University of Southern California Dissertations and Theses

Intra-lexical noun-verb dissociations: Evidence from Chinese aphasia

(USC Thesis Other)

Intra-lexical noun-verb dissociations: Evidence from Chinese aphasia

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back of the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.
UMI
A Bell & Howell Information Company
300 North Zeeb Road, Ann Arbor MI 48106-1346 USA
313/761-4700 800/521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
INTRA-LEXICAL NOUN-VERB DISSOCIATIONS:
EVIDENCE FROM CHINESE APHASIA
by
Sylvia Shengyun Chen
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Linguistics)
December 1997
Copyright 1997 Sylvia Shengyun Chen
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 9835107
UMI Microform 9835107
Copyright 1998, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES. CALIFORNIA 90007
This dissertation, written by
S y l v i a S h e n g y u n C h e n
under the direction of h.f.F. Dissertation
Committee, and approved by all its members,
has been presented to and accepted by The
Graduate School, in partial fulfillm ent of re
quirements for the degree of
DOCTOR OF PHILOSOPHY
Dean of Graduate Studies
Date P.'r.9.£! ? . ! ? . £F...^.;...A.?.? 1.
DISSERTATION COMMITTEE
Chairperson
V
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Be strong and of good courage;
be not afraid, neither be thou dismayed:
for the LORD thy God is with thee withersoever thou goest.
Joshua 1:9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Acknowledgments
This dissertation could not possibly be completed without the help from the
following people. First of all, I would like to give special thanks to my advisor
Elaine Andersen for her insightful comments and guiding me patiently in the
organization of the dissertation. I am also very grateful to Daniel Kempler for his
constructive comments on two of my earlier versions. I have also benefited
greatly from Audrey Li for her comments on my discussion of the Chinese
language, and her help for me to obtain a teaching assistantship in the Department
of East Asian Languages and Cultures at USC. Additionally, I thank Larry Hyman
for my scholarship in the first year. Without the financial support, I would not be
able to continue my study and research at USC. For my teachers in the
Department of Linguistics, I thank Bernard Comrie, Maria Luisa Zubizarreta,
Osvaldo Jaeggli, and Mark Seidenberg for their encouragement and discussion. I
am grateful to Pat Clancy for helping me develop the idea of the dissertation. For
years, I have received a great support from friends in the Department of
Linguistics at USC. Miss Laura Reiter helped me process all the petitions and
encouraged me when I was having a hard time. Laura Gonnerman helped me edit
a part of my dissertation. Robert Thornton helped me search the literature and
Miss Mary Dinh rushed the papers to me. I have deep gratitude for them.
I owe my gratitude especially to Elizabeth Bates, whose insights into
Neurolinguistics and cross-linguistic studies have contributed in many ways to the
development of this dissertation and to my view of Neurolinguistics. I also would
like to thank the support of friends from the Center of Research in Languages at
UCSD, in particular Meiti Opie, who always gives me timely help.
I am very grateful to Jian-Ching Mo at Cheng-Chi University, Rev. Joseph
Hsu at Fu-Jen University, Shuan-fan Huang at Taiwan University, Ovid Tzeng at
University of California, Riverside, Daisy Hung at Salk Institute for Biological
Studies, and Jerome L. Packard at University of Illinois for guiding me into the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
IV
field of Neurolinguistics. I thank Dr. Yi-nan Lien for training me in the field of
Speech Rehabilitation at Taiwan University Hospital. His training provides me
with the medical and clinical knowledge that is necessary for my research in
Neurolinguistics. I would like to give special thanks for the assistance of the data
collection from my friends: Miss Hsieh Fu-mei, Cheng Jing-chiu, Hu Mei-xiu
(Taiwan University Hospital), Miss Zhong Yu-mei and Li Shu-er (Taipei VA
Hospital), Miss Lin Li-ying and Chen Xiu-fang (Tri-Military Hospital).
I thank Frank Chang, Hongyi Li, Miaoling Xie, Ingrid Hsu for their support
and help in many ways. My brother Shenghao helped me collect the pictures of
the experiment. My sisters Zhenmin and Shengzhi generously shared their rooms
with me while I was collecting the data in Taipei. I am very grateful to my parents
for taking care of my little “Joy” while I was collecting data. Without their
support, I would not be able to start my dissertation. I thank all my friends
around me in these eleven years for their prayers and encouragement.
Finally, I would like to give thanks to my husband Daniel for being a
companion of my long-suffering. His prayer for my strength from God pushes the
birth of the dissertation. I would like to give my loving hug to my daughters “Joy”
and “Noel” for their warmest love that kept me running to the final race.
I am responsible for all the errors that inevitably remain in the dissertation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table of Contents
1 . Introduction..............................................
v
1
2. The Noun-verb Dissociation in Aphasia........................................ 16
2.1 The Morphological Account........................................ 17
2.2 The Syntactic Account................................................. 21
2.3 The Semantic-Conceptual Account............................. 26
2.4 The Lexical Account.................................................... 32
2.5 Previous Studies on Lexical Impairments of Chinese
Aphasics................................................................................ 36
2.6 Problems with Four Modular Accounts............................... 41
2.7 The Competition Model: Its Compatibility with
Lexical Processing in Chinese ............................................. 48
2.7.1 Semantic Factors...................................................... 50
2.7.2 Grammatical Factors ............................................... 51
2.7.3 Phonological Factors............................................... 53
2.7.4 Frequency ............................................................... 54
2.7.5 Uniqueness............................................................... 55
2.7.6 Cohort Structure...................................................... 56
3. Word Formation in Chinese: Linguistic Evidence for
the Level Analysis....................................................................... 59
3.1 Chinese Compounds: The Dilemma.................................. 61
3 .2 Five Linguistic Properties of Chinese Compounds
and the Interaction of Compounds with Phrases .............. 71
3.2.1 Linguistic Evidence for the Sublexical Level .......... 72
3 .2.1 a Components Are Semantically
Independent................................................. 72
3.2.1 .b Evidence for the Sublexical Level:
Grammatical Categories of Components 76
3.2.1 c Evidence for the Sublexical Level:
Productivity of Compounds....................... 83
3.2.2 Word Structures and Argument Structures ........... 88
3.2.2.a Word Structures......................................... 89
3.2.2.b Argument Structures................................... 92
3.2.3 The Interaction Compounds and Phrases............... 96
3.2.4 Summary................................................................ 97
4. Method......................................................................................... 99
4.1 Subject Selection: Background Information and
Criteria .............................................................................. 100
4.2 Testing Materials............................................................... 101
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
vi
4.3 Testing Procedure............................................................ 103
4.4 Data Analysis.................................................................... 104
4.4.1 Criteria for Noun-Verb Distinction........................ 104
4.4.2 Coding in Frequency Analysis............................... 107
5. Results and Discussion............................................................. 113
5.1 Differences in the Ability to Use Verbs and Nouns:
Lexical and Sublexical....................................................... 114
5.1.1 Evidence from Verbal and Nominal VN
Compounds............................................................ 115
5.1.2 Evidence from NN, NNN, VNN Nominal
Compounds............................................................. 129
5.2 Differences in Error Types: Lexical and Sublexical ......... 137
5.2.1 Error Analysis: Verbal VN Compounds................. 143
5.2.2 Error Analysis: Nominal VN Compounds ............. 147
5.2.3 Homophones: Verbal and Nominal VN
Compounds............................................................. 150
5.2.4 Reversal: Verbal and Nominal VN Compounds 152
5.2.5 Summary................................................................... 153
5.3 Differences in the Use of Alternative Word Types 154
5.3.1 Verbal and Nominal VN Compounds...................... 155
5.3.2 NN, NNN and VNN Nominal Compounds 163
5.3.3 Noun-verb Dissociations and Word Type
Frequency................................................................. 169
5.4 Summary.............................................................................. 173
6. Frequency Effects: Results and Discussion ................................ 175
6.1 Frequency Sensitivity: Whole Words and Components 176
6.1.1 Correct Responses: The Whole Word Level 177
6.1.2 Correct Responses: The Sublexical Level................ 178
6.1.3 Substitutions: The Lexical and Sublexical Levels 181
6.2 Frequency sensitivity. Word Structures .............................. 183
6.2.1 Word Type Analysis: Responses on
Single-Morpheme...................................................... 185
6.2.2. Word Type Analysis: Responses on
Two-morpheme......................................................... 187
6.2.3 Word Type Analysis: Responses on
Three-morpheme....................................................... 188
6.2.4 Word Type Analysis: Interactions Among Three
Word Types.............................................................. 190
6.3 Summary............................................................................... 194
7. General Discussion and Concluding Remarks.............................. 197
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7.1 Chinese Compounds: Unanalyzable or Analyzable............. 197
7.2 Lexical Representations of Chinese compounds:
A Single-factor Lexical Account or a Multi-factor
Connectionist Account........................................................ 199
7.3 The Compatibility of Chinese Facts with the Competition
Model ................................................................................... 201
7.4 A New Approach to the Old Problem................................. 208
7.5 Concluding Remarks ........................................................... 210
Appendix I Subject Information.......................................................... 213
Appendix II Experiment Stimuli: Noun-Verb.................................... 214
Appendix III Experiment Stimuli: Frequency..................................... 219
Appendix IV Sample Pictures of the Action vs.
Object Naming task....................................................... 225
Appendix V Lexical Errors of Broca’s vs. Wernicke’s: Verbal
VN Compounds & Nominal VN Compounds............. 227
References........................................................................................... 246
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Figures
Figure 1 Lexical model of lexical access......................................... 33
Figure 2. Percent production of verbal VN and nominal VN
compounds........................................................................ 117
Figure 3. Percent grammatically correct of VN noun and verb
compounds........................................................................ 119
Figure 4a. Percent correct of the verb vs. noun elements of VN
verbal compound................................................................ 121
Figure 4b. Percent correct of the verb vs. noun elements of VN
nominal compound............................................................. 122
Figure 5a. Lexicalization of the verb vs. noun elements of VN
verbal compound............................................................... 124
Figure 5b. Lexicalization of the verb vs. noun elements of VN
nominal compound.............................................................. 125
Figure 6a. Percent correct vs. lexicalization of NN compound 130
Figure 6b. Percent correct vs. lexicalization of NNN compound 133
Figure 6c. Percent correct vs. lexicalization of VNN compound 134
Figure 7a. Percent correct production of whole words by
frequency............................................................................. 178
Figure 7b. Percent correct production of sublexical elements by
frequency........................................................................... 179
Figure 8a. Percent substitution of whole words by frequency 182
Figure 8b. Percent substitution of sublexical elements by
frequency............................................................................ 183
Figure 9. Bi-directional feeding relation between phrases and
w ords................................................................................. 210
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Tables
Table 1. Word Structures: Nominal Compounds............................... 90
Table 2. Word Structures: Verbal Compounds.................................. 91
Table 3. Nominal Word Structures vs. Verbal Word Structures 108
Table 4. Six Types of Errors: Criteria for Data Coding...................... 139
Table 5. Four Types of Homophones..................................................141
Table 6. Reversal of Components........................................................143
Table 7. Numbers of Component Errors for Six Error Types:
Verbal VNs......................................................................... 146
Table 8. Numbers of Component Errors for Six Error Types:
Nominal VNs.........................................................................149
Table 9. Homophone Substitution Errors of Wernicke's
Aphasics................................................................................ 151
Table 10. Reversal of Components........................................................153
Table 11. Word Type Responses to VN-N & VN-V............................159
Table 12a. Percent of Word Type Responses to NN, NNN,
VNN Nominal Targets..........................................................166
Table 12b. Word Type Substitutions with Verb Pieces on NN,
NNN, and VNN Nominal Targets........................................167
Table 13a. Percent Response on Single-Morpheme Word Type 185
Table 13b. Percent Response on Two-Morpheme Word Type 188
Table 13c. Percent Response on Three-Morpheme Word Type 189
Table 14a. Lexical Variants by Broca’s & Wernicke’s Aphasics 192
Table 14b. Lexical Variants by Normal Controls............................... 193
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Abstract
Studies of inflectional languages have demonstrated that Broca’s aphasics
find it more difficult to produce verbs than nouns, while Wernicke’s aphasics show
the opposite profile. Explanations for this double dissociation include several
single-factor accounts such as grammatical accounts (e.g., verb deficits reflect the
greater morphological or syntactic complexity of verbs), semantic-conceptual
accounts (e.g., verbs are based on action meanings, which are stored in anterior
motor regions; nouns are based on object meanings, which are stored in sensory
cortex), and lexical accounts (verbs and nouns are stored in separate regions of the
brain, independent of their semantic content). In Chinese, many words are
compounds with a complex internal structure, including VN compound verbs like
“SHUI-JIAO” (literally sleep-sleep, meaning “to sleep”) and VN compound nouns
like “FEI-JI” (literally fly-machine, meaning “airplane”). Hence, noun-verb
dissociations between Broca’s aphasics and Wernicke’s aphasics may occur at the
lexical level and the sublexical level. Such sublexical dissociation would provide a
challenge to existing explanations. To explore this possibility, an object and action
naming study was conducted with Chinese Broca’s and Wernicke’s aphasics,
designed to elicit several different compound types (VN nouns, VN verbs, NN
nouns, VNN nouns, and NNN nouns). Results of the present study demonstrate
both lexical and sublexical noun-verb dissociations between Broca’s aphasics and
Wernicke’s aphasics, and provide evidence for the interaction of sublexical noun
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
verb dissociations with word structure frequency, e.g., sublexical noun-verb
dissociations are significant in non-major word types but are insignificant in major
word types. Additionally, a frequency effect is observed in the access of whole
words, components, and word structures, e.g., the more complex two-morpheme
word structure has higher accessibility than the simpler single-morpheme word
structure. The presence of noun-verb dissociations at the lexical and the intra-
lexical levels and their interaction with frequency effects cannot be explained by
any single-factor account discussed above. However, the complex phenomenon
observed in Chinese aphasia can be accounted for by connectionist models such as
the Competition Model, which consider the interaction of relevant factors at both
lexical and intra-lexical levels.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1
Chapter One
Introduction
The phenomenon of noun-verb dissociations has been used as a tool to
understand the organization of the lexicon in the brain. To date, numerous studies
have related this dissociation to two different aphasic syndromes: Broca’s and
Wernicke’s aphasia. Broca’s aphasia is a syndrome involving focal lesion(s) in the
anterior area of the left hemisphere of the brain. Patients with Broca’s aphasia
demonstrate relatively intact comprehension but dysfluent speech, which is very
often accompanied by omissions of grammatical words such as grammatical
markers, predicates, etc. Wernicke’s aphasia is a syndrome involving focal
lesion(s) in the posterior area of the left hemisphere of the brain. Patients with
Wernicke’s aphasia are very poor at comprehension but very fluent in their speech.
The results of studies on these subjects provide useful information concerning the
structure of the lexicon and its relation to other linguistic components such as
syntax. The implications of these studies, however, are limited by their bias
toward a single morphological type, i.e., inflectional languages (such as English
and Italian). This dissertation examines the structure of the lexicon in Chinese, a
language in which morphology is not inflected. It will present data from both
Broca’s and Wernicke’s aphasic patients to support the position that a Chinese
word consists of three levels of information. These levels are: the whole word
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2
level, the word component level, and the word-class substructure level
(henceforth the word structure level). This study proposes that the lexicon is
structured by the interaction of linguistic/cognitive properties (such as
grammatical category and frequency) and intra-lexical levels, which is different
from any proposal based on the studies of inflectional languages. This proposal
brings a different view to the organization of lexicon.
The dissertation addresses intra-lexical noun-verb dissociations in Chinese
aphasia. The term “intra-lexical” refers to different levels of representations within
a compound word. An intra-lexical noun-verb dissociation refers to the noun-verb
dissociation appearing at different levels within a compound word. The noun and
verb categories have long been thought to represent a basic distinction in language
(Sapir, 1921; Givon, 1984; Comrie, 1981). This basic distinction is also observed
in different levels of linguistic representations such as the level of roots (e.g. the
root “eat” of the inflected word “eat-ing” is a verb ), words (lexical nouns such as
“door” and verbs such as “open”), and phrases (NP vs. VP). This distinction is
further supported by language breakdown, where we found that Broca’s aphasics
suffer from verb deficits, while Wernicke’s aphasics suffer from noun deficits.
These lexical impairments provide us very important clues to understanding the
processing of nouns and verbs in differently distributed neural systems. To date,
several studies have reported that aphasics with damage in the anterior part of the
brain (diagnosed as Broca’s aphasia) display specific deficits in the production of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3
main verbs compared with their production of object names (Bates, Chen, Tzeng,
Li, & Opie, 1991; Caramazza & Hillis, 1991; Daniele, Giustolisi, Silveri,
Colosimo, & Gainotti, 1994; Lapointe, 1985; Miceli, Silveri, Villa, & Caramazza,
1984; Miceli, Silveri, Nocentini, & Caramazza, 1988; Glosser, Saykin, &
O’Connor, 1994; Osman-Sagi, 1987; Saffian, Schwartz, & Marin, 1980; Zingeser
& Bemdt, 1990). These patients often omit verbs in their spontaneous speech.
For example, in describing a picture in which a girl is giving flowers to a woman, a
patient omits the verb “to give” to yield an incomplete sentence “The girl is... the
flower” (from Saffran, Schwartz, & Marin, 1980). Alternatively, Broca's aphasics
sometimes replace a verb with the corresponding nominal form, for example,
saying “Bunny...tears” where normal speakers say “Bunny cries” (from Bates,
Friederici, Wulfeck, & Juarez, 1988).
In contrast, a complementary profile has been reported for fluent patients
(involving damage in the posterior area of the left brain), including Wernicke's
aphasics and some anomies. These patients display fewer problems with verbs and
more severe problems in the production of names for common objects in their
spontaneous speech. This often results in an overuse of pronominal forms and
circumlocutionary frames, for example, “this thing here, whatever it is called, it's
crying” (Bates, Chen, Tzeng, Li, & Opie, 1991). This problem with object naming
also causes the production of semantic paraphasias, for example, “baguette”
instead of “bunny” (Bates et al., 1988), and, in more severe cases, the production
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4
of neologisms and jargon. The noun-verb dissociation was originally reported for
spontaneous speech, but it has also been observed in experimental studies of object
and action naming outside of a sentence context (Miceli et al., 1984; Osman-Sagi,
1987). Specifically, when non-fluent and fluent aphasics were asked to name
simple objects and actions, they showed opposite patterns of naming difficulty;
Broca's aphasics made more errors on action-naming items, while Wernicke's and
anomies made more errors in object naming.
What is the basis of noun-verb dissociations? Bates et al. (1991) discuss
four different accounts for the noun-verb problem. These four accounts attribute
the noun-verb problem to a disruption of a specific linguistic component, e.g.
morphology, syntax, lexicon, or semantics. Because these accounts are proposed
to explain noun-verb dissociations observed in inflectional languages, in which
open class morphology are not complex, they all face a problem when applying
them to isolating languages, in which open class morphology are quite complex.
Their problem lies in the assumption of a single lexical entry for all the open class
words, including lexical items (i.e., uninflected words) and sublexical items ( i.e.,
parts of compounds or roots of inflected words). In languages where lexical and
sublexical forms are identical words, it is quite natural to postulate a single lexical
entry for all the lexical and sublexical open class items. In fact, for these cases,
separate lexical entries for lexical and sublexical forms are quite redundant. Take
the verb “start” as an example, the verb “start” may function as a lexical verb (i.e.,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5
an uninflected word) or a sublexical verb (i.e., a root verb of the past tense verb
“started”)- The proposal of the single lexical entry will put the lexical “start” and
the sublexical “start” into one lexical entry. However, the proposal of a dual
lexical entry will redundantly put the verb “start” into two separate lexical entries
depending on its function, i.e., a lexical or a sublexical element.
However, in languages where lexical and sublexical forms are not identical
(e.g. the Chinese language), two separate lexical entries are quite necessary to
represent lexical and sublexical items respectively. Take the Chinese verb “start”
as an example. The verb “start” has two forms in Chinese. One form is a
compound verb “kai-shi” (literally open-start, meaning “to start”); the other is the
component verb “shi” (to start). The component verb “shi” is a part of the
compound verb “kai-shi” (literally open-start, meaning “to start”). The compound
and its component differ from each other concerning their structures. The
compound verb “kai-shi” is a VV from, whereas the component verb “shi” is a
single V form. If we propose a single lexical entry for the lexical verb “kai-shi”
and the sublexical verb “shi,” that is, these two forms of the verb “start” are
represented in a lexical entry, then this proposal will miss the lexical-sublexical
distinction as well as the V-VV distinction between the verb “kai-shi” and the
component “shi.” To capture the co-existence of lexical and sublexical verbs, it is
necessary to postulate dual lexical entries (Bates et al., 1991; 1993).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6
Bates et al. (1991) reported that noun-verb dissociations between Chinese
Broca’s aphasics and Wernicke’s aphasics can appear in the internal structure of a
compound. Their finding argued for the existence of the sublexical level within a
Chinese compound, which was not predicted by the assumption o f single lexical
entry. However, the assumption of single lexical entry may not be a problem for
Chinese compounds because components of compounds may be treated as free
words of phrases. If this is the case, the reported sublexical impairments may just
be a kind of lexical impairments. In this regard, the question of whether
components of compounds are lexical or sublexical has to be answered before
discussing the implication of sublexical noun-verb dissociations on current lexical
theories. The status of components is closely related to the issue of compound-
phrase boundary. A review of the controversy on compound-phrase boundary can
help us to understand the controversy on the status of components.
In Chinese, compounds and phrases are very similar in their forms and their
meanings (see section 3.1). These similarities result in a very vague boundary
between compounds and phrases. Very little consensus has been reached
concerning the primary features which can be used to distinguish compounds from
phrases. Of all the studies on Chinese compounds, Chao (1968) was the first study
which discussed compounds in a very detailed fashion. He exhausted all possible
conditions to define the notion of “compound.” Following his work, Li &
Thompson (1981) summarized all the criteria into three basic criteria:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
7
(i) One or both of the constituents being bound morphemes
(ii) Idiomaticity of the meaning of the entire unit
(iii) Inseparability or limited separability of the constituents
Both criteria (i) and (iii) concern about the “boundedness” of components.
However, there is a dilemma when discussing the “boundedness” of components in
Chinese. In English, bound morphemes and free words can be clearly identified.
No controversy has been reported regarding the boundary between bound
morphemes and free words. For example, “blue” and “bird” are free words either
in the compound “bluebird” or in the phrase “the bird is blue,” whereas “-ed,”
“-ing,” and “-s” are always bound morphemes.
However, this clear boundary between bound morphemes and free words is
not available to Chinese speakers because there are two sets of bound morphemes
in Chinese. One set of bound morphemes are grammatical markers such as the
relative clause marker -de, the aspect marker -le, etc. These markers are like
bound morphemes in English which can be easily distinguished from free words.
Another set of bound morphemes belong to open class morphology. This set of
bound morphemes cause a major difficulty in defining compounds. Take the word
“tai-yang” (literally very-sun, meaning “sun”) as an example. There are two kinds
of morphemes in the word “tai-yang” (literally very-sun, meaning “sun”). The
component “tai” (meaning “very”) is a free word (e.g. tai da le, literally very big
aspect marker, meaning “too big”) which is parallel to “blue’ and “bird” of the
compound “bluebird.” The component “yang” (meaning ‘sun”) is not a free word
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
8
but a bound morpheme. Bound morphemes such as “yan” often cause the
confusion in discussion of compounds.
The confusion arises for two reasons. First, there is no list of bound open
class morphemes offered in any literature on Chinese morphology (including
dictionaries). As a result, to defend for a true bound open class morpheme,
Chinese speakers had to search all the possible cases to show that the open class
morpheme had never been used as free words. Very often these speakers
experienced a hard time in determining a true bound open class morpheme because
it is impossible to exhaust all the cases empirically. Moreover, many bound open
class morphemes may act as free words in a formal speech or writing because the
formal speech or writing is significantly affected by Classical Chinese, where single
morphemes are used as free words. For example, the so-called bound morpheme
verb “shi” (means to start), a sublexical variant of lexical compound “kai-shi”
(literally open-start, meaning “to start”), may be used as a free word in a formal
sentence as shown in example 1 .
1. yuan-zi-dan de fa-ming shi yu rei-dian.
atomic-boom Relative Clause Marker invent start from Sweden
“The invention of atomic bomb started from Sweden.”
Second, compounds which are made up of bound morphemes very often are not
fully integrated. Some syntactic operations such as the insertion of pronouns,
numerals, and classifiers may apply to compounds to yield the separation of
components within a compound. Because constituents of a compound can be
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9
moved, they are no longer called “bound morphemes.” For example, yu-mo
(humor) is a word borrowing from the foreign English word “humor” (Chao,
1968; Li & Thompson, 1981). Bothy** (as a verb) and mo (as a noun) are bound
morphemes that never appear alone in sentences as free words. However, the
compound yu-mo may take pronouns (e.g. “you”) and numerals (e.g. “one”) to
yield a phrase-like expression “yi* ni yi mo” (literally “humor you one humor,”
meaning “ humor you once”). Because “yu” and “/no” act as free words, they
cannot be called “bound morpheme.” The example of “yi*-mo” shows that we
may start with the criterion of “bound morpheme” to define compounds and may
end up with the finding that native speakers re-interpret these bound morphemes as
if these bound morphemes are free words.
Criterion (iii) was offered to include compounds that contain separable
bound morphemes. These compounds, as discussed by Li & Thompson (1981),
allow insertion of variety of lexical items between morphemes. These lexical items
include pronouns, adjectives, directional verbs, resultative verbs, adverbs,
negations, etc. Unfortunately, phrases may take the insertion of the same set of
lexical items as we observe in compounds. Because of this, the criterion “a limited
inseparability” actually makes no prediction concerning differences between
compounds and phrases.
These difficulties in defining the notion of compound lead Li & Thompson
(1981) away from the controversy to make the following claim:
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10
Fortunately, though, the definition of a compound is really
not a crucial issue for students of Mandarin. It is
important only to linguists analyzing the Mandarin lexicon
because it serves to delimit the domain of their studies.
Thus, we may consider as compounds all polysyllabic units
that have certain properties of single words and that can
be analyzed into two or more meaning elements, or
morphemes, even if these morphemes cannot occur
independently in Modem Mandarin. (P. 45-46)
Although Li and Thompson’s advice seems wise, it is not compatible with
what has been reported for aphasia. In their view, compounds are treated as
polysyllabic words by native speakers of Chinese. Therefore, no further
grammatical analysis into the internal structure of a compound should be carried
out by these speakers. If this assumption is correct, we would not expect to see
any sensitivity of native speakers to grammatical category of sublexical items.
That is to say, Broca’s and Wernicke’s aphasics who show lexical noun-verb
dissociations should not show any kind of sublexical noun-verb dissociations in
their production of compounds.
Recently, numerous studies demonstrated that aphasic subjects show
different preferences for grammatical categories of component morphemes and
word structures (Bates et al., 1991; Chen, 1990; Packard, 1990). Since none of
these aphasic subjects are linguists, their knowledge of grammatical categories of
component morphemes and word structures puts the proposal in question that
Chinese speakers process compounds as unanalyzable polysyllabic units.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
11
If compounds are not represented as unanalyzable polysyllabic units, the
question of how they are represented in the lexicon arises immediately. Bates et al.
(1993) discussed two very interesting phenomena that are relevant to criterion (ii).
Their discussion mainly revolved around verbal VN (or VO) compounds, which
are the most controversial ones. First, they argued that verbal VNs may function
as verbal compounds because the N of verbal VN does not carry a significant
semantic function. In Chinese, the VN form is one of the major devices for action
naming; the presence of nouns in the VN form is to fulfill a structural requirement
for verbal compounds. This morphological process is shown in the following
example: When the picture of “sewing a generic piece of fabric” (instead of sewing
clothes) was presented to elicit the action name “to sew,” normal Chinese-speaking
subjects invariably produced the most common VN “feng-yi-fu” (literally sew-
clothes-clothes, to sew) for the action “to sew” rather than a single verb “feng”
(literally sew) or a semantically specific VN “feng-bu” (literally sew-fabric,
meaning ‘to sew a fabric”). Their responses suggest that the noun of VN verb is
not semantically significant in the action naming (Bates et al., 1993).
Second, Bates et al. argued that a two-morpheme expression may carry
two readings. One is a phrase reading, and the other is a compound reading. The
differences are shown in their semantic idiomaticity (the second criterion stated by
Li & Thompson). The two readings are shown in the example jin-yu (gold fish).
The compound reading of jin-yu is goldfish. By contrast, the phrase reading of jin
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
12
yu is golden fish (the fish that is made of gold). Thus, the two morphemes jin yu
may be represented in a configuration, with the lexical compound on the top node
(for the lexical level) and two components at the sublexical level. The very same
expression jin-yu may also be represented in an NP configuration, with an NP on
the top node and two nouns at the lexical level.
Using “semantic idiomaticity” as a primary criterion to define compounds,
Bates et al. (1993) propose that compounds can never be separated, and
components of compounds are always bound morphemes. Moreover, any
expression, which consists of separable constituents, is not a compound but a
phrase. Their proposal provides us with a clear boundary between compounds and
phrases. This allows us to move forward to the discussion of the intra-lexical
activity of compounds.
The present study approaches the compound in terms of different levels of
representations. Lexical and sublexical items, together with word structures
constitute three levels of representations within a compound. The current study
looks at the data from noun-verb dissociations between Broca’s and Wernicke’s
aphasics. The results are then examined in terms of their compatibility with a
connectionist model like the Competition Model (Bates & MacWhinney, 1989;
Bates, Thai, & Marchman, 1991; MacWhinney, 1987).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
13
The discussion of noun-verb dissociations in this thesis is based on data
collected from object and action naming tasks. All the test items are true
compounds, including verbal VN for names of actions (mainly based on the
criterion of idiomaticity). In the discussion of noun-verb effects on word
production, the three levels within a lexical compound are shown to be interacting.
A clear picture of noun-verb dissociations is often obscured by the influence from
other processing factors such as frequency. With this consideration, this research
also includes a study of frequency effect on whole words, components and word
types. Results from the frequency study demonstrate significant frequency effects
on the processing of whole words, components and word structures. This finding
further supports the position that a Chinese compound may serve as an unanalyzed
whole word or may be decomposed into a word structure and components.
The remainder of this dissertation provides both theoretical background
and empirical findings for discussing the noun-verb dissociation. Chapter two
reviews the four accounts for noun-verb dissociations in aphasia. The four
accounts attribute the noun-verb problem to a deficit at the morphological level,
the syntactic level, the semantic-conceptual level and the lexical level. The
discussion of the four accounts provides both theoretical explanations and
empirical findings for each explanation. These four accounts, however, are
challenged by the lexical dysfunction of Chinese aphasics because noun-verb
problems of Chinese aphasics do not involve inflectional errors but involve intra-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14
lexical noun-verb dissociations. These intra-lexical impairments motivate an
explanation consistent with the Competition Model.
Chapter three examines the semantic independence of component
morphemes, the assignment of grammatical category to component morphemes
and lexical productivity to show that component morphemes are not syllables but
stem-like sublexical elements. It further investigates the presence of surface word
structure and argument structure to demonstrate the existence of word structures.
The chapter finishes with a discussion of the interaction between compounds and
phrases. The whole discussion leads to the conclusion that compounds, like
phrases, are made up of lexical items (i.e., component morphemes) and structures
(i.e., word structures).
Chapter four presents the methodology, including information of the
subjects, criteria for subject selection, testing materials, the testing procedure,
criteria for scoring subject responses, and the analysis. Chapter five presents and
discusses results of noun verb dissociations at the lexical, sublexical and word
structure levels. Chapter six presents results and discussion of frequency
sensitivity to lexical compounds, sublexical components, and word structures. The
presence of noun-verb dissociations and frequency effects at the intra-lexical levels
has far reaching implications for the organization of the lexicon. Chapter seven
addresses the implications of these findings. Specifically, findings from the present
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
15
study will challenge the four accounts that assume a single lexical entry for lexical
processing. Because a single lexical entry is not enough to capture these intra-
lexical effects, the present study suggests an alternative explanation that the
lexicon is organized in a cohort-like structure. In this proposal, a lexical item may
be activated as a whole chunk or as separate parts. This proposal is compatible
with the lexical theory proposed by the Competition Model. Very generally, the
Competition Model takes the view that the whole word is stored as distributed
representations with a permeable internal structure, at both the level of form and
the level of meaning. In this model, subcomponents of a complex word can be
activated together or separately, depending on the context (Bates, Thai, &
Marchman, 1991; MacWhinney, 1987; Plunkett & Marchman, 1991; 1993).
In conclusion, with evidence from non-inflected languages like Chinese,
this study argues for a connectionist approach to the structure of lexicon.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
16
Chapter Two
The Noun-verb Dissociation in Aphasia
The noun-verb dissociation is one of the puzzling findings in the literature
on language breakdown. Numerous studies of aphasia across languages have
reported evidence of selective impairment in nouns or in verbs as a result of focal
brain damage. Some patients have greater difficulty in accessing nouns, while
other patients have greater difficulty in accessing verbs. In many cases, this
differential dysfunction of verbs or nouns appears to correspond to different
aphasic syndromes. Broca's aphasics display specific deficits in the production of
main verbs compared with their production of object names. These patients often
omit verbs in their spontaneous speech or replace a verb with the corresponding
nominal form. In contrast, Wernicke’s and anomic patients display fewer problems
with verbs and more severe problems in the production of names for common
objects in their spontaneous speech, resulting in an overuse of pronominal forms,
the production of semantic paraphasias, and, in more severe cases, the production
of neologisms and jargon.
The selective impairment in nouns and in verbs suggests that nouns and
verbs are distinct in the cognitive or linguistic process associated with lexical
representation and access. This, in turn, indicates that noun-verb dissociations
must have a fundamental cognitive or linguistic basis. This chapter addresses both
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
17
theoretical explanations of noun-verb dissociations in aphasia and empirical
findings across languages. The following sections discuss the morphological
account (section 2.1), the syntactic account (section 2.2), the semantic-conceptual
account (section 2.3), and the lexical account (section 2.4). The empirical findings
used to support each account are also presented. Section 2.5 reviews literature on
intra-lexical noun-verb dissociations in Chinese aphasia. These preliminary findings
of noun-verb problems in Chinese aphasia provide a new aspect of noun-verb
dissociations which may challenge the four accounts. The discussion then moves
to integrate several factors from the Competition Model relevant to Chinese data.
2.1 The Morphological Account
The morphological account was proposed based on the correlation
between agrammatism (referring to aphasia involving the disruption of grammar)
and the main-verb problem in Broca’s aphasia. The main argument of the account
is that verbs are particularly difficult for agrammatic aphasics to produce, because
verbs tend to carry a heavy load of grammatical markings in languages like
English.
In the English language literature on grammatical impairments in aphasia,
the omission of grammatical markers and function words are two major symptoms
found in the speech of Broca’s aphasics (Bemdt & Caramazza, 1980; Caramazza
& Bemdt, 1985; Goodglass, 1968; Goodglass, 1976; Goodglass & Berko, 1960;
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
18
Goodglass & Hunt, 1958; Saffran, Schwartz, & Marin, 1980). Uninflected verbs
are common errors found in grammatical marker omissions. Problems with verb
inflections occur not only in English. They are also observed in other inflectional
languages. For example, there have been reports of verb inflection errors in motor
aphasia (i.e., Broca’s aphasia) in Russian (Luria, 1962; Luria & Tsvetkova, 1968;
Tsvetkova & Glozman, 1975) and Broca’s aphasia in Italian (Miceli et al., 1983;
Miceli et al., 1984; Miceli et al., 1988; Miceli & Caramazza, 1988).
Goodglass & Hunt (1958) reported a significant number of inflectional
errors, associated with the -s of the verb form in the speech of Broca’s aphasics.
A problem with inflectional morphology was suggested as a strong indicator for
the syndrome of agrammatism (Goodglass & Berko, 1960). Saffran et al. (1980)
also reported verb deficits when they examined the production of English-speaking
agrammatic patients. They observed two kinds of verb errors: omissions of verb
inflections and substitutions of verb inflections (e.g. “to give” was changed to
“giving”). Errors of inflectional markers were persistent and common in the
speech of agrammatic patients.
A problem with verbs exists in Russian aphasia too. Tsvetkova & Glozman
(1975) reported that Russian motor aphasics made errors in their use of verb
endings for gender, number, and tense/aspect. They also found that the
substitution of an infinitive verb for a conjugated one was very rare in Russian
motor aphasia. However, this type of substitution is very common in English.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19
Additionally, they reported a significant number of errors involving endings of
nouns and adjectives, such as agreement (a masculine subject but a feminine
adjective), number (e.g. in the sentence “This boy is small, not schoolboys, but
small children,” the first noun boy is a singular form while the other two nouns
schoolboys and children are plural forms/ and case endings [e.g. “I told my wife”
(with “wife” marked by an accusative case instead of a dative case)].
Russian and English are similar in some ways but they also differ in other
ways. For instance, they are similar in that both English-speaking and Russian-
speaking motor aphasics err at verb endings. They are different in that English-
speaking aphasic subjects preserve infinitive verbs but are impaired in conjugated
ones, while Russian-speaking aphasics show the opposite pattern. Moreover,
Russian-speaking aphasics also often commit errors of noun endings (because
nouns of Russian are also richly conjugated). English-speaking aphasics do not
encounter this problem because nouns in English do not carry rich endings. Thus,
in Russian, verb and noun endings are equally difficult for aphasics. This argues in
favor of a morphological account for lexical deficits displayed by aphasics.
Italian may be the best language to address the morphological account of
the verb problem. Verbs in Italian always carry inflections, which are far more rich
than nouns. Miceli et al. (1983) reported that two Italian-speaking agrammatic
patients tended to omit inflections of finite verbs and more severely, tended to omit
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
20
whole lexical verbs. The sample speech given by Miceli et al. (1983) demonstrated
errors in verb use:
Q: Come faceva la barba prima? (how did you use to shave?)
A: Macchinetta. (razor.)
Q: Elettrica? (electric?)
A. No, lame. (No, blades.)
Q: Mi descriva meglio. (tell me something more.)
A: Macchinetta (razor), pennello (brush), ...sapone (soap), ...insaponarre (to
lather),...viso (face), poi (then)... macchinetta (razor). (Miceli et al., 1983, p.84)
The speech sample shows that most verbs were omitted in the expression.
The only verb preserved is the infinitive verb “insapponare” which replaced the
target verb “insapono” (I lather). The verb ending problem was also reported in
several other studies by Miceli et al. (1984, 1988) and Miceli & Caramazza (1988).
In these studies, the relative sparing of nouns, but impairing of verbs, was
attributed to the difference in their morphological complexity. Verbs are more
complex than nouns in their morphological structures. The morphological account
can deal with noun-verb dissociations in terms of their morphological complexity
differences.
Because morphological complexity is the major factor in predicting the
severity of lexical impairments, the morphological account makes different
predictions from one language to another concerning the severity of lexical
impairments. For instance, because there is no overt inflectional marking on any
verb in Chinese, the account will predict no significant verb deficit in Chinese.
Moreover, because neither nouns nor verbs are inflected, both word classes do not
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
21
differ in their morphological complexity. The account, therefore, predicts that
neither nouns nor verbs will be impaired. However, as we shall see, Chinese
Broca’s aphasics, like Broca’s aphasics of inflectional languages, still have
problems with verbs. This observation is not predicted by the morphological
account (see discussion in section 2.5).
2.2 The Syntactic Account
The syntactic account was proposed on the basis of the assumption that the
main verb problem in Broca's aphasia is related to a breakdown of the grammatical
component. Unlike the morphological account, the syntactic account suggests that
the verb problem is attributable to the greater syntactic complexity of verbs. In
particular, verbs are syntactically more complex than nouns because verbs
determine the number of arguments and the relation between arguments in a
sentence; nouns function as arguments that are related by the verb (Lapointe,
1985). A variant of this argument can be found in Joanette & Brownell (1990),
who suggest that a deficit in main verbs may be only one manifestation of a more
general problem with predication, a grammatical function that links arguments.
Thus, the main verb problem in Broca's aphasia reflects an inability to assign
grammatical relations.
There are a couple of predictions that follow from the syntactic
explanation. First, it predicts a difficulty in verbs over nouns because of their
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
22
differences in syntactic complexity. It also predicts different accessibility between
verbs which differ in syntactic complexity and an identical accessibility between
verbs which do not differ in their syntactic complexity. Moreover, the explanation
predicts that Broca's aphasics, who suffer from a verb deficit, will demonstrate the
difficulty in activating arguments of the verb or assigning grammatical roles within
a sentence.
Recently, Bemdt, Haendiges, Mitchum, & Sandson (1997a; 1997b)
explored the relationship of verb retrieval to sentence processing by examining the
speech and comprehension of 1 1 chronic aphasic patients (including 5 Broca’s and
2 Wernicke’s, 3 anomic and I transcortical sensory).1 They conducted an
action/object naming task, a sentence formulation task, a story-telling task, a scene
description task and an argument assignment task. Three of their findings were
used to argue for the syntactic account.
Bemdt et al. (1997b) suggested the verb retrieval problem as a cause of
difficulty with sentence production. They presented three different lines of
evidence. First, they showed that two verb-impaired aphasics are very sensitive to
verb cueing. This finding was used as the evidence to argue for the verb retrieval
problem as a cause of difficulty with sentence production.
'One Broca’s patient did not participate in the second experiment of sentence production and
comprehension.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
23
In detail, Bemdt et al. (1997b) used syntactic well-formedness and
semantic well-formedness as two criteria for encoding “complete sentences” when
analyzing patients’ sentence production. Results demonstrated that the two verb-
impaired patients constructed complete sentences (80% and 83% syntactically
well-formed sentences and 70% and 67% semantically informative sentences)
when verb targets were given to them more often than when noun targets were
given (22% and 29% syntactically well-formed sentence and 17% and 8%
semantically informative sentences). By contrast, the noun-impaired patient made
only 45% syntactically well-formed sentences and 27% semantically informative
sentences when he was given target verbs, compared with 83% syntactically well-
formed and 54% semantically informative sentences when he was given noun
targets. Thus, two verb-impaired patients did produce more complete sentences
after the verb cueing.
Another line of evidence came from the story-telling task and the scene
description task. Bemdt et al. observed two sentence patterns in the speech of
patients. One pattern is a complete subject-verb-object sentence; the other is a
verb-object sentence. They also observed that verbs of subject-verb-object
sentence structures tended to be light verbs (i.e., “go,” “get,” “come,” “take,”
“give,” “have,” “do,” and “be”); however, verbs of verb-object patterns were
frequently substantive verbs ( “carry,” “eat,” “read,” “sell”). They suggested that
