
several parts of speech to enter. The functions are not equivalent.
Rule 3: if a large portion of words having Functions X and Y have one function
that can form the construction consisting of a modifier and the word it modi-
fies, then the compatibility of X and Y, by inference, is because some words
have several parts of speech. There is no way to determine whether the words
have equivalent functions. For example,〈numeral and measure word〉~
&〈attributive〉,〈subject or object〉&〈attributive〉(桌子 (desk), 节目
(program)  –​ 木头 (wood), 电视 (television)). These pairs of functions have
the highest compatible degrees of 0.75 and 0.94 but can form the immediate
constituents 木头桌子 (wooden desk) and 电视节目 (television program). Thus,
it can be inferred that the words 木头 (wood) and 电视 (television) have several
parts of speech, but it cannot be inferred that the two functions are equivalent.
The reason why we put forward this rule is that we find that the same word
forms in the positions of the modifier and headword construction have very
different syntactic expressions, for example:

(4) a. 也干净 (also clean),  也不干净 (also not clean),   也很干净

(also very clean)
b. 干净衣服 (clean clothes),  *不干净衣服 (* not clean
clothes),   *很干净衣服 (* very clean clothes)
(5) a. 都认真 (all serious),   都不认真 (all not serious),   都很认真
(all very serious)
b. 认真学习 (serious study)   *(不认真）学习 (*(not seriously)
study),   很认真学习 (very serious study)
(6) a. 我的木头 (my wood),   我的十根木头 (my ten pieces of wood)
b. 木头房子 (wooden house),   *+根木头房子 (*house of ten
pieces of wood)

木头 (wood) can be modified by numerals and measure phrases at the head-​

word position. This is the same as in the subject or object position, but it cannot
be modified by numerals and measure phrases at the attributive position, indi-
cating that its part of speech has changed (see Lv Shuxiang [1979], Zhang
Bojiang [1994] and Guo Rui [1997a]). In terms of grammatical meaning, 干净
(clean) and 认真 (serious) are statements in the head-​word position, which is
the same as in the predicate position. But they are modifiers in attributive or
adverbial positions. Just like in the subject or object position, 木头 (wood) is a
reference in the head-​word position, but is a modifier in the attributive position
(see Chapter 4). The two functions are highly compatible but do not reflect the
same part of speech. The fundamental reason for this is that the two parts of
speech do not highly differ from each other in the lexicon (see Chapter 4).
140

140  How to classify parts of speech

Rule 4: If the number of words x and y significantly decreases as their fre-
quency decreases, then by inference, the correlation between Functions X and
Y is because some words have several parts of speech.
The reason why we put forward this rule is that we find that the times of
appearance of one of the functions decrease significantly as the word fre-
quency decreases, whereas another function has no significant or negative
correlation with word frequency. As Table 6.3 shows, there is a positive correl-
ation between the compatible degree between 很 (very) ~ and 〈attributive〉
and word frequency because the number of words that have 〈attributive〉
decreases significantly as their frequency decreases, whereas 很 (very) ~ does
not decrease as the word frequency decreases but increases, presenting a nega-
tive correlation. In contrast, the compatible degrees between 很 (very) ~ and
不 (not) ~ and between 很 (very) ~ and ~ zhuo (着), le (了), guo (过) do not
have a positive correlation with word frequency. Meanwhile, we find that there
is a prominent positive correlation between a conversional word and word
frequency (see Table 6.4). We can say that the number of parts of speech of a
word has a positive correlation with word frequency,6 and thus are justified in
thinking that the compatibility of two functions is because some words have
several parts of speech (concurrent parts of speech or a conversional word).
Rule 5: An equivalent function is transferable; if Function X is equivalent to
Function Y, and Function Y is equivalent to Function Z, then Function X is
also equivalent to Function Z, thus forming an equivalent function cluster.
Although the compatibility degree among some functions is less than 0.5,
because they are equivalent to a certain identical function, the two functions
are transferable and thus equivalent. For example, the compatible degrees of
the two pairs of functions of 很 (very) ~ &〈complement〉 and 很 (very) ~ &
~〈real object〉 are less than 0.5, but 很 (very) ~ is equivalent to 不 (not)│没
(no) ~, but the latter is equivalent to 〈complement〉 and ~〈real object〉.
According to transferability, 很 (very) ~ is also equivalent to 〈complement〉
and ~〈real object〉.
Rule 6:  The generalization levels of grammatical function and parts of
speech represented by the grammatical function are consistent. The functions
whose generalization level is low in the equivalent cluster can determine the
equivalent functions at the same generalization level and represent different
subclasses. If the unidirectional compatible degree between the function
whose generalization level is high in an equivalent function cluster and the
function whose generalization level is low is greatly disparate, as shown in
Figure 6.2, then the function whose generalization level is low may represent
a subclass. Reference should be made to other characteristics such as redupli-
cation form and whether to classify subclasses or not.
As mentioned previously, the generalization level of a function such as a
syntactic constituent is higher than that of the function of a specific word or
a part of speech; the generalization level of the part of speech represented
by the former is also higher than that of the latter. For example, 〈subjec
t〉│〈object〉│〈attributive〉~, whose equivalent function has a high
 newgenrtpdf
 141
Table 6.3 The correlation between compatible degree and word frequency, and that between function and word frequency5

Frequency level Level 1 Level 2 Level 3 Level 4 Level 5 Total Correlational Marked­
coefficient ness
Number of 512 512 512 512 512 2560
words

Medium word 3113 535 167 48 4

frequency

Number Compati­ Number Compati­ Number Compati­ Number Compati­ Number Compati­ Number Compati­
of words bility of words bility of words bility of words bility of words bility of words bility

很(very) ~ & 282 0.55 169 0.33 126 0.25 85 0.17 45 0.09 707 0.28 0.925 +
attributive
很(very) ~ & 459 0.90 474 0.93 487 0.95 484 0.95 482 0.94 2386 0.93 -​0.946 +
不(not) ~
很(very) ~ & 着 399 0.78 397 0.78 404 0.79 360 0.70 307 0.60 1867 0.73 0.444 -​
(zhuo), 了(le),
过(guo)

Number pro­portion Number pro­ Number pro­ Number pro­ Number pro­ Number Pro­
of words of words portion of words portion of portion of portion of portion
words words words

很(very) ~ 498 97% 506 99% 508 99% 510 100% 504 98% 2526 99% -​0.873 -​
Attributive 292 57% 172 34% 126 25% 86 17% 46 9% 722 28% 0.931 +
 newgenrtpdf
142
Table 6.4 The correlation between number of conversional words and word frequency7

Frequency Level 1 Level 2 Level 3 Level 4 Level 5 total Correlation Marked-​

level coefficient ness

Number of 7622 7622 7622 7622 7622 38110

words

Medium 2901 401 114 33 6

frequency

Conversional Number Proportion Number Proportion Number Proportion Number Proportion Number Proportion Number Proportion
word
1308 17% 417 5.% 229 3.0% 136 1.8% 95 1.2% 2185 5.7% 0.996 ++
 143

How to classify parts of speech 143

generalization level, represents a notional word, but the compatible degrees
between 〈numeral〉~, whose generalization level is low, and “在 (in) ~” are
0.01 and 0. They belong to heterovalent functions at the low generalization
level and represent two subclasses. For another example, the unidirectional
compatible degrees of 很 (very) ~ and〈predicate〉are 0.99 and 0.19 respect-
ively, being greatly disparate. 很 (very) ~ has a function whose generalization
level is low and thus represents a subclass.

6.3  Classification criteria for Chinese notional words

6.3.1  Criteria for classifying major parts of speech of

Chinese notional words
According to the above-​mentioned rules, we determine the equivalence of
Chinese grammatical functions and form the following equivalent function
clusters:

Equivalent Function Cluster 1:  不 (not) ~, 没 (no) ~, ~ zhuo (着), le

(了), guo (过), 〈predicate〉, 〈complement〉, ~〈complement〉,
~〈real object〉, ~〈quasi-​object〉, 〈adverbial〉~, very~, 很不
(quite not)~, ~极 了 (extremely), ~ 得很 (nicely)
Equivalent Function Cluster 2: 〈subject〉, 〈object〉, 〈attribute〉~,
〈numeral and measure word〉~
Equivalent Function Cluster 3:〈modifier〉8
Equivalent Function Cluster 1 represents predicate words; Equivalent
Function Cluster Two represents notional words; Equivalent
Function Cluster 3 represents modifiers.

But an equivalent function cluster is not the criterion for classifying parts of
speech because some grammatical positions may allow several parts of speech
to enter. Therefore, although it has internal universality, it has no external
exclusivity. This can be demonstrated by the co-​occurrence9 of unequivalent
functions such as 不 (not) ~ and〈subject〉; for instance, 不去是应该的 (not
going is obligatory). Then, how can we decide which position allows several
parts of speech to enter? This can be decided by the hierarchy of co-​occurring
functions: the functions at the outer hierarchy do so. Under this circumstance,
the classification of parts of speech should give first priority to inner-​layer
functions, for example, 不∥去/​是应该的 (not //​going/​is obligatory). A deci-
sion can be made that the〈subject〉function in the outer layer allows several
parts of speech to enter. According to “不 (not) ~”, 去 (going) should be a
predicate word. For another example,〈predicate〉and 〈attributive〉~ are
not equivalent but can co-​occur in the following: 小王/​黄∥头发 (Little Wang/​
yellow//​hair). Its predicate function is at the outer layer. Therefore, we can
assume that the predicate position allows several parts of speech to enter.
According to its inner-​layer function 〈attributive〉~, 头发 (hair) should be
144

144  How to classify parts of speech

a notional word. For another instance, 〈attributive〉~ and 〈adverbial〉~
are not equivalent but can co-​occur in the following examples:

(7) a. 这本书的／及时∥出版 (The punctual //​publication/​of the

book)
b. *及时／这本书的∥出版 (*The punctual/​publication //​of the
book)
(8) a. (小王) 也/​黄∥头发 ((Little Wang) also/​yellow //​ hair)
b. *(小王) 黄/​也∥头发 (*(Little Wang) yellow/​also //​ hair)

These two examples indicate that both〈attributive〉~ and〈adverbial〉~

allow several parts of speech to enter. According to its inner-​ layer
function〈adverbial〉~, 出版 (publish) should be a predicate word; according
to its inner-​layer function〈attributive〉~, 头发 (hair) should be a notional
word (see Chapter 4).
That is to say, an equivalent function cluster only shows that the functions
in the same function cluster cannot distinguish one part of speech from
another. The classification criteria must be selected from the equivalent
function cluster and restrict the positions that allow several parts of speech
to enter, thus having external exclusivity. For example, only words that can
function as subjects or objects can be modified by attributives, and words
whose functions in Equivalent Function Cluster 1 are used as inner-​layer
functions, are classified into notional words. In addition, classification cri-
teria are also related to classification strategies. This issue will be discussed
in Chapter 7. Here we assume that it has been solved. After considering the
above-​mentioned issues, we can formulate the following criteria for classifying
all the major parts of speech of notional words:10

Predicate words: 不 (not) ~│没 (no) ~│很 (very) ~│很不 (very not) ~│~
〈object〉〈complement〉│〈complement〉│〈predicate〉Λ
*〈subject〉│〈object〉
Substantive words:  〈subject〉│〈real object〉│〈attributive〉~ Λ
*〈predicate word〉
Modifiers: 〈modifier〉 Λ (*〈predicate word〉│〈substantive word〉)
Substantive modification words:〈attributive〉Λ(*〈predicate
word〉│〈substantive word〉)
Predicate modification words:〈adverbial〉Λ(*〈predicate
word〉│〈substantive word〉)

6.3.2  Classification of basic classes

Rule 6 says that the generalization levels of grammatical function and parts
of speech represented by the grammatical function are consistent. Therefore,
predicate words, substantive words and modification words determined with
 145

How to classify parts of speech 145

the functions in the above-​mentioned same equivalent function cluster can be
further classified into such basic classes as nouns and verbs according to the
equivalent function cluster whose generalization level is low. We also infer the
classification values of a function according to the compatible degree among
functions and relevant rules. In the following, we first look at the compatible
degrees among the major functions of a modifier (see Table 6.5).
We must first make clear that we define grammatical functions by using the
concepts of parts of speech such as 〈numeral〉~, ~〈locative〉. Then do
we fall into a circular argument? Yes. The circular argument can be resolved
by listing all or some members of a class. For example, numerals and locatives
belong to a closed class11 and all their members can be listed. Some typical
members of measure words can be listed, for instance, 个 (piece), 斤 (jin),
项 (item) and so on. According to the fixed-​point theory, Bai Shuo (1995)
demonstrated that the use of not yet defined parts of speech to classify parts
of speech, in theory, does not involve a logically circular argument.
The present book holds that the use of parts of speech as an environ-
ment does not involve a logically circular argument: As mentioned before, a
grammatical position refers to the position in which immediate constituents
that have certain grammatical relations exist in a syntactic structure.
Fundamentally speaking, the grammatical position that uses parts of speech
as an environment is still decided by grammatical relations. For example,
X〈locality〉refers to a construction consisting of a modifier and the word
it modifies; the construction has its relative positional relations in a reference
system. 〈numeral〉X refers to the construction consisting of a modifier and
the word it modifies; the construction has its numerical measurement unit
relations. Therefore, the use of parts of speech to define the environment, in
essence, is to use the specific grammatical relations represented by a part-​of-​
speech combination, but grammatical relations are independent of parts of
speech. Therefore, there is no circular argument.
Among modifiers, the compatible degrees of more specific functions〈at
tributive〉and〈adverbial〉than〈modifier〉are 0.01 and 0.19 respectively,
thus leading us to conclude that they are not equivalent and form two equiva-
lent function clusters:

Equivalent Function Cluster 3.1:〈attributive〉

Equivalent Function Cluster 3.2:〈adverbial〉

Consequently, we can classify modifiers into determiners and adverbs:

Determiner: 〈attributive〉Λ (*〈predicate word〉│〈substantive word〉)

Adverb:  〈adverbial〉Λ* (〈predicate word〉│〈substantive word〉)
{马上(immediately), 亲自 (personally), 特意 (specially), 也 (also)}

Among determiners, the compatible degrees among the functions ~〈quan-

tity〉, for example, 三个 (three pieces), ~ X de (的) 〈noun〉,12 for example,
 newgenrtpdf
146
Table 6.5 The compatible degrees among specific grammatical functions of modification words

Grammatical function Number of words Compatibility Equal

value

x&y x y x&y Cx-​y grade Cy-​x grade Cx﹡y grade

~ measure word & ~ x de (的) noun1 54 64 6 0.11 –​ 0.09 –​ –​ 0.05 –​ –​ –​
~ measure word & ~ numeral, measure word, noun 54 11 5 0.09 –​ –​ 0.45 –​ 0.08 –​ –​ –​
~ measure word&numeral, measure word ~ noun 54 449 0 0.00 –​ –​ 0.00 –​ –​ 0.00 –​ –​ –​
~ numeral, measure word, noun & ~ x de (的) noun 11 64 3 0.27 –​ 0.05 –​ –​ 0.04 –​ –​ –​
~ numeral, measure word, noun & numeral, measure 11 449 0 0.00 –​ –​ 0.00 –​ –​ 0.00 –​ –​ –​
word ~ noun
~ x de (的) noun & numeral, measure word ~ noun 64 449 0 0.00 –​ –​ 0.00 –​ –​ 0.00 –​ –​ –​
~ measure word & quasi-​object 54 31 5 0.09 –​ –​ 0.16 –​ 0.06 –​ –​ –​
quasi-​object & ~ numeral, measure word, noun 31 11 0 0.00 –​ –​ 0.00 –​ –​ 0.00 –​ –​ –​
quasi-​object & numeral, measure word ~ noun 31 449 0 0.00 –​ –​ 0.00 –​ –​ 0.00 –​ –​ –​
quasi-​object & ~ x de (的) noun 31 64 13 0.42 –​ 0.20 –​ 0.16 –​ –​
 147

How to classify parts of speech 147

所有迟到的学生 (all the late students), ~〈numeral and measure word plus
noun〉, for example, 这三个学生 (these three students), 〈numeral and
measure word〉~ 〈noun〉, for example, 三台黑白电视 (three sets of black-​
and-​white television) are very low. We basically conclude that the functions
are not equivalent. Hence, we further classify determiners into four parts of
speech:

Numeral:  ~ 〈quantity〉Λ *(not ~│~〈numeral+ measure

words+noun〉{一 (one), 几 (several), 半 (half), 十 (ten)}
Demonstrative:  ~〈numeral+measure word+noun〉 Λ (*〈predicate
word〉│〈substantive word〉){每 (each), 任何 (any), 其他 (other),
这 (this)}
Numeral and measure word:  ~ X de (的)〈noun〉 Λ * ~〈numeral +
measure word + noun〉{许多 (many), 一切 (all), 俩 (both)}
Distinctive word:  〈numeral and measure word〉~ 〈noun〉Λ
*(〈predicate word〉│〈substantive word〉){高等 (high-​class),
公共 (public), 野生 (wild), 日常 (routine)}

The criteria of the latter half for the four types of determiner use the negative
values of conjunctive criteria in order to eliminate other parts of speech that may
have the functions of the former half. Among criteria for numerals, adjectives
and demonstratives may also occur in the position of ~〈measure word〉. For
example, 大块 (big piece), 每台 (each set) and (* not ~│~〈numeral + measure
word + noun〉) need to be used to eliminate adjectives and demonstratives.
Among criteria for demonstratives, predicate words and notional words may also
occur in the position of ~〈numeral+ measure word + noun〉, for instance, 雪白
(snow-​white) and 一双鞋 (a pair of shoes), and therefore should be eliminated
together. Moreover, classical Chinese demonstratives such as 本 (this) and 该
(these) cannot appear in the above-​mentioned positions and are different from
the function of a colloquial modern Chinese demonstrative, thus requiring other
classification criteria. For detailed discussion, see Chapter  1.5.10 in Volume
2. Among criteria for numerals and measure words, a demonstrative may also
occur in the position of ~ X de (的)〈noun〉. We use *~〈numeral+measure
word+noun〉to eliminate them. Among criteria for distinctive words, notional
words and predicate words may also occur in the position of 〈numeral and
measure word〉~〈noun〉. We use *〈predicate word〉│〈notional word〉to
eliminate them.
In terms of grammatical meaning, numerals and measure words indi-
cate quantity and actually have the same integral function as a phrase
formed by “numeral + measure word”. Compare:  许多迟到的学生 (many
late students) –​ 十个迟到的学生 (ten late students); 来了三个 (three came) –​
来了许多 (many came). This is the reason why we classify them into numerals
and measure words.
Actually, two types of modifiers can function as adverbials. One type
always functions as an adverbial before a modified constituent and is
148