verbs undermine the production of well-formed sentences. Their explanation is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
24
that simple sentence structures with light verbs are assigned argument structures
by a default strategy because frequencies of these light verbs are often in the upper
range (Bemdt et al., 1997b; Pinker, 1989). Moreover, sentences with light verbs
often carry simple syntactic structures in which verbs do not require inflections.
Sentences based on substantive verbs involve the activation of arguments
surrounding the verbs. When these patients experienced difficulties of verbs, they
also found difficulties in activating arguments of verbs. Subject arguments are
more difficult than object arguments for patients to access because subject
arguments are external arguments of verbs, while objects are internal arguments.
Internal arguments have a closer relation to verbs than external arguments (Bemdt
etal., 1997b; Grimshaw, 1990).
Finally, in the task of argument assignment, Bemdt et al. (1997b)
demonstrated the correlation of verb deficits with the syntactic ability in the
modality of comprehension. They asked aphasic patients to match a picture with
the sentence orally given in the test. Semantically reversible vs. nonreversible
sentences were two types of sentences under investigation (similar studies were
conducted by Bemdt, Mitchum, & Haendiges, 1996; Schwarts, Saffran, Fink,
Myers, & Martin, 1994). All of the patients with selective verb retrieval
impairment failed in sentence-picture matching of reversible sentences. This
suggested a strong association between verb retrieval impairments and
comprehension of sentences.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
25
While Bemdt et al. (1997b) argue for the syntactic role of verbs, there are a
couple of facts that need to be considered. First, light verbs and substantive verbs
are not different regarding their argument structures. Both of them require two
arguments. However, aphasic patients treat the two types of verbs differently.
This suggests that the syntactic complexity should only affect the retrieval of verbs
which frequencies are not in the upper range. Therefore, the account that
considers an interaction of frequency and syntactic complexity provides a better
explanation for differences in performance.
In the naming task, they found that verb deficits and noun deficits may
occur to different types of aphasics (Bemdt et al., 1997a). In the attempt to
explain why verbs are so difficult to Broca’s aphasics (whose syntactic ability is
impaired), the syntactic account leaves the noun deficit unexplained.
Regarding the claim that verbs determine sentence structures, just as the
two-verb impaired aphasics improved their sentence production after the verb
cueing, so too did the noun-impaired patient improve his sentence production after
the noun cueing. Both verb cueing and noun cueing help the production of
complete sentences. This suggests that both verb deficits and noun deficits should
affect sentence production.
Moreover, there are other possibilities behind the verb cueing effect. For
instance, there may be a semantic reason that patients reserve more attention for
finding additional vocabulary items when they do not have to spend their attention
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26
on the hardest work— verb retrieval (because verbs had been provided). In this
case, verb difficulties are not necessarily due to complex argument structures but
could be perceived as problems of retrieving the concept of action (see the
discussion on the semantic-conceptual account). There are reports of agrammatic
speech with only isolated nouns or with noun arguments preceding verbs, such as
NN or NNV (Gleason, Goodglass, Green, Ackerman, & Hyde, 1975). The data
indicated that noun arguments could be realized prior to the verb retrieval or
simply without it. It is possible that verbs enhance the realization of arguments.
However, the relation of a verb to its surrounding arguments is not always uni
directional. Isolated nouns and noun arguments may be activated before verbs are
accessed.
2.3 The Semantic-conceptual Account
This account attributes the noun-verb dissociation to contrasting
impairments at the semantic-conceptual level. Explanations at this level focus on
the different meanings conveyed by nouns and verbs. This approach rests on an
explicit neurological claim. It assumes that the brain regions responsible for the
semantic representations of actions lie closer to the motor cortex (the site of
damage for nonfluent Broca's aphasics); conversely, the representations that
underlie object names involve more input from sensory association areas (the areas
of damage for fluent aphasics).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
27
Indeed, there are a couple of studies demonstrating that the anterior
(motor) cortex participates in the activation of action meaning, while the posterior
(sensory) cortex plays a greater role in object meaning (Brown, Marsh & Smith,
1976, 1979; Damasio, 1989; Damasio, Grabowski, Tranel, Hichwa, Damasio,
1996; Decety, Perani, Jeannerod, Bettinardi, Tadary, Woods, Mazziotta, Fazio,
1994; Luria, 1962; Luria & Tsvetkova, 1968; Martin, Haxby, Laionde, Wiggs &
Ungerleider, 1995; Martin, Wiggs, Ungerleider, Haxby, 1996; Petersen, Posner,
Fox, Mintun, & Raichle, 1988; Wise, Chollet, Hadar, Friston, Hoffher, &
Frackowiak, 1991). The observation by Luria may be the earliest report on a
distinct neurological basis for action naming. Luria reported that patients with
frontal lobe damage anterior to Broca's area experience severe problems in action
naming, a symptom within the syndrome called “dynamic aphasia.”
Another line of evidence came from electrophysiological studies. They
found an anterior and posterior distribution for verbs and nouns, respectively
(Brown, Marsh, & Smith, 1976; 1979). Studies of metabolism in normal subjects
provided neurological support for the semantic-conceptual hypothesis (Damasio et
al., 1996; Martin et al., 1996; Petersen et al., 1988). For example, Petersen et al.
employed a reading test to elicit object names and action names. Different brain
regions became active when subjects read object compared to action names.
Specifically, during the silent reading of object names, activity appeared to be
restricted primarily to occipital and sensory association cortex; the classical
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2 8
language areas (i.e., Broca's and Wernicke's areas) were bypassed altogether.
When subjects were asked to read the same words aloud, the classical language
areas were active. By contrast, when subjects were asked to think of the action
associated with an object name, a region anterior to Broca's area showed a marked
increase in metabolic activity (Petersen et al., 1988). Because the approaches to
elicit object names and action names are different (to read object names but to
think of actions names); it raises a question of whether the activity of different
brain regions may be due to the contrast of lexical processing (reading aloud vs.
silent word generation) rather than action-object difference. Besides, there were
some studies which investigated the neural activation of normal subjects when
these subjects were asked to produce nouns only. Damasio et al.(1996) and
Martin et al. (1996) are two major studies discussed frequently in the literature.
There were also other studies examining the same issue by asking normal subjects
to name actions. Martin et al. (1995), Decety et al. (1994), and Wise et al. (1991)
are three major studies on this issue.
Damasio et al. (1996) reported that the retrieval of names for unique
persons, non-unique animals, and non-unique tools depends in separate regions in
the higher-order cortex of the left temporal lobe. They suggested conceptual
reasons for different neural regions for different kinds of entities used in their
experiment. For example, tools and animals can be distinguished in terms of
manipulability. Tools can be manipulated, but animals are not created for being
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
29
manipulated. Moreover, the distinction between unique persons and non-unique
entities (animals and tools) involves the fine physical characteristics and contextual
linkages of an entity, both which allow the mapping of unique items such as
familiar persons or places. They further stated that varied conceptual
specifications, which result from factors such as physical characteristics and
contextual complexity, are the driving forces and “principles” behind the
differences in location of the lexical intermediary regions identified in their study.
For instance, the conceptual description of a manipulate tool contains multiple
sensory channels (somatosensory, visual) and the hand motor patterns. The
activation of a manipulate tool was found to involve the region close to the
regions involved in visual motion (the left temporal-parietal) and hand motion
processing (the inferotemporal region). By contrast, when retrieving names of
persons, the lateral, inferior and mesial aspects of temporal-parietal lobe become
active because these regions are responsible for processing the fine physical
characteristics and contextual linkages of an entity.
Another study by Martin et al. (1996), which focused on the association of
different brain regions with naming animals and tools, demonstrated both
similarities and differences between these two naming tasks. Regarding similarity,
naming pictures of both animals and tools was associated with bilateral activation
of the ventral temporal lobes and Broca’s area. However, naming animals vs.
naming tools also involved the activation of different brain regions. In this study,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30
naming animals selectively activated the left medial occipital lobe~a region
involved in the earliest stages of visual processing. In contrast, naming tools
selectively activated a left premotor area and the left middle temporal gyms
(Martin et al., 1996).
The two studies reviewed above emphasized the relation of object names to
the activation of brain regions. The following three studies emphasized the relation
of action names to the activation of brain regions. Studies of these contrasting
topics enable us to make comparisons between nouns and verbs regarding the
activation of brain regions.
Wise et al. (1991) studied different regions responsible for nouns and verbs
using lexical judgments (the judgment between a noun-noun pair or a noun-verb
pair) and verb generations in which subjects were asked to think of, without
vocalizing, as many verbs appropriate to the presented noun. They observed two
brain regions responsible for the verb generation. One was the left posterior
superior temporal association cortex; the other involved the left promoter and
prefrontal cortex (including Broca’s area and the supplementary motor area).
The activation of the temporal lobe was also reported in Martin et al.
(1995). Specifically, a region in the middle temporal gyms located just in the
anterior to the area involved in the perception of motion became active when
subjects were asked to generate the name of an action associated with the object
(e.g. write when shown a picture of “pencil”). In the same task, they observed an
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
31
additional activation in the left inferior frontal lobe (Broca’s area) when subjects
attempted to name the action.
Decety et al. (1994) observed that different brain regions became active
when subjects were asked to take contrast tasks. One task asked subjects only to
perceive movements of an “alien” hand; the other asked them to imagine grasping
objects with their own hands. During the passive observation of hand movement,
the activation was mainly found in the visual cortical area. The activation was also
observed in the subcortical area involved in motor behavior such as basal ganglia
and the cerebellum. By contrast, during the motor imagination, inferior parts of
the frontal gyrus at the cortical level in both hemispheres were strongly activated.
Moreover, at the subcortical level, the caudate nucleus was found to be activated
on both sides. These areas are related to motor preparation and programming.
In summary, these five studies reviewed above show a neural contrast
between the activation of nouns and verbs. All these studies made use of items
which are in contrast with each other regarding semantic properties such as tools
vs. animals. Their results demonstrated a neural basis for the claim that noun-verb
differences and differences within nouns (or verbs) are attributable to conceptual
differences. Because the semantic-conceptual account is independent of any
morphological form used to convey noun and verb meanings, it leads to the
prediction that the noun-verb dissociation will be found in all natural languages
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32
including Chinese, a language in which the noun-verb distinction is not reflected in
inflectional morphology.
2.4 The Lexical Account
The lexical account attributed the noun-verb dissociation observed in fluent
and nonfluent aphasics to a breakdown in processing that is located entirely within
the lexicon (Caramazza & Hillis, 1990; 1991; Miceli et al., 1984; Miceli et al.,
1988). In this view, language processing is structured as a modular functional
model, where semantic and grammatical categories are located in separate function
modules. Within this model, the lexical representations of words are listed
repeatedly in modality-specific input and output lexicons; grammatical category
information for each lexical item is also represented repeatedly. Figure 1 diagrams
the model proposed by the lexical account (adapted from Caramazza & Hillis,
1990, p. 95).
Caramazza and his colleagues reported different kinds of selective lexical
dysfunction to support their model. First, Miceli et al. (1984) showed two types of
main verb production difficulties: the omission and nominalization of verbs. They
argued that the main verb problem is a lexical processing deficit unrelated to
semantic deficits or syntactic deficits. The semantic account was ruled out for two
reasons: the nominalization of verbs and uneven ratio of verb to noun in
agrammatic patients. They argued that nominalizations of verbs are semantically
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
33
related but lexically inappropriate forms, suggesting a preservation of semantic
representations.
Figure 1 Lexical model of lexical access.
Visual Input
I
Visual Processing
I
Auditory Input
1
Auditory Processing
I
Orthographic Input Phonological Input
Lexicon Lexicon
Verbs Nouns Verbs Nouns
Lexical - Semantic
System
Verbs Nouns Verbs Nouns
Orthographic Output Phonological Output
Lexicon Lexicon
Moreover, they reasoned that if the verb deficit is caused by a dysfunction
at the lexical-semantic level, the proportion of overall (correct and incorrect) noun
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
34
and verb responses to objects and actions should be approximately the same for
agrammatic, for anomic patients and for normal controls. In contrast, if the deficit
is a lexical impairment affected strictly by class information, we should expect
agrammatic patients to produce relatively fewer (correct and incorrect) verb
responses than either anomic or normal subjects in naming actions; yet, they should
produce the same number of noun responses as the other groups when trying to
name objects (Miceli et al., 1984, pp. 214-215). Results of their analysis showed
an uneven ratio of verbs and nouns in the overall verb and noun responses.
Agrammatic patients produced fewer (correct & incorrect) verb responses than
either anomic or normal subjects in naming actions, while the same agrammatic
group produced the same number of nouns as the other groups when trying to
name objects. Based on their findings, they argued that agrammatic patients were
not different from anomic and normal subjects at the semantic level. Their
differences resulted entirely from the difficulty for agrammatic patients to access
verb forms.
There are other cases showing dissociations of phonological input and
output within noun-verb dissociations. For example, some patients showed the
noun-verb dissociation only in production while others showed it only in
comprehension, suggesting that nouns and verbs were represented repeatedly in
four separate lexicons: a noun input lexicon, a verb input lexicon, a noun output
lexicon, and a verb output lexicon (Miceli et al., 1988). Similarly, Caramazza &
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
35
Hillis (1991) and Hillis & Caramazza (1994) described cases that showed noun-
verb dissociations in either speaking or writing only, leading them to propose
separate lexicons for nouns and verbs in orthographic or phonological modalities
(see also Bemdt & Zingeser, 1991).
The noun-verb dissociation has been seen as evidence for the proposal that
lexical representations may be organized in the lexicon by their grammatical
classes. This idea is compatible with reports of category-specific lexical and
morphological deficits; including the claim that derivational and inflectional
morphology can be selectively dissociated in aphasia (Miceli & Caramazza, 1988),
reports of patients who experience selective impairments to specific lexical
categories such as fruits and vegetables (Hart, Bemdt, & Caramazza, 1985), and
the patient whose knowledge of living things (but not objects) is spared in the
visual world but gravely impaired in the verbal domain (McCarthy & Warrington,
1985).
Evidence for each account is primarily based on languages in which
morphology is highly inflected. It would be interesting to know if these accounts
are applicable to languages without inflectional morphology. The Chinese
language provides a test case to this inquiry. Numerous studies are available on
lexical impairments in Chinese aphasia. A review of previous studies gives us
some ideas regarding the problems that these accounts may encounter.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
36
2.5 Previous Studies on Lexical Impairments of Chinese aphasics
The problem with closed class morphology has been one of the major
concerns in the literature on lexical impairments in aphasia. When examining the
Chinese language, deficits with closed class morphology are not sufficient to depict
the entire picture of lexical impairments from that aphasics may suffer. The major
reason for this insufficiency is that grammatical morphology of Chinese is very
opaque. There is no grammatical marker to indicate gender, number, or case. No
grammatical agreement is necessary in this language. Its closed class morphology
mainly consists of grammatical markers for verbs (such as the aspect marker -le,
the experiential marker -guo, the progressive suffix -zhe and the affix zai-, etc.),
the modifier -de (functioning as a relative clause maker such as “that,” an adjective
marker, a possessive marker, or nominalizer), grammatical markers for nouns
(such as the diminutive marker -zi), prepositions (also called co-verbs, because
they co-exist with verbs, with the same phonological and written forms), and
conjunctions.
All these grammatical markers are derived from open class words
diachronically and co-exist with their open class counterparts synchronically. For
example, the aspect marker -le co-exists with the verb “liao” (carrying the third
tone), which indicates the meaning “to complete.” All open-closed class pairs
share the same written forms, and some pairs even share the same phonological
forms. Prepositions and their counterpart verbs are examples of this kind. Tones
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
37
and vowels of closed class items are often neutralized if they are not
phonologically identical with their open class counterparts. The aspect marker -le
and the verb “liao” are examples of this case (Chao, 1968; Li & Thompson, 1980;
Ren, 1980; Wang, 1958; 1962; Zhu, 1980).
Given how impoverished morphology is in Chinese, it is possible that
Chinese aphasics suffer relatively less grammatical breakdown because they do not
have to process a large number of inflections. Empirical studies of Chinese aphasia
did not find such advantage in Chinese aphasia. Instead, numerous studies of
language breakdown in Chinese aphasia report lexical deficits in closed class items
as well as in open class items (Chen, 1984; 1990a; Chen, Tzeng, and Bates, 1990;
1992; Packard, 1990). For example, using a reading aloud task, Chen (1984)
observed that Broca’s aphasics, compared with normal controls, were less able to
read closed class items aloud. At the same time, Chen also observed substantial
omissions or substitutions of open class word parts. Further studies of
grammatical deficits of Chinese aphasics reported findings similar to Chen (1984)
concerning their impairments in grammatical markers and word parts (Chen,
1990a; 1993; Packard, 1990).
The problem with word parts reflects a language-specific morphological
property of the Chinese language. In this language, a word may consist of
meaningful components. Both the whole word and word parts carry their own
meanings and grammatical categories. For example, the whole word “fei-ji”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
38
(literally fly-machine, meaning “airplane”) is an object noun, which is composed of
a verbal component “fei” (to fly) and a nominal component “ji” (machine). This
example shows that the whole word and its components are different in their
grammatical categories and meanings. The difference of grammatical category
between a whole word and its components is particularly interesting for the
discussion of the relationship of verb deficits to grammatical breakdown.
In studying the language production of Chinese aphasics, Chen (1990a)
presented a Chinese agrammatic Broca’s aphasic who regularly omitted or
nominalized whole words as well as verbal components of compound words. For
example, the patient wrote a sentence “[ feij-ji tian-kong [fei] (literally [fly]-
machine sky-space [fly] (words in brackets are omitted ones),” in which the lexical
verb” fei” and the word component verb “fei” (to fly) in the nominal VN
compound “fei-ji” (literally flying-machine, meaning “airplane”) were omitted
simultaneously. The co-existence of sublexical and lexical verb errors suggested
that a central verb deficit may project to the lexical level as well as the sublexical
level. Packard (1990) also reported another form of sublexical errors involving
substitutions of nominal word parts. Sublexical errors were observed in speaking,
writing (including spontaneous writing and dictation), and reading aloud (Chen,
1984; 1990a).
Both Chen (1990a) and Packard (1990) investigated only a single case of
agrammatic Broca’s aphasic. To provide a detailed picture of noun-verb
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39
dissociations in Chinese aphasia, Bates et al. (1991) conducted a study of action
and object naming by investigating 10 Broca’s aphasics and 10 Wernicke’s
aphasics. Two major findings in their studies spoke directly to the issues raised by
the four accounts. First, they found a significant double dissociation between
Broca's aphasics and Wernicke's aphasics at the whole word level (for complex and
simple word forms), with Broca's performing better on nouns than verbs while
Wernicke's showed the opposite pattern. Second, they also found a dissociation
between nouns and verbs inside compound words at the level of component
morphemes (i.e., Broca's found it easier to lexicalize the nominal portion of a VN
compound, while Wernicke's found it easier to lexicalize the verbal component in
the same compound words).
Additionally, they also reported a preference for two-morpheme word
structures across subject groups. This preference for the two-morpheme word
structure was also reported in Chen, Andersen, Kempler, and Bates (1992) with a
similar object/action naming task. Unlike the study of Bates et al. (1991), Chen et
al. (1992) carefully controlled frequencies of testing items in their study by
providing high- vs. low-frequency whole words, high- vs. low-frequency
compound components, and three different word types (single morpheme, two-
morpheme and three-morphemes). They observed that all the subjects (three
Broca’s, two Wernicke’s, and two normal controls) tended to replace low-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
frequency items with high-frequency items. This frequency effect was significant
on both whole words and word components.
Chen et al. (1992) also reported a very strong tendency for all the subjects
to produce two-morpheme names for both action and object naming tasks— a
finding which was compatible with Bates et al. (1991). According to the Chinese
frequency dictionary (by Institute of Language Teaching and Research, 1985), the
two-morpheme word structure is the most frequent word type among all the
possible word types. Thus, the preference for the two-morpheme word structure
reflects a frequency effect on the access of word structure. This frequency effect
on word type retrieval does not directly relate to noun-verb dissociations.
However, it showed a significant role of word structure in lexical access. In
particular, it suggested that word structures, like lexical items, may be affected by
frequency independently.
Moreover, since each morpheme carries its own grammatical category,
these morphemes may be used to construct word-class substructures such as NN,
VN, W , etc. Because word-class substructures carry grammatical categories, it is
very likely that, like whole words and components, they are subject to noun-verb
dissociations. In addition to noun-verb dissociations at the whole word and the
word component levels, the noun-verb dissociation at the word structure level
(referring to word-class substructure level) will be investigated in the present
study.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
41
2.6 Problems with Four Modular Accounts
Two major findings in these previous studies of lexical impairments in
Chinese aphasics speak to the problems raised for the four modular accounts.
First, there is a significant double dissociation between Broca's aphasics and
Wernicke's aphasics at the whole word level (for complex and simple word forms),
with Broca's aphasics performing better on nouns than verbs, while Wernicke's
aphasics showed the opposite pattern. Second, there is also a dissociation at the
level of component morphemes between nouns and verbs inside compound words.
That is, Broca's aphasics found it easier to lexicalize the nominal portion of a VN
compound, while Wernicke's aphasics found it easier to lexicalize the verbal
component in the same compound words. These findings led Bates et al. (1991) to
draw conclusions summarized in the following:
(i) The morphological account has to be rejected because a double dissociation
between nouns and verbs is maintained in spite of the lack of grammatical
morphology in Chinese.
(ii) Any lexical account which assumes a single level of lexical representation has
to be revised, because it cannot account for the double dissociation observed at the
sublexical level (i.e., inside of VN compounds). However, the following
possibilities remain open.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
42
(iii) A semantic-conceptual account cannot be rejected at this point, although it is
clear that such an account must be extended to the meanings of individual
components within a compound.
(iv) It may be possible to entertain a revised lexical account, in which individual
components of compound words are listed separately in the lexicon. This is a
plausible alternative, but (a) it flies in the face of evidence suggesting that normal
Chinese subjects treat high-frequency compounds as whole words (Tzeng, in
preparation; Huang & Hsieh, 1989), and (b) it may constitute nothing more than a
restatement of the data.
These brief summaries highlight major problems of the four accounts when
applying them to the Chinese language. These problems may also extend to other
languages, such as Hungarian. The following discussion elaborates these problems
by considering cross-linguistic data.
The first problem with the morphological account involves the assumption
that verbs are more morphologically complex than nouns. This is a reasonable
assumption for languages like English or Italian, but it runs into problems in
Hungarian (where nouns carry as many morphological markings as verbs) and
Chinese (where there are no inflectional morphemes of any kind on nouns or
verbs). Osman-Sagi (1987) has shown that the double dissociation in object and
action naming can be observed in Hungarian, and Bates et al. (1991) have
demonstrated an equivalent double dissociation for Chinese. Hence it appears that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
43
the double dissociation cannot be explained in terms of the “morphological load”
carried by verbs.
The syntactic account could deal with the Hungarian and Chinese data,
since all languages have verbs that vary in their logical complexity. However, this
model has difficulty accounting for the fact that Broca's aphasics tend to preserve
the canonical order of words in their sentence production (Bates et al., 1988;
Chen, 1993; Tzeng & Chen, 1988). Although Bemdt et al. (1997b) argued a
causal relation between the accessibility of the main verb and access of arguments
surrounding the main verb, this causal relation had to allow an exception that
aphasic patients were able to produce well-formed sentences with “light” verbs as
main verbs of sentences. These sentences all bear appropriate arguments. The
ability to produce canonical word order and well-formed sentences (with light
verbs) means that (at some level) Broca’s aphasics are able to fill arguments and
assign positions in surface structure. Similarly, Shapiro & Levine (1990)
demonstrated that Broca's aphasics showed the normal effects of verb complexity
in sentence comprehension. This suggested that they did not have any difficulty
filling the arguments of a verb. Also, as I have pointed out, even if the verb
argument account can explain the verb deficit demonstrated by Broca's aphasics, it
provides no explanation for the opposite profile (the noun deficit) displayed by
fluent patients.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
44
Additionally, both the morphological and the syntactic accounts have one
problem in common. This problem has to do with the causal asymmetry of the
morphological and the syntactic accounts. While both of these models can account
for the verb deficit in non-fluent patients, they leave the noun problem in fluent
patients unexplained. Presumably, non-fluent patients' verb difficulty is caused by
agrammatism, while the fluent patients' problem with nouns is attributed to some
kind of (unrelated) semantic-conceptual deficit which (for unexplained reasons)
does not affect verbs to the same extent.
The last, deeper problem revolves around the size and nature of the unit
that is stored in a putative verb or noun lexicon (across or within modalities). This
problem is illustrated particularly well in Chinese. Because Chinese has no
inflectional or derivational morphology, it relies heavily on compounding to create
complex words.2 In fact, the overwhelming majority of Chinese words are
composed of two or more sublexical elements. With very few exceptions, each of
these individual elements has its own meaning and each is expressed as a single
syllable in the spoken language and a single character in the written language.
When these individual elements enter into a compound, their meaning often
undergoes modification, but there are no derivational changes in their physical
shape. This contrasts significantly with derivational processes in a language like
English, where complex words are built up using derivational markers like “-ment”
2 For an extensive discussion, see Bates et al. (1991).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
45
in “government,” or “-er” in “bottle-washer.” Compounds formed with two open
class morphemes are common in Chinese, and in many cases the compound itself
and the open-class morphemes that it contains belong to different grammatical
categories. For example, the compound “fei-ji” (literally fly-machine, meaning
“airplane” ) is a VN noun or the compound “shui-jiao” (literally sleep-sleep,
meaning “to sleep”) is a VN verb. Hence, category membership must be assigned
at two levels: the whole word level and the level of word components. If the
respective noun and verb lexicons are stored in separate parts of the brain, where
are VN compounds stored? Is a VN verb stored in the verb lexicon and a VN
noun in the noun lexicon? Or is the verb element stored in the verb lexicon while
the noun element is stored somewhere else? Should the lexical account be recast
at the sublexical level or at both the lexical and sublexical level? If we also allow
for the possibility of modality-specific lexicons, then we have to deal with a very
large number of separate lexical stores — a graphemic output sublexical verb store,
a graphemic output sublexical noun store, a graphemic output whole word verb
store, and so on, to the point where the Law of Parsimony really has been
stretched to its limits.
The same problem can be extended to the semantic-conceptual account. If
verbs are more affected in anterior patients because they draw heavily on “motor
meanings,” then what should happen to noun compounds that contain a verbal
element? Conversely, if nouns are more affected in posterior patients because they
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
46
draw heavily on “perceptual meanings,” then what consequences follow for verb
compounds that contain a nominal element? The semantic-conceptual account has
an advantage over the lexical account, in so far as it has some independent
motivation (e.g. there are good neuro-anatomical reasons why verb meanings
might be stored in the anterior cortex, in and around the motor strip). But such
stores will necessarily proliferate if we need to separate representations for lexical
and sublexical components.
The issue becomes even more complex because information about word
structures may impact lexical access independently. Word structures reflect the
grammatical categories of the stems within a word. For example, a nominal
compound often carries a Noun-Noun word structure; by contrast, a verbal
compound frequently carries a Verb-Noun word structure or a Verb-Verb
structure. These abstract structures may be isolated from semantic content and
phonological representations to constitute an independent factor. To handle the
interaction of lexical elements, sublexical elements, and word structures in lexical
access, we cannot, for example, simply replace a whole word theory (e.g. whole
nouns are stored in the posterior cortex while whole verbs are stored toward the
front) with a purely compositional theory (e.g. noun morphemes are stored in the
posterior cortex while verb morphemes are stored toward the front, with all
compound words created “on-line”).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
47
The attempt to generalize complex noun-verb dissociations to a single
factor (such as a morphological deficit, a syntactic deficit, a lexical deficit, or a
semantic-conceptual deficit) cannot fully capture the facts of noun-verb
dissociations in Chinese. Instead, an alternative multi-factorial account such as
that proposed by Bates & MacWhinney (1989) in their Competition Model seems
called for. The Competition Model has an advantage over the four single-factor
accounts in capturing the “size” and the “nature” of the lexicon because it
considers all the relevant processing factors and evaluates their impact on lexical
processing both within a language and across languages. For instance, within a
language, the factor “frequency” may play a significant role in predicting the
accessibility of open-class words. However, it plays no role in explaining the
problem of closed-class words because all these words are high-frequency items
(Gordon & Caramazza, 1982). Across languages, a factor may be crucial to one
language but totally irrelevant to another. Take the inflectional load as an
example. The inflectional load may influence lexical processing in languages which
have a rich inflectional system. In these languages, lexical problems may be
attributed to the overload of inflectional markings. However, the inflectional load
may be entirely irrelevant to languages which lack inflections. In cases like this, it
is inappropriate to attribute lexical problems to the inflectional load.
To date, the Competition Model has been successfully applied to many
languages including English, Italian, Turkish, Hungarian, Hindi, Serbo-Croatian,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
48
German, and Spanish (Bates & Wulfeck, 1989a; 1989b; also see papers in Brain
and Language, 1991). For each language the weights of factors are ranked in
different hierarchies. Unlike these other languages, Chinese has a morphology
which is not inflected. The impact of every individual factor proposed by the
Competition Model must be different for Chinese from for these inflectional
languages. Section 2.7 first provides a general introduction of the Competition
Model and its application to the noun-verb dissociation. It then moves to specific
factors of the Competition Model which are potentially relevant to the case of
Chinese.
2.7 The Competition Model: Its Compatibility with Lexical Processing in
Chinese
The Competition Model assumes a single lexicon for all the word classes
and no qualitative difference between these lexical items (Bates & Wulfeck, 1989b;
Bates & MacWhinney, 1989). A lexical item, whether it is a content word or a
grammatical morpheme, may serve as a cue to sentence meaning. The accessibility
of a cue is determined by its validity for the sentence meaning and its detectability.
The validity of an item is determined by its availability and its reliability. An item is
maximally high in availability if it is there whenever you need it. For example, both
Italian and English require agreement between the subject and verb; yet the
inflection for the agreement system has much higher availability to Italian-speaking
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
49
subjects than to English-speaking subjects because the agreement between the
subject and verb is more frequently inflected in Italian but less frequently inflected
in English. If the item always leads you to the correct conclusion when you rely on
it, it is maximal in its reliability. For example, in English, the inflectional
morpheme “-s” is a reliable cue to indicate that the subject is a third person and
singular noun. By contrast, the past tense “-ed” is not a reliable cue because the
inflectional marker “-ed” goes along with subjects which may differ in persons and
numbers. In real time processing, several available items, which are similar in form
or function but differ in their availability and reliability, can be activated at any
given time. The speaker/listener has to make a choice out of these items
depending on their availability and reliability. The validity of a cue (both its
availability and reliability) is a strong predictor for the cue selection.
In applying the Competition Model to noun-verb dissociations, both nouns
and verbs are cues to sentence meaning. Their accessibility is determined by their
validity, which is determined by an array of processing factors, including
grammatical properties, semantic content, phonological saliency, frequency,
uniqueness, cohort structure, assignability, etc. Of these processing factors,
semantic content, grammatical complexity, phonological, frequency, uniqueness,
and lexical cohort structure are of particular relevance to noun-verb dissociations
in Chinese.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
50
2.7.1 Semantic Factors
Semantic factors refer to the semantic content of lexical items. The role of
semantic content in lexical processing is best characterized in terms of different
accessibility between open class items and closed class items. Because open class
items bear more semantic content than closed class items, their accessibility among
aphasic and normal subjects is often higher than the accessibility of closed class
items (Bates, Friederici, & Wulfeck, 1987a; Bates, Friederici, & Wulfeck, 1987b).
In other words, two lexical sets may differ in their accessibility because of their
differences in semantic properties.
Turning to the noun-verb contrast, nouns and verbs differ in their semantic
properties. For example, concrete nouns often refer to objects, while verbs refer
to actions. The semantic-conceptual account provided neurological evidence for
this semantic-conceptual contrast. The account presented findings that different
brain regions became active when normal subjects were asked to think about
objects and actions associated with target words (Petersen et al. 1988; Damasio et
al., 1996; Martin et al., 1995; Martin et al., 1996). Their findings suggest that
semantic factors may play a significant role in processing nouns and verbs when
focal brain damage occurs.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
51
2.7.2 Grammatical Factors
The grammatical basis of noun-verb differences is often addressed in two
aspects: morphological structure complexity and argument structure complexity.3
In many languages, nouns and verbs differ in their morphological complexity.
Verbs are more complex than nouns in their lexical structures. For example, verbs
have a greater inflectional load than nouns, which has been given by the
morphological account as an explanation for the verb deficit observed in
agrammatic Broca’s aphasia.
Argument structure complexity refers to the number of arguments that a
lexical item can take. Nouns and verbs differ in the complexity of argument
structure. Verbs carry more complex argument structures than nouns because
verbs take arguments while nouns do not (they serve as arguments instead). This
distinction between nouns and verbs concerning their argument structure
complexity has been offered by the syntactic account as an explanation for the verb
deficit observed in agrammatic Broca’s aphasia.
The Competition Model considers the influence of specific morphological
structures of every individual language on the problem with complex structures, in
addition to the inflectional load offered by the morphological account. According
to the model, there are at least two ways to characterize word structure
3 The position that a lexical item appears in a sentence is another cue to tell grammatical function
of the lexical item. However, in the naming experiment, subjects are asked to produce names
without be given any positional cue. Thus, the positional cue is not a relevant grammatical factor
to the present discussion.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
52
complexity, depending on the morphological type of a language. In inflectional
languages, word structure complexity is manifested in the inflectional load of a
word (i.e., the number of inflectional markers appearing in a word). For example,
Italian verbs are structurally more complex than nouns because verbs carry a
greater inflectional load than nouns. In a non-inflectional language, structure
complexity cannot be characterized in terms of the inflectional load, rather it is best
characterized in terms of the number of morphemes that a word can bear. In the
case of Chinese, nouns and verbs do no significantly differ from each other
concerning the number of morphemes within a word. In general, the most
common word structure for verbs and nouns is the two-morphemes word
structure. However, both nouns and verbs may be as simple as single-morpheme
structures or as complex as three-morpheme structures. Thus, the word structure
complexity is not a useful feature to distinguish nouns from verbs. This leaves the
argument structure complexity as the only grammatical distinction between nouns
and verbs
When the syntactic account offered argument structure complexity as an
explanation for the verb deficit, it only considered argument structure of a lexical
verb. Additionally, argument structures of a sublexical verb observed in the
Chinese language may also be relevant to the verb deficit. This argument structure
of a sublexical verb is demonstrated very well by nominal VNN compounds, verbal
VN, and nominal VN. First, in these compounds, nouns are arguments of the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
53
verbs. For example, “zi” (meaning “character”) and “ji” (meaning “machine”)
serve as the object argument and the instrument argument of the verb “hit” in the
compound da-zi-ji (literally hit-character-machine, meaning “typewriter”).
Second, nominal VNN compound and the other two VN compounds show
different complexity of argument structure. The nominal VNN compound carries
two arguments, while the nominal VN compound and the verbal VN compound
take only one argument respectively. The syntactic explanation predicts the
difficulty of VNN compounds over VN compounds because the verb of VNN
takes more arguments than the verb of VN compounds. However, there will be no
differences between nominal VN and verbal VN because these two types of
compounds take the same number of argument.
2.7.3 Phonological Factors
The most common case of phonological effect on lexical accessibility is
observed in closed vs. open class words. The relative difficulty in accessing closed
class items, compared with open class items, is attributed to the lack of
phonological saliency in closed class items (Blumstein, Milberg, Dworetsky, 1984;
Goodglass, 1968; 1976; Kean, 1979, Milberg, Blumstein, & Dworetsky, 1988).
Phonological factors also are relevant to the noun-verb distinction. Studies on
phonological and form class relations showed that the syntactic classes of noun and
verb are distinguished in terms of the phonological classification of their vowels.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
54
Nouns are more likely to have back vowels than front vowels, while verbs are
more likely to have front vowels than back vowels (Sereno & Jongman, 1990).
Moreover, nouns and verbs differ in their stress patterns (Sereno & Jongman,
1995). These studies pointed out different distributions of phonological properties
in relation to the noun-verb contrast in English. They also open a possibility for
further investigation on Chinese that noun-verb distinctions may be predicted by a
phonological factor, such as the contrast of back vs. front vowels.
2.7.4 Frequency
Numerous studies have demonstrated this positive correlation between item
accessibility and item frequency (Bolota & Ferraro, 1993; Bresner & McCann,
1987; Gordon, 1983, Seidenberg, 1995; Monsell, 1991). Sereno & Jongman (in
press) reported a significant interaction between frequency effects and noun-verb
differences. Specifically, in a lexical decision word recognition task, response
latencies to uninflected nouns were significantly shorter than to uninflected verbs.
This result was attributed to a frequency difference between uninflected nouns and
uninflected verbs— the uninflected base form of nouns is more common than the
uninflected based form of verbs. Italian and Hungarian are a pair of languages in
which noun-verb differences are predicted by the frequency of morphological load.
In Italian, complex inflected verbs are more common than complex inflected nouns
(Bates, Friederici, & Wulfeck, 1987b); by contrast, in Hungarian, complex
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
55
inflected nouns are more common than complex inflected verbs (MacWhinney,
Pleh, Bates, 1985; MacWhinney, Osman-Sagi, & Slobin, 1991). In these cases,
frequency may function as a predictor of noun-verb differences.
Because Chinese is not an inflectional language, the inflectional load of
verbs and nouns cannot apply to Chinese verbs or nouns. However, among word-
class substructures that are assigned grammatical categories such as N, V, NN,
VN, W , NNN, VNN etc., NN structures function more frequently as nouns than
as verbs, whereas VN and W structures function as verbs more frequently than as
nouns (Huang, 1986; 1992; Huang & Hsieh, 1989). Thus, frequency may serve as
a clue to tell Chinese speakers that an NN lexical compound is likely to be a noun,
while W compounds or VN compounds are most likely verbs.
2.7.5 Uniqueness
This factor refers to the number of mapping relationships held between
forms and functions. It has a very close relation to the notion of frequency. Take
NN and VN word structures for a comparison. Both word types are high-
frequency word types. The NN word type associates primarily with the nominal
function; very few comparable grammatical functions are available to NN. Very
rarely NN can be a verb or an adjective as in “wu-se” (literally thing-color,
meaning “to hunt for” ) or mao-dun (literally spear-shield, meaning
“contradictory”). In contrast, the VN word type may function as a verb or a noun.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
56
In comparing NN with VN, the two word types differ in their uniqueness. The
mapping of NN to the nominal function is more consistent while the mapping of
VN to the nominal function and the mapping of VN to the verb function is equally
possible. Without considering the feature of “uniqueness,” NN and VN are not
different in their cue strength because both of them are high-frequency word types.
However, their cue strength becomes different if we consider the “uniqueness” of
their functions.
2.7.6 Cohort Structure
A cohort refers to the set that contains a lexical item and other items which
are lexically related to the lexical item. Two properties of the cohort are relevant
to the present discussion: (i) the size of the set and (ii) the level the cohort effect
involves (Bates & MacWhinney, 1989; Marslen-Wilson & Tyler, 1980).
The size of a cohort varies, depending on the nature of the cohort. For
example, closed class items belong to a small set because they are “closed”; the set
size of open class items, however, is an extremely large set because it constantly
takes in new members. Within the set of open class words, there are many subsets
in which lexical items are related in different ways. For example, lexical items
which share the initial phoneme may constitute a cohort (e.g. the list of words
which start with a specific consonant such as “s”).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
57
Based on the cohort structure discussed in the Competition Model,
consonant clusters, syllables, morphemes, words, or phrases are possible clues
which are used to form cohorts. These clues represent different levels at which
lexical items may associate or dissociate. When a subject attempts to retrieve a
lexical item, items within a cohort may come into competition with the target item.
The similar part will may cause the confusion of the subject to locate the target;
the dissimilar part will provide the subject with the feature that distinguishes the
target from the other items. Take the word “walk” as an example. Several words
are lexically related to “walk,” e.g. wall, walnut, walked, walking, etc. In a lexical
search, the subject finds the first three letters “wal” shared by these words, he/she
is still not able to decide the right word. The target word cannot be determined
until the distinctive part is found. Thus, the accessibility of a target lexical item
often depends on the number of lexical items that may compete with the target
lexical item, i.e., the set size of the target lexical item. Based on this idea, lexical
errors may tell us about (i) the number of neighbors that stay in the cohort of the
target (i.e., set size) and (ii) the level that the subject tends to associate with the
target.
The idea that lexical associations take place at different levels provides a
psycholinguistic basis for the proposal that a Chinese compound may be
decomposed into different levels of information. Like other languages, a Chinese
compound word may be associated at the single phoneme level, the consonant
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
58
cluster level, the syllable level, the morpheme level, the whole word level, or the
word structure level. The dissertation focuses on the morpheme level, the whole
word level, and the word structure level because these three levels are sensitive to
the assignment of grammatical category. Because these three levels carry
grammatical categories, noun-verb dissociations may appear at these three levels.
Before discussing noun-verb dissociations as the evidence for grammatical
category effects at the three levels, I will present several lines of linguistic evidence
for grammatical category effects at these three levels in chapter three.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
59
Chapter Three
Word Formation in Chinese:
Linguistic Evidence for the Level Analysis
Numerous studies of lexical impairments in Chinese aphasic patients have
reported a significant number of component errors (Chen, 1984; Chen, 1990a,
Packard, 1990) and a strong preference for high-frequency word structures (Chen,
Andersen, Kempler, & Bates, 1992). Some studies further demonstrated that
component errors may reflect noun-verb dissociations (Bates, Chen, Tzeng, Li, &
Opie, 1991; Bates, Chen, Li, Opie, & Tzeng,, 1993; Chen & Bates, in press).
Additionally, one study suggested that the preference for high-frequency word
structures is also relevant to noun-verb dissociations (Chen & Bates, in press).
These findings indicate that, in addition to a level for whole words, Chinese
compounds contain two intra-lexical levels; one for components and the other for
word structures. Like the whole word levels, these two intra-lexical levels are
sensitive to noun-verb dissociations.
Although neurolinguistic studies suggest the presence of two intra-lexical
levels within a compound, most Chinese linguists who are interested in formal
properties of compounds have never studied Chinese compounds in terms of an
intra-lexical approach. Instead, a lot of attention has been paid to the debate on
major defining features of compounds and the issue of the compound-phrase
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
60
boundary. Because Chinese compounds and phrases are similar concerning their
structures and vocabulary, these studies often fall into a dilemma that, no matter
what criteria one picks, there is no clear boundary between compounds and
phrases. As a result, there is a common view that the status o f compounds is a
very controversial issue in the field of linguistics (Chao, 1968; Li & Thompson,
1981; Tsao, 1979; Zhu, 1980, Ren, 1980).
In fact, the controversy of compound status is highly relevant to the
presence of intra-lexical levels. There are several lines of linguistic evidence
arguing that a Chinese compound can be decomposed into a level for components
and a level for word structures. The present chapter demonstrates several
linguistic arguments for the existence of two intra-lexical levels, based on
linguistic activities of normal native speakers. Arguments for the component level
are the semantic and grammatical independence of the component; arguments for
the word structure level are the effect of argument structure and the effect of
word-class substructure of a compound.
The decomposition of a compound leads to the analogy between
compounds and phrases, in the sense that both can be analyzed into vocabulary
(component morphemes for compounds and whole words for phrases) and
structures (word-class substructures for compounds and phrase structures for
phrases). This analogy helps us show that the problem of defining compounds is
due to a significant overlapping of vocabulary and structures between compounds
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
61
and phrases. More precisely, this overlapping of vocabulary and structure triggers
a rich bi-directional feeding between compounds and phrases, which is observed in
the re-analysis of compounds as phrases or the re-analysis of phrases as
compounds. Because compounds and phrases often involve a high degree of
interaction, a vague boundary between them is expected.
To give an idea of how the traditional approach fails to capture the true
nature of Chinese compounds, the following section first reviews the dilemma of
the attempt to separate compounds from phrases. The review further leads us to
an alternative view that a compound can be analyzed as a whole word chunk or
can be decomposed into a component level and a word structure level. This
decomposition process of compounds is supported by several lines of linguistic
evidence: (i) components carry their own grammatical categories and meanings,
(ii) components may be productively used to create lexically related compounds
(i.e., lexical productivity), (iii) word-class substructures carry their grammatical
categories, and (iv) word structures may also bear argument structures if there are
verbs in the structures. The last section discusses the interaction between
compounds and phrases emphasizing the bi-directional re-analysis process.
3.1 Chinese Compounds: The Dilemma
Chinese morphology is unique in its contrast between grammatical
morphology and open-class morphology. Its grammatical morphology is
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
62
extremely simple but its open-class morphology is extremely complex. The paucity
of grammatical morphology has been intensively discussed in the literature on
Chinese aphasia (see Bates et al., 1991, Chen, 1993; Chen & Bates, in press;
Packard, 1990, Tzeng, Chen, & Hung, 1991). The complexity of open class
morphology has received less attention in aphasia research. However, if we review
grammar books on Chinese morphology, we will find that open class morphology
is indeed very rich and complex. In this language, 80% of all words are
compounds, which are made up of two or more components. These components
are all single-syllable morphemes (Chao, 1968; Li & Thompson, 1981; Ren, 1980;
Zhu, 1980). Each component, like a word, carries its own meaning and
grammatical category. Compounds that are made up of components bear phrase
like structures.
In fact, Chinese compounds and phrases are similar in several ways. First,
neither compounds nor phrases are inflected. In English, inflectional markers such
as -ing are often presented in compounds with verbs such as man-eating (tiger).
Likewise, the inflectional marker -s is presented with the verb eat in the
corresponding phrase the tiger that eats the man. By contrast, in Chinese, man-
eating (tiger) is represented as chi-ren-hu (literally eat-mcm-tiger). There is no
grammatical marker at all in the Chinese compound. The subject argument (i.e.,
tiger), which stays outside of the compound in English, occurs instead inside of the
compound in Chinese. Likewise, the corresponding phrase chi ren de lao-hu
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
63
(literally eat man that preftx-tiger) contains no inflectional marker. The function
word -de is a relative clause maker, which is not an inflectional marker. The prefix
lao- functions as a derivational maker without significant semantic content (Chao,
1967). Second, compounds and phrases are alike in regard to their surface word
order. Take man-eating (tiger) as an example; in English, the NP phrase that
corresponds to man-eating (tiger) is “the tiger who eats the man,” which has an
SVO word order. This order differs from the OV(S) order of the compound
“man-eating’ (tiger). This surface word order contrast disappears in Chinese.
Both Chinese phrases and compounds share a VOS word order. The phrase is
“chi ren de lao-hu" (literally eat man who prefix-tiger, meaning “the tiger who
eats man”). Similarly, the compound is “chi-ren-hu” (literally eat-man-tiger,
meaning man-eating tiger) 4 These two aspects of similarity make it difficult to
distinguish compounds from phrases.
Because compounds and phrases are alike, some compounds may act as
phrases in one case and act as compounds in other case. For example, yu-mo
(humor) is a true word borrowing from the foreign English word “humor” (Chao,
1968; Li & Thompson, 1981). Both yu (as a verb) and mo (as a noun) are bound
4 There are two differences between this compound-phrase pair. One difference is the use of the
verb. The verb of the phrase (i.e., chi, meaning “/o eaf) is a verb used in Modem Chinese; the
verb of the compound (i.e., shi, meaning “to e a f ) is a verb used in Classical Chinese. Very
often, a classical verb is not productively used in Modem Chinese. It only appears as a part of a
compound in Modem Chinese. Another difference is the presence of a prefix in the lexical item
of phrases (e.g. lao-hu, literally preftx-tiger, meaning “tiger”) but the absence of the prefix in the
lexical item for compounds (e.g. hu, meaning “tig e f). There is no meaning change in lexical
items with/without prefixes (Li & Thompson, 1981).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
64
morphemes that never appear alone in sentences as free words. These properties
mak&yu-mo a highly integrated word.
Although the compound yu-mo is highly integrated, its constituents are
separable in a limited fashion. For example, it is acceptable to insert pronouns and
numerals between the two components to yield the phrase yu ni yi mo (literally
“humor you one humor," meaning “ humor you once”). This separability of
components leads Chinese linguists to adopt a less stringent definition such as
allowing a limited separability as one of defining features of compounds.
A less stringent definition for compounds raises a great controversy on the
compound-phrase boundary. O f all the compound types, the verbal VN compound
(see 2a-c) is the most controversial one because many verbal VN compounds are
no different than their phrase counterparts in regard to their vocabulary, their
surface word order, and their morphological complexity. The very same VN verbs
in 2a-c may also act like phrases in 2d-f. The distinction between VN compounds
and VN phrases lies in their semantic-specificity. Nouns of VN compounds are
generic, while nouns of phrases carry specific meanings.
Using the meaning preservation as a distinctive feature, we may show that
compounds and phrases differ in their syntactic freedom. More precisely,
components of VN phrases (see 2d-f), like free words, may take syntactic
transformations such as the topicalization to move nouns to the topic position such
as 3a-c.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
65
2. Verbal VN Compounds vs. Phrases
Compounds Phrases
a. tiao-wu
jump-dance
d. tiao wu
jump-dance
“to dance a dance’ “to dance”
b. kan-shu
look-book
“to read”
e. kan shu
look book
“to read a book”
c. chi-fan
eat-rice
“to have a meal”
f. chi fan
eat rice
“to eat rice’
After applying the topicalization transformation, meanings of nouns and verbs
remain unchanged. These nouns refer to specific objects/concepts. By contrast,
moving components of compounds results in the change of meanings. Once nouns
are moved, their generic interpretations are no longer preserved (see 3d-f)
Rather, these nouns have to refer to specific objects/concepts.
3. Verbal VN as Phrases (Huang, 1983)
a. Nei zhi wu, ta tiao le.
that classifier dance, he dance aspect marker
“He danced the dance.”
b. Nei ben shu, ta kan le.
that classifier book, he look aspect marker
“He read the book.”
c. Nei-wow fan Ta chi le.
tha.t-classifier rice, he eat aspect maker
“He ate the bowel of rice.”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
66
3. Verbal VN as Compounds (Continued)
d.*Nei zhi wu, ta tiao le.
that classifier dance, he dance aspect marker
“He has finished dancing.”
e.*Nei ben shu, ta kan le.
that classifier book, he look aspect marker
“He has finished reading.”
f.*Nei wan fan, ta chi le.
That -classifier rice, he eat aspec/ maker
“He had a meal.”
There are different views as to the “compoundhood” of these verbal VNs.
Some linguists who focus on the presence of bound morphemes in these verbal
VNs interpret examples 2a-c as verb phrases because none of the components are
bound morphemes. Because these components of verbal VNs are separable, some
linguists even suggest to exclude examples 2a-c from the list of compounds
(Huang, 1983).
However, there are also linguists who observe semantic ambiguity in verbal
VN compounds. They show that verbal VNs carry both compound readings and
phrase readings (Bates et al., 1993). Examples 2a-c demonstrate compound
readings, while examples 2d-f carry phrase readings. Their semantic contrast can
be shown in the topicalized sentences. More precisely, the semantically specific
nouns can be topicalized (see examples 3a-c), whereas the generic nouns cannot
(see example 3d-f).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
67
The claim that the noun “fan” of the compound “chi-fan” is generic is
observed in the following examples, in which the presence or the absence of the
noun “fan” does not affect the meaning of the VN “chi-fan.” When greeting,
Chinese people often express their concern by saying “did you eat yet?” For this
purpose, they may use either sentence 4a with the noun “rice” or sentence 4b
without the noun “rice,” both of which mean “did you eat yet?” These examples
show that the meaning of a generic noun is not significant.
4. Sentences
a. Ni chi fan le mei-you.
you eat rice aspect-maker not-have
“Did you eat yet?”
b. Ni chi le mei-you
You eat aspect-marker have-not
“Did you eat yet?”
If a generic noun is not semantically significant, for what purpose the noun
is included in a compound. Bate et al. (1993) argued that the generic object is
used for a structural requirement of a VN form for action names. For instance, the
generic noun “fan” (rice) has to appear after the verb “chi” (to eat) when naming
the action “to eat.” Likewise, the example 5 given by Bates et al. (1993) showed
that, in order to name the action “to sew,” the object noun “clothes” must be
present with the verb “to sew.”5 The single-morpheme word is not a preferred
s“Yi-fu” is a generic term of “clothes” in Chinese. This generic form was used by the subjects
to make a VN compound for the action name “to sew.”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
68
word structure in Mandarin Chinese. Even in example 3 b, where the generic noun
is dropped, the aspect marker -le must be attached to the verb to maintain a two-
morpheme word structure.
5. Feng-yi-fii
sew-clothes-clothes
“to sew” (from Bates et al., 1993)
The semantic ambiguity also holds in nominal compounds. For instance,
some nominal compounds, which are claimed to be true compounds, may carry
two readings. Examples 6a-c and examples 6d-f demonstrate the ambiguity.
6. Nominal Compounds 6. Nominal Phrases
a. hong-yu d. hong-yu
red-fish red-fish
“red snapper” “red fish”
b. jin-yu e. jin-yu
gold-fish gold-fish
“goldfish” “golden fish”
c. hong-ren f. hong-ren
red-man red-man
“a pearl in one’s eyes; Indian” “red man”
Examples 6a-c are uncontroversial lexical compounds because they refer to
objects. However, these compounds may be interpreted as phrases 6d-f. Lexical
compounds and phrases differ in their syntactic freedom. Lexical compounds
cannot take any syntactic operation, while phrases may take various syntactic
operations such as the insertion of the adjective marker -de (see 7a-c). If lexical
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
69
compounds attach the adjective marker -de, their meanings no longer refer to the
same objects (as referred by lexical compounds). The “red snapper” in 6a becomes
an ordinary red fish as in 7a. The goldfish in 6b becomes “a golden fish” in 7b.
Likewise, “red-man” in 7c is not an Indian but a man who is red (in skin or
clothes). Unlike compounds which do not allow any syntactic transformation,
phrases may attach the adjective marker -de while preserving the meanings of
phrases. Examples 7a-c and examples 6d-f are identical in their meanings
respectively.
7. Nominal Phrases
a. hong de yu
red adjective marker fish
“red fish”
b. jin de yu
gold adjective marker fish
“golden fish”
c. hong de ren
red adjective marker man
“red man”
These examples show that compounds may be interpreted as idiomatic
compounds and/or un-idiomatic phrases. The idiomatic and un-idiomatic readings
reflect a whole word reading and a literal reading in the interpretation of verbal
VN compounds and NN compounds. The literal reading may function as a phrase
reading if the component morphemes are also free words. Otherwise, the literal
reading remains as a “literal” translation if component morphemes never act as free
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
70
words. Because these two readings share the same form, no surface structural
difference is available to distinguish one from the other.
Therefore, because both readings are valid, it is a very likely that native
speakers may alternate between the two readings when they are asked to process a
compound. A subject could choose a whole word reading when the context
requires a compound reading. However, if the context asks for subjects to read
individual components, the subject might choose either a phrasal reading or a
literal reading. Because compounds and phrases can elicit two different readings
of the same structure, it is very difficult to distinguish one reading from the other
reading. However, it is possible to attribute the co-existence of two readings into
reading processes occurring at two different levels. In real-time lexical processing,
the whole word reading and the literal reading may be activated simultaneously.
This simultaneous activation of two readings is the major source for the
controversy on the boundary between compounds and phrases.
The traditional approach attempts to put the two readings of the same form
into either the category of phrase or the category of compound. This approach
fails because it does not distinguish compound readings from phrase readings.
However, the two readings of the same form may be captured by putting them into
different levels of representation. In fact, several linguistic phenomena
demonstrate that in addition to the whole word level, there are two additional
levels: a level for component morpheme (i.e., the sublexical level) and a level for
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
71
word structure. The compound reading takes place at the whole word level,
whereas, the literal reading (or the phrase reading in the case that component
morphemes act as free words) takes place at the sublexical level. The word
structure level is necessary in the case such as VN structural requirement for action
names. The following section demonstrates how these linguistic phenomena
provide evidence for the sublexical level and the word structure level.
3.2 Five Linguistic Properties of Chinese Compounds and the Interaction of
Compounds with Phrases
This section presents linguistic evidence for the sublexical level and the
word structure (or word class substructure) level as theoretical support for these
two levels which have been observed in the empirical study of noun-verb
dissociations. It will examine the semantic independence of component
morphemes, the grammatical category assignment of component morphemes,
lexical productivity, word-class substructure, and argument structure. Semantic
independence, grammatical category assignment, and lexical productivity will be
used to argue for a sublexical level. Word-class substructure and argument
structure will be used to argue for a word structure level. The fact that
compounds and phrases are alike concerning their structures, and that they share
many of the same vocabulary items, suggests that they may interact with each
other in some contexts. In this regard, I will present data to demonstrate a bi-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
72
directional feeding relation between compounds and phrases, including the
derivation of compounds from phrases and vice versa, and the re-analysis of
compounds as phrases and vice versa. The bi-directional feeding relation provides
another line of evidence for the sublexical level and the word structure level.
3.2.1 Linguistic Evidence for the Sublexical Level
Lexical items are represented at an independent level because they carry
their own meanings and grammatical categories. Moreover, these lexical items can
be productively used to create phrases or sentences. Following this idea,
sublexical elements require a separate level if they show the following properties:
(i) they carry their own meanings, (ii) they carry their own grammatical categories,
and (iii) they show productivity in compound formation.
3.2.l.a Components Are Semantically Independent
Semantic independence refers to the cases where native speakers are able
to identify the meaning of each individual part within a lexical compound. In
English, it is quite clear that compound parts are semantically independent. This is
because compound parts are all free words which also participate in the formation
of sentences.6 Take the compound “bath towel” as an example; “bath” and
“towel” are components of the compound “bath towel.” They are also free words
6 Components of some quasi-compounds are not necessarily free words, e.g. “cranberry,” where
the component “cran” is not a free word.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
73
in the sentence “I used the towel after taking a bath ” The Chinese word which
stands for “bath towel” is the compound “yu-jin” (literally bath-towel). Unlike the
English compound “bath towel,” both “ yu” and “ Jin" are bound morphemes which
carry their own meanings. In English, it is not necessary to create a sublexical
level for “bath” and “towel” because both of them are also free words. However,
in Chinese, bound morphemes such as “ yu” and “ jin” are never used as free words.
It is therefore necessary to create a separate level for these bound morphemes. If
there is no new level created for these components, they will be forced to stay at
the syllable level~a suggestion given by Li & Thompson (1981). However,
treating bound morphemes as syllables does not appropriately capture the nature of
these morphemes, because unlike some syllables, these bound morphemes carry
their own meanings. The interpretability of these bound forms suggests a special
mechanism or level responsible for interpretations of these morphemes.
The first evidence for semantic independence of word parts comes from the
presence of literal translation, which is often found in the study of Chinese
compounds. The following examples demonstrate that literal meanings and
compound meanings are constantly offered when Chinese linguists present
compounds (Chao, 1968; Li & Thompson, 1981).
8. Two Readings of Compounds
Literal Reading Compound Reading
a. zhang-fang account-room cashier's office (Chao, 1968, p. 382)
b. kai-dao open-knife operate on (Li & Thompson, 1981, p.77)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
74
In 8a, the compound “zhang-fang” has two levels of meaning which are not
identical, the literal meaning and the compound meaning. The presence of two
meanings is also true in 8b, in which the compound “kai-dao” is literally translated
as “open-knife” but its compound meaning is “to operate on.” The presence of
literal meaning suggests that the components of a compound are semantically
independent.
Components of compounds always carry their own meanings regardless of
the semantic relation between components and their whole compound. This
constant semantic independence is observed in compounds that bear a different
relation to their components. According to Li & Thompson (1981), there are
three forms of relations between compounds and their components: (i) the lexical
meaning is simply a composition of the component meanings; (ii) the lexical
meaning is a derivation from the meanings of components; (iii) no apparent
relation holds between a compound and its parts. The following three examples
represent each type of relation respectively.
9. a. cha-hu tea-pot “teapot” (Chen & Bates, in press)
b. qi-er stand-goose “penguin” (Lu, 1984)
c. hua-sheng flower-born “peanut” (Li & Thompson, 1981)
The contrast between the compound “cha-hu” (literally tea-pot, meaning “teapot”)
and the compound “qi-er” (literally standing-goose, meaning “penguin”) is a direct
mapping vs. derivational mapping. Let us first consider the example “penguin.”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
75
The literal meaning “standing-goose” depicts two major features of “penguin”;
penguins are usually standing and they own some properties of a goose. Thus,
meanings of components make up the meaning of the whole word by providing the
key features of the referent. The example “cha-hu” (literally tea-pot, meaning
“teapot”), however, involves a one-to-one mapping. Its meaning comes from
“tea” and “pot” directly. In cases like this, meanings of components are
completely transparent to the whole word. The last type (9c) exhibits the highest
idiomaticity, where two levels of meaning are provided but no apparent relation
exists between the two lexical levels. Although there are different mappings
between compounds and their components, these differences do not prevent the
components from receiving their own meanings. Components of compounds
which have the highest idiomaticity still carry literal meanings. This suggests that
meanings of compounds are not necessarily derived from their components and
vice versa.
The use of literal meanings is not only observed in written literature. It
appears in speaking too. A speech given by a senator from Taiwan provides a
clear example of the presence of literal meanings for components of the
compound.
10. Wo-men do shi luo-hua-sheng, tu-do la, yao luo-di-sheng-gen
I -plural all be fall-flower-grow, ground-bean particle will fall-ground-grow-
root
“we all are peanuts, (like) peanut, (we) were bom here and grew up here.”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
76
The speaker gave the first word “luo-hua-sheng” (peanut), and then provided the
whole word interpretation and the component interpretation immediately after.
“To-do” (peanut), which does not share any part with the word “luo-hua-sheng,”
is the whole word interpretation of “luo-hua-sheng” (i.e., peanut). In this case, no
component interpretation is involved. In contrast, the idiom “luo di sheng gen,”
which shares the two morphemes “luo” and “sheng” with the compound word
“luo-hua-sheng,” carries only component interpretations: “luo” (to fall) and
“sheng” (to bear). The interpretation for the whole idiom does not equal the
interpretation for the compound word “luo-hua-sheng” (peanut). The speech
activity tells us that the speaker may simultaneously activate both lexical and
sublexical levels when he speaks the word “luo-hua-sheng” (peanut).
3.2.l.b Evidence for the Sublexical Level: Grammatical Categories of
Components
This section demonstrates that components of a compound carry their own
grammatical category. This serves as another line of evidence for the presence of a
sublexical level. In turn, it supports the idea that compounds are generated by
word-class substructures, which are depicted by grammatical categories of
components. There are two ways to show that components carry their
grammatical category. First, the assignment of a grammatical category to a
component is observed in the grammatical analysis of compounds in Chinese
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
77
morphology. In studying the issue concerning the head of a compound, Huang
(1986) examined different compound types within their grammatical categories by
comparing the whole compound and its components. In his analysis, grammatical
categories were assigned at two levels to yield nominal NN, nominal VN, verbal
VN, nominal AN, etc. The first term “nominal” is a grammatical category for the
whole word, whereas, the rest of the terms, such as NN, VN, and AN depict
grammatical categories of components. The assignment of grammatical categories
to compounds and their components was also used in the study of lexical
acquisition in Chinese (Huang & Hsieh, 1989). In this study, a sensitivity to
grammatical categories inside of a compound is observed in both lexical
production and comprehension. Specifically, children showed preference for
nominal components for nouns (NN for nouns) and verbal components for verbs
( W for verbs).
Second, components of compounds may be assigned thematic roles (Chao,
1968; Li & Thompson, 1981). Based on the thematic role of the component, a
verbal VN compound may be structured as a Verb-Object compound. An NV
compound may be depicted as a Subject-Verb compound or an Object-Verb
compound depending on the thematic role of the nominal component. Because
components which carry thematic roles are nominal, and components which
provide predication relations between thematic roles are verbal, components of
compounds should have their own grammatical categories.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
78
Although Chinese linguists assume that each component of a compound
carries its own grammatical category, they do not actually spell out the clues that
are used to determine the grammatical category of a component. Because there is
no overt grammatical marker in Chinese to indicate grammatical category and no
syntactic clue is available inside the compound, the assignment of grammatical
category to a component is perplexing. Thus far, there are several studies
investigating how native speakers of Chinese determine grammatical roles of
arguments within a sentence (Chen, Tzeng, & Bates, 1990; Li, Bates & Liu, 1992;
Li, Bates, & MacWhinney, 1993). There are also studies investigating the
acquisition of grammatical categories such as noun and verbs (Tardif, 1996; Tardif,
Shatz, & Naigles, 1996). However, these studies focus on grammatical roles of
whole words. Very little attention has been paid to the assignment of grammatical
roles to word components.
A study of lexical development provides evidence for the assignment of
grammatical categories to components of a compound (Huang & Hsieh, 1989).
When comparing children with adults regarding their production of compound
types, Huang & Hsieh (1989) demonstrated that both children and adults showed a
strong tendency for producing NN compounds for nouns, VN and W for verbs,
and AA for adjectives. These four word types are the major word types for nouns,
verbs, and adjectives respectively. They also reported minor word types such as
NV, NA, VN, W , VA, AN, AV, and AA for nominal compounds; NN, NV, NA,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
79
VA, AN, AV, and AA for verbal compounds, and NN, NV, NA, VN, W , VA,
AN, and AV for adjectival compounds. In comparing the major word type with
minor word types, the major word type often consists of components whose
grammatical categories are identical with the target compound, whereas, the minor
word types consist of components whose grammatical categories are not identical
with their target compound. The nominal compound consists of all nominal
components (NN), as do the adjective compound and the verbal compound (for
the W case). The verbal VN contains a nominal component which is not identical
with the grammatical category of the compound. However, the head component
of the verbal VN is the verb, which is identical with the compound regarding the
grammatical category (Chao, 1968; Li & Thompson, 1981; Huang, 1986). The
consistency of grammatical category between a compound and its components
suggests that, like the compound, the components carry their grammatical
category. The grammatical category of components reflects the knowledge of
grammatical assignment at the component level. This also suggests that both
children and adults are sensitive to grammatical categories within a compound.
Additionally, in the pilot work for the present study, I made several
observations that may help us understand how a native speaker chooses a
grammatical category for a component. First, thirty Chinese-speaking college
students were asked to select a grammatical category for a single-morpheme word
out of six choices (noun, verb, adjective, adverb, preposition, auxiliary). In the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
8 0
test, they were also asked how they made a choice. They replied that they would
look for the meaning of the word first. They further explained that if the word was
an object or an action, it should be a noun or a verb. If the word did not fall into
the two major categories, they would seek another strategy, such as the relation of
the word to its adjacent word. For example, if the word is adjacent to an auxiliary
such as “hui” (meaning “be able to”), it is very likely that the word is a verb.
Moreover, the lexical items following classifiers are nouns. Subjects’ responses to
the selection of grammatical category suggest that semantic cues play a primary
role for a Chinese adult to determine the grammatical category of a word. The
positional cue comes after the semantic cue.
In the assignment of grammatical categories to components, syntactic clues
are not available because these clues cannot operate inside a compound. The only
clue left for determining grammatical categories of components is the semantic
one. The two kinds of semantic clues available within a compound are: the
meaning of the target component and the meaning of the adjacent component.
When the target component is not ambiguous, its meaning directly determines the
grammatical category of component. However, when the target component is
ambiguous, it is necessary to look for its adjacent component. The following
examples demonstrate how adjacent components help speakers to resolve the
ambiguity of the target component.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
81
The second component “guan” of the compound “guan-xin” is ambiguous.
One possible interpretation of “guan” is “to close” as in the phrase “guan men”
(meaning “to close door”); the other possible interpretation is “to involve” as in the
compound “guan-xin” (literally involve-heart, meaning “to concern”). When
speakers are asked to interpret the component “guan,” they often look for its
adjacent component. If the adjacent component of “guan” is “xin” (heart), the
component “guan” is interpreted as “to involve.” However, if the adjacent
component of “guan” is “men” (door), the component “guan” is interpreted as “to
close.”
Although “to close” and “to care” are different in their meanings, they do
not differ in their grammatical categories. Both “to close” and “to care” are
verbal. However, there are cases in which the adjacent morpheme may affect the
choice of grammatical category for the target morpheme. For instance, the
morpheme “hua” carries at least three interpretations: flower, to spend, and floppy.
Of these three meanings, the nominal interpretation “flower” is the most common
one. When there is no adjacent element available, the most likely interpretation for
“hua” is the nominal interpretation “flower.” However, if there is an adjacent
morpheme, such as “fei” (expense), the verbal interpretation (i.e., to spend) wins
over the nominal interpretation (i.e., flower). Moreover, if the morpheme “hua” is
adjacent to the morpheme “xin” (heart) to yield the compound “hua-xin” (literally
floppy-heart, meaning “floppy”), the stem “hua” is interpreted as “floppy,” an
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
82
adjective morpheme. This example of “hua” demonstrates that the grammatical
category of a component is strongly influenced by its adjacent morpheme.
Second, I observed an awareness of grammatical categories inside a
compound in the lexical development of my own child Noel. In the transition from
the one-word stage to the two-word stage, Noel started to distinguish NN from
W /V N concerning grammatical functions of these patterns. She used NN
patterns as nouns and W /V N patterns as verbs. Before the age of ten months,
she produced only NN for objects such as “mama” (meaning mother), qiu-qiu
(literally ball-ball, meaning “ball”), dian-deng (literally electricity-lamp, meaning
“lamp”), etc. At seventeen months, she started to produce VO or W (V ) for
actions, for instance, chi-fan (literally eat-rice, meaning “to eat”) or die-dao
(literally fall-fall, meaning “fall down”). The pattern for object nouns remains NN.
What Noel produces tells us that she has developed lexical compounds for
objects and actions. These lexical compounds can be called noun (for objects) or
verb (for actions). These two grammatical categories are distinguished in a
semantic-conceptual way (an object vs. action distinction), and in a grammatical
way (NN vs. V N /W distinction where NN is for noun and VN or VV is for verb).
It is interesting to note that there was not even a single case of nominal VN when
Noel started to use VN verbs. Thus, Noel is no different than children investigated
in the previous research, who prefer consistent grammatical category between a
compound and its components (Huang & Hsieh, 1989). This consistency may be
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
83
interpreted as a sensitivity to grammatical categories inside of the compounds and
a sensitivity to the relation of internal grammatical categories with the external
grammatical category.
3.2.1.C Evidence for the Sublexical Level: Productivity of Compounds
Components of compounds that can be freely combined to create a new
compound indicate a sublexical activity in the compound formation. The
combination of components usually follows constraints which may be structural or
semantic, depending on the morphological types of languages. Structural
constraints refer to word patterns that are used to construct compounds. For
example, in English, the NV compound has to be accompanied by an -ing (or an -
ed) to yield an NV-ing (or NV-ed) compound such as “man-eating,” “man-
drinking,” “man-taking,” “man-loving,” etc. Compounds can be productively
created by inserting words into the NV-ing word pattern. However, due to the
lack of inflection, word patterns with inflections cannot apply to the compound
formation in Chinese. Word patterns such as VN may be used to generate a verbal
compound (chi-fan, literally eat-rice, meaning “to eat”) as well as a nominal
compound (qi-er literally stand-goose, meaning “penguin”). Thus, in Chinese, a
total reliance on a structure (e.g. VN) may generate both verbal compounds (e.g.
fry-rice for the verb “to fry rice”) and nominal compounds (fry-rice for the noun
“fried rice”). This structural ambiguity does not cause a serious problem to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
84
English speakers because an un-inflected VN form often functions as a verbal VN
phrase, e.g. to fry rice. When a VN form functions as a nominal VN compound, it
is usually inflected such as fried rice1
The lack of an overt grammatical marker to distinguish one compound type
from the other compound type motivates us to look for an alternative analysis for
describing the productivity of Chinese compounds. Studies on language
processing in Chinese, including word/sentence production and word/sentence
comprehension, pointed out that Chinese subjects tended to rely on semantic cues
such as animacy over grammatical cues such as word order (Bates, Chen, Tzeng,
Li & Opie, 1991; Bates, Chen, Li, Opie, & Tzeng, 1993; Chen, Tzeng, Bates,
1990; Li, Bates, & MacWhinney, 1993; Li, Bates, & Liu, 1992; Li & Thompson,
1980). This semantic strategy is observed in compound formation too. The
semantic strategy means that the selection of a component morpheme is primarily
determined by its semantic relation to the other component morphemes within a
compound. In this sense, productivity is determined by the range of the variants
that a component morpheme can semantically relate to. In an actual lexical
7 Even if the VN is not inflected, its orthographic or phonological representations can tell the
lexical nature. For example, although “scapegoat” and “whiplash” appear without any
grammatical marker, these compounds are written as single words and carry a primary stress.
This advantage of English morphology is not available to Chinese speakers because Chinese is an
isolating language. Each component is written as an individual character. Thus, phrases and
compounds do not differ regarding their written forms. Moreover, in English, stressed syllables
tend to recur at regular intervals, while unstressed syllables do not (i.e., a stress-timed language,
see Ladefoged, 1975, p. 222). Unlike English, most Chinese morphemes (corresponding to
syllables of English) recur at regular intervals (i.e., a syllable-timed language by Ladefoged,
1975, p. 222). Because most morphemes like words are pronounced at regular intervals, the
speech interval is not a reliable feature to distinguish compounds from phrases.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
85
process, the productivity is shown in the search for plausible component
morphemes which can appear before/after the primary component morpheme. To
show how this idea works in compound formation, let us examine some lexically
related compounds in the following.
11. X-ji (nominal interpretation only)
a. huo-ji fire-chicken “turkey”
b. tu-ji soil-chicken “stew chicken”
c. gong-ji male-chicken “rooster”
d. mu-ji female-chicken “hen”
f. zha-ji fry-chicken “fried chicken/to fry chicken’
g. duen-ji stew-chicken “stewed chicken”
Examples 1 la-g share the second morpheme component “ ji” but differ in the first
morpheme component. Assuming that speakers have the “ji” (chicken) in mind,
they are about to choosing another morpheme component. Although there is great
freedom of choice, there is a semantic/pragmatic constraint when choosing an
appropriate combination of morpheme components. Examples lla-g demonstrate
a list of features that chickens may have. The component morphemes that modify
component “ ji” (meaning “chicken”) may belong to different grammatical
categories. For example, modifier components of “ji” may be nouns (i.e., initial
components of examples lla-b), adjectives (i.e. initial components of examples
1 lc-d), or verbs (i.e., initial components of examples 1 lf-g).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
86
The core component morpheme, which is surrounded by a set of variants,
may serve as a variant of another component morpheme. In examples 1 la-f, the
stem “ji” is the core. However, “ji” becomes one of the variants of the core “zha”
(to fry) as in examples I2a-d. This demonstrates that, in relating “ji” to the verb
“fry,” another group of lexically-related compounds is derived. In this case, the
verb (to fry) functions as the constant while the nouns serve as variants.
12. Zha-X (verbal and nominal interpretations)8
a. zha-ji fry-chicken fried chicken/to fry chicken
b. zha-yu fry-fish fried fish/to fry fish
c. zha-dan fry-bomb bomb
d. zha-you fry-oil fried oil
The formation of compounds is characterized as a productive process by
associating variant morphemes with the constant morpheme. The same association
holds in verbal compound formations. As shown in examples I3a-h and 14a-d, the
verb “to drink” can function either as a constant (13a-h) or as one of the variants
(14a-d).
8 For “zha-ji” and “zha-yu,” both verbal interpretations (to fry chicken and to fry fish) and
nominal interpretations (fried chicken and fried fish) are available to native speakers. The choice
of interpretation rests entirely upon the plausibility. For example, it is very likely to have both
“fried chicken” and “to fiy chicken” in case of “zha-ji.” However, it is not plausible for someone
to fry the oil, because the oil cannot function as an object. “You” of “zha-you” (fiy oil) more
likely functions as an instrument (i.e., the oil is used as a tool to fiy something).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
87
13. VX (verbal interpretation only)
a. he-shui drink-water “to drink”
b. he kafei drink-coffee “to drink coffee”
c. he-cha drink-tea “to drink tea”
d. he-guo-zhi drink-fruit-juice “to drink juice”
e. he-nui-nai drink-cow-milk “to drink milk”
f. he-qi-shui drink-gas-water “to drink soda”
g. he-jiu drink-wine “to drink wine”
h. he-dou-nai drink-bean milk “to drink soybean milk’
14. XN (verbal interpretations only)
a. he-shui drink-water “to drink”
b. sa-shui scatter-water “to sprinkle”
c. fang-shui release-water “to release water”
d. pen-shui spay-water “to spray”
Moreover, component morphemes can be joined with other elements; the
stem “pen” (to spray) joins with the stem “shui” (water). Yet, they can be
substituted. For example, the stem “pen” (to spray) is replaced with another stem
“fang” (to release) to yield another compound “fang-shui” (literally release-water,
meaning “to release water”). The fact that component morphemes can be freely
substituted by another component morpheme resembles the sentence formation
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
88
where lexical items can be freely substituted to yield alternative sentences. This
resemblance suggests that within the domain of compounds, component
morphemes are like lexical items because they are free to participate in word
formation.
Thus far, the section presents evidence from semantic independence, the
assignment of grammatical category, and compound productivity to argue for a
sublexical level for components of a compound. The following section discusses
evidence for the presence of word structure level in the access of the
representation of a compound.
3.2.2 Word Structures and Argument Structures
Since components have their own meanings and they carry their own
grammatical categories, it is possible to characterize the relation between these
components in terms of their meanings and their grammatical categories. When a
compound is characterized in terms of grammatical categories of their
components, it is plausible to construct a word structure for the compound in
terms of grammatical categories of its components. This structure is analogous to
the structure of a phrase. In detail, like the structure of a phrase, a word structure
is analyzed in terms of grammatical categories such as, N (noun), V (verb), A
(adjective), etc. NN, VN, W , etc. are examples of word structures.
Components, serving as content elements, are inserted into the word structure to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
89
derive a compound. This derivational process is similar to the process where
words are inserted into the phrase structure to yield a phrase.
Alternatively, a compound may be structured in terms of meanings of
components. This approach focuses on the grammatical relation between
components. The structure, which conveys the grammatical relation between
components, is called argument structure. Word structures at this level are similar
to deep structures of phrases. The following section elaborates some properties of
word structure and argument structure which are relevant to the issue of noun
verb dissociations.
3.2.2.a Word Structures
Because components carry their own grammatical categories, it is very
common for Chinese linguists to depict compounds in terms of grammatical
categories of components. Examples in Table 1 demonstrate possible nominal
word types and examples in Table 2 show various verbal word types. Notice that
these word types contain both the grammatical category of the whole word and
grammatical categories of its parts.
In comparing examples in Table 1 with examples in Table 2, there are a few
points which are relevant to noun-verb dissociations. First, there are more nominal
word types than verbal word types. Second, nominal word types and verbal word
types are not exclusive. For example, VN may appear as a verbal compound or a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
90
Table 1
Word Structures: Nominal Compounds
NN
(1) a. lan-hua
b. mei-hua
AN
(2) a. xian-hua
b. gan-zao-hua
VN
(3) a. xi-guan
b. wan-ju
VNN
(4) a. chi-ren-yu
b. xi-yi-fen
NNN
(5) a. yu-mao-qiu
b. huo-cai-he
W N
(6) a. duo-bi-qiu
b. tuo-la-ji
NVN
(7) a. ren-zao-hua
b. shou-la-mian
orchid-flower
plum-flower
fresh-flower
dry-dry-flower
suck-hole
play-instrument
eat-man-fish
wash-clothes-powder
feather-fiir-ball
fire-wood-box
hide-escape-ball
pull-pull-machine
man-make-flower
hand-pull-noodle
“orchid”
“plum flower”
“fresh flower”
“dry flower”
“straw”
“toy”
“piranha”
“detergent”
“badminton”
“match box”
“hide-escape-ball”
“towing truck”
“artificial flower”
“hand-made noodle”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
91
nominal one. Moreover, nominal compounds may contain a verb piece and the
verbal compound may contain a noun piece. Another feature of word type
variation, which is not shown in the list but relevant to the discussion, is that each
word type occurs at a different frequency in the language. Among possible word
types, NN is the major word type for compound nouns; VN and W are two major
word types for compound verbs (Huang, 1986; 1989). Frequency sensitivity found
in word types suggests that word structures, like lexical items, should be treated as
independent processing units.
Table 2
Word Structures: Verbal Compounds
Verbal VN
(1) a. tiao-wu
b. tiao-sheng
c. tiao-shui
d. tiao-fang-zi
Verbal W
(2) a. chi-wan
b. die-dao
Verbal NV
(3) a. tian-liang
b. tou-teng
Verbal V W
(4) a. zhan-qi-lai
b. xie-qing-cu
jump-dance
jump-rope
jump-water
jump-house-suffix
eat-finish
fall-fall
sky-bright
head-ache
stand-up-come
write-clean-bright
“to dance”
“to jump rope”
“to dive”
“to hopscotch”
“to have eaten”
“fall down”
“to dawn”
“headache”
“stand up”
“write clearly”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
92
The analogy of word structures to lexical items motivates the proposal
that word structures function like templates, which can be activated/inhibited
alone. When a compound is activated, the whole word, components, and word-
structure templates are activated. Components are inserted into a template to yield
a right word. Thus, lexical compounds, components, and word-structure
templates may be accessed independently while still interacting very closely.
3.2.2.b Argument Structures
Several linguistic analyses on compounds demonstrate not only that
sentences bear argument structures, but also that compounds may bear argument
structures (Fab, 1984; Di Sciullo & Williams, 1987). When there is a verbal part
and a nominal part within a compound, there is an argument structure. Nominal
components are arguments of the verbal element. The relation between arguments
constitutes the argument structure of the compound. The following examples
demonstrate possible argument structures within a Chinese word.
Nominal Compound
15. Verb-Agent
a. qi-er
stand-goose
“penguin”
b. yi-shi
cure-master
“doctor”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
93
16. Agent-Verb
a. tian-shi
heaven-send
“angel”
b. gou-shi
dog-eat
“dog food”
17. Instrument-Verb
a. suan-pan
count-dish
“abacus”
b. wan-ju
play-instrument
“toy”
18. Verb-Object
a. shao-bing
bum-cookie
“pan cake”
b. zha-ji
fry-chicken
“fried chicken”
19. Verb-Object-Agent
a. shi-ren-yu
eat-man-fish
“piranha”
b. hai-ren-jing
harm-man-monster
“trouble maker”
20. Verb-Object-Instrument
a. xi-yi-ji
wash-clothes-machine
“washing machine”
b. ge-cao-ji
cut-grass-machine
“mower”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
94
21. Verb-Object-Location
a. xi-yi-dian
wash-clothes-store”
“laundromat”
b. shi-pin-dian
eat-product-store
“grocery”
22. Agent-Verb-Object
a. ren-zao-rou
man-make-meat
“vegetarian meat”
b. fo-tiao-qiang
Buddha-jump-wall
“a name of dish”
The above examples inform us that, like phrases, compounds may carry
various argument structures. Examples 15-18 show that the argument (i.e., the
nominal element) in a VN compound may be an agent, an instrument, or an object.
Moreover, examples 19-22 show that verbal components of compounds may take
two arguments. Additionally, in many languages, the agent always appears outside
of compounds. For example, the agent “tiger” must appear outside of the
compound “man-eating” (Di Sciullo & Williams, 1987). However, in Chinese, the
agent is included in the compound in the very same expression (eat-man-tiger)
(Chen, 1990b; Chen & Shi, 1992). The fact that compounds takes various
arguments, including the external argument, suggests the presence of argument
structure with a nominal compound.
Verbal compounds contain argument structures as well. Their argument
structures are less complex than nominal compounds. Verb-Object is the most
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
95
common argument structure of verbal compounds (Chao, 1968; Huang, 1983).
Examples 23a-b demonstrate the verb-object argument structure of verbal
compounds. The morpheme “fan” is the object argument of the verb “chi” (to
eat), and the morpheme “jiao” is the object argument of “shui” (to sleep).
23. Verbal Compounds: Verb-Object
a. chi-fan
eat rice
“to eat”
b. shui-jiao
sleep-sleep
“to sleep”
Thus far, the discussion has demonstrated that a Chinese compound can be
decomposed into sublexical components and structures. These sublexical
components may be inserted into structures or substituted by other sublexical
components. If components can be manipulated with such freedom and lexical
structures can be productively used to create new compounds, there has to be a
level to represent these components and a level to represent word structures. A
single level for the whole word simply is not enough to capture the freedom of
components and the productivity of structures. This proposal resembles the
analysis of phrases, which often assume a level for lexical items and a level for
phrase types (Bates, Friederici, Wulfeck, Juarez, 1988; Bates, Harris, Marchman,
Wulfeck, & Kritchevsky, 1995; Marchman, Harris, Juarez, & Bates, 1992).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
96
3.2.3 The Interaction Compounds and Phrases
Compounds and phrases are similar because both carry a level for lexical
items (or sublexical items) and a level for structures. This similarity may trigger an
influence of phrases on compounds and vice versa. This influence of phrases is
observed in the re-analysis of compounds as phrases. For example, the English
word “humor” is translated as a VN compound yu-mo (meaning “humor”) in
Chinese. This compound is reanalyzed as a VO phrase, which allows the insertion
of the aspect marker -le in 24b, pronoun and numeral in 24c, or aspect marker -le,
pronoun and numeral in 24d. The influence of compound is observed in the re-
analysis of phrases as compounds. Examples 25a and 25b show that the relative
clause [yao fan de] (who beg for food] is re-analyzed as an object name, referring
to a “beggar.” In the re-analysis process, the function of relative clause marker
(or RCM) -de is changed into a derivational maker, functioning as a nominalizer.
24. Re-analysis of a Compound as a Phrase
a. yu-mo
humor-humor
“humor; to humor”
b. yu le mo.
humor aspect marker humor
“to pass a humor.”
c. yu ni yi mo.
humor you one humor
“to give you one time of humor.”
d. yu le ni yi mo.
humor aspect marker you one humor
“gave you one time of humor.”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
97
25. Re-analysis of a Phrase as a Compound
a. [yao fan de] ren.
[beg-rice RCM ] man
“The man who begs for food.”
b. nei ge [yao-fan-de]
that classifier \beg-rice-nominalizer]
“the beggar”