148  How to classify parts of speech

classified into adverbs; another type is usually placed after a modified con-
stituent and is called by Zhu Dexi (1982b) a quasi-​object. For example,
高一点 (a little higher), 等候片刻 (wait for a while) and 沉思许久 (meditate
for a long time). But in a negative expression, it is generally placed in front.
For example, 一点也不高 (not high at all), 片刻不得安宁 (cannot be peaceful
for a moment) and 许久未来 (for a long time not come). In our opinion,
a quasi-​object in its strict sense is actually a type of modifier and is iden-
tical to an adverbial placed in front in terms of its grammatical relations.
Or we can call this a “rear adverbial” (see Guo Rui 1997a). Phrases such as
许久 (for a long time) and 片刻 (for a moment) actually indicate quantity in
terms of their grammatical meanings, and their grammatical functions are the
same as other phrases that indicate quantity. Compare: 等了三天/​等了许久
(wait for three days /​wait for a long time), 高两厘米/​高一点 (two centimeters
high /​a little higher). Moreover, some words can actually appear both in ~
X de (的) 〈noun〉 and in the position of a quasi-​object, for instance, 一些
(some), 许多 (many), 不少 (a good many), 片刻 (for a moment) and 一点 (a
bit). Therefore, we can also classify into numerals and measure words the
words that appear in the position of quasi-​object. In this way, the properties
of numerals and measure words are not completely consistent. Some of them
have only the properties of substantive modification words such as 一切 (all),
少许 (a little) and 俩 (both); some have only the properties of predicate modi-
fication words (adverbs) such as 许久 (for a long time), 良久 (very long time)
and 不久 (soon); some have the properties of both substantive modification
words and predicate modification words such as 一些 (some), 许多 (many),
片刻 (for a moment) and 丝毫 (the slightest). The criteria for numerals and
measure words can be amended as -​X de (的) 〈noun〉│〈quasi-​object〉Λ
* ~〈numeral + measure word + noun〉.
Can the rest of adverbs be classified into smaller subclasses? Quite a
number of scholars have noticed that an adverb may have a virtual or a real
meaning. For instance, the meanings of 也 (also), 还 (still) and 究竟 (on earth)
are rather virtual, and those of 特意 (specially), 飞速 (rapidly) and 亲自 (per-
sonally) are rather real. But it is difficult to classify them into even smaller
types. Zhang Yisheng (1995a) classified them into adverbs or adverbial words;
Chen Yi (1989) classified them from adverbs into preattached words, but his
criteria lack internal universality and cannot effectively classify the two types
of words.
Among substantive words, the compatible degrees of 〈quantity〉~ and
~〈locality〉 are very high (see Table  6.6). We posit that they are equiva-
lent functions according to the rules proposed in Chapter 6.2.2.3. The com-
patible degrees of “在 (in) [〈notional words〉]~”13 and 〈quantity〉~,
“在 (in)[〈notional words〉]~” and ~〈locality〉] are also more than 0.5.
But the number of these two pairs of functionally overlapping words has
a prominently positive correlation with word frequency (see Table 6.7). “在
(in)[〈notional words]〉~” decreases prominently as their word frequency
decreases, but 〈numeral and measure word〉~ and ~〈locality〉 have no
 newgenrtpdf
 149
Table 6.6 The compatible degrees among specific grammatical functions of substantive words

Grammatical function Number of words Compatibility Equal

value

x&y x y x&y Cx-​y grade Cy-​x grade Cx﹡y grade

<numeral> ~ & 在 (in) [notional word] ~ 509 1313 6 0.01 –​ –​ 0.00 –​ –​ 0.00 –​ –​ –​
<numeral > ~& <numeral, measure word > ~ 509 21423 207 0.41 –​ 0.01 –​ –​ 0.01 –​ –​ –​
<numeral> ~ & ~ <locative> 509 22683 250 0.49 –​ 0.01 –​ –​ 0.01 –​ –​ –​
在 (in) [notional word] ~ & <numeral, measure 1313 21423 871 0.66 + 0.04 –​ –​ 0.04 –​ –​ –​?
word> ~
在(in) [notional word] ~ & ~ <locative> 1313 22683 1018 0.78 + 0.04 –​ –​ 0.04 –​ –​ –​?
<numeral, measure word> ~ & ~ <locative> 21423 22683 19553 0.91 ++ 0.86 + 0.80 + +
 newgenrtpdf
150
Table 6.7 The correlation between word frequency and compatible degree (substantive word)

Frequency level Level 1 Level 2 Level 3 Level 4 Level 5 Total Correlation Marked­
coefficient ness
Number of words 5541 5542 5542 5542 5542 27709

Medium frequency 1094 142 34 6 0

N Pro­ N pro­ N Pro­ N Pro­ N Pro­ N Pro­

u portion u portion u portion u portion u portion u portion
m m m m m m
b b b b b b
e e e e e e
r r r r r r

在 (in) [notional 450 8.1% 164 3.0% 81 1.5% 36 0.6% 155 2.8% 886 3.2% 0.959 ++
word] ~ &
<quantifier>
在 (in) [notional 554 10.0% 221 4.0% 110 2.0% 50 0.9% 160 2.9% 1095 4.0% 0.970 ++
word] ~ & ~
<dialect>
在 (in) [notional 558 10.6% 225 4.1% 111 2.0% 51 0.9% 162 2.9% 1137 4.1% 0.974 ++
word]
<quantifier> ~ 4289 77.4& 4232 76.4% 4126 76.4% 4279 77.2% 4387 79.2% 21423 77.3% -​0.037 -​
~<locational word> 4766 86.0% 4616 83.3% 4711 85.0% 4711 85.0% 4655 84.0% 23459 84.7% 0.652 -​
 151

How to classify parts of speech 151

prominent correlation with word frequency. According to Rule 4, we think
that the compatibility is because the words that can enter “在 (in)[〈notional
words〉]~” have several parts of speech. We cannot assert that these two pairs
of words have equivalent functions.
Thus, there are three equivalent function clusters among substantive
words:

Equivalent Function Cluster 2.1:  〈numeral〉~ (it may also take a

numeral as the environment)
Equivalent function cluster 2.2: in[notional word]~ equivalent function
cluster
Equivalent Function Cluster 2.3:  〈numeral and measure word〉~, ~
〈locational word〉

Hence, we classify measure words, locational words and nouns. The criteria
are as follows:

Measure word:  (一 (one)│几 (several) ~ Λ *〈subject〉{个 (piece), 次

(times), 年 (year), 斤 (jin)}
Locational word:  在 (in) [〈notional word〉]~ Λ * ~ 里 (inside)│以南
(to the south)
Noun:  〈subject〉│〈real object〉│〈attribute〉~│~ 里
(inside)│以南 (to the south) Λ (*〈predicate word >│〈measure
word〉│〈locational word〉){桌子 (desk), 面积 (area), 地步
(extent)}

Criteria for measure words:  some nouns can enter 一 (one)│几 (several)
~, for example, 这一地区 (this one region) and 几兄弟 (several brothers) can
be eliminated with *〈subject〉. Among criteria for locational words, some
nouns can also enter 在 (in)[〈notional word〉]~, for instance, 在教室 (in a
classroom), 在北京 (in Beijing) and 在操场 (at the sports ground). But these
words can also enter ~〈locality〉. We think that they have concurrently
the properties of nouns and locational words. Here we do not treat them as
conversional words with the homomorphic strategy but classify them into
nouns with * ~ 里 (inside) │ 以南 (to the south) in order to eliminate this
portion of nouns. The criteria for classifying nouns include the functions of
〈subject〉, 〈real object〉 and 〈attributive〉~, whose generalization level
is rather high, because 〈numeral and measure word〉~ and ~〈locality〉
have no internal universality. A noun is actually the residual class of substan-
tive words and so can be classified by eliminating measure words and loca-
tional words from substantive words.
Locational words can also be classified into locatives, place words and
time words; the classification can still be demonstrated with the previously
presented method. But due to space limitations here, the demonstration is
omitted and only classification criteria are listed as follows:
152

152  How to classify parts of speech

Locative: 在 (in) (〈substantive word〉~) Λ * ~ 里 (inside)│以南 (to the
south) {周围 (ambient), 以前 (before), 附近 (nearby)}
Time word:  在 (in) ~ Λ 等到 (until) ~│~ 以来 (since)│ ~ 的时候 (time
when){刚才 (just), 去年 (yesteryear), 最近 (recently)}
Place word:  在 (in) ~ Λ (* ~ 里 (inside) │以南 (to the south)│time
word〉){原地 (in place), 一旁 (aside), 民间 (folk)}

Among predicate words, the compatible degrees of 很 (very) ~ and 〈predi-

cate〉 are 0.99 and 0.19 respectively, being greatly disparate. Their gener-
alization level is diverse. According to Rule 6, 很 (very) ~ may represent a
subclass. The survey of other syntactic and morphological characteristics
shows that most reduplicated forms of double-​syllable or monosyllable pre-
dicative words that can enter 很 (very) ~ are AABB or AA (not neutral tone)
forms. Therefore, we think that 很 (very) ~, ~ 得很 (nicely), ~ 极了 (extremely)
and 很不 (quite not) ~, whose degrees of compatibility with 很 (very) ~ is
extremely high, represent a subclass of predicate words that differs from other
predicate words. The compatible degrees of 很 (very) ~ and 〈real object〉
are as low as 0.08 and 0.03 respectively. Most reduplicated forms of predicate
words that can enter ~〈real object〉 are also ABAB or AA (neutral tone)
forms. Although the two functions are equivalent (很 (very) ~ and 不 (not)
~ are equivalent; 不 (not) ~ and ~ 〈real object〉 are equivalent) according
to the transferability of equivalent functions, they are not equivalent at a
rather low generalization level. As for the co-​occurrence of 很 (very) ~ and ~
〈real object〉, for example, 很喜欢他 (like him very much), because ~〈real
object〉 is in the inner layer, they should be classified into verbs according
to the principle of “~〈real object〉 priority”. The criteria for classifying
adjectives are as follows:

Adjective: 很 (very)[不 (not)] ~ Λ (*很 (very) ~〈object〉){红 (red), 大

(big), 干净 (clean), 认真 (serious)}

Quite a large number of adjectives that can enter 很 (very) ~ can enter
~ 〈real object〉. There are two scenarios: (1) the real object that denotes
degree comparison appears only in ~〈real object〉 (comparative object)
+〈quasi-​object〉, for instance, 高他一头 (a head higher than him),
大他一岁 (one year older than him). Their degree just fits the degree of an
adjective. (2) The real object denotes a certain change on a part of a whole;
the subject and object have a relationship between part and whole, for
instance: (脸) 红了半边 ((face) gets red on its half side), (手上) 黑了一块
((hand) gets black a part), (你我)都白了头发 ((you and I) both have white
hair). At a rather low generalization level, 很 (very) ~ and ~〈real object〉
are not equivalent. But this phenomenon has its conditions and can be
properly reasoned by analogy. According to Rule 1, it should be regarded
as a syntactic phenomenon and still functions as adjective, which does not
change into a verb.
 153

How to classify parts of speech 153

Furthermore, it is difficult to classify verbs and state words primarily
because a state word belongs to a narrow type of function and does not have
its special type, whereas a verb has a great individual difference and can hardly
be separated from a state word with equivalent functions. Nevertheless, with
the double-​syllable predicate word excluded from adjectives, although most
of its reduplicated forms are ABAB, its grammatical meanings and phon-
etic forms are different. The latter half parts of reduplicated forms of 调整
(adjust), 请示 (ask), 商量 (consult), 研究 (study) are pronounced in a neu-
tral tone, and their grammatical meanings denote a small quantity, whereas
the reduplicated forms such as 雪白 (snow-​white), 通红 (red through), 矮胖
(stocky) and 干瘦 (skinny) are not pronounced in a neutral tone, and their
grammatical meanings denote a deeper degree. Most of the former can enter
不 (not) ~, 没 (no) ~, ~ 〈object〉, ~ 〈complement〉 and ~ zhuo (着)│guo
(过), whereas the latter cannot. Thus we classify the two types of predicate
words as follows:

Verb:  不 (not) ~│没 (no) ~│~〈object〉│~〈complem

ent〉│〈bound complement〉│(~ zhuo (着)│guo
(过)│(〈predicate〉│〈adverbial〉~) Λ *〈subject〉Λ (* 很
(very)[不 (not)] ~│(很 (very) ~〈object〉){吃 (eat), 看 (see), 研究
(study), 应该 (should)}
State word: 〈compound complement〉│(〈predicate〉 Λ (*〈subject〉│
〈object〉) Λ (*〈verb│ adjective〉){通红 (red through), 花白 (gray),
酷热 (sweltering hot)}

Most state words are the quantitative forms of adjectives. On these grounds,
a good many people think that they should be classified into adjectives. The
change from adjective into its corresponding quantitative form is a form of
word formation rather than a structural form, not being a syntactic phenom-
enon at all. Furthermore, a state word has prominently different syntactic
characteristics from an adjective; in particular it no longer has fundamental
functions of an adjective such as 很 (very) [不 (not)]~. Therefore, they should
be separated into another type of words.
The above discussion shows that the compatible degree of at least one
of the major functions determined as equivalent reaches the extremely high
critical value (c>=0.9), while the compatible degree among the functions
that cannot be determined as equivalent generally cannot reach this crit-
ical value. Put another way, all the functions whose compatible degree is
more than 0.9 are equivalent, with the only exception being 〈subject or
object〉&〈attributive〉. This shows the role of a compatible degree played
in determining an equivalent function.
Equivalent function clusters plus other part-​of-​speech exclusive restrictions
are the criteria for classifying parts of speech. The classification criteria are
both disjunctive and conjunctive. Disjunctive criteria refer to classification
criteria that play no distinctive role within the same equivalent function
154

154  How to classify parts of speech

cluster. Conjunctive criteria refer to functions that are used to exclude other
parts of speech.
We have discussed 13 out of 15 notional words, excluding pronouns and
onomatopoeia. A  pronoun is actually a type of word classified from the
perspective of discourse function and is not, strictly speaking, at the same
level as other parts of speech. From the perspective of a part of speech in
its usual sense, it should be classified into notional words, predicate words,
modifiers and other even smaller basic classes. Let us stop the discussion here.
Onomatopoeia is a special part of speech whose status in a part-​of-​speech
system we do not know sufficiently. We are not clear how to demonstrate it
with the methods presented in this book.

6.4  Brief summary

What we did in the above is actually to determine the correspondence between
grammatical function and part of speech. The determination of what gram-
matical function corresponds to what part of speech largely accomplishes the
part-​of-​speech classification  task.
The above analysis shows that the pairs between 不 (not) ~ and 〈attribu-
tive〉, 不 (not) ~ and 〈noun〉~, 很 (very) ~ and 〈attributive〉, 很 (very)
~ and 〈adverbial〉, and 〈subject or object〉 and 〈attributive〉 are
heterovalent functions and represent different parts of speech. Therefore,
words such as 研究 (study) and 调查 (investigate) that can be modified by 不
(not), directly modified by nouns and function as attributives actually have
the properties of verbs, nouns and distinctive words. Words such as 干净
(clean) and 新 (new) that can be modified by 很 (very) and directly function
as attributives actually have concurrently the properties of adjectives and dis-
tinctive words. Words such as 迅速 (rapidly) and 妥善 (appropriately) that
can be modified by 很 (very) and directly function as adverbials actually have
concurrently the properties of adjectives and adverbs. Words such as 木头
(wood) and 质量 (quality) that can function as subjects or objects and dir-
ectly function as attributives actually have concurrently the properties of
nouns and distinctive words. We do not just treat them as conversional words
according to the priority homomorphic strategy (see Chapter 7). Moreover,
because 〈predicate〉 and 〈complement〉 are equivalent functions, there
is no need to classify the so-​called “only complement word” that can only
function as a complement. Because words in different grammatical positions
may belong to different parts of speech, the statement by Li Jinxi (1924) that
“the part of speech of a word relies on a sentence” is reasonable to a cer-
tain extent. It is a pity that he did not establish the method for determining
what grammatical position expresses different parts of speech and what gram-
matical position expresses identical ones. The statement goes to extremes in
saying that “words have no definite parts of speech”, and that “there is no
part of speech without a sentence”.
 155

How to classify parts of speech 155

A part of speech is essentially the large category of expressional functions
of a word and the basic type that combines its semantic type with its gram-
matical function rather than its distributional type. Therefore, it is impossible
to classify parts of speech simply according to distributional similarity. But
distribution is the extrinsic manifestation of the properties of a part of speech
and can be inferred by determining the correspondence between distribution
and a part of speech. The correspondence between distribution and a part of
speech can be determined according to the compatibility with grammatical
functions and relevant rules, thus identifying the criteria for classifying parts
of speech of all words. However, after all, distribution is not the essence of a
part of speech; many factors influence the distribution of a word, and there
is a limitation to using distributional characteristics as criteria for classifying
parts of speech. For example, there are two kinds of words that can only
function as predicates. The first kind includes 交加 (occur simultaneously),
倍增 (multiply), 参半 (half) and 奇缺 (rare), and usually belongs to verbs; the
second kind includes 旖旎 (graceful), 婆娑 (dancing), 皑皑 (snow-​white) and
卓然 (outstanding), and generally belongs to state words. But it is difficult
to classify them functionally. This notwithstanding, the use of distributional
characteristics of a word as the basis can still classify the vast majority of
words into their corresponding parts of speech, and it is difficult to determine
the parts of speech of only a minority of words.
In the previous discussion, we demonstrated the internal classification of
the parts of speech of Chinese notional words. We may also demonstrate
the external classification, for example, from the typological perspective. For
discussions on this, see McCawley (1992) and Guo Rui (2001).

Notes
1 ~ denotes the position in which a subject appears; │ denotes disjunctive relation; Λ
denotes a conjunctive relation; * means that a certain function does not exist; the
word inside〈〉is a part of speech or syntactic constituent.
2 See Chapter 2.4 in Volume 2 for detailed discussion.
3 According to the above-​mentioned assumption, only functions that have extremely
high compatible degrees belong to equivalent functions. To be safe, here we use
the criterion that a heterovalent function should have a compatible degree of less
than 0.5.
4 Of course, the use of what substantive word as a vehicle of metaphor is related to
the typicalness of the substance and the social mentality and culture.
5 See Chapter 2 in Volume 2 for the correlation between word frequency and gram-
matical function
6 Many scholars have noticed a positive correlation between changes in lan-
guage form and in meaning and word frequency. This can be explained in this
way:  changes always occur in use. The more frequently a linguistic element is
used, the more possibly it changes; the less frequently it is used, the less possibly it
changes. Elements that have the least possibility to change have not been in use.
156

156  How to classify parts of speech

7 The multiple entries of polysemes and homographs are calculated as one entry.
Therefore, the statistical number of words (38110) is less than the total number of
words (43330).
8 The book calculated the compatible degrees among attributive, adverbial and
other grammatical functions but did not calculate attributives and adverbials as
modifiers. But only a few words can function as adverbials; an attributive basically
represent a modifier.
9 We regard the whole function of a phrase as the function of its head element.
For example, 苹果 (apple) in 大苹果好吃 (A big apple is delicious) functions as
a subject; the word 去 (go) in 我马上去图书馆 (I will go to library immediately)
functions as a predicate.
10 The use of these criteria should eliminate omission and transfer reference. For
example, there is omission in 我不 (I do not); we should not classify 不 (not) into
predicate words just because it functions as a predicate. 许多 (many) in 许多都坏了
(Many are ruined) is used as a transfer reference; we should not classify it into
notional words just because it functions as a subject.
11 The segment such as 一百二十三 (one hundred and twenty-​three) is a phrase.
12 ~ X de (的) 〈noun〉 refers to the de (的) construction and the construction that
the constituent which takes de2 (的) functions as an attributive.
13 [] denotes the constituent that may not necessarily appear.
 157

7 
Conversional words and nominalization

7.1  Conversional words and homographs of different parts of speech

A word that belongs to several parts of speech is a conversional word.
The treatment of conversional words and homographs is interrelated. The
following three phenomena are related to conversional words:

A. The same meaning (strictly speaking, the same generalization word)

concurrently belongs to several parts of speech. For example, 长期
(long-​term), 真正 (genuine) and 临时 (temporary) can only function as
attributives and adverbials, concurrently belonging to distinctive words
and adverbs. For another instance, 小时 (hour) can be used as 一个小时
(one hour of) and 一小时 (one hour), therefore being treated as a noun
that functions as a measure word concurrently.
B. A couple of associated meanings belong to different parts of speech.
For example, 典型 (representative) in 一个典型 (one representative) and
很典型 (very representative) has different meanings: the former is a noun,
whereas the latter is an adjective. For another example, 通知 (notifica-
tion) in 通知开会 (notify a meeting) and 写一个通知 (write a notice) has
different meanings: the former is a verb, whereas the latter is a noun.
C. Words that have no associated meanings but the same sounds and word
forms belong to different parts of speech. For example, 会 (hui, meaning
“can”) in 会游泳 (can swim) and 会 (hui, meaning “meeting”) in 开一个会
(hold a meeting) are homophones and homographs: the former belongs
to verbs, whereas the latter belongs to nouns. For another instance, 制服
(zhifu, meaning “uniform”) in 一套制服 (a suit of uniform) and 制服
(zhifu, meaning “conquer”) in 制服了对手 (conquer one’s opponent) are
homonymous: the former is a noun; the latter is a verb.