The influence of compounds on phrases is observed in the case that phrases may
undergo a kind of morpheme deletion process to yield compounds without
changing their word orders. For example, the noun phrases in 26a is written as a
compound 26b after several morphemes (such as country, to make, RCM, clothes
and instrument) are deleted.
26 a. mei-guo zhi-zao de xi yi-fu de
USA-country manufacture-make RCM wash clothes-clothes RCM
ji-qi
machine-instrument
b. mei-zhi-xi-yi-ji
USA-made washing-machine
The lexical expansion, opposite to the morpheme deletion process, may change a
lexical compound 26b into a phrase 26a. The feeding between compounds and
phrases is bi-directional (Chen & Shi, 1992).
3.2.4 Summary
To summarize, this chapter presents five linguistic properties of Chinese
compounds to argue for the existence of a sublexical level and a word structure
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
98
level. The semantic independence of the component morpheme, the assignment of
grammatical category to the component morpheme, and the freedom of the
component morpheme in compound formation are three linguistic arguments for
the claim that component morphemes require an independent sublexical level that
is separable from the whole word level. The presence of word-class substructure
and argument structure serves as evidence for the presence of the word structure
level. Because compounds and phrase are alike, they can feed each other. As a
result, a phrase may be derived from a compound and vice versa.
With these linguistic arguments in mind, the present study looks for
neurolinguistic evidence for the presence of the whole word level, the sublexical,
and word structure levels. It will investigate noun-verb dissociations between
Broca’s aphasia and Wernicke’s aphasia at these three levels. Specifically, if the
three levels are real to a Chinese speaker, we expect to find noun-verb
dissociations in Broca’s and Wernicke’s aphasia at these three levels. Moreover,
using frequency effects as supplemental evidence, we expect to find frequency
effects at three levels as well. The experiment to achieve the purpose of the
present study is elaborated in the following chapter.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
99
Chapter Four
Method
This chapter addresses the method of the present study. An experiment of
object and action naming was conducted to investigate two related issues: noun-
verb dissociations and frequency effects. The use of the object and action naming
test allows us to make a direct comparison with findings reported in the study by
Bates et al. (1991), with different subjects and modified materials. Modified
materials included more unproblematic testing items and excluded problematic
testing items used in the study of Bates et al. (1991).9 There are two groups of
brain damaged aphasic subjects and a group of normal controls in the present
study. The criteria for selecting subjects are provided in the following section,
which includes the background information of each aphasic subject, and the
background information of normal controls who are matched with aphasic subjects
on age, education, sex, and native language. This chapter also discusses the
testing procedure and two methods of data analysis. One method is the analysis
9 Because most verbal VNs may carry compound readings as well as phrasal readings, it is crucial
to design a test which can elicit compound readings but avoid phrasal readings. To achieve this
goal, the tester was instructed to ask subjects names of actions rather than description of actions.
This approach is proved successful by the theoretical and empirical support. First, there is a
distinction among languages that names are lexical, by contrast, descriptions are phrasal
(Russell, 1985). Moreover, numerous experimental studies have shown that object and action
naming tasks are able to demonstrate dissociable lexical impairments in aphasia (Bates et al.,
1991; Bemdt et al., 1997a; Bemdt & Zingerser, 1991; Hart et al., 1985; Luria, 1962; Miceli et
al., 1984; 1988; Osman-Sagi, 1987). From the theoretical and empirical support, the present
study chose an object and action naming experiment to address the issue of noun-verb
dissociations.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
100
for noun-verb dissociations and the other is the analysis for frequency effects.
Both analyses were based on the data collected from the experiment of object
naming and action naming.
4.1 Subject Selection: Background Information & Criteria
Subjects for this experiment included 10 Broca's aphasics, 10 Wernicke's
aphasics, and 9 controls that were roughly matched in sex, age, and education. All
aphasics were in-patients or out-patients at either Taipei Veteran Hospital, Taiwan
University Hospital, or Tri-Military General Hospital in Taipei. Controls were
selected in an informal interview. These normal controls showed intact speech,
language, and cognitive abilities. All subjects were native speakers of Mandarin
(some subjects were also able to speak Taiwanese or Hakka). Appendix I
summarizes demographic and neurological information for each subject in the
study. Aphasic subjects were classified mainly according to a Chinese version of
the Boston Diagnostic Aphasic Exam (henceforth BDAE). Additional inclusionary
and exclusionary criteria outlined by Bates and her collaborators (see the papers in
Brain and Language 41, 1991) were also used to assist in subject classification.1 0
Patients were classified as Broca's aphasia if they met the following definition:
reduced fluency and phrase length and a tendency toward omission of function
101 departed from the Bates et al. criteria in one respect: while they require that all patients be at
least 3 months post-onset, we tested some patients who were within 1-3 months from their stroke.
This compromise was necessary because of practical restrictions on hospitalization time and
availability of subjects.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
101
words (such as aspect markers, prepositions, classifiers, etc.) relative to normals in
the Chinese language. Patients were classified as Wernicke's aphasia if they fit the
following definition: fluent or hyperfluent expressive language, where fluency is
accompanied by marked word finding difficulties, semantic paraphasia, together
with clinical evidence of an impairment in language comprehension.
4.2 Testing Materials
The testing material of the present experiment is a revised version of object
and action naming used in Bates et al. (1991). Bates et al. adopted object and
action names from Miceli et al. (1988) and translated these names into Chinese.
As pointed out by Zhou et al. (1993), culture-specificity of some items (e.g. “to
take a shower” is not common, compared to “to take a bath in a bathtub”) and
problematic verbal VNs, where both phrase readings and compound readings are
available, are two important factors which have been neglected in Bates et al.
(1991). With this in mind, the present study revised testing materials by providing
object and action names which are encoded by true compounds. These objects and
actions are very common in the Taiwan society. Like the materials used by Miceli
et al. (1988) and Bates et al. (1991), the present study contained more object
names than action names because it proved somewhat more difficult to obtain a
consensus among normal speakers regarding the word that should be used to
describe the action in a picture (Bates et al., 1991, p. 215).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
102
Object names are represented with five word structures: four kinds of true
compounds and a single-morpheme word structure. The four nominal compound
types are: nominal NN compounds, nominal VN compounds, nominal VNN
compounds, and nominal NNN compounds. Single-morpheme word structure is
nominal N structure. Action names are represented by verbal VN only.1 1 There
are 28 nominal VN compounds, 33 verbal VN compounds, 62 NN compounds, 22
VNN compounds, 20 NNN compounds, and 50 single N words in the experiment.
Single-morpheme words will only be used in the discussion of frequency sensitivity
to word structure. Each item is depicted in one picture. The pictures used by
Miceli et al. (1988) and Bates et al. (1991) were black and white drawings.
However, these pictures often caused difficulties for subjects in recognizing the
actions or the objects (observed by the author when she tested both aphasic and
normal subjects). In an attempt to avoid this problem, most of pictures used in the
present experiment were color versions adopted from commercial catalogs in
Taiwan. Some pictures were black and white drawings because the color version
could not be found in the commercial catalogs. These black and white pictures
actually proved easier than the color ones to recognize in the pre-test (among both
aphasics and normals). In the actual test, none of the test items are unfamiliar to
Taiwanese subjects. Sample pictures are provided in Appendix IV. Appendix II
" The present study contained only VN action names. W compounds were not included in the
action naming task because it was very difficult to obtain a consensus among speakers regarding
the word that should be used to describe the picture for a W compound.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
103
provides ail the testing items in six word structures with their English translations.
Frequency of each testing compound and component is illustrated in Appendix III.
4.3 Testing Procedure
All the Broca’s and Wernicke’s aphasic subjects, who had been assessed
with the BD AE by speech therapists of local hospitals, were also given the naming
test from the BDAE. (Goodglass & Kaplan, 1983). This naming test consists of
six object names and six action names. The six object names are “chair,” “key,”
“glove,” “feather,” “hammock,” and “cactus.” The six action names are
“dripping,” “smoking,” “running,” “drinking,” “sleeping,” and “falling.” Aphasics
were selected as subjects if they could name three out of twelve items. This
criterion is necessary for excluding patients who do not provide names or provide
only unintelligible jargon. The instruction for both the BDAE naming test and the
naming test of the present experiment is “xian-sheng (xiao-jie, tai-tai), qing ni gao-
su wo tu-pian li dong- xi (dong-zuo) de ming cheng” [(Mr.(or Miss, Mrs.), would
you please tell me the name of the object (or the action) in the picture].
Two kinds of cues are allowed in the naming test. First, the tester points to
the target if the subject is distracted by a non-target item in the testing picture. For
example, the picture of “vase” is a vase with flowers in it. If the subject gives
“flower” instead of “vase,” the tester points to the vase and asks the subject the
name of the object “vase.” Second, the tester performs the action without any
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
104
verbal cue if the subject is not sure which action is to be named. For example, the
picture of “raising-hand” is a boy who raises his left hand with his mouth open. If
the subject gives a name like “smile” or “talking,” the tester would perform a
“hand-raising” action and ask the subject the name of the action demonstrated by
the tester.
4.4 Data Analysis
Two sets of data analysis were conducted in the present study. The first
set focused on the noun-verb distinction. The second set emphasized the
frequency sensitivity.
4.4.1 Criteria for Noun-Verb Distinction
Items were classified as “lexically correct” if the target word or an
acceptable synonym in the same word class was produced (i.e., a synonym of the
noun target; a synonym of the verb target). Note that acceptable synonyms did not
have to correspond to the same compound type as the target. For example, one of
the items in the NN category was designed to elicit the word “lan-hua” (orchid-
flower). However, responses on this item were scored as lexically correct if the
speaker produced “lan-hua” (the expected compound), “hua” (flower, a lone N) or
hu-die-lan (butterfly-butterfly-orchid, an NNN compound referring to a common
type of orchid in Chinese culture). These responses differ in their semantic
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
105
specificity (from a generic flower to a specific “butterfly orchid”), but they are all
acceptable responses for normal adult speakers of Chinese. Items were scored as
“lexically incorrect” if they were either not in the expected grammatical category
(nouns for verbs; verbs for nouns), not acceptable synonyms, or belonged to the
categories “total omission” (referring to no response or an empty circumlocution
such as this...this...I know...this is for that...) or “uninterpretable jargon.”
Following the criteria set forth by Bates et al. (1991), items were classified
as “grammatically correct” at the whole word level if the subject's response was in
the same grammatical category as the target, regardless of its semantic accuracy.
Thus, if the subject gave a verbal VN “zhu-fan” (literally cook-rice, meaning “to
cook”) or a single verbal V “chi” (eat) for the verbal target “chao-cai” (literally
stir-vegetable, meaning “to stir”), a verb lexicalization was credited. However, if
the subject responded to the same verbal target (namely chao-cai “to stir”) with a
nominal response such as “fan” (rice) or “guo-zi” (literally v/ck-suffix, meaning
“wok”), the response was scored as grammatically incorrect. The categories “total
omission” and “unintelligible jargon” were also scored as “unlexicalized or
grammatically incorrect.” Because there is no inflectional morphology in Chinese,
no morphological marking is available to a compound. Moreover, syntactic cues
are often lacking in the context of naming. The meaning of a lexical item is the
primary cue to tell the grammatical category of a lexical item. The raw data
collected in the present study showed a significant number of the initiating phrase
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
106
“this is.” This phrase cannot be used to determine grammatical category because it
may be followed by the name of the action or the name of the object. In this
regard, no grammatical category can be assigned to a jargon because there is no
semantic content in the jargon.
For both verbal VN compounds and four types of nominal compounds,
components were scored as “lexicalized at the sublexical level” if the components
had the same grammatical category as the components of the target regardless of
its semantic accuracy. Following this criterion, if a subject responds to “chi-fan”
(literally eat-rice) with another VN “he-shui” (literally drink-water), then both the
verbal component “he” (to drink) and the nominal component “shui” (water) are
credited as lexicalized elements. On the other hand, if the subject responds to “chi-
fan” (literally eat-rice) with a NN compound “shu-cai” (literally vegetable-
vegetable), only the nominal element “fan” (rice) is credited as a lexicalized
component “fan” (rice) and “cai” (vegetable) belong to the same grammatical
category). The verbal component “chi” (eat) is not credited because the subject
does not produce a verbal component. Thus, nonlexicalized components refer to
the ungrammatical components, omissions of components, and unintelligible
jargons.
To investigate the interaction between baseline word structure biases and
performance on the various compound items, responses on each target type were
classified into one of ten possible whole-word classes, which are six noun types at
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
107
the whole word level (VN, NN, VNN, NNN, W N , and N), and four verb types at
the whole word level (VN, W , VNN, and V). These word structures were
distinguished by the grammatical category of the whole word as well as the
grammatical category of the component. For any given item, responses of the
whole word (not including false starts or repetitions) were classified into one of
these ten categories. Examples of the six nominal word structures are given in
Table 3 (from a to f ) and examples of the four verbal types are also given in Table
3 (from g to j).
4.4.2 Coding in Frequency Analysis
Three kinds of analysis were conducted regarding frequency effects on
lexical production. These three analyses looked at frequency effects at the whole
word level, the component level, and the word structure level. Frequency of
occurrence of the whole word is based on the frequency count by Liu, Zhuang, &
Wang (1975). For this purpose, I divided all the compounds into three frequency
groups: high-frequency, mid-frequency, and low-frequency. High-frequency refers
to occurrence of 15 times or more (per million), mid-frequency refers to 5-14
times, and low-frequency refers to 0-4 times in the corpus. The number of items
for each group is: High-frequency: 45, mid-frequency: 40, low-frequency: 80.
At the whole word level, only high-frequency items and low-frequency
items were compared in the present study. The mid-frequency group was left out
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
108
Table 3
Nominal Word Structures vs. Verbal Word Structures
Nominal Word Structures
a. VN fei-er fly-goose “flying-goose’
b. NN xiang-yan perfume-smoke “cigarette”
c. VNN da-zi-ji hit-character-machine “typewriter”
d. NNN yu-mao-qiu feather-fur-ball “badminton”
e. W N ji-suan-ji count-count-machine “calculator”
f. N hua flower “flower”
Verbal Word Structures
g. VN shui-jiao sleep-sleep “to sleep”
h. W shui-zhao sleep-finish “to fall asleep’
i. VNN tang-tou-fa hot-head-hair “to perm”
j-v chang sing “to sing”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
109
to give a clear boundary between two frequency categories (namely high vs. low).
To show frequency-sensitivity of subjects, two analyses were conducted for each
lexical level: (i) a comparison of the percent correct of high-frequency with that of
low-frequency words, and (ii) a comparison of the ratio of high-for-low
substitution (high-frequency substitutes for low-frequency) with the ratio of low-
for-high substitution. No jargon data was included in the substitution analysis.
Similarly, two analyses were conducted at the component level. The first
analysis is a comparison of percent correct of high-frequency components with
percent correct of low-frequency components, and the second analysis is a
comparison of high-for-low substitution with low-for-high substitution. Frequency
of occurrence of components is based on the frequency count in the study by
Cheng (1982).1 2 Mid-frequency items were not used so as to give a clear cut-off
space between high-frequency components and low-frequency components. The
three frequency groups are: 500 times or more, 151-499 times, and 0-150 times in
the corpus. The number for each group is illustrated as follows: High-frequency:
138, mid-frequency: 85, low-frequency: 149.
Two analyses were conducted at the word structure level: word structures
without grammatical categories and word structures with grammatical category.
Word structures with grammatical categories were discussed in section 4.1 (see
I2The study by Cheng (1982) was based on the data in Liu, Zhuang, & Wang (1975). Thus,
there should not be any discrepancy between the frequency of whole word and the frequency of
component.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
no
pp. 108-109). The analysis of word structures without grammatical categories
looked at the distribution of different word structures produced by aphasic subjects
and normal subjects.1 3 Four word structures were under investigation, single
morpheme word, two-morpheme compound, three-morpheme compound, and
compounds with four or more morphemes. O f these four word structures, the
two-morpheme word structure is the most frequent word structure in the language,
followed by the single-morpheme word structure, and then followed by the three-
morpheme word structure. This information is provided by Xian-dai Hanyu Pinlu
Cidian (A Frequency Dictionary of Modem Chinese, 1985). The four word
structures appear in different frequencies given by Xian-dai Hanyu Pinlu Cidian (A
Frequency Dictionary of Modern Chinese (1985)), which provides the most
reliable frequency count for word structures. Like most dictionaries, this
dictionary counts word structure frequency in terms of the number of components
within a word. The frequency of each word structure and their ranks (which word
structure is the most frequent one and which is the least frequent one) are stated in
27.
Two frequency counts were given by the dictionary: the frequency of types
and the frequency of tokens. These two frequency counts differ in the order of
frequency. In the token-frequency count, single-morpheme word structures are
1 3 Word structures were not assigned with grammatical categories for two reasons. First, this
approach may exclude the possible confound caused by the grammatical category effect.
Moreover, in most Chinese frequency dictionaries, the grammatical category is not included
when counting the frequency of word structure.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Ill
the most frequent word structures; by contrast, in the type-frequency count, two-
morpheme word structures are the most frequent ones.
27. Word Structure Frequency (based on A Frequency Dictionary of Modern
Chinese (1985)
This discrepancy comes from the disproportionate frequency of function
words, which are often single morphemes (e.g. the aspect marker -le, the
nominalizer -de, the experiencer marker -guo, progressive markers zai- and -zhe,
and pronouns ni, wo, and (a, etc.). Since the present study discusses only content
words, the influence of function words should be excluded. Therefore, the type-
frequency gives a better characterization for the word structure frequency of
content words.1 4 Based on the type-frequency count, the two-morpheme word
structure has the highest frequency, followed by the single-morpheme word
structure, followed by the three-morpheme, the four-morpheme and the five-
morpheme word structures. The number of five-morpheme word structures
decreases dramatically compared with the other four word structures. In the
M The order of frequency does not differ between the token-frequency and the type-frequency
concerning the ranking of two-morpheme, three-morpheme, and four-morpheme word types.
1. Single-morpheme Word: 13,902 types
2. Two-morpheme Word: 63,527 types
3. Three-morpheme Word: 5,355 types
4. Four-morpheme Word: 4,555 types
5. Five-morpheme Word: 176 types 440 tokens
24,548 tokens
1,690,712 tokens
902,096 tokens
11,012 tokens
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
112
present analysis, the four-morpheme word structure and the five-morpheme word
structure were put into the category of “other” (which includes also
uncategorizable items such as phrasal expressions) because their frequency were
relatively low compared with the other three word structures.
In summary, the present study looks at subjects’ ability to produce nouns
and verbs in an object and action naming task. Results regarding noun-verb
dissociations are discussed in chapter five. Additionally, this study also examines
frequency effects at the word structure level, the whole word level, and the
component level. The study of frequency effect is presented in chapter six.
Results of frequency effects on the access of components and word structures
serve as supplementary support for the presence of an intra-lexical level.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
113
Chapter Five
Noun-Verb Dissociations: Results and Discussion
Bates et al. (1991) addressed the noun-verb problem in Chinese aphasia.
Tney compared two aphasic groups in their ability to use lexical verbs vs. nouns
and to use verbal components vs. nominal components. Although their research
provided reliable evidence for noun-verb dissociations at the lexical level between
Broca’s and Wernicke’s aphasics, noun-verb dissociations at the sublexical level
required further investigation. Moreover, Bates et al. reported a high preference
for two-morpheme structures across subject groups. This word structure effect in
lexical production suggests a possibility that word structures may be processed like
lexical items, both of which are subject to item access principles such as frequency,
uniqueness, etc. (Bates et al., 1995). To show the sublexical and word structure
effects, three sets of analyses on the data from the object and action naming test
were conducted in the present study. The first set of analyses explored the
production of lexical verb/noun and verbal component/nominal component items
using a “percent correct” analysis and a “grammatically correct” analysis; a second
set of analyses was performed to investigate the difference between the two
aphasic groups in the production of sublexical errors; and a last set of analyses
investigate differences between the two aphasic groups in the production of word
types which are assigned grammatical categories. Results of these analyses will be
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
114
used to address the issue of noun-verb dissociations at the lexical level, the
sublexical level, and the word structure level.
5.1 Differences in the Ability to Use Verbs and Nouns: Lexical and Sublexical
The present study extended the testing materials of Bates et al. to four
types of nominal compounds (VN nominal, NN, VNN, and NNN) and one type of
verbal compound (VN verbal). Of the four types of nominal compounds, the VN
compounds make very interesting pairs with verbal VN compounds in that they are
identical with regard to their grammatical categories of their components and their
surface form, but they differ in their grammatical category at the whole word level.
Moreover, nominal VN compounds are true compounds, which are free of the
problems caused by any criticism of compound-phrase confound. A comparison of
the two compound types will test the generality of the double dissociation that
Bates et al. reported at the sublexical level. After comparing these two compound
types for percent correct (i.e., production of the target) and percent grammatically
correct (i.e., production of some kind of noun for a noun target and some kind of
verb for a verb target, at both the lexical and sublexical levels), I then conducted
similar analyses over three additional noun compound types: NN (62 items), NNN
(20 items) and VNN (22 items) (see 5.1.2). None of these are subject to the
phrasal criticism that has been leveled against VN verbs, and all of them provide
further evidence for differential noun and verb biases in the two aphasic groups.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
115
Moreover, it is reported that structures and lexical items are alike regarding their
sensitivity to item-access principles such as frequency (Bates et al., 1995;
MacDonald, Pearlmutter, & Seidenberg, 1994; Trueswell, 1996). By applying the
similarity of structures and lexical items, if lexical items are sensitive to noun-verb
dissociations, word structures with grammatical categories should also be sensitive
to noun-verb dissociations. This would provide further evidence of a “noun bias”
in Broca's aphasics and a “verb bias” in Wernicke's aphasics. Additionally, a
significant number of component errors would indicate that sublexical elements
may be impaired when aphasic subjects attempt to access whole words.
5.1.1 Evidence from Verbal and Nominal VN Compounds
Whole word/Lexical level:
On the 33 verbal VN items, normal controls produced the target form or an
acceptable synonym from the same word class 97.9% of the time (mean = 32.3 out
of 33). This contrasts with 43% correct in Broca's aphasics (mean = 14.2) and
31.2% correct in Wernicke's aphasics (mean = 10.3). Hence, even though these
VN compounds are verbs at the whole word level, Broca's aphasics seem to
produce the target response or an acceptable synonym more often than Wernicke's.
When these scores were subjected to a simple one-way analysis of variance across
all three groups, the result was highly reliable (F(2,28)=82.25, p < .0001).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
116
However, an analysis directly comparing Broca's and Wernicke's aphasics just
missed significance (F(l,18)=3.39, p < .09).
On the 28 nominal VN items, normal controls produced the target form or
an acceptable synonym 95.6% of the time (mean = 26.8), compared with 43.6%
for Broca's (mean = 12.2) and a mere 16.1% for Wernicke's (mean = 4.5). A one
way analysis of variance across groups was highly reliable (F(2,28)=125.19, p <
.0001), and this time the comparison between Broca's and Wernicke's also reached
significance (F(l,18)=21.86, p < .0001). I then compared the two VN compound
types directly in a 2 (Broca’s vs. Wernicke’s) x 2 (compound type) multivariate
analysis of variance, using percent correct as the dependent variable to
accommodate for the difference in number of items on the two compound types.
This analysis determines whether there is indeed a double dissociation between
Broca's and Wernicke's aphasics on these two word types, at the whole word level.
There was a significant main effect of group (F( 1,18)= 12.40, p < .002), reflecting
worse performance overall by the Wernicke's aphasics. There was also a
significant main effect of type (F(l,18)=7.91, p < .012), indicating that nominal
VN compounds are more difficult across the board. Finally, there was a reliable
interaction of group and type (F(l,18)=9.13, p < .007), suggesting that there is
indeed a double dissociation between VN nouns and VN verbs at the whole word
level. However, because of the greater across-the-board difficulty of the VN
nominal items, this is not a full cross-over interaction (see Fig. 2). In fact, separate
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
117
comparisons within groups showed that the difference between VN nouns and VN
verbs did reach significance for Wernicke's (F(l,9)=28.14, p < .0001) but not for
Broca's (F(l,9)=0.02, n.s.).
Fig. 2: Percent production of target
compound
97.9
43 43.6
Broca's Wernicke’ s Normal
Patient Group
i I VN Verbs H i VN Nouns
Examination of the cell means suggests that although normals are close to
ceiling, they also have a harder time with VN nominal compounds compared with
VN verbs. However, a one-way MANOVA comparing the two word types within
the normal control group fails to reach significance, although there is a trend in this
direction (F(l,8)=2.90, p < .13).
In the investigation of the lexical error data, I found that Wernicke's
aphasics tended to make errors which were wrong in meaning, even though their
grammatical categories were correct. For example, the name “pear” was replaced
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
118
with “apple” in object naming, or the verb “to eat” was replaced by a generic verb
“to do” in action naming. In view of this fact, I think that it is necessary to
conduct an analysis which, on the one hand, can avoid the confound from semantic
deficits, and on the other hand, can truly measure subjects' ability to use
grammatical category. The analysis of “grammatically correct,” that is, whether
they could produce (lexicalize) a word that belonged to the same form class as the
target (a verb for a verb target; a noun for a noun target), was conducted for this
particular purpose.
Results using this dependent variable were quite illuminating. For normal
controls, nouns were produced on VN noun targets 99.6% of the time, and verbs
were produced on VN verb targets 99.0% of the time. This constitutes a proof
that our stimuli were successful. For Broca's aphasics, some kind of noun was
produced on VN nominal targets 84.3% of the time, while a verb of some kind was
produced on VN verbal targets only 59.7% of the time. For Wernicke's aphasics,
nouns were produced for VN nominal targets 63.6% of the time, while verbs were
produced for VN verbal targets at a mean rate of 79.7%. As illustrated in Fig. 3,
this appears to constitute a full cross-over dissociation at the whole word level. In
fact, a group (Broca’s vs. Wernicke’s) by type (VN nouns vs. VN verbs)
multivariate analysis of variance yielded no significant main effect of group
(F(l,18)=0.00, n.s.) or type (F( 1,18)= 1.93, n.s.), but there was a highly reliable
group by type interaction (F(l,19)=20.72, p < .0001). Furthermore, a series of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
119
Fig. 3: Percent lexicalization
on VN noun and verb compounds
Broca's Wernicke's Normal
Patient Group
I I VN verbs H VN nouns
additional analyses confirmed that this is a true cross-over interaction. In the
production of nouns on a VN noun target, Broca's aphasics’ results were
significantly better than Wernicke's aphasics (F(l,19)=l 1.05, p < .004) and
significantly worse than normals (F(l, 18)= 12.21 , p < .003). In the production of
verbs on a VN verb target, Broca's were significantly worse than Wernicke's
(F(l,19)=6.85, p < .02) and also significantly worse than normals (F(l,18)=56.39,
p < .0001). Wernicke's were worse than normals on both word types (For VN
nouns, F(l,18)=53.05, p < .0001; for VN verbs, F(l,18)=24.82, p < .0001).
Within Broca’s group, a separate MANOVA comparing the two target types
showed that nouns were lexicalized successfully more often than verbs
(F( 1,9)= 13.65, p < .005); a corresponding MANOVA within the Wernicke's group
showed that verbs were produced successfully more often than nouns (F(l,9)=7.08
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
120
p < .03). In short, when we look at production of the target form class rather than
production of the target word, we find a strong double dissociation between
Broca's and Wernicke's aphasics in the predicted direction. In view of the fact that
the VN noun and VN verb compounds are identical in surface form, this is a very
interesting confirmation of the noun-verb dissociation in these two aphasic groups.
Component/Sublexical Level:
Here again I conducted two analyses: “percent correct” (the production of
the target) and “percent grammatically correct” (i.e., production of some kind of
noun for a noun target, or some kind of verb for verb target), parallel to the whole
word level analysis. The analysis on “grammatically correct” replicates Bates et al.
(1991). Yet, this analysis focuses on whether the speaker is able to lexicalize the
verb and/or the noun element in each VN compound. Credit is given for
production of a verb or noun component whether or not the actual target
component is produced (i.e., whether or not it is the “right” form). Hence, this is a
sublexical variant of the analysis of “grammatically correct” at the whole word
level.
Percent Correct: VN Verbal
On verbal VN compounds, results from the “percent correct” analysis
demonstrated that mean lexicalization rates for nominal element were 63.03% for
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
121
Broca's aphasics and 44.55% for Wernicke's aphasics. The difference between the
two aphasic groups is significant (F (1, 18)=6.9802, p< .017), indicating that
Broca's aphasics are better able to produce nominal elements of verbal VN
compounds. On the same VN verbal compounds, mean lexicalization rates for the
verbal element were 51.2% for Broca's aphasics and 48.79% for Wernicke's
aphasics (see Fig. 4a). The difference between the two groups does not reach
significance (F (1, 18)= 0.1269, n.s.), reflecting only a trend that Broca's aphasics
show a higher ratio of verbal elements compared with Wernicke's aphasics.
Fig. 4a: Percent correct of the noun vs
verb elements of VN verbal compound
Broca's Wernicke's
Patient Group
I i verb element H i noun element
Percent Correct: VN Nominal
The rate of correct responses for verbal element is 45.36% for Broca's and
30.36% for Wernicke's. The difference between two subject groups is significant
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
122
(F (1, 18)=5.824, p< .027), suggesting that Broca's aphasics are better able to
produce verbal elements of nominal VN than Wernicke's aphasics. This pattern
holds in the lexicalization of the nominal element. Specifically, Broca's aphasics
produce 61.05% correct nominal elements in their response for nominal VN
compounds, whereas Wernicke's aphasics produce only 35.7% (see Fig. 4b). The
difference between the two groups is significant (F (1, 18)=12.5619, p< .0023).
Fig. 4b: Percent correct of the noun
vs. verb element of nominal VN compound
70r
Broca's Wernicke's
Patient Group
i ! verb element H I noun element
Thus, Broca's aphasics are better in producing all sublexical targets than
Wernicke's aphasics. This finding is similar to the result from the “percent correct”
analysis on lexical compounds. Since the superiority of Broca's aphasics over
Wernicke's aphasics may be due to the tendency of Wernicke's aphasics to produce
words which are wrong in meaning but correct in grammatical category, I
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
123
conducted another analysis on “grammatically correct” at the sublexical level,
which may characterize the actual difference between Broca's aphasics and
Wernicke's aphasics in their ability to employ grammatical categories such as noun
and verb.
Grammatical Correct: VN Verbal
On VN verbs (which are the same class of items studied by Bates et al.,
1991), mean lexicalization rates for the nominal element are 100% for normal
controls, 90% for Broca's aphasics, and 70.9% for Wernicke's aphasics. A simple
one-way analysis of variance over all three groups is significant (F(2,28)=21.79, p
< .0001). All of the two way comparisons are also reliable (Broca's vs.
Wernicke's, F(1,19)= 13.11, p < .001; Broca's vs. normals, F(l,18)=25.72, p <
.002; Wernicke's vs. normals, F( 1,18)=31.13, p < .0001). On the same VN verbal
compounds, mean lexicalization rates for the verbal element are 99.3% for
normals, 65.5% for Broca's, and 79.7% for Wernicke's. The one-way analysis of
variance over all 3 groups again reaches significance (F(2,28)= 17.71, p < .0001),
as do all of the two way comparisons (Broca's vs. Wernicke's, F(l,19)=4.57, p <
.05; Broca's vs. normals, F(l,18)=39.31, p < .0001; Wernicke's vs. normals,
F( 1,18)= 18.60, p < .0001). To test for the predicted double dissociation, we also
compared the two sublexical components directly in a 2 x 2 MANOVA (group by
sublexical element). There is no main effect of group (F(l, 19)=0.25, n.s.), which
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
124
Fig. 5a: Lexicalization of the noun
vs. verb elements of verbal V N compound
Broca's Wernicke's Normal
Patient Group
i i verb element H noun element
means that the two groups are equally successful (or unsuccessful). There is a
significant main effect of type (F(1,19)=4.87, p < .04), reflecting greater difficulty
overall on the verbal element in a VN compound. Most important for our
purposes here, there is a significant group by type interaction (F( 1,19)=21.79, p <
.0001), confirming that there is indeed a double dissociation at the sublexical level
for VN verbal compounds, as illustrated in Fig. 5a. That is, Broca’s aphasics are
better able to produce nominal components than Wernicke’s aphasics, while
Wernicke’s aphasics are better able to produce verbal components than Broca’s
aphasics. This finding constitutes a clear replication of Bates et al. (1991), using
different subjects and different materials.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
125
Grammatical Correct: VN Nominal
Turning now to the nominal VN compounds, success rates for
lexicalization of the nominal element are 100% for normal controls, 85.4% for
Broca's, and 73.2% for Wernicke's (see Fig. 5b). In a simple analysis of variance
over all three groups, the main effect of group is reliable (F(2,28)= 10.71, p <
.0001). In addition, two-way comparisons show that each of the aphasic groups is
significantly worse than normal controls (Broca’s vs. normal, F(l,18)=16.73, p <
.001; Wernicke's vs. normal, F( 1,18)= 18.68, p < .001). However, the difference
between the two aphasic groups just misses significance (F(l,19)=3.22, p < .09).
On the verbal element of nominal VN compounds, the success rates are 91.7% for
normal controls, 55.7% for Broca's aphasics and 49.3% for Wernicke's aphasics.
100-
90
§ 80
Fig. 5b: Lexicalization of the noun vs.
verb elements of nominal VN compound
100
o ou
S. 20
10
0
— ^ , 1 -------------^ 1 ■ 1 ------------------
Broca's Wernicke’ s Normal
Patient Group
i I verb element noun element
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
126
The analysis of variance across all three groups does reach significance
(F(2,28)=28.99, p < .001), and each of the aphasic groups is worse than normal
controls (Broca’s vs. normal, F(l,18)=49.64, p < .001; Wernicke's vs. normal,
F(l,18)=59.81, p < .001). However, the difference between Broca's and
Wernicke's does not even approach significance (F(l, 19)=0.88, n.s.).
It is obvious that nominal VN compounds do not yield the same kind of
sublexical dissociation that I observe for verbal VN items. To confirm this
impression, I carried out a 2 x 2 (group by sublexical element) MANOVA on
Broca's and Wernicke's only. There is no main effect of group (F(l,19)=2.82,
n.s.), indicating that both groups are equally impaired on these items at the
sublexical level. There is a large and reliable main effect of type (F(l, 19)=45.59, p
< .0001), indicating that all subjects have a harder time lexicalizing the verbal
component in a nominal VN compound. The group by type interaction does not
even approach significance (F(1,I9)=0.52, n.s.). Finally, I conducted individual
MANOVAs within the respective Broca’s aphasic and Wernicke's aphasic groups,
comparing the verb vs. the noun element in nominal VN compounds. In both
groups, subjects have a harder time producing the verbal unit (Broca's,
F(l,9)=38.25, p < .0001; Wernicke's, F(l,9)=14.32, p < .004). Although the
sublexical noun-verb dissociation for VN nouns is not significant, it is illustrated in
Fig. 5b for comparison with Fig. 5a., i.e., the sublexical noun-verb dissociation for
VN verbs.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
127
To summarize so far, like the contrast at the whole word level, the noun-
verb dissociation distinguishing Broca’s and Wernicke’s is not evident in the
analysis of the percent target correct. However, the noun-verb dissociation is
significant in the analysis of the percent “grammatically correct.” This discrepancy
primarily comes from the problem that Wernicke’s aphasics are not able to
produce semantically accurate names and therefore demonstrated a low ratio of
“percent correct.” Since the present study addresses the issue of the noun-verb
dissociation, which is a distinction of grammatical category, results of the analysis
on “grammatically correct” can exclude the confound of semantic deficits. Despite
the strong evidence for a double dissociation at the whole word level for VN
verbal compounds compared with VN nominal compounds, the two compound
types are not comparable at the sublexical level. On VN verbal compounds, the
present study replicates the sublexical dissociations reported by Bates et al. (1991).
But on VN nominal compounds, the same dissociation is not observed. At first
glance, this appears to provide support for the criticism raised by Zhou et al.
(1993), who claim that the sublexical dissociation reported in Bates et al. actually
reflects the phrasal nature of verbal VN items, and hence constitutes nothing more
than a by-product of the syntactic problem experienced by Broca's aphasics.
However, there are two aspects of our findings that mitigate against this
interpretation.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1 2 8
First, there appear to be differences between the two compound types in
baseline difficulty, reflected in the fact that normal controls are also less likely to
lexicalize the verb component in a VN nominal compound (i.e., 91.7% for the verb
element vs. 100% for the noun element). To determine whether this apparent
difference is statistically significant, we carried out a MANOVA comparing noun
and verb elements on VN nominal compounds for normals only. The difference
was highly reliable (F(l,8)=32.67, p < .0001).
Second, Broca's aphasics behave quite consistently across the two
compound types: they have a harder time lexicalizing the verbal element,
independent of grammatical class at the whole word level. This means that (a) the
deficit in verb lexicalization that Broca's aphasics show at the sublexical level is not
simply an artifact of the phrasal status of some VN verbs (against Zhou et al.,
1993), and (b) the disappearance of the double dissociation on nominal VN items
comes entirely from the Wernicke's aphasics. These fluent patients also had a
harder time producing nominal elements on verbal VN compounds, but they had a
harder time producing the verbal element on nominal VN compounds. This
suggests that the absence of a double dissociation at the sublexical level on VN
nominal compounds may be due to baseline differences in item difficulty that affect
performance in all three groups, but had a particularly severe impact on Wernicke's
aphasics. These baseline differences also influenced the performance of aphasics
when they were asked to give names for the other nominal compound types.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
129
The following section presents results of “grammatically correct” analyses
of nominal NN, VNN, and NNN compounds. Hence, they should provide further
insight into the production of complex word types by Broca’s and Wernicke’s
aphasics. It will examine the production of word types as alternatives to VN
targets, NN targets, VNN targets, and NNN targets. Results will help us to
determine whether performance is influenced by high-frequency competitors that
“attract” responses in a particular direction at the lexical and/or sublexical level.
We can also obtain additional evidence that Broca's aphasics disprefer words with
a verb component, while Wernicke's aphasics prefer words with verbal elements.
5.1.2 Evidence from NN, NNN, VNN Nominal Compounds
The three additional compound types at issue here are all nouns at the
whole word level, including VNN compounds. Hence they should provide further
insights into the production of complex word types by Broca's and Wernicke's
aphasics.
The Whole Word Level: NN, NNN, VNN Compounds
On the 62 NN compounds items, lexically correct responses (i.e., the
intended NN or an acceptable noun synonym) were produced 95.2% of the time by
normal controls, 55.3% by Broca's and 27.7% by Wernicke's. An analysis of
variance across all three groups did yield a significant main effect (F(2,28)=73.23,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
130
p < .0001). Both aphasic groups performed poorer than normals (Broca's,
F(l,18)=61.87, p < .0001; Wernicke's, F(l,18)=197.80, p < .0001). The
difference between the two aphasic groups was also significant (F( 1,19)= 18.15, p
< .0001) (see Fig. 6a).
Fig. 6a: Percent correct vs.
lexicalization of N N compound
Broca's Wernicke's Normal
Patient Group
i i percent correct ■ Lexicalization
In the analyses of VN nominal compounds and VN verbs, double
dissociations emerged most clearly when I used a less stringent criterion for
correct performance, accepting any response that was grammatically correct (i.e.,
in the same form class as the target word). Applying the same criterion here, I find
that normal controls produced some kind of nominal response to NN targets
97.1% of the time, compared with 86.9% for Broca's aphasics and 75.2% for
Wernicke's aphasics. The difference across all three groups was reliable
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
131
(F(l,28)=8.99, p < .001), and both aphasic groups were significantly worse than
normal controls (Broca's, F(l,9)=6.93, p < .02; Wernicke's, F( 1,9)= 17.96, p <
.001), but the difference between aphasic groups just missed significance
(F(l, 19)=3.82, p < .07). In other words, the two aphasic groups manage to
produce some kind of noun at the whole word level most of the time, and although
there is a tendency for Broca's aphasics to succeed more often, the difference is
only marginally reliable. When subjects failed to respond with the required noun,
they did so for one of two reasons: because they produced a word that is a verb at
the whole word level, or because they produced jargon or failed to respond at all.
Obviously, if there were differences in rates of jargon and omission, that might
skew the results. However, there were no differences between aphasic groups in
the percent of trials that fall into this “unclassifiable” category (Broca's = 22.4%;
Wernicke's = 25.3%; F(l,19)=0.25, n.s.). I looked at the proportion of all
responses to NN items that were verbs at the whole word level, and found that this
category accounted for only 0.4% of all responses by normals, vs. 3.6% by Broca's
aphasics and 15.3% by Wernicke's aphasics. In this case, the difference between
the two aphasic groups is reliable (F(l,19)=16.88, p < .001), constituting further
evidence that Wernicke's aphasics show a “verb bias” even on NN items, where
there is no verbal element at the lexical or the sublexical level (see also Fig. 6a).
On the 20 items designed to elicit NNN compounds, normals produced the
target form (or a noun synonym) 97.8% of the time, compared with 43% for
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
132
Broca's aphasics and only 16% for Wernicke's aphasics (see Fig. 6b). The analysis
over all three groups reached significance (F(2,28)= 72.33, p < .0001); both
aphasic groups were significantly worse than normal (Broca’s, F(l, 18)=52.35, p <
.0001; Wernicke's (F(l,18)=400.23, p < .0001). The difference between the
aphasic groups was also reliable (F(l, 19)= 11.41, p < .003). If we use the less
stringent criterion of “grammatically correct,” then normals produce some kind of
noun at the whole word level 100% of the time, in response to NNN targets,
compared with 81.5% for Broca's aphasics and 74% for Wernicke's aphasics (see
Fig. 6b). Once again, the difference between the two aphasic groups does not
reach significance (F(l,19)=0.78, n.s.), even though Broca's aphasics appear to be
somewhat more successful. However, when I look at the production of verb
alternatives in response to NNN targets, a group difference emerges once again.
Verb production at the whole word level is uncommon for Broca's aphasics on
these NNN items (3.5% of all responses), but Wernicke's aphasics produce some
kind of a verb at the whole word level 22.5% of the time. The difference between
Broca's and Wernicke's is reliable (F(l, 19)=14.34, p < .001), providing still more
evidence for the verb bias in Wernicke's aphasics. Once again, there is no reliable
difference between the two groups in jargon or omission (F(l,19)=3.83, p < .07),
although Broca's aphasics appear to be worse off on these items (15% jargon or
omission vs. 3.5% for Wernicke's aphasics).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
133
c o
N
_ ra
o
S
o
2
o
o
a >
E
©
CL
Fig. 6b: Percent correct vs.
lexicalization of NNN compound
97.8J00
Broca's Wernicke's Normal
Patient Group
J Percent Correct ■ Lexicalization
The most complex compound types employed in the present study were
VNN compounds, which also function as nouns at the whole word level. Using
the stringent criterion of percent correct production of the target word (or an
acceptable noun synonym), success rates for normals were 97.5%, compared with
only 23.6% for Broca’s aphasics, and 13.2% for Wernicke's aphasics (see Fig. 6c).
All group comparisons were reliably different: across the three groups
(F(2,28)=292.22, p < .0001); Broca's vs. normals (F(l,18)=712.53, p < .03);
Wernicke's vs. normals (F(l,18)=468.40, p < .0001), and Broca's vs. Wernicke's
(F(l, 19)=5.91, p < .03). It is clear that these items pose a substantial challenge for
both groups of aphasic patients. When I use the less stringent criterion of percent
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
134
grammatically correct (i.e., production of some kind of a noun at the whole word
level on nominal VNN compounds), the success rates were 99.5% for normals
(mean=21.9), 70.9% for Broca’s aphasics (mean=15.6) and 54.1% for Wernicke's
aphasics (mean=11.9). Hence performance is much better for all three groups
using this standard (see Fig. 6c). In this case, the difference between Broca’s
aphasics and Wernicke's aphasics is also reliable (F(l, 19)=4.75, p < .05).
C O
N
«
O
8
5
o
tr
o
o
c
< D
®
0.
Fig. 6c: Percent correct vs.
lexicalization of VNN compound
97.5 99 5
Broca's Wernicke's Normal
Patient Group
i I percent correct H I lexicalization
Again, responses were scored as “grammatically incorrect” if they were in
the wrong form class, or if they were total omissions or uninterpretable jargon. To
clarify the nature of the failures experienced by Broca's and Wernicke's aphasics on
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
135
VNN compounds, I carried out an additional analysis restricted to production of
words that are verbs at the whole word level. Such responses were observed only
0.5% of the time in normal controls, but both aphasic groups produced verbs
relatively often on these items (15.9% in Broca's and 24.1% in Wernicke's). In
contrast with the analyses of the two compound types that have no verbal element
( NN and NNN), the difference between Broca's aphasics and Wernicke's aphasics
missed significance for VNN targets (F(l, 19)=2.63, p < .13). Again, there was no
difference between Broca's and Wernicke's aphasics in the jargon/omission
category (F(l,19)=1.39, n.s.).
The contrast between NNN and VNN nominal compounds is of particular
interest here. Both item types are nouns at the whole word level; both are quite