A should belong to conversional words, which we call homonymous

conversional words. C should belong to homographs in different parts of
speech, and this is not controversial. There are different opinions on how to
deal with B.  Strictly speaking, a conversional word refers to the same gen-
eralization word that has concurrently different parts of speech; therefore,
158

158  Conversional words and nominalization

only A belongs to conversional words. But lexicology regards the associated
meaning as the different meanings of the same word, and a dictionary treats
it as one entry. Our word list also uses such a unit as one entry, and this
is comparatively economical (we call it a lexical word, and the unit induced
with the identity principle is called a grammatical word). Only A  belongs
to conversional words in theory, but considering that a dictionary treats
associated meaning as one entry and that a computer cannot distinguish
A from B in Chinese, we flexibly treat B as a conversional word and call it a
heteromorphic conversional word. In order to distinguish between the two
kinds of conversional words, we call a homonymous conversional word a
conversional word in its narrow sense, which, together with a heteromorphic
conversional word, is called a conversional word in its broad sense.
This shows that the conversional word in its narrow sense has something to
do with the determination of lexical words.

7.2  What is a real conversional word?

Whether a word belongs to conversional words or not depends on the
following three questions:

(1) Whether the word in two positions belongs to the same generalization
word (the identity problem).
(2) Whether a word belongs to several parts of speech or their

multifunctionality (whether they are transformed at the lexical level).
(3) What classification strategy is used.

If both the lexical meaning and the part of speech are different, then
different generalization words actually belong to different parts of speech.
Of course, they should be regarded as heteromorphic conversional words.
Therefore, Question (1) is not worth discussing. Provided that different gener-
alization words have different parts of speech, they should all be regarded as
conversional words. For example, 锁 (lock), 领导 (leader), 科学 (science) and
死 (death) should all be treated as conversional words. After determining the
identity, we should determine again whether the various functions of a word
differ in parts of speech. We do so according to the method for determining
equivalent functions presented in Chapter  6. This method proves that an
adjective that takes zhuo (着), le (了) and 过 (guo) does not function concur-
rently as a verb. It also proves that the transformation of parts of speech at the
syntactic level is not their conversion. For example, verbs such as 去 (going) in
去是应该的 (going is obligatory), which function as subjects, do not function
concurrently as nouns because they still have the basic functions of a verb,
such as taking objects and adverbials. In fact, a large number of verbs can
function as subjects. Moreover, we also prove that words such as 研究 (study)
and 调查 (investigate) that can be modified by 不 (not), be directly modified
by nouns and function as attributives actually have the properties of verbs,
 159

Conversional words and nominalization 159

nouns and distinctive words. Words such as 干净 (clean) and 新 (new) that
can be modified by 很 (very) and directly function as attributives actually have
concurrently the properties of an adjective and a distinctive word. Words such
as 迅速 (rapidly) and 妥善 (appropriately) that can be modified by 很 (very)
and directly function as adverbials actually have concurrently the properties
of an adjective and adverb. This problem, which has been solved previously,
will not be discussed here. We need to discuss classification strategies.

7.3  Classification strategies and conversional words

7.3.1  Classification strategies

The same generalization word that has the properties of several parts of
speech is not necessarily treated as a conversional word in its narrow sense.
Whether or not it is treated as a conversional word has something to do with
the strategy for classifying parts of speech. We first look at what classification
strategies are available.
As shown in Figure 7.1, the properties of two parts of speech can partially
overlap lexically. The words in Sphere a have the properties of words in Part
of Speech A; the words in Sphere b have the properties of words in Part of
Speech B; the words in Sphere c have the properties of words in both Parts
of Speech A and B. In theory, the following four classification strategies are
available:

I. The homogeneous strategy:  Starting from the properties of a part of

speech, all the properties of the part of speech of a generalization word
are treated equally to create a one-​ to-​
one correspondence between
parts of speech thus classified and their properties. If a generalization
word has the properties of several parts of speech, then we treat it as a
conversional word. As shown in Figure 7.1, there are two classes of parts

Properties of Part Properties of Part

of Speech A of Speech B

a c b

Figure 7.1 Properties of parts of speech

160

160  Conversional words and nominalization

of speech: Part of Speech A (A, C) and Part of Speech B (B, C), where
Part of Speech C functions concurrently as Parts of Speech A and B.
The advantage of the homogeneous strategy is that the properties of a
part of speech have one-​to-​one correspondence; its disadvantage is that
there are too many conversional words in their narrow sense.
II. The homomorphic strategy:  Starting from a generalization word, the
properties of all its parts of speech are treated equally. As long as the
properties of its parts of speech are different, they are classified into
different parts of speech. As shown in Figure 7.1, a generalization word
is classified into three parts of speech: A, B, C.
The advantage of the homomorphic strategy is that there are no
conversional words in their narrow sense; its disadvantage is that there
are too large numbers of parts of speech, and that there is no one-​to-​one
correspondence between a part of speech and its properties.
I II. The priority homomorphic strategy: Starting from a generalization word,
we give priority to the properties of a certain part of speech and do not
treat as a conversional word a word that has the properties of several
parts of speech; instead, we treat it as a priority part of speech, thus
creating a correspondence between a part of speech and the properties
of a priority part of speech. As shown in Figure  7.1, parts of speech
are classified into two classes: A and B. Because the order of priority is
different, Class C is grouped into the following two cases:
(1) The order of priority is:  A>B. According to this order of priority,
Parts of Speech A and C are grouped into Class A; Part of Speech B
is grouped into Class B.
(2) The order of priority is:  B>A. According to this order of priority,
Part of Speech A is grouped into Class A; Parts of Speech B and C
are grouped into Class B.
In the past, it was often said that “without a change in lexical meaning,
there is no change in a part of speech”, actually meaning the use of the
homomorphic strategy.
The advantage of the priority homomorphic strategy is that there are
no conversional words, and that there are not a large number of parts of
speech. Its disadvantage is that there is no entire one-​to-​one correspond-
ence between a part of speech and its properties.
I V. The consolidation strategy: A word that has the properties of two parts
of speech is consolidated into the one that has one part of speech. As
shown in Figure 7.1, a word belonging to Parts of Speech A, B and C are
consolidated into the one belonging to one part of speech.
The advantage of the consolidation strategy is that there are no
conversional words in their narrow sense, and that the number of parts of
speech is small, but its disadvantage is that there are too large numbers of
parts of speech, and that there is no one-​to-​one correspondence between
part of speech and its properties.
How should classification strategies be selected?
 161

Conversional words and nominalization 161

7.3.2  Selecting classification strategies

There are no right or wrong classification strategies but only good or bad ones.
Theoretically speaking, all the above-​mentioned four classification strategies
are feasible. What classification strategy is to be selected should be decided by
taking into consideration the whole grammatical system and the purpose for
selecting classification strategies. For example, starting from a practical pur-
pose, the homogeneous strategy should be used for the part-​of-​speech system
for information processing. Namely, we classify into verb-​noun conversional
words the lexical items that concurrently function as verbs and nouns. The
part-​of-​speech system in this book is not for a specific purpose, and we con-
sider the following two factors in selecting classification strategies:

(1) The simplicity of a part of speech: there are as small a total number of

parts of speech as possible and as small a total number of conversional
words in its narrow sense as possible.

To reduce the number of parts of speech, it is necessary to use the homo-

geneous strategy, the priority homomorphic strategy or the consolidation
strategy. To reduce the number of conversional words, it is necessary to use
the homomorphic strategy, the priority homomorphic strategy or the consoli-
dation strategy.
Why should the total number of parts of speech be reduced? The purpose
for classifying parts of speech is first and foremost to meet the needs of people
to learn and analyze a language. Too many parts of speech may increase the
memory burden and make people unable to have a sound command of
the whole part-​of-​speech system. If the homomorphic strategy is used for all
the parts of speech, then words with the properties of verbs, nouns, distinctive
words and adverbs should be classified into separate parts of speech. There
will be a total of hundreds of parts of speech if classified this way. Such a
huge number of parts of speech are obviously beyond the ken of an ordinary
person. Because the homomorphic strategy may increase the number of parts
of speech, we should avoid it.
Why should the number of conversional words be reduced? The reasons
are twofold. First, if a word belonging to two parts of speech is treated as
a conversional word, in formulating syntactic rules, we first and foremost
need to decide what part of speech the word in a sentence belongs to. Too
many conversional words may increase the burden of deciding parts of
speech. Second, if most words belonging to two or more parts of speech are
conversional words belonging to two or more parts of speech, then words
belonging to two or more parts of speech have not yet been differentiated,
and there is no need to classify them into two or more parts of speech.

(2) The simplicity of the syntactic rule: the grammatical functions of different

words in the same part of speech should be as few as possible.
162

162  Conversional words and nominalization

To make this happen, the use of the homogeneous strategy is needed to
treat the word that has concurrently different properties. For example, if the
homogeneous strategy is used to treat a word that has concurrently the prop-
erties of verbs, nouns and modification words as their concurrent parts of
speech, then the following rules may be obtained when constructing syntactic
rules with the part-​of-​speech system:

Noun + verb→subject-​predicate
Verb + noun → predicate-​object
Verb + verb → combination, predicate-​object, predicate-​complement,
link-​verb predicate

However, if the priority homomorphic strategy is used, then the same part-​
of-​speech sequence may produce various results and increase the ambiguity
of the syntactic analysis and its level of difficulty as a result. For example:

Verb + verb → subject-​predicate, consolidation, attributive-​headword,

adverbial headword
Verb + noun →subject-​predicate, predicate-​object, consolidation,
attributive-​headword
Verb + verb →subject-​predicate, consolidation, predicate-​object,
attributive-​headword, predicate-​complement, link-​verb-​predicate,
attributive-​headword, adverbial headword

The more widely the priority homomorphic strategy is used, the more
ambiguous is the sentence thus made, and the less efficiently is the sentence
described and analyzed. It may even make the description of syntactic rules
with parts of speech meaningless. Therefore, use of the priority homomorphic
strategy should be limited.
There is no correspondence in meaning between Chinese parts of speech
and syntactic constituents, and a sentence described with parts of speech
may be certainly too ambiguous. You may wonder why the use of the priority
homomorphic strategy should still be limited. In fact, the lack of a one-​to-​one
correspondence between Chinese parts of speech and syntactic constituents
is largely due to the fact that the Chinese grammatical system uses the pri-
ority homomorphic strategy to classify parts of speech. If the priority homo-
morphic strategy were not used to classify parts of speech, then there might
be a very strict correspondence between Chinese parts of speech and syntactic
constituents. Therefore, lacking a one-​to-​one correspondence between parts
of speech and syntactic constituents cannot explain the reason why the use of
the priority homomorphic strategy should be limited.
In addition, under special circumstances, the following should be
considered:

(3) Mental acceptability: when one part of speech contains more than one
property, it should be in agreement with the human mentality. If the
 163

Conversional words and nominalization 163

priority homomorphic strategy is used, then the status of two properties
of the same part of speech should be obviously higher or lower, and the
first priority is given to the property that has a higher syntactic status.1
For example, words such as 研究 (study) and 调查 (investigate) that are
used concurrently as the two parts of speech of verb and noun should
be classified into verb instead of noun with the priority homomorphic
strategy, which gives priority to verbs because the syntactic status of a
verb is higher than that of a noun.

Factors (1) and (2) are the most basic; Factor (3) is just supplementary. But
Factors (1) and (2) are contradictory. The simplicity of a part of speech tends
to ruin the simplicity of the syntactic rule. The simplicity of the syntactic
rule tends to ruin the simplicity of part of speech. Therefore, in selecting the
strategy for classifying parts of speech, we should consider both factors, redu-
cing their total cost to the lowest degree. Under any circumstances, the homo-
morphic strategy maximizes the total cost of the two factors (the number
of parts of speech increases; there is no correspondence between parts of
speech and grammatical properties). Therefore, we will not use it. We will only
choose the homogeneous strategy, the priority homomorphic strategy or the
consolidation strategy.
In fact, different strategies for classifying parts of speech cause a good
many controversies related to Chinese parts of speech. Strategies are not right
or wrong but only good or bad. Which strategy for classifying parts of speech
is the best depends on the specific circumstances.
Then, under what circumstance should the homogeneous strategy or the
priority homomorphic strategy be used? Our principles are as follows:

1. If overlapping members take the majority, then we should use the con-
solidation strategy.
2. If there are comparatively large numbers of words that have two parts
of speech or their conditions are identified (namely they can be con-
trolled by rules), then we use the priority homomorphic strategy; if not,
we use the homogeneous strategy. We do so because if the homogeneous
strategy is used for a large number of words that have two or more
parts of speech, a massive number of conversional words may appear.
Although the simplicity of the syntactic rule pays off, the part of speech
is complicated. Before the syntactic treatment, a massive number of
conversional words must be identified; once erroneous identification
of conversional words occurs, erroneous syntactic treatment may also
ensue. Therefore, when there is a large number of words that function
as two parts of speech, we would be better off using the priority homo-
morphic strategy.
3. If the priority homomorphic strategy is used for a small number of
words that function as two parts of speech, although the simplicity of
the part of speech pays off, that of the syntactic rule requires too heavy
a cost because more syntactic rules are needed to deal with the use of a
164

164  Conversional words and nominalization

conversional word. The use of the homogeneous strategy may simplify
syntactic rules, while the simplicity of a part of speech does not require
a heavy cost. Therefore, with the total cost considered, we should use
the homogeneous strategy. However, if a word that functions as two or
more parts of speech can be controlled by rules, then the priority homo-
morphic strategy is desirable.

In other words, the selection of which classification strategy to use has

something to do with the number of words that have concurrently the proper-
ties of two parts of speech. This is what Zhu Dexi (1985a) meant: there should
not be too many conversional words. We call this principle the systematic
principle. What are the bases for the systematic principle? If a word over-
lapped by two functions belongs to or basically belongs to the same batch,
then this means that the two functions do not differentiate in this language,
and there is no need to classify the word into different parts of speech. There
are no absolute criteria for deciding what proportion of words that have con-
currently the properties of two parts of speech is needed to select different
classification strategies, but some criteria have a greater tendency than others.
In the following, we shall discuss the specific criteria for selecting different
classification strategies.

7.3.2.1  Criteria for selecting the consolidation strategy

To consolidate words that have the properties of two parts of speech into
those that have the properties of one part of speech, the number of words
whose parts of speech overlap should not be less than 90% of the total number
of words that combine the two parts of speech. In other words, only when
the vast majority of words concurrently have the properties of two parts of
speech is there no need to classify them into two parts of speech.
The approximately inclusive relation can be regarded as a special case of
the consolidation strategy:
As shown in Figure 7.2, if the vast majority of words that have the prop-
erties of Part of Speech B have the properties of Part of Speech A, but only
a few words that have the properties of Part of Speech A have the properties
of Part of Speech B, then we regard Part of Speech B as a subclass of Part of
Speech A.

7.3.2.2  Criteria for selecting the priority homomorphic strategy

The priority homomorphic strategy is used only when two parts of speech
have advantages or disadvantages. The dominant hierarchy of parts of speech
is as follows:

Predicate word property > substantive word property > modification

word property
 165

Conversional words and nominalization 165

Properties of Properties of
Part of speech A Part of speech B

a c b

Figure 7.2 The approximately complete overlapping and approximately inclusive

relations between two parts of speech

Properties of Properties of
Part of speech A Part of speech B

a c b

Figure 7.3 The dominant hierarchy of parts of speech (Guo Rui, 1997)

The basis for this hierarchy is combination dominance. The combination of

constituents on the left-​hand side with those on the right-​hand side produces
a combined unit whose properties are identical with those on the left-​hand
side. Namely, the constituents on the left-​hand side are the core, for example:

Predicate word property + substantive word property→predicate word

property
Predicate word property + modification word property→predicate word
property
Substantive word property + modification word property→substantive
word property

When the overlapping part accounts for 40% to 80% of the dominant part
of speech, it should be included in the dominant part of speech, and only
words that are in a weak part of speech should be classified into another inde-
pendent part of speech.
When the overlapping part takes the majority of the weak part of speech,
it should not be included in the weak part of speech.
166

166  Conversional words and nominalization

Part of Speech A Part of Speech B

a c b

Figure 7.4 The overlapping part accounts for 40% to 80% of the dominant part of
speech (Part of Speech A)

Part of Speech A Part of Speech B

a c b

Figure 7.5 The overlapping part takes the minority (<20%)

7.3.2.3  Criteria for selecting the homogeneous strategy

If the overlapping part of two parts of speech has a small number of words
that belong to one of them (<20%), then it is desirable to use the homoge-
neous strategy to classify the overlapping part into conversional words.
Even though the overlapping part takes the majority (51%–​90%) of words
whose parts of speech are weak, it is desirable to use the homogeneous
strategy instead of the priority homomorphic strategy. But if the priority
homomorphic strategy has already been used to classify the overlapping part
of the dominant parts of speech because it takes a high proportion, then the
homogeneous strategy should not be used.
The above-​mentioned quantitative criteria refer to general conditions, but
in actual operation, the absolute number of the overlapping parts should be
taken into consideration. If the absolute number of words that have concur-
rently two parts of speech is rather big, then they should not be classified
into conversional words with the homogeneous strategy. In addition, if the
same batch of words has concurrently three or more parts of speech, which
classification strategy is to be used must be considered in all respects. Just
because of these reasons, in the above-​mentioned criteria, we leave the crit-
ical proportions 20%–​40% and 80%–​90% for the two classification strategies.
The critical proportions must be dealt with flexibly according to the absolute
 167

Conversional words and nominalization 167

number of overlapping parts of speech and the number of conversional words.
A detailed discussion is presented in the relevant subsections in Chapter 1 of
Volume 2.

7.3.3  Classification strategies for the major Chinese parts of speech

In the past, the Chinese grammar circle tended to use the priority homo-
morphic strategy. For example, they regarded predicates, attributives and
adverbials as the functions of an adjective. Following the above-​mentioned
new criteria, this treatment should be reconsidered.
The following sections will discuss the selection of classification strategies
for the major overlapping parts of speech.
First, let us look at the major overlapping parts of speech:

7.3.3.1  Classifying predicate words and substantive words

When a predicate word overlaps with a substantive word, Zhu Dexi (1985a)
proposed three criteria for determining the predicate word that has concur-
rently the properties of a substantive word.

1. Functioning as the object of a verb that can take a quasi-​predicate as its

object, for example, 有 (have), 作 (do), 加以 (conduct), 进行 (carry out)
and so on: 有研究 (have study), 有苦难 (have hardship), 进行研究 (con-
duct study).
2. Being directly modified by other nouns. For example, 历史研究 (history
study), 经济困难 (economy difficulty)
3. Directly modifying other nouns. For example:  研究方向 (research
interests). This criterion cannot identify a nominal adjective.