long and complex and both are relatively infrequent compound types (compared
with VN nouns, VN verbs and NN nominal compounds). However, the NNN
items are compatible at both lexical and sublexical levels (i.e., they are “nouns all
the way down”) while the VNN items require one verbal element. If my
arguments about double-dissociations at the sublexical level are correct, then I
should find that Broca's aphasics have more difficulty with VNN than NNN, while
Wernicke's aphasics should have trouble with both sets. The cell means cited
above provide support for this view, but to test it further, I conducted a
MANOVA on percent lexically correct, comparing groups (Broca's vs.
Wernicke's) and item types (VNN vs. NNN). Results included a significant main
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
136
effect of type (F(l,18)=9.0l, p < .01), due to the greater overall difficulty of VNN,
together with a significant main effect of group (F(l,18)=12.74, p < .002),
indicating that the worst performance overall was by Wernicke's aphasics. Most
important for our purposes here, there was a significant interaction between group
and item type (F(l,18)=5.01, p < .04), in the predicted direction. To confirm this
interpretation, I also conducted separate MANOVAs for each of the two aphasic
groups. For Broca's aphasics, production of a lexically acceptable response was
significantly more likely on NNN than VNN (F(l,9)=8.73, p < .02). For
Wernicke's aphasics, there was no difference in the overall difficulty of these two
item types (F(l,9)=0.68, n.s.). These results are presented in Fig. 6b and Fig. 6c.
In the same vein, I carried out a MANOVA comparing VNN and NNN
compounds in the production of a response that is grammatically correct (i.e.,
some kind of noun at the whole word level). In this analysis, there was a main
effect of item type (F( 1,18)= 17.06, p < .001), reflecting the greater overall
difficulty of VNN items. However, the main effect of group was not reliable
(F(l,18)=2.74, p < .12), and there was no significant interaction (F(l,18)=1.59,
n.s.). So the problem lies not with “nouniness” per se, but with the relative
difficulty that Broca's aphasics experience in producing a lexically correct form in
response to VNN items. It appears that the mere presence of a verbal element
within a nominal VNN compound poses a special challenge for Broca's aphasics,
reducing their performance well below the levels observed on NNN or NN.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
137
Because the VNN items are true compounds, and in no way subject to the verb
phrase criticism raised by Zhou et al., we take this as further evidence for a double
dissociation between nouns and verbs at the sublexical level. Conversely, all of
these data provide further evidence for a verb bias in Wernicke's aphasics. Even
though the NN, NNN, and VNN compounds are all classified as nouns at the
whole word level, Wernicke’s aphasics frequently respond by producing verbs.
This tendency will become clearer still as we move to the analysis of alternative
word types that are produced in response to these complex nouns (see section
5.3.2).
Another line of evidence for a noun-verb dissociation comes from the
analysis of errors, where subjects’ responses are lexically related to their target
words (the notion of “lexically related” refers to compound words which share
components). Lexically related errors indicate that subjects are able to activate
word parts. Noun-verb dissociations appear in the activation of word parts should
provide further evidence for the sublexical noun-verb dissociation.
5.2 Differences in Error Types: Lexical and Sublexical
Because the dissociation at the lexical level is quite solid, 1 will not discuss
its significance further. However, the dissociation at the sublexical level seems to
be confounded with several linguistic factors. The best way to address the issue at
the sublexical level is to look at either errors involving component omissions,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
138
component substitutions, or the reversal of components within a compound word.
The following section investigates these component errors. As we shall see, there
is an obvious difference between the two aphasic groups as to their tendency to
omit or substitute verbal vs. nominal components. Moreover, homophone errors
and reversals of component orders are two additional data which support
sublexical noun-verb dissociations. Homophone errors will show the “verb bias”
of Wernicke’s aphasics. Reversals of component orders will demonstrate that
components may be moved within a compound as independent units.
Table 4 shows she types of component errors. Four of them are
substitution errors. The other two types are omission errors. Substitution errors
of type 2 and type 4 involve a change of grammatical category. Type 2 represents
a noun-for-verb substitution (e.g., gou for diao), while type 4 represents a verb-
for-noun substitution (e.g., yao for shui). By contrast, substitution errors of type 1
and type 3 do not involve a change of grammatical category.
There are two possible interpretations for these substitutions. First,
substitution may just reflect a word for word substitution, even though the
substituted and the substitute are morphologically related. Second, the
substitution may reflect a partial success to retrieve the target. The subject
immediately activates another component to make up a well-formed compound,
given that single-morpheme words are not common lexical items. This kind of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
139
Table 4
Six Types of Errors: Criteria for Data Coding
type 1: VN—> VN jiao-hua—> xi-hua
water-flower wash-flower
qi-er— > fei-er
stand-goose fly-goose
type 2: V N ~> NN diao-yu—> yu-gou
hook-fish fish-hook
wan-ju- > ju-zi
play-instrument instrument-Suffix
type 3: VN--> VN tiao-sheng—> tiao-shui
jump-rope jump-water
zheng-long— > zheng-gao
steam-cage steam-cake
type 4: VN-- > VV tiao-shui— > tiao-yao
jump-water jump-jump
fei-ji— > fei-xing
fly-machine fly-walk
ty p e5 :V N --> N chao-cai— > -cai
fty-vegetable -vegetable
zhi-piao— > -piao
pay-bill -bill
type 6: VN — > V hua-chuan— > hua-
row-boat row-
qi-er— > qi-
stand-goose stand-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
140
lexical access involves an operation at the sublexical level. The first interpretation
is ruled out because most of the objects that the substitutes refer to do not appear
in the pictures. For example, “yu-gou” (fish-hook) does not appear in the picture
of “to fish.” Likewise, there is no object denoting “ju-zi” (literally saw-suffix,
meaning “saw” or literally instrument-suffix, an invented word) in the picture of
“wan-ju” (play-instrument). The second interpretation is acceptable because the
subject may activate a part of the target compound and associate the part with
another component to yield a compound that he/she is familiar with.
Homophone substitutions reflect subjects' success in retrieving
phonological or lexical forms but their failure to find their accurate grammatical
categories or semantic functions. There are four kinds of homophone errors (see
Table 5). Type I and type 3 do not involve the change of grammatical category.
For example, the verb component “hua” (to change) is replaced with another
homophone verb component (see examples o f type 1), and “er” as fairy substitutes
for “er” as goose (see examples of type 3). These pairs of homophone
components are identical in their grammatical category but different in their written
forms. The homophone pair of type 3 are like the English homophone pair of
“flower” and “flour.” The homophone pair of type I are like “saw” (as the past
tense o f ‘see”) and the verbal “sow."
Type 2 and type 4 are homophone substitutions involving the change of
grammatical category. The examples of type 2 resemble the contrast of “be” and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
141
“bee.” Examples of type 4 is compatible with the contrast between nominal “love”
and verbal “love.”
Table 5
Four Types of Homophones
Type 1 : verb to verb hua-zhuang — > hui-hua
change-cosmetic paint-paint
Type 2: verb to noun tou-qiu — > tou-fa
throw-ball head-hair
Type 3: noun to noun qi-er --> chang-er
stand-goose moon-fairy
Type 4: noun to verb xie-xin — > xin- ren
write-letter trust-serve
In type 4, the component “xin” may function as a nominal component of
verbal VN “xie-xin” (literally, write-letter, meaning “to write”) or a verbal
component of the verbal W compound “xw-ren” (literally, trust-serve, meaning
“to trust”). The Wernicke’s aphasic chose the verbal homophone (i.e., xin,
meaning “to trust”) to make a verb compound (i.e., xin-ren, literally trust-serve,
meaning “to trust”) when he was asked to provide the verbal VN “xie-xin”
(literally write-letter, meaning “to write”).
The substitution of the verbal component “xin” for the nominal “xin” may
have two indications. First, the subject is able to access the phonological form of
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
142
“xin.” Second, he displays a preference for the verbal function over the nominal
function. These two indications seem to argue against “the lexical account” but
ague for “the semantic-conceptual account.” Specifically, the subject is able to
access the lexical form, their problem with nominal components is not a failure at
the lexical level. Rather, the deficit may occur at the functional-semantic level.
There are several studies discussing the influence of nominal homophones
on the acquisition of verbal homophones in English, Italian, and Chinese (Caselli,
Casadio, & Bates, 1996) and on the use of verbs by bilingual subjects of Chinese
and English (Liu, 1996). The data of verbal “xin” and nominal “xin” suggest that
homophones of type 2 and type 4 should provide further insights into the “noun
bias” vs. the “verb bias” by Broca’s and Wernicke’s aphasics.
Reversals of component order indicate that components may be moved to
different positions as if they are free words although they do not involve
component errors. These reversals were considered significant because they
appeared also in speech/writing errors of normal speakers (Lai, 1984).
Example b in Table 6 shows that the components “qi” (to ride) and “lo”
(floor) are reversed to yield a non-word “lo-qi.” Examples in Table 6 are all
speaking errors. This order reversal of components was also observed in writing
(Chen, 1984). Chen (1984) reported that component order reversals appeared
more often in writing than in speaking. This may be because each component is an
isolating character in writing. However, the spatial isolation of character
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
143
disappears in a speech flow because, in speaking, the component is like a syllable
which functions as a part of the word.
Table 6
Reversal of Components
a. qi-er— >
stand-goose
“penguin”
b. qi-lo — >
ride-floor
“passage “
c. xiao-qing-wa— >
little-green-ffog
“frog”
er-qi
goose-stand
lo-qi
floor-ride
xiao-wa-qing
little-frog-green
5.2.1 Error Analysis: Verbal VN Compounds
This section analyzed patients' responses according to the six error types
displayed in Table 4. Only errors of verbal VN and nominal VN were computed
because these two compound types demonstrate a lexical noun-verb contrast and a
sublexical noun-verb contrast (within each compound type).
Three findings from the error analysis are relevant to my discussion. One
finding supports the presence of the sublexical level. The other two findings can
be used to argue for sublexical noun-verb dissociations.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
144
First, Table 7 shows the distribution of 101 (out of a possible 330)
occurrences of component errors in Broca's aphasics and 114 (out of a possible
330) of component errors in Wernicke's aphasics. These errors constitute about
one third of all the responses. This substantial number of component errors
indicate the freedom that components can be dropped or substituted as if they are
independent words. This independence of component strongly argues against the
proposal that compounds are processed as inseparable whole words (Zhou et al.,
1993) or that components of compounds are like syllables which do not carry any
significant grammatical function (Li & Thompson, 1981). It further constitutes
evidence for an independent sublexical level for components.
If components are grammatically more complex than syllables, they should
be sensitive to grammatical differences. Data used in the discussion are all
substitutions involving a change of grammatical category. There are two kinds of
substitutions: noun-for-verb substitutions and verb-for-noun substitutions. As
shown in Table 7, there are 16 instances of noun-for-verb substitutions in
Wernicke's aphasics, compared with 36 instances in Broca's aphasics. The contrast
demonstrates a higher noun preference in Broca's aphasics than in Wernicke’s
aphasics. Regarding verb-for-noun substitutions, Wernicke's aphasics made 14 of
these substitution errors, while Broca's aphasics committed such substitution errors
only two times. In comparing Wernicke’s aphasics with Broca's aphasics,
Wernicke's aphasics tended to produce more verb-for-noun substitutions, while
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
145
Broca’s aphasics tended to produce more noun-for-verb substitutions. This
contrast demonstrates another aspect of the noun-verb dissociation-this time in
substitutions. The noun preference found in Broca's aphasia and the verb
preference found in Wernicke's aphasia suggest a sublexical double noun-verb
dissociations between the two aphasic groups.
This contrast in the preference for noun vs. verb elements also holds in the
omission errors. As shown in Table 7, there are 27 instances of verb omissions in
Broca's aphasia but only 10 errors in Wernicke's aphasia. In contrast, there are 32
instances of noun omissions in Wernicke's aphasics but only 6 errors in Broca's
aphasics. This contrast of omission errors again indicates a noun preference in
Broca's aphasics and a verb preference in Wernicke's aphasics, arguing a double
noun-verb dissociation at the sublexical level.
Additionally, there are noun-for-noun (type 3) and verb-for-verb (type 1)
substitutions, which do not involve changes of grammatical category and therefore
cannot be used to show a noun-verb contrast. These errors are relevant to the
contrast of substitution vs. omission found in Broca's aphasics and Wernicke’s
aphasics. Table 7 shows that Wernicke’s aphasics produce 20 cases of verb-for-
verb substitutions and 22 cases of noun-for-noun substitutions, while Broca's
aphasics commit verb-for-verb substitutions 14 times and noun-for-noun
substitutions 16 times. Thus, instead of a contrasting ability to access nouns/verbs,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 7
Numbers of Component Errors for Six Error Types: Verbal VNs
146
Types Broca’s Wernicke's
1 V*N - > VN
16 22
2 VN -> N N 36 16
3 VN ~>VN 14 20
4 VN ~> VV 2 14
5 VN ~ > N 27 10
6 VN — > V 6 32
Total 101 114
* Underlined V or N are components involving substitutions.
there is an overall higher ratio of substitution errors in Wernicke's aphasics,
compared with Broca's aphasics. This is a pattern that has been reported
repeatedly in the literature (Bates, Friederici, Wulfeck, 1987b; Butterworth, 1979;
1983; Butterworth & Howard, 1987; Goodglass & Kaplan, 1983). Reversals in
this general tendency (with Broca's providing more substitutions) occur only in
those categories where a verb element is replaced by a noun or left out altogether.
The discussion above demonstrates a clear noun-verb double dissociation
between Broca's aphasics and Wernicke's aphasics in their production of
component errors in response to the tasks for eliciting verbal VN compounds.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
147
This finding is compatible with the results from the analysis of “grammatically
correct” responses. In my previous analysis, there seemed to be a difference
between verbal VNs and nominal VNs. For verbal VNs, the two aphasic groups
were sensitive to the noun-verb difference. However, it seemed harder to find a
noun-verb difference between the two aphasic groups on nominal VNs compared
with subjects' performance on verbal VNs. Although the lack of significance has
been attributed to the impact of the major word type (NN word type), we need to
find out whether the lack of significance is due to a qualitative difference between
nominal VN and verbal VN (nominal VNs are compound words, while verbal VNs
are phrases). If this qualitative difference holds, we should not find a sublexical
noun-verb contrast between Broca's aphasia and Wernicke's in their production of
nominal VNs. If the difference is not qualitative, we expect to find the same
sublexical contrast in nominal VNs, which may quantitatively differ from verbal
VNs.
5.2.2 Error Analysis: Nominal VN Compounds
Table 8 shows that Wernicke's aphasics make 90 component errors out of
280 attempts and Broca's aphasics make 71 component errors in their response to
nominal VN compounds. Their errors correspond to rates of approximately one-
third and one-fourth of the total. Compared with the one-third error rate on verbal
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
148
VNs, the error rate for nominal VNs represents a slight overall decrease. The
higher ratio of component error in verbal VNs over nominal VNs may be due to
the fact that verbal VNs have a higher separability than nominal VNs (see
discussion in 3.1). Although there is a quantitative difference between verbal VNs
and nominal VNs, component errors appear in both high separability and low
separability cases. This argues that components function as independent
processing units, which require a sublexical level to separate them from lexical
items.
While showing the independence of components, I am also interested in
determining whether there is a sublexical noun-verb dissociation within a nominal
VN compound— the very same question discussed in the preceding section for
verbal VNs. Looking at component errors in both verbal VNs and nominal VNs, I
find that aphasics respond to these two VN compounds in a very similar fashion.
That is, Broca's and Wernicke's aphasics show the same contrasting profiles at the
sublexical level regardless of the VN compound's status at the whole word level.
This similarity is shown in three ways. First, Wernicke's aphasics make more verb-
for-noun substitutions (N= 8) than Broca's aphasics do (N=0). Second,
Wernicke's aphasics omit nominal components more often than Broca's aphasics,
although their difference is quite small (Broca's made 3 such errors, while
Wernicke's made 5). The last similarity is also found in omission errors. Broca's
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
149
aphasics omit the verbal component 27 times, which is more frequent than 12
omissions made by Wernicke's aphasics.
Table 8
Numbers of Component Errors for Six Error Types: Nominal VNs
Types Broca’s Wernicke's
1 V*N— > VN
11 26
2 VN -> N N 19 23
3 VN_~> VN 11 16
4 VN -> V V 0 8
5 VN ~ > N 27 12
6 VN ~ > V 3 5
Total 71 90
" “ Underlined V or N are components involving substitutions.
For the category of noun-for-verb substitutions, both groups tended to
replace VN with NN, with a slight difference— Wemicke's aphasics produced a
slightly higher number of NN than Broca's aphasics (N=23 vs. N=19). The
tendency that both aphasic groups replaced VN with NN reflected a sensitivity to
the most frequent word type of nominal compounds. A slightly higher number of
replacing NN for VN in Wernicke’s aphasics constituted nothing more than a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
150
general trend that Wernicke’s aphasics made more substitution errors than Broca’s
aphasics. The finding also informs us that the NN effect is active not only at the
whole word level, where the substitutes do not share any component with their
targets, but also at the component level, where the substitutes share at least a
component with (or morphologically related to) their targets.
To summarize, results from the analysis of component errors show a higher
ratio of verb-for-noun substitution in Wernicke’s aphasia. This “verb bias”
constitutes one line of evidence for a noun-verb dissociation in nominal VNs,
which are true compounds. The lack of significant difference in noun-for-verb
substitutions prevents us from claiming a double dissociation at the sublexical
level. Instead, this insignificant difference reflects a major NN word type effect
interacting with a sublexical effect (a part of word was substituted). Moreover,
although substitution errors do not show a double noun-verb dissociation,
omission errors do show a tendency of such dissociation, i.e., Wernicke's aphasics
tend to omit nominal components more often than Broca's aphasics, while Broca's
aphasics tend to omit verbal components more often than Wernicke's aphasics.
5.2.3 Homophones: Verbal and Nominal VN Compounds
Table 9 shows that only Wernicke's aphasics make errors which are
homophones of their targets. No such error is observed in Broca’s aphasia. This
contrast may be partly due to the fact that substitution errors are substantially less
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
151
Table 9
Homophone Substitution Errors of Wernicke’s Aphasics
Type 1: verb to verb
1. hua-zhuang
— >
hui-hua
change-cosmetic paint-paint
2. gui-xia
— >
gui-zhong
kneel-down expensive-heavy
3. hua-xue
— >
hua-chuan
slide-snow row-boat
Type 2: verb to noun
1. tou-qiu
— >
tou-fa
throw-ball head-hair
Type 3: noun to noun
I. qi-er
— >
chang-er
stand-goose moon-fairy
2. wo-shi
— >
hu-shi
lie-room protect-person
3. wo-shi
— >
fang-shi
lie-room room-business
Type 4: noun to verb
1. xie-xin
— >
xin-ren
write-letter trust-trust
2. Jiao-hua
— >
hua-ming (fa-ming)
water-flower invent-bright
3. gang-qin
— >
gang-kai-shi
steel-piano just-open-start
4. wo-shou
— >
shou-lu
hold-hand keep-road
5. you-yong
— >
yong-che
swim-swim use-car
6. wei-qi
— >
lo-qi (qi-lo)
surround-chess floor-ride
7. wei-qi
— >
qi-lo
__smxotuQtLQhess____ _______________
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
152
frequent in Broca’s aphasia than in Wernicke’s aphasia, and may be partly due to
the fact that homophone pairs are less common than unspecified lexical pairs. Out
of the four error types, verb-for-noun is the most frequent error type, and noun-
for-verb is the least frequent error type. This once again demonstrates a “verb
bias” in Wernicke's aphasia. Although there is not enough data to run a statistical
analysis, it does demonstrate a trend toward a verb bias in Wernicke's aphasia.
5.2.4 Reversal: Verbal and Nominal VN Compounds
Reversing component order within a compound demonstrates that
components can be moved as free elements. This sort of data also serves as
evidence for the existence of the sublexical level. This error type is not as frequent
as the other error types in the present study. However, this error has been
reported in aphasia (Chen, 1984; 1990) and normal speakers (Lai, 1984).1 5 The
following Table demonstrates speaking errors of order reversal.
1 S While I was teaching Chinese at USC (1988-1992), I also observed that second language
learners were often confused with the order of components errors, in particular cases in which
reversing components yielded acceptable compounds (e.g. tooth-brush vs. brush-tooth).
Moreover, my first child made several unacceptable words involving the reversal of component
order. For example, she put the suffix “zi” (a diminutive marker) at the prefix position to yield a
nonword “zi-zhuo,” which, in her intention, meaned “zhuo-zi” (literally table-suffix, meaning
“table”). These errors of order reversal demonstrate that, like normal speakers and aphasics,
second language learners and native children both show a tendency to reverse the component
order.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
153
Table 10
Reversal of Components
qi-lo — > lo-qi
ride-floor floor-ride
qi-er —> er-qi
stand-goose goose-stand
xiao-qing-wa --> xiao-wa-qing
little-green-frog little-frog-green
5.2.5 Summary
I have shown that components of compounds are substituted and omitted
frequently by aphasic patients. These error types all suggest that components are
relatively free to move within a lexical domain. The freedom of components
provides significant evidence for the presence of a sublexical level of organization.
Additionally, Wernicke's aphasics tended to produce verbal errors (shown in
substitutions and homophones) more often than Broca's aphasics, while Broca’s
aphasics tended to drop verbal components more often than Wernicke's aphasics.
This result provides evidence for a noun-verb dissociation at the level of sublexical
components.
The following section will examine the production of word types as
alternatives to VN targets, NN targets, VNN targets, and NNN targets. Results
will help us to determine whether performance is influenced by high-frequency
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
154
competitors that “attract” responses in a particular direction at the lexical and/or
sublexical level. We may also obtain additional evidence that Broca's aphasics
have difficulty with words with a verb component, while Wernicke's aphasics
gravitate toward words with verbal elements.
5.3 Differences in the Use of Alternative Word Types
In analyzing nominal VN compounds, I reported a word type effect (NN)
which is biased against a full cross-over noun-verb dissociation. While doing the
same analysis, I was attracted by another interesting phenomenon: Broca's aphasics
and Wernicke's aphasics differ in their preference for verbal word types and
nominal word types. So far, this observation has not been reported in any
literature of noun-verb dissociations. This interesting observation deserves a closer
look.
There are two possible structural analyses in Chinese. One way is to assign
word structures by grammatical categories, resulting in patterns like NN, VN, etc.
Alternatively, a word may be characterized in terms of morpheme numbers. This
approach yields patterns like single-morpheme word structures, two-morpheme
word structures, and three-morpheme word structures. No grammatical category
is assigned in this analysis. Results for word structures organized by grammatical
category are useful for the discussion of noun-verb dissociations and its interaction
with frequency effects. Results for word structures without grammatical category
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
155
specifically address the issue of frequency effects, including its interaction with the
whole word level, the sublexical level, and the word structure level.
To show that word types may be interacting with noun-verb differences,
this section examines five word structures assigned grammatical categories.
Special attention is paid to the pair of VN word structures, which show a noun-
verb contrast at the whole word level (between verbal VN and nominal VN) as
well as the sublexical level (verbal component and nominal component within a
compound). Numbers of various word types produced in place of VN compounds
will also be compared between aphasic groups.
5.3.1 Verbal and Nominal VN Compounds
In section 3.2. l.b, I demonstrated that nouns and verbs differ in their major
word structures. Nouns frequently contain NN patterns, while verbs frequently
bear either W or VN patterns (Huang, 1992). This difference in the major word
structures suggests that the grammatical category of a compound may be
distinguished by its word structure. For instance, native speakers tended to treat
NNs as nouns and W s/VNs as verbs. Because of this special relation between
word structure and grammatical category, it is possible that the noun-verb
dissociation is observed at the level of word structure. Moreover, because the
major word type is high in its frequency of occurrence, the frequency effect may
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
156
facilitate the lexical access of aphasics. As a result, noun-verb dissociations
between Broca’s and Wernicke’s aphasics are less obvious when target words are
constructed with the major word type. In this regard, non-major word types may
serve as a better tool to reveal noun-verb dissociations.
Table 11 provides the ratio of the word types that patients and controls
produce in response to VN nominal and verbal compounds. Although no single
category predominates, it is clear that patterns of substitution differ for the two
target forms. It is also clear that Broca's aphasics exhibit a consistent “noun bias”
and Wernicke's aphasics continue to show a “verb bias” at this relatively fine
grained level of analysis.
Nominal VN: Word Type Analysis
First, considering the production of nominal VN word type (target) and
verbal VN word type for VN nominal compounds, Broca’s aphasics produce
47.5% nominal VN word types while Wernicke’s aphasics produce the same word
type only 27.9% of the time. This difference is significant (p<006). The
frequency of verbal VN word type is less. For Wernicke’s aphasics, there is 12.9%
verbal VN word type; for Broca’s aphasics the ratio decreases to 4.3%. The
difference between Broca’s and Wernicke’s aphasics reaches significance
(p<0001). Because, in the production of the verbal VN word type and the
nominal VN word type, Broca’s and Wernicke’s aphasics displayed a pattern that
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
157
looked like a double dissociation, I did a 2x2 MANOVA comparing nominal VN
vs. verbal VN in Broca’s vs. Wernicke’s. There was no main effect of group
(F(l, 19)=2.68, n.s.), but there was a significant effect of type (F(l, 19)=85.25, p <
.0001), indicating more correct nominal VN word type than incorrect verbal VN
word type. More important for my purposes here, there was a significant group by
type interaction (F(l,19)=20.02, p < .0001). Hence, the double dissociation
between Broca's and Wernicke's was reliable, in the predicted direction. This
finding led to a comparable sub-analysis for the verbal VN word type in the section
on verbal VN word type.
Second, considering the substitutions produced on VN nominal
compounds, normal controls make almost all of their substitutions from the high-
frequency NN category (mean = 2.33 per subject, approximately 8% of all
responses). This helps to explain why there are base rate differences in the
production of the V element in VN nominal compounds. The NN response is also
quite common for Broca's aphasics (mean=4.9, 17.5% of all responses) and for
Wernicke's aphasics (mean=4.7, 16.8% of all responses). When I compared the
use of this high-frequency alternative across the three subject groups, I found that
both patient groups produced NN more often than normal controls (Broca’s,
F( 1,18)= 12.37, p < .003; Wernicke's, F(l,18)=4.89, p < .05), yet Broca's and
Wernicke's aphasics did not differ from one another (F(l,19)=0.03, n.s.). We may
conclude that competition from the high-frequency NN alternative occurs for all
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
158
normal and patient groups, obscuring the predicted double dissociation at the
sublexical level.
I also looked at production of a lone V vs. a lone N in response to VN
nominal compounds. For normal controls, there was only one example of a lone N
and no example of a lone V. By contrast, both alternatives appeared among the
aphasics, but at different rates. For Broca's, a lone nominal element N was
produced 13.6% of the time, while a lone verbal element V was produced 1.4% of
the time. For Wernicke's, a single N was produced 5.7% of the time, while a
single V was produced 4.6% of all responses. A 2 x 2 MANOVA was conducted
comparing lone N vs. lone V for Broca's and Wernicke's aphasics only. There was
no main effect of group (F( 1,19)= 1.40, n.s.), but there was a significant effect of
type (F(l, 19)=7.68, p < .013), indicating the greater prevalence of single noun
responses on these VN nominal compounds. Most important for our purposes
here, there is a significant group by type interaction (F( 1,19)=5.39, p < .032).
Hence the double dissociation between Broca's and Wernicke's aphasics is reliable,
in the predicted direction.
Of all the substitution types in Table 11 for VN nominal compounds, the
most interesting alternative may be the production of W compounds in place of a
VN noun. These items are particularly interesting for my purposes here, because
they are “verbs all the way down,” in response to targets that are nouns at the
whole word level, with one nominal element at the sublexical level. There are no
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
159
responses in the VV category for normal controls. The mean for Broca's aphasics
is 0.2 (representing well under 1% of all responses), while the mean for Wernicke's
is 1.3 (constituting 4.6% of all responses). Although these alternatives are not
very frequent overall, the difference between Broca's and Wernicke's aphasics is
significant (F(l, 19)=6.15, p < .023), providing further evidence for a “verb bias” in
Wernicke's aphasics.
Table 11
Word Type Responses to VN-N & VN-V
Target VN-N VN-V
Responses
Nouns
B* W* N* B W N
VN:N 47.5 27.9 91.3 5.8 1.5
NN 17.5 16.8 8.3 18.2 7.0 1.0
N 13.6 5.7 0.4 9.6 3.6
N-zi 3.6 7.9 -- —
Verbs
VN:V 4.3 12.9 - 54.5 58.2 97.6
W 0.7 4.6 1.8 11.5 1.3
V 1.4 4.6 2.1 7.9
V-zi — 1.4 — —
* B=Broca’s aphasics, W= Wernicke’s aphasics, N= Normal controls
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
160
Verbal VN: Word Type Analysis
Verbal VN makes an interesting comparison with nominal VN. In the
discussion of nominal VN compounds, I reported a significant group (Broca’s
aphasics vs. Wernicke’s aphasics) by type (nominal VN word type vs. verbal VN
word type) interaction. This finding suggested that Broca’s and Wernicke’s
differed in their preference for nominal VN vs. verbal VN word types. However,
their different preference was not observed in the word type analysis of verbal VN
compounds. Results of verbal VN compounds showed a significant difference
between Broca’s and Wernicke’s only in their production of nominal VN word
type (Broca’s: 5.8% vs. Wernicke’s: 1.5%, p<004). There was no difference
between the two aphasic groups in the production of verbal VN word type on
verbal VN compounds (Broca’s: 54.5% vs. Wernicke’s: 58.2%).
Why does the production of VN word type appear as significant in nominal
VN compounds but appear as only a trend in verbal VN compounds? This
discrepancy may be explained by the interaction of word type frequencies with
noun-verb effects. Specifically, VN word type is a major word type (i.e., a high-
frequency word type) of verbal compounds but a minor word type (i.e., a low-
frequency word type) of nominal compounds. Results from word type analyses
show that, on the production of verbal VN word types, Broca’s and Wernicke’s
aphasics do not differ significantly because VN is the major word type of verbal
compounds. Broca's aphasics produce target word types 54.5% of the time, very
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
161
closely, Wernicke's aphasic produce target word types 58.2% of the time ( n.s.).
However, these two aphasic groups differ significantly when VN is not the major
word type. For example, on the production of nominal VN compounds, Broca’s
aphasics produce target word types 47.5% of the time, by contrast, Wernicke’s
aphasics produce target word types only 27.9% of the time. Their difference is
significant ( p<006). The major word type effect is also observed in the
production of NN word types for NN, NNN, VNN nominal compounds (see
Section 5.3.2).
In regard to the alternatives for VN verbal items, I found very few
substitutions of any kind for normal controls. They produced NN responses only
1% of the time compared with 8.3% for VN nominal targets. Of course this is not
surprising, since the target in this case is a verb instead of a noun at the whole
word level. The only other substitution observed for normal controls in this
category were W compounds (1.3% of all responses to VN verbal targets). This
confirms our suggestion that VN nominal compounds differ from VN verbs,
because the former experience competition from the high-frequency NN word
type, while the latter do not. By contrast, Broca's aphasics produced a relatively
large number of NN responses to VN verbal targets (mean=6.0, 18.2%); these
responses also occurred in Wernicke's, but they were less frequent (mean=2.3,
7%). The difference between aphasic groups is significant (F(l,19)=6.22, p <
.023).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
162
I also looked again at the production of a single V or a single N. These
responses never occurred at all among normal controls. For Broca's aphasics, a
lone N was produced 9.6% o f the time, while single V elements constituted 2.1%
of all responses to a verbal VN target. For Wernicke's aphasics, the corresponding
rates are 3.6% production of a lone N and 7.9% production of a lone V. Although
these are relatively low frequency response types, their distribution suggests
further evidence for a double dissociation. This impression is verified in a 2 x 2
MANOVA (group by V vs. N response). The analysis yields no main effect of
group (F(1,19)=0.0I, n.s.) or type (F(l,19)=0.73, n.s.), but there is a reliable
group by type interaction (F(l, 19)=9.19, p < .007).
Finally, I looked at the production of W compounds, a somewhat more
appropriate alternative for VN verbal targets. As noted, this did occur for normal
controls (although it is very rare). For Broca's aphasics, the W response occurred
1.8% of the time compared with 11.5% for Wernicke's. This difference was
statistically reliable (F( 1,19)= 10.24, p < .005). I also carried out a 2 x 2
MANOVA, comparing group (Broca’s vs. Wernicke’s) with NN vs. W responses
to VN verb targets. The two main effects were not significant, but there was a
reliable group by type interaction (F(l,19)=l 1.94, p < .003), indicating a double
dissociation, with more “double noun” responses in Broca's aphasics and more
“double verb” responses among Wernicke's aphasics.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
163
To summarize, the word type analysis helps to clarify the earlier
comparison of VN nouns and VN verbs, from several points of view. First, I have
confirmed the suspicion that VN nominal compounds are more difficult to produce
(even for normal controls) because of competition from the high-frequency NN
compound type. This fact makes it difficult to see the same sublexical dissociation
that emerges so clearly for VN verbal compounds. Second, I find clear evidence
for a noun bias among Broca's aphasics, in the production of a lone N and the
production of a double-noun response to either of these compound types. In
contrast, Wernicke's aphasics are more likely to produce a lone V and a double
verb response to VN compounds of any kind. This contrast of “noun bias” and
“verb bias” between Broca’s aphasics and Wernicke’s aphasics remains significant
in the other three nominal compounds. Moreover, I demonstrate a significant
interaction of noun-verb effects with word type frequency effects. This interaction
makes it difficult to see differences between Broca’s aphasics and Wernicke’s
aphasics concerning their production of verbal VN word type for verbal VN
targets, compared to differences between the two aphasic groups regarding their
production of nominal VN word type response to nominal VN targets.
5.3.2 NN, NNN and VNN Nominal Compounds
The three additional compound types at issue here are all nouns at the
whole word level, including VNN compounds. Hence, they should provide further
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
164
insights into production of complex word types by Broca's and Wernicke's
aphasics.
Sublexical analyses are fairly straightforward for VN nouns and VN verbs,
because there are only two possible slots within each target word, and they lie in a
clear, complementary distribution. By contrast, it is difficult to conduct an analysis
of omissions or substitutions at the sublexical level for NN, NNN and VNN
compounds, since it is difficult to assign a relationship between the “intended unit”
and the “observed unit.” For example, if the patient produces a lone V in response
to an NN target, how should that response be classified? Which N element is
omitted, and which is substituted? Given these problems, questions about “noun
intrusions” and “verb intrusions” at the sublexical level are best addressed by
looking at the competing forms produced by normals and aphasic patients in
response to these complex noun targets. Table 12a summarizes the alternative
word types produced by normals and aphasics in response to NN, NNN and VNN
compounds. At the whole word level, we already know that Wernicke's patients
tend to produce a pathological number of verbs in response to these nominal
targets. An examination of Table 12a tells us more about the sublexical structure
of their substitutions.
For normal controls, substitutions are relatively rare overall, as we have
already noted. Normals produced the target form 88.4% of the time on NN items,
87.2% on NNN, and 84.4% on VNN. Among the correct synonyms or incorrect
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
165
substitutions that they did produce, most are nouns at the whole word level (see
Table 12a). When looking at the internal structure of substitutions, normals
produced alternative words containing some kind of V element 3.2% of the time
on NN targets, 1.1% of the time on NNN targets, and 8.7% on VNN targets (not
including the production of target VNN). So, although substitutions are rare in
normals, there does appear to be some effort to include the verbal piece on items
designed to elicit a VNN nominal response.
In fact, if we look at the data in terms of percentage of all substitutions (as
opposed to percentage of all correct and incorrect responses), we find that
alternatives containing a verbal element account for 27.83% of all substitutions by
normals on NN items, 8.94% of substitutions on NNN items, compared with
55.06% of all substitutions on VNN (see Table 12b). This baseline fact will be
relevant to my interpretation of the aphasia data.
For Broca's aphasics, substitutions are more common overall, and (as we
saw earlier) they tend to be nouns at the whole word level. When looking at the
internal structure of word substitutions, we can see that substitutions containing a
verbal element account for 4.7% of all responses by Broca's aphasics on NN
targets, 5% of all responses on NNN targets, but 20.5% of all responses on VNN
targets (see Table 12a). If we look at this pattern expressed as a percentage of all
substitutions (as opposed to a percentage of all responses, correct or incorrect),
we find that Broca's produce substitutions with an internal verb 21.36% of the time
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
166
Table 12a
Percent of Word Type Responses to NN, NNN, VNN
Nominal Targets (as a percentage of total responses)
Targets NN NNN VNN
Responses B* W* N* B W N B W N
Nouns
NNN 3.1 3.1 6.3 37.0 18.0 87.2 5.5 3.6
NN 67.9 53.4 88.4 32.5 38.5 9.5 23.2 23.2 6.1
N 14.2 15.5 2.0. 10.5 10.0 1.7 6.4 4.6 1.0
VNN:N 0.8 0.5 — 0.5 3.5 — 31.4 16.8 84.4
VN:N 1.0 2.6 0.5 1.0 3.5 1.1 4.6 5.5 5.1
W N :N 0.2 — — -- 0.5 — — 0.5 3.1
Verbs
V W 0.3
w 0.2 1.1 — 0.5 7.0 — 0.5 3.2 —
V 0.5 2.1 -- 1.0 3.0 — 2.7 4.6 —
VNN:V 1.0 1.0 2.7 0.5 6.0 — 4.1 5.9 —
VN:V 1.0 6.9 — 1.5 6.0 — 8.6 9.6 0.5
W N :V — 0.3 — — — — — — —
* B=Broca’s aphasics, W= Wernicke’s aphasics, N= Normal controls
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
167
on NN items, 10.42% on NNN items, vs. 25.32% on VNN targets (see Table
12b). This is quite similar to the pattern displayed by normals. In fact, three
separate two way analyses comparing Broca's with normals on each target type
(using percent of all substitutions with a verbal element as the dependent variable)
yield no significant differences.
As we have already seen, Wernicke's aphasics produce more word
substitutions than any other group, and they are also more likely to produce
Table 12b
Word Type Substitutions with Verb Pieces on NN, NNN,
and VNN Nominal Targets (as a percentage of all substitutions)
Target Word Type Aphasic Type
Broca’s Wernicke’s Normal
NN 21.36 44.31 27.83
NNN 10.42 37.82 8.94
VNN 25.32 48.27 55.06
intrusions that are verbs at the whole word level. When looking at the internal
structure of their word substitutions, we can see that Wernicke's aphasics produce
substitutions containing some kind of verbal element 14.8% of the time on NN
targets, 29.5% on NNN targets, and 29.3% on VNN targets (see Table 12a).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
168
Expressed as a percentage of all substitutions within each category (see Table
12b), the corresponding figures are 44.31% on NN, 37.82% on NNN, and 48.27%
on VNN. Overall, these ratios appear to be higher for Wernicke's aphasics than
they are for normals or for Broca's aphasics, except for the VNN category. In this
category, the ratio is higher for normals than it is for Wernicke’s aphasics. I
carried out separate two-way comparisons looking at the percentage of all
substitutions containing a verbal element; first comparing Wernicke's with Broca's
and then comparing Wernicke's with normals. In the Wernicke’s/Broca’s
comparisons, there are significant differences in the ratio of substitutions with an
internal verb on NN targets (F(l,19)=8.14, p < .02), and on NNN targets
(F(l,19)=l 1.31, p < .003), but the comparison on VNN targets fails to reach
significance (F( 1,19)= 1.54, n.s.). In the Wernicke’s/normal comparisons, the
difference just misses significance on NN (F(l,18)=3.98, p < .07). The
comparisons are reliable on NNN items (F(l, I8)=7.53, p < .02) but do not reach
significance on VNN targets (F(l, 18)=2.59, n.s.).
In summary, results from word type analyses on NN, NNN and VNN
compounds demonstrate that Wernicke's aphasics produce a high ratio of
substitutions that contain a verbal element of some kind, a tendency that is most
striking on NN and NNN targets that are “nouns all the way down.” Broca's
aphasics produce relatively few verb intrusions, although it seems that they do
struggle to produce some kind of verb piece in response to VNN nominal
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
169
compounds -- as though, at some level, they are trying to match the desired
pattern. Normals produce very few substitutions of any kind, but when they do,
their responses are closer to those of Broca's aphasics.
5.3.3 Noun-verb Dissociations and Word Type Frequency
Results from word type analyses demonstrated that Broca's aphasics had a
stronger tendency to produce nominal word types than verbal ones, whereas
Wernicke's aphasics tended to produce verbal word types rather than nominal
ones. The “noun bias” of Broca’s and the “verb bias” of Wernicke’s were
observed across all compound types. However, there was a contrast between
nominal VN and verbal VN regarding the accessibility of VN word type.
Specifically, in their responses to nominal VNs, Broca’s and Wernicke’s aphasics
demonstrated the difference of accessing the VN word type, however, they did not
show such difference of accessing the VN word type in their responses to verbal
VN compounds. Why do verbal VNs and nominal VNs show such contrast in the
production of VN word type? One major factor that distinguishes VN nouns and
VN verbs is their relative standing in their grammatical categories. VN is the
major word structure for verbs but not for nouns. Huang (1992) investigated
percentages of possible word types within nouns vs. verbs. He reported that the
NN word type constituted 54% of all nouns while the VN word type was only
12%. As a major word type of verbs, VN word type constituted 35% of all
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
170
verbs.1 6 Thus, there is a difference between verbs and nouns as to their major
word types. The major word type here is parallel to high-frequency word type in
the language. In this sense, the VN pattern is a high-frequency word type for
verbs. Its frequency effect can affect only verbal VN compounds. The NN pattern
is a high-frequency word type for nouns only. Its impact is limited to nouns only.
Non-major word types, such as VN to nouns, will not have a significant influence
on the production of word type. This frequency effect has several predictions.
First, it predicts that, on the production of VN word types for verbal VN
compounds, differences between Broca’s aphasics and Wernicke’s aphasics will
decrease because of the high-frequency of VN word type, particularly, if frequency
effects dominate over noun-verb effects.
Second, it predicts that there will be no difference between the two aphasic
groups in the production of NN word types on NN compounds because NN, as
major word type of nominal compounds, is the most frequent word type among
nominal compounds. The frequency effect will help both Broca’s and Wernicke’s
aphasics to produce the target word type (i.e., NN).
Finally, it predicts that noun-verb dissociations should appear in subjects’
responses to non-major word types, including VN nominals, all nouns with verb
1 6 Huang (1992) reported that W and VN were two major word types for verbs. The W word
type constituted 44% of all verbs and the VN word type constituted 35% of all verbs. In my
dissertation, pictures for action names were designed to elicit verbal VNs. Hence, the W effect
was not so significant as the VN effect.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
171
pieces, and all verbs with only noun pieces. A summary of results in 28a-f from
Table 11 and Table 12a support the three predictions.
28. Summary of Results
a. Verbal VN as the Major Word Type of Verbal VN Targets
Broca's aphasics and Wernicke's aphasics did not differ significantly in their correct
production of verbal VN word type (Broca's 54.5% vs. Wernicke's 58.2%).
b. Nominal NN as the Major Word Type of Nominal VN Targets
Broca's aphasics and Wernicke's aphasics did not differ significantly in their
responses of nominal NN word type to nominal VN targets (Broca's: 17.5 % vs.
Wernicke’s: 16.8 %).
c. NN Nouns vs. W Verbs as Non-major Word Types of Verbal VN Targets
For NN nouns and W verbs as a pair of non-major word type of verbs, NN was
produced 18.2% of the time by Broca's aphasics and 7.0% of the time by
Wernicke's aphasics (p< .023). W was produced only 1.8% of the time by
Broca’s aphasics but 11.5% of the time by Wernicke's aphasics (p< 005). This
contrast demonstrated a noun-verb dissociation at the word structure level.
d. Lone N and Lone V as Non-major Word Types of Verbal VN Targets
For lone N and lone V as a contrast pair of non-major word type of verbs, Broca’s
aphasics produced a lone N 9.6% of the time, while single V elements constituted
2.1% of all responses to a verbal VN target. For Wernicke's aphasics, the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
172
corresponding rates were 3.6% production of a lone N and 7.9% production of a
lone V. A 2 x 2 MANOVA (group by V vs. N response) showed a reliable group
by type interaction (F(l, 19)=9.19, p < .007).
e. Nominal VN and Verbal W as Non-major Word Types of Nominal VN
Targets
VN nouns and W verbs were a contrast pair of non-major word types of nouns,
Broca's aphasics produced VN nouns 47.5% of the time and Wernicke's aphasics
produced them only 27.9% of the time (p<006). By contrast, for W verbs,
Broca's aphasics produced W verbs only 0.7% of the time and Wernicke's
aphasics produced W verbs up to 4.6% of the time (p<023). Broca's aphasics
and Wernicke's aphasics displayed a significant dissociation in the production of
non-major word types.
f. Nominal N and Verbal V as Non-major Word Types of Nominal VN
Targets
For lone N and lone V as a contrast pair of non-major word type of nouns, Broca's
produced a lone nominal element N 13.6% of the time but a lone verbal element V
only 1.4% of the time. By contrast, Wernicke's produced a single N was produced
5.7% of the time but a single V 4.6% of the time. A 2 x 2 MANOVA was
conducted comparing lone N vs. lone V for Broca's and Wernicke's aphasics only.
The result showed a significant group by type interaction (F( 1,19)=5.39, p < .032).
Hence, the double dissociation between Broca's and Wernicke's was reliable.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
173
5.4. Summary
Results from verbal compounds demonstrate a straightforward pattern of
double dissociation between non-fluent Broca's aphasics and fluent Wernicke's
aphasics in the production of nouns and verbs. By contrast, results from all the
nominal compounds do not show such a straightforward double dissociation, at
least not at the sublexical level. Instead, I find a strong bias toward the major
nominal word type NN, which dominates over the grammatical category effect.
Although the word type effect seems to prevail over the noun-verb dissociation,
there are some interesting component errors ( section 5.2) arguing for a sublexical
noun-verb double dissociation between Broca's aphasics and Wernicke's aphasics.
Additionally, the confound of major word type effects with the noun-verb effect
leaves only non-major word types as unproblematic cases for a further analysis.
The results of Table 1 1 and Table 12a all demonstrated noun-verb dissociations in
the production of non-major word types, as they were predicted.
Based on what has been found in the present study, I claim that although
the results vary somewhat depending on the compound type in question, they
provide further evidence for a double dissociation between non-fluent Broca's
aphasics and fluent Wernicke's aphasics in the production of nouns and verbs.