In fact, the third criterion shows the properties of not a substantive word but
a modification word. Therefore, the former two criteria are enough to identify
nominal verbs and nominal adjectives. Actually, there are only the former two
criteria for nominal verbs in Zhu Dexi’s Handouts on Grammar (1982b).
In Chinese, the proportion between a word that has concurrently the prop-
erties of a predicate word and one that has concurrently the properties of a
substantive word takes only 19% of the predicate words and 8% of the sub-
stantive words, far below the criteria for the consolidation strategy, which
should not be used. Neither should the predicate word be regarded as a sub-
class of substantive words. The overlapping proportion between predicate
words and substantive words should be treated as their conversional words
with the homogeneous strategy.
The above data are obtained with the joint investigation of verbs, adjectives
and state words within the predicate word. If the verbs, adjectives and state
words are investigated separately, then:
 newgenrtpdf
168
Table 7.1 The major overlapping parts of speech

Multi­ Overlapping Proportion Proportion Single Proportion Functional words Total

functional proportion that overlaps that overlaps functional
words previous words latter words word

predicate-​ substantive 2490 6.15% 19.08% 8.33% 37968 93.85% predicate+ 40458
substantive
predicate-​ distinctive 3895 12.13% 29.85% 16.97% 28207 87.87% predicate+ 32102
distinctive
predicate-​ adverb 469 3.49% 3.59% 55.70% 12954 96.51% predicate+ 13423
adverb
substantive-​distinctive 20081 63.18% 67.17% 87.51% 11702 36.82% substantive+ 31783
distinctive
distinctive-​adverb 106 0.88% 0.35% 12.59% 11965 99.12% distinctive+ 12071
adverb
distinctive-​adverb 284 1.21% 1.24% 33.73% 23221 98.79% Distinctive 23505
modification
+adverb
 169

Conversional words and nominalization 169

Multifunctional word Number Overlapping Percentage of Percentage of

proportion verbs nouns
Verb-​noun 2381 6.31% 23.12% 8.0%

Words that have concurrently the properties of verbs and nouns still take
a rather low proportion among verbs, existing in the critical range between
the priority homomorphic strategy and the homogeneous strategy. It is not at
all realistic to use the consolidation strategy to treat Chinese verbs as a sub-
class of nouns; instead, we can use the homogeneous strategy to treat them
as conversional words between verbs and nouns, for example, 研究 (research),
生产 (production), 管理 (management) and 学习 (study).
In particular, considering that some words are actually used as nouns far
more often than as verbs, it is unreasonable to classify them into verbs. For
example, the following table gives the data randomly retrieved from two pages
of language data in the Peking University CCL Corpus:

Properties of verb Properties of noun

manage 23 55
sleep 2 49

Under this circumstance, obviously it is more reasonable to treat them as

conversional words between verbs and nouns.
However, considering that there are a rather big absolute number of words
that have concurrently the properties of verbs and nouns, it is acceptable to
treat them as verbs with the priority homomorphic strategy. This book still
treats them as verbs (nominal verbs) in accordance with Zhu Dexi (1982,
1984b, 1985a, b).
The properties of adjectives and nouns also overlap. Out of 2,355 adjectives,
109 are nominal, accounting for 5% of the total, for example, 平衡 (balanced)
and 健康 (healthy). Even though the number is not huge, if adjectives are
the only consideration, we can well use the homogeneous strategy to classify
them into adjectives concurrently functioning as verbs. But because the pri-
ority homomorphic strategy is used for verbs that function as predicates, it
is desirable to use the same strategy, not classifying the adjectives into those
concurrently functioning as nouns.

7.3.3.2  Classifying substantive words and distinctive words

Too great an overlapping proportion (taking 67% of substantive words)
between the properties of substantive words and modification words (dis-
tinctive words) indicates that the degree of differentiation between modi-
fication words and substantive words is very low. It is advisable to use the
170

170  Conversional words and nominalization

priority homomorphic strategy to treat the overlapping part as a substan-
tive word (noun) instead of a conversional word. For example, words such as
木头 (wood), 质量 (quality) and 语言 (language) that can function as subjects,
objects or attributives should be treated as nouns instead of conversional
words between nouns and distinctive words.
A very high proportion, 87.5%, of words that have the properties of dis-
tinctive words have concurrently the properties of nouns. But compared with
the properties of nouns, the properties of distinctive words are weak; there-
fore, it is not advisable to use the priority homomorphic strategy to treat the
overlapping parts as distinctive words.

7.3.3.3  Classifying predicate words and distinctive words

The overlapping proportions between the properties of predicate words and
distinctive words are not high (6%, 19%, 8%), and the homogeneous strategy
should be used to treat them as conversional words between predicate words
and distinctive words.
If we investigate separately the verbs and adjectives among the predicate
words, the data are as follows:

Multifunctional word Number Overlapping Proportion to Proportion to

proportion predicate word distinctive word
property property

verb-​distinctive word 3189 29.6% 31.0% 87.4%

adjective (predicate 690 24.5% 29.3% 60.1%
word property)-​
distinctive word

The verbs and adjectives do not take a high total proportion or a high pro-
portion to dominant parts of speech, being in the critical range between the
homogeneous strategy and the priority homomorphic strategy. The homoge-
neous strategy should be used to treat them as conversional words between
verbs and distinctive words, for example, 成立 (found), 到达 (reach), 学习
(study) and 研究 (research), and conversional words between adjective and
distinctive words, for example, 新 (new), 红 (red), 干净 (clean) and 优秀 (excel-
lent). However, considering that they have a rather big absolute number, and
particularly that there are 285 adjectives among the 468 that have the highest
word frequency, this book still uses the priority homomorphic strategy to
treat them as adjectives.
The very high proportion of the overlapping parts to the weak parts
of speech indicates that the specialization degree of a distinctive word is
rather low.
 171

Conversional words and nominalization 171

7.3.3.4  Classifying predicate words and adverbs
The total overlapping proportion between properties of predicate words and
adverbs and the proportion to dominant parts of speech are very low. We
therefore use the homogeneous strategy to treat them as conversional words.
The very high proportion to properties of adverbs shows that their differenti-
ation degree is rather low.
With the following table, let us look at verbs and adjectives separately:

Multifunctional word Number Overlapping Proportion to Proportion to

proportion verb property adverb property

verb-​adverb 131 1.2% 1.3% 39.8%

adjective(predicate 277 10.3% 11.8% 45.7%
word property)-
adverb

The overlapping proportion between predicate words and adverbs and

their proportion to dominant parts of speech are very low. Ideally, we should
treat them as conversional words. In other words, we should treat words such
as 区别 (distinguish), 继续 (continue) and 重复 (repeat) as conversional words
between verbs and adverbs, while treating 认真 (serious), 努力 (laborious),
迅速 (rapid) and 紧急 (emergent) as conversional words between adjectives
and adverbs. However, considering that there are 172 adjectives among the
468 that have the highest frequency, taking as high a proportion as 37%, and
that words such as 认真 (serious) and 紧急 (emergent) have concurrently
the properties of distinctive words, this book treats them as adjectives with
the priority homomorphic strategy. To keep the classification strategies con-
sistent, we still use the priority homomorphic strategy to treat words that
have concurrently the properties of adjectives and adverbs as adjectives and
to treat only the overlapping parts of verbs and adverbs as conversional words
between verbs and adverbs.
The very high proportion of the overlapping parts to the weak parts of
speech indicates that the specialization degree of an adverb is not high.

7.3.3.5  Classifying noun and adverb

Multifunctional number Overlapping Proportion to Proportion to

word proportion noun property adverb property
noun-​adverb 345 2.4% 1.2% 41.0%
172

172  Conversional words and nominalization

The proportion of the overlapping part to the total number of words
belonging to nouns and adverbs and its proportion to the dominant parts
of speech are very low; therefore, the overlapping part should be treated as
conversional words. In other words, words such as 系统 (system), 正面 (front)
and 礼貌 (courtesy) should be treated as conversional words between nouns
and adverbs.

7.3.3.6  Classifying distinctive words and adverbs

Multifunctional number Overlap Proportion to Proportion to predicate

word proportion distinctive word word property
property
Distinctive 284 1.2% 1.2% 33.7%
word-​adverb

The proportion of the overlapping part to the total number of words belonging
to distinctive words and adverbs and its proportion to distinctive words are
very low. It is, therefore, advisable to treat them as conversional words. The
proportion to properties of adverbs is on the high side but does not reach
the classification criteria for the priority homomorphic strategy. Moreover,
because a distinctive word and an adverb both belong to modification words
and have no superior or inferior distinction between parts of speech, thus
lacking the preconditions for using the priority homomorphic strategy, it is
advisable to use the homogeneous strategy. In other words, words such as
共同 (common), 临时 (temporary) and 长期 (long-​term) that can function as
attributives and adverbials but cannot function as other constituents should
be treated as conversional words between distinctive words and adverbs.

7.3.3.7  Nouns that function concurrently as measure words

There are mainly two cases in which some nouns can function as measure
words have two: (1) functioning as capacity measure words such as 碗 (bowl),
桶 (barrel), 车 (carload) and 杯 (cup); (2) functioning as action measure words
such as 鞭子 (whip), 刀 (knife) and 枪 (gun). Only 109 out of 27,408 nouns
can be used in this way, taking just 0.4%. The number is very small, and the
conditions for this use can be identified. The former is limited to nouns that
denote objects that are often specially used as containers, whereas the latter
is limited to nouns that denote objects that are often specially used as a tool.
Because these nouns can be controlled with rules, the use of the priority
homomorphic strategy does not incur a loss in terms of syntactic rules but
pays off in terms of simplicity of part of speech. Consequently, we use the
priority homomorphic strategy and do not treat them as nouns functioning
concurrently as measure words.
 173

Conversional words and nominalization 173

The above-​ mentioned discussions are carried out against the back-
ground that there are no specific syntactic systems. In actual operations,
with the exception that it is not suitable to use the homomorphic strategy
because it brings about too huge a number of parts of speech, the homo-
geneous strategy, the priority homomorphic strategy and the consolidation
strategy are possible choices. Besides the above-​mentioned principles, the
selection of which strategy also depends on a matchable syntactic system.
For the “lexicon-​syntactic structure” in the generative grammar system, it is
desirable to use the homogeneous strategy. For example, in the case as shown
in Figure  7.6 below, it is possible to treat 顾虑 (worry) and 吩咐 (instruct)
(functioning as the objects of 有 (have)) as verbs that function concurrently as
nouns. Using the same classification strategy, 木头 (wood) and 质量 (quality)
(directly functioning as attributives) should be treated as nouns that function
concurrently as distinctive words; 成立 (establish) and 作废 (cancel) (directly
functioning as attributives) should be treated as verbs that function concur-
rently as distinctive words; 调查 (investigate) and 研究 (study) (functioning as
the objects of 有 (have) and directly as attributives) should be treated as verbs
that function concurrently as nouns and distinctive words. In addition, 干净
(clean) and 新 (new) (directly functioning as attributives) should be treated as
adjectives that function concurrently as distinctive words; 危险 (danger) and
温暖 (warm) (functioning as the object of a substantive-​object verb, directly
as attributives) should be treated as adjectives that function concurrently as
nouns and distinctive words; 迅速(rapidly) and 妥善(appropriately) (directly
functioning as adverbials) should be treated as adjectives that function con-
currently as adverbs; 强烈 (vehement) and 熟练 (skillful) (directly functioning
as attributives and adverbials) should be treated as adjectives that function
concurrently as distinctive words and adverbs (see Chapter 6).
However, except that 调查 (investigate) and 研究 (research) are treated as
verbs that function concurrently as nouns, and that 危险 (danger) and 温暖

Figure 7.6 The overlapping of verb property and nominal property

174

174  Conversional words and nominalization

(warm) are treated as adjectives that function concurrently as nouns, it is
rare to treat them as other kinds of conversional words. The homogeneous
strategy does not thoroughly treat words such as 调查 (investigate) as verbs
that function concurrently as nouns; what it thoroughly does is, as mentioned
earlier, treat as conversional words all the words that function concurrently
as several parts of speech. Some scholars emphasize that 调查 (investigate)
and 研究 (research) should be treated as verbs functioning concurrently as
nouns, but do not treat 成立 (establish) and 作废 (cancel) as verbs functioning
concurrently as distinctive words. They neither treat 木头 (wood) as a noun
functioning concurrently as a distinctive word nor 干净 (clean) and 新 (new)
as adjectives functioning concurrently as distinctive words. This shows that the
priority homomorphic strategy is often and commonly used, though nobody
realizes that. If we cannot accept that 成立 (establish) and 作废 (cancel)
should be treated as verbs functioning concurrently as distinctive words, and
that 干净 (clean) and 新 (new) should be treated as adjectives functioning
concurrently as distinctive words, then we cannot accept that 调查 (inves-
tigate) and 研究 (research) should be treated as conversional words but as
verbs. Of course, what classification strategy is used must be in accordance
with the syntactic system. This is not a matter of right or wrong but whether
a classification strategy is effective or not effective.

7.4  Nature of multiple functions of a Chinese part of speech

The earliest scholar who revealed the multiple functions of a Chinese part of
speech is Gao Mingkai. He said that “judging from whatever aspect, Chinese
notional words express their many part-​of-​speech meanings and have no fixed
part-​of-​speech characteristics” (Gao Mingkai, 1960: 38). “The vast majority
of Chinese words do not have only one kind of combination properties and
can be combined with many classes of words” (Gao Mingkai, 1957: 75).
Gao Mingkai thought that the multiple functions of a part of speech
are the reason why a Chinese modification word has no parts of speech. In
Chinese, the stem of a word is a lexeme that can be embodied as different
variants in the language’s syntactic category of part of speech. That is to say,
the same Chinese stem or word can have different part-​of-​speech meanings
at different occasions; thus, Chinese has the category of part of speech. But
just because these different part-​of-​speech meanings may be all the different
variants of the same word stem, Chinese notional words cannot be classified
into nouns, verbs, adjectives and others in accordance with their part-​of-​
speech meanings” (Gao Mingkai, 1963: 49).
Then Zhu Dexi further regarded the multiple functions of a part of speech
as an important feature of the Salmon grammar. He (1985b) held that Indo-​
European languages have a kind of simple one-​to-​one correspondence between
part of speech and syntactic constituent. Roughly speaking, a verb corres-
ponds to a predicate; a noun corresponds to a subject or object; an adjective
corresponds to an attributive; an adverb corresponds to an adverbial. But
 175

Conversional words and nominalization 175

Chinese or object predicate attributive adverbial

noun verb adjective adverb

Figure 7.7 The correspondence between a part of speech and a syntactic constituent

in English and Chinese

Chinese has no simple one-​to-​one correspondence between part of speech

and syntactic constituent, as shown in the following diagram:
An English adjective mainly functions as an attributive and belongs to sub-
stantive modification words, but a Chinese adjective mainly functions as a
predicate and belongs to predicate words. The functional equivalence of a
Chinese adjective to an English one is the distinctive word that also belongs to
substantive modification words. Therefore, the position of a Chinese adjective
in the above diagram should be that of a distinctive word; the adjective should
be put together with a verb, forming a large category of predicate words.
Moreover, because an English noun can generally function as an attribu-
tive, a connection line between nouns and attributives should be added.
The adjusted diagram for the correspondence between a part of speech and
a syntactic constituent is as follows:
After adjusting the position of an adjective, the multiple functions of
Chinese part of speech decrease because the function of two types of Chinese
modification words is still singular, and the multiple functions of a part of
speech are mainly reflected in predicate and substantive words.
What is the nature of the multifunction of a Chinese part of speech? It is
mainly threefold:

I. The positions of subject, object or predicate have a rather relaxed selec-

tional restriction on the parts of speech of a word.

We can see that the one-​to-​many correspondence between Chinese parts

of speech and syntactic constituents expresses itself in the fact that a predi-
cate word can function as a subject or object, and that a substantive word can
function as a predicate. The positions of subject, object or predicate do not
have a strict selectional restriction on the parts of speech of a word. The pos-
ition of subject or object allows the entrance of both substantive and predi-
cate words. The position of a predicate allows the entrance of both predicate
and substantive words.
In Chapter 4.3, we mentioned that 学习 (study), which functions as a sub-
ject in 学习很重要 (Study is very important), can take its object or adverbial
176

176  Conversional words and nominalization

English object predicate attributive adverbial

noun verb adjective adverb

subject or

Chinese object predicate attributive adverbial

substantive word predicate word distinctive word adverb

Figure 7.8 The correspondence between a Chinese part of speech and a syntactic

constituent

and thus remains a verb. In Chinese, it is very common that constituents with
predicate word properties function as a subject or object. We should not think
that words that function as a subject or object have changed in their parts of
speech.
In the example that noun phrases such as 小王黄头发 (Little Wang yellow
hair) function as a predicate, 黄头发 (yellow hair) can still be modified by an
attributive 一头黄头发 (a head of yellow hair) and therefore remains a noun
phrase.
In Chapter  4.3, we also mentioned that when constituents with predicate
word properties function as subjects, their parts of speech may not change
but their external expressional functions often have reference meanings. When
constituents with substantive word properties function as predicates, their
part of speech may remain a substantive word, but their external expressional
function often has statement meanings. That is to say, this kind of multifunc-
tion of a part of speech is actually caused by changes in its external expressional
function. The part of speech does not change; in other words, the selectional
restriction of the syntactic position is not strict. In English, the functioning of
a verb as a subject, object or attributive requires changes in form by adding
expressional function change markers such as “to, -​ing, -​ed” and so on. The
 177

Conversional words and nominalization 177

functioning of a noun as a predicate requires the addition of the link verb “be”.
This shows that Chinese has something in common with English. Namely the
expressional function of a word may change and then function as an unconven-
tional constituent. The difference lies in the fact that the changes in Chinese are
unmarked in most cases but those in English are marked. Zhu Dexi also said
that “the root causes for the two characteristics2 lie in the fact that Chinese parts
of speech have no marked form” (Zhu Dexi, 1985a: 9).
Because of the changes in expressional functions and parts of speech,
the loaned part of speech and the theory that the part of speech of a word
is based on sentence are reasonable. Simply because there is no distinction
between parts of speech at two levels, there is no clear explanation of their
relations. Thus we can conclude that Chinese words have no definite parts of
speech or none at all.

II. Classifying the parts of speech of a word that has concurrently two or
more parts of speech with the priority homomorphic strategy.

The multiple functions of some words are due to their different parts of
speech. For example, nominal verbs such as 检查 (examine), 管理 (manage)
and 调查 (investigate) have concurrently the properties of a verb and noun.
These words can function as predicates and have the properties of a predicate
word, but when they function as objects in such phrases as 进行检查 (conduct
examination), they show the properties of a noun (being able to be modified
by attributives or numerals and measure words) and have no characteristics
of a verb at all (not being able to take its object, adverbial, le (了), zhuo (着),
guo (过) or complement), therefore, having the properties of a noun. For
another example, 干净 (clean), 优秀 (excellent) and 好 (good) can function as
predicates and be modified by 很 (very), thus having the properties of a predi-
cate word and being adjectives, but when they function as attributives directly,
for example, 干净衣服 (clean clothes), 优秀学生 (excellent student) and 好人
(good person), they no longer have the characteristics of an adjective (not
being able to be modified by 很 (very) or 不 (not), and therefore having the
properties of a distinctive word. Words like 木头 (wood), 工人 (worker) and
手机 (cell phone) can freely function as a subject or object and be modified by
numerals and measure words, thus having the properties of a noun, but when
they function as attributives directly, they no longer have the characteristics
of a noun (not being able to be modified by numerals and measure words),
and actually have the properties of a distinctive word.
The multiple functions of a part of speech such as these are actually due to
these words originally having multiple parts of speech. Why should we regard
multiple parts of speech as the multiple functions of a part of speech? Other
languages such as English also have the phenomenon that one word has con-
currently multiple parts of speech, but why is this phenomenon not regarded
as multiple functions of a part of speech? This is because the current Chinese
part-​of-​speech classification system generally uses the priority homomorphic
178

178  Conversional words and nominalization

strategy and classifies into superior parts of speech words that have concur-
rently multiple parts of speech. For instance, nominal verbs such as 检查
(examine), 管理 (manage) and 调查 (investigate) that have concurrently
the properties of a verb and noun are classified into verbs; words like 干净
(clean), 优秀 (excellent) and 好 (good) that have concurrently the properties
of an adjective and distinctive word are classified into adjectives; words like
木头 (wood), 工人 (worker) and 手机 (cell phone) that have concurrently the
properties of a noun and distinctive word are classified into nouns.
In other words, such multiple functions of a part of speech actually result
from the selection of classification strategies. Strictly speaking, this belongs
to the phenomenon that one word has concurrently multiple parts of speech.
As mentioned previously, two Chinese modification words cannot have
multiple functions, but in reality, some Chinese words can belong to dis-
tinctive words and adverbs concurrently, for instance, 共同 (common),
长期 (long-​term), 自动 (automatic), 临时 (temporary) and 真正 (genuine).
However, our part-​of-​speech system uses the homogeneous strategy to treat
them as conversional words that belong to distinctive words and adverbs, thus
not being treated as the multiple functions of a part of speech.
English also has one word belonging to multiple parts of speech, but
English grammar generally uses the homogeneous strategy and therefore has
no multiple functions of a part of speech. With only one exception, which
is that because a great number of English nouns function as attributives,
English uses the priority homomorphic strategy and does not treat them as
nouns that function concurrently as adjectives, the multiple functions of a
noun are merged.
If Chinese also uses the homogeneous strategy to treat as conversional
words that concurrently belong to verbs and nouns, adjectives and distinctive
words, nouns and distinctive words, then the multiple functions of a Chinese
part of speech may also be greatly reduced. From this point of view, Chinese
and English have no substantial difference in language facts; the only diffe-
rence is their classification strategies.
The reason why Chinese generally uses the priority homomorphic strategy
is that it has a rather big number of words that have concurrently multiple
parts of speech. In other words, its parts of speech do not have a high degree
of differentiation. We shall discuss this issue in some detail in Chapter 3 in
Volume 2.