Furthermore, this dissociation also appears in the production of word types. All
these findings motivate the proposal that word structures, whole words, and
components are all sensitive to the noun-verb distinction. The presence of a noun
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
174
verb dissociation at three different levels cannot simply be explained in terms of the
assumption that nouns and verbs are stored in two different modules. Instead, it is
necessary to develop a connectionist approach to capture the fact that a
grammatical category contrast can simultaneously affect whole words, sublexical
items, and word structures.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
175
Chapter Six
Frequency Effects: Results and Discussion
The research has increasingly shown the importance of frequency in the
processing of inflections words and larger grammatical units. For example, at the
level of inflections, studies of child language acquisition demonstrate that case
markings are acquired by children in an early age because these markings are
frequently used in the language. Likewise, verb inflections are acquired by Italian
speaking children in a very early age due to its high frequency in the language
(MacWhinney, Pleh, & Bates, 1985). The frequency effect of inflections has also
been discussed in studies of brain-damaged patients and normal subjects in these
two languages (Bates, Friederici, & Wulfeck, 1987b; MacWhinney, Bates, Kliegl,
1984; MacWhinney & Osman-Sagi, 1991; MacWhinney, Osman-Sagi, Slobin,
1991). The frequency effect is also observed at the word level in the word
recognition task (Bresner & McCann, 1987; Seidenberg, 1995 for a review) and
the naming task (Bolota & Ferraro, 1993; Monsell, 1991). Recently, a radical step
has been taken to extend the frequency effect to larger grammatical units such as
phrasal structures. For example, MacDonald, Pearlmutter, & Seidenberg (1994)
reported that, instead of the frequency of a word in a language, the frequency that
the word occurred with a particular phrase affected the interpretation of the word.
Trueswell (1996) further showed that verb argument preferences affected the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
176
resolution of ambiguous sentences. Bates et al. (1995) also demonstrated that,
under highly constrained conditions in a film description task, Alzheimer's Disease
patients, who did not produce grammatical errors, found it difficult to produce
passive sentences for the films that normal subject would describe by using passive
sentences. They suggested that the difficulty was partly due to the low-frequency
of passive structures, compared with frequencies of active sentences (Bates,
Harris, Marchman, Wulfeck, & Kritchevsky, 1995).
If we are correct that Chinese compounds are processed at three levels,
then we should see frequency effects at each level. The presence of frequency
effects at these three levels will serve as additional support for the claim that
Chinese compounds contain three levels of information: the whole word level, the
component level, and the word structure level.
6.1. Frequency Sensitivity: Whole Words and Components
This section discusses results of frequency effects at the whole word level
and the component level. For this analysis, I used the same data that were used to
discuss noun-verb dissociations as well as additional data from subjects’ responses
to pictures of 50 single-morpheme Ns. Results of single-morpheme Ns will be
used to discuss the issue of “structural complexity.” There are two major analyses:
correct responses and substitutions. The “correct response” compares the ratio of
correct responses between high-frequency and low-frequency targets (see criteria
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
177
of high/low-frequency words in 4.4.2). The “substitution analysis” investigates
substitutions of higher frequency items for lower frequency ones and substitutions
of lower frequency items for higher frequency ones. The “percent correct” and the
“substitution” analyses are applied to whole words as well as sublexical elements,
but not to word structures. This is partly because there are only five target word
structures. The number is not enough for two frequency groups (high vs. low).
6.1.1 Correct Responses: The Whole Word Level
Fig. 7a showed that, out of 46 elicited high-frequency compounds vs. 80
low-frequency ones, normal controls showed only a slight difference: they
produced high-frequency targets more often than low-frequency targets (80.19%
to 77.5%). The preference for high-frequency words over low-frequency one
remained only as a tendency (normal: F (1, 16)= 1.37, n.s.). Both of the aphasic
populations studied demonstrated much more dramatic frequency effects.
Specifically, Broca's aphasics produced high-frequency targets 45.87 % of the time
and low-frequency targets 25.63% of the time; Wernicke’s aphasics produced
high-frequency targets 27.39% of the time and low-frequency targets 10.5% of the
time (see Fig. 7a). A simple one-way analysis of variance for each subject group
showed a significant distinction between high-frequency and low-frequency words
(Broca’ s: F(l, 18)=11.73, p< .005; Wernicke’ s: (1, 18)=19.35, p< .0005). Both
groups were better able to access high-frequency words than low-frequency ones.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
178
Fig. 7a: Percent correct production
of whole words by frequency
9 0 r
* 801 8 Q -i l 9 77.5
S 70[
“ S 60
g 50 45.87
o 40
o
£ 30
§ 20f
a. 10f
0i
i 1 25.63
■
1 1. ■
1
1 25.63 27.39
10.5
Broca's Wernicke's Normal
Patient Group
I i high frequency H low frequency
Why does frequency sensitivity at the whole word level appear in aphasics
but not in normal controls? The main reason for this lack of frequency effect is
that normals have such high accuracy that it is hard to find a frequency effect on
accuracy (i.e., a ceiling effect). However, the task is difficult enough to show
effects for the aphasic groups whose lexical access has been impaired. These
frequency effects at the lexical level argue that whole compound words are
independent units which are subject to item access principles.
6.1.2 Correct Responses: The Sublexical Level
Results from the analysis presented in Fig. 7b demonstrated that, at the
sublexical level, out of 138 high-frequency morphemes, normal controls produced
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
179
correct items 88.4% of the time. By contrast, out of 149 low-frequency items,
normal controls produced correct items 89.19% of the time. The simple one way
analysis of variance demonstrated that the difference was not significant (F (1,
I6)=0.34, n.s.).
Fig. 7b: Percent correct production of
sublexical elements by frequency
Broca's Wernicke’ s Normal
Patient Group
I I high frequency H I low frequency
In Broca's aphasics, their preference for high-frequency targets over low-
frequency ones also was less at the sublexical level. There were 50.8% correct
responses for high-frequency sublexical items in contrast to 48.93% correct
responses for low-frequency sublexical items. This preference reflected only a
slight tendency in a simple one way analysis of variance (F (I, 18)=0.15), n.s.).
Wernicke's aphasics showed a stronger tendency to produce high-frequency targets
more frequently than low-frequency ones. The tendency just missed significance
(F (1, 18)=4.2, p=0.055).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
180
At first glance, this finding does not appear to support the independence of
the sublexical level. However, after a more careful examination of the raw data, I
observe a large number of substitution errors which may explain why a “percent
correct” analysis fails to show frequency effects on the access of sublexical items.
The same reason may also apply to the whole word level. These substitution
errors include a significant number of lexical items where frequencies are even
higher than their targets. For example, a Broca’s aphasic patient produced “xian-
sheng” (lexical frequency: 570; sublexical frequencies: 1470-5253) for the target
“yi-sheng” (lexical frequency: 62; sublexical frequencies: 535-5253). The target
word “yi-sheng” (literally cure-person, meaning “doctor”) is a high-frequency
lexical item which, in turn, contains two high-frequency sublexical morphemes (yi
and sheng). At the whole word level, the frequency of the substitute is higher than
the target. At the sublexical level, the frequency of the substitute is also higher
than the target, while the final morpheme is correctly produced. Because of these
errors, the analysis of “correct responses” may not accurately reflect the nature of
any frequency effect. In a case like this, a comparison of response frequency with
target frequency may capture the frequency effect more accurately. To explore
whether subjects were sensitive to frequency, I further therefore conducted
additional two analyses: one for lexical substitutions and one for sublexical
substitutions. The analyses were based on the total number test compounds.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
181
There are 165 total lexical items, which are decomposed into 372 total sublexical
items for each subject (see chapter four).
6.1.3 Substitutions: The Lexical and Sublexical Levels
Fig. 8a presented the substitution frequencies for high-to-low and low-to-
high. Normal controls made very few substitutions at either level. Again, their
performance reflected a ceiling effect. However, for both groups of aphasic
subjects, the preference for high-for-low substitutions became significant at both
levels. Specifically, at the lexical level, Wernicke's aphasics produced 26.91%
high-for-low substitutions but only 10.73% low-for-high substitutions (F (1,
18)=62.26, p=0). Almost exactly parallel, Broca's aphasics produced 26.06%
high-for-low substitutions but only 10.75% low-for-high substitutions (F (1,
18)=20.90, P< .0005). Both groups of aphasics demonstrated a significant
difference between high-for-low substitutions and low-for-high substitutions. The
finding that both aphasic groups showed a significant preference for high-for-low
substitutions over low-for-high substitutions argued for the frequency effect at
lexical level.1 7
1 7 Low-for-high whole word substitutions may include a significant number of invented words
which frequency count is zero, for instance, the frequency of the invented word “qi-pao” (ride-
run) for the target “qi-ma” (ride-horse) is zero. Although “qi-pao” is an invented word, it is
interpretable by combining meanings of “qi” (ride) and “pao” (run).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
182
Fig. 8a: Percent substitution of
whole word by frequency
30
25f -
§
1 20 -
2
3
(O
26.06
26.91
< D
a
®
Q .
15
10|
si
0
I
1 |lOJ3
, 4-98_ 4.24
, . 1,1 ■ ,
Broca's Normal Wernicke's
Patient Group
I i high-for-low H low-for-high
At the sublexical level, Broca's aphasics produced high-for-low
substitutions 16.59% of the time and low-for-high 10.91% of the time. Similarly,
Wernicke’s aphasics produced high-for-low substitutions 28.82% of the time and
low-for-high 12.98% of the time (see Fig. 8b). The difference between the two
kinds of substitutions reaches significance in both populations (Broca's: F (1,
18)=8.30, p< .01; Wernicke's: F(l, 18)=109.41, p= 0). These results for Broca’s
and Wernicke’s aphasics provide a clear picture than the results of “percent
correct” analysis for sublexical elements. The significant preference for high-for-
low substitutions at the sublexical level demonstrated by Broca’s and Wernicke’s
aphasics provide the evidence for the frequency effect at the sublexical level.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
183
c
£
■5
2
3
C O
0
£ »
0
Q .
Fig. 8b: Percent substitution of
sublexical element by frequency
28.82
12.98
Broca's Wernicke's Normal
Patient Group
high-for-low Hi low-for-high
6.2 Frequency sensitivity: Word Structures
In section 5.3, I reported that noun-verb dissociations were significant on
major word types (such as NN nouns for object names or VN verbs for action
names) but insignificant on non-major word types (such as VN nouns for object
names and lone Vs for action names). Because major word types are also high-
frequency word types in the language, the major word type effect indirectly
suggests that the access of word structure is also frequency-sensitive.
Word types discussed in noun-verb dissociations are structures carrying
grammatical categories.1 8 An alternative approach can be used to depict word
1 8 There are two ways to depict word structures. First, word structures may be depicted in terms
of grammatical categories to yield word structures such as VN (Huang, 1986; 1992; Huang &
Hsieh, 1989). Alternatively, word structures may be characterized in terms of grammatical roles
to result in word structures such as VO. These two ways are related to each other closely because
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
184
structures. Using this approach, word structures are distinguished in terms of the
number of components within a compound. For example, the VN (or VO)
compound is a two-morpheme word structure. Word structures, which differ in
the number of morphemes, appear in the language in different frequencies. For
example, according to A Frequency Dictionary o f Modern Chinese (1985), the
two-morpheme word structure is the most frequent word structure in the language
(72.6%), followed by the single-morpheme word structure (15.9%), and followed
by the three-morpheme word structure (6%).1 9 In this approach, word structures
do not involve grammatical categories such as noun and verb. Results of the
analysis should not have interacting influences from the noun-verb effect. This
analysis allows us to track a simple frequency effect on the access of word
structure.
To examine possible frequency effects on word structure access, the
following section examined six different word types (single N, NN, nominal VN,
verbal VN and VNN, NNN). Responses of each subject were analyzed in terms of
the number of morphemes in a word. Three factors in the analysis are: subject
group, word type of targets, and word type of responses.
a component’s grammatical function can tell its grammatical category. For example, if a
component functions as an object (an agent, an instrument, or a location), it should be nominal.
1 9 The total is based on type frequencies of five word structures (single-morpheme, two-
morpheme, three-morpheme, four-morpheme, five-morpheme). For a detailed discussion, see
chapter four.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
185
6.2.1 Word Type Analysis: Responses on Single-Morpheme
Table 13a showed that, in response to 50 single-morpheme target pictures,
all three subject groups produced a very similar number of single-morpheme words
and two-morpheme words. Of their total responses, Broca's aphasics produced
single-morpheme words 47.6% of the time and two-morpheme words, 45.2% of
the time. Wernicke's aphasics produced single-morpheme words 45% of the time
and two-morpheme words 45.6% of the time. Normal controls produced single
morpheme words 48.89% of the time and two-morpheme words 48% of the time.
Table 13a
Percent Response on Single-Morpheme Word Type
Response Word Type Subject Types
Broca’s Wernicke’s Normal
Single-morpheme 47.6 45 48.89*
Two-morphemes 45.2 45.6 48
Three-morpheme 3.2 4.8 2.89
* The percentage of target word type by normal controls
After a detailed look at the raw data, I found that the relatively high ratio of two-
morpheme words could be attributed to two sources. The first source has
something to do with the picture— the pictures can elicit both single-morpheme and
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
186
two-morpheme words (for example, the picture “watch” can elicit the single-
morpheme word “biao” (meaning “ watch”) and the two-morpheme synonym
“shou-biao” (literally hand-watch, also meaning “watch”). When the two
candidates were competing, subjects were apt to select the two-morpheme word.
The second source has something to do with errors. When subjects were not able
to find the right words, they tended to find a two-morpheme substitute for the
target. In both situations, subjects all demonstrated a preference for the two-
morpheme word structure, which was the highest frequency word type in the
language. This serves as evidence for the frequency effect on word structures in
the word production.
When looking at the production of three-morpheme words for the N target,
the ratio of three morpheme words decreased dramatically; 3.2% for Broca’s
aphasics, 4.8% for Wernicke’s aphasics, and 2.89% for normal controls. These
results indicated a very little impact of three-morpheme word structure on the
word production, compared with the two-morpheme word structure.
To summarize, normal controls and aphasics do not differ in their
production of two-morpheme vs. single morpheme word structures. This suggests
that the two-morpheme word type is so influential that even when a simple single
morpheme is appropriate subjects are more likely to produce the two-morpheme
type.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
187
6.2.2. Word Type Analysis: Responses on Two-morpheme
Table 13b demonstrates that the preference for two-morpheme word
structures remains very strong in the task designed for two-morpheme targets
across the three subject groups. For two-morpheme targets (including NN, VN-N,
and VN-V), the production of two-morpheme structures is: 71.11% in Broca's
aphasia; 73.27% in Wernicke's aphasia; and 91.07% in normal controls. Of the
three subject groups, normal subjects produced the highest number of two-
morpheme word types across the three compound types. The production of two-
morpheme word types for two-morpheme targets remained high in Wernicke's
aphasia and Broca's aphasia (see Table 13 b), compared to their production of
single-morpheme for single-morpheme targets (see Table 13a). The fact that
normal controls make more two-morpheme word types than aphasics may be
attributed to the intact knowledge of the target word structures among normal
controls. Although the lexical knowledge of Broca’s and Wernicke’s aphasics is
impaired, the frequency effect may help them to produce two-morpheme word
structures. This frequency effect has also been observed in the overuse of two-
morpheme in response to single-morpheme targets (see Table 13a). The frequency
effect also holds in aphasic subjects’ responses to three-morpheme word types (see
Table 13c).
Because the three subject groups have a very strong tendency to produce
two-morpheme word types, their production of target word types (i.e., two-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
188
Table 13b
Percent Response on Two-Morpheme Word Type
Response Word Type Subject Response
Broca’s Wernicke’s Normal
Single-morpheme 13.71 16.63 1.39
Two-morphemes 71.11 73.27 91.07*
Three-morpheme 6.52 6.39 7.01
* The percentage of target word type by normal controls
morpheme) is the best performance among the production of ail the other word
types. This strong tendency to produce two-morpheme word structures reflects a
preference for high-frequency word types in the language.
6.2.3 Word Type Analysis: Responses on Three-morpheme
As indicated in Table 13 c, of the three-morpheme targets (NNN and
VNN), Broca's aphasics demonstrated almost equal tendency to produce two-
morpheme and three-morpheme word structures. They produced 38.07% two-
morpheme words for three-morpheme targets and 38% three-morpheme words for
three-morpheme targets. By contrast, Wernicke's aphasics showed a little higher
tendency to produce two-morpheme words compared with three-morpheme
words, with 56.39% two-morpheme word structures for three-morpheme targets
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
189
and 28.61% three-morpheme word structure for three-morpheme targets. Thus,
the preference for the two-morpheme word structure is stronger in Wernicke's
aphasia than in Broca's aphasia. This is similar to the pattern of responses reported
in two-morpheme targets. Unlike aphasic subjects, normal controls maintained
target three-morpheme word structures up to 84.05% of the time.
Table 13c
Percent Response on Three-Morpheme Word Type
Target Word Type Subject Type
Broca's Wernicke’s Normal
Single-morpheme 10.07 8.13 1.34
Two-morphemes 38.07 56.39 11.87
Three-morpheme 38 28.61 84.05*
* The percentage of target word type by normal controls
In summary, when subjects were asked to name objects designed to elicit
three-morpheme compounds, aphasics and normal controls chose different
preference orders. Normal controls were able to access the target word type,
suggesting that their knowledge of target word structures remained intact. For
aphasics whose lexical knowledge was impaired, they were greatly affected by the
frequency of word structures. As a result, they tended to produce high-frequency
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
190
word types, namely two-morpheme word types, and followed by single-morpheme
word types. This preference for high-frequency word structures provides another
line of evidence for frequency sensitivity at the word structure.
6.2.4 Word Type Analysis: Interactions Among Three Word Types
Thus, across three subject groups, the two-morpheme word type is a very
strong competitor among all word structures. In addition to this, it is interesting to
note that single-morpheme word types and three-morpheme word types differ in
their competition strength. Subjects’ responses to two-morpheme targets provide
us some ideas regarding the different influences that the single-morpheme word
type and the three-morpheme word type display.
As shown in Table 13b, Broca's aphasics and Wernicke’s aphasics are alike
because, in responses to two-morpheme targets, they all prefer single-morpheme
word types over three-morpheme word types. Unlike aphasics, normal controls
prefer three-morpheme word types over single-morpheme word types in their
response to two-morpheme targets. Thus, both Broca’s and Wernicke’s aphasics
show a preference for the single-morpheme word structure over the three-
morpheme word structure. Their performance is compatible with the frequency
dictionary; the frequency of single-morpheme word type is higher than that of
three-morpheme word type. Unlike aphasics, normal controls prefer three-
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
191
morpheme words over single-morpheme words in their response to two-morpheme
targets.
Why is the discrepancy between aphasics and normal controls? In addition
to word type effects, we need to consider the pathological and the linguistic
factors. Regarding the linguistic factor, the semantic-specificity is a crucial feature
to distinguish lexical differences between aphasics and normal controls. As
illustrated in Table 14a, single-morpheme words produced by aphasics (both
Broca's and Wernicke’s) are apparently reduced forms of targets, created by
deleting parts of compounds. For example, “er” (meaning “goose”) is a reduced
form of “qi-er” (literally standing-goose; meaning “penguin”). Very often, the
reduced form yields a generic form of the target. For example, the target “lan-
hua” (literally orchid- flower, meaning “orchid”) is reduced into a generic “hua”
(meaning “flower”).
In Table 14b, the three-morpheme words produced by normal controls are
often specific forms of targets. For example, the three-morpheme word “huo-chai-
bang” (literally fire-wood-stick, meaning “match stick”) refers to the stick of
“match” and the three-morpheme word “wan-ju-che” (literally play-instrument-car,
meaning “toy car”) refers only to “toys” which are cars. These two three-
morpheme words are more specific than the two-morpheme word “huo-chai”
(literally fire-stick, meaning “match”) and the two-morpheme word “wan-ju”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
192
Table 14a
Lexical Variants by Broca’s & Wernicke’s Aphasics
Broca’s Wernicke’s
Responses Targets Responses Targets
a. ge (N)
song
chang-ge
sing-song
“to sing”
a. diao (V)
hook
“to hook”
diao-yu
hook-fish
“to go fishing’
b. ma (N)
horse
qi-ma
ride-horse
“to ride horse”
b. kai (V)
open
“to open”
kai-deng
open-lamp
“to turn on”
c. bang (N)
shoulder
jian-bang
shoulder-shoulder
“shoulder”
c. hua (N)
flower
“flower”
lan-hua
orchid-flower
“orchid”
d. gang (N)
steel
gang-kui
steel-helmet
“helmet”
d. hua/fa (N)
hair
“hair”
tou-fa
head-hair
“hair of head”
e. ji (N)
machine
fei-ji
fly-machine
“airplane”
e. pen (N)
tub
“tub”
yu-gang
bath-tub
“bathtub”
f. er (N)
goose
qi-er
stand-goose
“penguin”
f. mo (V)
wipe
“to wipe”
mo-bu
wipe-cloth
“wipe towel”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
193
(literally play-instrument, meaning “toy”) respectively. Thus,
between aphasics and normal controls may initially be attribut
differences.
Table 14b
Lexical Variants by Normal Controls
Responses Targets
a. kai-dian-deng (VNN) kai-deng
open-electric-lamp open-Iamp
“to turn on” “to turn on”
b. jian-tou-fa (VNN) tang-fa
cut-head-hair hot-hair
“to cut hair” “to perm”
c. wan-ju-che (VNN) wan-ju
play-instrument-car play-instrument
“toy car” “toy”
d. xi-shui-guan (VNN) xi-guan
suck-water-hose suck-hose
“water straw” “straw”
e. huo-chai-bang(NNN) huo-chai
fire-wood-stick fire-wood
“match stick” “match”
f. di-qiu-yi (NNN) di-qiu (NN)
earth-ball-instrument earth-ball
“terrestrial globe” “earth”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
194
Pathologically, because the language ability of aphasic patients has been
impaired, these patients very likely show a tendency toward easy tasks but away
from difficult tasks. This tendency is particularly true in the word choice of
Chinese aphasic patients. In Chinese, the more specific a word meaning is, the
more complex its word structure will be. When aphasic patients are asked to
produce words, these patients often choose single-morpheme words, which
meanings are sufficient for the communication purpose; however, these words are
structurally less complex and semantically less specific than three-morpheme
words. Since normal controls do not suffer from any linguistic impairment, it is
quite nature for these normal subjects to provide semantically more specific lexical
items such as three-morpheme words simply for a better communication.
6.3 Summary
Thus far, results from the present study have demonstrated that frequency
effects appear at the whole word level, the component level, and the word
structure level. These frequency effects have different implications, especially for
the syntactic account and the lexical account.
First, the “substitution analysis” reported a significant frequency sensitivity
to both whole words and their components. This finding suggests separate levels
for representing lexical vs. sublexical items. Because the lexical account assumes
only a single level for lexical representations, the account has to be revised to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
195
capture the dual levels of lexical representations. The same problem of the lexical
account may be extended to frequency effects on word structure production
because the lexical account does not consider word structure as a separate part in
the lexicon either.
Second, results from the word type analysis are two-fold: (i) subject groups
have an overall preference for the two-morpheme word structure across word
types; and (ii) aphasic groups prefer the single-morpheme word structure over the
three-morpheme word structure. These two frequency effects on word structure
production challenges the syntactic account. Specifically, the difficulty with
grammatical structures, in particular the symptom of structure reduction found in
Broca's aphasia, is often attributed to a structured deficit. The present study shows
two types of structural “deficits”: one is the preference for the two-morpheme
word structure over the single-morpheme and three-morpheme word types; the
other is the preference for the single-morpheme word type over the three-
morpheme word type. The structure-reduction explanation may capture the
preference for the single-morpheme word type over the three-morpheme one.
However, it fails to capture the preference for the two-morpheme word structure
over the single-morpheme one because two-morpheme words are structurally more
complex than single-morpheme words. Thus, there is a conflict within the
structural deficit account.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1%
In conclusion, a significant frequency-sensitivity at the word structure level,
together with frequency sensitivities at the whole word and the component levels,
argues that whole words, components, and word structures are all frequency-
sensitive. All of them, therefore, should serve as independent processing units.
This proposal establishes a parallelism among word structures, sublexical items,
and whole words, which has a far reaching implication for the theory of lexical
representations and processing.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
197
Chapter Seven
General Discussion and Concluding Remarks
Results from the present study demonstrate that noun-verb dissociations
are significant at both the lexical level and the intra-lexical levels (i.e., the
component level and word structure level). Additionally, they show that
frequency effects are also evident at these three levels. These findings speak to
two controversial issues. The first issue addresses the current debate on the nature
of compound representations in Chinese. Are these compounds unanalyzable
whole words or are they analyzable words made up of components? My findings
support the position that compounds are analyzable. This conclusion has
important consequences for the second issue— whether a single-factor modular
model or a multi-factor connectionist model provides a better account for the
lexical impairments observed in Chinese aphasia.
7.1 Chinese Compounds: Unanalyzable or Analyzable
In chapter three, I discuss three different views concerning the status of
compounds. One view suggests that Chinese speakers do not actually look into the
internal structure of a compound. For these speakers, all Chinese compounds are
unanalyzable polysyllabic words (Li & Thompson, 1981). The second view
assumes that compounds contain internal structures. However, once a compound
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
198
is integrated, speakers will not go into the internal structure of the compound. The
integrity of a compound is shown in the structural inseparability of the compound
parts (Huang, T-C., 1983). The third view also assumes that compounds contain
internal structures. Apart from this point, however, the third view is very different
from the second view. The third view holds that speakers may simultaneously
access both the literal reading and the whole word reading for a compound.
What do my findings speak to these three views? First, the present study
shows that parts of compounds may be omitted or substituted independently. This
argues against the first view that compounds are represented as unanalyzable
words. It also argues against the second view that once a compound is integrated,
no internal structure is available to speakers. The finding, instead, argues for the
third view that the internal structure of a compound may be activated if the context
allows.
Further support for the third view is provided by the results of the
frequency study discussed in chapter six because frequency effects of whole words,
of parts and of word structures are observed in lexical access.
Finally, a significant noun-verb dissociation observed between 10 Broca’s
aphasics and 10 Wernicke’s aphasics again supports the view that compounds are
decomposable. In their differences, Broca’s aphasics tended to produce more N,
NN, and NNN patterns than Wernicke’s in their responses to N, NN, VN nouns,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
199
VN verbs, nominal VNN, and nominal NNN targets. In the same task, Wernicke’s
tended to produce more V, W , V W patterns than Broca’s aphasics.
7.2 Lexical Representations of Chinese compounds: A Single-factor Lexical
Account or a Multi-factor Connectionist Account
The performance of Chinese aphasics has important implications for
current lexical models. Recall my discussion in chapter two: current lexical models
may be divided into two kinds. One kind attempts to find a single/primary factor
for a symptom, the other kind attempts to search all possible relevant factors and
measure the weight of their influence. How do results from the present study
speak to these two alternative approaches?
First, results from the present study show that the access of a lexical item in
Chinese involves more than one factor. For example, very often the noun-verb
effect may interact with the frequency effects or the structural complexity effect.2 0
Their interactions are summarized in (i) to (iii). These findings provide us with
more arguments to disfavor single-factor accounts.
2 0 Additionally, in the languages such as English, noun-verb differences may be attributed to the
interaction of frequency with phonological factors. Sereno & Jongman (1990) report that, among
high frequency words, nouns are more likely to have back vowels than front vowels, and verbs
are more likely to have front vowels than back vowels. Their finding shows an interaction of
noun-verb differences with the phonological properties such as front vs. back and frequency.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
200
(i) Sublexical noun-verb dissociations interact with word structure frequency. This
is shown in an evident noun-verb dissociation in non-major word types and an
insignificant noun-verb dissociation in major word types (VN and NN).
(ii) Word structure frequency interacts with word structure complexity. This is
shown in a general tendency toward the two-morpheme word structure across
compound types.
(iii) Across compound types, both Broca’s aphasics Wernicke’s aphasics have a
stronger tendency toward a single-morpheme word compared with their
production of three-morpheme words. This reflects a structural complexity effect.
Second, single-factor accounts encounter more problems because they
cannot capture the access of a lexical item in Chinese that involves more than one
lexical level. Bates et al. (1991) pointed out that any lexical account which
assumes a single level of lexical representation had to be revised, because it could
not account for the double dissociation observed at the sublexical level (i.e., inside
of VN compounds).
The present study replicates what Bates et al.(1991) have found regarding
the presence of the sublexical level. Moreover, it provides us with an additional
finding concerning the noun-verb dissociation at the word structure level.
Specifically, the result shows that noun-verb dissociations are evident in the
production of word structures. This is shown in the tendency toward N, NN,
NNN in Broca’s aphasics and the tendency toward V, W , V W in Wernicke’s
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
201
aphasics. This finding is not compatible with the accounts that assume single-level
lexical representations.
7.3 The Compatibility of Chinese Facts with the Competition Model
Problems with the four single-factor accounts (discussed in chapter two)
revolve around two facts. The first is that a word may contain different levels of
information such as the level for whole words, the level for word components and
the level for word-class substructures. The second is that noun-verb dissociations
are affected by more than one factor. These factors often interact with each other.
In dealing with these two phenomena, connectionist models such as the
Competition Model have an advantage over the four single-factor accounts.
First, to handle the lexical and the intra-lexical levels, the Competition
Model assumes that whole words are stored as distributed representations with a
permeable internal structure. In this model, subcomponents of a complex word
can be activated altogether or separately, depending on the context. This model
has been applied to languages such as English. For example, the past tense may be
activated with the root as a whole chunk in the course of whole word learning;
however, it may also be activated as a separate past tense morpheme to attach to a
verb root in systematic treatment of verbs (Plunkett & Marchman, 1993).2 1 In a
2 1 For example, in the course of whole word learning stage, past tense verbs “walked” and “went”
are treated as whole words. However, after children acquire a systematic treatment of verb, they
may decompose the word “walked” into a verb root “walk” and a past tense morpheme “-ed.”
This knowledge of “past tense morpheme” is observed in their attaching the past tense “-ed” to
the irregular verb “went” to yield the incorrect verb “ wented.”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
202
similar spirit, the Competition Model offers an explanation of how a Chinese
compound may be activated as an unanalyzable whole chunk, its components may
be activated as separate sublexical elements, and the word structure may also be
activated independently. This explanation has an advantage over the four accounts
because it can capture the lexical and the intra-lexical facts in Chinese.
Second, the Competition Model offers an account which may capture the
complex interaction of factors underlying noun-verb dissociations in Chinese.
These factors include grammatical factors (such as structural complexity), semantic
factors (such as object vs. action distinction), frequency effects, and cohort
structure effects.2 2 The impact of each factor is shown in the major findings of this
study listed above (see also the discussion in Bates et al. (1991)).
For example, the observation of a stronger tendency of aphasics toward
single-morpheme words than toward three-morpheme words may be attributed to
a structural complexity effect. This factor is also observed by the syntactic
account. The difference between the syntactic account and my account is that in
the former the structural complexity is the primarily factor to distinguish nouns
from verbs. In my explanation, the structural complexity effect is one of many
factors that may contribute to noun-verb dissociations.
~The present study has not been designed to elicit the cohon effect. However, the cohort effect
is observed in subjects’ responses, in particular subjects with Wernicke’s aphasia. Because the
cohort effect seems very influential, it is necessary to conduct a research that investigates the
cohort effect specifically.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
203
Bates et al. (1991) suggest that noun-verb dissociations in Chinese may
rest on a semantic basis. This is because no grammatical marker is available to
determine grammatical category of a word in Chinese. The semantic difference
seems to be the only factor left for the distinction between nouns and verbs.
Moreover, numerous studies on sentence interpretations of Chinese normal
subjects and aphasics have demonstrated that Chinese subjects (aphasics or normal
subjects) primarily reply on semantic strategies (e.g. animacy) to determine the
grammatical role of a noun, such as a subject or an object, within a sentence
(Chen, Tzeng, & Bates, 1990; Li, Bates, & MacWhinney, 1993; Li, Bates, Liu, &
MacWhinney, 1992; Liu, Bates, & Li, 1992; Miao, Chen, & Ying, 1986). A
reliance on semantic properties (e.g. animacy) may contribute to the distinction
between nouns and verbs.
The finding that noun-verb dissociations are not significant in high-
frequency word types but are significant in low-frequency word types suggests that
the action-object distinction is not sufficient to account for noun-verb dissociations
between Broca’s aphasics and Wernicke’s aphasics. Instead, frequency is part of a
complex interaction of different factors that affect noun-verb dissociations.
Structural complexity, the action-object distinction, and the frequency
effect have been carefully investigated in this study. However, the cohort effect
has not been investigated because it is very difficult to control the size of a lexical
cohort that a lexical item belongs to. The raw data suggests that the cohort effect
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
204
may greatly affect noun-verb dissociations. For example, when a Wernicke’s
aphasic was asked to give a name for the target picture “ball point pen,” which was
“yuan-zi-bi” (literally atom-pen, meaning “ball point pen”) in Chinese, the aphasic
patient generously produced four different words for the target (none of them are
the target). As shown in examples 29a-d, these four compound words share the
first two components with the target.
29 Lexically Related Compounds
a. yuan-zi-ku atom-pants “elastic pants”
b. yuan-zi-lu atom-oven “nuclear “
c. yuan-zi-zao atom-oven “nuclear” (created word)
d. yaun-zi-dan atom-bomb “nuclear-bomb”
The subject was able to activate the first two components correctly. However, he
failed to provide the correct one after the first two-morpheme component because
he kept activating almost all the possible components in the languages that follow
the first two components. In view of this cohort effect, it is very likely that
speakers of Chinese (both aphasic and normal subjects) very often activate a set of
cohort at both lexical and sublexical levels.
In fact, it has been reported that normal speakers of Chinese also make
errors which are lexically related to the target (Lai, 1984). Their errors may be
divided into two kinds, depending on the direction of association (vertical or
horizontal).
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
205
Examples in 30a-i represent a horizontal association and examples in 3 la-i
represent a vertical association.
30. Horizontal Association
a. jin-tian
today-day
“today”
b. tian-sheng
day-born
“innate”
c. sheng-can
bom-produce
“labor”
d. can-jia
produce-holiday”
“maternity leave”
e. jia-qi
holiday-period
“vocation”
f. qi-wang
period-expect
“to expect”
g. wang-xiang
expect-suburb
“waiting for going home”
h. xiang-cun
suburb-country
“country”
i. cun-zhuang
country-village
“village”
31. Vertical Association
a. dian-deng
elect ric-Iamp
“lamp”
b. dian-shi
electric-vision
“television”
c. dian-hua
electric-speech
“telephone”
d. dian-ying
electric-shadow
_ • _ * 7
movie
e. dian-nao
electric-brain
“computer”
f. dian-zi
electric-Diminutive Marker
“electronics”
g. dian-qi
electric-instrument”
“electronic items”
h. dian-guo
electric-wok
“rice cooker”
i. dian-suan-ji
electric-count-machine
“calculator”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
206
In a horizontal association, the second component of the target is identical
with the first component of the next word. The second component “tian” of 30a
is identical with the first component “tian” of 30b, and the second component
“sheng” of 30b is identical with the first component ‘sheng” of 30c. This
horizontal association keeps going to the rest of examples. Unlike the horizontal
association, the vertical association show that all the compound words (3 la-i)
share the first component “dian.”
The cohort effect may affect noun-verb dissociations in two aspects. First,
the cohort size may influence the accessibility of nouns and verbs. For example, if
the cohort size of a noun target is larger than the cohort size of a verb target, it
may be more difficult to produce the correct noun than the correct verb for both
aphasic groups. This is because there are more competitors (i.e., lexically related
items) surrounding the noun target than the target verb. In a case like this, noun-
verb dissociations may be obscured by the cohort size effect.
Moreover, the cohort size may have a greater impact on one aphasic type
(e.g. Wernicke’s) than the other aphasic type (i.e., Broca’s). This size effect will
cause an overall poor lexical production by one aphasic type while leaving the
other aphasic type less affected. Indeed, as reported in literature, among these
subject groups, Wernicke's aphasics are especially sensitive to the cohort effect
(Bates et al., 1987b). This characteristic of Wernicke’s aphasics has been
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
207
intensively discussed in the syndrome called “paragrammatism,” in contrast with
“agrammatism” observed in Broca’s aphasia. While paragrammatic errors often
involve open class word errors; agrammatic errors are mostly structural errors or
errors of elements which are used to construct structure, e.g. closed class items.
These studies also reported that Wernicke's aphasics have a very strong tendency
for making substitution errors among open class items (Bates, Friederici, &
Wulfeck, 1987b; Butterworth, 1983; 1979; Butterworth & Howard, 1987; Miller
& Ellis, 1987). These substitution errors are often a variant of target (e.g. ”skut”
for “scout”) which are morphologically related to the target (Miller & Ellis, 1987,
p. 253). In contrast, Broca's aphasics do not produce such variations. They tend
to stay with a single lexical item (Bates, Friederici, & Wulfeck, 1987b). Because
Wernicke’s aphasics are distracted by lexically associated non-targets, their lexical
production on open-class words should be worse than Broca’s aphasics. Thus, the
effect of cohort size interacts with noun-verb dissociations. Although, the data
presented in this thesis contained a significant number of lexically related errors,
which might be attributed to the cohort effect, this was not systematically
investigated in the present study. However, the impact of the cohort effect on
intra-lexical activities certainly calls for a further investigation on the interaction of
the cohort effect with noun-verb dissociations.
Thus far, I have shown that the Competition Model is able to capture the
multi-levels within a Chinese compound and the multi-factors interaction within
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
208
the noun-verb effect. Based on the idea of the Competition Model, a Chinese
word contains three levels of information. When a speaker tries to access a word,
he/she may activate three levels of information simultaneously. However, if the
language ability of the subject is impaired, he/she may find difficulty in accessing
all three levels. As a result, he/she may activate only one level. For example,
sublexical errors occur when the subject activates only the component level. No
matter which level is activated, all the factors (including frequency, structural
complexity, action vs. object, and cohort effects) surrounding the lexical item
come to affect the lexical access. The impact of these factors may differ from one
another.
7.4 A New Approach to the Old Problem
The proposal presented above provides us with a new look at the status of
Chinese compounds. It has the advantage over traditional approaches because it
provides an explanation for the vague boundary between compounds and phrases.
As I discussed earlier in chapter three, compounds and phrases are closely related
to each other. Some may be identical. For example; the compound “goldfish”
may be interpreted as the phrase “golden fish.” Yet, some phrases are treated like
compound words (for semantic idiomaticity sake); for instance, the verb phrase
“chi fan” (literally eat rice, meaning to eat rice) can also be re-interpreted as a
generic verb “to eat,” indicating the name of the action. Moreover, to create a
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
209
compound, phrases (e.g. 32a) may become compounds (e.g. 32b) by deleting some
morphemes (in example 32a, underlined morphemes are those being deleted).
32 a. Mei-guo zhi-zao de xi yi-ffu de ii-qi.
USA-country make-make RCM wash clothes-clothes RCM machine-
instrument.
b. Mei-zhi-xi-yi-ji
USA-made washing-machine
All the approaches to distinguishing compounds from phrases fail because they
assume that a structure can carry only one function. For example, “chi-fan” can
only function as a compound or it can only serve as a phrase, but not both. In the
proposal based on the Competition Model, it is possible that a word may contain
two different readings. One is the reading at the whole word level the other is the
reading at the component level. Thus, the semantic interpretations of a word
resemble the interpretations of a phrase, both of which may have literal
interpretations and whole chunk interpretations. Literal interpretations are
reflected in a word-by-word reading for the phrase and a component-by-
component reading for the compound. The whole chunk reading is shown in the
whole phrase meaning and the whole word meaning. This similarity between
words and phrases triggers an active interaction between the word system and the
phrase system, as shown in Fig. 9 (Chen & Shi, 1992). The interaction is well
demonstrated in many cases of alternations between compounds and phrases (see
section 3.1). It is these alternations that trouble Chinese linguists who have
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
210
attempted to separate compounds from phrases without realizing the difficulty
caused by the ongoing alternations between whole word interpretations and literal
interpretations.
Fig. 9 Bi-directional feeding relation between phrases and words
Phrases Whole Word
Words Components
Phrase Structures Word Structures
7.5 Concluding Remarks
This study has addressed intra-lexical noun-verb dissociations in Chinese
aphasia. Results from the present study support the claim that a compound may be
treated as a whole word, or may be decomposed into components and word
structures. The presence of three levels challenges the morphological account, the
syntactic account, the lexical account and the semantic-conceptual account. The
morphological and the syntactic account fail because, on the one hand, noun-verb
dissociations show in aphasia of Chinese where verbs and nouns do not differ in
their morphological complexity, on the other hand, complex structures (e.g. two-
morpheme word structures) have a higher accessibility than simple structures (e.g.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
211
single-morpheme word structures). The lexical account and the semantic account
need to be revised because these two accounts cannot capture intra-lexical noun
verb dissociations. Consequently, they are forced to proliferate several
sublexicons to accommodate Chinese facts.
Compared with these four accounts, I find that the Competition Model has
the best compatibility with the facts from Chinese. One reason is that the
Competition Model allows different levels of information within a word. This
multi-level account may capture the presence of noun-verb dissociations at the
sublexical level and the word structure level. Moreover, it offers a multi-factor
explanation for the noun-verb dissociation. This multi-factor account captures (i)
a noun-verb effect— the tendency toward N, NN, NNN in Broca’s aphasia and the
tendency toward V, W , V W in Wernicke’s aphasia, (ii) a frequency effect— a
strong tendency toward the major word structure, namely two-morpheme word
structure, across compound types, (iii) an interaction of frequency effects with
noun-verb effects— an evident noun-verb dissociation in non-major word types and
an insignificant noun-verb dissociation in major word types (VN and NN), (iv) a
structural complexity effect— both Broca’s aphasics Wernicke’s aphasics have a
stronger tendency toward a single-morpheme word compared with their
production of three-morpheme words.
Based on the Competition Model, the present study provides an alternative
proposal to confront (instead of withdrawing from) the most controversial issue in
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
212
Chinese linguistics by emphasizing the alternation between compound
interpretations and literal interpretations.
In conclusion, Chinese compounds allow an activation of the whole chunk
as well as an activation of parts within a word. This parallel lexical processing is
manifested very well in noun-verb dissociations of Chinese aphasics. Although this
parallel lexical processing poses problems to the four accounts, it is totally
compatible with connectionist models such as the Competition Model. The co
existence of whole chunk readings (i.e., compound readings) and component
readings (i.e., literal readings) helps us deal with the difficulty in defining the
notion of compounds in Chinese.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
APPENDIX 1
SUBJECT INFORMATION
Age Sex Hand Post Etiology Lesion
edness Onset
Broca’s
#2 37 M R 8 y Head injury Lt F-T-P
#4 32 M R 5 m Head injury L't F
#7 37 M R 4 m CVA No CT report
#11 43 M R 5 m CVA No CT report
#21 69 F R 20 y CVA LtMCA
#23 36 M R 5 m Head injury L't F-T; R't F
#28 70 M R 5 m VA L't Frontal
#29 56 F R 4 m CVA L't Basal ganglia
#30 64 M R 3 m CVA L't Putamen
#31 62 M R 1 m CVA L't F-P
Wernicke’s
#6 47 M R 3 m CVA L't lateral ventricle/suius
#8 59 M R 14 m CVA L’ t T
#15 54 M R
3 y
CVA L't Basal ganglia
#17 62 M R 2.5 y CVA L’ t F-T-P
#20 67 M R 9 y CVA L't T-P
#22 69 M R 8 m Hemorrhage L't Putamen/intemal capsule
#25 68 M R 10 m CVA L't Temporal
#35 62 M R 2 m CVA L't Lacunar infarction
#36 63 M R 3 m CVA L't T-P
#37 47 M R I m CVA Brain Stem, Pons
Normal Control
#38 34 M R
#39 45 M R
#40 37 F R
#41 63 F R
#42 60 M R
#43 55 M R
#44 40 M R
#45 55 M R
#46 37 M R
M: male; F: female; R: right handedness; y: year (s); m: month(s); L’t= left hemisphere;
R’t= right hemisphere; F: Frontal Lobe, T: Temporal Lobe; P: Parietal Lobe
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
214
APPENDIX II
EXPERIMENT STIMULI: NOUN-VERB
VN (Verbal)
1. jiao-shui
water-water “to water”
2 . tiao-wu
jump-dance “to dance”
3. tou-qiu
pitch-ball “to pitch”
4. diao-yu angle-fish “to fish”
5. hua-chuan
row-boat “to row”
6 . tang-fa heat-hair “to perm”
7. ju-gong
bow-bow “to bow”
8 . kai-deng