7.5  Nominalization
Nominalization is related to conversional words and will be discussed here.
The debate over nominalization occurred from the 1950s to 1960s. It
means that changes take place in the properties of verbs or adjectives in the
positions of subject or object, or modified by an attributive. It has different
explanations:  (1) use as a noun; (2)  becoming a noun, or nominalized;
(3)  already being a noun; (4)  nominalization, transformed into an object.
 179

Conversional words and nominalization 179

For example, 看下棋 (watch playing chess) and 学习很重要 (study is very
important).
Zhu Dexi, Lu Jiawen and Ma Zhen (1961) criticized the idea of nominal-
ization mainly in terms of grammatical meaning and grammatical properties.
In terms of grammatical properties, one of the reasons for nominalization
is that verbs and adjectives in these positions have a series of grammatical
characteristics of nouns and lose some or all of the grammatical properties
of verbs and adjectives. The series of grammatical characteristics means that
verbs and adjectives can have anaphoric reference with nouns or pronouns,
be modified by attributives and form a coordinate construction together with
nouns. The loss of some or all of the grammatical properties of verbs and
adjectives means that they cannot be reduplicated and cannot function as
predicates. Zhu Dexi et al. demonstrated this as follows:

a. Words of different parts of speech not only have individuality that

distinguishes them from each other but also have some commonality. For
example, both verbs and adjectives can take le (了). The above-​mentioned
three points can be regarded as the commonalities of a verb, adjective
and noun.
b. The grammatical properties of a part of speech are reflected in a gener-
alization word; not all of its grammatical properties may be achieved in
one position. The lost grammatical properties are nothing but those not
achieved. Verbs have the properties of 不 (not) ~ and ~ le (了), but they
cannot be achieved at the same time.

In terms of grammatical meaning, the reason for nominalization is that

in these positions, verbs and adjectives have the meanings of transformed
objects, thereby being nominalized. Zhu Dexi et al. demonstrated three levels
of transformation into object:

1. The transformation into object with categorical meanings: verbs denote

names of behaviors and actions; nouns denote objects; adjectives denote
properties.
2. The transformation into object is reflected in the positions of subjects
and objects.
3. The transformation into object at the meaning level is represented by
什么 (what) or 怎么样 (how).

The transformation into object at the three levels is not consistent, for
example:

(1) 看下棋 (watch playing chess) —​看什么 (watch what) (Levels 1 and

2 are not consistent)
(2) 坐着也行,站着也行—​怎么样都行 (Sitting is OKay; standing is
OKay; whatever is Okay) (Levels 2 and 3 are not consistent)
180

180  Conversional words and nominalization

The transformation into object as nominalization is at Level 2. The incon-
sistency with the transformation into object at Level 1 cannot prove that there
is nominalization.
Zhu Dexi’s criticism mainly aims at nominalization but does not really
deny the transformation into object. Later on, he changed his point of view,
thinking that verbs and adjectives after verbs that take quasi-​predicate objects
are nominal, for example, 调查 (investigate) and 困难 (difficult) in 进行调查
(conduct investigation) and 有困难 (have difficulty). He called them nominal
verbs and nominal adjectives respectively (Zhu Dexi, 1985b).
In the 1990s, a number of scholars raised the nominalization issue once
again in the following five ways: (1) Semantic level and syntactic level should
be separated; nominalization is thought to be at the semantic level (Hu
Yushu & Fan Xiao, 1994). (2) Nominalization is neither acknowledged nor
denied, but emphasis is laid on the positions of subject and object, verbs
after attributives and changes in the properties of adjectives. Namely, the
properties of a predicate word are weakened but the properties of a noun
are enhanced (Zhang Bojiang, 1993, 1996). (3) It was held that some verbs
and adjectives in the positions of subject or object, or after attributives have
already been nominalized, although some of them are still predicate words,
the properties of a predicate word are weakened and turn into those of a
noun (Yang Chengkai, 1991). (4)  The “contradiction” between the head-
word of endocentric construction and general parts of speech was used
to demonstrate that nominalization exists (see Shi Guangan, 1981, 1988).
(5)  Nominalization is explained in terms of expressional function (see Li
Yuming, 1986; Xiang Mengbing, 1991). Respective discussions are as follows:
Hu Yushu and Fan Xiao (1994) proposed that nominalization and nominal
object transformation into a noun should be separated; nominalization refers
to the transformation of verbs and adjectives into nouns at the syntactic level;
the nominal object transformation into a noun refers to the transformation of
predicate meanings of verbs and adjectives into the “nominal object meaning”
(object meaning). Their argument is that the centers of the semantic struc-
ture of a sentence are the verb center and verb argument. The verb center is
the center of a predicate; the verb argument is the argument and coordinate-​
valence constituent. The verb argument has the nominal object meaning or ref-
erence meaning. For example: 骄傲使人落后 (Conceit makes one lag behind).
This explanation is still what Zhu Dexi meant by the transformation into
object at Level 2.  The question is:  how does the verb argument denote the
nominal object meaning? 坐着 (sitting) in Example (2) can hardly be regarded
as having nominal object meaning. For another example:

(3) a b
干净最重要 (To be clean is the most important). 干干净净的舒服
(To be very clean
is comfortable)
 181

Conversional words and nominalization 181

教书不容易 (Teaching is not easy) 天 天 练 才 学 得 会
(Learning is done by
practicing every day)
(4) 看下棋 (watch playing chess) 打算下棋 (intend to
play chess)
考虑参加不参加 (consider whether 觉得很好 (feel fine)
participating or not)

The subjects or objects in Examples b all have no nominal object meanings,

though they are verb arguments.
Although Ways (2) and (3) do not reach entirely the same conclusion, their
proof is basically the same. Here we discuss them together. Some scholars
hold that changes take place in the positions of subjects and objects and in the
properties of verbs and adjectives after attributives. They justify themselves
mainly as follows: (1) verbs and adjectives take tense and aspect constituents
such as le (了), zhuo (着), guo (过) or complete sentences, namely losing their
time meaning; (2) they cannot take modal adverbials such as 大概 (presum-
ably, 也许 (perhaps) and 必定 (certainly); (3)  numerals and measure words
cannot function as the objects of verbs and adjectives. Hence, they think that
verbs and adjectives in the positions of subject and object or after attributives
lose some properties of a predicative word and acquire some properties of
a noun.
These viewpoints are similar to the nominalization theory in the 1950s
and 1960s. Zhu Dexi, Lu Jiawen and Ma Zhen (1961) criticized such
viewpoints: not all the grammatical properties of a generalization word may
be achieved in one grammatical position. The lost grammatical properties are
nothing but those not achieved. This argument is mainly from a logical per-
spective and does not show that changes take place in the positions of subject
and object, and in the properties of verbs and adjectives after attributives,
but we cannot deny that changes may take place in their properties. Zhu Dexi
(1985b) also admitted that 研究 (research) in 进行研究 (carry out research) is
nominal. Then, can the three pieces of evidence proposed by these property
change theorists prove that verbs and adjectives in these positions really lose
some properties of a predicate word and have some properties of a noun?
In our opinion, they cannot. The three pieces of evidence rely on such an
assumption:  verbs and adjectives that take tense and aspect constituents,
modal adverbials and objects made of numerals and measure words reflect
the properties of a predicate word. The loss of these characteristics means the
loss of properties of a predicate word. We will explain that this assumption
cannot be valid.
Evidence One: Verbs and adjectives in the positions of subject and object
or after attributives cannot take tense and aspect constituents such as le (了),
zhuo (着) and guo (过). This is mainly related to the time meaning of a predi-
cate word. The common thinking is that at this time, verbs and adjectives
lose their time meaning. According to Guo Rui (1993, 1997b), the time
182

182  Conversional words and nominalization

meaning of a predicate word includes intrinsic time meaning and extrinsic
time meaning. Intrinsic time meaning refers to the time meaning decided
by the lexical meaning of a predicate word and reflects its intrinsic process
structure. The process structure refers to start, continuation and end, and
on this basis, verbs are classified into state verbs, action verbs and change
verbs. Extrinsic time meaning indicates the relationship between predicate
word elements and the external lapse of time. Depending on whether a predi-
cate word element is put into an external time lapse process, it is classified
into process and nonprocess elements. Intrinsic time meaning is the inherent
property of a predicate word, while extrinsic time meaning is not and is
the achieved property in the use of a predicate word. From this perspec-
tive, Evidence One cannot prove that verbs and adjectives in the positions
of subjects and objects, or after attributives lose their time meaning. In fact,
they exist merely in a certain type of special time meaning. Some examples
are analyzed as follows:

(5) 假若不幸而无论如何也不调谐,她会用她的气派压迫人们的眼睛,承
认她的敢于故作惊人之笔,像万里长城似的,虽然不美,而惊心动魄.
(老舍《四世同堂》) (Unfortunately, no matter how hard she tried,
if she is not attractive, she may use her imposing manner to attract
folk’s eyes, letting them admit that she is courageous to dress herself
up heart-​shockingly like the Great Wall. (Four Generations in One
House by Lao She)

The adjective 敢于 (courageous) cannot take tense and aspect constituents

such as le (了), zhuo (着) and guo (过). In fact, like “is, belong to, feel”, “cour-
ageous” does not take tense and aspect constituents under any circumstances.
This is the restriction on intrinsic time meaning, and no time meaning is lost.
These verbs have the infinite construction of a state verb; infinity is their time
meaning. Even in the position of predicate, they cannot take tense and aspect
constituents. Therefore, the infinity of intrinsic time meaning does not indi-
cate the loss of properties of a predicate word.

(6) 这本书的出版给我们带来了麻烦. (The publication of this book

brings us trouble.)

出版 (publication) in this example cannot take any tense and aspect

constituents because of extrinsic time meaning. Guo Rui (1997b) thought
that the properties of a predicate word can be classified into two types
according to its extrinsic time meaning. One type is the process element,
that is, the predicate word element observed in the external time process,
for example, 他在抽烟 (he is smoking), 他抽了烟了 (he smoked), 他抽过烟
(he has smoked) and 他抽烟呢 (he is smoking). They are usually negated
with 没 (no) or 没有 (not have). Another type is the nonprocess element
(or property element), for example, 他抽烟 (he smokes), 我抽烟 (I smoke),
你坐下 (you sit down), 猫吃老鼠 (a cat eats a mouse) and 地球绕太阳运行
 183

Conversional words and nominalization 183

(the earth turns around the sun). They are usually negated with 不 (not) or
its derivative forms 别 (don’t), 甭 (needn’t). Tense and aspect constituents
such as le (了), guo (过), ne (呢) and zai (在) are marked process elements,
whereas unmarked elements are usually nonprocess (property) elements. In
other words, a predicate word element in the predicate position differs in
the process element and the nonprocess (property) element. The property
element in the predicate position does not lose its time meaning but is a kind
of special time nonprocess. One may insist on the loss of time meaning, but
one cannot think that the loss of time meaning is the loss of properties of
a predicate word because it cannot be denied that the nonprocess element
that functions as a predicate loses the properties of the predicate word.
Similarly, verbs and adjectives in the positions of subjects and objects, or
after attributives, are often expressed as nonprocess elements (negated with
不 (not)). Their properties of predicate words should not be negated just
because they are not process elements. Actually, it is widely acknowledged
that verbs and adjectives in the positions of subjects and objects, or after
attributives, can still take such adverbials as 不 (not) and 及时 (punctually),
objects and sometimes subjects, indicating that they are still predicate words.
They do not lose their time meaning and the properties of predicate words,
but are expressed as nonprocess elements just as the predicate word elem-
ents in imperative sentences, volitional sentences, habitual sentences and uni-
versal sentences are expressed as nonprocess elements.
Fundamentally speaking, predicate words (verbs and adjectives) themselves
do not have a process meaning but merely express abstract actions, states or
attributes, all of which do not have a process meaning. This is just like nouns
themselves have no reference meaning but only express abstract objects (Li,
1997). The role played by such elements as le (了), guo (过), zai (在) and ne
(呢) in a predicate word element is similar to that played by numerals, measure
words and demonstratives in a noun element, making abstract things spe-
cific and tangible. Therefore, predicate words that do not take such tense and
aspect constituents as le (了), guo (过), zai (在) and ne (呢) are just their true
characteristics.
Similar to Evidence One, under the circumstances of Evidence Two, one
use of a predicate word is that it takes a modal adverbial. Because the modal
constituent is a sentence or a clause, and the constituent that is not in the
position of a predicate does not take a modal constituent normally, it cannot
be posited that it loses its properties of a predicate word. Fundamentally
speaking, a predicate word (verb and adjective) itself does not express any
modal meaning, which comes from its use. The true characteristic of a predi-
cate word is that it does not take a modal constituent.
Evidence Three itself does not conform to facts. We can still find some
examples that the predicate word in the positions of subject and object takes
numerals and measure words as its object:

(7) a. 去一次不够 (Going once is not enough)

b. 他建议去一趟 (He suggests going once)
184

184  Conversional words and nominalization

However, it is really difficult for verbs and adjectives after attributives to
take numerals and measure words as their objects. Conversely numerals and
measure word constituents can be put in front of verbs and adjectives (Yang
Chengkai, 1991):

(8) a. *这本书的出版多次 (*publishing the book many times)

b. 这本书的多次出版 (many times publishing of the book)

Constituents with the properties of a predicate word in the position of a

predicate can also be added to their front with numerals and measure word
constituents:

(9) a. 他三天没吃饭 (He did not eat for three days)

b. 他一次也没去 (He did not go even once)
c. 我多次告诉他…… (I told him many times . . . )
d. 我三次去上海,都没有找到他 (I went to Shanghai three times
but did not find him)

Should numerals and measure word constituents be placed before or after

predicate words? What are the rules? We are not very clear on this. Maybe
they are related to pragmatics and focus arrangement. We are sure that from
these examples, the front position of numerals and measure word constituents
cannot prove that they lose their properties of a predicate word.
Such being the case, the changes in the properties of verbs and adjectives in
the positions of subjects and objects, or after attributives, are not real changes
in their properties but merely one of the properties of a predicate word.
Verbs and adjectives do not lose the properties of predicate words, which are,
instead, their true characteristics, namely expressing an abstract action, state
or attribute.
Ways (4) and (5) can be discussed together. Shi Guan’gan (1981) thought
that because the construction of 这本书的出版 (the publication of this book)
is neat and nominal, and that the classification of 出版 (publication) into a
verb violates the endocentric construction theory, it should be regarded as
a noun. Li Yuming (1986) thought that nominalization is the “positional
meaning” reflected by a verb and adjective in their positions of subjects
and objects, namely the meaning brought by a grammatical position. Xiang
Mengbing (1991) thought that nominalization was actually a reference. There
is no contradiction between the headword of 这本书的出版 (the publication
of this book) and the parts of speech of the whole construction consisting of
a modifier and the word it modifies because the whole construction still has
the properties of a verb.
Our points of view are similar to those of Li Yuming and Xiang Mengbing.
Namely we think that nominalization is actually a reference, but do not con-
sider it as the properties brought by a grammatical position. Instead, we
regard it merely as the temporary properties of a predicate word itself. We
 185

Conversional words and nominalization 185

neither think that the whole phrase 这本书的出版 (the publication of this
book) has the properties of a verb. Its properties are actually nominal. Then
how should we explain the contradiction between the headword and the parts
of speech of the whole construction consisting of a modifier and the word
it modifies? Chapter 4.3 mentions that an expressional function has intrinsic
(inherent) and extrinsic (temporary) levels. Correspondingly, a part of speech
also has lexical and syntactic levels. 出版 (publication) in 这本书的出版 (the
publication of this book) is at the extrinsic level or nominalization at the syn-
tactic level. Though the word 出版 (publication) is still a verb at the lexical
level, it is a noun at the syntactic level. The nominality of the whole phrase
这本书的出版 (the publication of this book) is in agreement with the part
of speech of the headword 出版 (publication) at the syntactic level, thus not
violating the endocentric construction theory. The nominalization of verbs
and adjectives in the position of subjects or objects is also the reference at
the extrinsic level or the nominalization at the syntactic level. The parts of
speech at the lexical level are still verbs and adjectives.
Our points of view are summarized as follows:

“The nature of nominalization” includes the following two cases:

1. The reference of an extrinsic expressional function and nominaliza-
tion at the syntactic level.
(10) a. 去不合适 (Going is inappropriate)
b. 不去不合适 (Not going is inappropriate)
c. 去广州不合适 (Going to Guangzhou is inappropriate)
d. 你去不合适 (Your going is inappropriate)
马上去不合适 (Going immediately is inappropriate)
(11) a.  社会的公正(是社会进步的基础) (Social justice (is the basis
of social progress)).
b. 社会的不公正(在任何时代都存在) (Social injustice (exists
in any era)).
c. 社会的绝对公正(是社会进步的基础) (Absolute social
justice (is the basis of social progress)).
去 (going) and 公正 (justice) still have the general characteristics
of a verb and adjective, and do not change their properties at the
lexical level.
2. Some words that have the properties of verbs, adjectives and nouns
embody the properties of a noun at the position of subjects or
objects. Their functioning as subjects or objects no longer has the
general characteristics of a verb or adjective but has the general
characteristics of a noun. For example:
(12) a. 进行调查 (conduct investigation)
b. *进行不调查 (*carry out no investigation)
c. *进行调查这个问题 (*conduct investigation of this
problem)
186

186  Conversional words and nominalization

d. *进行他们调查 (*carry out they investigate)
e. *进行马上调查 (*carry out immediate investigation)
f. 进行仔细的调查 (carry out careful investigation)
g. 进行社会调查 (carry out social investigation)
(13) a. 保持稳定 (remain stable)
b. *保持很稳定 (*remain very stable)
c. *保持不稳定 (*remain unstable)
d. *保持稳定得很 (*remain stable nicely)
e. 保持社会稳定 (maintain social stability)
f. 保持社会的稳定 (maintain societal stability)

Here, 调查 (investigation) and 稳定 (stability) solely reflect the properties

of a noun; thus, we can say that they have two properties of a verb (adjective)
and a noun. They can be regarded as changes that have taken place in intrinsic
expressional function. In other words, the properties of parts of speech at the
lexical level have already changed indeed, but we merely do not treat them as
conversional words with the priority homomorphic strategy.

Notes
1 High or low syntactic status mainly depends on the following: (1) after a word is
combined with another, with which its integral property is identical. For example,
the phrase formed by combining a noun with a distinctive word is a noun phrase
because the status of the noun is higher than that of a distinctive word. (2) The dir-
ection of change in a part of speech. For instance, most distinctive words change
from a noun or a verb; there are few changes in the opposite direction. Therefore,
the statuses of a noun and a verb are higher than those of a distinctive word. See
Chapter 4.3.3.
2 This refers to the two characteristics that there is no one-​to-​one correspondence
between a part of speech and a syntactic constituent, and that the sentential struc-
tural rule by and large agrees with the phrasal structural rule.
 187

8 
Conclusions

Because a grammatical position has its selectional restriction on words,

Chinese has its parts of speech, and it is necessary to classify them.
There is an intricate correspondence between parts of speech and the dis-
tribution of words; therefore, we cannot find any parts of speech that have
internally universal, but externally exclusive, distributions. Essentially, a part
of speech is not a distributional type; rather, its intrinsic basis is the expres-
sional function and semantic type. In essence, large categories (substantive
words, predicate words, modification words) are classes of words classified
according to their expressional functions. The basic categories of parts of
speech (nouns, verbs, adjectives) are classes that combine semantic types with
syntactic functions. Semantic types reflect the categorization of a language
and therefore must be taken into consideration in classifying parts of speech.
Only in this way can we give a more reasonable explanation for the bases of
the selectional restriction of a grammatical position on word. The expres-
sional functions used as the essence of parts of speech refer to the meaning
representation pattern of a word, including the four basic types: statements,
references, substantive modifications and predicate modifications. Semantic
types refer to categorized semantic meanings, such as entity, position, meas-
urement unit, demonstration, behavior, action, attribute, state and so on.
Both the expressional function of a word and its semantic type are its gram-
matical meanings. We classify notional words into 13 parts of speech, each of
which has its own grammatical meaning: a noun denotes an entity reference;
a positional word denotes a position reference (a place word denotes an abso-
lute space position reference; a time word denotes an absolute time position
reference; a locative denotes a relative position reference); a measure word
denotes a measurement unit or a grade unit reference. An adjective denotes
a property statement; a state word denotes a state statement; a verb denotes
a behavior or action statement. A numeral denotes a numerical value modi-
fication; a numeral and measure word denotes a quantitative modification;
a demonstrative denotes a demonstration modification; a distinctive word
denotes a property modification; an adverb denotes a condition modification.
The grammatical meaning of a word is the root factor that restricts its distri-
bution, and this is the reason why we can classify parts of speech according
to distribution.
188

188 Conclusions
Expressional functions can be classified into the intrinsic expressional
function and the external expressional function. Correspondingly, we can
also classify parts of speech into those at the lexical and the syntactic level.
The lexical and syntactic levels often cause changes in a part of speech, indi-
cating that grammar is dynamic. One of the reasons why it is difficult to clas-
sify Chinese parts of speech is that Chinese has a predominantly dynamic
grammar.
The classification of parts of speech according to the distribution of a
word is only a convenient theory. Strictly speaking, the parts of speech of a
word are inferred from its distributive characteristics and already exist before
“being classified into them”.
Although expressional functions are the intrinsic bases for classifying parts
of speech, they are not directly observable and therefore cannot be used as
criteria for classifying parts of speech. We still use word distribution as a cri-
terion for classifying parts of speech, but do so not purely according to the
distributional difference but rather through the “representation-​expression”
relationship among distribution, expressional function and semantic type
(distribution reflects expressional function and semantic type, which are
represented as distribution). We use the distribution compatibility and the
correlation principle to analyze which distributional differences represent the
differences in a part of speech and which do not. In this way, grammatical
functions that have equal classification values are collected into clusters, with
each equivalent function cluster representing one part of speech. The classi-
fication criteria are selected from the equivalent function cluster of a part of
speech.
The relationship between a grammatical function and a part of speech
is intricate and complicated but has a prototypical connection:  a substan-
tive word functions as a subject or object; a predicate word functions as a
predicate or complement; a modification word functions as a modifier. Other
connections are marked.
Conversional words are mainly concerned with the identity of a word, the
multiple functions of a part of speech and the classification strategy. Different
generalization words have different parts of speech and, of course, should be
treated as conversional words (or homographs with different parts of speech).
Whether a part of speech has multiple functions or not is mainly distinguished
through distribution compatibility and the correlation principle. If a gener-
alization word has the properties of different parts of speech, whether it is
treated as a conversional word or not has something to do with the classifica-
tion strategy. On the whole, there are four classification strategies: the homo-
geneity strategy, the homomorphical strategy, the priority homomorphical
strategy and the consolidation strategy. The strategy selection should be
considered in an all-​around way. The principle is to minimize the total costs
of part-​of-​speech simplicity and syntactic simplicity.
Because a Chinese word commonly has multiple functions, Chinese scholars
more commonly use priority the homomorphical strategy to classify parts of
speech, lest there should be too many conversional words. The so-​called “no
 189

Conclusions 189
change in lexical meaning, no change in a part of speech” actually means
the use of the priority homomorphical strategy. However, this does not mean
that the priority homomorphical strategy has no shortcomings. Because
there is no one-​to-​one correspondence between a part of speech of a word
and its grammatical function, the parts of speech classified with the priority
homomorphical strategy are not so effective for syntactic analysis. Therefore,
in the areas of formal grammar and Chinese information processing, it is
advisable to use the homogeneity strategy to classify Chinese parts of speech.
In contrasting the Chinese part-​of-​speech system with that of another
language, the classification strategy should be taken into consideration. The
differences in a part-​of-​speech system, which are caused by using different
classification strategies, should not be regarded as the differences in the parts
of speech themselves of the two languages. For example, Chinese grammar
has the characteristic that “there is no one-​to-​one correspondence between a
Chinese part of speech and a syntactic constituent”. To a greater extent, this
is actually caused by using the priority homomorphical strategy to classify
Chinese parts of speech. As a matter of linguistic fact, if we use the homo-
geneity strategy to classify Chinese parts of speech, then Chinese will have
no such common correspondence between parts of speech and syntactic
constituents.
A grammatical position has its selectional restriction on a word, and we
actually use it to classify parts of speech. We can say that a part of speech is
the basis for syntactic analysis. But because Chinese has no one-​to-​one corres-
pondence between parts of speech and syntactic constituents, a part of speech
may not play such a big role in syntactic analysis as in Western languages
such as English. We cannot use the category of a part of speech to write out
the quite complete and basically dis-​ambiguous phrasal structural rules. How
should we look at this? The following are our opinions:

(1) There are two causes for no one-​to-​one correspondence between Chinese
parts of speech and syntactic constituents: (a) The selectional restriction
of Chinese syntactic positions such as subjects, objects and predicates on
the parts of speech of a word is rather loose. The positions for subjects
and objects allow words that have predicate word properties to enter.
The position for predicates also allows words that have substantive word
properties to enter. (b) Most modern Chinese grammatical systems use
the priority homomorphical strategy to classify parts of speech, and thus
classify words that function concurrently as several parts of speech into
dominant parts of speech. Both causes may make it difficult to describe
syntactic rules with a part-​of-​speech marker.
(2) We should not think that the reason why Chinese words have no part-​of-​
speech distinction is that there is no one-​to-​one correspondence between
parts of speech and syntactic constituents. The observation that a language
has no parts of speech actually means that its grammatical positions have
no selectional restriction on words, implying that a certain position allows
any words to enter. A syntactic constituent is only a grammatical position
190

190 Conclusions
that has a rather high generalization level. Although Chinese has no one-​
to-​one correspondence between parts of speech and syntactic constituents,
the grammatical position that has a rather low generalization level is quite
strict with the selectional restriction on words. We still have to admit that
the grammatical position has its selectional restriction on a word. So long
as there is selectional restriction, we have to admit that a Chinese word
has its part-​of-​speech distinction. In other words, the distinction may not
necessarily be shown in a syntactic constituent, but may well be shown in
the main in a rather specific grammatical position, not to mention that a
syntactic constituent can show part-​of-​speech distinctions. For example,
a distinctive word cannot function as a complement or a predicate and
cannot take an object, a complement or an adverbial, thus distinguishing
between a verb and an adjective.
(3) It is a fact that Chinese has no one-​to-​one correspondence between parts
of speech and syntactic constituents, but we cannot distort the fact for the
sake of yielding to syntactic analysis. Some people think that once a verb
or adjective functions as a subject or an object, it becomes a noun, but
this idea is not true to fact.
(4) We cannot count on Chinese to have a completely neat correspondence
between parts of speech and syntactic constituents. Even English does
not have a complete correspondence. For example, the + adj. can appear
in the position of a subject. But here it remains an adjective because it has
an adjective’s general properties. For instance, it can be modified by an
adverbial and have comparative and superlative degrees.
(5) To overcome the insufficiency of a part of speech for syntactic analysis,
we can separate syntactic constituents by attaching conditions. Although
different Chinese parts of speech may sometimes occupy the same
grammatical position, if looked at in detail, the two have very different
structures. For example, verbs and some nouns may function as predicates.
But nouns that function as predicates usually denote judgment and are
easily distinguished with language sense from verbs that function as
predicates. A noun predicate can well be called a quasi-​predicate so as to
be distinguished from a verb predicate.
(6) Part-​
of-​
speech classification can be combined with grammatical
characteristic description. Although a part of speech has its limitations
for syntactic analysis, we can describe the grammatical characteristics of
the words in a word bank in addition to tagging their parts of speech, for
example, whether a word can function as subject or be modified by 不
(not) or not, thus offsetting the limitations of parts of speech in syntactic
analysis.

The fundamental reason why the four categories of parts of speech classified
according to the four positions of expressional functions are universal in the
world languages is that the basic work mechanisms of human languages com-
bine statements with references to express meaning and transmit information.
 191

Conclusions 191
That is to say, a human language expresses its meaning by combining the
object-​representation constituent with the assertion-​ representation con-
stituent. It has its division of labor for object-​representation constituents and
assertion-​representation constituents, there being consequently distinctions
between nouns and verbs. Furthermore, there is a division of labor among
modification constituents, object-​representation constituents and assertion-​
representation constituents. Then a modification word appears as a part of
speech. If there is division of labor for object-​modification constituents and
assertion-​modification constituents, then there is a distinction between sub-
stantive modification words (distinctive words or adjectives with the proper-
ties of a modification word) and predicate modification words (adverbs).
I abandoned the view that distribution is the essence of a part of speech,
which I had firmly held for many years, and then proposed the view that the
essence is the expressional function/​semantic type because my long-​time part-​
of-​speech research shows that the former is internally inconsistent and has
insoluble internal contradictions. Instead, the latter is internally consistent
and can be used as the starting point to determine the classification system
and select classification criteria according to the correspondence between dis-
tribution and a part of speech, thus making the part-​of-​speech classification
demonstrable and not relying on sense perception to do it. In this way, our
attention is focused not on identifying a single distributive characteristic that
is internally universal and externally exclusive but on clustering the grammat-
ical functions that have the same classification value through the “reflection-​
representation” relationship between distribution and expressional function/​
semantic type, thereby finding the classification criteria. This method is con-
sistent with the form and meaning mutual verification method persistently
proposed by Zhu Dexi. In other words, we seek the mutual agreement of form
and meaning to a maximal extent and think that only the things that have the
mutual agreement of form and meaning exist in language reality and are valu-
able. From this perspective, the view that distribution is the essence of a part
of speech actually holds that a part of speech has only a form, while things
that have only forms should be valueless. But a part of speech is valuable. This
means that we should change our view and regard the expressional function/​
semantic type as the essence of a part of speech. The expressional function/​
semantic type represents meaning, whereas distribution represents form. The
agreement between the two is a part of speech. Based on this methodological
view, this book strictly uses distributional criteria to classify parts of speech.
Therefore, although it asserts that a part of speech is not a distributional type
in essence, scholars who adhere to the view that distribution is the essence of
a part of speech may rely more heavily on distribution in its classification.
192

Bibliography

Ai Wen & Jiang Wenqin (1980). Attempt to divide notional words of modern Chinese.
Journal of Wenzhou Normal College, nos. 1/​2.
Arnauld, Antoine & Lancelot, Claude (1660). Grammaire générale et raisonnée
(trans. Zhang Xuebin & Liu Li). Checked and annotated by Yao Xiaoping.
Changsha: Hunan Education Publishing, 2001.
Bai Shuo (1995). Computer aided discovery of linguistic knowledge. Beijing: Science  Press.
Bhat, D. N.  S. (1994). The adjectival category. Amsterdam and Philadelphia:  John
Benjamins.
Bloomfield, Leonard (1926). A set of postulates for language science. Language,
2(3), 153–​164. (Also in Liu Runqing (Ed.). An anthology of modern linguistic
masterpieces, vol. 1. Beijing: Surveying and Mapping Press, 1988.)
Cao Bohan (1995). Comments on morphology and parts of speech, issues of Chinese
parts of speech. Beijing: Zhonghua Book Company.
Chen Aiwen (1986). Research on Chinese parts of speech and division experiments.
Beijing: Peking University Press.
Chen Baoya (1985). Study of syntactic structure. Journal of Southwest Normal College,
no. 2.
Chen Baoya (1999). Methodology of Chinese linguistics in twentieth century: 1898–​
1996. Jinan: Shandong Education Press.
Chen Chengze (1922/​1982). A draft of Chinese grammar. Beijing: Commercial Press.
Chen Enquan (1987). Dividing parts of speech in Mandarin. Journal of Lanzhou
University (Social Science Edition), no. 3.
Chen Guanglei (1994). Chinese morphology. Shanghai: Xuelin Press.
Chen Guanglei (1996). The functions of Chinese parts of speech. Journal of Shanghai
University, no. 1.
Chen Ningping (1987). Extensions of modern Chinese nouns –​studying the boundary
between verb and noun in modern Chinese. Studies of Chinese Language, no. 5.
Chen Wangdao (1939/​1984). From diversity to unity. Language Weekly, no. 33. (Also
in Chinese Grammatical Reform Series. Beijing: Commercial Press.)
Chen Wangdao (1941). Reply to the critique of discussions of Chinese grammatical
reform. In Collected Papers of Chen Wangdao, vol. 3. Shanghai: Shanghai People’s
Publishing, 1981.
Chen Wangdao (1943). Study of grammar. In Collected Papers of Chen Wangdao, vol.
3. Shanghai: Shanghai People’s Publishing, 1981.
Chen Wangdao (1978). Brief introduction to grammar. Shanghai: Shanghai Educational
Publishing House.
 193

Bibliography 193
Chen Xiaohe (1999). Study of Chinese parts of speech from the perspective of auto-
matic syntactic analysis. Language Teaching and Research, no. 3.
Chen Yi (1989). A discussion on adding words before specialized verbs. Studies of
Chinese Language, no. 1.
Cheng Zenghou (1988). A review of course in general linguistics. In Hu Mingyang
(Ed.), Selected readings of Western linguistics masterpieces. Beijing:  Renmin
University of China Press.
Comrie, B. (Ed.). (1987). The world’s major languages. London and Sydney: Croom
Helm.
Conrad, Конрад Н. И. (1952). On Chinese (trans. Peng Chunan). Linguistic Issues, no.
3. (Also in Studies of Chinese Language, nos. 9–​11.)
Croft, W. (1991). Syntactic categories and grammatical relations: the cognitive organiza-
tion of information. Chicago and London: University of Chicago Press.
Dixon, R. M.  W. (1977). Where have all the adjectives gone? Studies in Language,
1(1),  19–​80.
Dixon, R. M.  W. (2004). Adjective classes in typological perspective. In R. M.  W.
Dixon & A. Y. Aikenvald (Eds.), 1–​49.
Dixon, R. M. W. (2010). Basic linguistic theory, vol. 2, Grammatical topics. Oxford:
Oxford University Press.
Dixon, R. M. W. & Aikhenvald, A. Y. (Eds.). ( 2004). Adjective classes: a cross-​linguistic
typology. Oxford: Oxford University Press.
Draguno, A. (1958). Study of modern Chinese grammar: parts of speech I (trans. Zheng
Qingzu). Beijing: Science Press.
Fan Xiao (1990). Functional division of words. Journal of Yantai University, no. 2.
Fang Guangtao (1939). Systems and methods. Language Weekly, no. 28.
Fang Guangtao (1956). Some fundamental problems in the study of Chinese parts of
speech (outline). In Collected Papers of Fang Guangdao. Beijing: Commercial Press.
Forey, P. L. (1983). A review of cladistic systematics. In Zhou Mingzhen, Zhang
Miman & Yu Xiaobo (Comps.), Translation collections of cladistic systematics.
Beijing: Science Press.
Fries, C. C. (1952). The structure of English (trans. He Leshi). Beijing:  Commercial
Press, 1964.
Fu Donghua (1938). A proposal for the new system of Chinese grammar. Language
Weekly, no.16. (Also in Series of Chinese Grammar Reforms. Beijing: Commercial
Press, 1984.)
Fu Zidong (1956). Division and identification of parts of speech. Studies of Chinese
Language, no. 3.
Gao Gengsheng (1995). Assumptions on dividing Chinese parts of speech. Journal of
Chinese Linguistics, no. 6.
Gao Mingkai (1953). Differences in Chinese parts of speech. Studies of Chinese
Language, no. 10.
Gao Mingkai (1954). Further discussions on differences in Chinese parts of speech.
Studies of Chinese Language, no. 8.
Gao Mingkai (1955). The third discussion on differences in Chinese parts of speech.
Studies of Chinese Language, no. 1.
Gao Mingkai (1957/​1986). On Chinese grammar. Beijing: Commercial Press.
Gao Mingkai (1960). Speech at the May Fourth Scientific Symposium at Peking
University in 1959. Linguistic Series, 4th series. Shanghai: Shanghai Educational
Publishing House
194

194 Bibliography
Gao Mingkai (1963). On parts of speech in modern Chinese research. Journal of Anhui
University, no. 1, 35–​52.
Gao Mingkai & Lin Yongyou (1963). From “nominalization of verbs and adjectives”
to parts of speech in Chinese. Journal of Peking University (Humanities Edition),
no. 2, 49–​66.
Givón, T. (1984). Syntax: a functional and typological introduction, vol. 1. Amsterdam:
John Benjamins.
Gong Qianyan (1997). The history of Chinese grammar. Beijing:  Language &
Culture Press.
Guo Rui (1990). On categorical systems of grammar. In Papers from the second sem-
inar on modern Chinese grammar. Huadong Normal University.
Guo Rui (1993). Process structures of Chinese verbs. Studies of Chinese Language, no. 6.
Guo Rui (1997a). Types of expressional functions and related issues. Linguistic Series,
19th series.
Guo Rui (1997b). Process and non-​process: two extrinsic time types of Chinese predi-
cate constituents. Studies of Chinese Language, no. 3.
Guo Rui (1999). Part of speech tagging in dictionary of modern Chinese. Studies of
Chinese Language, no. 2.
Guo Rui (2000). Transformation of expressional functions and functions of Chinese
character de (的). Contemporary Linguistics, no. 1.
Guo Rui (2005). Re-​ thinking on conversional words. In Papers from the 50th
Anniversary of Symposium on Parts of Speech Expert Discussion. Wuhu:  Anhui
Normal University Press.
Guo Rui (2010). Typology and system of Chinese parts of speech. In Papers from the
16th Symposium on Modern Chinese Grammar. Hong Kong:  City University of
Hong Kong Press.
Guo Rui (2011). Zhu Dexi’s study of Chinese parts of speech. Chinese Language
Learning, no. 5, 13–​26.
Guo Rui (2012). Gao Mingkai’s study of Chinese parts of speech. In Essays on
Linguistics, vol. 46.
Guo Rui (2015). Revisions of part of speech tagging in the 6th edition of Dictionary
of Modern Chinese. Journal of Chinese Dictionaries, 1, 167–​173.
Guo Rui (2017). System of Chinese parts of speech from the perspective of the
Amsterdam model. In Discussions on Chinese and Japanese studies  –​the 60th
birthday celebration of Professor Yang Kairong. Tokyo: Asahi Press.
Guo Shaoyu (1979). A new study of Chinese grammatical rhetoric. Beijing:
Commercial Press.
Guo Xiliang (1996). System of ancient Chinese parts of speech. In Xie Jifeng & Liu
Guanghe (Eds.), Xin Huo Collections. Taiyuan:  Shanxi University Associated
Press.
Halliday, M. A. K. (1985). An introduction to functional grammar. London: Edward
Arnold.
Harris, Z. S. (1946). From morpheme to utterance. Language, 22, 161–​183. (Also Li
Zhenlin (trans.), Linguistics Materials (1963), 6th issue.)
Harris, Z. S. (1951). Methods in structural linguistics. Chicago:  University of
Chicago Press.
Hengeveld, K. (1992). Non-​verbal predication: theory, typology, diachrony. Functional
Grammar Series 15. Berlin: De Gruyter Mouton.
Hengeveld, K. (2007). Parts-​ of-​
speech systems and morphological types. ACLC
Working Papers, 2(1),  31–​48.
 195

Bibliography 195
Hengeveld, K., Rijkhoff, J. & Siewierska, A. (2004). Parts-​of-​speech systems and word
order. Journal of Linguistics, 40(3), 527–​570.
Hengeveld, K. & van Lier, E. (2008). Parts of speech and dependent clauses in
functional discourse grammar. Studies in Language (Special issue:  Parts of
Speech: Descriptive Tools, Theoretical Constructs), 32(3), 753–​785.
Hengeveld, K. & van Lier, E. (2010). An implicational map of parts-​ of-​
speech.
Linguistic Discovery, 8(1), 129–​156.
Hockett, C. F. (1958). A course in modern linguistics (trans. Suo Zhenyu & Ye Feisheng).
Beijing: Peking University Press.
Hopper, Paul J. & Thompson, S. A. (1984). The discourse basis for lexical categories in
universal grammar. Language, 60(4), 703–​752.
Householder, F. W., Jr. (1952). Review: methods in structural linguistics. International
Journal of American Linguistics, 18(4), 260–​268.
Hu Mingyang (1992). Prologue for Shi Youwei’s call for flexibility. In Shi Youwei’s Call
for Flexibility. Haikou:  Hainan Publishing House. (Also in Language Planning
(1993), no. 2.)
Hu Mingyang (1995). A survey of modern Chinese parts of speech. Studies of Chinese
Language, no. 5.
Hu Mingyang (1996a). A review of the study of modern Chinese parts of speech. In
Hu Mingyang (Ed.).
Hu Mingyang (1996b). Issues on conversional words. In Hu Mingyang (Ed.).
Hu Mingyang (1996c). A survey of conversional words of verbs and nouns. In Hu
Mingyang (Ed.).
Hu Mingyang (Ed.). (1996). A survey of issues of parts of speech. Beijing:  Beijing
Language Institute Press.
Hu Yushu & Fan Xiao (1996). “Nominalization” of verb and adjective. Studies of
Chinese Language, no. 2.
Jin Zhaoxin (1922/​1983). Studies of Chinese grammar. Beijing: Commercial Press.
Juilland, A. G. & Elliott, E. (1957). Perspectives of linguistic science. In Monographs
Series on Language and Linguistics. (Chinese version in Linguistics Materials, 1964,
no. 3).
Kotov, A. M. (1986). Functional hierarchy of Chinese syntactic phenomena. Journal
of Fudan University, no. 6.
Langacker, R. W. (1987b). Foundations of cognitive grammar, vol. 1, Theoretical
prerequisites. Stanford, CA: Stanford University.
Li Jinxi (1924/​1992). Newly published Chinese grammar. Beijing: Commercial Press.
Li, Y. H. (1997). Structure and interpretations of nominal expression. Paper presented
at the 9th North American Conference on Chinese Linguistics, Victoria, Canada.
Li Yuming (1986). A new explanation of “nominalization”. Journal of Huazhong
Normal University, no. 3.
Li Yuming (1996). Status of parts of speech of non-​predicate adjective. Studies of
Chinese Language, no. 1.
Li Zuofeng (1995). Notional words of classical Chinese. Beijing: Language & Culture
Press.
Liao Qiuzhong (1991). A review of linguistic categorization: typical examples in lin-
guistic theory. Foreign Linguistics, no. 4.
Liu Danqing (1991). A view of macroscopic study of grammar from the perspective of
Chinese parts of speech. Jiangsu Social Sciences, no. 2.
Liu Danqing (1987). Issues of word and non-​word in contemporary Chinese diction-
aries. Lexicographical Studies, no. 5.
196

196 Bibliography
Liu Danqing (1994). Preliminary exploration of “Weibo words”. Chinese Language
Learning, no. 3.
Liu Danqing (1996). Correlation between part of speech and word length –​the second
comment on a series of discussions on phonetic level of Chinese grammar. Journal
of Nanjing Normal University, no. 2.
Lu Bingfu (1981). Conversional words of verbs and nouns. Lexicographical Studies,
no. 1.
Lu Bingfu (1992). From parts of speech of “dance” and “inevitable” to distinctions
between “suddenly” and “abruptly”. Studies of Language and Linguistics, no. 1.
Lu Bingfu (1993). Core-​derived grammar. Shanghai:  Shanghai Educational
Publishing House.
Lu Bingfu (2010). Considerations of parts of speech and nominalization. In Papers
from the 16th Symposium on Modern Chinese Grammar. Hong Kong:  City
University of Hong Kong Press.
Lu Chuan (1991). Issues of parts of speech in modern Chinese information grammar.
Grammar Research and Exploration, 5th series. Beijing: Language & Culture Press.
Lu Jianming (1980). Grammatical property of “degree adverb + adjective + ‘de (的)’ ”
structure. Language Teaching and Research, no. 2.
Lu Jianming (1982). A discussion on independent use of adverbs in modern Chinese.
Language Teaching and Research, no. 2.
Lu Jianming (1983). Distinctions between attributive and adverbial. Chinese Language
Learning, no. 2.
Lu Jianming (1985). 多 (more) and 少 (less) as attributives. Studies of Chinese Language,
no. 1.
Lu Jianming (1986). A tentative discussion on verb as predicate in modern Chinese.
Series of Chinese Language, 2nd series. Beijing: Foreign Language Teaching and
Research Press.
Lu Jianming (1987). Study of conditions for inserting adjectives in the middle of
numeral-​measure word phrase. Language Teaching and Research, no. 4.
Lu Jianming (1991a). Phenomenon of referent of Versachlichung in modern Chinese.
Studies in Language and Linguistics, no. 1.
Lu Jianming (1991b). A brief comment on time words in modern Chinese. Language
Teaching and Research, no. 1.
Lu Jianming (1991c). Opinion on intransitive verbs in modern Chinese. Grammar
Research and Exploration, 5th series. Beijing: Language & Culture Press.
Lu Jianming (1993a). Chinese grammar study in 1980s. Beijing: Commercial Press.
Lu Jianming (1993b). On dividing Chinese parts of speech. Humanities, 69th and 70th
series, Scientific Institute of Humanities of Yonsei University (Ed.). Seoul: Yonsei
University Press.
Lu Jianming (1994). Issues on conversional words. Studies of Chinese Language, no. 1.
Lu Jiawen (1982). Criteria and levels of dividing modern Chinese parts of speech.
Zhongzhou Academic Journal, no. 6.
Lu Yingshun (1998). Dividing parts of speech –​the combination of generalized morph-
ology with prototype theory. In New ideas of language research. Shanghai: Shanghai
Educational Publishing House.
Lu Zhiwei (1938). Chinese monosyllabic words (mimeographed version by Yenching
University). In Monosyllabic Words of Beijing Dialect (renamed in 1951), Beijing:
People’s Publishing House.
Lu Zhiwei (1957). Chinese word formation (rev. version in 1964). Beijing: Science Press.
 197

Bibliography 197
Lu Zongda (1955). Dividing Chinese words, issues of Chinese parts of speech.
Beijing: Zhonghua Book Company.
Lv Shuxiang (1955). Some principles on Chinese parts of speech, issues of Chinese
parts of speech. Beijing: Zhonghua Book Company.
Lv Shuxiang (1962). On “identification of language units”. Studies of Chinese
Language, no. 11.
Lv Shuxiang (1979). Issues of Chinese grammatical analysis. Beijing: Commercial Press.
Lv Shuxiang & Rao Changrong (1981). A tentative discussion on non-​predicate adjec-
tive. Studies of Chinese Language, no. 2.
Lv Shuxiang & Zhu Dexi (1951). Talks on grammar and rhetoric. The People’s Daily,
June 6 to December 15.
Ma Biao (1994). An attempt to use statistic method to divide parts of speech. Studies
of Chinese Language, no. 5.
Ma Jianzhong (1898/​1983). Ma’s grammar. Beijing: Commercial Press.
Ma Qingzhu (1991). Influence factors of dividing parts of speech and their definition
principles. Grammar Research and Exploration, 5th series. Beijing:  Language &
Culture Press.
Ma Qingzhu (1995). Reference verbs and declarative nouns. Grammar Research and
Exploration, 7th series, Beijing: Commercial Press.
Magnusson, R. (1954). Studies of the theory of parts of speech. Copenhagen:  Ejnar
Munksgaard.
McCawley, J. D. (1992). Justifying parts-​of-​speech assignments in Mandarin Chinese.
Journal of Chinese Linguistics, 20(2), 211–​ 246. (Also in Study of Functional
Grammar of Chinese Language (trans. Zhang Bojiang). Nanchang:  Jiangxi
Education Publishing House).
Mo Pengling (1990). Re-​understanding issues of parts of speech. Journal of Nanjing
Normal University, no. 1.
Mo Pengling & Shan Qing (1985). A statistical analysis of syntactic functions of three
types of notional words. Journal of Nanjing Normal University, no. 2.
Mo Pengling & Wang Zhidong (1988). A preliminary exploration of fuzzy
clustering analysis of words. Journal of Changzhou Institute of Engineering and
Technology, no. 3.
Малдер крылов (1954). Chinese has parts of speech. Studies of Chinese Language, no. 6.
Quirk, R., Greenbaum, S., Leech, G. & Svartvik, J. (1972). A grammar of contem-
porary English. London: Longman.
Quirk, R., Greenbaum, S., Leech, G. & Svartvik, J. (1985). A comprehensive grammar
of the English language. London and New York: Longman.
Ren Ying (2010). Dividing parts of speech:  exploring identity of meanings and
functions  –​essence of parts of speech reconsidered. In Papers from the 16th
Symposium on Modern Chinese Grammar. Hong Kong:  City University of
Hong Kong.
Rozhdestvensky, Ю. В. (1958). Henri Maspero’s view of Chinese grammar. Translation
Series of Language Study, 1st Series (trans. Cao Jing), Department of Chinese,
Nankai University (Ed.). Tianjin: Nankai University Press, 1984.
Schachter, P. (1985). Parts of speech system. In T. Shopen (Ed.), Language typology
and syntactic description. Cambridge: Cambridge University Press.
Shao Jingmin (1995). Analysing coordination valence of disyllabic “v + n” structure.
In Shen Yang & Zheng Dingou (Eds.), Study of modern Chinese valence grammar.
Beijing: Peking University Press.
198

198 Bibliography
Shen Jiaxuan (1997). Tagging patterns of syntactic functions of adjectives. Studies of
Chinese Language, no. 4.
Shen Jiaxuan (2007). Chinese nouns and verbs. Journal of Sino-​Tibetan Languages,
no. 1, 27–​47.
Shen Jiaxuan (2009a). My views of Chinese word classes. Linguistic Sciences, no. 1, 1–​12.
Shen Jiaxuan (2009b). Half a step forward, re-​discussion on Chinese nouns and verbs,
essays on linguistics (3–​22), vol. 40. Beijing: Commercial Press.
Shen Jiaxuan (2015). Typology of parts of speech and Chinese nominalisation.
Contemporary Linguistics, no. 2, 127–​145.
Shen Jiaxuan (2016). Nouns and verbs. Beijing: Commercial Press.
Shen Yang (1996). Dynamic types of compound words in modern Chinese. In Luo
Zhensheng (Ed.), Study of language and character in computer era. Beijing: Tsinghua
University Press.
Shi Anshi (1980). A further discussion on dividing Chinese parts of speech. Language
Study Series, Department of Chinese Linguistics, Nankai University (Ed.).
Tianjin: Tianjin People’s Publishing House.
Shi Anshi & Zhan Renfeng (1988). Introduction to linguistics. Beijing:  Higher
Education Press.
Shi Dingxu (2003). Substantialisation and nominalization of verbs. In Grammatical
Study and Exploration, vol. 12. Beijing: Commercial Press.
Shi Dingxu (2005a). On referential and asserting properties of verb elements. Chinese
Language Learning, no. 4.
Shi Dingxu (2005b). The ambiguity of v-​n constructions and their disambiguation.
Language Teaching and Linguistic Studies, no. 3.
Shi Dingxu (2007). Criteria, methods and results of dividing noun and verb. Language
Teaching and Linguistic Studies, no. 4, 3–​12
Shi Dingxu (2009). Issues on dividing Chinese parts of speech. In Essays on linguistics
(93–​110), vol. 40. Beijing: Commercial Press.
Shi Guangan (1981). The part of speech of “publication” in “the publication of this book”
under the theory of “centripetal structure”. Newsletter of Chinese Language, no. 4.
Shi Guangan (1988). Centripetal and centrifugal structures of modern Chinese.
Studies of Chinese Language, no. 4.
Shi Youwei (1991). Parts of speech:  puzzles of linguistics  –​a preliminary study of
patterns of relational parts of speech. Grammar Research and Exploration, 5th
series. Beijing: Language & Culture Press.
Shi Youwei (1994). Crux of issues of parts of speech and their strategies –​a tentative
treatment of flexibility of Chinese parts of speech. In Hu Mingyang (Ed.), A survey
of issues of parts of speech. Beijing: Beijing Language Institute Press.
Shi Youwei (1997). Parts of speech of 出品 (product) and others  –​a discussion on
Chinese parts of speech. Chinese Teaching in the World, no. 3.
Shi Yuzhi (1992). Symmetry and asymmetry of affirmation and negation. Taipei: Taiwan
Student Book Company.
Simpson, J. M. Y. (1979). A first course in linguistics. Edinburgh: Edinburgh University
Press.
Song Shaonian (1998). Reference and nominalization of predicate constituents of
classical Chinese. In Guo Xiliang (Ed.), Classical Chinese Grammar Series. Beijing:
Language & Culture Press.
Sun Hongkai (1982). History of Derung language. Beijing:  Ethnic Nationality
Publishing House.
 199

Bibliography 199
Takahasi Yasuhiko (1997). Relationship between noun and locative. Chinese Teaching
in the World, no. 1.
Tang Tingchi (1992). Chinese parts of speech:  basis and function of division.
Three series of Chinese Morphology and Syntax. Taipei:  Taiwan Student Book
Company.
Taylor, J. R. (1989). Linguistic categorization: prototypes in linguistic theory (2nd ed.).
Oxford: Clarendon Press, l995.
Taylor, J. R. (1991). Category theory of language, foreign languages and their teaching
(trans. Rong Pei). Journal of Dalian Institute of Foreign Languages, no. 6.
Thompson, S. A. (1988). A discourse approach to cross-​linguistic category “adjec-
tive”. In J. A. Hawkins (Ed.), Explaining language universals (167–​185). Oxford and
New York: Basil Blackwell.
Vogel, P. M. & Comrie, B. (Eds.). (2000). Approaches to the typology of word classes.
Berlin and New York: De Gruyter Mouton.
Wang Hongjun (1994). A view of words and phrases from the perspective of characters
and their groups –​a discussion on Chinese parts of speech division criteria. Studies
of Chinese Language, no. 2.
Wang Hongqi (1991). A review of Chinese parts of speech studies. Logical Language
Writing Series, 4th series. Beijing: Peking University Press.
Wang Li (1943). Modern grammar of China. Beijing: Commercial Press, 1985.
Wang Li (1944). Chinese grammatical theory, collected papers of Wang Li, vol. 1,
Jinan: Shandong Education Press, 1984.
Wang Li (1955). Issues on whether Chinese has parts of speech. Journal of Peking
University, no. 2.
Wang Li (1960). Dividing Chinese notional words. Linguistics Series, 4th series.
Shanghai: Shanghai Educational Publishing House.
Wang Li (1989). History of Chinese grammar. Beijing: Commercial Press.
Wells, R. S. (1947). Immediate constituents. Language, 23, 81–​117. (Also in Language
Materials (1963) (trans. Zhao Shikai), no. 6.)
Wen Lian (1995). Bases and criteria for dividing parts of speech. Studies of Chinese
Language, no. 4.
Wen Lian & Hu Fu (1954). Discussions on dividing parts of speech. Studies of Chinese
Language, nos. 2/​3.
Wetzer, H. (1996). The typology of adjectival predication. Berlin and New  York:  De
Gruyter.
Whaley, L. J. (1996). Introduction to typology:  the unity and diversity of language.
Thousand Oaks, CA: Sage.
Xiang Mengbing (1991). Parts of speech of 出版 (publication) in 这本书的出版
(the publication of this book):  re-​understanding “nominalization” of verbs and
adjectives. Journal of Tianjin Normal University, no. 4.
Xiao Guozheng (1991). A survey of referential uses of object predicate in modern
Chinese. Chinese Language Series, 4th series (ed. Zhang Zhigong). Beijing: Foreign
Language Teaching and Research Press.
Xing Fuyi (1981). Distinctions of parts of speech. Lanzhou: Gansu People’s Publishing
House.
Xing Fuyi (1989). Four points of distinction of parts of speech. Language Teaching
and Research, no. 3.
Xing Fuyi (1991). Thinking about issues of parts of speech. Grammar Research and
Exploration, 5th series. Beijing: Language & Culture Press.
200

200 Bibliography
Xing Fuyi (1997). Chinese grammar. Changchun: Northeast Normal University Press.
Xing Gongwan (1956). Morphological method and modern Chinese word formation.
Journal of Nankai University, no. 2.
Xing Hongbing (1999a). Statistics on uses of modern Chinese parts of speech. Journal
of Zhejiang Normal University, no. 3.
Xing Hongbing (1999b). Analysing words whose parts of speech are difficult to divide and
tag. Computational Linguistics Series (eds. Huang Changning & Dong Zhendong).
Beijing: Tsinghua University Press.
Xu Tongqiang (1994a). “Character” and Chinese syntactic structure. Chinese Teaching
in the World, no. 2.
Xu Tongqiang (1994b). “Characteristics” and methodology of Chinese language
study. Chinese Teaching in the World, no. 3.
Xu Tongqiang (1997). On language. Changchun:  Northeast Normal University
Press.
Xu Shu (1991). Conversional words and problems in dealing with them. Grammar
Research and Exploration, 5th series. Beijing: Language & Culture Press.
Yang Chengkai (1991). Parts of speech division principles and “normalization” of predicates.
Grammar Research and Exploration, 5th series. Beijing: Language & Culture Press.
Yang Chengkai (1992). Typological study of generalized predicative objects. Studies
of Chinese Language, no. 1.
Yang Chengkai (1994). A review of modern Chinese grammatical meta-​theory. Studies
of Language and Linguistics, no. 2.
Yang Chengkai (1996). Theoretical study of Chinese grammar. Shenyang:  Liaoning
Education Press.
Yang Zhenwu (1994). Self-​ reference and transfer reference. Research on Ancient
Chinese Language, no. 3.
Yang Zhenwu (1995). “N’s V” in modern Chinese and “V of N” in classic Chinese”.
Linguistic Research, nos. 2/​3.
Yang Zhenwu (1996). Causes and rules of nominalization of Chinese predicate
constituents. Studies of Chinese Language, no. 1.
Yin Guoguang (1997). Studies of parts of speech in Master Lv’s spring and autumn
annals. Beijing: Huaxia Publishing House.
Yu Min (1955). Morphological change and grammatical environment, issues of
Chinese parts of speech. Beijing: Zhonghua Book Company.
Yuan Yulin (1995). Family similarity of categories of parts of speech. Social Sciences
in China, no. 1.
Yuan Yulin (1998). Analysing Chinese parts of speech based on prototype, cognition
research and computational analysis of language. Beijing: Peking University Press.
Yuan Yulin (2000). Quasi-​axiomatic system of parts of speech in Chinese. Studies of
Language and Linguistics, no. 4.
Yuan Yulin (2005). Fuzzy division of Mandarin words based on membership degree.
Social Sciences in China, no. 1.
Yuan Yulin (2009). Realising parallelism between Chinese and English in grammar –​
on noun/​verb and reference/​statement, subject/​topic, sentence/​utterance segments.
Journal of Sino-​Tibetan Languages, no. 3.
Yuan Yulin (2010). A cognitive investigation and fuzzy division of word-​class in Mandarin
Chinese. Shanghai: Shanghai Educational Publishing House.
Yuan Yulin, Ma Hui, Zhou Ren & Cao Hong (2009). A handbook for dividing Chinese
parts of speech. Beijing: Beijing Language and Culture University Press.
 201

Bibliography 201
Zhan Weidong (1998). On attributive and headword construction of “np of vp”.
Chinese Language Learning, no. 4.
Zhan Weidong (2009). Three questions on parts of speech:  a reflection on
learning and using Chinese parts of speech. In Essays on Linguistics, vol. 40.
Beijing: Commercial Press.
Zhang Bojiang (1993). Constructing “v of n” structure. Studies of Chinese Language,
no. 4.
Zhang Bojiang (1994). Explaining functions of flexible uses of parts of speech. Studies
of Chinese Language, no. 5.
Zhang Bojiang & Fang Mei (1996). A study of functional grammar of Chinese lan-
guage. Nanchang: Jiangxi Education Publishing House.
Zhang Gonggui (1983). Relationship between part of speech and sentence con-
stituent with some problems about parts of speech considered. Journal of Nanjing
University, no. 4.
Zhang Shoukang (1985). Word formation and morphological method. Wuhan: Hubei
Education Press.
Zhang Yujin (2001). Syntax of oracle-​bone inscriptions. Shanghai: Xuelin Press.
Zhao Yuanren (1948/​1995). Grammar of Beijing spoken Chinese (comp. Li Rong).
Beijing: China Youth Publishing House.
Zhao Yuanren (1968a/​1979). A grammar of spoken Chinese (trans. Lv Shuxiang).
Beijing: Commercial Press.
Zhao Yuanren (1968b/​2002). A Grammar of spoken Chinese (rev. ed.). Retranslated by
Ding Bangxin. Hong Kong: Chinese University of Hong Kong Press.
Zhou Ren (2015). Reflections on “multiple categories” in Mandarin’s parts of speech.
Linguistic Sciences, no. 5, 504–​516.
Zhu Dexi (1956). A study of modern Chinese adjectives. Studies of Language and
Linguistics, no. 1.
Zhu Dexi (1960). Speech at the May Fourth Scientific Symposium at Peking University
in 1959. Linguistics Series, 4th series. Shanghai: Shanghai Educational Publishing.
Zhu Dexi (1961). Discussions on “de (的)”. Studies of Chinese Language, no. 12.
Zhu Dexi (1966). About “Discussions on de (的)”. Studies of Chinese Language, no. 1.
Zhu Dexi (1979). Syntactic issues related to the verb 给 (give). Dialect, no. 2.
Zhu Dexi (1982a). Grammatical analysis and grammatical system. Studies of Chinese
Language, no. 1.
Zhu Dexi (1982b). Lectures on grammar. Beijing: Commercial Press.
Zhu Dexi (1982c). Preface to Series of Chinese grammar. In Series of Chinese grammar.
Beijing: Commercial Press.
Zhu Dexi (1983). Self-​reference and transfer reference –​grammatical functions and
semantic functions of Chinese nominalization marker of de (的), zhe (者), suo (所),
and zhi (之). Dialect, no. 1.
Zhu Dexi (1984a). Defining centripetal structure. Studies of Chinese Language, no. 6.
Zhu Dexi (1984b). Distinctions between attributive and adverbial and antithesis of nom-
inal and predicate. Linguistics Series, 13th series. Beijing: Commercial Press.
Zhu Dexi (1985a). Answers to questions on Chinese grammar. Beijing:  Commercial
Press.
Zhu Dexi (1985b). Grammaticised verbs and noun-​verbs in modern written Chinese.
Journal of Peking University, no. 5.
Zhu Dexi (1987). What are objects of study of modern Chinese grammar? Studies of
Chinese Language, no. 5.
202

202 Bibliography
Zhu Dexi (1988). Issues on nouns with verb characteristics in the Pre-​Qin Chinese.
Studies of Chinese Language, no. 2.
Zhu Dexi (1990). A note on distinctions between noun and verb in the Pre-​Qin Chinese.
In Collected Papers in Commemoration of Mr Wang Li. Beijing: Commercial Press.
Zhu Dexi (1991). Word meaning and parts of speech. Grammar Research and
Exploration, 5th series. Beijing: Language & Culture Press.
Zhu Dexi (1993). A view of nominalization of state adjectives from the perspective of
dialect and history. Dialect, no. 2.
Zhu Dexi, Lu Jiawen & Ma Zhen (1961). “Nominalization” of verbs and adjectives,
Journal of Peking University, no. 4.
 203

Index

Note: Page numbers in bold refer to tables and page numbers in italics refer to figures.

AABB reduplication 34–​36 19–​20; questions on 1–​2; selectional

ABAB reduplication 34, 109 restriction of grammatical position
adjective reduplication 35–​36 and 49–​51, 94–​95, 113; selectional
adverb reduplication 37 restriction of grammatical positions
Amis language, transformation of on words as evidence of
expressional functions in 96 49–​51; studies of 3–​5, 13;
Aristotle 92 transfer reference of modification
auxiliaries, as expressional function constituent by extraction in
79–​80 100–​103; transformation of
expressional functions in 96. See also
Bloomfield, Leonard 114–​15 Chinese parts of speech classification;
parts of speech classification
Chen Chengze 5, 9 Chinese parts of speech classification:
Chen Wangdao 5, 10, 130 colloquial modern Chinese and
Chinese Grammar Synopsis (Lv classical Chinese hierarchies and
Shuxiang) 3, 9 24–​27; criteria conditions for 107;
Chinese parts of speech 75; controversy criteria for classifying parts of
on existence of between linguists speech issue 9–​11; de (的)/​di (地) and
41–​49; controversy over relationship 37–​38; difficulties in classification
between syntactic constituents and 2, 188; distinctive words and
4–​9; correspondence between syntactic adverbs overlapping words and 172;
constituent in English.and 174–​75, expressional functions and 82–​85;
175; correspondence between syntactic grammatical function as criteria
constituent.and 176, 176; de (的)/​ for 108–​109, 111, 128–​29, 154–​55;
di (地) functions in Chinese syntax grammatical meaning of words as
103–​104; formation and development criteria for 109–​11; Hocuspocus
of framework for 3–​4; Gao Mingkai vs. God’s Truth School linguists on
critics 44–​47; Gao Mingkai vs. purpose of 51–​52; individual and
KoHpaд on existence of 41–​44; as generalization word grouping for
inhomogeneous system mixed with 27–​28; layer-​by-​layer classification
different hierarchical constituents 13, 12–​13; lexical meaning of a word as
24; internal hierarchies of 13, 24–​27; not meeting criteria for 110–​11; major
lack of morphological change in 2, functions as classification criteria
4, 9, 129–​30; major overlapping parts of
108–​109; multiple functions of speech classification strategies
174–​78, 188; nominalization in 167–​74, 168, 173; marker theory
178–​86; overview and examples of 19, (Croft) and 104–​105, 104–​106;
204

204 Index
morphology as classification criteria major overlapping parts of speech
41–​49, 95, 107–​109; natural classification strategies 167–​74, 168;
classification 52–​53; need for both priority homomorphic classification
distribution and semantic bases for strategy for 159, 160–​64, 188–​89;
89–​90; notional words compatibility priority homomorphic classification
degrees of main grammatical strategy selection criteria 164–​65,
functions 136–​38; notional words 165–​166; real conversional
equivalent function clusters as words defined 158–​59. See also
criteria for 143–​53; noun and adverb nominalization
overlapping words and 171–​72; nouns Croft, W. 104–​105, 105
functioning concurrently as measure
words and 172; operational procedures de (的)/​di (地): functions of in Chinese
for 12–​13; overview of 19; predicate syntax 103–​104; grammatical
and adverb overlapping words functions and classification and 37–​38
and 171; predicate and distinctive discourse function, as part of speech
overlapping words and 170; predicate 91–​92
and substantive overlapping words distribution as classification: clustering
and 167–​69; priority homomorphic analysis 68–​73, 69, 70, 70–​72;
classification strategy for 159, distribution nature theory 73–​74;
160–​64, 188–​89; purpose of as effectiveness of/​limitations of 54–​56,
reflection of attitude towards language 91–​95, 123–​27, 125–​126, 130–​32;
51–​53; selectional restriction of external exclusivity and internal
grammatical position and 49–​51, universality and 58–​60, 63, 122,
94–​95, 113, 187; semantic type and 128–​29, 144; overall similarity
syntactic distribution and 88, 90, clustering view method 68–​73, 69,
187; substantive and distinctive 70, 70–​72; paradoxes of overall
overlapping words and 169–​170; distribution 56–​57; paradoxes of
substantive and predicate overlapping partial distribution view 57–​63,
words and 167–​69; word identity 61–​62; paradoxes of prototype theory
determination and 13–​14, 27–​28. 63–​68, 130; paradoxes of similarity
See also Chinese parts of speech; theory 63; paradoxes of the single-​
distribution as classification; item distribution view 56; possible
expressional function; grammatical classifications obtained with the two
function as classification criteria; parts criteria of object and 很(very) 61–​63,
of speech classification 62; scholar’s studies on distribution
Chinesische Grammatik (Gabelentz) 3 as not essence of parts of speech
clustering analysis 68–​73, 69, 70, 70–​72; 91–​93; scholar’s study on 128–​31, 129;
similarity theory 68–​73, 69–​72, 70 selectional restriction of grammatical
constituents: vacant constituents 99, 99 position and 49–​51, 94–​95, 113, 187;
continuity, between word and phrase 21 selectional restriction of syntactic
conversional words 1, 11–​12, positions on words and 73–​75
188; classification strategies for double-​line system 4–​6
159, 159–​60, 188; classification
strategy selection 159, 161–​64, English 33, 75, 85; correspondence
188; consolidation classification between part of speech and syntactic
strategy selection criteria 164, 165; constituent in Chinese.and 174–​75, 175;
homogeneous classification strategy parts of speech classification in 112;
for 159, 159–​164; homogeneous transfer reference of modification
classification strategy selection constituent by extraction in 100–​103;
criteria 166–​67; homomorphic transformation of expressional
classification strategy for 159, 160–​62; functions in 96; unmarked or marked
homonymous (in its narrow sense) correlations in 105
and heteromorphic (in its broad expressional function 76–​77, 187;
sense) conversional words 158–​59; auxiliaries as 79–​80; bases and criteria
 205

Index 205
for classifying parts of speech 95–​96; of functions and 132–​33; compatible
Chinese parts of speech and 82–​85; degrees among specific grammatical
correlations among part of speech, functions of modification words 146;
syntactic constituents and 104, compatible degrees among specific
104–​106; discourse function and grammatical functions of substantive
91–​92; external exclusivity and words 149; compatible degrees of
internal universality and 58–​60, main grammatical functions of
63, 122, 128–​29, 144; hierarchies Chinese notional words 136–​38;
of expressional functions 81–​82; correlation between compatible degree
intrinsic and extrinsic part of speech and function and word frequency
characteristics and classification 141; correlation between number
93–​95, 188; lexical item differences of conversional words and word
and need for classification and 87–​88; frequency 142; correlation between
modification as 78–​81; part-​of-​speech word frequency and compatible degree
classification system of predicate, (substantive word) 150; de (的)/​di
substantive, predicate modification (地) 37–​38; effectiveness of 123–​27,
and substantive modification words 125–​126, 131; equivalent function
85–​88; patterns of classification clusters as criteria for Chinese
87–​88; possible classifications notional words classification 143–​53;
obtained with the two criteria equivalent function determination
of object and 很(very) 61–​63, 62; and compatibility with 132–​34, 134;
relationships among expressional equivalent function determination
functions 80–​81; selectional restriction methods and rules 134, 134–​35,
of grammatical position and 94–​95; 139–​40, 143; equivalent functions
semantic bases of parts of speech 91; and 131–​32; by generalized or specific
statement and reference as 77–​78, distribution 119–​22; grammatical
80–​82, 100–​103; transformation of function defined 114–​18; grammatical
96–​104; transformation of (de (的)/​ meaning as reason for 123, 131;
di (地) functions in Chinese syntax) lexical meaning and 124; methods of
103–​104; transformation of (lexical classification by 112–​13; morphology
transformation of intrinsic) 97; and 108–​109; selectional restriction of
transformation of (nominalization) grammatical position and 49–​51,
100; transformation of (reference/​self-​ 94–​95, 113; specific function and
reference and transfer reference) general function and 118–​19; word
100–​103, 101; transformation of formation and distribution and
(syntactic transformation) 124–​27, 125–​126
97–​99, 99; transformation of (vacant Guo Rui 139, 155, 181–​82
constituents) 99, 99; universal and
rational grammar and 93 Harris, Z.S. 114–​15
extraction 103 hierarchical classification method 13,
19; colloquial modern Chinese and
finite principle 23–​24 classical Chinese hierarchies and 24–​27
functions of a word: materials used to Hocuspocus group of linguists 51–​52
investigate 38–​39; words with special Hopper, Paul J. 91
uses and 39–​40 Hungarian, transformation of
fuzzy clustering analysis method 11 expressional functions in 96

Gao Mingkai 6, 9, 41–​50, 94, 174 independent application word property

Givón, T. 93 21
God’s Truth School 51–​52 Indo-​European languages 6, 33, 43,
grammatical function as classification 46–​49, 95, 117, 174
criteria 18, 95, 107; in Chinese parts integrity principle: expansion criteria in
of speech classification 108–​109, 21–​22; integrity criteria 22
111, 128–​29, 154–​55; compatibility Italian 108
206

206 Index
Jin Zhaozi 9 syntactic constituent in English and
Chinese.and 174–​75, 175; discourse
KoHpaд, H.И. 41–​44 function and 91–​92; generalization
words and 38; grammatical function
Langacker, R.W. 93 as criteria for 112–​14; Hocuspocus
Latin 116–​17 vs. God’s Truth School linguists on
layer-​by-​layer classification 12–​13 51–​52; intrinsic characteristics vs.
locative reduplication 37 extrinsic characteristics and 93–​94;
Lu Jianming 1–​2 intrinsic expressional function and 95,
Lu Jiawen 12–​13; nominalization and 107, 113–​14; lexical item differences
179–​81 in expressional function and 87–​88;
Lv Shuxiang 3, 5, 9, 24, 47–​49, 128; major functions as classification
intermediate word level system of criteria 129–​30; marker theory (Croft)
classification 7 and 104–​105, 104–​106; meaning
and 92, 109–​11; morphology as
Ma Jianzhong 3, 9, 41, 109 classification criteria 95, 107–​109;
Ma Zhen: nominalization and 179–​81 natural classification, objective natural
Ma’s Grammar (Ma Jianzhong) 3, 5, 9, order and general reference system
41, 109 in 52–​53; need for both distribution
Magnusson, R. 92 and semantic bases for 89–​91, 91;
marker theory (Croft) 104–​105, 104–​106 observation and 107; scholar’s studies
Maspero, N. 41–​42 on distribution as not essence of parts
measure word reduplication 36 of speech 91–​93; selectional restriction
Mo Pengling 13, 130 of grammatical position and 94–​95;
modification, as expressional function syntactic constituent and distribution
78–​79 in narrow sense and 112–​13; time
morphology 41–​49; as classification stability and 93; word meaning (lexical
criteria 41–​49, 95, 107–​109; defined and grammatical) and 109–​111, 113.
108; lack of morphological change in See also Chinese parts of speech
Chinese 2, 4, 9, 108–​109 classification
predicate words classification 167–​69
National Seventh Five-​year Plan prototype theory 11, 63–​68, 130
Research Project 1–​2
natural classification 52–​53 reduplication: AABB reduplication
neighbor linkage hierarchies 69, 69–​70 34–​36; ABAB reduplication 34;
New Chinese Grammar (Li Jinxi) 9 adjective reduplication 35–​36;
nominalization 100, 178–​86. See also adverb reduplication 37; locative
conversional words reduplication 37; measure word
Nootka 51, 103 reduplication 36; noun reduplication
noun reduplication 36 36; numeral reduplication 36;
numeral reduplication 36 state word reduplication 36; verb
reduplication 34–​35
On Chinese Language (KoHpaд) 41 Russian 117
On Latin Language (Varro) 108
onomatopoeia 154 Saussure, Ferdinand de 54–​55
selectional restriction: distribution
parts of speech classification 51; nature theory and 73–​75; of
bases and criteria for 95–​96; grammatical position 49–​51, 94–​95,
comprehensiveness and 107; 113, 187
conditions for criteria for 107; Shan Qing 13, 130
correlations among expressional Shi Youwei 11, 63–​64
function, syntactic constituents and similarity theory 63–​68; clustering
104, 104–​106; correspondence between analysis 68–​73, 69–​72, 70
 207

Index 207
Spanish, transformation of expressional and structural meaning and 28–​30;
functions in 96 individual and generalization word
stand-​by application word properties 21 grouping for classification and
state word reduplication 36 27–​28; lexical transfer reference and
statement and reference, as expressional syntactic transfer reference 30–​32;
function 77–​78, 80–​81 part of speech classification and
substantive words classification 167–​69 generalization words and 38; word
syntactic constituents 4–​9; double-​line identity determination 27–​28; word
system 4–​6; one-​line system 6–​7; identity determination case studies
three-​line system 7–​8 28–​33; word meaning and 28–​33;
syntactic transfer reference 30–​32, 39 word-​formation, morphology and
syntax and 33–​34, 34
Tagalog 51, 103 word parsing: standby application and
Thompson, S.A. 91 independent application and 13
time stability 93 word properties 21, 94–​95; finite
transparency principle 22; functional principle and 23–​24; independent
transparency criteria 23; structural application word property 21; stand-​
transparency criteria 22–​23; by application word properties 21;
transparent meaning criteria 23 transparency principle and 22–​23

US descriptive linguistics 54, 114 Yu Min 44, 44, 45

Yuan Yulin 66
vacant constituents 99, 99
Varro, W.T. 108 Zhang Shoukang 34
verb reduplication 34–​35 Zhu Dexi 1–​2, 61, 76, 96, 167; extraction
concept of 103; grammatical
Wang Li 109–​11, 129 characteristics as criteria for
Wen Lian 95 classification by distribution and
word identity determination 128–​29, 129; nominalization and
12–​14, 24; constituent meaning 179–​81