open-lamp “to turn on”
9. qi-chuang up-bed “to get up”
10. shui-jiao
sleep-sieep “to sleep”
11. chang-ge sing-song “to sing”
12. xie-zi
write-character “to write”
13. qi-ma
ride horse “to ride”
14. xi-zao
wash-bath “to bathe”
15. tan-qin
play-musical instrument “to play”
16. chi-fan
eat-rice “to eat”
17. chao-cai
fiy-vegetable “to stir fry”
18. zou-lu
walk-road “to walk”
19. tiao-sheng
jump-rope “to jump rope”
20. zhao-xiang
shine-photo “to photo”
21. tiao-shui
jump-water “to dive”
22. hua-xue
slide-snow “to ski”
23. kao-rou roast-meat “to roast”
24. hua-zhuang
put on-cosmetics “to make-up”
25. huai-yun
bear-pregnancy “to be pregnant”
26. wo-shou
hold-hand “to shake (hands)'
27. you-yong swim-swim “to swim”
2 8 .shuo-hua
speak-speech “to speak”
29. ju-shou
raise-hand “to raise(hand)”
30. dian-huo
light-fire “to light”
31. chou-yan
smoke-cigar “to smoke”
32. an-ling
push-bell “to ring”
33. wan-yao
bend-waist “to bow”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
215
VN (Nominal)
1. suan-pan
2. fei-ji
3. qi-er
4. tiao-qi
5. mo-bu
6. zheng-long
7. yi-sheng
8. tou-shou
9. wan-ju
10. zhi-piao
11. wo-shi
12. zhao-pian
13. fa-piao
14. shui-yi
15. yin-zhang
16. wei-qun
17. nao-zhong
18. qing-tie
19. bei-zhen
20. jian-dao
21. tuo-xie
22. xi-guan
23. chao-fan
24. wei-qi
25. yu-gang
26. diao-deng
27. hu-shi
28. diao-chuang
NN (Nominal)
1. dian-hua
2 . huo-che
3. yi-fu
4. bao-zhi
5. qi-che
6 . tie-gui
7. dian-nao
8. tou-fa
9. chuang-hu
10. gang-qin
count-dish
fly-machine
stand-goose
jump-chess
wipe-rag
steam-cage
cure-man(professional)
pitch-hand
play-instrument
issue-ticket
lie-room
shine-slice
distribute-ticket
sleep-clothes
print-seal
wrap-skirt
alarm-clock
invite-card
put-needle
cut-knife
pull-shoe
suck-hose
fry-rice
surround-chess
bath-tub
hang-lamp
protect-stafF
hang-bed
“abacus”
“airplane”
“penguin”
“Chinese checker”
“wiper”
“steamer”
“doctor”
“pitcher”
“toy”
“check”
“bedroom”
“photo”
“invoice”
“pajamas”
“seal”
“apron”
“alarm clock”
“invitation card”
( i
pm
“scissors”
“slippers”
“straw”
“fry rice”
“Japanese chess; go”
“bathtub”
“art lamp”
“nurse”
“hammock”
tele-speech
fire-car
clothes-clothes
newspaper-paper
gas-car
steel-trail
tele-brain
head-hair
window-window
steel-musical instrument
“telephone”
“train”
“clothes”
“newspaper”
“car”
“railroad”
“computer”
“hair”
“window”
“piano”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
216
11. dian-ying
12. mao-bi
13. shu-bao
14. di-qiu
15. zu-qiu
16. er-duo
17. qian-bi
18. jian-bang
19. huo-cai
2 0 . mian-bao
21. niu-nai
22. ya-chi
23. yu-mi
24. ping-guo
25. yi-ba
26. lan-qiu
27. xin-zang
28. bang-qiu
29. mi-feng
30. jiang-you
31. nai-fen
32. la-zhu
33. mian-bei
34. jin-yu
35. che-lun
36. qi-shui
37. pi-bao
38. yu-yi
39. lan-hua
40. cha-ye
41. huo-ji
42. mao-yi
43. shou-zhuo
44. mao-xian
45. hua-ping
46. hai-xing
47. tou-kui
48. zhu-gan
49. yao-shi
50. you-tong
51. bei-ke
52. lian-ou
53. bing-gan
tele-shadow
fur-pen
book-bag
ground-ball
foot-ball
ear-shape
lead-pen
shoulder-shoulder
fire-wood
wheat-dumpling
cow-milk
tooth-tooth
jade-rice
apple-fruit
tail-tail
basket-ball
heart-organ
stick-ball
honey-bee
sauce-oil
milk-powder
wax-candle
cotton-cover
gold-fish
car wheel
gas-water
leather-bag
rain-clothes
orchid-flower
tea-leave
fire-chicken
fur-clothes
hand-bracelet
fur-thread
flower-bottle
sea-star
head-helmet
bamboo-pole
key-key
post-pail
shell-shell
lily-root
cookie-dryness
movie
“brush pen”
“school bag”
“earth”
“football”
“ear”
“pencil”
“shoulder”
“match”
“bread”
“milk”
“tooth”
‘com”
“apple”
“tail”
“basketball”
“heart”
“baseball”
“bee”
“soybean sauce”
“milk powder”
“candle”
“comforter”
“gold fish”
“wheel”
“soda”
“bag”
“raincoat”
“orchid”
“tea”
“turkey”
“sweater”
“bracelet”
“knitting wool”
“vase”
“sea star”
“helmet”
“bamboo pole”
“key”
“mailbox”
“shell”
“lily root”
“cookie”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
217
54. dan-gao
55. cha-hu
56. lun-tai
57. nai-ping
58. wang-qiu
59. ban-ma
60. hua-pen
61. cao-mei
62. fan-shu
egg-cake
tea-pot
wheel-tire
milk-bottle
net-ball
stripe-horse
flower-pot
grass-berry
westem-yam
“cake”
“teapot”
“tire”
“bottle for milk’
“tennis”
“zebra”
“flower pot”
“strawberry”
yam
NNN (Nominal)
1. huo-che-zhan fire-car-station “train station”
2. mei-gui-hua rose-rose-flower “rose”
3. shui-long-tou water-dragon-head “faucet”
4. shou-dian-tong hand-electric-tube “flashlight”
5. yu-mao-qiu feather-fur-ball “badminton”
6 . shui-guo-dao water-fruit-knife “knife”
7. ma-xi-tuan horse-show-group circus
8. xian-ren-zhang fairy-person-palm “cactus”
9. ji-qi-ren machine-machine-man “robot”
10. qian-bi-he lead-pen-case “pen case”
11. jin-yu-gang gold-fish-tub “fish tank”
12. ri-guang-deng sun-light-lamp “fluorescent laj
13. mao-tou-ying cat-head-eagle “owl”
14. shen-fen-zheng
body-identity-certificate “ID”
15. le-se-tong trash-trash-pail “trash pail”
16. san-jiao-xing three-angle-shape “triangle”
17. ri-ben-ren sun-root-man “Japanese”
18. niu-zai-ku
cow-kid-pant
44 * _ _
jeans
19. huo-cai-he fire-wood-case “match case”
20. xiang-pi-quan rubber-skin-circle “rubber band”
VNN (Nominal)
1. ji-cheng-che
2. shou-yin-ji
3. hua-zhuang-pin
4. pen-shui-chi
5. zhao-xiang-ji
6 . xi-fa-jing
count-distance-car “taxi”
receive-sound-machine “radio”
put-cosmetics-product “cosmetics’
spring-water-fountain “fountain”
shine-picture-machine “camera”
wash-hair-essence “shampoo”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
218
7. ji-shi-bu
8. chui-feng-ji
9. xiang-ri-kui
10. you-yong-chi
11. Iu-yin-dai
12. tiao-se-pan
13. xi-chen-qi
14. da-zi-ji
15. gua-hu-dao
16. you-yong-qnan
17. xi-yi-ji
18. pai-qi-guan
19. kai-guan-qi
20. xi-yi-fen
21. feng-yi-ji
22. luo-di-chuang
record-event-note
blow-wind-machine
ace-sun-sunflower
swim-swim-pool
record-sound-belt
mix-color-dish
suck-dust-instrument
hit-character-machine
shave-mustache-knife
swim-swim-circle
wash-clothes-machine
pass-gas-hose
open-can-instrument
wash-clothes-powder
sew-clothes-machine
fall-ground-window
“notebook”
“hair dryer”
“sunflower”
“swimming pool”
“audio tape”
“palette”
“vacuum cleaner”
“typewriter”
“shaver”
“life buoy”
“washing machine”
“muffler”
“can opener”
“detergent”
“sewing machine”
“window”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
APPENDIX III
219
EXPERIMENT STIMULI: FREQUENCY
VN (Verbal) Component Frequency Compound Frequency
1. jiao-shui 10-1981 5
2 . tiao-wu 201-207 20
3. tou-qiu 333-832 8
4. diao-yu 40-321 7
5. hua-chuan 32-449 1
6 . tang-fa 17-140 0
7. ju-gong 10-17 6
8. kai-deng 1916-146 1
9. qi-chuang 2102-170 16
10. shui-jiao 335-1012 51
11. chang-ge 310-297 49
12. xie-zi 536-764 4
13. qi-ma 107-724 1
14. xi-zao 173-21 18
15. tan-qin 93-147 1
16. chi-fan 713-302 81
17. chao-cai 8-161 0
18. zou-lu 1281-917 37
19. tiao-sheng 201-69 0
20. zhao-xiang 601-1645 2
21. tiao-shui 201-1981 1
22. hua-xue 107-133 8
23. kao-rou 314-248 2
24. hua-zhuang 1396-18 4
25. huai-yun 219-34 12
26. wo-shou 117-1536 20
27. you-yong 182-60 46
28. shuo-hua 5640-914 166
29. ju-shou 656-1536 28
30. dian-huo 1659-539 11
31. chou-yan 82-150 11
32. an-ling 193-37 1
33. wan-yao 98-71 10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
220
VN (Nominal) Component Frequency Compoun
1. suan-pan 507-145 11
2 . fei-ji 464-1198 3
3. qi-er 240-84 5
4. tiao-qi 201-21 0
5. mo-bu 59-431 1
6 . zheng-Iong 104-58 4
7. yi-sheng 535-5253 62
8 . tou-shou 333-1536 30
9. wan-ju 222-678 27
10. zhi-piao 424-187 18
11. wo-shi 48-359 12
12. zhao-pian 601-632 57
13. fa-piao 3345-187 1
14. shui-yi 335-512 2
15. yin-zhang 307-227 3
16. wei-qun 327-25 2
17. nao-zhong 136-12 3
18. qing-tie 797-20 3
19. bei-zhen 854-119 4
2 0 . jian-dao 80-96 4
2 1 . tuo-xie 68-96 6
2 2 . xi-guan 404-796 4
23. chao-fan 8-302 1
24. wei-qi 327-21 2
25. yu-gang 21-9 I
26. diao-deng 19-146 1
27. hu-shi 354-773 26
28. diao-chuang 19-170 1
NN (Nominal) Component Frequency Compound Frequency
1. dian-hua 1085-914 89
2 . huo-che 539-894 49
3. yi-fii 512-762 186
4. bao-zhi 692-377 85
5. qi-che 207-894 141
6 . tie-gui 297-45 5
7. dian-nao 1085-209 34
8 . tou-fa 1357-140 62
9. chuang-hu 174-111 17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
221
10. gang-qin 104-137 37
11. dian-ying 1085-758 126
12. mao-bi 324-343 12
13. shu-bao 1063-315 10
14. di-qiu 4064-832 136
15. zu-qiu 558-832 52
16. er-duo 187-98 63
17. qian-bi 56-343 39
18. jian-bang 92-61 46
19. huo-cai 539-33 10
20 . mian-bao 55-551 22
21. niu-nai 174-44 22
22. ya-chi 149-66 25
23. yu-mi 84-142 27
24. ping-guo 13-1564 10
25. yi-ba 77-146 21
26. lan-qiu 86-832 18
27. xin-zang 2482-50 19
28. bang-qiu 93-832 17
29. mi-feng 38-49 13
30. jiang-you 19-322 12
31. nai-fen 44-117 10
32. la-zhu 59-39 26
33. mian-bei 125-1083 1
34. jin-yu 765-821 3
35. che-lun 849-179 5
36. qi-shui 207-1981 6
37. pi-bao 315-551 9
38. yu-yi 311-512 2
39. lan-hua 122-750 8
40. cha-ye 172-266 8
41. huo-ji 539-116 1
42. mao-yi 324-512 7
43. shou-zhuo 1536-3 2
44. mao-xian 324-485 4
45. hua-ping 750-120 6
46. hai-xing 1189-397 I
47. tou-kui 1357-2 1
48. zhu-gan 94-16 5
49. yao-shi 9-13 8
50. you-tong 62-37 5
51. bei-ke 48-46 4
52. lian-ou 35-1 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
222
53. bing-gan 31-21 8
54. dan-gao 131-18 5
55. cha-hu 172-53 3
56. lun-tai 179-37 2
57. nai-ping 44-120 I
58. wang-qiu 60-832 1
59. ban-ma 24-724 2
60. hua-pen 750-28 4
61. cao-mei 558-2 2
62. fan-shu 4-6 2
NNN (Nominal) Component Frequency Compound Frequency
1. huo-che-zhan 539-894-376 11
2 . mei-gui-hua 15-16-750 5
3. shui-long-tou 1981-78-1357 2
4. shou-dian-tong 1536-1085-37 2
5. yu-mao-qiu 34-324-832 7
6 . shui-guo-dao 1981-1564-96 0
7. ma-xi-tuan 724-290-419 3
8. xian-ren-zhang 55-11290-118 3
9. ji-qi-ren 1198-8-1129 1
10. qian-bi-he 56-343-28 3
11. jin-yu-gang 765-321-9 1
12. ri-guang-deng 2264-933-14 2
13. mao-tou-ying 64-1357-12 2
14. shen-fen-zheng 1423-2998-379 1
15. le-se-tong 31-28-16 1
16. san-jiao-xing 2454-343-1114 25
17. ri-ben-ren 2264-521-11290 52
18. niu-zai-ku 174-75-21 1
19. huo-cai-he 539-33-28 2
20. xiang-pi-quan 39-315-89 13
VNN (Nominal) Component Frequency Compound Frequency
1 . ji-cheng-che 933-403-894 9
2. shou-yin-ji 696-1050-1198 21
3. hua-zhuang-pin 1396-18-653 6
4. pen-shui-chi 70-1981-51 3
5. zhao-xiang-ji 601 -1645-1198 6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
223
6. xi-fa-jing 173-140-813 0
7. ji-shi-bu 730-2820-12 1
8. chui-feng-ji 162-827-119 2
9. xiang-ri-kui 1534-2660-4 1
10. you-yong-chi 182-60-51 2
11. lu-yin-dai 112-1050-93 0
12. tiao-se-pan 140-1052-14 1
13. xi-chen-qi 404-38-582 0
14. da-zi-ji 914-764-1198 I
15. gua-hu-dao 26-14-96 0
16. you-yong-quan 182-60-89 0
17. xi-yi-ji 173-512-119 2
18. pai-qi-guan 332-1584-79 0
19. kai-guan-qi 1916-31-582 0
20. xi-yi-fen 173-512-126 4
21. feng-yi-ji 69-512-1198 5
22. luo-di-chuang 417-4064-17 1
N (Nominal) Component Frequency Compound Frequency
1. cao (grass) 320 123
2. iian (face) 263 203
3. men (door) 896 430
4. shu (book) 1063 423
5. qian (money) 526 370
6 . hai (ocean) 1189 343
7. hua (flower) 750 312
8. shu (tree) 375 154
9. zui (mouth) 149 101
10. xin (letter) 867 363
11. shou(hand) 1537 638
12. ma (horse) 724 251
13. yu (fish) 321 193
14. shui (water) 1981 951
15. fan (rice) 302 80
16. he (river) 555 106
17. gong (bow) 21 144
18. tui(leg) 109 71
19. jian (sword) 58 25
20. jiao (foot) 295 116
21. fo (Budda) 100 30
22. yan(salt) 130 48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
224
23. zhong (clock) 306 50
24. mao (cat) 64 54
25. miao (temple) 87 53
26. gou (dog) 88 66
27. cong (onion) 13 4
28. yang (sheep) 93 47
29. li (pear) 16 7
30. she (snake) 22 1
31. wan (bowl) 79 71
32. yao (medicine) 188 59
33. suan (garlic) 3 I
34. sun (bamboo) 10 0
35. dao (knife) 96 0
36. biao (watch) 37 32
37. dan (egg) 131 35
38. jing (whale) 13 9
39. qiu (ball) 832 101
40. guo (cooker) 48 39
41. cai (vegetable) 161 64
42. lei (tears) 85 26
43. deng (lamp) 146 15
44. chuan (boat) 449 254
45. rou (meat) 248 104
46. qi (flag) 41 23
47. che (car) 894 203
48. zhuo (desk) 230 112
49. xie (shoes) 96 30
50. bi(pen) 343 112
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
225
APPENDIX IV
SAMPLE PICTURES OF
THE ACTION VS. OBJECT NAMING TASK
1. yu “fish” (N) 2. tou-fa “hair” (NN)
3. qi-er “penguin” (VN-N) 4. xi-zao “to take a bath” (VN-V)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
226
APPENDIX IV (Continued)
SAMPLE PICTURES OF
THE ACTION VS. OBJECT NAMING TASK
5. suan-pan “abacus” (VN-N)
7. kai-guan-qi “can opener” (VNN)
6 . chao-cai “to stir fry” (VN-V)
8 . xian-ren-zhang “cactus” (NNN)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
227
APPENDIX V
LEXICAL ERRORS OF BROCA’S VS. WERNICKE’S:
VERBAL VN COMPOUNDS & NOMINAL VN COMPOUNDS
1. jiao-hua water-flower ‘to water”
Wernicke’s responses
xi-hua
hua-ming
mai-shui
dao-shui
guan-
wash-flower
invent-bright
buy-water
pour-water
pour-
; ‘ to wash flowers’
‘to invent”
‘ to buy water”
‘to pour”
‘ to pour”
Broca’s responses
-shui
jiao-qiu
sa-shui
hua-shui
-water
water-ball
sprinkle-water
flower-water
“water”
invented word
to sprinkle”
invented word
2. tiao-wu jump-dance ‘to dance’
Wernicke’s responses
tiao-zou jump-walk
tiao- jump
tiao-quan-quan jump-circle-circle
“to jump away”
jump
“to jump in circle”
Broca’s responses
wu-dao dance-dance
wu-dao dance-dance
‘dance” (noun)
‘dance” (noun)
3. tou-qiu throw-ball
Wernicke’s responses
qiu-ti
pi-xie
bang-qiu
pa-ya-kiu
diu-qi-zi
ball-carry
leather-shoe
bat-ball
hit-wild-ball
throw-chess-suffix
‘to pitch”
invented word
“shoe”
“baseball”
“to play baseball”
“to throw chess”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
228
Broca’s responses
-qiu -baU “ball”
tou-shou throw-hand “pitcher”
bang-tou bat-throw invented word
tou-shou throw-hand “pitcher”
bang- bat “bat”
bang-qiu bat-ball “baseball”
tou-shou throw-hand “pitcher”
bang-shou bat-hand invented word
»-yu angle-fish “to fish”
Wernicke’s responses
yu-gou fish-hook “fishhook”
diao angle “to angle”
jing-yu whale-fish “whale”
diao-chen angle-sink invented word
pi-yu skin-fish invented word
Broca’s responses
diao-gan angle-rod “fishing rod”
yu-gou fish-hook “fishhook”
5. hua-chuan row-boat ‘to row’
Wernicke's responses
hua- row
hua-qi-che row-gas-car
diao angle
hua- row
‘ to row
‘to row a car”
‘to angle”
‘to row”
Broca’s responses
chuan boat “boat”
tui push “to push”
hua-shui row-water “to swim”
zuo-chuan sit-boat “to take a boat”
6. tang-fa heat-hair “to perm”
Wernicke’s responses
mao hair “hair”
xi-fa wash-hair “to wash hair”
zuo-fa do-hair “to do the hair”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
juan-fa curl-hair “to curl hair”
229
Broca’s responses
-fa
tou-fa
tou-fa
tou-fa
xi-fa
hair
head-hair
head-hair
head-hair
wash-hair
‘hair”
‘hair”
‘hair”
‘hair”
‘to wash hair”
7. ju-gong
jing-li
bow-bow
respect-polite
Wernicke’s responses
-li polite
gui-zhong expensive-heavy
Broca’s responses
“to bow”
“to bow”
incomplete word
“precious” (for “gui-xia,r
meaning kneel-down)
ju-bing bow-ice invented word
-deng open-lamp “to turn on”
Wernicke’s responses
guan-dian close-electric incomplete word
da- hit incomplete word
kai- open incomplete word
kai- open incomplete word
dian-deng electric-lamp “lamp”
kai-men open-door “to open door”
kai- open incomplete word
ban-kai move-open “to open”
Broca’s responses
-deng lamp “lamp”
dian-hua electric-speech “telephone”
da-deng hit-lamp “to light”
guan-deng close-lamp “to turn off”
kai-guan open-close “switch”
an-deng press-lamp “press lamp”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
9. qi-chuang up-bed “to get up”
230
Wernicke’s responses
ti-yu body-exercise “physical education”
qi-lai up-come “come up”
zao-qi early-up “to getup early”
Broca’s responses
-chuang bed “bed”
10. shui-jiao sleep-sleep
Wernicke’s responses
shui-di sleep-ground
Broca’s responses
-jiao sleep
11. chang-ge sing-song
Wernicke’s responses
qi-ge start-song
chang-xi sing-opera
chang-gua-qu sing-song-classifier
chang sing
Broca’s responses
chang sing
-ge song
-ge song
12. xie-zi write-character
Wernicke’s responses
xin-ren trust-trust
xie- write
xie- write
Broca’s responses
xie-mao-bi write-brush-pen
“to sleep”
invented word
incomplete word
“to sing”
invented word
“to sing opera”
“to sing”
“to sing”
‘to sing”
‘song”
‘song”
‘to write’
to trust
‘ to write”
‘to write”
xie-xm write-letter
“to write with
a brush pen”
“to write letter”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
231
13. qi-ma ride horse “to ride”
Wernicke's responses
qi-pao ride-run invented word
ma horse “horse”
sai-ma compete-horse “horse race”
qi- ride “to ride”
Broca’s responses
ma horse “horse”
ma-chang horse-field invented word
ma horse “horse”
ma horse “horse”
pao-ma run-horse invented word
ma horse “horse”
sai-ma compete-horse “horse race”
14. xi-zao wash-bath “to bathe”
Wernicke’s responses
xi-che wash-car “to wash car”
xi-shui wash-water invented word
-shen -body incomplete word
Broca’s responses
xin-ku body-body “body” (Taiwanese)
xi-shen wash-body “to bathe”
xi-yi wash-clothes “to wash clothes”
15. tan-qin play-musical- “to play”
instrument
Wernicke’s responses
gang-qin steel-piano “piano”
gang-kai-shi just-open-start “just start”
tan-er-ji play-ear-machine invented word
yao-qin wave-piano invented word
feng-qin wind-piano “organ”
tan- play “to play”
Broca’s responses
gang-qin steel-piano “piano”
gang-qin steel-piano “piano”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
232
gang-qm
gang-qin
gang-qin
16. chi-fan
steel-piano
steel-piano
steel-piano
eat-rice
Wernicke’s responses
jia (Taiwanese) eat
Broca’s responses
fan rice
17. chao-cai fry-vegetable
Wernicke’s responses
cai
chao
yong-cai
cha (Taiwanese)
vegetable
stir-fry
use-vegetable
stir-fry
Broca’s responses
rou meat
huo-guo fire-pot
chao 6 7
18. zou-lu walk-road
Wernicke’s responses
xing-zou
zou-zou
zou
zou
walk-walk
walk-walk
walk
walk
Broca’s responses
zou-yi-zou
zou
walk-one-walk
walk
19. tiao-sheng jump-rope
Wernicke’s responses
tiao-shui jump-water
sheng rope
tiao jump
piano
piano
“piano”
“to eat”
“to eat”
“rice”
“to stir fry “
“vegetable”
“to stir fiy”
“do vegetables”
“to stir fry”
“meat”
“hot pot”
“to stir fiy”
“to walk”
“to walk”
“to take a walk”
“to walk”
“to walk”
“to take a walk”
“to walk”
“to jump rope”
“to dive”
“rope”
“to jump”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
233
tiao-san
tiao-gao shengzi
tiao sheng-zi
tiao-wu
Broca’s responses
tiao
qiao-qiao-ban
sheng-zi
tiao-qiu-qian
jump-umbrella “to parachute”
jump-high rope-suffix “high jump, rope”
20. zhao-xiang
jump rope-suffix
jump-dance
jump
up-up-board
rope-suffix
jump-swing-swing
shine-photo
to jump over a rope
“to dance”
“to jump”
“seesaw”
“rope”
invented word
Wernicke’s responses
zhao-pian shine-slice
xiang-pian image-slice
zhao-ren shine-man
jing-xiang enter-image
Broca’s responses
xiang-ji
zhao-xiang-ji
dian-ji
zhao-pian
image-machine
shine-image-machine
electric-machine
shine-slice
“to photo’
“photos”
“photos”
“take photos of man”
invented word
“camera”
“camera”
“engineer”
“photos”
21. tiao-shui jump-water “to dive”
Wernicke’s responses
tiao jump
tiao jump
tiao-gao jump-high
tiao jump
Broca’s responses
cha-shui
tiao-jin
hua-shui
penetrate-water
jump-enter
row-water
“ to jump”
“ to jump”
“ high jump”
“ to jump”
invented word
“ jump into”
“to swim”
22. hua-xue slide-snow ‘to ski’
Wernicke’s responses
hua-che slide-car
xue-chu snow-remove
invented word
invented word
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
234
qun-xin (chong-xue) rush-snow
hua-chuan row-boat
invented word
“to row a boat’
Broca’s responses
hua-shui
xue
slide-water
snow
‘ to surf”
‘snow”
23. kao-rou roast-meat ‘to roast*
Wernicke’s responses
rou-qi
tao(kao)
chi-rou
zhu-rou
meat-instrument
roast
eat-meat
pig-meat
invented word
“to roast”
“to eat meat”
“pork”
Broca’s responses
-rou meat “meat”
cai-huo wood-fire “firewood”
cao- stir-fry “to stir fry”
-rou meat “meat”
yan-rou smoke-meat “smoked meat”
ji-rou chicken-meat “chicken”
24. hua-zhuang put on-cosmetics “to make-up”
Wernicke’s responses
hui-hua paint-paint “painting”
lian face “face”
Broca’s responses
sai-hong cheek-red “brush”
huang cosmetics “cosmetics”
25. huai-yun bear-pregnancy “to be pregnant”
da-du-zi big-tummy-suffix “to be pregnant” (variant)
Wernicke’s responses
da-shui big-water “flood”
yun-yu pregnancy-raise “to raise”
-yun pregnancy incomplete word
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Broca’s responses
-yun pregnancy
-fu woman
da-pi-gu big-butt-butt
yun-fu-zhuang pregnancy-
woman-clothes
26. wo-shou hold-hand
Wernicke’s responses
hu-shou protect-hand
shou-lu watch-road
zhua-shou grasp-hand
Broca’s responses
Ia-shou hold-hand
ke-shou ?-hand
la-shou hold-hand
you-shou right-hand
27. you-yong swim-swim
Wernicke’s responses
you-shui swim-water
you-shou swim-hand
yong-che use-car
Broca’s responses
ti-yu body-raise
28. shuo-hua speak-speech
Wernicke’s responses
-hua speech
shuo speak
shu (shuo) speak
ting listen
Broca’s responses
-yan speech
wue (Taiwanese) speech
incomplete word)
incomplete word
“big butt”
“maternity dress”
“to shake hands”
“protector”
invented word
“to grasp”
“to hold hands”
invented word
to hold hands”
“right hands”
“to swim”
“to swim”
invented word
“to use a car”
“physical education”
“to speak”
incomplete word
“to speak”
“to speak”
“to listen”
incomplete word
“speech”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
236
29. ju-shou raise-hand
Wernicke’s responses
shou, zhengli qi-lai hand, right-stand
up-come
shou-zhi hand-point
Broca’s responses
shou hand
30. dian-huo light-fire
Wernicke’s responses
huo fire
shui-huo water-fire
ca-huo scratch-fire
da- hit-
Broca’s responses
huo fire
chai-huo wood-fire
chai-huo wood-fire
huo-chai fire-wood
dian-deng electric-lamp
huo-chai fire-wood
huo-zi fire-suffix
huo-zi fire-suffix
31. chou-yan smoke-cigar
Wernicke’s responses
xi-xiang- inhale-perfume
yan-wu smog-fog
na take
Broca’s responses
hun-tao cigar-head
yan smog
“to raise hand”
“hand raise up”
“finger”
“hand”
“to light”
“fire”
“water fire”
“to ”
“hit”
“fire”
“fire wood”
“fire wood”
“match”
“lamp”
“match”
invented word
invented word
“to smoke”
incomplete word
“smog”
“to take”
“cigar tips”
( t n rir , n
smog
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
32. an-Iing push-bell “to ring”
237
Wernicke’s responses
dian-ling electric-bell
da-dian hit-electric
ling-zai bell-suffix
kan look
Broca’s responses
ying (ling) bell
dian-ling electric-bell
meng-ling door-bell
dian-ling electric-bell
33. wan-yao bend-waist
Wernicke’s responses
wan bend
VN (Nominal)
1. suan-pan count-dish
Wernicke’s responses
zhu-suan
suan-zhong
zhu-sheng
suan-shu
pan-zi
bead-count
count-clock
bead-count
count-count
dish-suffix
Broca’s responses
bi-suan
2. fei-ji
pen-count
fly-machine
Wernicke’s responses
ji-che
fei-xing
fei-gong
fei-zi
machine-car
fly-walk
fly-work
fly-suflBx
“buzzer”
incomplete word
“bell”
“to look”
“bell”
“buzzer”
“bell bell”
“buzzer”
“to bow”
“to bend”
“abacus”
“abacus”
invented word
“abacus”
“arithmetic”
“dish”
“to count”
“airplane”
“motorcycle”
“to fly”
invented word
invented word
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
238
Broca’s responses
dao-ji big-machine invented word
-ji -machine incomplete word
-ji -machine incomplete word
fei-chuan fly-boat “airship”
3. qi-er stand-goose “penguin”
Wernicke’s responses
er-qi goose-stand reversed word
fei-ga fly-goose “fly-goose”
chang-er girl-girl “moon fairy”
qi- stand- incomplete word
Broca’s responses
-er
da-fei-er
fei-er
goose
big-fly-goose
fly-goose
4. tiao-qi jump-chess
Wernicke’s responses
qi-zi
qi-pan
chess-suffix
chess-dish
incomplete word
“big flying-goose”
“flying goose”
“Chinese checker”
‘chess”
‘chess board’
Broca’s responses
wei-qi
-qi
xiang-qi
qi-zi
xiang-qi
xia-qi
wei-qi
S. mo-bu
surround-chess
-chess
elephant-chess
chess-suffix
elephant-chess
down-chess
surround-chess
wipe-rag
Wernicke’s responses
bu-shan fabric-shirt
xi-cai-bu
cha-bu
mo-
wash-veggie-fabric
wipe-fabric
wipe
“Japanese chess”
incomplete word
“Chinese chess”
“chess”
“Chinese chess”
“to play chess”
“Japanese chess; go’
“wiper”
‘clothes”
invented word
“wiper”
incomplete word
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
239
Broca’s responses
-bu -fabric incomplete word
cha-bu wipe-fabric “wiper”
6. zheng-Iong steam-cage “steamer”
Wernicke’s responses
iong-zi cage-suffix “cage”
-long cage incomplete word
zheng-gao steam-cake “steamed cake”
zheng-bu-dao steam-not-arrive “could not steam”
Broca’s responses
deng-long lamp-cage “paper lamp”
-long cage incomplete word
san-long three-cage “three baskets”
7. yi-sheng cure-person “doctor”
Wernicke’s responses
yi-shi cure-master “doctor”
xiang-sheng first-person “Mr.”
-li principle incomplete word
yi-jia clothes-hang “hanger”
8. tou-shou pitch-hand “pitcher”
Wernicke’s responses
shou-ti hand-carry invented word
9. wan-ju play-instrument “toy”
Wernicke’s responses
ju-zi instrument-suffix invented word
shi-ju food-instrument “food-ware”
wan- play incomplete word
Broca’s responses
wan- play incomplete word
-ju instrument incomplete word
-mu -wood incomplete word
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
10. zhi-piao issue-ticket “check”
240
Wernicke’s responses
-piao -bill incomplete word
piao-zi bill-suffix “bill”
ya-zhi-piao process-pay-bill “to cash”
fa-piao deliver-bill “invoice”
zhi-mian pay-face invented word
Broca’s responses
fa-piao deliver-bill “invoice”
-piao -bill incomplete word
fa-piao deliver-bill “invoice”
fa-piao deliver-bill “invoice”
11. wo-shi lie-room “bedroom”
Wernicke’s responses
hu-shi protect-person “nurse”
fang-shi house-business “sexual intercourse”
wo-chuang lie-bed ‘to lie on bed”
Broca’s responses
fang -house incomplete word
12. zhao-pian shine-slice “photo”
Wernicke’s responses
zhao-zi shine-suffix invented word
zhao-xiang shine-image “to take picture”
ka-pian card-slice “card”
zhao-shui shine-water invented word
Broca’s responses
di-pian bottom-slice “film”
13. fa-piao distribute-ticket “invoice”
Wernicke’s responses
-piao -bill incomplete word
piao-zi bill-suffix “ticket”
fa-yin deliver-sound “to make sound”
fa-xin deliver-letter “to deliver letters”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
241
piao-zi
fa-mian
Broca’s responses
guo-piao
shou-piao
piao-zi
14. shui-yi
bill-suffix
grow-flour
?-bill
receive-bill
bill-suffix
sleep-clothes
Wernicke’s responses
han-yi
ha-yi
bu-yi
shui-zhao
sweat-clothes
sweat-clothes
fabric-clothes
sleep-complete
“bill”
“yeast grow’
invented word
invented word
“ticket”
“pajamas”
invented word
invented word
“clothes”
“fall asleep”
Broca’s responses
yi-fu clothes-clothes “clothes”
shui-jiao sleep-sleep “to sleep”
zhi-yi paper-clothes “paper clothes”
-yi -clothes incomplete word
yi-fu clothes-clothes “clothes”
15. y in-zhang print-seal “seal”
Wernicke’s responses
yin print “to stamp”
zhang chop “chop”
Broca’s responses
zhang chop “chop”
zhang-zi chop-suffix “chop”
16. wei-qun wrap-skirt “apron”
Wernicke’s responses
-qun -skirt incomplete word
yi- clothes incomplete word
wei-Iei surround-fold “to surround”
Broca’s responses
chu- kitchen- incomplete word
wei-jin surround-towel “scarf”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
242
17. nao-zhong alarm-clock ‘alarm clock’
Wernicke’s responses
zhong clock
dei-zhong ?-clock
long-zhong use-clock
shi-zhong hour-clock
“clock”
invented word
invented word
“clock”
Broca’s responses
zhong
zhong
zhong
zhong
zhong
shi-zhong
18. qing-tie
clock
clock
clock
clock
clock
hour-clock
Wernicke’s responses
xi-zi
qing-ke
joy-word
invite-guest
Broca’s responses
-tie -card
-tie -card
“clock”
“clock”
“clock”
“clock”
“clock”
“clock”
invite-card “invitation card”
“wedding words”
“ to host”
incomplete word
incomplete word
19. bei-zhen
put-needle “pin”
Wernicke’s responses
kou button
bei-hong don’t-comfort
bing- (Taiwanese) wear-
“to button”
“don’t cheat”
incomplete word
20. jian-dao cut-knife “scissors”
Wernicke’s responses
dao-zi knife-suffix “knife”
Broca’s responses
zhi-jia-dao point-nail-knife “nail cutter”
jian cut “ to cut”
jian cut “ to cut”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
21. tuo-xie pull-shoe “slippers”
243
Wernicke’s responses
tou-zi pull-suffix
hu-tuo protect-pull
xie-zi shoe-suffix
xie-zi shoe-suffix
xie-zi shoe-suffix
Broca’s responses
tuo-pi pull-skin
tuo-ba pull-handle
Iiang-xie cool-shoe
22. xi-guan suck-hose
Wernicke’s responses
shi-yong-guan make-use-hose
su- (Taiwanese) suck
Broca’s responses
su-jiao-guan moid-glue-hose
23. chao-fan fry-rice
Wernicke’s responses
fan rice
chao-mian fry-noodle
chi-fan eat-rice
zhong clock
mi-fan rice-rice
chi-fan eat-rice
Broca’s responses
fan rice
dan-chao-fan egg-fry-rice
fan rice
24. wei-qi surround-chess
Wernicke’s responses
-qi
xiang-qi
-chess
elephant-chess
invented word
invented word
“shoe’
“shoe’
“shoe’
“to peel skin”
“mop”
“sandal”
straw”
invented word
“to suck”
“plastic hose”
“fried rice”
“rice”
“fried noodle”
“to eat”
“clock”
“rice”
“to eat”
“rice”
“egg fried rice”
nee
“Japanese chess; go”
incomplete word
“Chinese chess”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
244
-qi -chess “chess”
xuan-zhan-qi choose-fight-chess invented word
mi-qi ?-chess invented word
lo-qi floor-ride reversal word
qi-lo ride-floor “pathway”
Broca’s responses
qi-lu chess-road “chess playing strategy”
-qi -chess incomplete word
xiang-qi elephant-chess “Chinese chess”
25. yu-gang bath-tub “bathtub”
Wernicke’s responses
yu-qi bathe-instrument invented word
yu-shi bathe-room “bath-room”
yu- bathe- incomplete word
Broca’s responses
yu-gang fish-tank “aquarium-tank”
pen basin “basin”
yu-shi bathe-room “bathroom”
jin-gang gold-tub “golden tub”
26. diao-deng hang-Iamp “art lamp”
Wernicke’s responses
da-deng hit-lamp “ light is on”
you-deng oil-lamp “oil lamp”
deng lamp “lamp”
dian-deng electrical-lamp “electrical lamp”
Broca’s responses
deng lamp “lamp”
deng lamp “lamp”
tai-deng counter-lamp “desk lamp”
27. hu-shi protect-stafT “nurse”
Wernicke’s responses
hu-tao protect-cover “glove”
hu-Ii protect-reason “the study of nurse”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
245
28. diao-chuang hang-bed
Wernicke’s responses
diao-chuan hang-boat
chaung-zi bed-suffix
diao-qi-lai hang-start-come
Broca’s responses
chuang bed
wang net
“hammock”
invented word
invented word
“to hang up”
‘bed”
“net”
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
246
REFERENCES
Altmann, G.T.M., & Steedman, M. 1988. Interaction with context during human
sentence processing. Cognition, 30, 191-138.
Baker, M. 1985. The Mirror principle and morphosyntactic explanation. Linguistic
Inquiry, 16, 3 73-415.
Balota, D., & Ferraro, F. 1993. A dissociation of frequency and regularity effects
in pronunciation performance across young adults, older adults, and individuals
with senile dementia of Alzheimer type. Journal o f Memory & Language, 32, 573-
592.
Bates, E., Chen, S., Tzeng, O., Li, P., & Opie, M. 1991. The noun-verb problem in
Chinese aphasia. Brain and Language, 41, 203-233.
Bates, E., Chen, S., Li, P., Opie, M., & Tzeng, O. 1993. Where is the boundary
between compounds and phrases in Chinese? A reply to Zhou et al. Brain and
Language, 45, 94-107.
Bates, E., Friederici, A , & Wulfeck, B. 1987a. Sentence comprehension in
aphasia: A cross-linguistic study. Brain and Language, 32, 16-97.
Bates, E., Friederici, A , & Wulfeck, B. 1987b. Grammatical morphology in
aphasia: Evidence from three languages. Cortex, 23, 545-574.
Bates, E., Friederici, A., Wulfeck, B., & Juarez, L. 1988. On the preservation of
word order in aphasia. Brain and Language, 33, 323-364.
Bates, E., & Goodman, J. 1996. On the inseparability o f grammar and lexicon:
Evidence from acquisition, aphasia and real-time processing. (Tech. Rep. No.
9603). Lo Jolla, CA: University o f California, San Diego, Center for Research in
Language.
Bates, E., Harris, C., Marchman, V., Wulfeck, B., & Kritchevsky, M. 1995.
Production of complex syntax in normal aging and Alzheimer’s Disease. Brain and
Language, 10,487-539.
Bates, E., & MacWhinney, B. 1989. Functionalism and the competition model. In
B. MacWhinney & E. Bates (Eds.), The cross-linguistic study o f sentence
processing. New York: Cambridge University press.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
247
Bates, E., Thai, D., & Marchman, V. 1991. Symbols and syntax: A Darwinian
approach to language development. In N. Krasnegor, D. Rumbaugh, R.
Schiefelbusch, & M. Studdert-Kennedy (Eds.), Biological and behavioral
determinants o f language development. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Bates, E., & Wulfeck, B. 1989a. Comparative aphasioloy: A cross-linguistic
approach to language breakdown. Aphasiology, 3(2), 111-142.
Bates, E., & Wulfeck, B. 1989b. Cross-linguistic studies of Aphasia. In B.
MacWhinney & E. Bates (Eds.), The cross-linguistic study o f sentence processing.
New York: Cambridge University press.
Bemdt, R., & Caramazza, A. 1980. A redefinition of the syndrome of Broca's
aphasia: Implication for a neurological model of language. Applied
Psycholinguistics, 1,225-278.
Bemdt, R., Haendiges, A., Mitchum, C., & Sandson, J. 1997a. Verb Retrieval in
Aphasia: 1 Characterizing single word impairments. Brain and Language, 56, 68-
106.
Bemdt, R., Haendiges, A., Mitchum, C., & Sandson, J. 1997b. Verb Retrieval in
Aphasia: 2 Relationship to sentence processing. Brain and Language, 56, 107-
137.
Bemdt, R., Mitchum, C., & Haendiges, A. 1996. Comprehension of reversible
sentences in “agrammatism”: A meta-analysis. Cognition, 58, 289-308.
Bemdt, R., & Zingeser, L. 1991. Grammatical class effects in word production:
Finding the locus of the deficit. Brain and Language, 41, 597-600.
Blumstein, S., Milberg, W., & Dworetsky, B. 1984. Phonological factors in lexical
access in aphasia. Abstracts of Academy o f Aphasia, Los Angeles.
Bradley, D. 1978. Computational distinction o f vocabulary type. Unpublished
doctoral dissertation, Cambridge, MA: MIT.
Bradley, D., Garret, M., Zurif, E. 1980. Syntactic deficits in Broca's aphasia. In D.
Caplan (Ed.), Biological studies o f mental process. Cambridge, MA: MIT Press.
Brown, W., Marsh, J., & Smith, J. 1976. Evoked potential waveform differences
produced by the perception of different meanings of an ambiguous word Journal
o f Electroencephalograph and Clinical Neurophysiology, 41, 113-123.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
248
Brown, W., Marsh, J., & Smith, J. 1979. Principal component analysis of ERP
differences related to the meaning of an ambiguous word. Journal o f
Electroencephalograph and Clinical Neurophysiology, 46, 706-714.
Butterworth, B. 1979. Hesitation and the production of verbal paraphasia and
neologisms in jargon aphasia. Brain and Language, 8, 133-161.
Butterworth, B. 1983. Lexical representation. In B. Butterworth (Ed.), Language
production: Development, writing, and other language processes. London:
Academic Press. Vol. 2.
Butterworth, B., & Howard, D. 1987. Paragrammatisms. Cognition, 26, 1-37.
Caramazza, A., & Bemdt, R. 1978. Semantic and syntactic processes in aphasia: A
review of literature. Psychological Bulletin, 85, 898-918.
Caramazza, A., & Bemdt, R. 1985. A multi-component view of agrammatic
Broca's aphasia. In M.-L. Kean (Ed.), Agrammatism. Orlando, FL: Academic
Press.
Caramazza, A., Bemdt, R., & Brownell, H. 1982. The semantic deficit hypothesis:
Perceptual parsing and object classification by aphasic patients. Brain and
Language 15, 161-189.
Caramazza, A., & Hillis, A. 1990. Where do semantic errors come from? Cortex,
26, 95-122.
Caramazza, A., & Hillis, A. 1991. Lexical organization of nouns and verbs in the
brain. Nature, 349, 788-790.
Caselli, C., Casadio, P., & Bates, E. 1996. A cross-linguistic study o f the
transition from first words to grammar. (Tech. Rep. No. 9701). Lo Jolla, CA:
University of California, San Diego, Center for Research in Language.
Chao, Y-R. 1968. A grammar o f spoken Chinese. Berkeley, CA: University of
California Press.
Chen, S. 1984. An investigation o f Speech in Chinese aphasia: With special
reference to four function words. Unpublished MA thesis. Taipei: Fu-Jen Catholic
University.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
249
Chen, S. 1990a. Morphological Problem in Chinese. Paper presented in The 2nd
North America Chinese Linguistics Conference. Philadelphia: University of
Pennsylvania.
Chen, S. 1990b. Bi-directional feeding relation between Morphology and Syntax:
Nominal VN compounds and relative clauses. Ms. Los Angeles, C A: University of
Southern California.
Chen, S. 1993. On grammatical deficits in Chinese aphasia. In F. R. Eckman (Ed.),
Confluence: Linguistics, L2 acquisition, and speech pathology. Amsterdam: John
Benjamins.
Chen, S., Andersen, E., Kempler, D, & Bates, E. 1992, October. On the structure
preservation in Chinese aphasia. Abstracts of Academy ofAphasia, Toronto,
Canada.
Chen, S., & Bates, E. in press. The dissociation between nouns and verbs in
Broca's and Wernicke's aphasia: Evidence from Chinese. Special issue on Chinese
aphasia, Aphasiology.
Chen, S., & Shi, D-X. 1992. On the feeding relation between syntax and
morphology: Evidence from Chinese nominal V-N compounds. In S. Wang & F-P.
Tsao (Eds ), Proceedings o f the Third International Symposium on Chinese
Languages and Linguistics, Hsinzhu.
Chen, S., Tzeng, O., & Bates, E. 1990. Sentence interpretation in Chinese aphasia.
In H. Burmeister & P. Rounds (Eds.), Variability in SecondIxtnguage
Acquisition: Proceedings o f the Tenth Meeting o f the Second Language Research
forum. Eugene, Oregon: University of Oregon.
Chen, S., Tzeng, O., & Bates, E. 1992. Word order preservation in Chinese
aphasia. Proceedings o f the First International Symposium on Sinology.
Singapore.
Cheng, C.-M. 1982. Analysis of present-day Mandarin. Journal o f Chinese
Linguistics, 10,281-358.
Chomsky, N. 1970. Remarks on nominalization. In R. Jacobs and P. Rosenbaum,
(Eds.), Readings in English transformational grammar. Waltham, MA: Ginn.
Colheart, M., Patterson, K., & Marshall, J. (Eds.). 1980. Deep dyslexia. London:
Routledge & Kegan Paul.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
250
Comrie, B. 1981. Language universals and linguistic typology. Chicago, WS:
University of Chicago Press.
Damasio, A. 1989. Time-locked multi-regional retroactivation: A system-level
proposal for the neural substrates of recall and recognition. Cognition, 33, 25-62.
Damasio, H., Grabowski, T., TraneL, D., Hichwa, R., & Damasio, R 1996. A
neural basis for lexical retrieval. Nature, 380,499-505.
Daniele, A., Giustolisi, L., Silveri, C., Colosimo, C., & Gainotti, G. 1994.
Evidence for a possible neuroanatomical basis for lexical processing of nouns and
verbs. Neuropsychologia, 32, 1325-1341.
Decety, J., Perani, D., Jeannerod, M., Bettinardi, V., Tadary, B., Woods, R.,
Mazziotta, J., & Fazio, F. 1994. Mapping motor representations with positron
emission tomograph. Nature, 371,600-602.
Dell, G., & Juliano, C. 1990. Connectionist approaches to the production o f
words. (Cognitive Science Tech. Rep. CS-91-05, Learning Series). Urban, E L :
University of Illinois, The Beckman Institute.
Di Sciullo, A., & Williams, E. 1987. On the Definition o f Word. Cambridge, MA:
MIT press.
Elman, J., & McClelland, J. 1986. Interactive processes in speech perception: The
TRACE model. In D. Rumelhart & J. McClelland (Eds.), Parallel distributed
processing: Explorations in the microstructure o f cognition. Cambridge, MA:
MIT Press.
Fab, N. 1984. Syntactic Affixation. Doctoral dissertation, Cambridge, MA: MIT,
Givon, T. 1984. Syntax: A functional-typological introduction. Amsterdam: John
Benjamins. Vol. 1 .
Gleason, J., Goodglass, H., Green, E., Ackerman, N., & Hyde, M. 1975. The
retrieval of syntax in Broca’s aphasia. Brain and Language, 2,451-471.
Glosser, G., Saykin, A., Sperling, M., & O’Connor, M. 1994. Dissociations
between object and action naming following anterior temporal lobetomy. Brain
and Language, 47, 546-548.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
251
Goodglass, H. 1968. Studies on the grammar of aphasics. In S. Rosenberg and J.
Koplin (Eds ), Development in appliedpsycholinguistic research. New York:
Macmillan.
Goodglass, H. 1976. Agrammatism. In H. Whitaker & H.A. Whitaker (Eds ),
Studies in neurolinguistics. New York: Academic Press. Vol. 1 .
Goodglass, H. 1993. Understanding aphasia. San Diego, CA: Academic Press.
Goodglass, H., & Berko, J. 1960. Agrammatism and inflectional morphology in
English. Journal o f Speech and Hearing Research, 3. 257-267.
Goodglass, H., & Hunt, J. 1958. Grammatical complexity and aphasic speech.
Word, 14, 197-207.
Goodglass, H., & Kaplan, H. 1983. Boston diagnostic aphasia examination.
Philadelphia, PA: Lea and Febiger, 2nd edition.
Gordon, B. 1983. Lexical access and lexical decision: Mechanisms of frequency
sensitivity. Journal o f verbal learning and verbal behavior, 22(1), 22-44.
Gordon, B. & Caramazza, A. 1982. Lexical decision for open- and closed-class
items: Failure to replicate differential frequency sensitivity. Brain and Language,
15, 143-160.
Grimshaw, J. 1990. Argument structure. Cambridge, MA: MIT Press.
Grosjean, F., & Itzler, J. 1984. Can semantic constraint reduce the role of word
frequency during spoken-word recognition? Bulletin o f the Psychonomic Society,
22, 180-182.
Hart, J.; Bemdt, R., & Caramazza, A. 1985. Category-specific naming deficit
following cerebral infarction. Nature, 316,459-460.
Hillis, A., & Caramazza, A. 1994. Lexical organization o f the brain. (Tech. Rep.
No. 43). Baltimore, MD: The Johns Hopkins University, Cognitive
Neuropsychology Laboratory.
Hinton, G., & Shallice, T. 1991. Lesioning a connectionist network: Investigations
of acquired dyslexia. Psychological Review, 98, 74-95.
Huang, C-T. J. 1983. Phrase structure, lexical integrity and Chinese compounds.
Journal o f the Chinese Language Teacher's Association, 19, 53-78.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
252
Huang, S-F. 1986. Chinese as a headless language. The Second International
Conference on Sinology, Japan.
Huang, S-F. 1992. Lexical innovation and semantic change. Ihe Third Symposium
on Chinese Languages and Linguistics, Hsinzhu.
Huang, S-F., & Hsieh, M-L. 1989. Acquisition and comprehension of Mandarin
compound morphology. The International Conference on Syntax Acquisition in
the Chinese Context. Hong Kong.
Institute of Language Teaching and Research. 1985. Xiandai Hanyu Pinlu Cidian
(A frequency dictionary o f Modern Chinese). Beijing: Beijing Language Institute
Press.
Joanette, Y., & Brownell, H. (Eds.). 1990. Discourse ability and brain damage:
Theoretical and empirical perspectives. New York: Springer-Verlag.
Kean, M-L. 1979. Agrammatism, a phonological deficit? Cognition, 7,69-83.
Lai, C. S. 1984. Speech errors as evidence fo r models o f speech production: A
case from Mandarin Chinese. Unpublished M. A. thesis. Taipei: Fu-Jen Catholic
University.
Lapointe, J. 1985. A theory of verb form use in the speech of agrammatic aphasics.
Brain and Language, 24, 100-155.
Ladefoged, P. 1975. A course in phonetics. New York: Harcourt Brace
Jovanovich.
Li, C., & Thompson, S. 1981. Mandarin Chinese: A functional reference to
Chinese grammar. Berkeley: University of California Press.
Li, P., Bates, E., & Liu, H. 1992. Cues as functional constraints on sentence
processing in Chinese. In H-C. Chen & O., Tzeng (Eds.). Language processing in
Chinese. Amsterdam: Elsevier Science Publisher.
Li, P., & Bates, E., & MacWhinney, B. 1993. Processing a language without
inflections: A reaction time study of sentence interpretation in Chinese. Journal of
Memory and Language, 32, 169-192.
Liu, H. 1996. Lexical access and differential processing in nouns and verbs in a
second language. Doctoral dissertation, University of California, San Diego.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
253
Liu, H., Bates, E., & Li, P. 1992. Sentence interpretation in bilingual speakers of
Chinese and English. Applied Psycholinguistics, 13,451-484.
Liu, Y-M., Zhuang, Z-R., & Wang, S-Z. 1975. Frequency count o f40,000
Chinese words. Taipei: Lucky Publisher.
Lu, S-C. (Ed.). 1984. Ci-hui (A Collection o f Chinese words). Taipei: Wen-hua
Tu-shu Gongsi (Culture Books Inc.).
Luria, A. 1962. Higher cortical Junctions in man. New York: Basic Books. 2nd.
Luria, A., & Tsvetkova, L. 1968. The mechanism of dynamic aphasia1 .
Foundation o f Language, 4, 296-307.
MacDonald, M., Pearlmutter, N., & Seidenberg, M. 1994. The lexical nature of
syntactic ambiguity resolution. Psychological review, 101(4), 676-703.
MacWhinney, B. 1987. The competition model. In B. MacWhinney (Ed.),
Mechanisms o f language acquisition. Hillsdale, NJ: Lawrence Erlbaum
Associates.
MacWhinney, B. 1991. In J. Miller (Ed.), Research on child language disorders:
A decade o f progress. Austin, TX: PRO-ED.
MacWhinney, B., & Bates, E. (Eds.). 1989. Cross-linguistic studies of sentence
processing. New York: Cambridge University Press.
MacWhinney, B., Bates, E., & KliegL, R. 1984. Cue validity and sentence
interpretation in English, German, and Italian. Journal of verbal learning and
verbal behavior, 23, 127-150.
MacWhinney, B., & Osman-Sagi, J. 1991. Inflectional marking in Hungarian
aphasics. Brain and Language, 41, 165-183.
MacWhinney, B., Osman-Sagi, J., & Slobin, D. 1991. Sentence comprehension in
aphasia in two clear case-marking languages. Brain and Language, 41,234-249.
MacWhinney, B., Pleh, C., & Bates, E. 1985. The development of sentence
interpretation in Hungarian. Cognitive Psychology, 17, 178-209.
Marchman, V. 1993. Constraints on plasticity in a connectionist model of the
English past tense. Journal o f Cognitive Neuroscience, 5,215-234.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
254
Marchman, V., Harris, C., Juarez, J., & Bates, E. 1992. Functional constraints on
the production o f complex syntax. (Tech. Rep. No. 9205). La Jolla, CA:
University of California, San Diego, Center for Research in Language.
Marslen-Wilson, W., & Tyler, L. 1980. The temporal structure of spoken language
understanding. Cognition, 8, 1-71.
Martin, A., Haxby, J., Lalonde, F., Wiggs, C., & Ungerleider, L. 1995. Discrete
cortical regions associated with knowledge of color and knowledge of action.
Science, 270, 102-105.
Martin, A., Wiggs, C., Ungerleider, L., & Haxby, J. 1996. Neural correlates of
category-specific knowledge. Nature, 379, 649-652.
Martin, N., Saffran, E., Dell, G., & Schwartz, M. 1991, October. On the origin of
paraphasic errors in deep dysphasia: Simulating error patterns in deep dyslexia.
Paper presented at the Deep Dyslexia Meeting, London.
Martin, N., Saffran, E., Dell, G., & Schwartz, M. 1994. Origins of paraphasia in
deep dyslexia: Testing the consequences of a decay impairment to an interactive
spreading activation model of lexical retrieval. Brain and Language, 47, 52-88.
McCarthy, R., & Warrington, E. 1985. Category-specific in an agrammatic patient:
The relative impairment of verb retrieval and comprehension. Neuropsychologia,
23, 709-727.
Miao, X., Chen, G., & Yang, H. 1986. The role of word order and animacy in
sentence comprehension in Chinese. In M. Zhu (Ed.), Studies in child language
development. Shanghai.
Miceli, G. & Caramazza, A. 1988. Dissociation of inflectional and derivational
morphology. Brain and Language, 35, 24-65.
Miceli, G., Mazzucchi, A., Menn, L., & Goodglass, H. 1983. Contrasting cases of
Italian agrammatic aphasia without comprehension disorder. Brain and Language,
19, 65-97.
Miceli, G., Silveri, C., Nocentini, U., & Caramazza, A. 1988. Patterns of
dissociation in comprehension and production of nouns and verbs. Aphasiology, 2,
351-358.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
255
Miceli, G., Silveri, M., Villa, G., & Caramazza, A. 1984. On the basis for
Agrammatics' difficulty in producing main verbs. Cortex, 20, 207-220.
Milberg, W., Blumstein, S., & Dworetzky, B. 1988. Phonological and lexical
access in aphasia. Brain and Language, 34, 279-293.
Miller, D. & Ellis, A. 1987. Speech and writing errors in ‘‘neologistic
jargonaphasia”: A lexical activation hypothesis. In M. Coltheart, G. Sartori, & R.
Job (Eds.), The cognitive neuropsychology o f language. Hillsdale, NJ: Lawrence
Erlbaum Associates.
Mitchum, C., & Bemdt, R. in press. Verb retrieval and sentence construction:
Effects of targeted intervention. In G. Humphreys and J. Riddoch (Eds.),
Cognitive neuropsychology and cognitive rehabilitation. Hillsdale, NJ: Lawrence
Erlbaum Associates.
Monsell, S. 1991. The nature and locus of word frequency effects in reading. In D.
Besner & G. Humphreys (Eds ), Basic processes in reading visual word
recognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Osman-Sagi, J. 1987. Action naming in Hungarian aphasic patients. In Abstracts of
the Second World Congress o f Neuroscience IBRO. [Neuroscience, Supplement to
Vol. 22, p. S509.].
Packard, J. 1990. Agrammatism in Chinese: A case study. In L. Menn & L. Obler
(Eds.), Agrammatic aphasia: Cross-language narrative sourcebook. Amsterdam:
John Benjamins.
Petersen, S., Posner, M., Fox, P., Mintun, M., & Raichle, M. 1988. Positron
emission tomographic studies of the cortical anatomy of single word processing.
Nature, 331, 585-589.
Petzesky, D. 1979. Russian morphology and lexical theory. Ms., Cambridge, MA:
MIT.
Pinker, S. 1989. Leamability and cognition: The acquisition o f argument
structure. Cambridge, MA: MIT Press
Plaut, D.C. 1994. Parallel distributed processing challenges the strongest
modularity hypothesis, not the locality assumption. Behavioral and Brain Science,
17, 77-78.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
256
Plaut, D C. in press. Double dissociation without modality: Evidence from
connectionist neuropsychology. Journal o f Clinical and Experimental
Neuropsychology.
Plunkett, K., & Marchman, V. 1991. U-shaped learning and frequency effects in a
multi-layered perception: Implications for child language acquisition. Cognition,
38,43-102.
Plunkett, K., & Marchman, V. 1993. From rote learning to system building:
Acquiring verb morphology in children and connectionist nets. Cognition, 48,21-
69.
Ren, X-L. 1980. Hanyu goucixue (Chinese morphology). Shanghai: The
Commercial Publisher.
Russell, B. 1985. Description. In A. Martinich (Ed.). The philosophy o f Language.
London: Oxford University Press.
Saffran, E., Schwartz, M., & Marin, O. 1980. The word order problem in
agrammatism II: Production. Brain and Language, 10,263-280.
Sapir, E. 1921. Language: An introduction to the study o f speech. New York:
Harcourt Bruce Jovanoich, Inc.
Schwartz, M., Saffran, E., Bloch, D., & Dell, G. 1994. Disordered speech
production in aphasic and normal speakers. Brain and Language, 47, 52-88.
Seidenberg, M. 1995. Visual word recognition: An overview. In J. Miller & P.
Eimas (Eds.), Speech, language, and communication. San Diego, CA: Academic
Press.
Seidenberg, M., McClelland, J., & Petterson, K. 1987, July. A distributed
developmental model of visual word recognition, naming and dyslexia. Paper
presented at the Experimental Psychological Society (UK.) Symposium on
Connectionism, Oxford.
Seidenberg, M., McClelland, J., & Patterson, K. 1989. A distributed
developmental model of visual word recognition and naming. Psychological
Review, 96, 523-568.
Selkirk, E. 1982. The syntax o f words. Cambridge, MA: MIT Press.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
257
Sereno, J., & Jongman, A. 1990. Phonological and form class relations in the
lexicon. Journal o f Psycholinguistic Research, 19(16), 387-404.
Sereno, J., & Jongman, A. 1995. Acoustic correlates of grammatical class.
Language and Speech, 38, 57-76.
Sereno, J., & Jongman, A. in press. Processing of English inflectional morphology.
Memory and Cognition.
Shapiro, L., & Levine, B. 1990. Verb processing during sentence comprehension
in aphasia. Brain and Language, 38,21-47.
Tardif, T. 1996. Nouns are not always learned before verbs: Evidence from
Mandarin speakers’ early vocabularies. Developmental Psychology, 32, 492-504.
Tardff T., Shatz, M., & Naigles, L. 1996. The influence of caregiver speech on
children’s use of nouns versus verbs: A comparison of English, Italian, and
Mandarin. Abstracts o f the With International Congress for the Study o f Child
Language. Istanbul.
Trueswell, J. 1996. The role of lexical frequency in syntactic ambiguity resolution.
Journal o f Memory and Language, 35, 566-585.
Tsao, F-P. 1979. A functional study o f topic in Chinese: The first step toward
discourse analysis. Taipei: Student Book Store.
Tsvetkova, L., & Glozman, J. 1975. A neurolinguistic analysis of expressive
agrammatism in different forms of aphasia. Linguistics, 154/155, 61-76.
Tzeng, O. in preparation. On-line processing o f Chinese compounds: Evidence
from lexical decision, manuscript in preparation, University of California,
Riverside.
Tzeng O., & Chen, S. 1988, October. Aphasia in Chinese. Symposium on Aphasia
in Non-European Languages. Abstracts o f Academy o f Aphasia, Montreal.
Tzeng, O., Chen, S., & Hung, D. 1991. The classifier problem in Chinese aphasia.
Brain and Language, 41, 184-202.
Wang, L. 1958. Hanyu shigao (History o f the Chinese language). Beijing: Kexue
Chubanshe (Science Publisher).
Wang, L. 1962. Gudai hanyu (Ancient Chinese). Beijing, China. 2 Vols.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
258
Wise, R., Chollet, F., Hadar, U., Friston, K., Hoffner, E., & Frackowiak, R. 1991.
Distribution of cortical neural networks involved in word comprehension and word
retrieval. Brain, 114, 1803-1817.
Zhou, X-L., Ostrin, R., & Tyler, L. 1993. The noun-verb problem and Chinese
aphasia: Comments on Bates et al. 1991. Brain and Language, 45, 86-93.
Zhu, D-X. 1980. Yufa jiangyi (Chinese syntax). Shanghai: The Commercial
Publisher.
Zingeser, L., & Bemdt, R. 1990. Retrieval of nouns and verbs in agrammatism and
anomia. Brain and Language, 39, 14-32.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
IMAGE EVALUATION
TEST TARGET (Q A -3 )
' /
V -
✓
/>
' / r
1 . 0
U ■ 2 .8
us
I 2- 5
■ » B32
|2 .2
H r
1 1 I u s
V s -
l.l L „
I I
II1,8
1 .2 5
I I
1 .4 1
1.6
" " I I I--r ~ ~ = " =
150mm
IIV M G E . I n c
1653 East Main Street
Rochester, NY 14609 USA
Phone: 716/482-0300
Fax: 716/288-5989
0 1993. Applied Image. Inc.. All Rights R eserved
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Frequency in sentence comprehension

PDF

Morphology and the lexicon: Exploring the semantics-phonology interface

PDF

The syntax of the Chinese BA-constructions and verb compounds: A morphosyntactic analysis

PDF

Genetic dissections of the cerebellar circuitry involved in classical eyeblink conditioning

PDF

Evaluating heaviness: Relative weight in the spoken production of heavy-NP shift

PDF

On the representation and licensing of Q and Q-dependents

PDF

In the event of focus

PDF

Language and context: toward a model of psycholinguistic competence

PDF

Effects of working memory and semantic impairments on speech in Alzheimer's disease.

PDF

Definiteness types in Spanish: A study of natural discourse

PDF

The Nine Songs: A reexamination of shamanism in ancient China

PDF

Metaphor and metonymy: A comparative study of Chinese and Western poetics

PDF

Language differentiation in early trilingual development: Evidence from a case study

PDF

Cathay revisited: The Chinese tradition in the poetry of Ezra Pound and Gary Snyder

PDF

Justus Georg Schottel and linguistic theory

PDF

Grotesque transformations of the mythical woman: From August Strindberg to Nissim Aloni

PDF

Emotions: Linguistic representation and cultural conceptualization

PDF

The dramatic technique of Cyril Tourneur

PDF

Satirical social criticism in the novels of John Dos Passos

PDF

Greek drama and the African world: A study of three African dramas in the light of Greek antecedents

Asset Metadata

Creator Chen, Sylvia Shengyun (author)

Core Title Intra-lexical noun-verb dissociations: Evidence from Chinese aphasia

Degree Doctor of Philosophy

Degree Program Linguistics

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag language, linguistics,OAI-PMH Harvest,psychology, cognitive,psychology, experimental

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-322287

Unique identifier UC11351032

Identifier 9835107.pdf (filename),usctheses-c17-322287 (legacy record id)

Legacy Identifier 9835107.pdf

Dmrecord 322287

Document Type Dissertation

Rights Chen, Sylvia Shengyun

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA