PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

Q$55E&

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/C|RCIDaleDue.indd-p.1

DIFFERENTIAL EFFECTS OF ATTENTION IN SECOND LANGUAGE
ACQUISITION OF VERB-NOUN COLLOCATIONS

By

Ye Fan

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Linguistics and Germanic, Slavic, Asian, and African Languages

2005

ABSTRACT

DIFFERENTIAL EFFECTS OF ATTENTION IN SECOND LANGUAGE
ACQUISITION OF VERB-NOUN COLLOCATIONS

By

Ye Fan

This study aims to investigate how the acquisition of a verb's collocational features is
influenced by different levels of attention and whether the effect of attention is mediated
by collocational complexity and proficiency level. Four levels of attention were studied:
(1) semantic processing: learners were only asked to understand passages with the target
collocations embedded; (2) memorizationfor recall: learners were instructed to memorize
the target collocations in the passages for a later recall test; (3) rule given: learners were
provided with the target collocational rules and studied how the rules applied to the
instances in the passages; (4) rule given plus negative evidence: learners were provided
with the target rules, studied how they applied to the instances in the passages; moreover,
they were informed of what were impossible noun collocates for the target verbs.

94 Mandarin speakers enrolled in English 2nd- and 4th-level courses in a university
in China participated in the research. The target learning items were four partially

artificial English verbs, which displayed two degrees of collocational complexity. Within

each proficiency level, participants were randomly assigned to one of the four attentional
conditions as specified above and received a three-day treatment. On the fourth day, all
the participants did a test that consisted of three parts: (1) determining the basic meaning
of the target verbs; (2) writing down as many noun collocates as possible for the target
verbs; (3) judging whether a sentence containing one of the target verbs was good or not.
The results indicated that, overall, learners in the two rule-oriented conditions (3 and
4) excelled in various parts of the test: recall of phrases that appeared in the passages.
production of new collocates for the target verbs, and judgment of bad collocations.
Learners under the memorization for recall condition demonstrated certain advantages in
storing old phrases, but not in other areas. The semantic processing condition turned out
to be the least efficient for learning L2 collocations. Moreover, it was found that negative
evidence in L2 collocations could help to reduce overgeneralization errors. This study
was not able to detect an interaction between attention and collocational complexity, but
an interaction between attention and proficiency level did emerge. The memorization for
recall and rule given plus negative evidence conditions were less effective with Level 2
leamers than with Level 4 learners. And this differential effect of attention was accounted

for in terms of leamers' processing features and capacity.

Dedicated to Yu

ACKNOWLEDGEMENTS

I want to thank many people who have contributed directly or indirectly to the
writing of this dissertation.

First and foremost, I am eternally grateful to my committee chair, Professor Susan
6355. Two years ago, when Professor Gass said that she was willing to be my committee
chair, I felt so excited and deeply honored. Today I feel not only honored but also
fortunate to be her student and have the opportunity to learn so much from her. Despite
her extremely busy schedule, she is always ready to guide me, help me, and listen to my
joys and worries. She gives me the freedom to think independently, challenges me to
think through the issues, and encourages me when I make progress. I will never forget the
hug she gave me after my first paper presentation at a conference. I was too nervous to let
Professor Gass listen to that presentation, but I hope in the future I will have more
achievements to show to her. Professor Gass, thank you so much! Your wisdom, humor
and enthusiasm have been, and will continue to be, a great source of inspiration and
support for me.

I also wish to express my wholehearted thankfulness to the other members serving

in my committee, Professor Barbara Abbott, Professor Yen-Hwei Lin and Professor

Charlene Polio.

When I just came to USA for my PhD study, perhaps because of many expected and
unexpected changes in my life, I felt somewhat confused and low-spirited. However,
these negative feelings were always replaced by joy and a strong desire for knowledge,
whenever I walked into Professor Abbott‘s semantics and pragmatics class. Throughout
my entire period of study at Michigan State University, I have benefited immensely from
her profound knowledge in linguistics, her kind smiles, and her encouragements.

Working with Professor Lin has also been a great pleasure for me. Inside her
phonology class, I learnt how to read and review others' works, how to defend my own
arguments, and what it means by being an outstanding researcher. Outside her class,
whatever questions I have, I can always seek very clear and inspiring answers from her.
She reads my papers carefully, teaches me how to improve my writing patiently, and
sends feedback quickly. Her way of teaching and researching sets up a good model for
me in my future career.

It is also so fortunate for me to have Professor Polio, a distinguished researcher in
second language acquisition (SLA), in my committee. I still have a very clear memory of
my first meeting with Professor Polio in her office. After asking about my major research

interests, she immediately helped me search for relevant authors and articles on the

vi

Internet and recommended good books to me. From that day on, her great insights into

SLA and her critical and constructive comments on my works have helped me make a lot

of progress. I want to thank her for being such a great support in my research.

This dissertation has also benefited markedly from Professor Debra Hardison, who

gave me very useful advice on a paper that is related to this dissertation, and Ms. Huimin

Tu, who provided me with expert assistance in statistical analyses. Acknowledgement

must also be made to all the students, teachers and friends who volunteered their time and

energy to participate in this study and help with the experiment. Without their unfailing

support, this project will not be possible.

Finally, I want to express deepest thanks to my family. My parents. my

parents—in-law and my sister have given me endless love and care during my PhD study. I

will never get to this point without the sacrifice they have made for me. I am also blessed

to have a lovely daughter, Qingyi Sun, who is always a great joy for me. Last but not

least, my eternal thanks go to my wonderful husband, Yu Sun, who let me pursue my own

dream, encouraged me with his foresight and sense of humor during the hard times. I am

sure the emails between him and me in the past several years will make one of the most

beautiful writings in the world.

vii

TABLE OF CONTENTS

LIST OFTABLEan
LIST OF FIGURES .............................................................................. xv
] INTRODUCTION ............................................................................. 1
1.1. Background ................................................................................ 1

1.2. Purpose and significance of this study ............................................... 3

2 LITERATUREREVIEW..................................; ................................. 6
2.1. The nature of collocations .............................................................. 6
2.1.1. Types of collocations ............................................................. 7

2.1.1.1. Grammatical and lexical collocations ................................ 7

2.1.1.2. Free collocations, restricted collocations and idioms. . . . . . ....8

2.1.1.3. Summary of the types of collocations ............................... 12

2.1.2. Collocations and rules ......................................................... 12
2.1.3. Significance of collocational knowledge .................................... 16

2.2. Basic features of attention ............................................................. 19
2.3. Acquisition of L1 collocations ........................................................ 22
2.3.1. Implicit learning ................................................................ 23
2.3.2. Implicit learning of L1 collocations ......................................... 27
2.3.2.1. Process of chunking .................................................... 27

2.3.2.2. Results of chunking .................................................... 31

2.3.2.3. Summary ................................................................ 33

2.4. Acquisition of L2 collocations ........................................................ 33
2.4.1. Problems in production .......... ' .............................................. 34
2.4.2. Attention in learning L2 collocations ....................................... 41

2.5. More issues about attention .......................................................... 47
2.5.1 Attention in learning L2 syntax and morphology ........................ 47

2.5.1.1. Attention at the level of noticing-

the Noticing Hypothesis ............................................... 47

2.5.1.2. Empirical evidence for the Noticing Hypothesis ................. 51

2.5.1.3. Attention at the level of rule understanding ....................... 54

viii

2.5.2. Attention in learning L2 word meaning .................................... 59
2.5.3. Attention and different aspects of language ............................... 65
2.5.3.1. Complexity of different language aspects .......................... 65
2.5.3.2 The interaction between attention and complexity. . . ..69
2.5.4. Attention and proficiency level .............................................. 76
2.5.5. The role of negative evidence ................................................. 81
2.5.6. Summary of the role of attention in SLA .................................. 86
THE PRESENT STUDY ................................................................... 88
3.1. Research questions ...................................................................... 88
3.2. Method .............................................................................. i ..... 93
3.2.1. Participants ...................................................................... 93
3.2.2. Linguistic items to be investigated .......................................... 95
3.2.3. Material and procedure ....................................................... 99
3.2.3.1. Treatment ............................................................... 100
3.2.3.2. Test and scoring ....................................................... 108
3.2.3.3. Exit questionnaire .................................................... 113
3.3. Research hypotheses .................................................................. 114
RESULTS .................................................................................... 120
4.]. Acquisition of the basic meaning of the target verbs ........................ 121
4.2. Production of collocates of the target verbs ................................... 121
4.2.1. Classification of the IL production: procedure and method ......... 122
4.2.2. An overall picture of the collocate production .......................... 127
4.2.2.1. Number of collocates ................................................ 128
4.2.2.2. Production of collocates for each target verb ................... 130
4.2.3. Production of old collocates ................................................ 131
4.2.4. Production of new collocates - overall comparison ................... 138
4.2.5. Production of new target-like collocates ................................. 142
4.2.6. Production of non-target-Iike collocates ................................. 149
4.2.7. Production of semi-target-like collocates ................................ 154
4.2.8. Summary of collocate production .......................................... 156
4.2.9. NS judgment of IL production ............................................. 158
4.3. Judgment of collocations of the target verbs .................................... 160
4.3.1. Judgment of good collocations .............................................. 160
4.3.2. Judgment of bad collocations ............................................... 164

4.4. Summary of the test results ......................................................... 170

4.5. Rule detection .......................................................................... 172
4.5.1. Semantic processing group .................................................. 172
4.5.2. Memorization for recall group .............................................. 173

4.6. Rule recall .............................................................................. 173

4.7. Three-day treatment .................................................................. 176

5 ANALYSIS AND DISCUSSION .......................................................... 179

5.1. Attention and acquisition of the basic meaning ................................. 180

5.2. Attention and production of collocates ............................................ 184
5.2.1. Production of old collocates ................................................ 184

5.2.1.1. Overall effects of attention .......................................... 184
5.2.1.2. The interaction between attention and proficiency level. . ...189
5.2.1.2.]. The memorization for recall group (Level 2) ............ 190
5.2.1.2.2. The rule given plus negative evidence group
(Level 2) ......................................................... 192
5.2.2.Production of new target-like collocates .................................. 194
5.2.2.1. Overall effects of attention .......................................... 194
5.2.2.2. The interaction between attention and proficiency level.....l96
5.2.3. Production of non-target-like collocates ................................. 199
5.2.3.1. The rule given group vs. semantic processing group ........... 200
5.2.3.2. The rule given plus negative evidence group vs.
rule given group ..................................................... 200
5.2.3.3. The memorization for recall group vs.
semantic processing group ........................................ 205
5.2.4.NS judgment of IL production ............................................... 207

5.3. Attention and judgment of collocations .......................................... 208
5.3.1. Judgment ofgood collocations..............................................208
5.3.2. Judgment of bad collocations ............................................... 209

5.4. The lack of interaction between attention and complexity ..................... 211

5.5. Summary ................................................................................ 215

APPENDICES ..................................................................................... 221
1 Appendix A Treatment exercises for semantic processing group
(Day 1- Day 3) ............................................................. 221

Appendix B Treatment exercises for memorization for recall group

(Day 1- Day 3) .............................................................. 233
Appendix C Treatment exercises for rule given group

(Day 1- Day 3) .............................................................. 236
Appendix D Treatment exercises for rule given plus

negative evidence group (Day 1- Day 3) ............................... 242
Appendix E Test .......................................................................... 250
BIBLIOGRAPHY ................................................................................ 253

xi

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 11

Table 12

Table 13

Table 14

Table 15

LIST OF TABLES

Collocational continuum .......................................................... 9
Participants information ........................................................ 95
Matching of the pseudo-forms with the real words ....................... 99
Distribution of target collocations in the reading passages ............ 101
Glossing for the target verbs ................................................. 102
Summary of negative evidence .............................................. 107
Daily treatment procedure (Day 1 to Day 3) .............................. 108
Sentences with a good collocation in the judgment test ................. 111
Sentences with a bad collocation in the judgment test .................. 112
A summary of research hypotheses ......................................... 119

Descriptive statistics of quantity of collocate production by
complexity * level * attention ................................................. 128

Post-hoe LSD test of the effect of attention on the number
of collocates ...................................................................... 130

Production of collocates for each target verb ............................. 131

Descriptive statistics of old collocates by complexity * level
* attention ........................................................................ 132

Production of old collocates for each target verb ........................ 133

xii

Table 16

Table 17

Table 18

Table 19

Table 20

Table 21

Table 22

Table 23

Table 24

Table 25

Table 26

Table 27

Post-hoe Scheffé test of the effect of attention on the
production of old collocates .................................................. 134

Post-hoe Scheffé test of the effect of attention on production
of old collocates (Level 2) ..................................................... 137

Post-hoe Scheffé test of the effect of attention on production
of old collocates (Level 4) ..................................................... 138

Descriptive statistics of new collocates by complexity * level
* attention ........................................................................ 139

Post-hoe Scheffé test of the effect of attention on production
of new collocates ................................................................ 141

Descriptive statistics of new target-like collocates by complexity *
level * attention ................................................................. 143

Post-hoc Scheffé test of the effect of attention on production of new
target-like collocates ........................................................... 145

Post-hoe LSD test of the effect of attention on production of new
target-like collocates (Level 2) ............................................... 147

Post-hoe Scheffé test of the effect of attention on production of
new-target—like collocates (Level 4) ......................................... 148

Descriptive statistics of non-target-like collocates by complexity *
level * attention ................................................................. 150

Post-hoe Scheffé test of the effect of attention on production of
non-target-like collocates ...................................................... 152

Descriptive statistics of semi-target-like collocates (l) .................. 155

xiii

Table 28

Table 29

Table 30

Table 31

Table 32

Table 33

Table 34

Table 35

Table 36

Table 37

Table 38

Table 39

Table 40

Table 41

Table 42

Table 43

Descriptive statistics of semi-target-like collocates (2) ................. 156

Production of each sub-type of collocates by level * attention. . . . . ...157
NS Judgment of new target-like collocates ................................ 158
NS Judgment of non-target-like collocates .................... ........... 159
NS Judgment of semi-target-like collocates (1) ........................... 159
NS Judgment of semi-target-like collocates (2) ........................... 159

Descriptive statistics of the judgment of good collocations by
complexity * level * attention ................................................ 161

Number of participants who incorrectly judged a good sentence.. . . 163

Descriptive statistics of the judgment of bad collocations by
complexity * level * attention ................................................. 167

Post-hoc Scheffé test of the effect of attention on the judgment of bad
collocations ....................................................................... 169

Summary of the test results ................................................... 171

Detection of rules by participants in memorization for recall

condition ......................................................................... 173
Descriptive statistics of rules that can be completely recalled. . . . . ...174
Descriptive statistics of rules that can be partially recalled ........... 175
Descriptive statistics of rules that cannot be recalled .................. 175

Non-understanding of negative evidence in rule given and
rule given plus negative evidence groups ................................... l 77

xiv

Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9

Figure 10

Figure 11
Figure 12
Figureis
Figure 14

Fig“ re 15

LIST OF FIGURES

Mean number of collocates produced by participants .................. 129
Production of old collocates (Level 2) ...................................... 135
Production of old collocates (Level 4) ...................................... 136
Production of new collocates ................................................. 139
Production of old vs. new collocates ........................................ 142
Production (if new target-like collocates (Level 2) ....................... 146
Production of new target-like collocates (Level 4). .- ..................... 146
Production of non-target-like collocates (Level 2) ....................... 153
Production of non-target-like collocates (Level 4) ........................ 153

Production of each sub-type of collocates by

level * attention ................................................................. 157
Judgment of good collocations (Level 2) .................................. 162
Judgment of good collocations (Level 4) ................................... 162
Judgment of bad collocations (Level 2) ..................................... 170
Judgment of bad collocations (Level 4) .................................... 170
Processing of the meaning of the target verbs ............................ 182

XV

Chapter 1
Introduction
1.1. Background

Acquisition of L2 vocabulary is a gradual and incremental process of learning
several different knowledge components. According to Nation (2001 ), there are nine
aspects of what is involved in knowing a word: spoken form, written form, word parts,
form and meaning, concept and referents, associations, grammatical functions,
collocations and constraints on use. Among these parts, one of the most notoriously
problematic for L2 learners is collocations, that is, knowledge about What words or types
of words occur with a particular word, for instance. the knowledge that the typical noun
collocates of the adjective underdeveloped are words like territories and areas (more
below on the term collocation).

Recognition of L2 leamers' collocational deficit can be traced back at least to the
19305. Palmer (1933) listed a selection of verb-noun combinations (e. g., ask a question,
do a favour, give trouble, have patience) and pointed out: "While these are fairly regular
they show the learner (what sooner or later in the course of his study he must come to
know) that this particular verb may be followed by this particular object Without such

information the learners tends to form such combinations by guess work or on the

analogy of his mother tongue, and we can imagine him coming up with such unusual

expressions as 'to make a question, to perform a favor, to do trouble, to keep patience. . ."'

(p. 8). A similar point was made by Pawley and Syder (1983): "It is a characteristic error

of the language learner to assume that an element in [a linguistic] expression may be

varied according to a phrase structure or transformational rule of some generality, when

in fact the variation (if any) is much more restricted"(p. 215).

Bahns and Bldaw (1993) used cloze and translation exercises to test the productive

knowledge of English verb-noun collocations among German EFL learners (e. g., keep a

diam; take a call, pay compliments). The obtained data showed that collocational

competence did not develop at the same rate as the knowledge of vocabulary in general.

Similar conclusion can also be drawn from the research by Amaud and Savignon (1997).

who compared the vocabulary knowledge of French advanced learners of English with

that of native speakers (NSs) of English through a word checklist test and a

multiple-choice test. The learners were found to be able to attain nativelike performance

with rare low-frequency words, but not with complex lexical units, which included noun

phrase (e.g., a damp squib), verb equivalent (e.g., make no bones about), adverbial or

prepositional phrase (e.g., at a pinch), etc.

Marton (197 7) looked at both production and comprehension data. It was reported

that although L2 learners of English could comprehend and translate certain phrases in
English, they could not thereafter produce them. Granger (1998) asked 56 French
speakers of English to choose, from a list of 15 adjectives in each case, the acceptable
collocates of 11 amplifiers (e.g., readily available, bitterly cold) and also to mark it with
an asterisk if they felt one adjective was more frequently associated with the amplifier
than all the others. Leamers' answers indicated that their sense of frequency in L2
collocations was weak and partly misguided.

To sum up, there has been general recognition of the problems facing learners in
achieving native-like selection of collocations. The problems, which are interrelated with
each other, mainly include: (1) overgeneralizing a word's possible collocates; (2)
discrepancy between comprehension and production; (3) poor or wrong sense of
frequency; (4) slow progress in this area.

1.2. Purpose and significance of this study

Why is the development of L2 collocational competence so difficult for learners?
How do learners acquire a word's collocational features? This study aims to provide some
answers by looking at the initial stages in learning L2 collocations and examining how
the learning is influenced by different levels of attention, one of the most important

factors determining what in the input will be noticed and become assimilated into the

interlanguage (IL) system (Gass, 1988a, 1997; Schmidt, 1990, 2001).

The target items in the current study are verb-noun phrases. This subset of collocation

is chosen because from a writer's point of view they represent the proposition core of a

fully formed sentence and additionally they can provide the closest point of contact with

other published studies on the acquisition of L2 collocations.

The research questions of this study can be briefly summarized as follows:

RQl: Which of the following attentional conditions leads to the best gains in learning a

verb's possible noun collocates (in all conditions, learners were asked to understand the

passages containing the target collocations): (l) attention directed towards irrelevant parts;

(2) instructed to memorize the target collocations for a later recall test; ( 3) provided with

the collocational rules underlying the target items and studying how the rules work in the

passages; (4) provided with the collocational rules underlying the target items, studying

how the rules work in the passages, and also provided with negative evidence (NE), that

is, what are impossible collocates for the target words?

RQ2: Is the effect of the four attentional conditions mediated by complexity of

collocational patterns?

RQ3: Does attention have differential effects on different proficiency levels?

The answers to these questions can contribute to second language acquisition (SLA)

research in two ways. First, previous research on learning L2 collocations has been
focusing on identifying the errors in IL production and analyzing the causes There have
been very few studies that directly examine the dynamic learning process, that is, how
leamers' knowledge of a particular word develops from complete unfamiliarity, through
recognition of the word and some idea of its meaning, to knowledge of its collocational
features. This study aims to fill in this gap and throw light on some of the unexplored
problems. Second, this study can enhance our understanding of the role of attention in
SLA. There are still controversy and uncertainties over which level of attention is
required for L2 learning and how attention interacts with language area and proficiency
level. Moreover, previous research on attention is limited to the learning of syntax,
morphology and basic word meaning. Since collocational knowledge is viewed as the
essence of all fluent and appropriate language use (e.g., N. Ellis, 2001; Pawley & Syder,
1983), we cannot have a thorough understanding of the role of attention in SLA without

exploring its effect in learning collocations.

CHAPTER 2
Literature review

In this chapter, I will first provide a description of the two crucial terms in the title
of this research, that is, collocation and attention. On the basis of this description, I will
then proceed to review previous literature on acquisition of L1 and L2 collocations, some
of which is related to the issue of attention. Although this study only focuses on
verb—noun phrases, my review about collocation and its acquisition is not limited to this
category, because I consider all types of collocations as essentially the same in the sense
that they are all about word sequences. Finally, I will look at the most widely investigated
and discussed issues concerning the role of attention in SLA, which have inspired the
current study. These issues include: (1) attention in learning L2 syntax and morphology;
( 2) attention in learning L2 word meaning; (3) attention and different aspects of language;
(4) attention and proficiency levels; (5) the role of NE.
2.1. The nature of collocations

Collocation is a term that has been used and understood in different ways. Some
researchers restricted it to an idiom whose meaning is not obvious from its components,
like blow the gafl (e. g., Palmer, 1933); some used it to refer to a phrase in which a word

is used in a specialized sense only found in the context of certain types of words, like

blow a fuse (e.g., Bahns & Eldaw, 1993). This study adopts a broad notion of collocation,
which is defined as “the co-occurrence of two or more lexical items as realizations of
structural elements within a given syntactic pattern” (Cowie, 1978. p.132). In other words,
the term collocation is used to describe any generally accepted grouping of words into
phrases or clauses, for example, compose music, begin to. as it were (see Howarth, 1998;
Moon, 1997; Nation. 2001 for a similar approach). In this broad sense, collocations can
occur in a variety of forms and with a variety of relationships between the words that
make up them. I will first review two major dimensions that have been proposed to
classify collocations, and then discuss the rules underlying the collocational phenomenon
and the significance of collocations in natural language use.
2.1.1. Types of collocations
2.1.1.1. Grammatical and lexical collocations

Following Benson, Benson, and Ilson (1986), collocations fall into two categories:
grammatical collocations and lexical collocations.

A grammatical collocation generally consists of a dominant open class word (noun,
adjective or verb) and a preposition or particular structural pattern such as an infinitive or
a clause. The major types of grammatical collocations are: (1) noun + preposition/ to

infinitive/ that clause (e. g., access to, pleasure to, agreement that...); (2) preposition +

noun (e. g., in advance, to somebody’s advantage); (3) adjective + preposition /to

infinitive/ that clause (e. g., aware of necessary to, afraid that...); (4) a verb combining in

different ways with a preposition, an infinitive with to, an infinitive without to, a verb

form in -ing, that clause, an interrogative word, etc. (e.g., adjust to, begin to. keep doing,

think that..., ask why... ).

A lexical collocation, on the other hand, normally does not contain infinitives or

clauses. It typically consists of open class words (noun, adj ective. verb or adverb). Six

major types have been identified by Benson et al.: (1) verb + noun/pronoun/prepositional

phrase (e. g., compose music, perform a task, draw up a bill); (2) adjective + noun (e. g..

slight decline, rough estimate); (3) noun + verb ( e. g.. The economy (is) booming. Bees

buzz); (4) noun + noun (e.g., a bunch offlowers, a swarm ofbees); (5) adverb + adjective

(e.g., deeply grateful. sound asleep); (6) verb + adverb (e.g., increase dramaticallr;

appreciate sincerely).

2.1.1.2. Free collocations, restricted collocations and idioms

Another way to classify collocations is to set up a continuum -- free collocation,

restricted collocation and idiom (figurative idiom and pure idiom), which is derived from

using criteria like restricted collocability, semantic specialization, and idiomaticity (see

Aisenstadt, 1981; Arnold, 1986; Howarth, 1998). It has to be noted that each of the

criteria is gradable, so there is no clear-cut borderline between the categories (Howarth,

1996). Table 1 presents some examples for each category.

Table 1 Collocational continuum

 

 

 

 

free collocation restricted collocation figurative idiom pure idiom
lexical make a cake, make a suggestion, blow your own have a chip
collocation forge a blade forge unity trumpet on one's
shoulder
grammatical under the desk, under attack, under the under the
collocation on the wall on the increase micro-scope weather

 

 

 

 

 

 

Free collocations, also referred to as open collocations or free combinations, are

made up of elements used in their literal sense. The constituent words are not bound

specifically to each other and they occur freely with many other semantically compatible

lexical items. For example, but a book (TV, piano. car... ), in the suitcase (room, house,

bo.\'...).

On the other end ofthe continuum is the most opaque and fixed set: figurative and

pure idioms. A figurative idiom has a metaphorical meaning in temis of the whole and a

current literal interpretation (e. g., blow your own trumpet). A pure idiom has a meaning

that cannot be determined from the meanings of the components and it does not have a

reasonable literal meaning (e. g., have a chip on one's shoulder).

In the middle of the collocational continuum category is restricted collocations,

which has one component, usually called a node word (Stubbs, 1995), used in a

specialized sense and accompanied only by certain types of collocates. Three major types
of cases can be identified for restricted collocations:

First, some words have a very narrow and specific meaning. so they occur almost
entirely in the company of one or two other words, or a narrow set of words. For instance,
the adjective blond normally occurs with the noun hair, or a noun referring to something
made of hair (e.g., locks, wig) or some entity with hair (e.g., student, doll). According to
Allerton (1984), this limitation of possible collocates is caused by the fact that blond has
a primary meaning of "light yellow-coloured" and also a secondary meaning of "as
applied to hair". In other words, it is part of the meaning of blond that it applies to hair. In
certain extreme cases, the secondary meaning ofa word is so limiting that in typical use,
it only displays one or two collocational patterns, for instance, torrential rain, cartterize
wound, shrug shoulders, curry favor:

Second, sometimes a word is used in its metaphoric sense, so its possible collocates
are limited to certain lexical items or sets rather than all the items that are semantically
compatible with the word's basic meaning. For example, the literal sense of the verb
germinate is "to cause a seed to start growing". In its metaphoric use, the notion drawn
from the concrete and tangible domain of plant (the source domain) is used to depict a

more abstract target domain (e. g., hope/mistrust/a good idea + germinate). Although

10

there are many other things that can start developing, the target domain in this case is
typically related to something in one's mind (see Collins Cobuild English Dictionary,
1995)

Third, when a word is used in its delexical sense, the range of its possible collocates
seems to be delimited for no good semantic reasons. Delexicalisation is the tendency of a
word, especially a verb, to lose its original meaning and become lexically "not full" when
it is used in conjunction with certain other words (Allerton, 1984; Carter & McCarthy,
1988, p153). For instance, in have a shower, the general meaning of the verb have, such
as "to possess, to receive, to hold", is dropped, and the meaning of the phrase falls upon
the noun shower. Verbs often considered delexicalised include give, have, make, take. do,
hold. keep and set, all of which are of high frequency. and the first four are the most
common ones used in this way (Sinclair & Fox, 1990). Because a delexical verb is
somewhat semantically bleached, sometimes there seems to be no obvious reason why
certain nouns are eligible as its collocates but some not. For instance, why can we say
give a smile/ cry/ hug/ promise, but not *give a look/ question? (There are more
discussions about this issue in 2.1.2.) Words of medium frequency may also be used in
this secondary and abstract sense, in which case again the range of the collocates is

determined by a level beyond semantic compatibility. For example, we may pay attention

/a call/ respects/compliments, but not *pay greeting/welcome.

In short, a restricted collocation has one component used in a special sense (a very

narrow, metaphorical, or delexical sense), which causes its collocates to be limited to

certain types of words or even several words.

2.1.1.3. Summary of the types of collocations

As demonstrated above, in its broad sense, collocations cover a very wide range of

phenomenon. This is best reflected in Nation's (2001) identification of three points along

a scale of collocation: "At one end we have idioms like a red herring, you're telling me

and be that as it may which are largely fossilized and opaque. In the middle we have

groups like take medicine, for exan'zple and little did .r know which allow some

substitution, are sometimes grammatically unique. are not necessarily adjacent and are at

least partially transparent. At the other end we have items like as a result. it is assumed

that, Where was I? which are grammatically well-formed, allow a lot of substitution and

grammatical change, and are transparent"(p. 335).

2.1.2. Collocations and rules

One level of restriction on the co-occurrence of words is the syntactic level, which

can be captured by rules. Then how about the other levels of restrictions? To what extent

are these restrictions governed by rules? In a paper on English multi-word items, Moon

12

(1997) claimed: "Just as a syntactic view of language observes rules underpinning
grammatically well-formed utterances, a collocationist view of language observes the
strong patterning in the co-occurrence of words. This itself can be seen to some extent as
rule-govemed and motivated (that is, it reflects some subliminal or underlying system or
process of analogy), however prolific the rules are and difficult to codify" (p.42). In
Doughty and Williams (1998), lexical item was categorised as form, but collocation was
categorized as rule like devoicing, agreement, etc. In this part, I am going to discuss, on
the basis of previous literature, to what extent collocations (free collocations, restricted
collocations and idioms) are governed by non—syntactic rules (simply called rule in the
following discussions). The examples that will be used are all verb-noun phrases, since
they are the items targeted in the current experiment, but the analysis also applies to other
types of collocations.

For free collocations, the rule underlying the combination is the mutual semantic
compatibility of its constituent words. As Allerton (1984) put it, " Individual words are
only combined when the ideas they express come together at some point in the physical
and/or mental experience of the speaker" (pp. 20- 21). For example, we may say that
resemble has no selectional restrictions whatsoever on its following noun, but indulge is

more restricted in that it requires a subject with the feature [+human], except in fairy tales

13

or other creative work.

For idioms, synchronically speaking, the word selection is purely arbitrary (Allerton,

1984). Although there are some attempts to categorize and describe them, the mismatch

between the actual meaning and the summed meanings of the constituents seems to be a

great challenge to the rule-based view of language (Cornell, 1999). In other words, due to

the semantic opaqueness and complete frozenness of an idiom, it is very hard to find any

systematic rule on this ostensibly unruly world.

As to restricted collocations, the situation is quite complex. Sometimes, the

collocates of the node word can be defined in terms of semantic feature or semantic field

(see Stubbs. 1995). For instance, in its metaphoric use, warp often takes a noun related to

one‘s mind: personality,judgment, etc; F orge, in its metaphoric use, often co-occurs with

nouns from the semantic field of “relationship":friendship. unity. bond, link, alliance, etc.

(see Collins C obuild English Dictionary, 1995).

However, there are three factors that make things more complicated.

First, a collocational rule might be probabilistic and there are some seemingly

arbitrary gaps in the patterns. For example, one might state that commit is often

accompanied by a noun with the semantic feature of “bad or illegal”, like crime, sin,

oflence, murder, suicide (Stubbs, 1995). However, it is odd to say *commit

l4

treachery/burglary/thefi, although these nouns also fit the “bad or illegal” category.

Second, for a collocationally active word, the rules underlying word co-occurrence
might be numerous and seemingly unrelated. For example, in its delexical use, take can
be collocated with nouns or noun phrases (I) that refer to physical actions (e.g., look, step,
shower: breath), (2) that refer to a time when you are not working (e. g., holiday break,
the day ofl), (3) that refer to an attitude or opinion that you have or are expressing (e. g.,
interest. offence, a foolish attitude). (4) that refer to formal declarations (e.g., oath, vow)...
(see Collins C obuild English Dictionary, 1995). Learners need to associate these
unrelated sets of nouns with the same node verb and also avoid overgeneralizations. This
is obviously a difficult task.

Third, on some occasions it seems hard to codify a rule. One example is the choice
of delexical verbs. As pointed out by Allerton (1984), the contrast among delexical verbs
is somewhat neutralized because they have lost their own distinctive meaning; as a result,
sometimes there seems to be no good reason why a particular delexical verb is preferred
to another and the choice is partly arbitrary. For example, why do we have to say put a
question, but give an answer? What is the rule regulating the choice between put and give?
Although it is usually possible to find a semantic justification for this preference (for

example, as suggested by Allerton, the answerer of a question is performing a service to

the questioner by answering, so give is preferred; whereas a question involves placing an
utterance in front of the listener for his/her attention, so put is used), "it would usually
also have been possible to devise a rationale for some other item" (Allerton, 1984, p.33).

In conclusion, at this point, we can only say that collocation is a system that is partly
governed by rules, which might be probabilistic and numerous. Sometimes the language
simply seems to dictate that such-and-such a combination is or is not acceptable, which is
at best only partly explicable in semantic terms.
2.1.3. Significance of collocational knowledge

The Chomskyan view of language posits that linguistic competence is basically an
autonomous syntactic component that is capable of generating all the grammatical
sentences of a language. However, with the rapid development of large corpora of texts
together with increasingly sophisticated text analytic software, it is becoming more and
more clear that NSs do no exercise the creative potential of syntactic rules to the full
extent, and words display strong clustering tendencies and intricate links with each other.
For instance, in the Bank of English, a collection of over 300 million words of written and
transcribed oral English texts, the adjective torrential collocates with rain in 99% of the
instances, and much less often with a semantically related word such as downpour or

storm (Moon, 1997).

16

Sinclair (1987, pp. 319-325; 1991, pp. 110—115), who directed the COBUILD
project, proposed two principles to account for the structural patterning of lexis: the open
choice principle and the idiom principle. The open choice principle assumes that
language is a result of many complex choices where the only constraint is syntactic
well-formedness. However, according to the idiom principle, language users have
available to them a large number of memorized or semi—preconstructed phrases that
constitute single choices and there are constraints and limitations beyond
grammaticalness. Sinclair further argued that most of the text is interpretable by the
idiom principle and it is the idiom principle that takes precedence. As to the open-choice
analysis, it could be "imagined as an analytic process which goes on in principle all the
time, but whose results are only intermittently called for" (1991, p. 114). In other words.
language use is not as open-class as is supposed with generative grammar.

Besides Sinclair, there are many other researchers who view collocational
knowledge as the basis of language use. Becker (1975) claimed. " We speak mostly by
stitching together swatches of text that we have heard before; productive processes have
the secondary role of adapting the old phrases to the new situations". Zemick and Dyer
(1987) proposed a "phrasal approach", which incorporates into the lexicon both

individual words and entire phrases like at noon/midnight/five o'clock, look/snifl/laugh at.

According to Pawley and Syder (1983), the best explanation for nativelike selection and
nativelike fluency (i.e., selecting the most appropriate way to express certain ideas from a
variety of possibilities created by syntax and producing language fluently) is that the
user's mind is stored with familiar collocations, which express a wide range of familiar
concepts and speech acts. By retrieving these collocations as wholes or as "automatic
chains" from the long-term memory (LTM), the user "minimizes the amount of
clause-intemal encoding work to be done and frees himself to attend to other tasks in
talk-exchange, including the planning of larger units of discourse" (p.192). Similarly,
Cowie (1994) stated that "native-like proficiency of a language depends crucially on
knowledge ofa stock of prefabricated units" (p. 3168), that is. units assembled to be
wholes.

To sum up, the collocational level of linguistic analysis shows that language
production is not a process solely governed by syntactic rules; instead, the storage of
collocations and the knowledge of the general sequential probabilities of words play a
significant role in appropriate and fluent language use. Then how is the complex
collocational system acquired? What is the role of attention in the process? I will now

move to a review of the second crucial term in the title of this research, that is, attention.

18

2.2. Basic features of attention

People are constantly exposed to large amounts of sensory and cognitive
information. It is our attentional mechanism that controls and reduces the input.
According to Sergeant (1996) and Sanders (1998), attention operates at three stages of
information processing: (1) taking in and processing auditory and visual stimuli; (2)
executing central control and making decision, for example, allocating attention to
competing task demands; (3) making response and monitoring through sustained
attention. These stages correspond to three uses of the concept of attention: (1) attention
as selection of the information to be processed and stored in memory; (2) attention as
capacity for dealing with information; (3) attention as efforts or energy made to process
input and produce output (Robinson. 2003).

There are three critical features of attention that are most relevant to this study:

First, attention is a limited mental resource. Kahneman (1973) described attention as
a single finite pool of resources, in which information competes for limited resources.
Wickens (1984, 1989, 1992) expanded Kahneman's view and proposed multiple, separate
pools of resources (e. g., visual, auditory, vocal, manual), which can be allocated to tasks
somewhat independently of one another. According to the multiple-source approach, it is

easier to simultaneously carry out attention-demanding activities of different modalities

than those of the same modality. In other words, there is some flexibility to the capacity
limitations of attention and resources compete within but not between separate attentional
pools.

Second, attention is selective. Broadbent (1958, 1971) assumed that specific stimuli
must be selected from the incoming information early because attention is a system of
limited capacity. However, under the multipleosource model of attention, selection is a
result of meeting task demands rather than capacity limitation and it takes place after
stimuli have been fully analysed (Allport, 1987).

Third, attention is not a unitary phenomenon. Farah (1992) posited that the
activation of attention has a graded property and information represented in neural
networks may be "partially represented" (p.40). C owan (1995) also claimed that we must
theoretically be able to speak of the possibility of graded amounts of awareness and
attention (p. 200). Posner and Petersen ( 1990) decomposed the human attentional system
into three separate yet interrelated networks: alertness, orientation, and detection.
Alertness is general readiness to deal with incoming information; orientation is to allocate
attention towards some types of sensory stimuli while excluding others; detection is the
process of selecting or cognitively registering a specific bit of information with or

without awareness. According to Posner and Petersen, alertness can work independently

20

or modulate orientation, and orientation makes detection more likely but does not
guarantee it. This analysis of attention implies that the activation of attention is a
continuum, which can range from alertness to orientation, to detection outside awareness
and to detection with awareness. Robinson (2003) described the following stages of
information processing, which differentiated peripheral attention with focal attention.
First, sensory information is detected through peripheral attention and held temporarily in
the sensory register. Then, certain stimuli are further selected by focal attention, where
two types of rehearsal processes operate: maintenance rehearsal and elaborative rehearsal.
Maintenance rehearsal refers to cycling certain representations in short—term memory
(STM) through a “phonological loop” (Baddeley. 1986) and it can lead to the storage of
the representations in an unanalyzed form in LTM. Elaborative rehearsal is a more
effortful learning process, in which the learner may attempt to find a rule by consciously
retrieving old instances and comparing them with the present input, or test a previously
formed hypothesis.

In the process of learning L2, learners are exposed to influx of input, and it is
attention that sorts out the information and brings order to it. According to cognitive
accounts of SLA, L2 development is mostly driven by what learners pay attention to in

the input. The allocation of attention is claimed to be a necessary condition for selecting

21

items from the input, keeping it active in working and STM, encoding it into LTM and
also retrieving it from LTM (e.g., Long, 1996; Schmidt, 1994). It also allows leamers to
notice a gap between their IL and the target language (TL), or a gap between what they
can produce and what they need to produce, both of which are considered as crucial for
triggering changes in the IL system (Gass, 1988a, 1997).

Although it is generally agreed that some kind of attentional process is required for
input to initiate development in the IL system, opinions vary as to the amount and type of
attention that are necessary for the acquisition of certain L2 items. Which level of
attention is necessary for acquisition to take place. focal attention or detection that
requires minimally low levels of peripheral attention? What are the differential effects of
attention on different language items and on learner proficiency levels? These questions
have stimulated a lot of debate and discussions in the field of SLA. mostly focusing on
the area of learning L2 syntax and morphology. Before reviewing these discussions,
which have motivated the research questions of the current study. l will first provide an
overview of previous research on how individuals acquire L1 and L2 collocations, some
of which is related to the issue of attention.

2.3. Acquisition of L1 collocations

As suggested in 2.1, text and corpus studies have revealed the significance of

22

collocations as basic, intermediary units between the levels of lexis and grammar. There
are certain collocational constraints that learners must adhere to. How do people develop
the collocational competence in L1 ? According to N. Ellis (1997), this is predominantly
an automatic and implicit process of analyzing word sequence and frequency information,
like learning phonotactics. This section will first consider some of the basic issues about
implicit learning and then look at how the mechanisms of implicit learning operate in L1
collocation acquisition.

2.3.1. Implicit learning

The theory of implicit learning is largely formulated on the basis of three types of
empirical experiments within the field of cognitive psychology.

The first type is the learning of artificial grammars. Subjects were exposed to strings
of letters that were generated by a grammar or rule system (e. g.. MXRMXT. VMTRRR)
and instructed to memorise the examples for a later memory test. It was found that after
the treatment subjects could make correct judgment about whether NEW strings
conformed to the underlying grammar or not at above chance levels although they were
generally unable to verbalise the corresponding rules (e.g., Reber, 1967, 1976; Reber,
Kassin, Lewis & Cantor, 1980).

The second type of experiments is sequence learning in which subjects were exposed

23

to long symbol strings that displayed recurrent patterns and then asked to predict future
sequence. It turned out that subjects were sensitive to the patterns and could make right
prediction, but again they could not verbalise the patterns (e. g., Cleeremans & Jimenez,
1998; Curran & Keele, 1993).

The third type is control of complex system on computer. Subjects were asked to
control an output variable (e. g., production of a simulated factory) by operating input
variables (e. g., amount of raw material). Again, they could do this successfully without
conscious knowledge of the formula used to relate input variables to output (e. g., Berry &
Broadbent, 1984; Dienes & Fahey. 1995).

In short, according to Reber (1993). Winter and Reber (1997). in these types of
experiments, complex knowledge is acquired mostly without conscious operations and
without explicit knowledge of both the process of acquisition and the knowledge that is
acquired, and this process is labeled as implicit learning.

There are two issues about implicit learning that have been widely discussed:

The first issue is the process. In implicit learning, the structural nature of the input is
mastered naturally, simply and without conscious Operation (N. Ellis, 1994a). This
process contrasts with explicit learning, in which the individual pays attention to

particular aspects of the stimulus, consciously forms and tests hypotheses in order to

24

search for certain structures (N. Ellis, 19943). Some researchers (e.g., R. Ellis, 1994a, p.
i 390; McLaughlin, Rossman, & McLeod, 1983, p. 142) suggested that the distinction
between implicit learning and explicit learning is that of peripheral and focal attention.
That is, the former involves mere exposure to the rule-govemed input with no focused
attention to the rules per se; whereas the latter requires focal attention to the regularities
underlying the surface stimuli.

The second issue about implicit learning is the nature of the knowledge acquired in
this way. Reber (1989) concluded that what is acquired in implicit learning is abstract,
rule-based knowledge, on the grounds that the acquired information exceeds what can be
verbally expressed and can be applied to stimuli that are not presented during the
treatment stage. However, some researchers suggested that the results of implicit learning
are knowledge based on memory for fragmentary instances, or a combination of both
memory-based and rule-based knowledge (e.g., Dulany, Carlson, & Dewey, 1984; Vokey
& Brooks, 1992). Dulany et al. (1984) quoted the artificial grammar learning experiment
by Reber and Allen (1978) and showed that during the retrospection phase some subjects
mentioned that "first and last letters, bi grams, the occasional tri gram, and recursions"
(p202) were important in their grammaticality judgment. Dulany et a1. further argued that

subjects had explicitly remembered fragments of strings and their judgment was based on

25

similarity of instances rather than unconscious representations of a formal grammar.
Vokey and Brooks (1992) reported that resemblance between the instances used in
training and those in transfer test affected test judgment more than rule conformity. They
suggested that dual knowledge and rule and instance models of representation (Knowlton
& Squire, 1996; Logan, 1988) could explain this phenomenon. In other words, the effects
of implicit learning are attributable to a synergy of both abstract rule-based knowledge
and concrete instance-based knowledge. This view was adopted by several other
researchers (e.g., Mathews, Buss, Stanley, Blanchard-Fields, Cho, & Druhan, 1989,
Shanks & St. John, 1994). Moreover, it was argued that processingdemands of tasks can
predispose learners to rely more heavily on one type of knowledge than on the other.
When subjects had seen few example strings. they tended to memorise them and use
them as the standard against which to make judgments about the grammaticality of new
strings; but when they had seen many exemplars, subjects would induce more abstract
representation (McAndrews & Moscovitch, 1985; Meulemans & Van der Linden, 1997).
To sum up, experiments in cognitive psychology showed that people can acquire the
underlying rule of complex stimulus without paying focused attention to the rule per se at
the point of learning. The product of this implicit learning process is still a controversial

issue, but many researchers maintained that it is a blend of rule-based and instance-based

26

knowledge.

2.3.2. Implicit learning of L1 collocations

As previously mentioned, N. Ellis (1997) viewed acquisition of collocations as

predominantly an implicit process, that is, we develop collocational competence through

meaning-focused receptive and productive language use rather than overtly exploring any

underlying collocational patterns. According to N. Ellis, as long as individuals are

exposed to and also attend a sufficient amount of speech stream, analysis of the frequency

and sequence information will take place automatically. In this part, I will review some of

the research related to the details in this implicit learning process.

2.3.2.1. Process of chunking

There is one mechanism claimed to play a central role in acquisition of collocations,

that is, chunking (N. Ellis, 1996, 1997). The term chunking was coined by George Miller

in his review of STM (Miller, 1956). Miller distinguished "bits" ofinformation from

"chunks" of information. Bits of information can be weld together and formed into a

chunk, which itself becomes a unit of memory organization. Chunks that have already

been formed in memory can be further brought together and develop into a larger chunk.

Chunking implies the ability to build up such structures recursively and develop

permanent sets of associative connections in LTM (N. Ellis, 1996; Newell, 1990, p. 7).

27

Chunking is a pervasive feature of human memory and people chunk at a constant
rate (Newell, 1990, p. 7), especially when the stimuli elements are frequently observed
occurring together (Nation, 2001, p. 320). In the language we hear or read, there are
many recurring phrases or clauses. On the one hand, the more certain units are repeated
in STM, the greater the LTM for these items (see Melton, 1963). On the other hand, our
phonological LTM is automatically tuned to the regular sequence, which allows these
recurring patterns to be more readily perceived from the input. As N. Ellis (1997) put it,
"...our phonological memory systems automatically. and often unconsciously, abstract
regularities or patterned "chunks" from the collective evidence of the stream of speech to
which we are exposed" (p. 124).

Many empirical studies of language acquisition (e.g., Clark. 1974; Nattinger &
DeCarrico, 1992; Nelson, 1973; Peters. 1983) have shown that attended speech is
chunked automatically by L1 learners. Children use a large number of unanalyzed or
partly analyzed chunks of language as if they form a single unsegmented unit before they
learn the meanings of the individual words (e.g., "How are you", "Give me", "go on").
This special sub-group of collocations has been named as lexical phrases (Nattinger &

DeCarrico, 1992), holophrases (Corder, 1973), formulaic speech (Wong-Fillmore, 1976),

and lexical chunks (Lewis, 1993). Hakuta (1976) and Krashen and Scarcella (1978)

28

Lav-gr
“nub.

"J’W‘r'
unfit»

't
.u-
,7 '7

r urn

.\.. 1
w" ~

,t.\-. ,

ll” l l

l
5 in."
r...
r
n' W
1., b
r
I, , '
.1 \r‘
‘ H;
r
'—
- . 1
up. ,.
-
t t
‘.
'0
‘v
‘ V\\|k

further divided these chunks into routines and patterns. Routines are utterances learned as
memorized and unanalysed chunks, such as at any rate, by and large, and as it were.
Patterns are utterances that have been partially analyzed and possess one or more open
slots, such as _ sorry to _, and __ago.

In a study of 18 children producing their first fifty words, Nelson (1973) found a
sharp contrast between the so-called referential and expressive children. The expressive
group produced 6 to 18 phrases during the research period, while the referential group
produced only 0 to 5 phrases. These speech units include "expressions useful for dealing
with people" (e. g., go away, stop it, do it. I want it) and "may be considered as preformed
or stereotyped units... which can be combined into larger units but are not themselves
analyzable. Thus, they are equivalent to words in their linguistic function"(p.25). Clark
(1974) found that her son Adam, at age 2;9 to 3;0, often incorporated a part ofa
preceding adult utterance into his own utterance. For example.

Mother: That's upside down.

Child: No, I want to upside down.

Bolinger (1976) claimed that "learning goes on constantly - but especially with young
children - in segments of collocation size as much as it does in segments of word size. . ."

(p8). Pawley and Syder (1983) proposed that most words are stored for multiple times,

29

once as an individual word and numerous times as a part of a larger stored chunk, for
example, a phrase, a clause or even a unit of longer length.

Studies of word association (asking individuals to say what is the first word that
comes to their mind when they hear a stimulus word) also indicate that in L1 acquisition
there is a period when a word's collocation is most prominently represented in a person's
mind. One of the most extensive studies of L1 leamers' word association is that of
Entwisle (1966). The study found that (1) clang associates (i.e., giving responses that are
phonologically similar to the stimulus words, e.g., bitter 9 litter, present 9 pleasant)
were rare and most frequently found among first-grade children; (2) the major difference
between younger and older children lay in the shift from syntagrnatic response (e.g.,
bitter 9 taste, on 9 top) to paradigmatic responses (e.g.. bitter 9 sweet, on 9 off). It is
generally assumed that salient representations oflexical items (phonology, semantics,
syntax, etc.) tend to drive word association responses (N. Ellis, 1996), so Entwisle's study
indicates that in early stage of L1 acquisition, the phonological features of lexical items
are most strongly represented in one's mind, then there is a focus on a word's collocations
(most probably as a result of chunking at the level of phrases), and finally syntactic and

semantic classifications begin to emerge.

30

2.3.2.2. Results of chunking

The language input that is implicitly chunked by individuals can develop in three
directions, as identified in previous literature:

First, it might remain to be "unanalyzed wholes" and function as integrated wholes.
During online processing, these chunks can be accessed quickly and contribute to the
development of fluency, because they do not require a lot of internal computation.

Second, memorized smaller unanalyzed chunks might be combined with other
chunks to form into larger chunks (Nation. 2001. pp. 319-320), which can also be
mobilized in a particular communicative context and enable fluency to be achieved.

Third, acquired chunks may be analyzed later. As N. Ellis (1996. 1997) argued, the
analysis may lead to knowledge of the statistical frequencies and sequential probabilities
of the words. This process and the resultant knowledge is implicit in the sense that the
learner does not consciously analyze the sequence infomration and he/she knows which
word goes with which "by feel" rather than any explicit rule.

Moreover, some researchers suggested that the analysis of chunks may also lead to
the creation of novel constructions and acquisition of syntactic features or rules. In a
study of the development of inflection in the Harvard children, Cazden (1968) wrote that

"the child begins to operate with stored fragments of speech he has heard... which are

31

somehow tagged liberally for semantic information on the verbal and nonverbal context
and only later are gradually subjected to analysis for the acquisition of productive
rules"(p.437). N. Ellis (1997) argued that through automatic implicit analysis of a word's
sequential position relative to other words in the phrases that have already been stored,
children can learn the grammatical word class of a word and grammatical structures.
Skehan (1998, p. 56) suggested that chunk analysis may generate rules and the rules can
be further used to create other exemplars. which are then available for access as integral
wholes. Peters (1983) provided a detailed description of how unanalyzed chunks can
eventually become analysed into formulaic frames with slots and then into the syntactic
patterns of the language. For instance. a child is able to accumulate a number of phrases
that begin with the same unit all: all clean, all done, all (It); all gone, all through, all wet.
Then he/she might make an abstraction from "unit+unit" to "unit+list" (i.e., all + clean,
done, din; gone, through. wet). The next move is the generalization from "unit+ closed
list" to "unit +open class" by detecting the common feature of the words in the closed list
(p.45). Once this productive pattern is achieved, the child is able to create novel
Combinations. Peters regarded this step as the most crucial abstraction in the early
acquisition of syntax. Similarly, Nattinger and DeCam'co (1992, p.12) suggested that

initially a child may frequently use a phrase like I-want-to-go as if it were a single unit.

32

 

 

Gradually, when he/she encounters other chunks with similar patterns (e.g.,
I-want—to-get-up, I -want-my-ball), he/she may detach the pattern from the chunks and
acquire regular syntactic rules.
2.3.2.3. Summary

On the basis of previous research, we may conclude that acquisition of L1
collocations is an implicit process. People acquire a word's collocational features through
mere exposure to numerous instances, without consciously searching for or paying focal
attention to the formal collocational regularities. In this process, attended speech is
automatically chunked by L1 learners. The chunked units and long-term knowledge base
of word sequence can serve as memorized fonnulas that can be retrieved fast in language
use and also be used as raw material for later analysis that leads to very abstract
knowledge in frequencies and sequential probabilities of the words. Moreover, as argued
by some researchers. syntactic patterns can emerge from the chunk analysis. This account
0f collocation acquisition corresponds to the view that knowledge acquired through
implicit learning is both exemplar-based and rule-based (see the discussion in 2.3.1).
2.4. Acquisition of L2 collocations

Although acquisition of L1 collocations seems to be quite "effortless", acquisition of

L2 collocations appears to be a much more difficult process. It is by no means easy for a

33

non-native speaker (NNS) to distinguish which of the grammatically well-formed L2
sentences are nativelike, normal or unmarked and which are unnatural or highly marked.
As noted by Gass and Selinker (2001), "The problem for the learner is to learn how not to
be innovative and stick to the standard combinations" (p392). This section is going to
review previous research on acquisition of L2 collocations, which is concerned with ( l)
the problems in the collocations produced by L2 learners; ( 2) the process of leaming L2
collocations, especially the role of attention in it.
2.4.1. Problems in production

A major theme ofthe research on L2 collocations was to identify the problems in
leamers' production and classify their causes. The data often came from corpora of
leamers' argumentative or narrative writing (e. g. C hi, Wong & Wong, 1994; Granger.
1998; Lennon. 1996; Lorenz, 1999; Nesselhauf, 2003) or was elicited from experimental
tasks like Ll-L2 translations (e.g., Bahns & Eldaw, 1993; Farghal & Obiedat, 1995;
Gabrys-Biskup, 1992) and cloze tests (e. g., Bahns & Eldaw, 1993; Farghal & Obiedat,
1995). There are three important findings in these studies:

First, many problems in L2 collocation production originate from L1 influence.
When learners learn a word for a concept, they often assume that they can extend the use

of that word to the same contexts as its corresponding item in L1 and this literal

34

translation process may cause errors when L1 and L2 do not match (Bahns, 1993; Chi,
Wong & Wong, 1994; Farghal & Obiedat, 1995; Gabrys-Biskup, 1992). For example,
*drive a bookshop (transfer from Polish kierowac sklepem) (Gabrys-Biskup, 1992); *do
exchange (transfer from Chinese zuojiaoyi) (Chi, Wong & Wong, 1994). In addition to
literal translation, Gabrys-Biskup (1992) found that assumed formal similarity between
L1 and L2 could account for some of the errors made by German learners of English. For
instance, crack nuts was rendered as crunch nuts or crunk nuts, an attempt at anglifying
the German lexeme knacken.

L1 influence is also reflected in the overuse and underuse of certain items. Granger
(1998) reported on a study of a subcorpus of the 1C LE ( lntemational Corpus of Learner
English) database and a NS corpus of English. The comparison of the two corpora
showed that L2 collocations with a direct translation equivalent in L1 were regarded as
safe bets and overused by learners, but those without a direct L1 translation equivalent
were underused. For instance, phrases with amplifiers like completely and totally were
overused by French-speaking learners of English, but highly was underused.

What is interesting is that learners seem to transfer an L1 collocation to L2 in a
Systematic way. Kellerman and Jordens (1977) suggested that psychologically marked

features tend to be perceptually more complex and are thus less likely to be transferred

35

than unmarked features. So when L1 and L2 are perceived as distant by learners, only
core meanings (i.e., most frequently used, literal, and concrete meanings, psychologically
unmarked features) will be transferred. As to the peripheral meanings (i.e., more abstract
meanings, psychologically marked features), they might be rendered by means of
synonyms or paraphrases. This tendency was demonstrated in an experiment reported in
Kellerman (1979). Kellerman gave Dutch learners of English a list of Dutch sentences
with various meanings of the word breken (break) (e.g., hi] brak :ijn word (He broke his
word), hij brak zijn been (He broke his leg)) and asked them which of the translation
equivalents could be used in English. He found that when breken was used in a sense
closest to the core (e.g., hij brak :i/‘n been (He broke his leg) ), learners tended to accept
its English direct translation equivalent, but when breken was used in a more peripheral
sense (e.g., hij brak zijn word (He broke his word)). they were less likely to accept the
eCInivalent. Similarly, Gabrys—Biskup (1992) reported cases in which the learners were
reluCtant to transfer the metaphorical use of an L1 (Polish) word to L2 (English),
although sometimes the target L2 collocation was exactly a word-for-word equivalent of
the L1 one. For instance, czysty przypadek can be literally translated (as a (pure

Coincidence; however, Polish students still preferred to translate it as a plain/ total/

Straightforward coincidence rather than a pure coincidence.

36

Second, problems in L2 collocation production are also associated with intralingual
factors. One factor is the frequency of collocations. Hussein (1990) found that in a cloze
test which asked learners to complete a sentence with one of the four provided words,
students majoring in English at a university in Jordan did quite well with collocations
commonly used in everyday life, e. g., have a seat, fine arts, missing link, second thought,
spare parts. This result is not surprising, because as discussed in 2.3.2.1, the chances of
retaining certain items in LTM increase with their frequency in input. The other
intralingual factor that influences collocation production is the type of collocations (free
collocations, restricted collocations, idioms). This issue will be the focus of the following
disc ussion.

Free collocations seem to be the easiest to acquire. Putting aside syntactic
Well-formedness, we may say that the only co-occurrence constraint for this category is a
matter of semantic compatibility (see the discussion in 2.1.2). Allerton (1984) claimed
that " . .. this level of word cooccurrence restriction is not something the speaker has to

learn about. Once the meanings of the words are known, the limitation follows
aut0matically"(p. 21). For example, learners will not produce odd combinations like
a"Pacifiz theihouse, *tall information if they know the meanings of the individual words.

Idioms are more difficult for learners both in comprehension and in production.

37

/

Because the meaning of an idiom is non-transparent, learners might not be able to
recognize immediately its intended meaning (Cooper, 1999; Hatch & Brown, 1995;
Laufer, 1997). Moreover, the fixed nature of an idiom can be a problem for learners; even
misuse of a small particle (e.g., article) can make a speaker sound ridiculous (e.g., *She
let a cat out of a bag) (O'Dell, 1997). Overall, idioms present a special learning problem
and NNSs use much fewer idioms than NSs (Howarth, 1998).

Restricted collocations. in which the word selection does not solely depend on
syntactic principles and semantic compatibility, also pose great challenges for learners. In
a study of a corpus comprising of English assignments written by university students in
Hong Kong, Chi, Wong and Wong (1994) found that learners had difficulty in using

delexical verbs (have, make, take. do, get). There were confusions of these verbs with
each other (e. g., *do effort (make), *make discussions (have)) and also confusion of these
Verbs with other types of verbs (e.g.. *make solution (find), * get goal (reach)). Howarth
( 1 998) made a comparison between NS corpus of written British English and NNS data
drawn from a set of 10 essays written by 10 non-native postgraduate students as
assignments. It was concluded that NNSs of English, whatever their L1 is, used about
500/0 less restricted verb-noun collocations (and idioms) than NSs did, and the delexical

and figurative use of a verb were the most difficult for learners. Moreover, Howarth

38

identified two possible sources for learner errors: 1) blending two collocations that had a
semantic similarity in their noun element (e.g., pay attention, make an effort 9 *pay
effort); 2) filling in a collocational gap within a partially overlapping cluster (e. g., assign
a role to, assign a value to, attach a value to 9 *attach a role to). In a study of the data
from the German subcorpus of 1C LE. Nesselhauf (2003) differentiated two types of
restricted collocations: combinations where the verb (in the given sense) can only take a
limited number of nouns (e. g, run a risk). and combinations where the verb (in the given
sense) can take a wider range of nouns (e. g. exert + control, pressure, influence, authoriti‘.
power). It turned out that learners were mostly aware of the restriction in combinations of
the former type, but tended to be too creative with combinations ofthe latter type.

Third, in addition to L1 and L2 factors, production of L2 collocations is also found
to be influenced by proficiency levels. In Kellerman's (1979) study ofthe phrases of
breken, which was referred to earlier. a U-shaped learning curve was discovered, where
beginners and advanced learners were most willing to transfer the use of breken in

inStances further from the core meaning. It was suggested that perhaps beginners were
more daring and more willing to try something uncertain, advanced learners might have

learnt various senses of breken, while intermediate learners were more cautious.

In short, previous empirical research on L2 collocations has been trying to find out

39

what forms are misused, overused or avoided and why. These findings can provide us
with some insights into the learning process, but only indirectly. According to Gass
(1988a), there are five stages involved in the conversion from the input to the output: (1)
apperceived input (the bit of the language that a learner is exposed to and also notices), (2)
comprehended input (the part of the input that a learners understands). (3) intake (the part
of the linguistic material that is internalized or assimilated by the learner), (4) integration
(the development of L2 grammar or storage of the information contained in the input),
and (5) output (learner's production of that language). Under this model of SLA, we may
reinterpret the production data from the perspective of learning process. For example. the
learner errors in L2 collocations, as identified above, tell us that during the intake stage
leamers' hypothesis formation about a word’s collocational behavior is influenced by
theh‘ Ll (e.g., accepting a word-by-word L2 translation of an L1 collocation), by features
in L2 (e.g., making overgeneralization errors with a word used in its restricted sense) and
also by proficiency level (e.g., making more conservative hypothesis about a word's
possible collocates because of the past experience with the language). Moreover,
according to Gass, among the factors that may determine what output will take place are
(1 ) confidence in one's ability to produce correct target language, (2) degrees of strength

of knowledge representation (perhaps related to the automaticity of language processing).

40

 

Leamers' underuse and overuse patterns, as summarized above, indicate that factors like

the congruence between L1 and L2 (whether the word-by—word translation of a L1

collocation is acceptable in L2), types of the L2 collocations (free vs. restricted
collocations) and frequency of L2 collocations can influence how fast certain items can
be accessed and how confident one is in using these items. To sum up, analysis of the
problems in learner production can reflect certain aspects of the later stages in the
Ieaming process. In the next part, I will review the discussions and empirical studies that
concern the initial stages in Ieaming L2 collocations, more specifically, how different

levels and types of attention (peripheral vs. focal attention) influence the Ieaming of L2

(:01 locations.

2.4.2. Attention in learning L2 collocations

In previous literature. there have been some discussions about the best way to learn
01‘ teach collocations, which are related to the issue of attention. One recurrent view is
that Ieaming L2 collocations is predominantly a byproduct of natural exposure to L2. ,
TlLat is, by repeatedly coming across the same words occurring together in the same order,
individuals can store implicitly learned collocational patterns. And there is no need to pay
f0cal attention to a word's collocational features per 86 by making and testing conscious

hYpothesis or other means. Under this view, acquisition of L1 collocations and L2

41

 

l_

collocations are basically the same. According to Mackin (1978), learners cannot hope to
learn all the collocations they should know in any systematic way and the only way to
acquire some degree of collocational competence is years of reading, observation and
Ieaming of the language. Similarly, Gabrys- Biskup (1992) suggested that "probably the
best and easiest way for students to acquire the collocational system of a foreign language
is to be extensively exposed to a live language spoken and used by native speakers at a
certain period of time" (p.40). Kennedy (2003) expressed the same view, but from a
teacher's perspective. According to him, the frequency with which individuals experience
words and groups of words significantly influences the extent to which these linguistic
items are associated, stored and retrievable from memory; therefore the most important
t218k in teaching collocations is to maximize opportunities for learners to acquire them
thI‘OUgh "autonomous implicit Ieaming activities such as reading". The arguments by
SChrnidt (1995) and N. Ellis (1997) more explicitly addressed the issue of attention.
Schrnidt (1995) claimed that although the abstraction of more complex rules requires
undivided attention, simple item Ieaming or the Ieaming of very simple patterns (e. g.,
VOC abulary words plus their collocation) seems to be possible when attention is divided.
This comment implies that acquisition of L2 collocations does not require focal attention

to the relevant items. According to N. Ellis (1997), collocation acquisition (both L1 and

42

 

L2) is mainly an automatic and implicit process that takes place as leamers' vocabulary
extends and as they practise hearing and producing the words. As he put it, ". .. general
Ieaming mechanisms of chunking and sequence analysis, operating in the particular
domain of phonological memory" allow the acquisition of formulas, phrases, idioms, and
word collocation information (p.133).
There have also been researchers who emphasized or admitted'the value of
pedagogical attempt that aims to induce focal attention to collocations per se (e.g., Bahns,
1 993; Lewis, 2000; Moon, 1997; Nesselhauf, 2003). Lewis (2000) pointed out that too
often students and teachers look for and record only individual new words in texts and
this way ofleaming is to misidentify the constituent chunks. According to him, it is
P€dagogically helpful to (1) ensure that students are aware of word partnerships (e.g., by
having them keep a record of what words are in the company of a word); (2) raise
Students' consciousness awareness of the lexical nature of language and the importance of
Collocational knowledge. Some researchers suggested that it is beneficial to focus on the
following types of collocations: (1) collocations frequent in a neutral register and any
SDecial register that is of use to the learner (Moon, 1997; Nesselhauf, 2003); (2)

Collocations which do not have a direct equivalent in L1 (Bahns, 1993; Nesselhauf,

2003).

43

./

Even among the researchers who think that collocational Ieaming is essentially an
implicit process, there are some who acknowledged the potential value of focal attention
to this language area. N. Ellis (1997) noted that acquisition could be speeded up by
making the underlying collocational patterns more salient through explicit instruction or
consciousness-raising. Kennedy (2003) stated that although teaching explicitly the
complex system of collocations is not easy, some explicit instruction in frequently
occurring collocations is almost certainly worthwhile.

In spite of these discussions, to my knowledge, there have been very few empirical
studies that examined the role of attention in Ieaming L2 collocations. There is one study
relevant to this issue. Boers (2000) conducted three experiments with intemrediate
learners of English in Belgium and found that classroom activities aimed at enhancing
language leamers' metaphor awareness could facilitate retention of unfamiliar figurative
expressions. In the first experiment, two parallel groups of pupils were asked to read a
"3)“, which included many expressions to describe anger: cry it out, bottle up the
emOtions, ventilate one's emotions, blow off steam... After reading the text, the
experimental group received vocabulary notes organized along various metaphoric

themes, for instance,

Anger as a hot fluid in a container: anger welled up inside me, she erupted...

44

 

q...
:n
ll:
(fir-1
“\O
P‘n»

‘—

l‘ ‘4

 

 

 

 

Anger as fire: an inflammatory remark, she exploded...

The control group received the same vocabulary input, but the vocabulary notes were
organized along pragmatic or functional lines, for instance,
To describe acute and sudden anger: she exploded, she erupted...

To describe anger as a process: anger welled up inside me, an inflammatory

remark. ..

After reading the notes, both groups could ask for further clarification and
participated in a guided class discussion about anger and conflicts. Finally the
participants were given a cloze test with ten items meant to elicit the lexis studied and

they were encouraged to list several answers per gap if they could. The results of the test
Showed that the experimental group was significantly more likely to reproduce the lexis
Studied than the control group. The advantage of the experimental group was also found
in the other two experiments, which concerned other types of figurative expressions
(Experiment 2: items that describe downward and upward trends such as creep up. go
dOthilI, skyrocket, crash; Experiment 3: phrases using orientational metaphors, such as

Cheer up, feel down (GOOD IS UP, BAD IS DOWN). However, Experiment 3 also found

45

—————

 

that the experimental group was not any significantly better than the control group in
filling the gaps meant to elicit the expressions that had not been included in the treatment.
Boer made two major points about these results: (1) identifying source domains of the
metaphoric use of a word provides a framework for lexical organization and encourages
deep-level cognitive processing, which in turn promotes memory storage; (2) there is "no
successful transfer of the cognitive semantic approach" when learners try to tackle new
phrases, as indicated by the results of the novel items in Experiment 3.

In Boer's experiments, both the experimental and control groups paid some focal
attention to the target collocations, because the target was presented separately and as a
unit for them and there were special Ieaming activities focusing on these phrases.
However, the two groups paid focal attention to the targets on different levels. For the

CXperimental group. the metaphoric relations underlying the phrases were explicitly

revealed and explained to the learners. For the control group, learners focused on

rnapping the word sequence with a particular meaning. The fact that the experimental

gr Oup excelled seemed to indicate that focal attention at the level of knowing the

Underlying structure and organization is more effective for collocation Ieaming than

attention at the level of meaning comprehension.

To sum up, there have been very few studies on the role of attention in L2

46

collocation Ieaming. Most of them are discussions about the best way to learn or teach
collocations. One dominant view is that learning collocations mainly results from natural
exposure to the language. However, some researchers suggested that pedagogical
intervention is beneficial (e. g., encouraging learners to chunk phrases and learn them as a
\ whole, enhancing leamers' awareness of word partnerships). In other words, focal
attention in this area can be facilitative. There has been a lack of empirical research that
directly tests these assumptions and explores the exact role of attention in learning L2
collocations. The current study aims to deal with this issue that hasn't received
satisfactory treatment yet. In the next section. I will give an overview of the abundant
research on attention in Ieaming other language areas. which essentially shapes the
questions of my study.
2.5. More issues about attention
2.5.l Attention in learning L2 syntax and morphology
25.1.]. Attention at the level of noticing» the Noticing Hypothesis
Most of the research that explicitly addressed the issue of attention in SLA was
bElsed on syntax and morphology Ieaming and they were frequently centered on the
NOticing Hypothesis (Schmidt 1990, 1994, 1995; Schmidt & Frota, 1986). According to

the strong version of the hypothesis, although learning without intention or explicit

47

 

 

metalinguistic knowledge of what has been learned is possible, Ieaming without
awareness at the level of noticing, which is essentially isomorphic with focal attention, is
impossible. Here "noticing" is used as a technical term equivalent to "apperception"
(Gass, 1988a), to "detection within selective attention" (Tomlin and Villa, 1994) and to
"detection plus rehearsal in short term memory" (Robinson, 1995b). As to the content of
noticing, Schmidt argued that it must be whatever evidence is relevant for a particular
Ieaming domain rather than the input in a global sense. For instance. to acquire syntax
one must pay attention to word order and meaning. The weak version of the Noticing
Hypothesis claims that even if noticing is not necessary, it is certainly helpful and
contributes to Ieaming and retention. As stated by Schmidt (2001 ). ". .. since many
features of L2 are likely to be infrequent. non-salient, and cornmunicatively redundant,
intentionally focused attention may be a practical (though not theoretical) necessity for
successful language learning" (p. 23).

Many other researchers have put forth arguments for the importance of focused
attention or attempts to induce focal attention. According to Long (1991 ),focus on form,
that is, paying focal attention to certain linguistic features in the context of using the
language meaningfully, may be required for Ieaming; otherwise, certain items in the input

may go unnoticed, unprocessed and thus not learned (e. g., the grammatical tense verb

48

 

morphemes that tend to be ignored by learners, probably because of the high salience of

temporal adverbs). Terrell (1991) suggested that explicit grammar instruction, defined as

"the use of instructional strategies to draw the students' attention to, or focus on, form

and/or structure" (p. 53), can function as (1) an advanced organizer, which provides

learners with comprehension strategies that highlight crucial grammatical elements; (2) a

meaning-form focuser that highlights non-salient and communicatively redundant

relations. MacWhinney (1997) pointed out that explicit instruction can help students

narrow down certain hypotheses and consolidate the memory trace of particular items.

"Students who receive explicit instruction, as well as implicit exposure to forms, would

seem to have the best of both worlds" (p.278). According to Peters (1998), in every

aspect oflanguage Ieaming (grammar, phonology. semantics, pragmatics, vocabulary,

discourse). learners must attend to and notice any crucial source of variation. Peters gave

the following example: In English, both I turned the covers down and 1 turned down the

covers are acceptable, and there is no difference in meaning that is determined by the

position of the direct object; however, learners must also notice that I turned it down is

possible, but I turned down it is possible only in the sense of I turned down the road.

Williams (1999) claimed that learners do not simply absorb input and abstract the

underlying regularities. "Even rules that might be thought of as simple and transparent

49

 

 

 

....

-<»bd

 

a. ~\

‘r...

 

 

(e.g., plural marking of nouns) might fail to be induced if the relevant forms and
meanings are not segmented out of the input stream and intentionally associated together
in an instance" (pp. 40-41).

However, not all researchers agree that noticing is the attentional level required for
Ieaming L2 syntax. Some claimed that adults can best learn L2 syntax by attending to
meaning alone and grammatical patterns will be abstracted from the input incidentally
and subconsciously (e.g., Krashen, 1992, 1994; Schwartz, 1993; Schwartz &
Gubala-Ryzak, 1992; Zobl, 1995). In other words, it is not necessary to pay focal
attention to word order, to search for or learn syntactic rules. Tomlin and Villa (1994)
argued that there is a lot of evidence showing that information can be cognitively
detected even though the individual is not aware of its occurrence. For example,
subliminally presented words may activate their meanings without the subject being
aware of them (Marcel, 1983). Moreover, according to Tomlin and Villa, detected
information can be further used for processing, hypothesis formation and testing. That is,
conscious registration is not the minimally necessary level; instead, detection with
peripheral attention is the level by which SLA must operate.

It has to be noted that the usefulness of unconscious detection for learning is

controversial. According to Holender (1986) and Shanks and St. John (1994), subliminal

50

 

 

exposure effects are not likely to last for more than a few hundred milliseconds and they

are rapidly lost from memory. Kellogg & Dare (1989) pointed out that "the degree of

elaboration resulting from unattended encoding appears to be too limited to have any

substantive influence on human cognition or behavior "(p. 412). Robinson (2003) made a

similar point by saying that cases of unconscious detection are evidence of automatic,

unaware activation of LTM representation, not of Ieaming.

2.5.1.2. Empirical evidence for the Noticing Hypothesis

In the field of empirical research, there has been a large number of findings that lend

support to the Noticing Hypothesis.

First. studies of leamers' introspective or retrospective data (e.g., the data from diary,

think-aloud activity, interview, questionnaire) indicated that noticing is crucial for L2

development. Schmidt's diaries of Ieaming Brazilian Portuguese showed that the forms

appearing in his interactions with NSs were closely connected with his recorded noticing

in the diary (Schmidt & Frota, 1986). Cases of incorrect use could also be attributed to

wrong analysis of what he heard. Alanen (1992) asked English-speaking participants to

learn locative suffixes and a phonological phenomenon of Finnish under one of four

conditions: (1) input enhancement, in which the relevant features were highlighted by

italicization, (2) rule provision, (3) rule provision and enhancement, (4) control (mere

51

 

 

”I’M

exposure with no rule or enhancement). All participants described their thoughts as they
went along. Subsequent unexpected tests showed that, generally speaking, learners who
reported that they paid attention to the target forms had acquired them, whatever the
treatment, and there were no learners who acquired the targets without having noticed
them. In a study of Ieaming Spanish stem-changing verb form. Leow (1997) asked
leamers to think aloud while completing an L2 crossword puzzle. He found that learners
who stated some types of rule about the stem change performed more accurately on
subsequent tasks than those who mentioned the forms but did not state any rule. And the
latter group did better than those who did not mention the forms.

Second. many experimental or quasi-experimental classroom studies have been
conducted. comparing learners who focused on meaning alone with learners for whom
certain strategies were used to induce noticing. for example, input enhancement (i.e,
highlighting the target features by color-coding. italicization, underlining, etc.), input
flooding (i.e., artificially increasing the number of target forms in the input), instruction
to search for certain rules, provision of pedagogical rules (e. g., Alanen, 1995; Day &
Shapson, 1991; DeKeyser, 1995; Doughty, 1991; Jourdenais, Ota, Stauffer, Boyson &
Doughty, 1995; Leeman, Arteagoitia, Fridman, & Doughty, 1995; Lightbown, 1991;

Lightbown & Spada, 1990; Master, 1994; Robinson & Ha, 1993; Shook, 1994; Spada,

52

 

-
r

Cl
[.

.n‘

u'u

d..,

41..

'l
t.

lila-

.5”

Jr-n' .
with ,

._“
hi: '9'

 

1987; Spada & Lightbown, 1993; White, 1991; White, Spada, Lightbown, & Ranta, 1991;
also see Long, 1996, Norris & Ortega, 2000, for a review). The latter group generally
performed better and the advantages seemed to be durable. The following are three
examples:

Doughty (1991) compared three conditions of Ieaming object-of—preposition-type
relative clauses: an exposure only group, a meaning-oriented group with enhancement,
and a rule-oriented group which was informed of the rule. For 10 days, all participants
read sentences containing the targets. The post-tests showed that both the meaning-
oriented group with enhancement and rule-oriented group were superior to the control
group in their ability to relativize.

White (1991) compared the perfomiance ofthree uninstructed classes with two
classes who received two weeks' fomial instruction in question formation (a week of
explicit rules followed by corrective feedback on learner errors). The immediate posttest
(sentence correction activity) indicated that instructed learners were significantly more
accurate. In the second phase of the study, which included more testing activities (written
and oral communication tasks) and a delayed test 5 weeks after the instructions,
instructed learners still outperformed the uninstructed ones.

Jourdenais et al. (1995) investigated how English learners learned Spanish preterit

53

 

 

and imperfect verbs under one of two conditions: with highlighted input and with no
typographical modification. The enhanced group produced more target items in
subsequent writing task than the group without enhancement.
2.5.1.3. Attention at the level of rule understanding

In the preceding discussion, I mentioned various types of techniques that aim to
induce noticing or focal attentive processing of certain forms. These techniques display
different degrees of explicitness, which might lead to variegated levels of focal attention.
Following DeKeyser (1995), treatment can be considered as explicit if a pedagogical rule
is given or learners are asked to search for certain metalinguistic generalization on their
own. Techniques like input flood and input enhancement are relatively implicit, because
they only make forms more salient with no explicit instruction about how to process the
input (Doughty & Williams. 1998). An issue that is frequently discussed in the literature
is: Do more explicit techniques lead to better results? Or, more specifically, does explicit
rule formulation in the student leads to higher levels of achievement?

According to Schmidt (2001), ". . .the objects of attention and noticing are elements
of the surface structure of utterances in the input -— instances of language, rather than any
abstract rules or principles of which such instances may be exemplars" (p. 5). He further

claimed that rule understanding, is not necessary for Ieaming, but can be facilitative. The

54

 

first part of the claim does not seem to be controversial. There is a lot of evidence
showing that conscious rules are not required for L2 Ieaming. For example, in the study
by Green and Hecht (1992), German ESL learners could correct many errors, although
they could not articulate the rules or gave incorrect rules. However, the second part of
Schmidt's claim, which concerns the facilitative role of rule understanding, is a much
more debated issue.

On the one hand, a significant body of research showed that rule-based condition is
beneficial: it helps learners bring order to the input and also enables them to generalize
the knowledge to novel instances. Hulstijn and Hulstijn (1984) found that learners with
explicit knowledge oftwo Dutch word order rules had significantly higher perfomiance
scores than learners who were unable to verbalize the rules. although the latter perfomied
at better-than-chance levels on one of the target structures. In the classroom studies by
Scott (1989), an explicit group was provided with French rules about relative pronouns
and subjunctive and an implicit group read texts flooded with the target items. On written
posttests, the explicit group showed a significant advantage. Dekeyser (1995) focused on
the Ieaming of categorical and prototypical rules in an artificial language under one of
two Ieaming conditions: (1) explicit-deductive Ieaming (rule teaching followed by

thousands of illustrative picture and sentence combinations); (2) implicit-inductive (mere

55

exposure to thousands of picture and sentence combinations). As far as acquisition of
categorical rules is concerned, the two Ieaming groups performed at similar high levels of
accuracy in supplying the morphemes in sentences for previously viewed pictures, but the
explicit group was significantly more accurate in producing new sentences describing
new pictures. DeKeyser interpreted this result as meaning that implicit Ieaming was more
item-dependent and memory-based than explicit Ieaming. (The results about the
prototypical rule will be discussed in detail later in 2.5.3.2). De Graff (1997) investigated
the Ieaming of simple and complex morphological and syntactic forms of an artificial
language in about 150 hours of computer-based treatment. The explicit group was
presented with rules and also corrective feedback with grammatical explanation, but the
implicit group only rehearsed the sentences that contained the target structures. The
explicit group perfonned better than the implicit group, on all test sessions (immediate
and delayed posttests) and also on all task types (grammaticality judgment. cloze test. and
correction). Norris and Ortega (2000) summarized the results from the experimental and
quasi-experimental studies on the effectiveness of L2 instruction (published between
1980 and 1998) by using meta-analysis (i.e., "findings from all available primary research
studies are converted to comparable values by estimating the magnitude of an observed

relationship or effect, typically referred to as the effect size" (p. 163)). Comparisons of

56

average effect sizes from 49 sample studies indicated that treatment involving an explicit
focus on the rule-govemed nature of L2 structures (including deductive and inductive
approach) is more effective than treatments that do not include such a focus.

However, there is also evidence revealing the limitation of rules. For example, in
DeKeyser (1995), the advantage of rule instruction over mere exposure was restricted to
categorical rules describing simple patterns. Rule instruction could cause learners to
overgeneralize (Alanen, 1995; Robinson, 1996). And rules might be less effective if not
combined with exemplars (N. Ellis. 1993). The most serious problem is that knowledge
of rules does not necessarily guarantee target-like performance. For example, Seliger
(1979) found no relationship between perfomiance on the task and having a conscious
rule ofthe a/an allomorphs ofthe indefinite article in English. Macrory and Stone (2000)
observed students who had 4 to 5 years of instruction of French in a British secondary
school. The students were able to verbalize rules about the present perfect tense, but in
spontaneous conversation they omitted the auxiliary most of the time, except in formulaic
utterances.

Some studies concluded that the most important factor for intake is not the presence
of explicit information or directions to search for rules; instead, what plays a major role is

structured input activities, which is presented by means of practice and involves learners

57

.CL

1“
will

h;
I 5;

.t
.“

c“

 

in making form-meaning connections. VanPatten and Oikkenon (1996) investigated the
acquisition of Spanish preverbal direct-object pronouns in three conditions: ( l) explicit
rule and structured input, (2) explicit rule only, (3) structured input only. The two groups
with structured input outperformed the rule only group in both interpretation and
production. This indicated that it was the structured input activities not the rules that were
responsible for the benefits. This result was replicated by Farley (2003) and Benati
(2003). Sanz and Morgan-Short (2004) examined the acquisition of Spanish word order
by four Ieaming groups comprised of [+/— rule explanation] (i.e, whether grammar
information is provided) and [+/— explicit feedback] (i.e., whether grammar information
accompanies corrective feedback or not]. All groups improved significantly and similarly
on interpretation and production tests. The lack of effects for rule explanation and explicit
feedback leads the authors to conclude that explicit information may not necessarily
enhance L2 development. What is sufficient for acquisition is a feature shared by all
treatments, that is, structured input activities in which learners need to make responses by
processing the target forms for meaning, for example, tasks requiring learners to identify
the correct visual cues for a written or oral sentence or to comprehend written input.

In sum, there are mixed findings about whether explicit rule presentation (i.e., a

condition facilitating focal attention at the level of rule understanding) is always superior

to the Ieaming conditions that do no overtly involve rules (e. g., mere exposure to
numerous instances, memorizing instances, engaging in activities that demand
form-meaning connections). One possible reason for the complicated picture is that the
effect of attention is mediated by other factors. And this issue will be the focus of 2.5.3
and 2.5.4.
2.5.2. Attention in learning L2 word meaning

Like acquisition of L2 syntax and morphology, acquisition of L2 word meaning also
depends on the nature of the attentional activities carried out by the learner. There is one
view widely supported by psychologists and SLA researchers: more elaborate processing
of a new word (e.g., establishing rich associations between the fomial and semantic
features of a word, paying close attention to the context in which it is used, describing its
semantic relations to other words) leads to higher retention than less elaborate processing
(e.g., attending to only one of these dimensions) (see Anderson 1995, ch.6; Baddeley
1997, ch.7; Laufer & Hulstijn, 2001; Skehan, 1998). According to N. Ellis (1997),
although the acquisition of a word's form, its Input/Output (I/O) lexical specifications, its
grammatical class information and its collocations, all mainly involves implicit process of
sequence analysis, the acquisition of a word's semantic and conceptual features and

form-meaning mapping result from .explicit Ieaming processes, in which cognitive

59

Ieaming strategies and deep, elaborative processing are very effective (e.g., inferring

word meanings from context, semantic or imagery mediation between the L2 word and

L1 translation). In other words, in terms of Ieaming L2 word meaning, focal attention that

involves more elaborative rehearsal is more effective than attention that involves less or

no elaborative rehearsal.

However, what exactly constitutes a level of processing in lexical acquisition? How

do we know one type of processing is more elaborate than another? Some SLA

researchers have operationalized the general labels of attention and elaboration in

concrete constructs. Nation (2001, p. 63) described three crucial processes that enhance

the retention of a word: noticing, retrieval and creative (generative) use. Noticing a word

is the first step, which may be induced by various factors: pedagogical intervention, the

salience of a word in the textual input. negotiation of meaning in interaction, awareness

of the gap between what one needs to produce and what one can, etc. Retrieval is to recall

a word's meaning after perceiving its form or to access a form in order to express certain

meaning. Each retrieval can strengthen the path linking form and meaning and make

subsequent access to the word easier. Creative (generative) use refers to the cases when

old words are met or used in ways that are different from previous encounters, which can

also lead to mental elaboration that facilitates learning. According to Nation, noticing,

60

 

retrieval and creative (generative) use are three steps, with the later steps including the
earlier ones. In other words, noticing is a prerequisite for successful retrieval, and
retrieval is a prerequisite for successful creative use. Laufer and Hulstijn (2001) added a
motivational dimension to the constructs of vocabulary acquisition. Under their proposal,
depthfof processing is operationalized in three concrete task-specific constructs: need,
search, evaluation. Need refers to leamers' desire to meet task demands, which may be
externally imposed or self-imposed. Search is the attempt to get the meaning of a new L2
word or find a L2 word for a particular concept, which is equal to retrieval in Nation's
proposal. Evaluation involves comparing a given word with other words, comparing a
particular sense of a word with its other senses, or assessing whether a word can be used
in a context or not. We might say evaluation is similar to Nation's creative (generative)
use. but it covers a wider range of concepts. Laufer and Hulstijn assumed that when
everything else is equal, tasks with a higher involvement load, that is, with more amount
of need, search or evaluation, will be more effective for vocabulary retention than those
with a lower load.
In empirical research of Ieaming L2 word meaning, a large number of experiments

showed that more deliberate and effortful Ieaming conditions lead to more gains than the

less effortful ones (e. g., Hulstijn 1992; Hulstijn & Laufer, 2001; Laufer, 2000, 2003;

61

 

Newton, 1995; Paribakht & Wesche, 1997; Smith, 2004; Watanabe, 1997). This indicates
that more focal attention to a lexical item can strengthen the form-meaning mapping.

Watanabe (1997) investigated two groups of students Ieaming 16 target words by
reading texts. For one group, the words were glossed in the margin, either in the common
single gloss format or in a multiple-choice formati For the other group, students read the
same passages without glosses. Three unexpected vocabulary posttests showed that the
former group acquired significantly more new words than the latter. According to
Watanabe. under the non-glossing condition, learners might understand the text without
needing to know the new words, especially when the contexts are very rich, and this
might cause reduced attention to the target words and impede Ieaming. However, under
the glossin g condition, as suggested by Watanabe, learners tended to see the target word
as well as its definition when referring to the gloss and might also return to the word in
the text to see whether the provided meaning fit in the context. In other words, margin
glossing increases the possibility of noticing, which is a crucial first step in Ieaming.

On the other hand, glossing seems to lose its advantage when compared to
conditions that require learners to make more mental efforts to get the meaning of a new
word. For example, asking learners to infer meaning of a new word through context was

found to be superior to margin glossing in a study by Hulstijn (1992). Students who

62

 

looked up a new word in an electronic dictionary, which involved searching the right
form and determining which sense is appropriate when the word had several senses,
performed significantly better than those who were provided with margin gloss, both on
immediate and delayed vocabulary retention tests (Laufer, 2000).

Smith (2004) reported on lexical acquisition in a computer-mediated communicative
environment. By interacting with one another in a synchronous mode over a local area
network, pairs of NNSs of English cooperated with each other to complete jigsaw and
decision-making tasks that were embedded with unknown target lexical items. According
to the results of immediate and delayed receptive posttests (recognizing a new word) and
productive posttests (labeling objects with new words), new words that were negotiated
for meaning, that is, those involving noticing and retrieval (in Nation's term), or need and
search ( in Laufer and Hulstijn's term), were retained significantly better than those items
that were not negotiated. This finding is consistent with that of Newton (1995), in which,
items that were negotiated for meaning through confirmation checks, clarification
requests, etc. were better remembered than non-negotiated items, even by learners who
were simply observing the negotiation.

Besides meaning negotiation, exercises that attract leamers' focal attention to new

words and demand the processes of retrieval and generative use (or search and

63

evaluation) also prove to enhance Ieaming. Paribakht and Wesche (1997) compared
vocabulary Ieaming through reading and through reading coupled with exercises that
required the use of new words. Both conditions showed significant gains over three
months, but the reading plus exercises group acquired more words and also developed
deeper knowledge of the new words. Laufer (2003) reported three experiments that
compared the gains from reading and the gains from word-focused tasks. In the reading
condition, learners read a text with marginal gloss or read and looked up unknown words
in the dictionary. In the word-focused task condition, leamers completed given sentences,
created novel sentences, or incorporated words in a composition. Laufer found that more
words were acquired through tasks than through reading even when a particular new
word was noticed and looked up in the dictionary in the reading condition.

In conclusion, to the extent that L2 lexical acquisition is about mapping form to
semantic and conceptual representations, there is plenty of empirical evidence showing
that vocabulary acquisition is facilitated by deeper level of attention. Here attention is
used in the sense of selection and effort (see the discussion in 2.2). After a new word is
picked out from the input and noticed, the more efforts made to establish and strengthen
form-meaning links (either from form to meaning or from meaning to form), the richer

associations made between the word and other old knowledge, the higher are the chances

that the new word will be acquired.
2.5.3. Attention and different aspects of language

Recently, researchers begin to be interested in differential effects of attention on
different parts of language. VanPatten (1994) suggested that “perhaps different aspects of
language are processed and stored differentially" (p. 31). Schmidt (1995) also claimed
that “different aspects [of language] may require more or less [attention]” (p.14). While
challenging the Noticing Hypothesis in the area of syntax, Truscott (1998) indicated this
hypothesis might be applicable to other areas, for example, pragmatic knowledge and
lexical Ieaming.
2.5.3.1. Complexity of different language aspects

One feature that is often used to distinguish various aspects of language is the
complexity of a particular language phenomenon. Many factors have been proposed as
influencing the degree of complexity, all affecting the amount of attention and processing
effort expended in Ieaming (e. g., DeKeyser, 1998, 2003; Doughty & Williams, 1998; R.
Ellis, 1990; Hulstijn & de Graff, 1994; Krashen, 1982; Williams & Evans, 1998).
Although they are mostly motivated by the difference within syntactic or morphological
phenomena, some of them can also be applied to characterize the complexity in

collocational behavior, which will be discussed in Chapter 3. In the following part, I will

65

provide a brief overview of some of the major factors:

(1) Involvement of permutations, additions and deletions. It has been suggested that
the more permutations, additions or deletions are involved, the more complex a structure
is. For example, Krashen (1982, pp. 97-98) claimed that wh-question formation, which
requires extensive permutations of word order, is formally complex, but the addition of
the morpheme for third person agreement is simple.

(2) Salience. Salience of particular aspects ofa structure is related to the ease with
which structural correlates of rules can be noticed and learned, so in this sense structures
involving non-salient elements are more complex and thus more difficult to learn than
those involving salient elements. For example, wh-question fomiation is visually and
acoustically more salient in preposition stranding (e.g., Who did John give the book to?)
than in preposition pied—piping (e.g.. To whom did John give the book?), in terms of
frequency of occurrence and positions ofthe wh- word and preposition. This is claimed to
be the reason why the former type of structure is easier to be grasped by L2 learners than
the latter type (Bardovi-Harlig, 1987; Robinson, 1996).

(3) Size of context. Forms or structures whose applications depend on larger context
are more complex than those whose applications depend on immediate context. For

example, the alternation between a/an in English, which is determined by immediate

66

linguistic context, is simpler than the use of perfect tense, which is not always governed
by features of the immediate context (Green & Hecht, 1992).

(4) Reliability. Forms or structures whose use can only be captured by probabilistic
rules with several or many exceptions are more complex than those whose use can be
explained in terms of categorical rules. For example, the patterns in irregular past tenses
of English are probabilistic and a number of phonological or semantic characteristics can
only determine the likelihood that a certain morpheme or allomorph will appear
(DeKeyser, 1998). So this phenomenon is complex in the sense that it cannot be reduced
to an economical rule. Sometimes the use of a form or structure is so complex that it
cannot be described with satisfactory rule formulations without extensive yet incomplete
lists ofexceptions (Doughty & Williams. 1998). One example ofthis category. as
illustrated by Westney (1994), is the verbs that do or do not enter into dative alternation
in English.

(5) Semantic complexity. Forms or structures whose use is tied to numerous or
abstract semantic categories are more complex than those expressing simple and concrete
semantic concepts. For example, Krashen (1982. pp. 97-98) considered the use of plural
forms in English as functionally simple, that is, the morpheme denotes a very clear

semantic concept. However, the choice between definite and indefinite articles in English

67

is semantically complex because the underlying rules are related to semantic and
discourse concerns that are too numerous for the learner to acquire easily (Doughty &
Williams, 1998).

It has to be noted that because many factors can contribute to the degree of
complexity, it is not always easy to determine whether a particular language phenomenon
is simple or complex. For example, Krashen (1982) viewed the third person -s in English
as formally simple because it involves very simple addition, but R. Ellis (1990, p. 173)
pointed out that this form might involve a long-distance relationship with the
grammatical number of the subject. In other words. the use of this morpheme may be
considered as complex in terms of size of context.

Another issue that deserves attention is that complexity might be an individual issue.
As noted by DeKeyser (2003), complexity can be described as the ratio of a rule's
inherent linguistic complexity to the learner's ability to deal with such a rule. A rule that
seems to be moderately difficult to one student may seem easy for another who has more
language Ieaming experience.

In sum, numerous factors are conspiring to determine whether a particular aspect of

language is complex or not.

68

2.5.3.2. The interaction between attention and complexity

In this part, I will consider the controversy over the differential effects of attention
on different aspects of language. Central to the debate is this question: how does the
complexity of language areas influence the effectiveness of explicit Ieaming or
instruction (i.e., paying focal attention to certain aspects of language through rule
search/test or attempts to induce leamers' focal attention to certain aspects of language
through input enhancement, rule presentation, etc)?

Reber (1989, 1993) and Krashen (1982, 1985. 1994) claimed that implicit Ieaming is
more effective than explicit Ieaming when the stimuli domain is complex and conscious
rule Ieaming will be more effective in JUST those instances in which rules are simple. It
was pointed out that complex rules are more likely to be forgotten or imperfectly recalled
(Reber, 1993) and they can also result in confusion, which may cause learners to fall back
on Ieaming individual exemplar sentences (MacWhinney, 1997).

According to N. Ellis (1996), it is not true that explicit Ieaming or instruction is
ineffective for all complex areas. He distinguished two types of complex materials. When
the material to be learned is relatively complex but there are only a limited number of
variables and the critical features are salient, explicit hypothesis testing is helpful and

learners will also be able to transfer the knowledge to novel situations. However, "when

69

the material to be learnt is more randomly structured with a large number of variables and
where the important relationships are not obvious, then explicit instructions only interfere
and an unselective mode of Ieaming is more effective" (p.114).

The view held by Hulstijn and de Graff (1994) is contrary to that of Reber and
Krashen. They hypothesized that the advantage of explicit instruction is greater in the
case of complex rules than in the case of simple rules. The rationale behind the
hypothesis is: simple formal phenomena may be salient enough to be discovered by L2
learners themselves. even when explicit instruction is not available; but for complex
domain, explicit instruction can save learners considerable time in discovering the
underlying intricate rules.

However, it should be noted that Hulstijn and de Graff also assumed that "the
advantage of explicit instruction is greater when language production can only be based
on rule application, than when it can be based not only on rule application but also on the
retrieval of individually stored items" (p. 105). Although Hulstijn and de Graff separated
the issue of "item vs. rule learning" from the issue of "complexity", we may argue that
these two are actually related to each other sometimes, because some complex
phenomena are essentially lexical that can be treated as item Ieaming. One example is the

English verbs that do or do not enter into dative alternation, which was mentioned earlier.

70

For this type of material, Hulstijn (1995) recommended lexical learning (i.e., simply
memorizing individual instances) rather than rule Ieaming. It appears, then, under the
view of Hulstijn and de Graff, the advantage of explicit instruction is generally greater in
the case of complex rules than in the case of simple rules. but explicit instruction of
complex rules might be less effective than asking learners to store individual instances
when the target Ieaming items are essentially lexical.

In short, researchers seem to have different views about the differential effects of
attention on complexity. It is very likely that the interaction between attention and
complexity is quite intricate. Now we will turn to some empirical studies on this issue,
which again present a very complicated picture.

Carroll and Swain (1993) examined the Ieaming of L2 English dative alternations, a .
very complex phenomenon (see the discussion above), under five conditions. Upon
making an error, Group A was provided with explicit metalinguistic information about the
rule, Group B was simply told that they were wrong, Group C got recast, Group D was
asked if they were sure about their response, and there was also a group who did not get
any feedback. Subjects were tested twice on the feedback items plus novel items. It was
found that although all types of feedback were better than no treatment, those subjects

who received metalinguistic feedback performed better in extending to novel cases than

71

those who got implicit feedback. This means explicit learning can be facilitative when the
stimuli domain is very complex, which is contrary to the claims made by Reber and
Krashen.

Gass, Svetics and Lemelin (2003) compared the Ieaming of three parts in L2 Italian
under two conditions. The three parts are: lexicon (the meaning of five lexical items),
syntax (wh- question formation in biclausal sentences) and morphosyntax
(order/placement and agreement features of direct and indirect object pronouns). The
lexicon is considered as less complex than the syntax and morphosyntax, because the
target L2 words have semantic counterparts in L1 and the only difference lies in the
phonetic shape of the words. However, to loam syntax and morphosyntax, as pointed out
by the authors, learners need to isolate forms and learn the correctness of those fomis.
The two Ieaming conditions in this experiment are: [-focused attention] (mere exposure,
attention directed towards irrelevant parts) and [+focused attention] (rule presentation
coupled with examples or instruction to guess word meaning]. Analysis of
pretest-to-posttest gains showed that [+focused attention] condition facilitated
development in lexicon, syntax and morphology, but its greatest effect size was noted in
syntax/morphosyntax (the most complex area) and the least in lexicon (the most simple

area). The authors suggested that with the lexicon learners are able to use their own

72

internal mechanisms to generate attention, but for Ieaming very complex and abstract

areas, learners cannot solely rely on their own internal devices and explicit instruction
may be necessary. This result lends support to the hypothesis by Hulstijn and de Graff
and does not support the claims by Reber and Krashen.

Dekeyser (1995) (see previous discussion 2.5.1.3) focused on the Ieaming of two
types of rules in an artificial language: (1) categorical rule (the simpler area); (2)
prototypical rule in the sense that there was a choice between two possible morphemes
and the choice of the allomorph was probabilistic (the more complex area). There were
two Ieaming conditions: (1 ) explicit-deductive learning (rule teaching followed by
thousands ofillustrative picture and sentence combinations); (2) implicit -inductive (mere
exposure to thousands of picture and sentence combinations). This experiment made two
major findings. First, as far as acquisition of categorical rules is concerned, the explicit
group was significantly more accurate in producing new sentences describing new
pictures. Second, as far as acquisition of prototypical rules is concerned, there was no
difference between the two conditions in production of sentences describing old and new
pictures, although the implicit learners appeared to be more sensitive to the probabilistic
nature of those rules. To sum up, explicit instruction is superior to the implicit condition

in Ieaming simple rules, but this superiority seems to be leveled in the Ieaming of more

73

complex rules.

Similar results to those of DeKeyser were obtained by Robinson (1996). There were
two target rules in Robinson's study: (1) hard rule -- a rule about how to form pseudo
clefts of location (e.g., Where Marv and Johan live is in Chicago not in New York); (2)
easy rule -- a rule describing the fact that subject-verb inversion is allowed in sentences
where adverbials of movement or location are fronted (e. g., Into the house John ran/ran
John). Four Ieaming conditions were examined: (1) implicit -- learners were asked to
memorize sentences; (2) rule-search-- learners were asked to search for the underlying
rules; (3) incidental -- learners read sentences for meaning; (4) instructed —- learners were
taught with the rules and learned how to apply the rules to the sentences. In this
experiment, instructed learners outperfomied all others in learning simple rules and
implicit learners did not outperform other learners on complex rules.

To sum up, previous studies showed that in Ieaming simple areas, focal attention to
target items or structures (often induced by explicit instruction) is more effective than
lower level of attention (e.g., mere exposure to instances). However, the answer is less
clear for learning more complex area. The studies by Gass, Svetics and Lemelin (2003)
and Carroll and Swain (1993) observed strong effect of focal attention in learning

complex areas, but in the studies by Dekeyser (1995) and Robinson (1996), this effect

74

seemed to be diminishing because no significant difference was found between the
focused attention condition and the non-focused attention condition(s) in learning
complex items.

This mixed result can be understood if we view complexity as a continuum,
moreover, as a personal continuum (see previous discussion about complexity as an
individual issue). At one end of the continuum are the target items that seem very simple
to the learners. Learners can characterize the input data themselves, thus the impact of
externally driven focal attention is not that strong. In the middle of the continuum there
are some aspects that are more complex and require more efforts to figure out. In this
case, externally driven focal attention may speed up the natural process. The target items
in the experiment by Gass et al. may be something ranging from the simple end to the
middle of the continuum, at least to the subjects. Towards the other end of the continuum
are some extremely complex items that might be resistant to explicit instruction or
Ieaming, at least for some learners. As a result, explicit instruction does not make a
difference, as demonstrated in the experiments by DeKeyser and Robinson. A similar idea
was expressed by DeKeyser (2003), according to whom, explicit instruction is not
necessary when the grammatical rule is very easy, not effective when the rule is very

difficult, and effective when the rule is easy, moderately difficult or difficult. In short, at

75

this point, we may suggest that the effect of externally imposed focal attention depends
on the side of the complexity continuum the target Ieaming items sit on. However, it is
not easy to determine the degree of complexity and the issue is further complicated when
complexity is considered as an individual issue.

2.5.4. Attention and proficiency level

There is another issue relevant to the present study, that is, differential effects of
attention on proficiency levels, or current state of IL system.

L2 leamers' progress is characterized by specific stages of development, with certain
structures acquired later than others (Pienemann, 1987, 1989; Pienemann & Johnston,
1986). Although the developmental stages that have been identified cover only a very
small part of structures in English and Gemian and even less of other languages, and
some features do not seem to have a fixed acquisition sequence (see Li ghtbown, 1998),
there is a clear tendency that forms demanding more processing complexity and capacity
(e. g., those involving a lot of manipulation of elements, or discontinuity created by
extraction) are acquired later than those that are easier to process (e.g., structures
involving less manipulation of elements and respecting continuity) (see Hawkins, 1989;
Tarallo & Myhill, 1983; Wolfe—Quintero, 1992). For instance, learners seem to acquire

unanalyzed formulas before they are able to move elements around to establish

76

long-distance relationship.

For Pienemann, each developmental stage is a prerequisite for the next, because
learner production is constrained by speech-processing strategies that are related to
cognitive factors (e. g., perceptual salience and continuity of elements) and learners need
to shed certain constraints before they are able to move on to a new stage. Doughty and
Williams (1998) pointed out that if learners still need to expend most of their attentional
resources for other parts of comprehension or production and have little attention to free
up for noticing a particular new form, then focusing on that form will be ineffective. In
other words, it is useless to attract leamers' focal attention to a form for which they are
not developmentally ready yet. In a discussion about the relationship between explicit
knowledge (i.e., metalingual and conscious knowledge about language and about the uses
to which language can be put) and implicit L2 knowledge (i.e., intuitive and unconscious
knowledge which is available for automatic use), R. Ellis (1997) suggested that explicit
knowledge derived from formal instruction maybe converted into implicit knowledge.
but only if the learner has reached a level of development that enables him/her to
assimilate the new material. Schmidt (1990) posited that one of the factors influencing
noticing is readiness to notice. There are two implications of this claim. First, if a learner

is ready to notice some features, then formal instruction can speed up this natural process

77

or perhaps even without pedagogical intervention he/she may notice those features
spontaneously. Second, if a learner's attention is directed to something that he/she is not
ready to notice, then it might fail to be incorporated into the IL system. In short, all these
discussions suggest that a learner's proficiency level is predisposing him/her to be ready
to attend to something while ignoring other things, which in turn influences the effect of
any extemally-imposed focal attention.

A number of experiments have been carried out to investigate the interaction
between attention (usually externally driven focal attention) and proficiency.

VanPatten (1990) asked learners of L2 Spanish to count the number of occurrences
ofa lexical item (noun), a free morpheme (a definite article). or a bound morpheme (third
person plural), while comprehending the meaning ofthe text simultaneously: Lower-level
students could not do both types of processing at the same time. but the advanced-level
students could process the form and meaning of the noun and article simultaneously.
although they could do not do this with the third person plural. This result indicates that
advanced learners are more effective input processors.

In a review of experiments that explored the effects of formal instruction, Long
(1988) found that even if learners had grasped explicit knowledge about certain linguistic

features and also practised, the target structures were still not obvious in communicative

78

language use unless learners were developmentally ready.

Pienemann (1989) conducted a series of studies to examine the effects of

instructional intervention on developmental rules (i.e., rules displaying certain acquisition

sequence) among young learners of German as L2 and adult learners of German as

foreign language. It was found that only those who were at a stage immediately before

the stage represented by the target items benefited from the instruction. For learners who

were at lower stages, the intervention could bring about confusion, avoidance of the

target structures or even caused some learners to perform worse than before.

Williams and Evans (1998) looked at the acquisition ofparticipial adjectives and

passives in classrooms by three groups: (1) the flood group -- learners received written

input that had been modified to include three times the number of target structures

contained in the original material and their teacher deliberately used more target items in

discussions; (2) the instructed group -- learners received the same input as the flood

group; moreover, their attention was drawn to the target forms as a result of rule

explanation and corrective feedback from the teacher; (3) the control group -- learners

worked with basically the same materials as the other two groups, but got no instruction

or artificial increase in the number of target forms. One interesting result emerged from

this experiment: although the instructed group showed greater increases in the accurate

79

use of participial adjectives than the flood and control groups, in terms of passives, the
instructed group lost its superiority and the input flood was almost as effective as the
instruction. According to the authors, one possible reason for the result is that the passive
is more complex than the participial adjective and many participants of the experiment
might have just begun to notice the passive and were not yet ready for the Ieaming of the
form. This explanation received further support from other data of the study: the students
who made the greatest gains after instruction were those who already had some
knowledge of the target item (as shown in the pretest); those who had extremely low
scores in the pretest generally made little progress after the treatment. Moreover, this
pattem could also be observed among the few students who achieved Ieaming gains in
the control condition. Williams and Evans viewed this as evidence that learners who are
ready to notice may find the relevant data wherever they can even without instruction.

In the experiment by Gass, Svetics and Lemelin (2003), referred to earlier in 2.5.3.2,
the researchers questioned the differential effects of attention on both language areas
(lexicon, morphosyntax, syntax) and proficiency levels (first-, second- or third-year
students in Italian classes at a university). For lst-year students, [+focused attention] had
a significant impact on all language areas. For 2nd-year students, [+focused attention]

produced significant results in lexicon, and marginally significant results in

80

morphosyntax and syntax; For 3rd-year students, [+focused attention] produced no
significant difference in any of the areas, although the results approached significance for
syntax. In short, the focused attention condition appeared to have the greatest effect in ,
early stages of Ieaming, and a diminished role with higher-level learners. As suggested by
the authors, one possible reason is that hi gh-level learners have mastered more
knowledge of L2 and they can use their own internal resources to find out and understand
aspects of the language.

To sum up, we can get two conclusions from previous studies. First, attempts to
draw leamers' focal attention to certain features are facilitative only when learners are
proficient enough to accommodate those features. Otherwise. the attempts will be futile
or counterproductive. Second. when learners are ready to learn something. they might
notice the relevant features themselves even without external interventions (although
external force can speed up the natural process).

2.5.5. The role of negative evidence

This section will focus on the role of NE in SLA. NE is relevant to the issue of
attention in the sense that it directs leamers’ attention to what is impossible or a problem
area.

Some studies reported that NE often fails to be noticed by adult learners. For

81

example, Hawkins (1985) found that adult classroom leamers often did not perceive NE
as such when working with a NS on a problem-solving task. Slimani (1992) observed 6
ESL classes to adult students and found that 36% of the grammatical and lexical items
presented by teachers were ignored by students and the majority of them were error
corrections. Some studies claimed that students who were corrected demonstrated no
gains, or were similar to the comparison group who did not get corrective feedback or
corrected less (e.g., Cohen & Robbins, 1976; Robb, Ross & Shortreed, 1986; Semke,
1984).

However, most research showed that NE does facilitate Ieaming.

First, L2 learners not only notice but also benefit from NE. at least in the short temi.
Wren (1982) found that an advanced adult ESL student could only self-correct 14% of
her errors, but during individual tutorial sessions she could correct 83% of the errors in
response to a teacher’s feedback. Pica (1988) reported that when interlocutors showed
signs of noncomprehension, beginning ESL acquirers modified their speech towards the
direction of the correct forms 31% of the time; intermediate learners did so 51% of the
time in the study by Pica, Holliday, Lewis and Morgenthaller (1989). Lightbown and
Spada (1990) presented data from intensive communicative ESL classrooms in Quebec.

Learners who received error correction achieved greater accuracy in the production of

82

some of the target structures (e.g., the use of the correct pattern There is in place of the
wrong structure caused by Ll influence *It has...) The high performance on there is/are
structure was maintained when learners were tested one year later (Li ghtbown, 1991). In
natural setting, Oliver (1995) found that child L2 learners did make use of the negative
feedback from their NS peers in the conversation and the NE was incorporated by the
NNS into their IL system.

Second, NE might be required in the sense it can make learners aware of certain
deviant forms that are often not noticed through PE. Logically speaking, PE alone can
only tell learners what is allowed, but not what is disallowed, and leamers seem to need
NE to avoid overgeneralizations. Gold (1967) illustrated the following example in
English: we can say The man is sick; I visited the sick man; however, although The child
is afraid is OK, * l comforted an afraid child today is not. In this case, according to Gold,
NE is necessary for those learners who make overgeneralizations to unleam the errors.
White (1991) examined the acquisition of adverb placement restrictions in English by
two groups of N83 of French. Before the treatment, both groups accepted SVAO
structures as a possible English order as a result of L1 influence. Then one group was
given information about the adverb placement constraint in English, while the other was

instructed in question formation. After two weeks' instruction, the adverb placement

group learned that SVAO was not permitted in English, but the question group did not,
although this effect disappeared when participants were tested again one year later. In a
follow-up study, Trahey and White (1993) provided two classes with an input flood of the
use of English adverbs, which lasted two weeks. The input flood enabled learners to
know that SAV order is possible, but it was not sufficient for learners to learn the
ungrammaticality of SVAO order. These experiments indicated that NE is required for
learners to remove overgeneralization errors, which are caused by an analogy within L2
or with Ll.

Another line of research is the differential effects ofNE. The following are the
major assumptions and findings on this issue.

First. the usefulness of NE may be related to language areas. While dismissing the
usefulness of NE in Ieaming syntax, Schwartz (1993) claimed that providing NE in the
lexicon and morphology might indeed be successful. According to her, Ieaming the
lexicon involves item-by-item Ieaming, the inventory of morphological forms is
comparable to those found in the lexicon and "no clustering is posited in these areas”
(p159); therefore, NE might be useful for Ieaming the lexicon and morphology. Although
previous empirical studies do not support Schwartz' view about the uselessness of NE in

Ieaming syntax (see the discussion above), it stands to reason that NE might produce

84

different effects for Ieaming different parts of language, because as reviewed in 2.5.3,
focal attention has different effects on different language areas and NE is exactly a means
to draw leamers' focal attention to a particular area.

Second, the perception of interactional corrective feedback is related to the target of
the feedback and the nature of the feedback type. Mackey, Gass and McDonough (2000)
found that learners were more accurate in their perceptions about lexical. semantic and
phonological feedback than about morphosyntactic feedback. The authors proposed the
following reasons to explain this result: (1) in a conversational context, individuals
usually focused on meaning rather than form. so morphosyntactic feedback that generally
dealt with non-meaning-bearing elements in the experiment was not noticed; on the other
hand. phonological. lexical and semantic feedback. which were more often related to
meaning. were more likely to be noticed; (2) most of morphosyntactic feedback was
provided through recast (i.e., "reforrnulations of an incorrect utterance that maintains the
original's meaning" (Gass & Selinker, 2001, p.458)), which was an implicit type of NE
and might be perceived by learners as another way of saying the same thing.

Third, the effectiveness of NE is influenced by proficiency levels. As noted by
Spada and Li ghtbown (1993), “error correction strategies are more effective at different

times in leamers’ development (p219)”; Mackey and Philp (1998) reported that implicit

85

negative feedback in the form of recasts was more effective for advanced learners than

beginning learners. The differential effects of NE on proficiency levels might be

associated with leamers' readiness to notice, understand and assimilate the information

passed by the NE (see the discussion about the interaction between attention and

proficiency level in 2.5.4).

To sum up, as claimed by Schachter (1991), “the kind of knowledge to be learned,

the kind of evidence presented to the learner, the situation in which the Ieaming takes

place and the cognitive capabilities ofthe learner all play a part in the efficient or

non-efficient use of negative evidence" (p.99).

2.5.6. Summary of the role of attention in SLA

On the basis ofthe preceding literature review, we may draw the following

conclusions about the role of attention in SLA:

(1)Attention, a continuum from the peripheral to focal. plays a crucial role in SLA.

(2) As far as Ieaming L2 syntax is concerned, many experiments show that noticing,

which involves the allocation of focal attention, can facilitate Ieaming. Rule

understanding, a higher level than noticing, is not required, but it can promote L2

development at least on some occasions.

(3) As far as Ieaming L2 word meaning is concerned, many experiments show that more

86

elaborate manipulation and analysis of a lexical item, which requires focal-attentive

processing, contributes to the retention of the form-meaning connection.

(4) As far as Ieaming L2 collocation is concerned, very few empirical studies have been

carried out to explore the role of attention in this area. Some researchers suggested that

Ieaming L2 collocations is predominantly an implicit process that does not require any

focal attention to the collocational features per se. But there were also some who

maintained that some type of focal attention can enhance the Ieaming of L2 collocations.

(5) Attention can have differential effects on different language areas. The most optimal

attentional condition may be different for learning syntax, morphology, lexicon,

phonology and pragmatics, and also different for learning simple areas and complex areas.

However, we still do not know the exact picture ofthe differential effects.

(6) The effect of focused attention is mediated by proficiency levels. Any attempt to draw

leamers' focal attention to certain features in L2. for example. rule explanation or

instruction to search for a rule, is effective only when learners are developmentally ready.

(7) On balance, the available evidence indicates that NE promotes L2 Ieaming. However,

the perception and usefulness of NE can be influenced by factors like the target of the NE,

types of the NE, and learner proficiency levels.

87

CHAPTER 3
The Present Study
3.1. Research questions

In Chapter 2, I provided an overview of the major issues surrounding the role of
attention in SLA, most of which were based on evidence from Ieaming syntax,
morphology and basic word meaning. The current study aims to explore some parallel
issues in the area of Ieaming L2 collocations. As listed in Chapter 1, there are three
research questions, which are repeated here and described in more detail below:

RQI: Which ofthe following four attentional conditions leads to the best gains in
Ieaming a verb's possible noun collocates?

( 1) semantic processing ([8]) -- Learners are asked to understand passages that are
embedded with non-highlighted target collocations and then their attention is directed
towards aspects unrelated to the target collocational features;

(2) memorization for recall ([M]) -- Learners are asked to understand passages that
are embedded with highlighted target collocations and then they are instructed to
memorize the target collocations for a later recall test;

’(3) rule given ([R]) -- Learners are asked to understand passages that are embedded

with highlighted target collocations and then they are provided with the collocational

88

rules underlying the target items and study how the rules apply to the instances in the
passages;

(4) rule given plus negative evidence ([R+NE]) -- Learners are asked to understand
passages that are embedded with highlighted target collocations and then they are
provided with the collocational rules underlying the target items and study how the rules
apply to the instances in the passages. Moreover, they are presented with negative
evidence, that is, what are impossible noun collocates for the target verbs?

This research question is tied to the following issues: (1) the effectiveness of
focused versus peripheral attention to target items; (2) the usefulness of attention at the
level of rule understanding; (3) the role of negative evidence. The four experimental
conditions. as identified above, are designed to induce four levels of attention, which are
largely a consequence of the processing demands of particular task requirement.

It is intended that learners under the semantic processing condition have peripheral
attention to target collocations. in the sense that they are only exposed to the word
sequence but are not consciously searching for any collocational regularities nor
chunking out the target combination as a unit. It has to be noted that attention is
ultimately what takes place in the participants’ brain, not necessarily what the researchers

intend it to be. As pointed out by Gass et a1 (2003), "one cannot rule out the possibility

89

that learners, because of their own individual needs and interests, paid attention to
something despite the fact that we did not experimentally draw their attention to that
something"(p. 526). Therefore, at this point, it can only be claimed that the semantic
processing condition aims to limit the possibility of any self-directed focal attention to
the target collocations. There are two factors contributing to the achievement of this aim.
First, in this condition, leamers’ attention is drawn towards an aspect that is not directly
related to the target collocational features, thus the likelihood of their focal attention to
the targets is much reduced. because as reviewed in 2.2., attention is capacity robbing and
limited and it is difficult for a learner to do several types of mental processing at the same
time, especially when the processing draws on the same resource pool. Second. the
internal need to pay focal attention to a word's collocational feature is not strong, because
this type of attention is not required for meaning comprehension. For example, once
learners know that the basic meaning of wield is “to have and use something”. they will
not have any difficulty in understanding combinations like wield sticks, wield power, and
there is no need for them to focus on or reflect upon what type of noun frequently follows
wield. As suggested by Gabrys-Biskup (1992), learners do not make a conscious effort to
understand or memorize new collocations, especially those with transparent meaning,

since those collocations do not cause any perception problem; as a result, the new

90

collocations very often pass unnoticed, with no or only a weak memory trace left. In short,
due to the lack of external direction and internal need, the semantic processing group is
very likely only paying peripheral attention to the target collocations. This assumption
will be further tested in an exit questionnaire (see later description in 3.2.3.3.).

For learners in the memorizationfor recall condition. their attention is directed
towards the target collocations through the following means: ( l) the target collocations
are highlighted in the passages and there is prior instruction which asks them to pay
attention to the targets when reading the passages; (2) the memory task encourages
learners to chunk the target collocation as a unit rather than merely noticing individual
words, which in turn can enhance the long-term storage ofthe collocations. Compared to
the semantic processing condition, memorization for recall can induce a more focal level
of attention to the target collocations. Moreover, learner's knowledge is based more on
memory for 'co-occurring word sequence than on underlying rules, although one cannot
preclude the possibility that by repeating those combinations, learners might detect
certain patterns.

For learners in the rule given condition, their attention is directed towards the target

collocations at an even deeper level than the memorization for recall condition, that is,

attention at the level of rule understanding. Besides input enhancement, there are explicit

91

discussions about the collocational rules, which tell learners how to organize the input.

This is likely to facilitate the development of a knowledge base in which the deductive

application of rules plays a large part. Meanwhile, it may also lead to the long-term

storage of the target instances.

Learners in the rule given plus negative evidence condition are supposed to have the

deepest level of focal attention to the target items, because besides getting the treatment

like that in the rule given condition, they are informed of what impossible collocates are,

thus their attention is directed towards both the collocational patterns displayed in the

positive evidence and the areas where they might make errors.

To sum up, the four experimental conditions semantic processing, memorizationfor

recall, rule given and rule given plus negative evidence are designed to elicit four levels

of attention to the target collocations respectively: (i) peripheral attention (mere

exposure), (ii) focal attention at the level of instance chunking and storage, (iii) focal

attention at the level of rule understanding, (iv) focal attention at the level of rule

understanding and knowledge of what are impossible combinations.

RQ2: Is the effect of the four attentional conditions mediated by complexity of

collocational patterns?

This question is motivated by previous debate over the interaction between attention

92

and complexity of a particular language phenomenon. Two degrees of collocational

complexity are investigated: a) a verb collocated with two types of nouns, which can be

related to each other by metaphor (e.g., wield weapon, wield power); b) a delexicalised

verb accompanied by two types of nouns, which refer to disparate and unrelated things

(e.g., make some progress, make a speech). There will be more detailed discussions

about this issue in 3.2.2.

RQ3: Does attention have differential effects on different proficiency levels?

This question is inspired by previous discussions about the interaction between

attention and learner proficiency level. This study will compare the Ieaming gains of

participants from two proficiency levels (see further description in 3.2.1.).

3.2. Method

3.2.1. Participants

10] Chinese students who were non-English majors in a university in China were

recruited to participate in this research. The experiment was carried out at the beginning

of a spring semester (i.e., the second semester of an academic year). At that time, 49 of

the participants were first-year university students who had just enrolled in a college

English course (Level 2); 52 were second-year or first-year students who had just

93

enrolled in a college English course (Level 4)‘. Before being admitted to the university,
all the students had studied English at school for six to nine yearsz.

Within each proficiency level, participants were randomly assigned to one of the four
attentional conditions as specified in 3.1. The condition assignment procedure went like
this: subjects of a particular level seated themselves randomly in a classroom. Then they
were asked to count themselves off as A‘s, B’s, C ’s and D's. A’s were put into semantic
processing group, B’s into memorization for recall group, C’s into rule given group, and
D's into rule given plus negative evidence group (See Mackey & Gass (2005) for
description of this method). Because 7 students dropped out during the multiple-session
experiment. the actual number of participants who were present throughout the whole
procedure and whose data could be used is 94. Table 2 shows the number of participants

of each proficiency level and treatment condition.

 

' At the time of the experiment, there were four levels of English courses offered in that university: Level 2,
Level 3, Level 4, and Advanced English, with Levels 2-4 having the largest enrollment. In order to recruit
as many participants as possible and also maximize the proficiency gap among the learners so that a
difference in the treatment effect is more likely to be detected, Level 2 and Level 4 learners were selected
for the experiment.

2 In China, the standard of English teaching in the elementary and middle schools varies greatly with
regions. So new students at the university are required to take an English placement test, which determines
which level of course they should begin with.

94

 

Table 2 Participants information

 

 

 

 

semantic memorization rule given rule given plus Total
processing for recall negative evidence
Level 2 13 12 9 10 44
Level 4 13 12 12 13 50
Total 26 24 21 23 94

 

 

 

 

 

 

 

3.2.2. Linguistic items to be investigated
This study aims to investigate the acquisition of the collocational patterns of four
verbs: wield, warp, make, take. In the input material presented to the participants, each of
the verbs collocated with two types of nouns; moreover, these verbs displayed two
degrees of complexity in terms of collocational behavior. The following is a detailed
description of the target collocational patterns (see Collins C obuild English Dictionary
and Collins C obuild English Language Dictionary for the collocational patterns).
(1) easy type: wield, warp
wield occurs with
-- nouns referring to a concrete tool or piece of equipment that can be carried and used
(e.g., stick, weapon, knife) (The examples in the parentheses are the nouns that
appeared in the passages read by the participants. The same is true with the
examples isted in the following parentheses.)
-- nouns associated with the concept of power (e.g., control, power, influence)

warp occurs with

95

-- nouns referring to an object that can be damaged by heat or water (e. g., floor, door,
bookshelf?
-- nouns related to one's mind (e.g., mind, judgment, view)
(2) difficult type: make, take
make, used as a delexical verb, occurs with
-- nouns related to speech action (e.g., claim, speech, suggestion)
-- nouns related to Changes (e.g., change, progress, recovery)
take, used as a delexical verb, occurs with
-- nouns referring to physical actions (e. g., look, breath, step)
-— nouns referring to a particular role. especially one of leadership (e.g., lead, role,
power)

The collocational patterns of wield and warp are considered as easier to acquire than
those of make and take for the following two major reasons:

The first reason lies in the difficulty in figuring out the target patterns. When
learners encounter various noun collocates of wield and warp, the contrast between nouns
denoting concrete objects and those about abstract concepts can be quite salient, for
instance, stick/weapon/knife vs. control/power/influence. This salient distinction enables

learners to divide the possible collocates into two clear-cut categories from the very

96

beginning, a process that can facilitate the further detection of what specific type of

concrete or abstract noun can be associated with the verb. However, for the noun

collocates of make and take targeted in this research, there is no such salient distinction;

moreover, there are no close synonyms among them, which may further impede the

clustering and categorization process. For instance, the noun collocates of take shown to

the participants were look, lead, breath, role, step, power: All of them are about

something not tangible and there is no synonymous relationship between any two of them.

In this case, learners might not be exactly sure about how to classify these noun

collocates, at least not at the beginning. To confimi this assumption about the difficulty in

discovering the underlying collocational patterns, three English teachers in China, who

are NNSs of English, were presented with all the target verb-noun collocations and asked

to rank the difficulty in classifying the noun collocates of each ofthe four verbs. All of

them agreed that it was easier to make the classification for wield and warp than for make

and take.

The second reason for the distinction is related to the difficulty in storing the

patterns in LTM. The two types of noun collocates of wield and warp, one concrete and

the other abstract, can be easily associated with each other through metaphor. For

instance, both power and weapon can follow wield, and power can be regarded as a type

97

of weapon. This metaphorical relationship can facilitate the retention of the two types of
collocates. However, for the delexical use of make and take, the two types of noun
collocates targeted in the study are unrelated with each other. For instance, the noun
collocates of make --speech and change -- refer to two disparate things. Obviously, it is
more difficult to encode two unrelated items in LTM and associate them with one verb
than to store two related items. Moreover, as previously described, in their delexical use,
the basic meaning of make and take is somewhat reduced, so the semantic connection
between a delexical verb and its following noun is quite loose, which can make the
long-temr storage of the combination even harder.

In fact. besides the above two reasons. which are related to semantic complexity,
there is another factor that causes the collocational features of make and take to be more
difficult to learn, that is. the two verbs can take numerous types ofnouns. However. to
avoid confounding these two factors (the difficulty in acquiring a particular pattern and
the quantity of patterns to be acquired), this study will only target two collocational
patterns for each of the four target verbs.

Moreover, since participants surely knew make and take before the experiment, in
order to ensure that any Ieaming gain is a result of the experimental treatment rather than

other sources, pseudo-forms voke, dain, mand, nert were used to replace wield, warp,

98

 

make and take. The meaning and collocational behavior of the pseudo-word are identical
with its real counterpart, and the difference only lies in form. These pseudo-words were
created and matched with the real ones on the basis of the following criteria so that they
look more like real English words and the negative effect caused by the use of artificial
words could be minimized: first, each form is a stressed syllable of a real English word
(provoke 9 voke, disdain 9 dain, cornrnarul 9 mand, inert 9 nert); second, all forms
are of CVC or CVCC phonological structure, just like wield, warp, make and take; third,
matching of the pseudo-forms with the real ones was randomized so that it was different
across the four attentional conditions and across the two proficiency levels (see Table 3).

Table 3 Matching of the pseudo-forms with the real words

 

semantic memorization rule given rule given plus

processing for recall negative evidence

 

Level 2 Matching method Matching method Matching method Matching method

 

A B C D
Level 4 Matching method Matching method Matching method Matching method
C D A B

 

 

 

 

 

 

Note:

Matching method A: mand = wield, nert = warp, voke = take, dain = make
Matching method B: mand = make, nert = wield, voke = warp, dain = take
Matching method C: mand = take, nert = make, voke = wield, dain = warp

Matching method D: mand = warp, nert = take, voke = make, dain = wield

3.2.3. Material and procedure

The experimental procedure consisted of three parts: treatment (Day 1 to Day 3). test

99

(Day 4) and exit questionnaire (immediately after the test). In order to make the
following description of the material and procedure easier to understand, I will use Wield,
warp, make and take when illustrating examples, although in the material used by the
participants, these words were actually replaced by pseudo-forms mand, nert, voke and
dain.
3.2.3.1. Treatment

The treatment lasted three consecutive days. Every day all the participants were
asked to complete four tasks: (1) reading two passages that contained the target
collocations, (2) doing reading comprehension exercises, (3) doing an exercise that
helped them review the basic meaning ofthe four target verbs, (4) doing an exercise that
aimed to manipulate their attention to the target collocations in different ways. The
following part will describe each task in more detail:
(1) Task 1 -- reading passages

Task I asked participants to read two passages, each of which contained 230-341
words. Passage 1 included one collocational pattern of each of the four target verbs, and
Passage 2 included the other collocational pattern of each verb. Table 4 displays the

distribution of the targets in the passages:

100

Table 4

Distribution of target collocations in the reading passages

 

 

 

 

 

 

 

 

 

 

 

 

 

wield warp make take

Day 1 Passage wield sticks warp the floor make no take a look
1 changes
Passage wield control a warped mind make the claims take the lead
2

Day 2 Passage wield warp the front make a speech take a deep
1 weapons door breath
Passage wield power warp his make slow take the role
2 judgment progress

Day 3 Passage wield great warp make some take a step
1 influence bookshelves suggestions
Passage wield knives warp a make afull take power
2 erson's view recovery

 

Moreover, for all the participants, the basic meaning of the target verbs was

provided in English in the margin. The reason for the margin glossing is as follows:

vocabulary Ieaming typically proceeds from a word’s core meaning to the more refined

collocational knowledge (Henriksen. 1999; Schrnitt, 2000), so it is more meaningful to

analyse the collocation data from the students who have acquired the basic meaning; in

order to increase the amount of usable data in this study, margin glossing, a way of

promoting noticing of a new word and subsequent meaning acquisition, was used;

otherwise, learners might simply ignore the new word because it might not be crucial for

the comprehension of the whole text or the aim of a specific exercise. Table 5 shows the

basic meaning of the target verbs that was provided to the participants.

101

 

Table 5 Glossing for the target verbs

 

 

 

 

 

 

 

 

target word basic meaning

wield to have and/or use something
warp to change something for the worse
make to produce something

take _ to get something

 

Note: The reference for the basic meaning of the target verbs comes from Cambridge
International Dictionary of English, Longrnan Dictionary of Contemporary English, and
Encarta World English Dictionary.

The content of the passages was the same for all the attentional groups. But the

instruction for the memorization for recall. rule given and rule given plus negative

evidence groups asked learners to pay attention to the target verbs and their noun

collocates, which were highlighted in the passages. The semantic processing group did

not get this type of instruction or textual enhancement.

(2) Tasks 2 and 3

Task 2 asked learners to answer some questions related to the main idea ofthe

passages. Task 3 asked learners to match five words (including the four target verbs) with

five provided meanings. Like margin glossing, Task 3 aimed to encourage learners to

process the basic meaning of the target verbs and increase the amount of data that could

be used for further analysis of collocation. When learners did tasks 2-3, they could go

back to the corresponding passages for answers. The content of Tasks 2 and 3 was the

same for all the groups.

102

(3) Task 4 - attention manipulation exercise

After finishing Tasks 1, 2 and 3, learners of different attentional conditions were

asked to do different exercises, which exposed them to the target collocations again, yet

aimed to elicit different levels of attention to the targets.

For the semantic processing group, the sentences containing the target verbs were

picked out from the passages and presented to learners according to the order they

appeared in the passages. Participants were asked to read those sentences and find a word

with a particular meaning. The correct answer was a word after the target word. For

instance,

1. Rebels surrounded a govemment building. while wielding sticks and

throwing Stones. In this sentence. which word means "to cause something to

move rapidly through the air"?

 

2. Tips about how to prevent heat from warping the floor are available on the
lntemet.

In this sentence, which word refers to "a computer network"?

 

This design aims to make students read the target collocations again, but without paying

focal attention to the target collocational features.

103

For the memorization for recall group, the sentences containing the target verbs
were also picked out from the passages and presented to learners, but with those
containing the same target verb grouped together. Learners were asked to memorize the
highlighted target collocations for a later recall test, without using paper or pencil. For
example,

1. Rebels surrounded a government building, while wielding sticks and
throwing stones.

2. It makes me feel like someone is trying to wield control over us.

For the rule given group. leamers were provided with the rules underlying the target
collocations, as those stated in 3.2.2, and asked to find out how the target collocations
displayed in the passages matched these rules. For example.

I. wield

In the two passages, the verb wield is followed by TWO types of nouns:

Type 1: nouns referring to a tool or a piece of equipment that can be canied and
used.

Type 2: nouns related to the concept of power.

For instance,

104

1).. .rebels who surrounded a goverrnnent building, while wielding sticks and

throwing stones...

In this sentence, the noun following wield falls into Type , that is, nouns

about

 

2) It makes me feel like someone is trying to wield control over us.

In this sentence, the noun following wield falls into Type _, that is, nouns

about

 

Do you think the two types of nouns are somewhat related to each other? If yes,

in what way?

 

In order to make the exercise more interesting and challenging. on Day 2, the
collocational rules were not explicitly given to the rule given group. Instead, learners
were asked to recall and complete the target rules they learned on Day 1 with the help of
sample sentences from the passages. For example,

I. wield

In the passages, the verb wield is followed by TWO types of nouns:

Type 1: nouns referring to a __ or a piece of that can be carried

and

Type 2: nouns related to the concept of

 

105

Are the two types of nouns somewhat related to each other? If yes, in what way?

 

The following two sentences from the passages might help you fill in the
blanks above.

1)All kinds of horrible scenes occurred to her mind: men wielding weapons or
taking away victims by force

2) It seems that he always wants to wield power over me.

For the rule given plus negative evidence group, in addition to the treatment like that
in the rule given group, leamers were also provided with negative evidence which
showed that some noun collocates were inappropriate for the target verb, although they
were semantically compatible with the basic meaning ofthe verb. For wield and warp,
learners were informed that, generally speaking, when the verb is used in the basic
meaning as shown in the passages, it is followed ONLY by two types of nouns. (This type
of statement was not applicable to make and take, due to their collocationally active
nature.) Moreover, for all the four target verbs, participants were told that a particular
sentence was bad and asked to think about the reason. For example,

Notice you CANNOT say:

*He wields a new method to teach his students.

106

Why do you think is this sentence bad?

 

Table 6 summarizes the negative evidence presented to the rule given plus negative

evidence group from Day 1 to Day 3:

Table 6 Summary of negative evidence

 

 

 

 

 

 

 

 

 

 

wield warp make take
Day 1 *wield low *warp his health *make two novels *take a lot of
prices comment from
Day 2 *wield $100 *his performance *make a lot of *take any response
has been warped articles from
Day 3 *wield a new *the environment... *make a very good *take any helpfrom
method has been warped idea

 

To sum up, during the treatment all the participants encountered a target verb's
collocations 4 times every day (2 times in reading passages and 2 in the attention
manipulation exercise). The difference lay in the level of attention to the target
collocations.

Before the material was used by the participants, it was proofread by three NSs of
English, with the pseudo-forms re-written as their real counterparts. Any parts regarded
as inappropriate by more than one NS was modified according to their suggestions.
Appendices A, B, C and D3 show the three-day treatment material presented to the
semantic processing, memorization for recall, rule given, and rule given plus negative

evidence groups respectively. Table 7 is a summary of the daily treatment procedure from

 

3 The material shown in the appendices is the version in which mand. nert, voke and dain match wield.
warp, take and make respectively.

107

Day 1 to Day 3.

Table 7 Daily treatment procedure (Day 1 to Day 3)

 

 

 

 

 

 

Treatment semantic memorization for rule given rule given plus
processing recall negative evidence

7 minutes 1. Read two 1. Read two 1. Read two 1. Read two
passages passages passages passages

(instructed to pay (instructed to pay (instructed to pay
attention to the attention to the attention to the
highlighted target highlighted target highlighted target
collocations) collocations) collocations)

5 minutes 2. Answer 2. Answer 2. Answer 2. Answer
comprehension comprehension comprehension comprehension
questions questions questions questions

1 minute 3. Match the 3. Match the target 3. Match the target 3. Match the target
target verbs verbs with their verbs with their verbs with their
with their corresponding corresponding corresponding
corresponding meaning meaning meaning
meaning

5 minutes 4. Find a word 4. Memorize the 4.a) Rules 4.a) Rules
with a particular target collocations underlying the underlying the

meaning (not
the target verb)
from the
sentences
containing the
target

collocation

 

 

for a later recall

ICSI

 

target collocations
are presented;

b) Find out how
the use of the
target verbs in the
passages matches

these rules.

 

target collocations
are presented;

b) Find out how
the use of the
target verbs in the
passages matches
these rules;

c) NE is provided.

 

3.2.3.2. Test and scoring

The day after the completion of the treatment, that is, on Day 4, participants did a

test (see Appendix E), which consisted of three parts, testing three aspects of Ieaming

108

 

gains: (1) knowledge of the basic meaning of the target verbs, (2) collocation production,
(3) collocation judgment.

Part I of the test. Part I of the test was a multiple-choice exercise, which asked
learners to choose a word or phrase that had a meaning closest to that of the target verb,
for example,

warp __ a) to make (something) worse b) to divide (something) into parts

c) to look for d) to go through
The maximum score for Part I was 4. with 1 point for each target verb. This part aims to
see whether learners had acquired the basic meaning of the target verbs. If and only if the
basic meaning was acquired, the acquisition of the target collocational pattems (i.e., data
from Parts 11 and III of the test) was further analysed.

Part II of the test. Part II asked learners to write down as many noun collocates as
possible for each of the target verbs within 4 minutes. Leamers' production was divided
into four types: (1) old collocates, that is, collocates that appeared in the passages read by
the participants; (2) new target-like collocates, that is, collocates that followed the target
rules and did not appear in the passages; (3) non-target-Iike collocates, that is, collocates
that did not follow the target rules; (4) semi-target-like collocates, that is, collocates that

were half way between consistent and inconsistent with the target rules. An English

109

teacher in. China and the primary researcher (both NNSs of English) worked on the
classification of the IL production. When their opinion differed, another English teacher
in China, who was also a NNS of English, was consulted, and the majority opinion was
adopted. The number of collocates for each type was calculated and compared across the
four attentional conditions.

Moreover, as mentioned in the literature review, collocational rules can be
probabilistic; for make and take. their possible collocates are far beyond the types
targeted in the current study. As a result, in the test, there were occasions when the IL
production followed the target rules but was not accepted or commonly used by N85 or
when the production did not follow the target rules but was accepted by NSs. Therefore,
two NSs of English were asked to determine ifthe collocates produced were (1) good, (2)
bad, or ( 3) OK but not the best way to say it, with the pseudo-forms replaced by their real
counterparts. When the judgment of the two NSs differed, another NS was consulted and
the majority opinion was adopted.

Part III of the test. Part III asked learners to judge whether a sentence was a possible

English sentence or not. If bad, leamers were supposed to identify the error4. (When

 

" In the original design, Part III of the test also asked learners to translate the sentences into Chinese if they
considered them good. However, in a similar study on L2 verb-noun collocations by Fan (2004) (more
details about this study in Chapter 5), it was found that once learners acquired the basic meaning of the
target verbs, they did not have difficulty in translating the sentences they correctly judged as good. So in
order to reduce experiment time and subject fatigue, the translation part was omitted in the current design.

110

learners did Part 111, they were not allowed to change the answers in Part I or II.) All the

sentences contained a good or bad verb-noun phrase of one target verb. For each of the 4

target verbs, there were 5 sentences, 2 with a good collocation, 2 with an inappropriate

collocation and l with a nonsense collocation.

The good collocations represented the two collocational patterns displayed in the

treatment input, but were not the same as those that appeared in the passages read by the

participants (see Table 8).

Table 8 Sentences with a good collocation in the judgment test

 

wield He is so well trained that he can wield these heavy swords with ease.

He doesn't seem to know how to wield authority.

 

warp The wood is warped by damp conditions.

Years of living alone have warped her personality.

 

make He made the shortest speech 1 have ever heard.

.lack made a shift in his political views.

 

 

take After dinner, Mary took a walk in the park.

 

 

The actor wanted to take the part of Hamlet.

 

The bad collocations contained in the sentences for judgment were deviant from the
target rules, although the noun collocate was compatible with the basic meaning of the
target verb. Moreover, those bad collocations were semantically parallel to the negative
evidence provided to the rule given plus negative evidence group during the treatment, in
the sense that the tested bad noun collocate and the one shown in the negative evidence

had a synonymous or hyponymy relationship, or belonged to the same lexical field. For

111

instance, the rule given plus negative evidence group was informed that they could not

say *He wields a new method to teach his student, and the judgment test included this

parallel sentence: *T he company wielded a new marketing strategy to sell its product. In

the two sentences, the inappropriate noun collocates, method and strategv, were

near-synonyms. Table 9 displays the sentences with a bad collocation in thejudgment

1681.

Table 9 Sentences with a bad collocation in the judgment test

 

wield *The girl wielded several methods to solve the problem.

* The man wielded his own money to support these poor children.

 

warp *The company's operations have been warped since the new manager came.
*The air quality in this area has been warped by the pollution from the

factories.

 

make *Mary made great interest in music.

*The writer made two books last year.

 

 

take *He didn't take any reply from his girlfriend.

 

 

*He didn't take any support from his family.

 

To sum up, there were altogether 20 sentences for judgment. 8 with a good

collocation, 8 with an inappropriate collocation and 4 with a nonsense collocation. When

learners made a correct judgment, judging a good collocation as good or judging a bad

collocation as bad for the reason intended by the researcher, that is, recognizing the

inappropriateness of the target verb-noun combination, they got 1 point, otherwise 0 point.

The maximum score for the judgment of good collocations and bad collocations was both

8 points.

112

Before the experiment, all the sentences for judgment, with the pseudo-forms

re-written as their counterparts in real English, were read by 2 N 85 of English to ensure

that they fell into the right category, that is, they are sentences with a good or bad

collocation.

3.2.3.3. Exit questionnaire

Immediately after the test, participants were asked to answer some questions in

English or Chinese. There were altogether three questions and each attentional group

answered some of them.

The first question was for the semantic processing group only: "When you read the

passages in the past three days, did you pay special attention to the nouns or noun phrases

that follow nert, voke. mand or dain?" This aims to see whether it is true that learners in

that condition did not pay any focal attention to the target collocations as supposed by the

researcher.

The second question was for the semantic processing and memorization for recall

groups: "When you read the passages, did you feel that nert, voke, mand and dain had

always been followed by particular TYPES of nouns? If yes, please write down what you

noticed for each of the four words." This aims to see to what extent learners under the

two non-rule-oriented conditions could detect the underlying collocational rules on their

113

own.

The third question was for the rule given and rule given plus negative evidence
groups: "In the past three days, you were informed about what TYPES of nouns follow
nert, voke, mand and dain. Please write down these patterns." This aims to see how much
rule knowledge is actually gained by learners under the two rule-oriented conditions, and
how the recall of rules is related to leamers' performance in the test.

3.3. Research hypotheses

There are altogether 6 research hypotheses concerning the effect of attention.
Hypothesis 1 is about acquisition ofthe basic meaning ofthe target verbs; hypotheses 2
to 4 are related to the production of collocates; hypotheses 5 and 6 are about judgment of
collocational well-formedness.

vaothesis 1. Attention to collocational features does not have a significant main
effect on acquisition of the target verbs' basic meaning. During the 3-day treatment, all
the participants, whatever their level of attention to the target collocations, encountered
each of the four target verbs for at least 5 times per day and also did a daily exercise
focusing on the basic meaning of these verbs. These encounters and exercises are very
likely sufficient for learners to notice the four words (although not necessarily their

collocational features) and retain the form-meaning associations.

114

Hypothesis 2. In terms of producing old collocates for the target verbs (i.e.,
collocates that appeared in the passages), the memorization for recall, rule given and rule
given plus negative evidence groups will significantly outperform the semantic
processing group. And this trend will be found among students of both proficiency levels
and for both easy and difficult collocational rules.

The reasons are as follows: The treatment in the memorization for recall condition
directly encourages the storage of target collocations in LTM. The rules provided in the
rule given and rule given plus negative evidence conditions can help learners organize the
target phrases in the input, elicit focal attention to them and thus promote memorization,
although participants might not intentionally try to memorize those target collocations.
For lower-level students. very complex collocational rules. that is, those of make and take,
might be less effective as an input organizer, but the rule-oriented treatment can still
induce focal attention to the target collocations and increase chances of LTM storage. As
to the participants in the semantic processing condition, their attention to the target
collocations is very likely peripheral (see previous discussion in 3.1 .), and as a result,
their memory trace of the target collocations tends to be the weakest.

Hypothesis 3. As to new target-like collocates (i.e., collocates that follow the target

rules and did not appear in the passages), generally speaking, the rule given and rule

115

given plus negative evidence groups will write significantly more of this type of

collocates than the memorization for recall and semantic processing groups. However, for

lower-level students Ieaming complex collocational patterns, that is, those of make and

take, there might be no significant difference between the two rule-oriented groups and

the two non-rule-oriented groups.

This hypothesis is based on the following rationale Rules presented to the rule given

and rule given plus negative evidence groups can serve as an aid for making

generalizations to novel cases. As to the memorization for recall and semantic processing

groups, it is quite hard for them to find the correct target rules. The memorization for

recall group might detect certain underlying rules as a result of repeating the phrases, but

the amount of input provided to them might not be enough for the learners to induce the

correct target rules. The semantic processing group might not sense any recurrent patterns

at all because of a lack of focal attention to the targets. Therefore, the two rule-oriented

groups will enjoy more advantages than the two non-rule-oriented groups in terms of

making target-like extensions. However, for lower-level students, complex rules of make

and take might be too hard for them to acquire and thus less useful as the basis for

transfer, so in this case, the superiority of the two rule-given conditions might not exist.

In other words, I view the collocational rules of make and take as sitting on the "too

116

confusing" side of the complexity continuum for the lower-level participants (see
literature review 2.5.3).

Hypothesis 4. In terms of non-target-like collocates, that is, collocates that do not
follow the target rules, generally speakin g, the rule given plus negative evidence group
will write significantly fewer of this type of collocates than the rule given group, which in
turn will write fewer than the memorization for recall and semantic processing groups.
However, the significant difference among the four attentional conditions might not hold
for lower-level students Ieaming complex collocational patterns.

The following are the reasons for this hypothesis. Positive evidence alone only
explicitly tells learners which sentences are possible. but not which sentences are
impossible. Negative evidence can fill in this gap, help learners narrow down their
hypothesis and be more conservative and cautious about a word's possible collocates. so
the rule given plus negative evidence group will produce the fewest non-target-like
extensions. For the rule given group, although there is no explicit information about what
is not allowed, the revealed recurrent patterns might serve as a type of indirect negative
evidence, which makes learners somewhat prefer to observe the patterns in the input. As
to the two non-rule-oriented groups, due to the failure in discovering the target rules, very

likely there will be more production that violates the target rules. However, it has to be

117

noted that, as previously mentioned, for lower-level students, the complex collocational
rules of make and take are hard to acquire, so in this case the rule-oriented treatment
might be less effective and the difference among the four attentional conditions can
disappear.

Hypothesis 5. In terms of judging good collocations, attention to target collocational
features does not have a main effect. As stated in Hypothesis 1, all the participants will
have acquired the basic meaning of the four target verbs, so the good collocations, in
which the noun collocate is semantically compatible with the basic meaning of the target
verb, tend to be accepted by all the learners, and there won't be significant difference
among the four attentional conditions.

Hypothesis 6. In temis of judging bad collocations. generally speaking. the rule
given plus negative evidence group will outperform the rule given group, which in turn
will outperform the memorization for recall and semantic processing groups. However,
again the significant difference among the four attentional conditions might not hold for
lower-level students Ieaming complex collocational patterns. The reasons for this
hypothesis are the same as that stated in Hypothesis 4.

Table 10 presents a summary of these research hypotheses.

118

Table 10 A summary of research hypotheses

 

Tasks in the test

The effect of attention to the target collocations

 

Acquisition of the
meaning of the

target verbs

no main effect

 

Production of old

collocates

[M], [R], [R+NE] >S

 

Production of
new target-like

collocates

[R], [R+NE] > [M]. [S]
but for lower-level participants Ieaming complex rules, no main
effect

 

 

Avoidance of
non-target-like
collocates in

production

[R+NE] > [R]> [M], [S]
but for lower-level participants Ieaming complex rules, no main
effect

 

Judgment of good

no main effect

 

 

 

collocations

Judgment of bad [R+NE] > [R]> [M], [S]

collocations but for lower-level participants Ieaming complex rules, no main
effect

Note:

[S] = semantic processing condition, [M] = memorization for recall condition,

[R] = rule given condition, [R+NE] = rule given plus negative evidence condition;

The ">" mark between conditions indicates more superior performance, and the

9

indicates no significant difference.

119

 

H H

CHAPTER 4

Results

This chapter will report the results of the experiment in seven parts:

(1) acquisition of the basic meaning of the target verbs;

(2) production of collocates of the target verbs;

( 3) judgment of collocations of the target verbs;

(4) summary of the test results;

(5) detection of target rules by the semantic processing and memorization for recall

groups;

(6) recall of target rules by the rule given and rule given plus negative evidence

groups;

(7) participants' performance during the three—day treatment.

The first four parts focus on the results of the experiment. The fifth and sixth parts

are related to the exit questionnaire. And the last part aims to characterize leamers'

answers to some of the major exercises during the three-day treatment, which can be

helpful to our interpretation of the test results. In order to simplify the report, the target

words in the experiment are written as wield, warp, make and take in this chapter rather

than the original pseudo-forms.

120

4.1. Acquisition of the basic meaning of the target verbs

Research Hypothesis 1 states that attention to collocational features does not have a
significant main effect on acquisition of the target verbs' basic meaning. This hypothesis
is confirmed by the result. In the first part of the test, which examined knowledge of the
basic meaning of the target verbs. all the participants got 4 points, the maximum score. In
other words, after the treatment, leamers in all the attentional conditions acquired the
basic meaning of the target verbs, and different levels of attention to the target
collocations did not produce significantly different results in terms of Ieaming basic word
meaning.
4.2. Production of collocates of the target verbs

The second part of the test asked participants to produce as many noun collocates as

possible for the four target verbs within 4 minutes. Perhaps due to a misunderstanding of
the task requirement, 3 learners wrote down the meaning or collocational rules ofthe
target verbs rather than possible collocates. So their data could not be used and the actual
pool for this part was 91 participants. The results of the collocate production will be
reported according to this outline:

(1) classification of the IL production: procedure and method;

(2) an overall picture of the collocate production;

121

(3) production of old collocates;

(4) production of new collocates -- overall comparison;

(5) production of new target-like collocates;

(6) production of non-target-like collocates;

(7) production of semi-target-like collocates;

(8) summary of collocate production;

(9) NSjudgment ofIL production.

4.2.1. Classification of the IL production: procedure and method

As described in Chapter 3, leamers' production of collocates was classified into old

and new collocates according to whether they appeared in the passages read by the

participants or not. And the new collocates were further sub-categorized into three types:

(i) new target-like collocates (new collocates that follow the target rules); (ii)

non-target-like collocates (collocates that do not follow the target rules); (iii)

semi-target-like collocates (collocates that are half-way between consistent and

inconsistent with the target rules). This section will describe the method of assigning the

IL production to each category, which was determined on the basis of the available data.

The first problem encountered in the process of classification was how to deal with

the formal errors produced by some learners. When writing the noun collocates, some

122

learners misspelled a word or misused a morpheme, for example. occasion was rendered
as *occassion, speech as *speach, comprehension as *compehension, excitement as
*exciteness, suggestion as *suggests, and alternation as *altering. Since the major aim of
the production test was to examine participants' knowledge about what MEANING
determined the compatibility of nouns with the target verb, formal errors were not taken
into account. There was only one occasion when it was quite hard to guess what the
learner meant with a misspelled word. that is, "ffance, which was produced for take. This
single case was excluded from the analysis.
After fomial errors were modified, the collocates were divided into old and new.

This was not a difficult process. The more challenging task was the further classification
ofthe new collocates. which was done by two raters. with an interrater reliability
calculated at 97.26% (320 out of 329 types of new collocates). As to the 9 disagreements,
as described in Chapter 3, a third rater was consulted and the majority opinion was
adopted. The following were the criteria used by the raters to make the classification.

( 1) Collocates of wield:

Rule 1 --- nouns referring to a concrete tool or piece of equipment that can be carried
and used.

If the noun produced for wield, like the target nouns in the reading passages, had

123

both of the following semantic features, it was regarded as consistent with the rule (i.e.,
target-like): [which is typically used as a weapon or a tool for making or repairing
something], [which can be carried, that is, used with one's hands]. For example, gun,
hammer, sword, saw, screwdriver.

Moreover, there were two groups of nouns, which were hard to judge whether they
were completely consistent or inconsistent with this target rule. In the first group, the
nouns denote certain concrete objects that can be carried and are not typically used as
weapons or tools, but may be used for these functions, for example, pen, pencil, book,
umbrella, cup, glass. These nouns were labeled as semi-target-like (1), that is, nouns
sharing partial semantic features with the target noun collocates (( l ) indicates the first
type of semi-target-like collocates). In the second group, the nouns produced reflect some
general terms that semantically dominate the target-like nouns, or are superordinates of
the target-like nouns, for instance, tool, equipment, machine, apparatus, instrument,
device. This group of nouns was classified as semi-target-like (2) ((2) indicates the
second type of semi-target-like collocates).

Rule 2 -- nouns associated with the concept of power.

If the noun produced for wield met ONE of the following standards, it was

considered as consistent with the target rule: (i) the noun is a synonym of one of the

124

target noun collocates in the reading passages (i.e., control, power, influence), for
example, force, strength, authority, energy (Chamber's Dictionary of Synonyms and
Antonyms was used for reference); (ii) the noun's definition involves the concept of power,
for example, violence (definition: extreme force), leadership (definition: the action of
leading a group of people or an organization) (Cambridge International Dictionary of
English and The New Oxford Dictionary of English were used for reference).

If the noun produced for wield was consistent with neither of the two above rules, it
was categorized as non-target-like, for example, car. classroom. library, money, water,
essay, source, idea, time, method, thought, relationship, chance, impression.

(2) Collocates of warp:

Rule 1 -- nouns referring to an object that can be damaged by heat or water.

When asked to write down noun collocates for warp, the participants produced
many nouns refening to concrete objects. It was found that all these objects could be
damaged by heat or water. However, some of the nouns (e. g., desk, shelf table, chair)
obviously have a closer semantic relationship with the three target nouns in the reading
passages (i.e., bookshelf floor, door), in that all of them are very often made of wood.
Some of the nouns produced by the participants (e. g., book, food, paper, road) do not

share this semantic feature with the target nouns. In order to differentiate the two types of

125

production, nouns with ALL of the following semantic features were categorized as

target-like: [concrete objects], [which can be damaged by heat or water], [which is often

made of wood], for example, desk, shelf table. chair, stick, board, cabinet (see

Cambridge Dictionary of American English). Nouns that met all the features except the

third one were categorized as semi-target-like (l), for example. book, paper, car, picture,

plastic, steel, water.

Moreover, like wield, participants also produced some general terms which were

difficult to judge whether they completely complied with the target rule or not, for

example, thing, object. These general nouns were classified as semi-target-like (2).

Rule 2 -- nouns related to one's mind.

lfthe noun produced for warp denoted or involved certain mental activities, it was

regarded as consistent with this rule, for example, decision, emotion, mood, thought.

belief feeling.

If the noun produced for warp did not meet either of the two above rules, it was

regarded as non-target-like, for example, atmosphere, condition, consequence,

environment, performance, situation.

(3) Collocates of make

Rule 1 -- nouns related to speech action.

126

Rule 2 -- nouns related to changes.

If the noun produced for make denoted speech action (e.g., talk. advice, lecture,
comment, statement), or expressed a kind of change (e. g., alteration, destruction,
revolution, improvement), it was considered as target-like. Otherwise, it was
non-target-like (e.g., idea, machine, influence, product, book. method, power: story).

(4) Collocates of take

Rule 1 -- nouns referring to physical actions.

Rule 2 -- nouns referring to a particular role, especially one of leadership.

If the noun produced for take described certain physical action (e.g., blowfall, gaze,
jump, kick. sigh. touch), or was a synonym of lead, role. or power. three of the target
nouns in the reading passages (e.0., apart (in a play). leaders/rip), it was classified as
target-like. Otherwise, it was non-target-like (e. g., monev. book. help. idea. chancejoh,
support, box, car).

To sum up, the collocates produced by participants were divided into old, new
target-like, non-target—like, semi-target-like (1) and semi-target-like (2). Before looking at
the production details of each sub-type of collocate, I will first present a general picture
of the collocate production.

4.2.2. An overall picture of the collocate production

127

4.2.2.1. Number of collocates

Table 11 and Figure 1 show the mean number of collocates produced by participants
by complexity * level * attention, in which "simple" represents the simple collocational
behavior of wield and warp, "complex" represents the complex collocational behavior of
make and take, and [S], [M], [R] and [R+NE] represent the semantic processing,
memorization for recall, rule given and rule given plus negative evidence conditions

respectively. These abbreviations will also be used in the other tables and figures.

Table 11 Descriptive statistics of quantity of collocate production by

complexity * level * attention

 

 

 

 

 

 

 

 

 

 

 

 

Complexity _ Level Attention i[Mean l Std. deviation ; N
Simple T2 [S] i5.9167 1 4.20948 .72“
:[M] 4.5455 é2.16165 ~ 11
[R] 7.5000 3.25137 8
: [R+NE] : 5.0000 * 2.05480 ; 10
4 . [5] , 5.3077 7 4.42313 13
’ [M] 1 5.6667 1.66969 1 12
1 [R] 8.7500 f 2.52713 12
d...-. __ ._ g 1 [R+NE] 1 10.0769 5 1.84669 1 13
Complex . 2 i [S] 1 5.8333 i 3.24271 1 12
j [M] [4.1818 §199089 ? 11
. g [R] 16.3750 5 4.53360 8
g 1 [13+er] 14.4000 ' 1.64655 10
14 [[sjmfl " 14.5385 3.84308 [13
i . [M] 14.8333 1.80067 i 12
‘ [R] 4.7500 2.34036 1 12
[R+NE] 6.7692 2.83296 , 13

 

 

 

Note: mean: mean number of collocates produced;

N= number of participants.

128

Mean number of collocates produced by participants

 

 

§18
§16
:14.
812 ,
‘610 loomplexrules
E 8 ~' 7 ‘Ds’nplerules ‘
s 6 ’
a 4 ,
‘ 1': 2
8 o
z

 

Level " attention conditions

Figure 1 Mean number of collocates produced by participants

An analysis of variance (ANOVA) of the number of collocates was performed, with
attention and level as two between-subjects factors and complexity as a within-subjects
factor (All the statistical tests were performed on SPSS 13.0 statistical software and the
alpha level was set at 0.05). There were three major findings from the test:

(1) Level did not have a significant main effect (F (1, 83) =2.239, p=.138. partial Eta
squared (np2)= .026);

(2) Complexity had a significant main effect (F (1, 83) = 38.751, p= .000,'r]p2= .318),
with the collocates produced for wield and warp (simple rules) being significantly more
than the collocates for make and take (complex rules);

(3) Attention had a marginally significant effect (F (3, 83) =2.697, p=.051, 111:2

129

= .089). As shown in Table 11 and Figure l, in general, the rule given and rule given plus
negative evidence groups tended to produce the greatest number of collocates, which was
followed by semantic processing group. Mernorization for recall group produced the
fewest collocates. Post-hoe Least Significant Difference (LSD) test revealed significant
difference between memorizationfor recall and rule given, between memorization for
recall and rule given plus negative evidence, with the memorization group producing

significantly fewer collocates than the two rule-oriented groups (see Table 12).

Table 12 Post-hoe LSD test of the effect of attention on the number of collocates

 

 

 

 

 

Attention I Attention Mean difference l Std. Error l Sig.
[S] [M] T .5539 f .78984 [ .485
, [R] ‘ -14450 . .82011 1 .082
_ [R+NE] ‘ 4.4243 .78984 f .075
[M] M “I I [S] I‘M—.5539 T "T789811“ "T7185 M
i [R] —1.9989(*_) .83581 - .019
? [R+NE] -l.9783(*) .80613 5 .016
[R] i I [S] i T ' 1.4450?" I “1”.820l1 8.7.082“
E [M] ‘ 1.9989(*) * .83581 .019
' [R+NE] g .0207 , .83581 1 .980
[RINE] “T[[sij—mf14243T—W I ’7 .78984 Eli—.075—
‘ [M] " 1.9783(*) ’ .80613 .016
j [R] , l-.0207 .83581 ] .980

 

 

 

Note: mean: mean number of collocates produced

* The mean difference is significant at the .05 level.

4.2.2.2. Production of collocates for each target verb

Table 13 shows the total tokens/ types of collocates produced by all participants for

each target verb. These statistics provide us with a general impression of the effect of

130

complexity in acquiring collocational features. Generally speaking, more tokens of old

and new target-like collocates were produced for wield and warp (simple rules) than for

make and take (complex rules), and more tokens of non-target-like collocates were

produced for make and take (complex rules) than for wield and warp (simple rules). A

more detailed statistical analysis will be made in 4.2.3 -4.2.7.

Table 13 Production of collocates for each target verb

 

 

 

 

 

 

 

 

 

Target tokens/ types of collocates produced by all participants Total
verbs

Old New collocates

collocates Target-like Non-target-like Semi(l) Semi(2) Total
wield 138/6 44/15 67/39 23/7 33/6 167/67 305/73
warp 136/6 83/32 55/46 19/12 6/4 163/94 299/100
make 124/6 33/16 85/60 0/0 0/0 1 18/76 242/82
take 97/6 29/15 105/77 0/0 0/0 134/92 231/98
Total 495 189 312 42 39 582 1077
tokens

 

 

 

 

 

 

 

 

 

Note: semi = semi-target-like collocates

4.2.3. Production of old collocates

According to Research Hypothesis 2, in terms of producing old collocates for the
target verbs (i.e., collocates that appeared in the passages), the memorization for recall,
rule given and rule given plus negative evidence groups will significantly outperform the
semantic processing group, and this trend will be found among students of both
proficiency levels and for all the target verbs. This prediction is partially confirmed.

Table 14 presents the descriptive statistics of the production of old collocates by

131

complexity * level * attention.

Table 14 Descriptive statistics of old collocates by complexity * level * attention

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Complexity 1 Level Attention Mean Std. deviation N
Simple [2 [S] .7500 .96531 12
l [M] 12.5455 2.33939 . 11
; i[R] . 4.0000 3.02372 '8
a [R+NE] 112.4000 4 1.77639 10
_ 1' Total ] 2.2683 [ 2.29155 41
[4 I [S] ; 1.2308 1 .83205 13
: : [M] 14.0000 1 1.90693 . 12
i [R] 14.0833 1 1.62135 1 12
! [R+NE] 15.2308 1 1.96443 13
. , Total [ 3.6200 é 2.19359 50
' Total 1, [S] g 1.0000 i .91287 [25
.[M] '33043 32.20402 '23
7 [R] 4.0500 2.21181 i 20
" [R+NE] , 4.0000 ‘ 2.33550 1 23
Total 3.0110 2.32615 91
2:83.81... ‘ ’2? I 7"[s] 10000 ”1:41-421"? Ti 12 I
: [M] 2.4545 42.25227 1 11
[R] ’ 3.6250 . 3.77728 , 8
: [R+NE] : 1.8000 ‘ 1.93218 1 10
i Total ' 2.0976 ; 2.45769 ' 41
[TUTTI—[ls]— _ i T3846“? 1650445 I: i3“
I [M] 3 3.3333 1 1.87487 12
1 [R] :30000 1 1.85864 . 12
; [R+NE] [4.1538 2.23033 1 13
[ [Total ’ 2.7000 2.23379 50
| Total [[8] __..__ [.6800 1.10755 725*
1 [M] 2.9130 2.06514 1 23
l [R] 3.2500 2.71206 f 20
! [R+NE] 3.1304 2.37992 1 23
i Total 2.4286 2.34352 91

 

 

 

Note: mean: mean number of old collocates produced

132

N= number of participants.

An ANOVA of the number of old collocates was performed, with level and attention
as between-subjects factors and complexity as a within-subjects factor. The statistical test
showed five major results:

(1) Proficiency level had a significant main effect (F (l,83)=5.243, p=.025, npz
=.059), with learners of Level 4 producing significantly more number of old collocates
than learners of Level 2;

(2) Collocational complexity had a significant main effect (F (l. 83)=10.208, p=.002,
hp: = .1 10), with learners producing significantly more number ofold collocates for wield
and warp (simple rules) than for make and take (complex rules). As previously shown in
Table 13, the total number of old collocates produced for wield, warp, make and take
were 138, 136, 124 and 97 respectively. Table 15 shows what old collocates were
produced for each target verb and how many participants produced them.

Table 15 Production of old collocates for each target verb

 

Target Old collocate (number of participants who produced that collocate)
verbs
wield [power (44) stick(27) knife (24) weapon (21) control (15) influence (7)
warp floor (45) mind(32) bookshelf (27) door (19) view (7) judgment (6)
make speech (47) change(24) suggestion (22) progress recovery (9) claim (9)

(13)
take power (23) step (22) look (16) role (15) breath (I 1) lead (10)

 

 

 

 

 

 

 

 

 

 

 

 

 

(3) Attention had a significant main effect (F (3, 83): 12.769, p= .000, 1']p2 = .316);

133

To locate the difference among the four attentional conditions, a post-hoe Scheffé

test was performed, which revealed a significant difference between semantic processing

and each of the other three conditions (i.e., memory for recall, rule given and rule given

plus negative evidence), with the semantic processing group producing significantly

fewer old collocates than the other groups (see Table 16). This is exactly what was

predicted by Research Hypothesis 2.

Table 16 Post-hoe Scheffé test of the effect of attention on the production of old

 

 

collocates

Attention iAttention iMean difference [ Std. Error 1 Sig.
[S] :[M] -2.2687(*) 1 .50877 , .000

[R] " -2.8100(*) j .52827 , .000

3 [R+NE] 9 —2.7252(*) = .50877 '1 .000
[M]? T ”I T? [8]” WW”? 2268—7111 I ' I T” iii—50877" ind“

j [R] f —.5413 § .53838 .799
2-.- 3-1113113].-- _____,:i5_65 , L 41929--. ____.;$§L
[R] f [S] 2.8100(*) : .52827 , .000

3 [M] g .5413 .53838 .799

% [R+NE] 1 .0848 _ .53838 .999
[Rt-NEE— T i [SI—m T VIT—I217252Fffiv I MTG—5530877 ._ "T600“

1 [M] 3 .4565 . .51926 1 .856

i [R] [-0848 l .53838 g .999

 

 

 

Note: mean= mean number of old collocates produced;

* The mean difference is significant at the .05 level.

There was another interesting phenomenon. Although as shown in Table 15, all the

target nouns in the reading passages were recalled by Ieamers, some of the nouns, that is,

(I) mind (collocate for warp), (2) recovery, progress, claim (collocates for make), (3)

I34

lead, role (collocates for take) were only written by learners from memorization for recall,
rule given and rule given plus negative evidence groups; and none of the participants
from the semantic processing group recalled these nouns.

(4) The interaction between attention and collocational complexity was not
significant (F (3, 83) = .578, p= .613, 11,2 = .020);

(5) The interaction between attention and proficiency level was significant (F (3, 83)

= 3.118, p=.030, npz = .101) (See Figures 2 and 3);

Production of old collocates (Level 2)

9
U“
l
1
l

.03
courts

---. .. ' +simples rules
_ . +oomplex rules

-§

 

Mean number of old
collocates
N
01 N U!
l

 

 

[S] [M] [R] IR+ NE]
Attentional conditions

Figure 2. Production of old collocates (Level 2)

135

5 z

4, f—fl/J i “2“,- a___ g

3 }\./ i i-Q—simple rules
i

2

// , i—I—oomplex rules E

 

 

 

 

collocates

Mean number of old

 

 

[S] [M] [R] iR+ NE]
Attentional conditiions

Figure 3. Production of old collocates (Level 4)

Since the effect of attention was mediated by level, separate ANOVAs were
conducted for the results of Level 2 and Level 4, with attention as a between-subjects
factor and complexity as a within-subjects factor.

For Level 2 students, it was found that (l) attention had a main effect (F(3, 37) =
3.271, p= .032, npz = .210), (2) complexity did not have a main effect (F(l , 37) =.777,
p=.384, npz = .021), (3) attention and complexity did not have significant interaction (F(3,
37) = .679, p=.571, npz = .052). A post-hoc Scheffé test of attention demonstrated
significant difference between rule given and semantic processing conditions, with the
former group significantly outperforming the latter; but there was no significant
difference between memorization for recall and semantic processing, or between rule
given plus negative evidence and semantic processing (see Table 17). Moreover, as shown

136

in Figures 2 and 3, the increasing trend from rule given to rule given plus negative
evidence group, which was found with Level 4, did not hold for Level 2. Instead, for
Level 2, the rule given plus negative evidence group produced fewer old collocates than

the rule given group and even the memorization for recall group.

Table 17 Post-hoc Scheffé test of the effect of attention on production of old
collocates (Level 2)

 

 

 

 

 

 

 

 

 

 

 

Attention Attention Mean difference Std. error Sig.
[S] [M] -1.6250 .87326 .340
[R] -2.9375(*) .95487 .036

[R+NE] -12250 .89575 .604

[M] [S] 1 1.6250 .87326 .340
[R] -13125 .97208 .614

[R+NE] [.4000 ' .91407 1.979

[R] [S] 12.93750) .95487 i .036
[M] 1 1.3125 .97208 | .614

i [R+NE] 1.7125 .99233 .407

[R+NE] ([3] __ 71.2250 .89575 [.604
i [M] l-.4000 .91407 i .979

g [R] - -1.7125 .99233 [.407

 

 

 

Note: mean = mean number of old collocates

* The mean difference is significant at the .05 level.

For Level 4 students, it was found that ( l) attention had a significant main effect (F(3,
46) = 17.237, p=.000, npz = .529); (2) complexity also had a main effect (F (1, 46) =
13.066, p= .001 , 1192 = .221); (3) the effect of attention was not mediated by complexity
(F(3, 46) = .153, p=.927, npz = .010). A post-hoc Scheffé test of attention showed a

pattern that was consistent with the overall result, that is, the semantic processing group

137

produced significantly fewer old collocates than each of the other three conditions (see

Table 18).

Table 18 Post-hoe Scheffé test of the effect of attention on production of old
collocates (Level 4)

 

 

 

 

 

 

‘ Attention Attention Mean difference Std. error Sig.
[S] [M] -2.8590(*) .5766] .000
[R] -2.7340(*) .5766] .000

[R+NE] -3.8846(*) .56496 .000

[M] [S] 2.8590(*) .5766] .000
[R] .1250 .58803 .997

[R+NE] -1.0256 .5766] .378

[R] [S] 2.7340(*) .57661 .000
[M] -.1250 .58803 .997

[R+NE] -1.1506 .5766] .277

[R+NE] [S] 3.8846(*) .56496 .000
[M] 1.0256 .5766] .378

[R] 1.1506 1 .57661 .277

 

 

 

 

 

 

Note: mean = mean number of old collocates;

* The mean difference is significant at the .05 level.

In short, the effect of attention on the production of old collocates was mediated by
the factor of proficiency level, with Level 2 learners behaving differently from the overall
pattern. This interaction relationship was not predicted by Hypothesis 2.

4.2.4. Production of new collocates - overall comparison

Table 19 and Figure 4 present the statistics for the production of new collocates.

138

Table 19 Descriptive statistics of new collocates by complexity * level * attention

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Complexity Level Attention Mean Std. deviation N
Simple 2 [S] 5.1667 4.58918 12
[M] 2.0000 2.32379 1 l
[R] 3.5000 1.92725 8
[R+NE] 2.6000 1.50555 10
4 [S] 4.0769 4.88981 13
[M] 1.6667 2.26969 12
[R] ' 4.6667 3.05505 12
[R+NE] 4.8462 1 .67562 13
Complex 2 [S] 4.8333 3.48590 12
[M] 1.7273 2.24013 11
[R] 2.7500 2.25198 8
[R+NE] 2.6000 1.64655 10
4 [S] 4.1538 4.07934 13
i [M] 1.5000 2.19504 12
[R] 1.7500 2.17945 12
. [R+NE] 2.6154 1.75777 13
Note: mean = mean number of new collocates
N= number of participants
Production of new collocates
12 -..- - _ -
3
g 10 ,._
E .3: 8 *
g g 6 .complex rules
E = ‘ Bsimple rules
3 8 4 ’ ‘
S
3 2
E
0
aaaaaasa
6% £7

 

Level * Attention conditions

 

 

Figure 4 Production of new collocates

139

 

An ANOVA of the number of new collocates was performed, with level and
attention as between-subj ects factors and complexity as a within-subjects factor. There
were three major findings:

(1) Proficiency level did not have a significant main effect (F(1, 83) =.000, p=.983,
npz = .000).

(2) Collocational complexity had a significant main effect (F (l , 83) = 16.926, p
= .000, npz = .169), with the mean number of new collocates for wield and warp (simple
rules) being significantly more than that of make and take (complex rules). As previously
shown in Table 13, the total number of new collocates produced for wield, warp, make
and take were 167, 163, 118 and 134 respectively.

3) Attention had a significant main effect (F(3, 83) =4.254, p=.008, npz = .133).

As shown in Table 19 and Figure 4, overall, the semantic processing group produced
more new collocates than the other three groups; the memorization for recall group wrote
the fewest new collocates. A post-hoe Scheffé test of attention showed that a significant
difference lay between memorization for recall and semantic processing, with the
memorization for recall group producing significantly fewer extensions than the semantic

processing group (see Table 20).

140

Table 20 Post-hoe Scheffé test of the effect of attention on production of new

 

 

 

 

 

collocates

Attention Attention Mean difference Std. error Sig.
[S] [M] 5.6452(*) 1.58550 .008
[R] 2.7300 1.64627 .437
[R+NE] 2.6017 1.58550 .446
[M] [S] -5.6452(*) 1.58550 .008
[R] -2.9152 1.67778 .394
[R+NE] -3.0435 1.61819 .323
[R] [S] -2.73 00 1.64627 .437
[M] 2.9152 1.67778 .394

[R+NE] -. 1283 1.67778 1.000
[R+NE] [S] -2.6017 1.58550 .446
[M] 3.0435 1.61819 .323

[R] .1283 1.67778 1.000

 

 

 

 

 

 

Note: mean = mean number of new collocates

 

* The mean difference is significant at the .05 level.

Before ending this section and turning to the result of each sub-type of new
collocates, we will look at the ratio of old to new collocates in each level * attention
group: 2[S] = .175, 2[M] = 1.341, 2[R] = 1.22, 2[R+NE] = .808, 4[S] = .196, 4[M] =
2.316, 4[R] = 1.104, 4[R+NE] = 1.237 (See Figure 5). As shown by the numbers, for both
rule given and rule given plus negative evidence groups, they produced similar number of
old and new collocates. For memorization for recall group, old collocates tended to
occupy a much larger percentage than new collocates. In contrast, for the semantic
processing group, old collocates took up a very small percentage and new collocates

dominated their production.

141

Production of old vs. new collocates

 

 

 

'ihéw collocates
-’i 1[jlold collocates ]

Adda—b

 

owbaaoooroaoaoo
|
1911].
1
will.
‘1 ,1.
- 1
l l
‘11il
1 1.
‘111.
1:-‘
l
111'.
1,.
1111‘
1‘
[ill
[‘11

Mean number of collocates

 

«9.949.999.9949
«.9 .9

Attention ' Level conditions

Figure 5 Production of old vs. new collocates
4.2.5. Production of new target-like collocates

Now we will turn to the first sub-type of new collocates produced by participants,
that is, new target-like collocates (collocates that follow the target rules but which did not
appear in the passages). Research Hypothesis 3 states that, in general, the rule given and
rule given plus negative evidence groups will write significantly more new target-like
collocates than the memorization for recall and semantic processing groups, but for
lower-level students Ieaming complex collocational patterns (i.e., those of make and take),
there might be no significant difference between the two rule-oriented and the two
non-rule-oriented groups. As will be shown below, this hypothesis is partially confirmed.
Table 21 presents the descriptive statistics of the production of new target-like collocates

by complexity *level * attention.

142

Table 2]

Descriptive statistics of new target-like collocates by complexity *

level * attention

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Complexity Level Attention Mean Std. Deviation N
simple 2 [S] .3333 .77850 12
[M] ] .4545 1.03573 1 11
[R] 1 1.5000 1.51186 } 8
[R+NE] 1.0000 1.33333 j 10
Total .7561 1.19959 1 41
4 [S] l .8462 .98710 i 13
[M] . .8333 1.46680 g 12
[R] ; 2.9167 [2.39159 12
[R+NE] j 3.0769 1 1.44115 13
Total 1.9200 3 1.92555 [50
Total [8] .6000 - .91287 25
[M] ! .6522 1 1.26522 23
[R] 2.3500 2.15883 1 20
[R+NE] i 2.1739 1 1.72290 23
Total ’ 1.3956 [1.73128 . 91
complex 2 [S] .2500 ; .62158 1 12
[M] 1 .0909 ; .30151 ; 11
[R] 1.2500 1 1.83225 1 8
[R+NE] ; .7000 E .94868 1 10
Total 1-5122 1.05171 j 41
4 [S] 1.0769 5.27735 ‘13
[M] 1 .0833 i .28868 g 12
[R] 1 .7500 1 1.21543 1 12
[R+NE] 2.3077 : 1.70219 1 13
Total .8200 . 1.39518 150
Total [S] .1600 f .47258 25
[M] .0870 .28810 ’ 23
[R] .9500 1.46808 20
[R+NE] 1.6087 1.61637 23
Total .6813 1.25503 91

 

 

Note: mean = mean number of new target-like collocates

N= number of participants.

143

 

An ANOVA of production of new target-like collocates was performed, with level
and attention as between-subjects factors and complexity as a within-subjects factor.
There were five major findings:

(1) The effect oflevel was significant (F(1, 83) =8.969, p=.004, n,"- = .098), with
Level 4 learners producing significantly more new target-like collocates than Level 2
learners;

(2) The effect of complexity was significant (F(1, 83) = 21.213, p= .000, 1],,2 = .204),
with learners producing significantly more new target-like collocates for wield and warp
(easy rules) than for make and take (complex rules). As previously shown in Table 13, the
total number of new target-like collocates produced for wield (44) and warp (83) were
both more than those for make (33) and take (29).

(3) The effect of attention was significant (F(3, 83) = 12.012, p=.000, 11,,2 = .303);

A post-hoc Scheffé test of attention showed that the two rule-oriented groups (i.e.,
rule given and rule given plus negative evidence groups) produced significantly more
new target-like collocates than the two non—rule-oriented groups (i.e., semantic
processing and memorization for recall groups) (see Table 22). This pattern is the same as

the prediction made in Hypothesis 3.

144

Table 22 Post-hoe Scheffé test of the effect of attention on production of new
target-like collocates

 

 

 

 

 

 

 

Attention Attention Mean difference Std. error Sig.
[S] [M] .0104 .30204 1.000
[R] -1.2700(*) ] .31362 .002
[R+NE] -1.5113(*) 1 .30204 .000
[M] [S] -.0104 ] .30204 1.000
[R] -l.2804(*) 1 .31962 1 .002
[R+NE] -1.5217(*) ; .30827 .000
[R] [S] 1.2700(*) 1 .31362 .002
[M] 1.2804(*) 1 .31962 .002
1 [R+NE] —.2413 i .31962 .903
[R+NE] ] [S] 1.5113(*) .30204 ..000
1 [M] 1.5217(*) : .30827 .000
i [R] .2413 E .31962 .903

 

 

 

 

 

Note: mean = mean number of new target-like collocates

* The mean difference is significant at the .05 level.

(4) The interaction between attention and complexity was not significant (F(3, 83) =
1.324. p=.272, 11,} = .046).

(5) The interaction between attention and level was significant (F(3. 83) = 3.321,
p=.024, npz = .107) (see Figures 6 and 7).

Separate ANOVA tests of the number of new target-like collocates were conducted
for Level 2 and Level 4, with attention as a between-subjects factor and complexity as a

within-subjects factor.

145

Production of new target-like collocates (Level 2)

 
  

 

‘5 1.6
g 1.4
l: p -22 ___ --.--L
l . ' I .i
i E g 0.8 5' —I- 39:62:32;
1 .§ 2 0.6 ___--f . -2--- - “L---
l :1 g 0.4
1 l:
g g 0.2
g 0

[31 [M] [R] [R+NE]

Attentional conditions

Figure 6. Production of new target-like collocates (Level 2)

Production of new target-like collocates (Level 4)

..'.. 3.5 - - --— -~ - -

8t

5 3 ‘

u: g 2.5 ,

95 g 2 " ‘ +simple rules
i E 1.5 +oomplex rules
5:1

c

c 0.5

8

z 0

 

 

 

[51 [M] [R] 1R+ NEl
Attentional conditions

.“‘—___...__ ,-

Figure 7. Production of new target-like collocates (Level 4)
For Level 2 students, it was found (1) the effect of attention was significant (F (3, 37)
=3.186, p=.035, n; = .205), (2) the effect of complexity was not significant (F(1, 37) =

1.7367, p=. l 96, npz = .045), (3) the effect of attention was not mediated by complexity

146

(F(3, 37) = .116, p= .950, npz = .009). A post-hoc LSD test found that the rule given group
significantly outperformed the two non-rule-oriented groups, but the rule given plus
negative evidence group did not have any significant superiority over the two
non-rule-oriented groups (see Table 23). In fact, as shown in Figures 6 and 7, the

increasing trend from rule given to rule given plus negative evidence, which was found

with Level 4, did not hold for Level 2.

Table 23 Post-hoc LSD test of the effect of attention on production of new
target—like collocates (Level 2)

 

 

 

 

 

 

Attention 3 Attention Mean difference 1 Std. Error Sig.
[S] i [M] .0189 E .37468 1 .960
[R] -1.0833(*) .40970 , .012

' [R+NE] -.5583 I .38433 1 .155

[M] [S] -.0189 .37468 .960
[R] -l.1023(*) .41708 .012

[R+NE] —.5773 .39219 .150

[R] [S] l.0833(*) .40970 ; .012
[M] 1.1023(*) .41708 .012

[R+NE] .5250 ‘ .42577 : .225

[R+NE] , [S] .5583 ; .38433 .155
[M] .5773 : .39219 .150

i [R] -.5250 E .42577 g .225

 

 

 

Note: mean = mean number of new target-like collocates

* The mean difference is significant at the .05 level.

For Level 4 students, attention had a significant main effect (F(3. 46) =1 l .585,
p=.000, npz = .430), and so did complexity (F( l , 46) =26.247, p=.000, npz = .363). The

interaction between attention and complexity approached significance (F(3, 46) =2.538,

147

p=.068, npz = .142). A post-hoc Scheffé test of attention revealed a result consistent with
the overall pattern, that is, the two rule-oriented groups (rule given, rule given plus
negative evidence) produced significantly more new target—like collocates than the two

non-rule-oriented groups (semantic processing, memorization for recall) (See Table 24).

Table 24 Post-hoe Scheffé test of the effect of attention on production of
new-target-like collocates (Level 4)

 

 

 

 

 

 

 

Attention i Attention Mean difference 1 Std. error Sig.
[S] 1 [M] .0032 .46060 g 1.000
{ [R] -l.3718(*) ' .46060 i .042
l [R+NE] -2.2308(*) 1 .45129 i .000
[M] i [S] -.0032 .46060 1 1.000
[R] -1.3750(*) .46972 ‘ .047
[R+NE] -2.2340(*) (- .46060 § .000
[R] 3 [S] 1.3718(*) 9 .46060 i .042
9 [M] 1.3750(*) .46972 1 .047
[R+NE] -.8590 .46060 ‘ .336
[R+NE] ; [S] 2.2308(*) . : .45129 : .000
[M] 2.2340(*) , .46060 * .000
3 [R] .8590 1 .46060 T .336

 

 

 

Note: mean = mean number of new target-like collocates

* The mean difference is significant at the .05 level.

In short, in terms of the overall effect of attention on production of new target-like

collocates, the result is consistent with the prediction in Research Hypothesis 3, that is,

the rule-oriented groups outperformed the non-rule-oriented groups. However, when it

came to the interaction between attention and level/complexity, the hypothesis was not

confirmed. According to the hypothesis, attention would not show a main effect for Level

148

2 learners Ieaming complex collocational rules. But in reality, the rule given plus
negative evidence group of Level 2 lost its advantage over the two non-rule-oriented
groups with both simple and complex rules, while the rule given group kept its significant
advantage with both types of rules.
4.2.6. Production of non-target-like collocates

The second sub-type of new collocates produced by participants is non-target-like
collocates (i.e., collocates that do not follow the target rules). According to Research
Hypothesis 4, in general. the rule given plus negative evidence group will write
significantly fewer number of non-target-like collocates than the rule given group, which
in tum will write fewer number ofnon-target-like collocates than the memorizationfor
recall and semantic processing groups. but the significant difference among the four
attentional conditions might not hold for lower-level students Ieaming complex
collocational patterns (i.e., those ofmake and take). The result is not completely the same
as this hypothesis. Table 25 presents the descriptive statistics of production of

non-target-like collocates by complexity * level * attention.

149

Table 25 Descriptive statistics of non-target-like collocates by complexity *

level * attention

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Complexity Level Attention Mean Std. deviation N
Simple 2 [S] 3.6667 3.84550 12
[M] .9091 1.04447 ;11
[R] 1.1250 1.55265 18
[R+NE] .6000 .84327 110
Total 1.6829 2.57332 :41
4 [S] 2.6154 4.15408 [13
[M] .8333 ' 1.52753 1 12
[R] .1667 E38925 312
[R+NE] .5385 [.77625 l13
Total 1.0600 [241939 150
Total [S] 3.1200 [3.96148 ;25
[M] .8696 11.28997 :23
[R] .5500 1.09904 20
[R+NE] .5652 ;.78775 23
Total 1.3407 2.49542 .91
Complex 2 [S] 4.5833 3.67939 '12
[M] 1.6364 2.06265 5 11
[R] 1.5000 2.44949 8

[R+NE] '
1.9000 : 1.72884 . 10
Total 2.5366 [2.88182 .41
4 [S] 4.0769 [3.98877 ‘ 13
[M] 1.4167 ‘223437 '1 12
[R] 1.0000 1.95402 512
[R+NE] .3077 .63043 [13
Total 1.7200 2.85028 F50
Total [S] 4.3200 3.77183 '25
[M] 1.5217 2.10777 [23
[R] 1.2000 2.11760 20
[R+NE] 1.0000 1.44600 23
Total 2.0879 2.87769 91

 

 

150

 

Note: mean= mean number of non-target-like collocates

N= number of participants.

An ANOVA of the number of non-target-like collocates was performed, with
attention and level as between-subjects'factors and complexity as a within-subjects factor.
There were five major findings:

(1) The effect ofproficiency level was not significant (F(1, 83) =1 .671. p= .200, npz
= .020).

(2) The effect of collocational complexity was significant (F(1, 83) =14.598. p=.000,
1],,3 = .150), with the non-target-like collocates for make and take (complex rules) being
significantly more than those for lt'ield and warp (simple rules). As previously shown in
Table 13, the number ofnon-targct-like collocates for make (85) and take (105) were
both more than those for wield (67) and warp (55).

(3) The effect of attention was significant (F(3, 83) =8.883, p= .000, r],,2 = .243);

A post-hoe Scheffé test of attention showed that significant difference lay between
semantic processing and all the'other three groups (i.e., memorization for recall, rule
given. and rule given plus negative evidence). The three groups which included some
degree of focal attention to the target collocations produced significantly fewer

non-target-like collocates than the semantic processing group (see Table 26).

151

Table 26

Post-hoe Scheffé test of the effect of attention on production of

non-target—like collocates

 

 

 

 

 

 

 

 

 

 

 

Attention Attention Mean difference Std. error Sig.
[S] [M] 2.5243(*) .65413 .003
[R] 2.8450(*) 5 .67920 .001
[R+NE] 2.9374(*) 1 .65413 .000
[M] [S] -2.5243(*) 1 .65413 .003
[R] .3207 :.69220 1.975
[R+NE] .4130 1 .66762 1 .943
[R] [S] -2.8450(*) f .67920 1 .001
[no] 43207 3.69220 [.975
[R+NE] .0924 5 .69220 1 .999
[R+NE] [S] -2.9374(*) .65413 .000
[hi] 44130 1.66762 1.943
[R] 40924 1.69220 p.999

 

 

Note: mean = mean number of non-target-like collocates

* The mean difference is significant at the .05 level.

(4) The interaction between attention and complexity was not significant (F(3, 83)
= .636, p: .594, npz = .022);
(5) The interaction between attention and level was not significant (F(3. 83) = .1 l3.

p= .952, 71p: = .004) (see Figures 8 and 9).

152

 

A.- ..-- ._ -“‘... "a‘pfigw*-_._.é __,__, -

Production of non-target-Iike collocates (Level 2)

' 5

E. 4.5 ‘

g 4

c

3133': -3--_3-_____-_

3_ 25 . +simplerules -
i E § '2 '+oomplex rules
' E .3 1.5
. g = 1
f g 0.5

2 O

 

[31 [M] [R] lR+ NEl
Attentional conditions

Figure 8 Production of non-target-like collocates (Level 2)

 

Production of non-target-Iike collocates (Level 4)

, 4.5
5 4

9

g m 3.5

33 3

E

.52 2.5 —e—simple rules
g, 8 2 +complex rules
1531.5 --

:—

l: 1 -

g 0.5

E 0

 

 

 

[31 [Ml [R1 [R+NE]

Attentional conditions

 

 

 

Figure 9 Production of non-target-like collocates (Level 4)
To sum up, the results with production of non-target-like collocates were different
from Research Hypothesis 4. The hypothesis predicted that (l) the rule given plus

negative evidence group would produce significantly fewer non-target-like collocates

153

than the rule given group, which in turn would produce significantly fewer

non-target-like collocates than the memorization for recall and semantic processing

groups; (2) the effect of attention would be mediated by level and complexity. However,

it turned out that (1) although the trend predicted by the hypothesis was observed (i.e., the

number of non-target-like collocates increased with the order of [R+NE], [R], [M] and

[8]), there was no statistically significant difference among the rule given plus negative

evidence, rule given and memorizationfor recall groups, and they all significantly

outperformed the semantic processing group; (2) the effect of attention was independent

from that of complexity and proficiency level.

4.2.7. Production of semi-target-like collocates

The last type of new collocates produced by participants is semi-target-like

collocates.

Table 27 presents the descriptive statistics of production of semi-target-like

collocates (1) by level * attention, that is, nouns sharing partial semantic features with the

target noun collocates. Because all the semi-target-like collocates (1) were for wield and

warp, the factor of complexity was irrelevant here.

154

Table 27 Descriptive statistics of semi-target-like collocates (1)

 

 

 

 

 

 

 

 

 

 

 

Level Attention Mean Std. Deviation N
2 [S] .5833 .79296 12
[M] .4545 1 .93420 11
[R] .5000 1 .75593 8
[R+NE] .5000 E1.52705 10
Total 1 .5122 1.74572 41
4 [S] i .3077 .63043 13
[M] 1 .0000 1 .00000 1 12
[R] 1.8333 , 1.11464 1 12
[R+NE] 1 .5385 .51887 L13
Total 1 .4200 a .73095 1 50
Total [S] [.4400 1.71181 125
[M] f .2174 1 .67126 i 23
[R] 1 .7000 .97872 1 20
[R+NE] .5217 .51075 123
Total 1 .4615 1 .73496 191

 

 

 

Note: mean = mean number of semi-target-like collocates ( l)

N: number of participants.

An ANOVA was performed, with level and attention as two independent factors and
the number of semi-target-like collocates (l ) as a dependent factor. Neither attention nor
level had a significant main effect (Attention: F (3, 83) =1 .355, p=.262. 11,3 = .047; Level:
r11. 83) =.338, p=.563, t1,2 = .004).

Table 28 presents the descriptive statistics of production of semi-target-like
collocates (2) by leve1*attention, that is, nouns reflecting some general terms that
semantically dominate the target-like nouns or are superordinates of the target-like nouns.

Like the case with semi-target-like collocates (1 ), all the semi-target-like collocates (2)

155

were for wield and warp only, so the factor of complexity was irrelevant here.

Table 28 Descriptive statistics of semi-target—like collocates (2)

 

 

 

 

 

 

 

 

 

 

 

 

Level Attention 1Mean Std. Deviation N
2 [S] 1.5833 .66856 12
[M] 1 .1818 .60302 11
[R] 1 .3750 .74402 8
[R+NE] .5000 .7071 1 1 10
Total .4146 .66991 41
4 [S] .3077 .63043 1 13
[M] 1 .0000 1 .00000 1 12
[R] - 1.7500 11.05529 112
[R+NE] 1 .6923 1.94733 1 13
Total .4400 1.81215 1 50
Total [S] 1.4400 1.65064 125
[M] ; .0870 1 .41703 123
[R] .6000 1 .94032 120
[R+NE] .6087 .83878 :23
Total 1 .4286 1 .74748 191

 

 

 

Note: mean = mean number of semi-target-like collocates (2)

N: number of participants.

An ANOVA was performed. with level and attention as two independent factors and
the number of semi-target-like collocates (2) as a dependent factor. It was found that
neither attention nor level had a significant main effect (Attention: F (3, 83) =2.223,
p=.092, n; = .074; Level: F (l, 83) =.031, p=.86, n; = .000).

To sum up, for the production of both types of semi—target-like collocates, neither
attention nor proficiency level had a significant main effect.

4.2.8. Summary of collocate production

156

Table 29 and Figure 10 provide an overview of the production of each sub-type of

collocates by each level * attention group.

Table 29 Production of each sub-type of collocates by level * attention

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mean number of collocates produced
9V3" old new non-target- semi-target- semi-target Total
ttention target-like like like (1) -like (2)
1.75 0.5833 8.25 0.5833 0.5833 11.75
2[M] 5 0.5455 2.5455 0.4545 0.1818 8.7273
2[R] 7.625 2.75 2.625 0.5 0.375 13.875
2[R+NE] 4.2 1.7 2.5 0.5 0.5 9.4
4[S] 1.6154 0.9231 6.6923 0.3077 0.3077 9.8462
4[M] 7.3333 0.9167 2.25 0 0 10.5
4[R] 7.0833 3.6667 1.1667 0.8333 0.75 13.5
4[R+NE] 9.3846 5.3846 0.8462 0.5385 ‘ 0.6923 16.8462
Production of each sub-type of collocates
by level * attention
13 - .. - ..W.
3 16 . -.---- .. - MW - ....-- l'semi-target-like
H 14 .. ..- _ - oollomtes(2)
g . asemi-target-like
3 12 1. collocates(1)
“a 10 -_--..... -.~——,-- , Clnon-target-like
t. . g 21.: oollomtes
8 8 - .
E .111: .11.. 1 I target-like
2 6 7' __1 7111" collocates
g 4 .. 1__ _ 1——-—~ .. .. -1 111—1 Iold collocates
z 2 1 —1 V _-
e -r-r--.; .. 2:: 1:. 1
O 3 1t 1 . - - 1. ”a 1
’\ \
.6 a“ .4- .48 4° .8“ .45 38>
8“ t8"

Level * attention conditions

Figure 10 Production of each sub-type of collocates by level * attention

157

4.2.9. NS judgment of IL production

As described in Chapter 3, all the new collocates produced by participants were
presented to two NSs of English, who categorized the IL production as Y (good
combinations), 0 (OK, but not the best way to say it) or N (bad combinations). The
interrater reliability was 87.54% (288 out of 329 TYPES (not tokens) of collocates). Not
surprisingly, in most of the cases which caused different judgments (35 out of 41 cases),
one NS judged the combination as 0, while the other judged it as Y/N. Only 6 cases were
judged as Y by one NS, while N by another. For those controversial cases, a third NS was
consulted and the majority opinion was adopted. If the answers by the three NSs tumed
out to be Y, N, 0, then that particular item was labeled as 0. Tables 30 to 33 show the)
NS judgment of new target—like collocates, non-target-like collocates, semi-target-like

collocates ( l) and semi—target-like collocates (2) respectively.

Table 30 NS Judgment of new target-like collocates

 

 

 

 

 

Y(= good 0 (= OK, but not the N (= bad
combinations) best way to say it) combinations)
collocates of wield 9 4 2
collocates of warp 15 15 2
collocates of make 7 7 2
collocates of take 4 8 3

 

 

 

 

 

 

Note: The number in the table indicates how many types (not tokens) of a verb's new
target-like collocates are judged as Y, O or N.

158

Table 31 NS Judgment of non-target-Iike collocates

 

 

 

 

 

Y(= good O (= OK, but not the N (= bad
combinations) best way to say it) combinations)
collocates of wield 0 4 35
collocates of warp 0 4 42
collocates of make 22 i 18 20
collocates of take 41 3 33

 

 

 

 

 

 

Note: The number in the table indicates how many types (not tokens) of a verb's non-

target-like collocates are judged as Y, O or N.

Table 32 NS Judgmentof semi-target-like collocates (1)

 

 

 

 

Number of collocates for each category
Y(= good O (=‘OK, but not the N (= bad
combinations) best way to say it) combinations)
collocates of wield 1 l S
collocates of warp O 2 1 10

 

 

 

 

 

 

Note: The number in the table indicates how many types (not tokens) of a verb's semi-

target-like collocates (l) are judged as Y, O or N.

Table 33 NS Judgment of semi-target-like collocates (2)

 

 

 

 

Number of collocates for each category
Y(= good O (= OK. but not the N (= bad
combinations) best way to say it) combinations)
collocates of wield 3 3 O
collocates of warp l 3 O

 

 

 

 

 

 

Note: The number in the table indicates how many types (not tokens) of a verb's semi-
target-like collocates (2) are judged as Y, O or N.

There are three important pieces of information conveyed by Tables 30 - 33:
First, most target-like collocates were judged as good or OK by NSs (wield: 86.67%,
warp: 93.75%, make: 87.5%, take: 80%).

Second, among the non-target-like collocates, the situation was somewhat different

159

for wield, warp on the one hand and make, take on the other hand. For wield and warp,
most non-target-like collocates were judged as bad (wield: 89.74%; warp: 91.30%) and
there was no non-target—like collocates judged as good. However, for make and take,
there were much more non-target-like collocates judged as good or OK (make: 66.67%;
take: 57.14%).

Third, most semi-target-like collocates (1), that is, nouns sharing partial semantic
features with the target nouns, were judged as bad (wield: 71.43%; warp: 83.33%);
however, all the semi-target-like collocates (2), that is. superordinates of the target nouns,
were judged as good or OK.

4.3. Judgment of collocations of the target verbs
4.3.1. Judgment of good collocations

According to Research Hypothesis 5. in terms of judging good collocations, attention
to target collocational features does not have a main effect. This hypothesis makes the
right prediction. Table 34 presents the descriptive statistics of the judgment of good
collocations by complexity * level * attention. Figures 11 and 12 show the results for

Level 2 and Level 4 respectively.

160

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 34 Descriptive statistics of the judgment of good collocations by
complexity * level * attention

Complexity Level Attention Mean Std. deviation N
Simple 2 [S] 3.3846 .86972 13
[M] 3.4167 .66856 12

[R] 3.4444 .52705 9

[R+NE] 3.0000 .66667 110

Total 3.3182 1.70785 .44

4 [8] 3.6923 ‘.63043 113

[M] 3.9167 I.28868 12

[R] 3.6667 1.49237 112

[R+NE] 3.9231 1.27735 : 13

Total 3.8000 1.45175 150

Total [S] 3.5385 i.76057 126

[M] 3.6667 3.56466 ;24

[R] 3.5714 1.50709 21

[R+NE] 3.5217 1.66535 .23

Total 3.5745 7.63057 :94

Complex 2 [S] 2.9231 7 1.03775 , 13
[M] 3.4167 1.66856 . 12

[R] 3.0000 1.70711 9

[R+NE] 3.4000 i.69921 10

Total 3.1818 1.81477 .44

4 [8] 3.1538 1.80064 L 13

[M] 3.2500 1.75378 12

[R] 3.5000 .79772 112

[R+NE] 3.8462 1.37553 113

Total 3.4400 7.73290 [50

Total [8] 3.0385 .91568 326

[M] 3.3333 .70196 24

[R] 3.2857 .78376 21

[R+NE] 3.6522 .57277 23

Total 3.3191 .77893 94

 

 

 

 

 

 

 

Note: mean = mean score of judgment of good collocations;

The maximum score for judging good collocations of easy and complex patterns

l6]

 

were both 4;

N = number of participants.

 

N co .5.
AU:

d

Judgment of good collocations (Level 2)

' +simple rules
1+oomplexrulesf1

l
D

 

 

 

good collocations

Mean score of judgment of
O

Gin-sinminooin

 

[S] [M] [R]

Attentional conditions

[R+NE]

Figure 1] Judgment of good collocations (Level 2)

amt».A
#01

good collocations

Mean score of judgment of
p

 

Judgment of good collocations (Level 4)

 

+oomplex rules

 

oat-e'olm'moou:

[R1
Attentional conditions

Figure 12 Judgment of good collocations (Level 4)

An ANOVA of the judgment of good collocations was performed, with attention

and level as between-subjects factors, complexity as a within-subjects factor. It was

162

found:

(1) Proficiency level had its significant main effect (F(1, 86) = 11.563, p = .001, npz

= .l 19), with Level 4 outperforming Level 2.

(2) Collocational complexity had its significant main effect (F(1, 86) = 7.618,

p=.007, npz = .081 ), with the mean score for wield and warp (simple rules) being

significantly higher than the mean score for make and take (complex rules). Table 35

displays the number of participants who incorrectly judged a good sentence. As shown in

the table, there were more participants who made errors with the good sentences

including make and take than those who made errors with the good sentences including

wield and warp. And learners seemed to perfomi the worst with the two sentences

containing take.

 

 

 

 

 

 

 

 

 

 

Table 35 Number of participants who incorrectly judged a good sentence
Level*attention wield I wield 2 warp I warp 2 make I make 2 take I take 2
2[S] 2 1 3 2 3 5 3 ' 3
2[M] l 2 l 3 2 2 2 l
2[R] l 2 0 2 2 3 2 2
2[R+NE] 3 1 3 3 0 l 4 1
4[S] 0 1 2 1 1 l 6 2
4[M] 0 1 O O 1 1 7 0
4[R] O 1 0 3 0 2 O 5
4[R+NE] 0 0 O 1 O 0 O 2
Total 7 9 9 15 9 15 24 16

 

 

 

 

 

 

 

 

 

 

Note: wield 1 = the first good sentence containing wield in the judgment test;

163

 

wield 2 = the second good sentence containing wield in the judgment test;
The same is true with warp 1, warp 2, make 1, make 2, take 1, take 2;
See Table 8 in Chapter 3 for the order of the good sentences used in the

judgment test.

(3) Attention did not have its significant main effect (F(3, 86) =1.l4l , p=.337, npz
=.038). As previously shown in Table 34, the mean score in all attentional conditions
approached 4, the maximum point. This result is consistent with Research Hypothesis 5.
4.3.2. Judgment of bad collocations

Among the sentences for judgment, there were 8 containing a bad collocation, 2 for
each target verb. Leamers' response to a bad sentence could be classified into 6 major
types:

(l)judging it as good;

(2) judging it as bad, but for reasons unrelated to the target collocations. For example,
one student from the rule given plus negative evidence condition (Level 2) thought it‘al‘p
could not be used in passive voice, so judged the following sentence as wrong:

* The company 's operations have been warped since the new manager came.

(3) judging it as bad, simply pointing out that the target verb-noun combination was
inappropriate, without discussing which collocational rules were violated. For instance,
one student from the rule given condition (Level 2) wrote that the combination made two

books in the following sentence seemed odd:

164

*T he writer made two books last year.
Some students only underlined the target collocations, signaling that this was the wrong
part.

(4) judging it as bad, pointing out that the target verb-noun combination was
inappropriate, but using an incorrect collocational rule as the standard for judgment. For
example, some students mistakenly used the rules for make to judge the combination of
take; One student from memorization for recall condition (Level 2) underlined the phrase
wielded several methods in a bad sentence and said that wield should be followed by
concrete objects. Obviously, the rule written by the student was inaccurate compared to
the target rules: "nouns referring to a concrete tool or piece of equipment that can be
carried and used or nouns associated with the concept of power".

( 5) judging it as bad. pointing out that the target verb-noun combination was
inappropriate, using ONE OF the two target rules (In the experiment design, there were
two collocational rules for each target verb). For example, one student from rule given
plus negative evidence condition (Level 2) judged the following sentence as bad:

* He didn 't take any reply from his girlfriend.
and the reason was that reply was not a physical movement. This was only concerned

with one of the target collocational rules of take.

165

(6) judging it as bad, pointing out that the target verb-noun combination was
inappropriate, using BOTH OF the two target rules. For example, one student from rule
given plus negative evidence condition (Level 4) judged the following sentence as bad:

*He didn '1 take any support from this family.

Meanwhile. it was pointed out that support had nothing to do with a role or physical
activities. In this case, the two target rules of take were both mentioned.

Among the six types of responses. obviously, the first and second were wrong, so
they were assigned 0. The sixth type was perfect in that it identified the error and also
provided complete and‘correct reasons. so it was assigned 1 point, the maximum score.
As to learners who gave the third, fourth and fifth types of responses, they also saw the
inappropriateness of the target combinations, although they did not provide a reason or
provided a wrong or incomplete semantic reason. Since the major purpose of the
judgment test was to see whether learners could recognize or feel the collocational
problems, rather than whether they possessed accurate judgment criteria, the third to fifth
types of responses were also assigned 1 point. If they were given partial points instead,
the scoring method might have unfairly favored the two rule-oriented groups, who were
provided with explicit rules for sentence judgment.

Now we will turn to the statistical results. Research Hypothesis 6 states that in terms

166

of judging bad collocations, in general, the rule given plus negative evidence group will
outperform the rule given group, which in turn will outperform the memorization for
recall and semantic processing groups, but the significant difference among the four
attentional conditions might not hold for lower-level students learning complex
collocational patterns. As will be shown below, the predicted overall effect was found in
the experiment, but the interaction relationship between attention and proficiency level/
complexity was not. Table 36 presents the descriptive statistics of the judgment of bad

collocations by complexity* level * attention.

Table 36 Descriptive statistics of the judgment of bad collocations by
complexity * level * attention

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Complexity ___ .. MLevel Attention 1 Mean l Std.__deviation : N

Simple 2 [S] 1.8462 f.98710 7:13
[M] 2.4167 1.79296 :12 .

[R] f 1.4444 1 1.66667 fi 9

[R+NE] [2.8000 . 1.22927 1 10

Total 1 1.2955 1 1.43995 £44

4 [S] 1.3077 1.48038 113

[M] ‘2500 .62158 112

[R] 1.0833 .99620 [12

[R+NE] 3.2308 [1.01274 13

Total 1.2400 1.46469 50

Total [S] .5769 .80861 26

[M] .3333 .70196 24

[R] 1.2381 1.30018 21

[R+NE] 3.0435 1.10693 23

Total 1.2660 1.44561 94

 

 

167

 

Table 36 (cont'd)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Complexity Level Attention Mean Std. deviation N
Complex 2 [S] .5385 .51887 13
[11] 1.0833 .79296 12

[R] l 7778 ‘L20185 9
[R+NE] 12.9000 1.10050 10
(Complex) Total 1 1.4773 1.24804 144
4 [S] 1.8462 .80064 [13

[14] 1 L0833 .90034 112

[R] ;1 6667 .98473 112

[R+NE] 13.3846 1.04391 S 13
Tons 1176001 L36367 :50

Toun [S] 1.6923 1.67937 {26
[14] 51.0833 7.82970 ?24

[R] 71.7143 :1.05560 {21

[R+NE] 23.1739 '1 1.07247 123

'Toun 11.6277 11.31152 :94

Note: mean = mean score ofthejudgment ofbad collocations;

The maximum score forjudging bad collocations of easy and complex
patterns were both 4.

N: number of participants.

An ANOVA ofthcjudgment of bad collocations (with attention and level as
between-subjects factors and complexity as a within-subject factor) showed:

(1) Proficiency level did not have its significant main effect (F(1, 86) =.001, p=.971,
11,2 = .000);

(2) Collocational complexity had its significant main effect (F(1, 86) =8.587, p=.004,
npz = .091), _with the mean score of wield and warp (simple rules) being significantly

lower than that of make and take (complex rules);

168

(3) Attention had its significant main effect (F(3, 86) = 53.648, p= .000, npz = .652).

A post-hoe Scheffé test of the effect of attention showed that the two rule-oriented
groups (rule given and rule given plus negative evidence groups) significantly
outperformed the other two non-rule-oriented groups (semantic processing and
memorization for recall groups). Moreover, the rule given plus negative evidence group
significantly outperformed the rule given group (See Table 37). These patterns are

consistent with Hypothesis 6.

Table 37 Post-hoe Scheffé test of the effect of attention on the judgment of bad

 

 

 

 

 

 

 

collocations
Attention Attention Mean difference : Std. error 1 Sig.
[S] [M] -.0737 F .21226 1.989
[R] -.84l6(*) f .22000 3 .004
[R+NE] -2.4741(*) .21464 .000
[M] [S] .0737 1 .21226 .989
[R] -.7679(*) a .22406 i .011
[R+NE] -2.4004(*) .21880 E .000
[R] [S] .8416(*) 3 .22000 1 .004
[M] .7679(*) 5 .22406 .011
[R+NE] -1.6325(*) .22632 I .000
[R+NE] [S] 2.4741(*) .21464 .000
[M] 2.4004(*) .21880 .000
[R] 1.6325(*) .22632 .000

 

 

 

 

 

Note: mean: mean score of judgment of bad collocations;
* The mean difference is significant at the .05 level.

(4) The interaction between attention and complexity was not significant (F(3, 86) =

1.561, p=.205, 11,2 = .052).

169

(5) The interaction between attention and level was not significant (F(3, 86) =.948,

p= 421.11.." = .032) (see Figures 13 and 14)-

Judgment of bad collocations (Level 2)

9°
0:

 

(a)

 

 

' M

_s
U'I-‘U'INU’t

 

—-—— 1

+simple rules ' .
. -I— complex ruIes '- .

, :f/ ,

[S] [M] [Rl [R+NE]

Attentional conditions

 

 

 

bad collocations

 

l..._—.
nr—_._.-

Mean score of judgment of
O

 

C

Figure 13 Judgment of bad collocations (Level 2)

Judgment of bad collocations (Level 4)

 

u. 4 ~ ~ ~ ~ w
o
g 3.5 ..
s 2 3
a. O
'0 '5 2.5 ~ ,
§§ 2 +simple rules
:3 =8- 1 5 +oomplex rules
§° '
.3 1 '
I:
3 0.5
I 0

 

 

 

[S] [M] [R] [R+ NE]
Attentional conditions

Figure 14 Judgment of bad collocations (Level 4)
4.4. Summary of the test results

Table 38 summarizes the major results of this experiment.

170

.oocooobmw 2805:me 0: 8:8me x88 ...: of. moogqbotoa ootoasm 808 8:865 #88 ..A.. 2:.

 

 

 

 

 

 

 

 

 

682
532225 Hoobo HE AS: A moosooozoo
539885 63$:me o: anmawv. 0: 295m Axanoo E9: on V: A 523: can we EoEwwE.
“Segue—Bo.
5388 A 235 =A>H Hoobo :38 on team .«0 Eofiwwg
838:8
Hoowo ofi—LowomEEom
588 o: Soho :88 o: .3 cowosvoom
532605 E
539825 538.53 Soto HE A 838:8 oxaoomofléo:
Emu—mama o: ogommnwa o: 5383 A 295m 52: o: E): .7: AmZ+E mo oocecwo><
E d); A

88835 324.5 .E E Z .9: A 880:8 3:888
88888 8 E ..E A E H= 8388 A 288 =AZ 82$: .3 38 oo 8828.5

EA .
88835 _m2+mH._a.HE ”2 EA 88888
scapegoat 8 EAE a 8388 A 29% = A2 52+”: .7: .2): so 8 8885
35>
330825 cocofioofi Soho “owns o5 mo wgofi
83$:me o: “coo—mafia o: Soho :88 o: Ema: o: Soho :88 o: o8 mo cosmmmscoaw
bio—QEooanozaoce. _o>o_ «nocao:< 59.5—98.5 _ .264 not—53¢ «m8 2: 5 9.2?

 

 

 

 

 

 

 

8:59. :8 on. we Pafifism an 2.38

171

4.5. Rule detection
4.5.1. Semantic processing group

In the exit questionnaire, only 2 students from the semantic processing group
(altogether 26 students) said that they paid some attention to the target noun collocates
(one from Level 2, the other one from Level 4). This result is consistent with the
assumption of the research design, that is, learners in the semantic processing condition
were only exposed to the target collocational sequence, and their focal attention to the
target collocations was limited to the greatest extent.

When asked to write down any collocational rule they had noticed for the four target
verbs. among the 26 students in semantic processing condition, 13 could not write
anything; 8 only listed some possible nouns or simply repeated the definition ofthe verb
(e.g., "warp is followed by something that is made worse"); and 5 wrote some rules or a
combination of rules and nouns. The rules figured out by the 5 learners were either
irrelevant or incomplete and no one discovered any correct target rule. The rules
relatively close to the targets were: 1) wield can be followed by something concrete (4
students), 2) warp can be followed by something abstract (1 student); 3) warp can be
followed by thought and objects (1 student); 4) make can be followed by something

abstract (1 student); 5) take can be followed by something abstract (1 student).

172

4.5.2. Memorization for recall group

When asked to write down any collocational rule they had noticed for the four target
verbs, among the 24 students in memorization for recall group, 7 wrote nothing, 3 listed
some possible noun collocates, 14 listed some rules. Obviously, compared to the semantic
processing group, the memorization group had more students who were able to induce
some rules. Moreover, their rules were more target-like. Table 39 displays the rules
discovered by the learners on their own. The number in the parentheses indicates the

number of participants who produced that rule.

Table 39 Detection of rules by participants in memorization for recall condition

 

 

 

 

 

 

Target verbs Detected rules (number of participants who produced that rule)
wield a) weapon, concrete b) abstract nouns c) nouns related to power
objects, or device(8) (2) (2)
warp a) concrete objects, b) abstract nouns c) nouns about personality.
furniture (9) (1) view, thought. feeling (5)
make a) abstract nouns (3)
take a) abstract nouns (4) b) action or
physical action (2)

 

 

 

 

 

4.6. Rule recall

1n the exit questionnaire, learners from the rule given and rule given plus negative
evidence conditions were asked to recall the collocational rules they learned during the
treatment. Their answers could be divided into 3 major types:

(1) No recall of the target rules, that is, giving no answer, simply listing some nouns,

173

or producing a completely wrong rule.

(2) Partial recall of the rules, that is, giving incomplete rules or using synonyms to
replace some of the words in the given rules. For example, some students wrote that
wield could be followed by a tool or piece of equipment, while omitting the part "that
could be carried and used". Some students said that warp could be followed by nouns
related to "feeling" or "thought" rather than using the original word in the given rule:
"mind".

( 3) Production of complete and correct rules.

Tables 40, 41 and 42 show the mean number of rules that can be completely recalled,
partially recalled and not recalled respectively within each level * attention condition.

The total number of rules to be recalled is 8, with 2 for each target verb.

Table 40 Descriptive statistics of rules that can be completely recalled

 

 

 

 

,_L§V.6_'.’::\13931192__ ,- -. “.1433“ f 51813688892 1._,__E___
2 [R] , 3.6250 I 2.19984 8
2 [R+NE] 4.7000 1.63639 ; 10
4 [R] . 4.7273 2.00454 3 11

..‘l [R+NE] [ __ 6.9231 1.49786 13
Total L 5.1905 2.14406 42

 

 

 

 

 

Note: Mean = mean number of rules that can be completely recalled

N = number of participants

174

Table 41 Descriptive statistics of rules that can be partially recalled

 

 

 

 

 

 

 

 

Level * Attention Mean Std. Deviation N

2 [R] 1.6250 1.06066 8
2 [R+NE] 1.6000 1.26491 10
4 [R] ; 1.3636 1.12006 11
4 [R+NE] ' .6923 .85485 13
Total 7 1.2619 . 1.10563 42

 

Note: Mean = mean number of rules that can be partially recalled
N = number of participants

 

Table 42 Descriptive statistics of rules that cannot be recalled

 

 

 

 

 

Level * Attention 1 Mean Std. Deviation L N

2 [R] in 2.7500 [ 2.121327 8
2 [R+NE] 1 1.9000 l 1.79196 1 10
4 [R] . 1.9091 [ 1.70027 i 11
4 [R+NE] . .3846 ‘_ 96077; 13
361811 _ _W— TM W ill—”1.5952l ......2 1.799511 1 42

 

 

 

Note: Mean = mean number of rules that cannot be recalled

N = number of participants

ANOVAs were perfomied on the results of rule recall, with level * attention
condition as the independent factor. There were three findings:

(1) In terms ofthe mean number of rules that can be completely recalled, there was
significant difference among the four level*attention conditions (F (3, 38) = 6.430,
p= .001, npz = .336). A post-hole Scheffé test revealed significant difference between
4[R+NE] and all the other three groups (i.e., 2 [R], 2 [R+NE] and 4 [R]).

(2) In terms of the mean number of rules that can be partially recalled, there was no

significant difference among the four level*attention conditions (F (3, 38): 1.897,

175

p= .146, 11,2 = .130).

(3) In terms of the mean number of rules that cannot be recalled, there was
significant difference among the four level*attention conditions (F (3, 38) = 4.004,
p= .014, npz = .240). A post-hoe Scheffé test found that the significant difference lay
between 2 [R] and 4 [R+NE].

4.7. Three-day treatment

Now we will retum to the three-day treatment before the test. There are three parts
in the treatment exercises, the answers of which are especially helpful to our
understanding of the test results.

First, the exercise which asked learners to match the target verb with its meaning.
All the participants. whatever their attentional condition. got the right answers.

Second, the exercise which asked learners from the two rule-oriented groups to
match the given collocational rules with the instances in the reading passages. Learners
did quite well in this. Only 1 student from the rule given plus negative evidence condition
(Level 2) left this part blank for wield and warp on Day 1, and 2 students from the rule
given plus negative evidence condition (Level 4) mismatched the two rules for take with
the two target phrases on Day 1.

Third, understanding of the negative evidence provided. As described in Chapter 3,

176

when the rule given plus negative evidence group was informed that a particular sentence
containing an inappropriate collocation of the target verb was bad, they were also asked
"Why do you think is this sentence bad?" Most students referred to the collocational
inappropriateness in the answers. But there were also some answers indicating that the
student did not pick up the message conveyed by the negative evidence, for example,
leaving it blank, providing an irrelevant reason, saying "I don't know", or insisting that
the sentence was good. Table 43 shows the number of participants who gave these

unsatisfactory answers.

Table 43 Non-understanding of negative evidence in rule given and rule given

plus negative evidence groups

 

 

 

 

 

 

 

 

 

Leveli‘attention/Day Number of participants who did not seem to understand the
intention of the negative evidence
wield warp make take
2 [R+NE] Day 1 3 3 3 1
Day 2 2 2 4 1
Day 3 0 2 0 2
4 [R+NE] Day 1 0 2 4 3
Day 2 3 4 2 1
Day 3 l 1 2 l

 

 

 

 

 

 

 

There are two notable features in Table 43:
1) On Day 3, for both proficiency levels, the number of participants who misunderstood
the negative evidence was reduced compared to the previous two days;

2) There was not very salient difference between Level 2 and Level 4, in terms of the

W7

number of participants who did not seem to understand the meaning of the negative

evidence.

In this chapter, I have reported the major results of the experiment. The following

chapter aims to provide possible explanations for these results and discuss their

implications.

I78

CHAPTER 5
Analysis and discussion

This chapter will evaluate and interpret the results of the experiment according to
the following outline:

(1) Attention and acquisition of the basic meaning.

(2) Attention and production of collocates. (There will be analysis of the overall
effect of attention and its interaction with proficiency level in the production of old
collocates, new target-like collocates and non-target-like collocatess. Moreover, the NS
judgment of IL production will be discussed.)

(3) Attention and judgment ofcollocations. (The effect of attention on thejudgment
of good collocations and bad collocations will be discussed separately.)

(4) The lack of interaction between attention and collocational complexity.

(5) Integration ofthe analysis.

The first three parts focus on the effect of attention and its interaction with
proficiency level that have been observed in various tasks of the experiment test. The

fourth part explains why in the current research the effect of attention was not mediated

 

5 Semi—target-like collocates are not discussed in the analysis, because they are considered as borderline
cases. And different from the cases that fall into quite clear-cut categories, these borderline cases might not
be able to offer a clear window onto the relationship between the type of attention learners pay to the input
and what learners can gain from the input.

179

by the factor of collocational complexity. The last part aims to integrate the

interpretations and conclusions made in previous parts and relate them to the three

research questions put forward in Chapter 3. Again, for the sake of simplicity, the target

verbs in the experiment will be written as wield, warp, make and take rather than the

original pseudo-forms.

5.1. Attention and acquisition of the basic meaning

In Part I of the test, all the participants got the right answers when asked to choose

from four options a word/phrase which had a meaning closest to that of the target verb.

To interpret this result, we need to consider two issues in the experimental procedure: (I)

what type of meaning knowledge was examined by the test; (2) how learners processed

the basic meaning of the target verbs during the treatment.

First, we will analyze what kind of meaning knowledge was assessed by Part I of the

test. Laufer and Goldstein (2004) conceptualized the knowledge of meaning-form

relationship as a hierarchy of four degrees of strengths, in which some degrees of

knowledge are considered as more advanced than others: passive recognition, active

recognition, passive recall and active recall. This hierarchy is based on two distinctions:

(1) active knowledge (i.e., providing the form for a given meaning) vs. passive

knowledge (i.e., providing the meaning for a given form); (2) recall (i.e., being able to

I80

recall the form or meaning of a word) vs. recognition (i.e., being able to recognize the

form or meaning or a word in a set of choices). Laufer and Goldstein further argued that

if active knowledge is more difficult to acquire than passive knowledge and if recall is

more difficult than recognition, then the most advanced knowledge in this hierarchy is

active recall and the least advanced is passive recognition. In the present experiment, Part

I of the test asked learners to choose the meaning of a word from a set of options, which

falls into the category of passive recognition. If the assumption by Laufer and Goldstein

is correct, we may say that the type of meaning knowledge assessed by the test was the

least advanced and the easiest to achieve.

Second, we need to take a closer look at how learners processed the basic meaning

of the four target verbs during the three-day treatment. Each day the participants

encountered the target verbs in the manners shown in Figure 15:

181

35> 538 o5 3 3:52: 3. .3 35.535

 

 

 

 

35> 598 o5 .3
33.58 o5 8 c2385
.3 m_o>o_ :55ch

8 wo_ o>3 Ema:

3023 «no? 5385a

e E 3353:00 “owns
o5 9 33:85 .8352
833358 3 BER
55 8555 so wEoQ
”v 8%

 

 

wcmcooE ~55 53, 35>
women“ 38 ofi :39:
8 $3.52 “5:32

333 8555 so mEoQ
”m 85

 

 

 

 

35> 3:0 @32on o5
5m @5mb EwBE ofi
:238 8 2% 353 95
N owommom E 333:5
35> Hows“ So.“ o5 («o
@3958 o5 35585
”N doom

 

 

2 88:

 

Li,

 

35> o5 .3 $35on o5
o8 353E Emcee of
2:38 8 2% 353 25
fl owemmem E 333:5
35> 538 .58 o5 .«o
9:558 o5 wcmmmoooi
; new

 

 

182

 

During the lst and 2nd steps, the margin glossing could promote noticing of the
target verbs and their basic meaning. At step 3, the nature of the task required all the
participants to pay focal attention to the meaning of the target verbs. At step 4, as
indicated in the above figure, leamers' attention to the basic meaning of the target verbs
was different as a result of different task demands. Under the semantic processing
condition, learners were asked to look for a word with a particular meaning in a sentence
containing the target verb. To do this, they might need to review the meaning of the target
verb to see whether it was the word that matched the given meaning or perhaps they
could immediately identify the answer while ignoring the target verb at all. Under the
memorization for recall condition. very likely the memorization of the phrases was based
on meaning comprehension, so learners needed to process the meaning of the target verbs
again. As to the rule given and rule given plus negative evidence groups, the collocational
rules presented to them were actually closely related to the basic meaning of the target
verbs, because, as mentioned in Chapter 2 (2.1.1.2.), a word's collocational feature can be
regarded as about a word's finer shade of meaning. It appears, then, that the two
rule-oriented groups were processing the meaning of the target verbs at a deeper level
than the other two groups, moving from rough categorization and a vague idea about the

general meaning to more precise mastery of meaning.

183

 

Considering the fact that in the test all the participants could recognize the meaning

of the four target verbs, we may suggest that passive recognition of a word, which only

requires a quite shallow level of meaning knowledge, can be attained when the following

conditions are all satisfied: (1) noticing and encountering its use in a sentence (in this

experiment at least 4 times each day and for 3 consecutive days); (2) being informed of

its basic meaning (in this experiment, margin glossing was provided 2 times each day and

for 3 consecutive days); (3) doing exercises that explicitly focus on its basic meaning (in

this experiment 1 exercise each day and for 3 consecutive days). Any more elaborate

level of meaning processing. as the one experienced by the rule given and rule given plus

negative evidence groups, does not seem to be necessary for successful passive

recognition.

5.2. Attention and production of collocates

5.2.]. Production of old collocates

5.2.1.1. Overall effects of attention

Overall, the memorization for recall, rule given and rule given plus negative evidence

groups produced significantly more old collocates than the semantic processing group.

Before interpreting this result, let us first review the four groups' level of attention to the

target collocations:

184

 

(1) semantic processing group -- peripheral attention (mere exposure to the target word

sequence and meaning comprehension);

(2) memorization for recall group -- focal attention at the level of phrase chunking and

storage (meaning comprehension and maintenance rehearsal, i.e., cycling the target

phrases in STM through a “phonological loop”);

(3) rule given group -- focal attention at the level of rule understanding (meaning

comprehension and elaborative rehearsal, i.e., explicitly Ieaming and testing the

underlying patterns);

(4) rule given plus negative evidence group -- focal attention at the level of rule

understanding and knowledge of what are impossible combinations (meaning

comprehension and elaborative rehearsal).

In short, the men-torizationfor recall. rule given and rule given plus negative evidence

groups directed more focal attention to the target phrases and had more elaborate

processing of them than the semantic processing group.

The superiority of the three focal attentional conditions over the non-focal attentional

condition in producing old collocates is consistent with the view summarized in Chapter

2. That is, processing new lexical information more elaborately and from more

dimensions will lead to higher retention than processing new lexical information less

185

elaborately. It also fits in with the argument by Gass (1988a) that one factor determining
whether a particular instance of comprehended input will become intake or not is the
level of input analysis a learner achieves. According to Gass, an analysis at the level of
meaning, that is, having a general understanding of the message, might not be as useful
for intake as a mini-linguistic structural analysis (e. g., analysis at the level of syntax). In
this study, participants in the semantic processing group made an analysis of the target
phrases only at the level of meaning comprehension, and it seems that the target phrases'
record in their memory tended to be fleeting. In contrast, the other three groups made
richer encoding of the collocations, especially the two rule-oriented groups who
performed certain structural analysis of the verb's collocational features, so their retention
of the target phrases tended to be much more long-lasting.

At this point, I am going to look at another two issues in the result: (1)The two
rule-oriented groups produced more old collocates than the memorization for recall group;
(2) but the superiority lacked statistical significance.

First, why were the rule groups able to produce more old collocates than the
memorization group? As previously analyzed, the major cognitive process under the
memorization for recall condition is maintenance rehearsal, which involves holding

information over a brief period of time. For the two rule-oriented conditions, the

186

dominating process is elaborative rehearsal, which involves the Ieaming and testing of
rules and the embedding of new information into old knowledge system. It has been
assumed that maintenance rehearsal simply prevents forgetting during the process of
maintenance or primes existing units (i.e., making a particular representation active)
without transforming the information into a deeper code, and it is unlikely to lead to
substantial long-term Ieaming; on the other hand, elaborative rehearsal can result in an
increase in the depth at which an item is encoded, promote the establishment of new
associations between two previously separate items, and thus facilitate long—term Ieaming
(see Baddeley, 1997; Craik & Lockhart, 1972; Craik & Watkins, 1973). These views can
explain the fact that the two rule-oriented groups recalled more old collocates than the
memorization group.

But why was the gap between the rule-oriented and memorization condition so close
that there was no statistical significance? This might be related to the complex interaction
between maintenance and elaborative rehearsal. As reviewed in Chapter 2 (2.3.1. and
2.3.2.2.), when representations of input accumulate in memory, inductive Ieaming
mechanisms might automatically extract generalizations, so repetition may create new
Ieaming rather than simply perpetuating the status quo. Craik and Lockhart (1972)

pointed out whether repetition of an item strengthens the memory trace or merely

187

postpones forgetting depends on what the subject is doing with his rehearsal and only
deeper and richer processing will lead to an improvement in memory. A similar view was
expressed by Baddley (1997). According to him, although priming and associative
Ieaming have been discussed as if they were quite separate processes, in reality many
situations involve some elements of both and the distinction between maintenance and
elaborative rehearsal is not as absolute as at first seemed likely. In the current experiment,
the phrase memorization task might have enabled learners to come up with certain
explicit or implicit rules (i.e., rules that can or cannot be verbalized) that help them
organize and retain the input, so learners in the memorization condition were not simply
repeating the phrases. The research design is not able to detect leamers' knowledge of
implicit rules, but the result ofthe exit questionnaire did show that some participants
from the memorization group acquired a mixture of instances and explicit collocational
rules. In short, the lack of statistically significant superiority of the rule-oriented groups
over the memorization group in recalling old phrases can be attributed to the possibility
that members under the memorization condition possess a blend of instances and
rule-based knowledge.

Before concluding the analysis of the overall effect of attention on the production of

old collocates, I would like to discuss another interesting phenomena reported in Chapter

188

 

4. That is, the semantic processing group was the most inferior not only in the mean

NUMBER of old collocates produced but also in the TYPES of old collocates produced.

Some of the target noun collocates were able to be recalled ONLY by learners from the

memorization for recall, rule given and rule given plus negative evidence groups, and not

the semantic processing group: (1) mind (collocate for warp), (2) recovery, progress,

claim (collocates for make). (3) lead, role (collocates for take). These nouns share one

common feature: they all denote something abstract, with the target verb used in its

figurative or delexical sense. We may get two implications from this result: First, when a

verb is used in its literal sense and the following noun denotes something concrete, the

verb+noun combination might be stored in LTM through mere semantic processing,

although. as analyzed earlier, the retention chance is much lower compared to the focal

attention conditions; Second, when a verb is used in its figurative or delexical sense and

the following noun denotes something abstract, there is a higher demand on focal

attention for learners to keep the verb+noun combination in LTM.

5.2.1.2. The interaction between attention and proficiency level

In the production of old collocates, the effect of attention was mediated by the factor

of proficiency level. It was found that for Level 2 learners, the rule given group still

significantly outperformed the semantic processing group; however, the advantage of the

189

 

 

A_.z._ Q—L-fi- . .N . _

memorization for recall and rule given plus negative evidence conditions over the
semantic processing condition was not statistically significant. This part attempts to
analyze the reasons why the behavior of the memorization for recall and rule given plus
negative evidence groups of Level 2 was deviant from the overall pattern.

5.2.1.2.]. The memorization for recall group (Level 2)

One issue to be addressed here is why, different from Level 4, for Level 2
participants the superiority of the memorization for recall group over the semantic
processing group fell short of statistical significance. This differential effect of attention
might be related to the different processing features of Level 2 and Level 4 learners.

A number of authors have argued that to master complex cognitive tasks. two types
of operations are required to be integrated -- controlled processes and automatic
processes" (see e.g., Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977). Controlled
processes are subject to the attentional control of a person. They require relatively large
amounts of processing capacity and time, so they are slow, inefficient, and limited by the
capacity of STM. Automatic processes take much less processing energy, so they are fast,
efficient, not limited by the STM capacity and difficult to suppress or modify. According

to Shiffrin and Schneider, the development of cognitive skills involves a shift from

 

° These two concepts were touched upon in previous parts of this dissertation, and here a more formal and
complete characterization of them is provided.

190

 

controlled to automatic processing, which results from repeated exposure and rehearsal.

Novice learners need to pay careful attention to every component of the task. But

advanced learners, as a result of practice, may be able to process more information

automatically and controlled processes are free to be allocated to hi gher-level subtasks or

other components of the task. From these perspectives of information processing, L2

Ieaming is thought to be regulated by controlled processes, in which complex skills

become automatic only after the earlier use of controlled processes (see e. g., McLaughlin,

Rossman, & McLeod, 1983; McLeod & McLaughlin, 1986).

In the current study, when Level 2 learners (the lower-level leamers) were asked to

memorize the target phrases for a later recall test. they might still need to consume most

of their attentional resources on certain surface elements. like the form and meaning of

each individual word. That is, there were many elements in the input that were still

subject to controlled processing. However, for Level 4 Ieamers (the hi gher—level learners),

the mastery of more advanced knowledge and skills in English might have enabled them

to process more input effortlessly and automatically so that some of their attentional

resources were freed up for deeper-level processing of the input, for example, discovering

the underlying regularities of the collocations. In other words, Level 2 learners tended to

process the target phrases in a more shallow level than Level 4 learners. Because it is

191

 

deeper coding rather than sheer repetition of the phrases that enhances their long-term

retention (see 5.2.1.1), the memorization tasks were less effective with Level 2 learners

than with Level 4 learners.

5.2.1.2.2. The rule given plus negative evidence group (Level 2)

The second question to be answered is why for Level 4 the rule give plus negative

evidence group had statistically significant advantage over the semantic processing group

in producing old collocates, but for Level 2 this is not the case. This differential effect of

attention might be again associated with the different processing features and capacity of

Level 2 and Level 4 learners.

Chapter 2 (2.2 and 2.5.4) has reviewed two important features of attention: (1)

Attentional capacity has its limitations and once this limit is exceeded Ieaming is

impossible; (2) The effect of attention interacts with proficiency level in the sense that

extemally-driven focus on a new item is effective only when learners are proficient

enough to release attention for that particular item during input processing. As suggested

in 5.2.1 .2.1, for hi gher-level learners, the amount of knowledge they have acquired

enables them to handle more information routinely and "effortlessly" than lower-level

learners and a subtask that once took up a lot of processing capacity may become so

automatic that it requires little processing energy, thus advanced learners have more

192

 

attentional resources than the less advanced ones to process new information. In other
words, the attentional resources of an L2 learner can be constrained by the degree of
automaticity he/she has over comprehension and production in their L2.

In this study, it is possible that when facing the input presented for the rule given
plus negative evidence condition (i.e., positive evidence, rules and negative evidence),
Level 4 learners could attend to, deal with and organize it quite effectively within the
given time, including storing in LTM the collocates that appeared in the passage.
However, for Level 2 learners, those task demands may have overloaded their limited
capacity system in such a way that their storage of the old collocates was distracted rather
than facilitated. As suggested by Baddeley and Hitch (1974), once the working memory
limits were exceeded. there would be a trade-off between the total amount of information
that could be retained and the efficiency in processing the information that comes in later.
As a result, some of the stored information would be forgotten, and/or processing would
be erroneous or be slowed down.

In sum, Level 2 learners were able to process and assimilate less amount of English
input stimuli than Level 4 learners within the same length of time, and the input presented
in the rule given plus negative evidence condition may have exceeded their processing

capacity, so certain negative effects were found (in this case, the retention of old

193

 

collocates was influenced). However, for Level 4 learners, who tended to be more
efficient input processors, the amount of information provided in the rule given plus
negative evidence condition was still manageable, so no side effects were observed.
5.2.2. Froduction of new target-like collocates
5.2.2.1. Overall effects of attention

As reported in Chapter 4, in general, the rule given and rule given plus negative
evidence groups produced significantly more new target—like collocates (i.e., new
collocates that were consistent with the target rules) than the memorization forrecall and
semantic processing groups. To understand the implications of this result, we need to
look at the fundamental difference between the rule given, rule given plus negative
evidence groups on the one hand and the memorizationfor recall, semantic processing
groups on the other hand.

For the rule given and rule given plus negative evidence groups, they were presented
with collocational rules and guided in explicit discussions about the target collocational
_ features. As described in Chapter 4, during the treatment stage, most learners in the two
conditions could follow the discussions quite well and complete the rule matching
activity satisfactorily. Moreover, when they were asked to recall the target rules in the

exit questionnaire, most of them could write down parts or all of the rules. In other words,

194

 

the two experimental conditions were successful in inducing focal attention at the level of
understanding collocational rules.

For the memorization for recall and semantic processing groups, no rules were
explicitly given to them. The exit questionnaire shows that only a small number of
learners in the semantic processing condition could verbalize any collocational rules, and
those verbalized rules were not the same as the target rules. As to the memorization for
recall group, although there were more participants explicitly talking about certain
collocational rules in the questionnaire, very few of the patterns found by learners
themselves were the same as the target rules.

In short, the basic difference between the rule given. rule given plus negative
evidence groups and the memorization for recall. semantic processing groups was
whether the learners possessed any explicit knowledge of the underlying collocational
rules. The significant superiority of the two rule-oriented groups over the two
non-rule-oriented groups in the production of new target-like collocates suggests that
focal attention at the level of explicit rule knowledge can speed up the acquisition of
certain recurrent pattern in collocational behavior and learners do use the rules to make
new creations. Without the rule-level focus, learners tend to ignore the underlying

collocational regularities or fail to find the correct patterns, at least when the encounters

195

 

with the target collocations are limited to a few times, so, not surprisingly, their
production contains fewer new collocates that follow the re-occurring pattern. Here, I do
not rule out the possibility that with numerous encounters and longer Ieaming time, even
learners in the non-rule-oriented condition may automatically develop a sense of a word’s
frequent type of collocates. However, compared to the rule-oriented condition, this
Ieaming process is obviously much slower and less efficient.

To sum up, the positive effect of explicit rule formulation observed in this part of L2
collocation Ieaming is quite similar to what was found in some studies of L2 syntax
Ieaming (see the review in 2.5.1 .3.). That is. rule-based condition helps learners bring
order to the input and also enables them to generalize the knowledge to novel instances.
5.2.2.2. The interaction between attention and proficiency level

Like the production of old collocates, in the production of new target-like collocates.
the effect of attention was also mediated by proficiency level and it was again Level 2
learners who deviated from the overall pattern. As displayed in Chapter 4, for Level 2
learners, only the rule given group produced significantly more new target-like collocates
than the memorization for recall and semantic processing groups, and the rule given plus
negative evidence group did not have any significant difference with the two

non-rule-oriented groups, although there was a trend toward an advantage for the rule

196

given plus negative evidence condition. This result might be again related to the
possibility that the addition of negative evidence overloaded the processing capacity of
Level 2 learners, who were less effective information processors (see previous discussion
in 5.2.1.2.2.), and it seems that the excessive amount of input not only inhibited Level 2
leamers' LTM storage of old collocates but also interfered with their learning of the target
collocational rules, which serve as an important basis for making new extensions in
collocations.

However, as shown in Chapter 4, there was no significant statistical difference
between the rule given plus negative evidence (Level 2) group and the rule given (Level 2)
group in terms ofthe mean number of rules that could be completely recalled, partially
recalled or not recalled in the exit questionnaire. In other words, the rule given plus
negative evidence (Level 2) group did as well as the rule given (Level 2) group in
verbalizing the collocational rules. Then why the former group did not have the same
superiority over the two non-rule-oriented conditions as the latter in producing new
target-like collocates? And in what sense did the addition of negative evidence interfere
with the Ieaming of target rules among the rule given plus negative evidence (Level 2)
group? The remaining part of this section will attempt to examine these two questions.

As stated in 5.2.1 .2.1, controlled processing can be considered as laying down the

197

 

"stepping stone" for the automatic processing as L2 learners makes progress towards
more and more difficult levels. Following this line, we may speculate that the acquisition
of collocational rules proceeds in several stages:

First, learners notice the underlying collocational rules. In this experiment, this is
aided by external instructions.

Second, the rules might have been memorized or stored, but learners are not sure
about how to apply them yet or the operation of the rules is still a controlled and effortful
process. At this stage, learners might be able to verbalize the rules when they are required
to do so, but target-like performance in language production is not guaranteed (also see
previous literature review in 2.5.1.3.). This way oflooking at Ieaming is similar to the
argument by Gass (19883) that some input may be processed and put into storage ifit is
not yet possible to integrate it into the IL system and this storage can take the form of
some kind of explicit representation ofL2 items and rules.

Third, the originally controlled and attention-demanding operations of the rules
become much faster and more automatic. At this stage, learners not only have rule
knowledge but also become more efficient in making use of the rules to generate new
instances in production.

It has to be noted that the stages proposed above may not be distinguished very

198

 

clearly in real acquisition, because the process of automatization is a gradual and
continuous transition from fully controlled processing to fully automatic processing (see
Schneider & Detweiler, 1988).

In this experiment, it is possible that for the rule given plus negative evidence (Level
2) group, the information flood in the input did not make learners learn more. Instead, it
delayed their progress from Stage 2 to Stage 3 in the acquisition of collocational rules. As
a result, learners could verbalize the rules, but couldn't apply the rules in production as
well as the rule given (Level 2) group did and their production of new target-like
collocates failed to show statistically significant superiority over the two
non-rule-oriented (Level 2) groups.
5.2.3. Production of non-target-like collocates

As reported in Chapter 4, the memorization for recall, rule given and rule given plus
negative evidence groups produced significantly fewer number of non-target-like
collocates (i.e., collocates deviating from the target rules) than the semantic processing
group. The following analysis will focus on three major questions in this result:

(1) Why was the rule given group able to produce significantly fewer non-target-like
collocates than the semantic processing group?

(2) Why did the rule given plus negative evidence group have no significant

199

superiority over the rule given group in avoiding non-target-like collocates?

(3) Why was the memorization for recall group able to produce significantly fewer

non-target-like collocates than the semantic processing group?

5.2.3.1. The rule given group vs. semantic processing group

The superiority of the rule given group over the semantic processing group in

avoiding non-target-like collocates has the following indications: although theoretically

speaking, the presented rules only tell learners what are allowed in L2 but not what are

disallowed, they seem to have the function of helping learners realize that they should be

cautious with new creations and not all the nouns that are semantically compatible with

the basic meaning of the verb are appropriate as its collocates. In other words, explicit

knowledge of the recurrent patterns make learners prefer to produce collocates that

comply with those patterns. On the other hand, the semantic processing group only paid

peripheral attention to the target collocations, which was insufficient to induce awareness

of any recurrent patterns. Without rules serving as constraints, it is not surprising that

learners under the semantic processing condition produced much more collocates that did

not observe the target rules.

5.2.3.2. The rule given plus negative evidence group vs. rule given group

Different from the rule given group, the rule given plus negative evidence group was

200

 

presented with not only the collocational rules, but also the information that some nouns
semantically compatible with the basic meaning of the verb may not be suitable as its
collocates. Moreover, according to Chapter 4, the majority of the participants (both Level
2 and Level 4) could understand the message conveyed by the negative evidence,
considering their answers to the question "Why do yOu think is this sentence bad?".

In short, the rule given plus negative evidence group learned the combinational
restrictions both from positive evidence (recurrence of certain patterns) and from
negative evidence. Theoretically speaking, it should have more keen awareness of the
restrictions than the rule given group and produce fewer non-target-like collocates. Then
why did its superiority over the rule given group fall short of statistical significance?

One possibility is that negative evidence is not useful at all. This is not likely,
because we did observe a trend towards an advantage for the rule given plus negative
evidence group. Moreover, in terms of j udging bad collocations (an item in the judgment
test), the rule given plus negative evidence group achieved significantly higher mean
score than the rule given group. These facts indicate that negative evidence is useful in
constraining leamers' hypothesis. But why did it show obvious facilitative effect in
judgment test but not in production test?

This is an issue related to task-induced variation. Many studies have reported that IL

201

may vary when learners perform different tasks. For example, in a study on second-verb
ellipsis in sentences like Stella is eating a banana and Alice an apple, M. Schmidt (1980)
found that in free speech learners always included the second verb, but they increasingly
omitted it in tasks like written sentence-combining and grammaticality judgment.
According to Hyltenstam (1984), different tasks (i.e., written composition,
grammaticality judgments and oral picture description task) resulted in both quantitative
difference in IL (i.e., more or less target-like behavior) and qualitative differences (i.e.,
different or even contradicting rules).

Various reasons have been put forward to account for the observed task-induced
variation. M. Schmidt (1980) argued that substantial differences arise depending on the
degree of monitoring that different tasks permitted R. Ellis (1994a, p. 140) claimed that
the variation is related to whether the task permits or encourages the use of metalingual
knowledge or whether it taps "communicative" behavior. According to Lococo (1976),
leamers' perception of the task might be one factor influencing the results. Learners
might focus on accuracy in a translation task and on expressing ideas in a picture
description task. Tarone (1982) maintained that when learners pay attention to their
speech, controlled processes come into play, so their performance is likely to be different

from that in informal settings. Taking these views as a whole, we may draw the following

202

conclusion: Task-induced variation may occur when a particular language item hasn't
been automatized yet. In that case, if a task allows enough time for controlled processing,
encourages the application of explicit rules, or is perceived to be focusing on accuracy
rather than just communication, IL production tends to be more target-like; otherwise, the
IL performance may deviate from the target norm.

In the current experiment, when learners were asked to judge whether a sentence
was a good English sentence or not and also identify any error, it might be a natural
tendency for them to draw on their explicit knowledge about how English should work
and then use that as the criterion for judgment. In other words, the judgment test is a
format that encourages the use of metalingual knowledge. On the other hand, the
production test, which asked Ieamers to write as many noun collocates as possible within
a given time, was more implicit in tapping leamers' rule knowledge. Learners might have
focused on retrieving those nouns they considered as semantically suitable for the verb,
while ignoring the collocational knowledge they have stored, unless the knowledge has
been automatized or the experimental time was long enough for controlled processing. So,
compared to the judgment test, target-like performance in the timed production test relies
more on a learner's ability to apply knowledge automatically.

The fact that negative evidence showed its significant effect in the judgment test but

203

not in the production test seems to imply that learners have stored the collocational

restrictions conveyed by negative evidence, but it hasn't been acquired to the extent that it

can be applied very fast or automatically in production. In other words, just like the

gradual assimilation of the target collocational rules (see previous discussions in 5.2.2.2.),

the integration of the information passed by the negative evidence may also develop in

several phases: (1) understanding of the knowledge contained in the negative evidence -)

(2) storage ofthe knowledge, which might be utilized in ajudgment test or in fully

monitored production -) (3) being able to operate the knowledge automatically in

production. In this research, the participants under the rule given plus negative evidence

condition might have only reached the second stage. so the effect of negative evidence

was noticeable only in judgment but not in production.

This explanation leads to our final question in this section: why did the participants

in the rule given plus negative evidence condition fail to reach the third stage in acquiring

the information conveyed by the negative evidence? This might have to do with the

timing of the focused attention, which was considered in Chapter 2 (2.5.4.). In the current

experiment design, the negative evidence was presented immediately after the flood of

positive evidence and rules. At that time, learners might not be developmentally ready for

it yet. That is, they might need more time to digest the information conveyed by previous

204

items (i.e, the positive evidence and rules), consolidate their knowledge base and

gradually develop some emerging hypotheses about the collocational restrictions, before

they are able to make full use of the negative evidence. As a result, the "premature"

presentation of the negative evidence only led to rule or item storage that may be used in

the judgment test, but not the type of automatic knowledge application demanded by the

timed production test. As pointed out by previous research, direct instruction should be

delayed until learners have shown at least some emerging knowledge of the form

(Doughty & Williams, 1998) and "learner's emergent grammars modulate their

apperception of the input and therefore the use they might make of explicit instruction"

(Mackey, Philp, Egi, Fujii & Tatsumi, 2002, p. 203).

5.2.3.3. The memorization for recall group vs. semantic processing group

The third question with regard to the production of non-target-like collocates is why

the memorizationfor recall group was able to produce significantly fewer non-target-like

collocates than the semantic processing group. Previous analysis shows that very few

members in the memorization group had explicit knowledge of the target collocational

rules. So it seems that this group should have behaved like the semantic processing group

in making overextensions, due to the lack of constraints from any rules or any correct

rules. However, the reality is that the memorization for recall group was as good as the

205

two rule-oriented groups and produced significantly fewer non-target-like collocates than

the semantic processing group.

This result might be related to the task demands for the memorization for recall

group in the treatment and test. During the treatment stage, this group was asked to

memorize the target phrases and told that there would be a recall test of the phrases later.

Then, in the test they were asked to produce as many noun collocates of the target verbs

as possible. It might be a natural tendency for the learners to try their best to recall all

those phrases they read before. As a result, the time left for making new creations was

reduced, and when learners did try to produce novel collocates, they might have tried to

produce those nouns that came from the same lexical field as the old ones (e.g., (warp) a

bookshelf -) (warp) a desk). Because of these factors. the chance of producing

non-target-like collocates was much lower for the memorizationfor recall group than for

the semantic processing group, who was poor in both rule and instance knowledge. This

explanation can gain some support from the results reported in Chapter 4: Overall, the

memorization for recall group produced the fewest number of noun collocates and,

different from other groups, among their production, old collocates tended to occupy a

much larger percentage than new collocates.

206

5.2.4. NS judgment of IL production

Most new target-like collocates, that is, new collocates that are consistent with the
target rules, were judged as good or OK by NSs. This is not surprising. However, it has to
be noted that there were still some target-like collocates judged as bad by NSs. This can
be accounted for by the nature of collocations as summarized in Chapter 2. That is,
collocation is a PARITIALLY rule-govemed system. Certain rules can capture the
common features of the allowed combinations, but some phrases that comply with the
rules seem to be excluded arbitrarily. Therefore, the rule-based extensions made by
participants in this study were not always accepted by NSs. This complex nature of
collocations indicates that both acquisition of rules and accumulation of instances might
play a crucial role in learning L2 collocations. With this dual knowledge base, learners
can develop a good sense of what are frequent combinations and what are infrequent.

Another issue is the NS judgment of non-target-like collocates. The results are
somewhat different for wield, warp and make, take. For wield and warp, most
non-target—like collocates were judged as bad and no non-target-like collocates were
judged as good. However, for make and take, much more non-target-like collocates were
judged as good or OK. This result has to do with the different collocational features of

wield/warp and make/take. When wield and warp are used in the target basic meaning,

207

their collocates are mostly limited to the two types of nouns displayed in the passages.
But make and take are more collocationally active and can be combined with nouns from
a variety of semantic fields. What were targeted in this experiment were only two types,
so even if learners produced something outside the scope of the target types, there was
still a good chance for the IL production to be acceptable.
5.3. Attention and judgment of collocations
5.3.1. Judgment of good collocations

In terms of the mean score of judging good collocations, attention did not have
significant main effect and all the participants did this part very well. This might be
related to the fact that they have successfully acquired the basic meaning ofthe target
verbs. As shown in Chapter 3 (3.2.3.2.), in the eight good sentences used forjudgment,
except two phrases (i.e., take a walk, make a shift) in which the basic meaning of the
target verb and the meaning of the following noun seem to have remote semantic and
logic relationship, the combinations of the target verbs and their noun collocate are
semantically quite transparent (e.g., wield heavy swords, wield authority, the wood...
warped, warp her personality, make the shortest speech, take the part of Hamlet).
Knowing the meaning of the words, very naturally the participants tended to judge the

good sentences as good.

208

 

This result is consistent with the finding made by Fan (2004). In that study, 51
Mandarin speakers enrolled in 2nd-, 3rd—, and 4th-level English courses in a university in
China were exposed to passages containing the margin glossing and collocational
patterns of germinate, muster, warp, which were used in their metaphoric sense. The
participants were placed into one of three attention conditions: [- focused attention]
(attention directed towards aspects unrelated to the target collocational patterns, a
condition comparable to semantic processing in this study), [+focused attention]
(attention directed towards the targets in the form of collocational rule search, a
rule-oriented condition resembling rule given in this study) and [++focused attention]
(attention directed towards the targets in the form of collocational rule search and
provision of negative evidence, a rule-oriented condition resembling rule given plus
negative evidence in this study). In the test of that study, all the participants did very well
in acquisition of the basic meaning of the target verbs and there was no significant
difference among the three attentional groups in terms of the judgment of good
collocations.

5.3.2. Judgment of bad collocations
In terms of the mean score of judging bad collocations, the rule given plus negative

evidence and rule given groups significantly outperformed the semantic processing and

209

memorization for recall groups. This again matched the result in Fan (2004), which found
that the two rule-oriented groups (i.e., [++focused attention] and [+focused attention])
significantly outperformed the non-rule-oriented group (i.e., [-focused attention]) in
judging bad collocations. These results reconfirm the conclusions made in 5.2.3., that is,
explicit knowledge of rules helps learners narrow down their hypotheses concerning a.
word's possible collocates and enables them to recognize certain inappropriate
combinations.

Moreover, the rule given plus negative evidence group significantly outperformed the
rule give group in judging bad collocations, which shows the facilitative effect of
negative evidence in preventing over-extensions. The detailed explanation was offered in
5.2.3.2., so it will not be repeated here. In Fan (2004), there was also a trend towards the
advantage ofthe [++ focused attention] group (i.e., the rule-oriented group provided with
both positive and negative evidence) over the [+focused attention] group (i.e., the
rule-oriented group provided with positive evidence only) in judging bad collocations,
but the superiority showed no statistical significance. This might be caused by these
factors: (1) Like the current study, in Fan (2004), the negative evidence was presented
immediately after the positive evidence, which might be too early for learners to fully

integrate the information (see previous discussions in 5.2.3.2.). Moreover, in this study,

210

 

after learners were informed that a particular sentence was bad, they were asked "Why do
you think is this sentence bad", which was useful in promoting further thoughts on
leamers' side and facilitating Ieaming; but in Fan (2004), learners were simply informed
that a sentence was bad, without any further prompting question. As a result, the effect of
the negative evidence might have been compromised. (2) The number of participants in
Fan (2004) might not be large enough for it to be sensitive to the difference between the
[+focused attention] and [++focused attention] group.
5.4. The lack of interaction between attention and complexity

As previously analysed, focused attention to the target collocations at the level of
rule understanding had a facilitative main effect on the production of old collocates, new
target-like collocates, non-target-like collocates and the judgment of bad collocations.
However, in none of these tasks was the effect of attention mediated by the complexity of
the target collocational rules, which was different from the prediction made in the
research hypotheses.

One possible reason for the lack of interaction between attention and collocational
complexity in this study is that the claimed difference between the collocational-pattems
of make, take and those of wield, warp did not exist at all. But this is not likely, because

as reported in Chapter 4, wield, warp did produce results that were significantly different

211

from make, take. In terms of producing old collocates, producing new target-like
collocates and avoiding non-target-like collocates, leamers' performance with wield and
warp (simple patterns) was significantly better than that with make and take (complex
patterns). These results are not surprising. considering that all these tasks could be
facilitated by rule knowledge, and it is more difficult and effortful to discover, encode
and assimilate the rules of make and take than those of wield and warp. Meanwhile, there
is an interesting phenomenon that needs to be mentioned here. That is, in judging bad
collocations, leamers' performance with wield and warp (simple patterns) was inferior
rather than superior to that with make and take (complex patterns). This might be related
to the fact that in the sentences for judgment make and take were used in relatively more
abstract sense and had more distant semantic relationship with the noun collocate. so
learners tended to judge them as bad. In short, the claimed complexity difference between
the collocational patterns of make, take and those of wield, warp did exist.

A more tenable explanation for the lack of interaction between the effect of attention
and collocational complexity is that for the participants of this study wield/warp and
make/take were both located in the middle of the complexity continuum, although they
differ in the degree of complexity. As reviewed in Chapter 2 (2.5.3.2.), the effect of

externally imposed focal attention may be influenced by the side of the complexity

212

 

...—..._ .:_‘fr _-‘ . .—

continuum the target learning items sit on. When the Ieaming material is located on one
of the two extremes, that is, very simple or very complex, the effect of focal attention to
the underlying rules might not be that obvious. For very complex material, rule-oriented
attention might only cause confusion; for very easy material, learners might detect the
rules themselves, so the superiority of externally-driven focused attention will be again
decreased. The effect of focused attention is most obvious when the material to be
learned is in the middle of the complexity continuum. That is, the language pattern has
certain degree of complexity and requires some efforts to be figured out, but the
explanations of the pattern or rule are not too complicated and are able to be assimilated
by the learners. These features might be exactly what the target items of this study have:

First, for the rules underlying the collocational behavior of wield and warp, although
they are easier to figure out than those of make and take, the accurate target rules are still
quite difficult to be found out by learners on their own and it is by no means easy to
delimit the exact semantic border of the noun collocates on the basis of the limited
number of input instances. For example, exposure to the noun collocates of warp in the
passages (i.e., stick, weapon, knife) may make learners think that warp can be followed
by noun collocates that denote concrete objects, but the input instances may not be

sufficient for learners to hypothesize that warp can be followed by a specific type of

213

 

 

concrete object, that is, a concrete tool or piece of equipment that can be carried and used.

As shown in the exit questionnaire, no learners in the non-rule-oriented conditions (i.e.,

semantic processing and memorization for recall groups) could figure out the target rules

of wield or warp on their own, although some of their findings were quite close to the

targets

Second, for the rules underlying the collocational behavior of make and take,

although they are more complex than those of wield and warp, it seems that all the

participants in this research, both Level 4 and Level 2, were able to cope with the explicit

instruction of the rules, and my research hypotheses seem to have underestimated Level 2

leamers' ability to take in and make use ofthe quite complex rules. There are two factors

that may have contributed to the success by Level 2 learners: (l ) There were only 2 rules

for make and take respectively and the rules and corresponding examples were repeated

for multiple times during the treatment stage; (2) The Level 2 learners were university

students who had learned English for at least 6 years, and they may have accumulated

enough knowledge and skills to deal with the complex target rules, although they were

less advanced than the Level 4 learners.

To sum up, for the participants of this study, both the hard and easy items targeted in

this study fall into the middle range of the complexity continuum rather than at the very

214

hard or easy extreme. Therefore, focused attention at the level of rule understanding was
effective with both types of items.
5.5. Summary

I will now put together previous analyses of the experiment results and relate them
to the three research questions of this study.

The first question is concerned with what type of attentional condition leads to the
best gains in Ieaming a verb's possible noun collocates.

The results of the present study offer evidence that focused attention at the level of
rule understanding is the most facilitative. Explicit knowledge of collocational rules has
multiple functions: it can help learners bring order to the input, retain the phrases they are
exposed to, and narrow down hypotheses about a word's collocational possibility; and the
rules are also used by learners in making new extensions. Within a length oftime as short
as the present study. mere semantic processing of the word sequences is insufficient to
induce the discovery of the recurrent patterns. Therefore, learners under this condition are
inferior in both LTM storage of the old phrases and new extensions; they also tended to
over-generalize a word's collocational behavior. It is possible that with the increase of
exposure to a word's collocations, learners under the semantic processing condition might

detect certain regularities on their own, but obviously this process is much slower.

215

 

In this study, focused attention at the level of phrase chunking and memorization
only showed its superiority in transferring the old phrases into LTM. However, the
advantage of this condition may be beyond this for the following two reasons: (1)
Collocation is a system characterized by rules and also a lot of seemingly arbitrary
exclusions and inclusions. Dual knowledge of rules and instances might help to nurture a
sense of combination frequency and make leamers' production more native-like and
idiomatic. (2) Explicit rule knowledge is useful when a word's collocational features can
be captured by a small number of rules, but it is not necessarily so when the rules are
miscellaneous and numerous. for example, those for delexical verbs. In the latter case, a
more realistic way of Ieaming might be simply paying some special attention to the
phrases and gradually develop the knowledge of the word's frequent collocates. It has to
be pointed out that these suggestions concerning the usefulness of a combination of
rule-level and instance-level focal attention are speculative and further research is needed
to address this issue. Moreover, we should always keep in mind that real situations very
often involve components of both types of processes, as indicated by previous
discussions about the interaction between maintenance and elaborative rehearsal.

A final point to make with regard to the best type of attentional condition is related

to the role of negative evidence. From the current study, we can conclude that negative

216

 

evidence in L2 collocations can be properly perceived by learners and is utilized by them.
It can speed up the process of recovering from an originally over-extended collocational
grammar.

At this point, we can also offer an answer to the question raised at the beginning of
this dissertation, that is, why is the development of L2 collocational competence so
difficult for learners? According to this study, focal attention is a crucial mechanism in
the acquisition of L2 collocations. However, knowing a word's collocational features is
not indispensable for meaning comprehension and those features tend to escape leamers’
focal attention (see discussion in 3.1.). Therefore. the development of collocational
knowledge is relatively slow and learners tend to rely solely on the compatibility of basic
meaning or L1 knowledge to combine words, which can cause various types of
production errors (see the review in 2.4.1). Moreover, as mentioned in 2.4.1, learners
make fewer errors with highly-frequent phrases. This might also be related to the
attentional issue, that is, frequent exposure to certain items can induce leamers' focal
attention to them and thus promote Ieaming.

The second research question I seek to address is whether the effect of attention is
mediated by complexity of collocational patterns. This study observed no interaction, but

this may not be used as evidence for a simple "no" answer to the question. It is possible

217

that although the target items in this study display a difference in collocational
complexity, none of them was either too easy or too hard for the participants. If there
were certain very easy or very difficult items involved in the experiment, an interaction
between attention and complexity may have been found. Further research is required to
explore this issue.

The third research question is whether attention has differential effects on different
proficiency levels. The results of this study support a "yes" answer; moreover, it turns out
that the differential effects can be best accounted for with an information-processing
approach, which considers the difference between lower-level and hi gher-levelileamers in
the efficiency of input processing, that is, the ability to attend to. deal with and organize
new information. Within the same amount of time, higher-level learners are able to deal
with more information than lower-level learners, because advanced learners can handle
more knowledge and skills automatically and have more attentional resources available
for new information, while lower-level learners are more dependent on controlled
processes which take up a lot of processing energy and time. In this study, Level 4
learners could deal with the input provided in the rule given plus negative evidence
condition, but for Level 2 learners, the information seemed to have overloaded their

processing capacity, and caused the interference with the retention of old collocations and

218

also the assimilation of the instructed rules. Moreover, for hi gher-level learners, more
attentional resources may be freed up for processing the input at a deeper level and
abstracting regularities from linguistic stimuli, which can promote the transfer of
language material into LTM. As a result, the memorization for recall condition was more
effective with Level 4 students than with Level 2 students in facilitating the retention of
the target phrases.

In short, in the current study, the lower- and hi gh- level learners seemed to have
processed the same input treatment differently, both in amount and depth. Therefore, the
effect of attention varied.

Besides these, there are another two issues emerging from the study that are related
to the theme of processing capacity. They are about the infomiation processing features
of learners as a whole, not just those of lower-level or hi gher-level learners.

The first is the integration of certain items into IL system. This is a dynamic and
interactive process, which requires building up a series of well-leamed, automatic
procedures so that controlled processes are freed for new Ieaming. In the current study,
the Ieaming of target collocational rules by the rule given plus negative evidence (Level 2)
group and the assimilation of negative evidence by the rule given plus negative evidence

(Level 2 and Level 4) groups seemed to have only progressed to the point of controlled

219

 

processing. As a result, the knowledge representation was only strong enough to be
reflected in tasks that tend to tap metalingual knowledge (e. g., rule recall and sentence
judgment), but not a task that has higher demands on automatic processing (e. g.,
language production within limited time).

The second is the timing of focused attention. Focused attention to particular items
is more effective when learners have acquired and consolidated enough background
knowledge to be developmentally ready to handle those items. In this study, the negative
evidence was provided immediately after a great amount of positive evidence rather than
after learners were given enough time to fully digest previously learned infomiation. This
presentation of negative evidence might have been too early, so its effect was obvious
only in thejudgment test, a format that encouraged learners to make use of explicit rule
knowledge they have stored.

To sum up, in this research, I found evidence that focal attention at the level of rule
understanding and provision of negative evidence can speed up the Ieaming of a verb's
possible collocates. Moreover, the effect may not be immediately visible in language
production and it will be most noticeable when the amount of input stimuli and timing of

focused attention are compatible with learners' processing capacity and features.

220

Appendix A

Treatment exercises for semantic processing group (Day 1- Day 3)

Day 1 (semantic processing group)

Task 1: Please read the following two passages in 7 minutes. After this, you will be

asked to do some exercises related to the passages.

Passage 1

 

 

For many of us, the Internet is a very important part
of our life. We use it for communication, shopping, or
getting all types of information, from news about rebels
who surrounded a government building, while mandingl
sticks and throwing stones, to tips about how to prevent
heat from nerting2 the floor. However, the network has
also caused some problems.

With the Internet placing a wealth of research
papers at our fingertips, some students plagiarize3. To
prevent this from happening, two graduate students at
the University of California have devised a special
computer program. "We will run students' paper through
the program, voke4 a look at every sentence. and
compare each one with every other term paper on the
web," said John, one of the program designers.

A professor who teaches biology at U.C. Berkeley
told his class that he would use the program. However,
numerous students copied passages from others' work
without citations5 anyway. He ran all 300 papers through
the program and found 45 of them, or 15 percent of
students, had cut significant amounts of material from
the lntemet, dained6 no changes and turned in the
writing as their own work. Sometimes, students might be
falsely blamed. In that case, they can have the
opportunity to defend themselves. Competition is tough
at famous universities like the University of California.
Some students welcome the Internet research watchdog7,

considering it as justified. (230 words)

 

l.mand: to have and/or use something

2.nert: to change something for the

\VOTSC

3. plagiari:e: to take (words. ideas.
etc.) from someone else’s work and
use them in one's own work without
admitting one has done so

4.voke: to get something

5.cttation: a short note recognizing a
source of information or of a quoted
passage

6.dain: to produce something

7.watchdog: a person or organization

that defends against illegal practices

 

221

 

Passage 2

 

Nowadays in many hospitals, when a patient calls
for help, it's never a problem finding the nurse, because
nurses all wear special badges'. Since the invention of
the badge, many US. hospitals have voked2 the lead in
using it. It is actually a part of a wireless system, which
uses infrared3 signal to track nurses' movements on the
floon

But some nurses worry that the devices will be used
to listen in on private conversations and monitor their
movements. "You know, it can be used to listen through
a wall unit to nurses or even to patients in other rooms.
I'm not wearing mine. It makes me feel that someone is
trying to mand4 control over us. I will still meet my
patients' needs but I will not wear this badge." said
Annette, one of the dozens of nurses who dained5 the
claim that they would never wear the badges and staged
an active protest against them.

However, Michael, a hospital official, said, "We are
never interested in listening to gossip“, gossip about a
man with a nerted7 mind, or someone's love affairs...
What we are most concerned about is how long it takes
nurses to respond to calls. We are interested in patient
care." Many officials believe that once the nurses
actually find out how useful the system can be for them

and the patients. the system will become very popular.
(230 words)

 

 

l. badge: a small flat object worn to
show a person’s name, rank, job or
membership in an organization

2. voke: to get something

3. infrared: a type oflight that feels

warm but cannot be seen

4. mand: to have and/or use something

5. (lain: to produce something

6. gossip: talk about other people's
private lives
7. nert: to change something for the

“'OTSC

 

Please go on to the following exercises. You may go back to the two

passages for answers.

222

 

 

WVV“_ 7...

Task 2: Please answer the following questions in one short sentence. You will have 5

minutes to complete this task.

Passage 1 .

1. What is the purpose of the computer program devised by the two graduate students at
the University of California?

2. What does this program check for in a student's paper?
3. Are there any students who welcome this special program?

Passage 2

1. What is the function of the badge worn by nurses in some hospitals?
2. Why do some nurses refuse to wear the badge?
3. According to hospital officials, what is the major purpose of using the monitoring

system?

Task 3: Please match the words in the left column with the meanings provided in the

right column.

l.nert _ a) to create something

2.mand _ b) to talk about others' private life
3.gossip _ c) to use something

4.voke _ d) to get something

5.dain _ e) to make something worse

Task 4: Please read the following sentences and look for a particular word according to
the instruction. You have 5 minutes to complete this task.
1. Rebels surrounded a government building, while manding sticks and throwing stones.

l
o

In this sentence, which word means "to cause something to move rapidly through the air")

 

2. Tips about how to prevent heat from nerting the floor are available on the lntemet.

In this sentence, which word refers to "a computer network"?

 

223

 

3. We will run students' paper through the program, voke a look at every sentence, and
compare each one with every other term paper on the web.

In this sentence, which word means "a network "?

 

4. ...15 percent of students had cut significant amounts of material from the lntemet,
dained no changes and turned in as their own work. Sometimes, a student might be
falsely blamed.

In these sentences, which word means "to say that someone did something wrong"?

 

5. Since the invention of the badge, many US. hospitals have voked the lead in using it.
It is actually a part of a wireless system, which uses infrared signal to track nurses'
movements on the floor.

In these sentences, which word refers to "a type of light"?

 

6. "It makes me feel like someone is trying to mand control over us. I will still meet my
patients needs but I will not wear this badge".

In these sentences. which word refers to "to satisfy something"?

 

7. Annette was one of the dozens of nurses who dained the claim that they would never
wear the badges and staged an active protest against it.
In this sentence, which word means "to show that you disagree with

something "I"

8. We are never interested in listening to gossip, gossip about a man with a nerted mind,
someone's love affairs... What we are most concerned about is how long it takes nurses to
respond to calls.

In these sentences, which word means "to reply ".7

 

224

Day 2 (semantic processing group)

Task 1: Please read the following two passages in 7 minutes. After this, you will be

asked to do some exercises related to the passages.

Passage I

 

 

It was six o’clock in the evening. Andrea left her office and drove
her car onto the road to home. It seemed that everything went wrong
that day. That morning, she had found rainwater had nertedl the front
door of her house and it was hard to shut it. In the afternoon. her boss
dained2 a speech that was long and boring...When she was thinking
about all these unpleasant experiences, she saw an old lady standing by
the road. with a sign saying "Brockbourne". Normally, Andrea would
never pick up a hitchhiker3 when she was alone, thinking it was
dangerous. However, the cold weather and the coming darkness made
her feel sorry for the lady. Andrea stopped her car, and the lady hurried
over. When the lady got into her car, the old lady said in a very weak
voice, "Thank you -- I'm just going to Brockboume."

Something in the way the lady spoke, and the way she never turned
her head and stared continuously into the darkness ahead. made Andrea
feel a little bit nervous. Careful not to turn her head. Andrea looked
sideways4 at her passenger. She studied the hat. the dirty dress. the arms
with thick black hairs...thick black hairs? Hairy arms? Andrea's blood
froze. This wasn't a woman. It was a man. All kinds of horrible scenes
came to her mind: men mandingS weapons or taking away victims" by
force She didn't know what to do. But then, an idea came to her.

After voking7 a deep breath, Andrea suddenly stopped her car.

"My God!" she shouted, "A child! Did you see the child? I think I hit
her!"

The "old lady" was clearly shaken by the sudden stop. "I didn't see
anything," she said.

"I'm sure it was a child!" insisted Andrea. "Could you just get out
and have a look? She held her breath. Would her plan work?

It did. The passenger slowly opened the car door and climbed out
to check. As soon as she was out of the car, Andrea drove madly away...
(341 words)

 

1. nert: to change
something for worse
2. dain: to produce

something

3. hitchhiker: people
who travel by
getting rides in

others‘ cars

4. sideways: toward

one side

5. mand: to have
and‘or use
something

6. victim: a person
who is harmed or
killed

7. voke: to get

something

 

225

 

Passage 2

 

 

Sharon was married five years ago, but it's not a
happy marriage. According to Sharon, "It seems that he
always wants to mandl power over me. I was feeling...
like depressed and even though I don't really show it.
Self-interest has completely nerted2 his judgment. I
cannot tolerate it anymore. But I am also very worried
about my kids if we get divorced3. "

"The effects of divorce on kids are problems that
could last for many years and emerge in adulthood."
Author Judith Wallerstein said, "When man-woman
relationships move to center stage, all the ghosts of the
parents' divorce might come out of the basementl."

Wallerstein studied 93 children over a generation.
Her findings haven't been published in a medical journal,
only in her book. She says children of divorce are more
likely to abuse5 drugs and dain6 slow progress in their
studies. Also, 40-percent of them avoid marriage
themselves, and when they do marry, the marriages fail at
nearly twice the usual rate. She also says, "These children
define themselves as having a great deal of difficulty
trusting and are very frightened that their relationships
will fail.

But some researchers say Wallerstein might not be
able to draw this conclusion from a small study. Factors
other than JUST divorce should be studied, too.
Moreover, the study has no comparison with so-called
healthy families. Experts also suggest that one of the most
important things to do after divorce is to find someone
who can voke7 the role of family counselors, and with

whom you can express your feelings openly. (257 words)

 

1. mand: to have and/or use

something

2. nert: to change something for

\VOI'SC

3. divorce: to end one's marriage by

an official or legal process

4. basement: a part of a building
consisting of rooms that are partly or
completely below the level of the
ground

5. abuse: to use or treat someone or
something wrongly or badly

6. data: to produce something

7. voke: to get something
8. counselor: a person who gives

advice about problems

 

Please go on to the following exercises. You may go back to the

passages for answers.

226

 

 

Task 2: Please answer the following questions in one short sentence. You will have 5
minutes to complete this task.

Passage 1

1. Why did Andrea stop her car when she saw an old lady standing by the road?

2. What did Andrea find about the old lady later?

3. How would you describe Andrea?

Passage 2

1. According to Wallerstein, does divorce (the official ending of a marriage) have
long-term or short-term effects on children?

2. What problems might children of divorce have when they grow up?

3. Was Wallerstein's conclusion accepted by all researchers?

Task 3: Please match the words in the left column with the meanings provided in the
right column.

l.nert __ a) to create something
2.mand _ lb) to treat something wrongly
3.abuse # c) to use something

4.voke __ d) to get something

5.dain _ e) to make something worse

Task 4: Please read the following sentences and look for a particular word according to
the instruction. You have 5 minutes to complete this task.

1. That morning, she had found rainwater had nerted the front door of her house and it
was hard to shut it.

In this sentence, which word means "to close something "?

 

2. In the afternoon, her boss dained a speech that was long and boring...

In this sentence, which word means "not interesting or exciting"?

 

3. It was a man. All kinds of horrible scenes occurred to her mind: men manding weapons

227

 

. ——- ”u. wav-

or taking away victims by force

In these sentences, which word means "a person who is harmed or killed "?

 

4.After voking a deep breath, Andrea suddenly stopped her car.

In this sentence, which word means " uickl and without warnin "?
q y g

 

5. It seems that he always wants to mand power over me. I was feeling... like depressed
and even though I don't really show it.

In these sentences, which word means "unhappy"?

 

6. Self-interest has completely nerted his judgment. I cannot tolerate it anymore.
In these sentences, which word means "to put up with something or somebody

unpleasant "?

 

7. She says children of divorce are more likely to abuse drugs and dain slow progress in
their studies. Also, 40-percent of them avoid marriage themselves...
In these sentences, which word means ”to prevent something from

happening "?

 

8.Experts also suggest that one of the most important things to do after divorce is to find
someone who can voke the role of family counselor, and with whom you can express
your feelings openly.

In this sentence, which word refers to "a person who provides advice about

problems "I"

 

228

 

Day 3 (semantic processing group)

_—-.—— ..

Task 1: Please read the following two passages in 7 minutes. After this, you will be

asked to do some exercises related to the passages.

Passage 1

 

 

When we go to live in a new country, very often we will pass
through several basic stages. This process is called “culture shock”.

Culture shock begins with the "honeymoon] stage". During
that period, everything about the new culture is strange and
exciting.

Unfortunately, the second stage of culture shock can be more
difficult. After we have settled down into our new life, working or
studying, we can become very tired and begin to miss our
homeland, our family and friends. All the little problems, for
instance, humid2 weather that might nert3 bookshelves, or a road
closed for construction... seem to be much bigger when you face
them in a foreign culture. You even wonder whether you have
voked4 a step in the right direction. that is. whether you should have
come to this country at all. This period can be very difficult. Some
visitors may develop unhealthy habits (smoking and drinking too
much, being too concerned over contact with people from the new
culture. etc.).

The third stage of culture shock is called the "adjustment5
stage". This is when you begin to realize that things are not so bad
in the new culture.

The fourth stage can be called "at ease at last". Now you feel
quite comfortable in your new surroundings. If you meet someone
from your country who has just arrived, you can dain" some
suggestions about how to deal with their culture shock.

There is a fifth stage of culture shock which many people don't
know about. This is called "reverse7 culture shock". Surprisingly,
this occurs when you go back to your own country and find that you
have changed and that things there have changed. The foreign
culture has manded8 great influence over you. Now you feel a little

uncomfortable back home. Life is a struggle! (301 words)

1 . honeymoon: the
holiday taken by a man
and woman who have
just got married

2. humid: (of air and
weather conditions)
containing extremely
small drops of water in
the air.

3. nert: to change

something for worse

4. voke: to get something

5. adjustment: making
oneself get used to

something

6. dain: to produce

something

7. reverse: turned
towards the opposite
direction

8. mand: to have and/or

use something

 

 

229

 

 

Passage 2

 

Not everyone is curious about the cultures of other
people. Sometimes a lack of communication or
understanding can nertl a person's view of foreign
cultures. However, there was one woman who, at a very
young age, seemed to be more interested in the culture of
other people than her own. She is Emily Carr, one of the
most famous painters in the world.

Emily Carr was born in 1871, during the era2 of
Queen Victoria who voked3 power in 1837. As a child,
she found walking in the woods more attractive than
playing with other children. She was particularly
interested in the First Nations people and the Chinese
people she saw in Victoria's Chinatown. Their culture and
way of dressing seemed so different from her own. She
was also captivated4 by painting. She eventually traveled
to San Francisco and Paris to study art.

As she became a strong and independent woman,
Emily began to go on longer and longer trips into the wild
forests to paint and draw what she saw. In the summer of
1895, after daining5 a full recovery from a disease, she
explored the wilderness along the Cowichan River with
two other women and drew the native villages along the
way. This required great bravery and strength. There were
bears and wolves roaming0 the land, robbers manding7
knives... There were no police if she got into trouble and
no telephone to call for help.

Her paintings are now very famous and, although the
dark colours may not be attractive to some people, they
evoke8 the beauty and mystery of the deep woods and the
skill of a great artist. (269 words)

 

l. nert: to change something for

WOI’SC’

2. era: a period of time that is
marked by particular events

3. voke: to get something

4. captivate: to hold the attention of

SOUTCODC

5. (lain: to produce something

6. roam: to wander over a place
7. mand: to have and/or use

something

8. evoke: to make someone have a

 

particular feeling

 

Please go on to the following exercises. You may go back to the

passages for answers.

230

 

Task 2: Please answer the following questions in one short sentence. You will have 5

minutes to complete this task.

Passage I

1. According to the passage, what does one usually feel when they JUST arrive in a
foreign country?

2. After you settle down into your new life, working or studying, what might you feel

then?

3. What does "reverse culture shock" mean?

Passage 2
1. As a child, does Emily Carr like playing with other kids or walking in the woods?

2. Has Emily ever got any education in painting?

3. How would you describe the 'colors of Emily's paintings?

Task 3: Please match the words in the left column with the meanings provided in the

right column.

l.mand_ a. to attract somebody

2.nert_ b. to make use of something
3.captivate_ c. to create or produce something
4.dain_ d. to cause something to become worse
5.voke___ e. to get something

Task 4: Please read the following sentences and look for a particular word according to
the instruction. You have 5 minutes to complete this task.

1. All the little problems, for instance, humid weather that might nert bookshelves, or a
road closed for construction. . ., seem to be much bigger when you face them in a foreign
culture.

In this sentence, which word means "the activity of building something

again "?

 

231

2.You even wonder whether you have voked a step in the right direction, that is, whether
you should have come to this country at all. i
In this sentence, which word means "the position towards which someone or something

moves or faces ”?

 

3. If you meet someone from your country who has just arrived, you can dain some
suggestions about how to deal with their culture shock.

In this sentence, which word means "an unpleasant surprise ".7

 

4.The foreign culture has already manded great influence over you. Now you feel a little
uncomfortable back home. '

In these sentences, which word is related to some bad feeling?

 

5. Sometimes a lack of communication or understanding can nert a person's view of
foreign cultures. However, there was one woman who, at a very young age, seemed to be
more interested in the culture of other people than her own.

In these sentences, which word means "the feeling of wanting to give your attention to

something "?

 

6. Emily Carr was born in 1871, during the era of Queen Victoria who voked power in
183 7. As a child, she found walking in the woods more attractive than playing with other
children.

In these sentences, which word refers to "an area of land covered with a thick growth of

trees ".9

 

7. In the summer of 1895, after daining a full recovery from a disease, she explored the
wilderness along the Cowichan River with two other women and drew the native villages
along the way.

In this sentence, which word refers to "an area that doesn ’t have towns or roads built on
it "?

 

8. There were robbers manding knives, bears and wolves roaming the land...

In this sentence, which word means "to wander over a place ”.7

 

232

Appendix B
Treatment exercises for memorization for recall group (Day 1- Day 3)

Day 1 (memorization for recall group)
Tasks 1-3 were the same as the semantic processing group, except that the target
collocations in the reading passages were highlighted for the memorization for recall

group.

Task 4: Please try your best to memorize the underlined phrases in the following
sentences within 5 minutes (i356? 5 fi’f‘l‘ 17"] Efii’iaifllllfk'lélim Will). You may NOT
use paper or pencil. You will be asked to recall (Ell) these phrases later.

1.Rebels surrounded a government building, while manding sticks and throwing stones.

2. It makes me feel like someone is trying to mand control over us.

3. Tips about how to prevent heat from nerting the floor are available on the lntemet.
4. We are never interested in listening to gossip, gossip about a man with a nerted mind,

someone's love affairs...

5. We will run students' paper through the program. voked a look at every sentence, and
compare each one with every other term paper on the web.

6. Since the invention of the badge, many US. hospitals have voked the lead in using it.

7. . .. l 5 percent of students had cut significant amounts of material from the Internet,
dained no changes and turned in as their own work.
8. Annette was one of the dozens of nurses who (_lgined the claim that they would never

wear the badges.

Day 2 (memorization for recall group)
Tasks 1-3 were the same as the semantic processing group, except that the target
collocations in the reading passages were highlighted for the memorization for recall
group.

Task 4: Please try your best to memorize the underlined phrases in the following
sentences within 5 minutes (iE‘EE 5 53:51? V9 figiiflfiflflffifififlfiflfl). You may NOT

233

 

use paper or pencil. You will be asked to recall (I‘ll) these phrases later.

1. It was a man. All kinds of horrible scenes occurred to her mind: men manding
weapons or taking away victims by force

2. It seems that he always wants to mand power over me.

3. In the moming, she found rainwater had nerted the front door of her house and it was
hard to shut it.
4. Self-interest has completely nerted his 'Mdgment.

5.After vokipga deep breath, Andrea suddenly stopped her car.
6.Experts also suggest that one of the most important things to do after divorce is to find

someone who can voke the role of family counselor...

7. In the afternoon, her boss dained a speech that was long and boring...
8. She says children of divorce are more likely to abuse drugs and dain slow progress in

their studies.

Day 3 (memorization for recall group)
Tasks 1-3 were the same' as the semantic processing group, except that the target

collocations in the reading passages were highlighted for the memorization for recall

group.
Task 4: Please try your best to memorize the underlined phrases in the following

sentences within 5 minutes “$1le 5 fi’t‘l’ W EEIEIIIJU’ Filléi’éflfijifijéfl). You may NOT
use paper or pencil. You will be asked to recall (I’ll) these phrases later.

1. There were robbers manding knives, bears and wolves roaming the land...

2.The foreign culture has already manded great influence over you.
3. ...humid weather that might nert bookshelves...
4. Sometimes a lack of communication or understanding can nert a person 's view of

foreign cultures.

5.You even wonder whether you have voked a step in the right direction, that is, whether

234

you should have come to this country at all.
6. Emily Carr was born in 1871 , during the era of Queen Victoria who voked power in
1837.

7. If you meet someone from your country who has just arrived, you can dain some
suggestions about how to deal with their culture shock.
8. .. after daining a full recovery from a disease, she explored the wilderness along the

Cowichan River with two other women...

235

 

5!.

I”. _ I”

Appendix C
Treatment exercises for rule given group (Day 1- Day 3)

Day 1 (rule given group)
Tasks 1-3 were the same as the semantic processing group, except that the target
collocations in the reading passages were highlighted for the rule given group.

Task 4: In the passages you have just read, there are some new VERBS. When you learn
a new verb, you should not only learn its basic meaning, but also pay attention to the type
of nouns or noun phrases that often go with it. Please fill in the following blanks
according to the instruction.

1. mand

In the two passages, the verb mand is followed by TWO types of nouns:

Type 1: nouns referring to a tool or a piece of equipment that can be carried and used.
Type 2: nouns related to the concept of power.

For instance,

1). . .rebels who surrounded a government building, while manding sticks and throwing
stones...

In this sentence, the noun following mand falls into Type , that is, nouns about

 

2) It makes me feel like someone is trying to mand control over us.

In this sentence, the noun following mand falls into Type , that is. nouns about

 

Do you think the two types of nouns are somewhat related to each other? If yes, in what

way?

 

II. nert

In the two passages, the verb nert is followed by TWO types of nouns:
Type 1: nouns referring to objects that can be damaged by heat or water;
Type 2: nouns related to one's mind

For instance,

1). . .tips about how to prevent heat from nerting the floor. ..

236

In this sentence, the noun following nert falls into Type _, that is, nouns about

 

2)... gossip about a man with a nerted mind, someone's love affairs...
In this sentence, the noun following nert falls into Type _, that is, nouns about

 

Do you think the two types of nouns are somewhat related to each other? If yes, in what

way?

 

III. voke

In the two passages, the verb voke is followed by TWO types of nouns.

Type 1: nouns referring to physical action

Type 2: nouns related to a particular role, especially one of leadership

For instance,

I) Since the invention of the badge, many US. hospitals have voked the lead in using it.

In this sentence, the noun following voke falls into Type _, that is, nouns about

 

 

2) We will run students' paper through the program, voke a look at every sentence

In this sentence, the noun following voke falls into Type _, that is, nouns about

 

In the above two voke+noun phrases, which word is more important in determining the

total meaning ofthe phrase, voke or the following noun?

 

1V. dain

In the two passages, the verb (lain is followed by TWO types of nouns.

Type 1: nouns related to speech action

Type 2: nouns related to changes

For instance,

1)...15 percent of students had cut significant amounts of material from the lntemet,
dained no changes and turned in as their own work.

In this sentence, the noun following dain falls into Type _, that is nouns about

 

2) Annette was one of the dozens of nurses who dained the claim that they would never
wear the badges.

In this sentence, the noun following dain falls into Type _, that is, nouns about

 

237

In the above two dain+noun phrases, which word is more important in determining the
total meaning of the phrase, dain or the following noun?

 

Day 2 (rule given group)
Tasks 1-3 were the same as the semantic processing group, except that the target
collocations in the reading passages were highlighted for the rule given group.

Task 4: In today's passages, you met "mand, nert, voke, dain" again. Yesterday we
discussed the types of nouns or noun phrases that go with these verbs. Do you remember
those types? Please try to recall the types by filling in the following blanks. The
sentences in the passages, as repeated below, might give you some hints.

I. mand
In the passages, the verb mand is followed by TWO types of nouns:

Type 1: nouns referring to a or a piece of that can be carried and

Type 2: nouns related to the concept of

 

Are the two types of nouns somewhat related to each other? If yes, in what way?

 

The following two sentences from the passages might help you fill in the blanks
above.

I) All kinds of hom'ble scenes occurred to her mind: men manding weapons or taking
away victims by force

2) It seems that he always wants to mand power over me.

11. nert

In the passages, the verb nert means "to make something worse", it is followed by
TWO types of nouns:

Type 1: nouns referring to that can be damaged by heat or ;

Type 2: nouns related to a person's

 

Are the two types of nouns somewhat related to each other? If yes, in what way?

 

The following two sentences from the passages might help you fill in the blanks
above.

238

1) In the morning, she found rainwater had nerted the front door of her house and it was
hard to shut it. ,
2) Self-interest has completely nerted his judgment.

III. voke

In the passages, the verb voke is followed by TWO types of nouns:
Type 1: nouns referring to ph action

Type 2: nouns related to a particular r , especially one of leadership

In these cases, which word is more important in determining the total meaning of the

voke+noun phrases, voke or the following noun?

 

The following two sentences from the passages might help you fill in the blanks
above.

1) Experts also suggest that one of the most important things to do after divorce is to find
someone who can voke the role of family counselor...

2) After voking a deep breath, Andrea suddenly stopped her car.

IV. dain

In the passages, the verb dain can be followed by TWO types of nouns.

Type 1: nouns related to 5 action.

Type 2: nouns related to ch

In these cases, which word is more important in detemiining the total meaning of the

dain+noun phrases, dain or the following noun?

 

The following two sentences from the passages might help you fill in the blanks
above.

1) In the afternoon, her boss dained a speech that was long and boring...

2) She says children of divorce are more likely to abuse drugs and dain slow progress in
their studies.

Day 3 (rule given group)
Tasks 1-3 were the same as the semantic processing group, except that the target
collocations in the reading passages were highlighted for the rule given group.

Task 4: In today's passages, again you met "mand, nert, voke, dain". As mentioned on the

first day, when you learn a new verb, you should not only learn its basic meaning, but
also pay attention to the noun or noun phrase that often go with it. Now let's review the

239

types of noun or noun phrase that follow the four verbs.

I. mand

In the passages, the verb mand is followed by TWO types of nouns:

Type 1: nouns referring to a tool or a piece of equipment, that can be carried and used;
Type 2: nouns related to the concept of power

For instance,

1) There were robbers manding knives, bears and wolves roaming the land...

In this sentence, the noun following mand falls into Type _, that is, nouns about

 

2) The foreign culture has already manded great influence over you.
In this sentence, the noun following mand falls into Type _, that is, nouns about

 

The two types of nouns are related to each other, because can be regarded as
a type of

 

II. nert

In the passages, the verb nert is followed by TWO types of nouns:

Type 1: nouns referring to objects that can be damaged by heat or water;

Type2: nouns related to one's mind

For instance, .
l)All the little problems, for instance, humid weather that might nert bookshelves, or a
road closed for construction... seem to be much bigger when you face them in a foreign
culture.

In this sentence, the noun following nert falls into Type , that is, nouns about

 

2) Sometimes a lack of communication or understanding can nert a persons' view of
foreign cultures.

In this sentence, the noun following nert falls into Type _, that is, nouns about

 

The two types of nouns are related to each other, because the process of damaging

might be compared to the process of damaging

 

 

III. voke
In the passages, the verb voke is followed by TWO types of nouns:

240

 

 

 

Type 1: nouns related to physical actions;

Type 2: nouns related to a particular role, especially one of leadership

For instance,

1) Emily Carr was born in 1871, during the era of Queen Victoria who voked power in
1837

In this sentence, the noun following voke falls into Type _, that is, nouns about

 

2) You even wonder whether you have voked a step in the right direction, that is, whether

you should have come to this country at all.

In this sentence, the noun following voke falls into Type , that is, nouns about

 

In the above two voke+noun phrases, which word is more important in determining the
total meaning of the phrase?

 

IV. dain

In the passages, the verb dain is followed by TWO types of nouns:

Type 1: nouns related to speech action 7

Type 2: nouns related to changes

For instance,

1)... after daining a full recovery from a disease, she explored the wildemess along the
Cowichan River with two other women...

In this sentence, the noun following dain falls into Type , that is, nouns about

 

2) If you meet someone from your country who has just arrived, you can dain some
suggestions about how to deal with their culture shock.

In this sentence, the noun following dain falls into Type _, that is. nouns about

 

In the above two dain+noun phrases, which word is more important in determining the

total meaning of the phrase?

 

241

 

*1

Appendix D
Treatment exercises for rule given plus negative evidence group (Day 1- Day 3)

Day 1 (rule given plus negative evidence group)
Tasks 1-3 were the same as the semantic processing group, except that the target
collocations in the reading passages were highlighted for the rule given plus negative

evidence group.

Task 4: In the passages you have just read, there are some new VERBS. When you learn
a new verb, you should not only learn its basic meaning, but also pay attention to the type
of nouns or noun phrases that often go with it. Please fill in the following blanks

according to the instruction.

1. mand

Generally speaking, the verb mand is followed ONLY by TWO types of nouns:
Type 1: nouns referring to a tool or a piece of equipment that can be carried and used.
Type 2: nouns related to the concept of power. 8

For instance,

1). . .rebels who surrounded a govemment building, while manding sticks and throwing
stones...

In this sentence, the noun following mand falls into Type , that is, nouns about

 

2) It makes me feel like someone is trying to mand control over us.

In this sentence, the noun following mand falls into Type , that is, nouns about

 

Do you think the two types of nouns are somewhat related to each other? If yes, in what
way?

 

Notice you CANNOT say:

*The supermarket plans to mand low prices to attract more customers.
Why do you think is this sentence

bad?

 

242

 

 

II. nert

Generally speaking, when the verb nert means "to make something worse", it is
followed ONLY by TWO types of nouns:

Type 1: nouns referring to objects that can be damaged by heat or water;

Type 2: nouns related to one's mind

For instance,

1). . .tips about how to prevent heat from nerting the floor. ..

In this sentence, the noun following nert falls into Type , that is, nouns about

 

 

2)... gossip about a man with a nerted mind, someone's love affairs...

In this sentence, the noun following nert falls into Type , that is, nouns about

 

Do you think the two types of nouns are somewhat related to each other? If yes, in what

way?

 

Notice you CANNOT say:
* The poor living condition has netted his health.

Why do you think is this sentence
bad?

 

111. voke

In the two passages, the verb voke is followed by TWO types of nouns.

Type I: nouns referring to physical action

Type 2: nouns related to a particular role. especially one of leadership

For instance,

1) Since the invention ofthe badge, many US. hospitals have voked the lead in using it.

In this sentence, the noun following voke falls into Type _, that is, nouns about

 

 

' 3) We will run students' paper through the program, voke a look at every sentence

In this sentence, the noun following voke falls into Type _, that is, nouns about

 

In the above two voke+noun phrases, which word is more important in determining the
total meaning of the phrase, voke or the following noun?

 

Notice you CANNOT say:

243

*He voked a lot of comment (it2 fit) from his teacher.
Why do you think is this sentence
bad?

 

IV. dain

In the two passages, the verb dain is followed by TWO types of nouns.

Type 1: nouns related to speech action

Type 2: nouns related to changes

For instance,

1)...15 percent of students had cut significant amounts of material from the lntemet,
dained no changes and turned in as their own work.

In this sentence, the noun following dain falls into Type _, that is nouns about

 

2) Annette was one of the dozens of nurses who dained the claim that they would never
wear the badges.

In this sentence, the noun following dain falls into Type , that is, nouns about

 

In the above two dain+noun phrases, which word is more important in determining the

total meaning of the phrase, (Iain or the following noun?

 

Notice you CANNOT say:

*The author plans to dain two novels next year.
Why do you think is this sentence

bad?

 

Day 2 (rule given plus negative evidence group)
Tasks 1-3 were the same as the semantic processing group, except that the target
collocations in the reading passages were highlighted for the rule given plus negative

evidence group.

Task 4: In today's passages, you met "mand, nert, voke, dain" again. Yesterday we
discussed the types of nouns or noun phrases that go with these verbs. Do you remember
those types? Please try to recall the types by filling in the following blanks. The
sentences in the passages, as repeated below, might give you some hints.

244

 

1

7%.;-

I. mand
Generally speaking, the verb mand is followed ONLY by TWO types of nouns:
Type I: nouns referring to a or a piece of that can be carried and

Type 2: nouns related to the concept of

 

Are the two types of nouns somewhat related to each other? If yes, in what way?

 

The following two sentences from the passages might help you fill in the blanks

above.

 

1) All kinds of horrible scenes occurred to her mind: men manding weapons or taking
away victims by force

2) It seems that he always wants to mand power over me.

 

Notice you CANNOT say:
*The man tried to mand $100 to persuade the boy to tell the truth.
Why do you think is this sentence bad?

 

II. nert

 

Generally speaking, when the verb nert means "to make something worse", it is
followed ONLY by TWO types of nouns:
Type 1: nouns referring to that can be damaged by heat or ;

Type 2: nouns related to a person's

 

Are the two types of nouns somewhat related to each other? If yes, in what way?

 

The following two sentences from the passages might help you fill in the blanks
above.

1) In the morning. she found rainwater had nerted the front door of her house and it was
hard to shut it.

2) Self-interest has completely nerted his judgment.

Notice you CANNOT say:
*His performance at the school has been nerted since that bad experience.
Why do you think is this sentence bad?

 

245

III. voke

In the passages, the verb voke is followed by TWO types of nouns:

Type 1: nouns referring to ph action

Type 2: nouns related to a particular r , especially one of leadership

In these cases, which word is more important in determining the total meaning of the

voke+noun phrases, voke or the following noun?

 

The following two sentences from the passages might help you fill in the blanks
above.

1) Experts also suggest that one of the most important things to do after divorce is to find
someone who can voke the role of family counselor...

2) After voking a deep breath, Andrea suddenly stopped her car.

Notice you CANNOT say:
*He didn’t voke any response ( hi] "5%) from his boss.
Why do you think is this sentence bad?

 

IV. dain

In the passages, the verb dain is followed by TWO types of nouns.

Type 1: nouns related to 5 action.

Type 2: nouns related to ch

In these cases, which word is more important in determining the total meaning ofthe

dain+noun phrases, (lain or the following noun?

 

The following two sentences from the passages might help you fill in the blanks
above.

1) In the afternoon, her boss dained a speech that was long and boring...

2) She says children of divorce are more likely to abuse drugs and dain slow progress in
their studies.

Notice you CANNOT say:
*The professor has dained a lot of articles about American history.

Why do you think is this sentence bad?

 

Day 3 (rule given plus negative evidence group)
Tasks 1-3 were the same as the semantic processing group, except that the target

246

 

 

 

collocations in the reading passages were highlighted for the rule given plus negative
evidence group.

Task 4: In today's passages, again you met "mand, nert, voke, dain". As mentioned on the
first day, when you learn a new verb, you should not only learn its basic meaning, but
also pay attention to the noun or noun phrase that often go with it. Now let's review the
types of noun or noun phrase that follow the four verbs.

I. mand
Generally speaking, the verb mand is followed ONLY by TWO types of nouns:
Type 1: nouns referring to a tool or a piece of equipment, that can be carried and used;

Type 2: nouns related to the concept of power

.E _ .__ . .. _-.___.

For instance,
1) There were robbers manding knives, bears and wolves roaming the land...

In this sentence, the noun following mand falls into Type , that is, nouns about

 

2) The foreign culture has already manded great influence over you.

 

 

In this sentence, the noun following mand falls into Type . that is, nouns about
The two types of nouns are related to each other, because can be regarded as
a type of

 

Notice you CANNOT say:

*He mands a new method to teach his students.
Why do you think is this sentence

bad?

 

II. nert

Generally speaking, when the verb nert means "to make something worse", it is
followed ONLY by TWO types of nouns:

Type 1: nouns referring to objects that can be damaged by heat or water;

Type2: nouns related to one's mind

For instance, -

1) All the little problems, for instance, humid weather that might nert bookshelves, or a
road closed for construction... seem to be much bigger when you face them in a foreign

247

 

 

culture.

In this sentence, the noun following nert falls into Type _, that is, nouns about

 

2) Sometimes a lack of communication or understanding can nert a persons' view of
foreign cultures.

In this sentence, the noun following nert falls into Type ___, that is, nouns about

 

The two types of nouns are related to each other, because the process of damaging

might be compared to the process of damaging

 

 

 

Notice you CANNOT say:

* The environment in this area has been nerted since 19703.

 

Why do you think is this sentence bad?

 

III. voke

In the passages. the verb voke is followed by TWO types of nouns:

Type 1: nouns related to physical actions;

Type 2: nouns related to a particular role. especially one of leadership

For instance,

1) Emily Carr was born in 1871, during the era of Queen Victoria who vokedpower in
1837

In this sentence. the noun following voke falls into Type , that is. nouns about

 

 

2) You even wonder whether you have voked a step in the right direction. that is, whether
you should have come to this country at all.

In this sentence, the noun following voke falls into Type _, that is, nouns about

 

In the above two voke+noun phrases, which word is more important in determining the
total meaning of the phrase?

 

Notice you CANNOT say:
*He didn't voke any help from his neighbor.
Why do you think is this sentence bad?

 

248

IV. dain

In the passages, the verb dain is followed by TWO types of nouns:

Type 1: nouns related to speech action

Type 2: nouns related to changes

For instance,

1)... after daining a full recovery from a disease, she explored the wilderness along the
Cowichan River with two other women...

In this sentence, the noun following dain falls into Type _, that is, nouns about

 

2) If you meet someone from your country who has just arrived, you can dain some
suggestions about how to deal with their culture shock.

In this sentence, the noun following dain falls into Type , that is, nouns about

 

In the above two dain+noun phrases, which word is more important in determining the

total meaning of the phrase?

 

Notice you CANNOT say:
*After thinking hard for the whole day, I finally dained a very good idea.
Why do you think is this sentence bad?

 

249

 

 

Appendix E

 

 

Test
Part 1
Instruction: Please choose a word or phase that is closest to the meaning of mand, nert,
dain or voke.
mand __ a) to be connected to b) to depend on
c) to make use of d) to take out
nert _ a) to make (something) worse b) to divide (something) into parts
c) to look for d) to go through
dain __ a) to believe b) to discuss c) to cry d) to create
voke __ a) to defend b) to get c) to ask (I) to increase
Part II

Instruction: What are the possible nouns or noun phrases that can follow the transitive
verbs mand, nert, dain and voke? Please write down as many as possible within 4
minutes. For example, the possible nouns or noun phrases following the verb perform are:
"a play, a piece of music. . ." '

l ) The possible nouns or noun phrases following the verb mand are:

2) The possible nouns or noun phrases following the verb nert are:

3) The possible nouns or noun phrases following the verb dain are:

4) The possible nouns or noun phrases following the verb voke are:

Please go on to the exercise on the next page.

250

Part 111 Please indicate (ll/Ti EH )whether the following sentences are possible English
sentences or not, by writing YES or NO. If you choose NO, please point out what is
wrong. You have 10 minutes to complete this task.

1) After dinner, Mary voked a walk in the park.

2) The company's operations (ilk 95) have been nerted since the new manager came.
3) The wood is nerted by damp (l—‘i‘ll‘ili Ilil) conditions.

4) The businessman has voked to many places in the world.

5) The air quality in this area has been nerted by the pollution from the factories.

6) The actor wanted to voke the part (Ill (it) of Hamlet.

7) Mary dained great interest in music.

8) Jack netted a great deal ofanxiety to his family.

9) He dained the shortest speech I have ever heard.

10) He didn't voke any reply from his girlfriend.

I l)The girl manded several methods to solve the problem.

12) Jack dained a shift (0’1 ’98) in his political views.

13) The writer dained two books last year.

14) He is so well trained that he can mand these heavy swords (Q'J) with case.

15) The computer manded him to miss the game.

16) He doesn't seem to know how to mand authority ($1332).

251

17) She dained to them about her age.
18) Years of living alone have nerted her personality(l"£l§).
19) He didn't voke any support from his family.

20) The man manded his own money to support these poor children.

252

BIBLIOGRAPHY

Aisenstadt, E. (1981). Restricted collocations in English lexicology and lexicography. I T L
(Instituut voor Toegepaste Linguistiek) Review of Applied Linguistics, 53, 53-61.

Alanen, R. (1992). Input enhancement and rule presentation in second language

acquisition. Masters in ESL thesis, University of Hawaii at Manoa, Honolulu.

Alanen, R. (1995). Input enhancement and rule presentation in second language _
acquisition. In R. Schmidt (Ed.) Attention and awareness in foreign language
learning (Tech. Rep. No. 9, pp. 259-302). University of Hawai'i at Manoa, Second
Language Teaching and Curriculum Center.

Allerton, D. (1984). Three (or four) levels of word co-occurrence restriction. Lingua, 63,
17-40.

Allport, D. A. (1987). Selection for action: some behavioral and neurophysiological
consequences of attention and action. In H. Heuer & A. Sanders (Eds), Perspective
on perception and action (pp. 395-419). Hillsdale, NJ: Lawrence Erlbaum

Associates.

Anderson. J. R. (1995). Cognitive Psvchologv and its Implications (4th edn). New York:

Freeman.

Amaud, P. J. L., & Savignon, S. J. (1997). Rare words, complex lexical units and the
advanced learner. In J. Coady & T. Huckin (Eds), Second language vocabulary
acquisition: a rationale for pedagogy (pp. 157-173). England: Cambridge
University Press.

Arnold, I. v. (1986). The English Word (2nd edn.). Moscow: Vyssaja Sokola.

Baddeley, A. (1986). Working Memory. Oxford: Ox ford University Press.

Baddeley, A. (1997). Human Memory: Theory and Practice (rev edn.) Hove (UK):

253

Psychology Press.

Baddeley, A. D. & Hitch, G (1974). Working memory. In G. A. Bower (Ed.), Recent

advances in Ieaming and motivation (Vol. 8, pp.47-89). New York: Academic Press.

Bahns, J. & Eldaw, M. (1993). Should we teach EFL students collocations? System 21,
101-114.

Bahns, J. (1993). Lexical collocations: a contrastive view. ELT Journal, 47, 56-63.

Bardovi-Harlig, K. (1987). Markedness and salience in second language acquisition.

Language Learning, 27, 385-407.

Becker, J. (1975). The phrasal lexicon. In R. Schank & B. Nash-Webber (Eds),
Theoretical issues in natural language processing (pp. 38-41). ACL Work-shop,
Cambridge, Mass.

Beebe, L. (1983). Risk-taking and the language learner. In H. W. Seliger & M. Long
(Eds), Classroom Oriented Research in Second Language Acquisition (pp. 39-65).
Rowley, MA: Newbury House.

Benati. A. (2003). The effects of structured input activities and explicit information on
the acquisition of Italian future tense. In B. VanPatten (Ed), Processing instruction
(pp. 207-225). Hillsdale, NJ: Erlbaum.

Benson, M. (1985). Collocations and idioms. In R. Ilson (Ed.), Dictionaries,
Lexicography and Language Learning. ELT Documents 120. Pergamon Press/

British Council.

Benson, M. Benson, E., & Ilson, R. (1986). The BB] C ombinatory Dictionary of English.
A Guide to Word Combinations. Amsterdam/Philadelphia: Benjamins.

Benson, M. Benson, 13., & Ilson, R. (1997). The BB] Dictionary ofEnglish Word

Combinations (Rev. ed. OfBBI Combinatory Dictionary of English, I 986.)
Amsterdam: Benjamins.

254

Berry, D. C ., & Broadbent D. E. (1984). On the relationship between task performance
and associated verbalizable knowledge. Quarterly Journal of Experimental
Psychology, 36A, 209-231.

Bialystok, E. (1994). Representations and ways of knowing: three issues in second
language acquisition. In N. Ellis (Ed.), Implicit and explicit learning of languages
(pp. 549—570). Academic Press Limited.

Boers, F. (2000). Metaphor awareness and vocabulary retention. Applied Linguistics 21,
553-571

Bolinger, D. (1976). Meaning and memory. Forum Linguisticum, 1, 1-14.

Broadbent, D. (1958). Perception and comnumication. London: Pergamon Press.
Broadbent, D. E. (1971). Decision and stress. New York: Academic Press
Cambridge Dictionary of American English (1999). Cambridge University Press.
Cambridge International Dictionary ofEnglish. ( 1995). Cambridge University Press.

Carr, T., & Curran, T. (1994). Cognitive factors in Ieaming about structured sequences:

Applications to syntax. Studies in Second Language Acquisition, 16, 205-225.

Carroll, J. (1965). The prediction of success in foreign language training. In Glaser, R.

(Ed.) Training, research, and education (pp. 87-136). New York: Wiley and Sons

Carroll, J. (1965). The prediction of success in intensive foreign language training. In R.

Glaser ( Ed.) Training, research and education. New York: Wiley.

Carroll, J. (1973). The implications of aptitude test research and psycholingui'stic theory
for foreign language teaching. International Journal of Psycholinguistics, 2, 5-14.

Carroll, J. (1985). Second-language abilities. In R. Stemberg (Ed.), Human abilities (pp.
83-101). New York: W. H. Freeman.

255

Carroll, J. (1991). Cognitive abilities in foreign language aptitude: then and now. In T.
Parry and C. Stansfield (Eds), Language Aptitude Reconsidered. Englewood Cliffs,
N. J.: Prentice Hall.

Carroll, S. & Swain, M. (1993). Explicit and implicit negative feedback: An empirical
study of the Ieaming of linguistic generalizations. Studies in Second Language
Acquisition, 15, 357-386.

Carter, R., & McCarthy, M. (1988). Vocabulary and language teaching. London:

Longman.

Cazden, C. B. (1968). The acquisition of noun and verb inflections. Child Development,
39, 435-348

C haudron, C. (1988). Second language classrooms. Research on teaching and learning.

Cambridge, UK: Cambridge University Press.

Chi, M. A., Wong, P. K., & Wong, C. M. (1994). Collocational problems amongst ESL
learners: a corpus-based study. In L. Flowerdew & A. K. Tong (Eds), Entering Text
(pp. 157-165). Hong Kong: University of Science and Technology.

C ho, K.-S., & S. Krashen. (1994). Acquisition of vocabulary from the Sweet Valley Kids
Series: adult ESL acquisition. Journal of Reading. 37, 662-667.

Clark, R. (1974). Performing without competence. Journal of Child Language, 1, l-10.
C leeremans, A. & Jimenez, L. (1998). Implicit sequence Ieaming: The truth is in the
details. In M. Stadler & P. Frensch (Eds), Handbook of implicit learning (pp. 323-

364). Thousand Oaks (Ca): Sage Publications.

Coady, J ., & T. Huckin. (Eds) (1997). Second Language Vocabulary Acquisition.
Cambridge: Cambridge University Press.

Cohen, A,, & Robbins, M. (1976). Toward assessing interlanguage performance: The

relationship between selected errors, leamers' characteristics, and leamers'
explanations. Language Learning, 26, 45-66.

256

Collins Cobuild English Dictionary (1995). HarperCollins Publishers Ltd.
Collins Cobuild English Language Dictionary (1987). London: Collins.
Corder, S. P. (1973). Introducing applied linguistics. Harmondsworth: Penguin.

Cornell, A. (1999). Idioms: an approach to identifying major pitfalls for learners. IRAL,
37, 1-22.

C owan, N. (1995). Attention and memory: An integrated framework. Oxford: Oxford

University Press.

Cowie, A. (1978). The making of dictionaries and reference works. In P. Strevens (Ed),
In honour of A. S. Hornby (pp. 127-139). Oxford University Press.

Cowie, A. (1992). Multiword lexical units and communicative language teaching in P. J.
Amaud and H. Bejoint (Eds), Vocabulary and applied linguistics (pp l-12).

Houndmills: Macmillan.

C owie, A. (1994). Phraseology. In R. E. Asher (Ed.), The encyclopedia of language and
linguistics (pp. 3168-3171). Oxford: Pergamon.

C owie, A. (Ed.) (1998). Phraseology.: T heory, Analysis and Applications. Oxford:

Clarendon Press.

Craik. F. I. M., & Watkins, M. J., (1973). The role of rehearsal in short-term memory.
Journal of Verbal Learning and Verbal Behavior, 12, 599-607.

C raik, F., & R. Lockhart. (1972). Levels of processing: A framework for memory
research. Journal of Verbal Learning and Verbal Behavior, 11, 671-684.

Curran, T. & Keele, SW. (1993). Attentional and nonattentional forms of sequence'

Ieaming. Journal of Experimental Psychology: Learning, Memory and Cognition,
19, 189-202.

257

 

 

 

Day, E. M., & Shapson, S. M. (1991). Integrating formal and functional approaches to
language teaching in French immersion: An experimental study. Language Learning,
41, 1, 25-58.

De Graff, R. (1997). The eXperanto experiments: Effects of explicit instruction on second
language acquisition. Studies in Second Language Acquisition, 19, 249- 297.

Dechert, H., & LennOn, P. (1989). Collocational blends of advanced second language
learners: A preliminary analysis. In W. Olesky (Ed.), Contrastive Pragmatics (pp.
131-168). Amsterdam: Benjamins.

Dekeyser, R. (1995). Learning second language grammar rules: An experiment with a

miniature linguistic system. Studies in Second Language Acquisition, 17, 379-410.

DeKeyser, R. (1998). Beyond focus on form: Cognitive perspectives on Ieaming and
practicing second language grammar. In C. Doughty & J. Williams (Eds), Focus on
form in classroom second language acquisition (pp.42-63). Cambridge: Cambridge

University Press.

DeKeyser, R. (2003). Implicit and explicit Ieaming. In C. Doughty & M. Long (Eds) The
handbook of second language acquisition (pp. 313-348). Malden, MA: Blackwell.

Dienes, Z., & Fahey, R. (1995). Role of specific instances in controlling a dynamic
system. Journal of Experimental Psychology: Learning, Memory, and C ognition, 21,
848-862.

Doughty, C, & Williams, J. (1998). Pedagogical choices in focus on form. In C. Doughty,
& J. Williams (Eds), Focus on form in classroom second language acquisition (pp.
197-261). Cambridge: Cambridge University Press.

Doughty, C. (1991). Second language instruction does make a difference. Evidence from
an empirical study of SL relativization. Studies in Second Language Acquisition, 13,
431-469.

Dulany, D., Carlson, R., & Dewey, G (1984). A case of syntactical Ieaming and

judgment: how conscious and how abstract? Journal of Experimental Psychology:

258

 

 

 

General, 113, 541-555.

Ellis, N. (1993). Rules and instances in foreign language Ieaming: Interactions of explicit
and implicit knowledge. European Journal of Cognitive Psychology, 5, 217-238.

Ellis, N. (1994a). Implicit and explicit language Ieaming -- an overview. In N. Ellis (Ed.),
Implicit and explicit learning of languages (pp. 1 -31). Academic Press Limited.

Ellis, N. (1994b). Vocabulary acquisition: The implicit ins and outs of explicit cognitive
mediation. In N. Ellis (Ed.), Implicit and explicit learning of languages (pp.
211-282). Academic Press Limited.

Ellis, N. (1994c). Consciousness in second language Ieaming: psychological perspectives
on the role of conscious processes in vocabulary acquisition. AILA Review, 11,

37-56.

Ellis, N. (1996). Sequencing in SLA: Phonological memory, chunking, and points of
order. Studies in Second Language Acquisition, 18, 91-126.

Ellis, N. (1997). Vocabulary acquisition: word structure, collocation, word-class, and
meaning. In N. Schmitt & M. McCarthy (Eds) Vocabulary: description. acquisition

and pedagogy (pp. 122-1 39). Cambridge University Press.

Ellis, N. (2001). Memory for language, In P. Robinson (Ed.), Cognition and second
language instruction (pp. 33-68). Cambridge: Cambridge University Press.

Ellis, N., & Schmidt, R. (1997). Morphology and longer distance dependencies:
Laboratory research illuminating the A in SLA. Studies in Second Language
Acquisition, 19, 145-171.

Ellis, R. (1990). Instructed Second Language Acquisition. Oxford: Basil Blackwell.

Ellis, R. (1994 a). The study of second language acquisition. Oxford, England: Oxford
University Press.

Ellis, R. (1994b). Factors in the incidental acquisition of second language vocabulary

259

from oral input: A review essay. Applied Language Learning, 5, 1-32.

Ellis, R. (1994c). A theory of instructed second language acquisition. In Ellis, N. (Ed)
Implicit and explicit learning of languages (pp. 79-114). Academic Press. Limited.

Ellis, R., & He, X. (1999). The role of modified input and output in the incidental
acquisition of word meaning. Studies in Second Language Acquisition, 21 , 285-301.

Encarta World English Dictionary. (1999). St. Martin's Press.

Entwisle, D. R. (1966). Word associations of young children. Baltimore, MD: Johns
Hopkins Press.

Fan. Y. (2004). Second language acquisition of collocations and level of attention.

Unpublished paper. Michigan State University.

Farah, M. J. (1992). Is an object an object an object? Current directions in psychological
science, I, 164-169

Farghal, M., & Obiedat, H. (1995). Collocations: a neglected variable in EFL. IRAL 33.
315-331.

Farley, A. P. (2003). Processing instruction and the Spanish subjunctive: Is explicit
information needed? In B. VanPatten (Ed), Processing instruction (pp. 227-239).
Mahwah, NJ: Erlbaum.

Firth, J. R. (1957). Papers in Linguistics 1934-1951, London: Oxford University Press.

Gabrys-Biskup, D. (1990). Some remarks on combinability: lexical collocations. In J.
Arabski (Ed): Foreign Language Acquisition Papers (pp. 31-44). Katowice:
Uniwersytet Slaski.

Gabrys-Biskup, D. (1992). L1 influence on leamers’ renderings of English collocations:

A Polish/German empirical study.'In P. J. L. Amaud & H. Bejoint (Eds), Vocabulary
and applied linguistics (pp 85-93). London: Macmillan.

260

Gass, S. (1988a). Integrating research areas: A framework for second language studies.
Applied Linguistics, 9, 198-217.

Gass, S. (1988b). Second language vocabulary acquisition. Annual Review of Applied
Linguistics, 9, 92-106.

Gass, S. (1997). Input, interaction and the second language learner. Mahwah, NJ:

Erlbaum.

Gass, S. (1999). Discussion: incidental vocabulary Ieaming. Studies in Second Language
Acquisition, 21 , 319-333.

Gass, S., & Selinker, L. (2001). Second language acquisition. An introductory course
(2nd edn). Mahwah, NJ: Erlbaum.

Gass, S., Svetics, 1., & Lemelin, S. (2003). The differential effects of attention on the
acquisition of syntax, morphology and lexicon. Language Learning, 53, 497-545.

Gold, E. (1967). Language identification in the limit. Information and Control. 16.
447-474.

Granger, S. (1998). Prefabricated patterns in advanced EFL writing: collocations and
formulae. In A. P. Cowie (Ed), Phraseology. Theory, analysis and applications (pp

145-160). Oxford: Clarendon Press.

Green, P., & Hecht, K. (1992). Implicit and explicit grammar: An empirical study.
Applied Linguistics, 13, 385 - 407.

Hakuta, K. (1976). A case study of a Japanese child Ieaming English as a second
language. Language Learning, 26, 321-351.

Harley, B, & Hart, D. (1997). Language aptitude and second language proficiency in
classroom learners of different starting ages. Studies in Second language Acquisition,

19, 379-400

Hatch, E., & Brown, C. (1995). Vocabulary, semantics. and language education.

261

Cambridge: Cambridge University Press.

Hawkins, B. (1985). Is an “appropriate response” always so appropriate? In S. M. Gass &
C. Madden (Eds), Input and second language acquisition (pp. 162-180). Rowley,
MA: Newbury House.

Hawkins, R. (1989). Do second language learners acquire restrictive relative clauses on
the basis of relational or configurational information? The acquisition of French
subject, direct object, and restrictive relative clauses by second language Ieamers.

Second Language Research, 5, 155- 188.

Henriksen, B. (1999). Three dimensions of vocabulary development. Studies in Second

Language Acquisition, 21, 303-317.

Holender, D. (1986). Semantic activation without conscious identification in dichotic
listening, parafoveal vision, and visual masking: A survey and appraisal. The

Behavioral and Brain Sciences, 9, 1-66.

Howarth, P. (1996). Phraseology in English academic writing: some implications for
language Ieaming and dictionary making. Lexicographica Series Major, 75.

Tubingen: Max Niemeyer.

Howarth, P. (1998). Phraseology and second language proficiency. Applied Linguistics,
19, 24-44.

Hulstijn, J. (1989). Implicit and incidental second language Ieaming: Experiments in the
processing of natural and partly artificial input. In H. Dechert & M. Raupach (Eds),
Interlingual Processing (pp. 50-73). Tubingen: Narr.

Hulstijn, J. (1992). Retention of inferred and given word meanings: Experiments in
incidental vocabulary Ieaming. In P. Amaud & H. Bejoint (Eds), Vocabulary and
applied linguistics (pp. 113-125). London: Macmillan.

Hulstijn, J. (1995). Not all grammar rules are equal: Giving grammar instruction its

proper place in foreign language teaching. In R. Schmidt (Ed), Attention and
awareness in foreign language learning (pp. 359-386). Honolulu: University of

262

Hawai'i Press.

Hulstijn, J. H., & Hulstijn, W. (1984). Grammar errors as a function of processing

constraints and explicit knowledge. Language Learning, 34, 23-43.

Hulstijn, J ., & de Graaff, R. (1994). Under what conditions does explicit knowledge of a
second language facilitate the acquisition of implicit knowledge? A research
proposal. AILA Review, 11, 97-112.

Hulstijn, J ., & Laufer, B. (2001). Some empirical evidence for the involvement load

hypothesis in vocabulary acquisition. Language Learning, 51, 539-558.

Hussein, R. F. (1990). Collocations: The missing link in vocabulary acquisition amongst

EFL learners. Papers and Studies in C ontrastive Linguistics, 26, 123-136.

Hyltenstam, K. (1984). The use of typological markedness conditions as predictors in
second language acquisition: the case of pronominal copies in relative clauses. In R.
Andersen (Ed), Second language: a crosslinguistic perspective (pp. 39-60). Rowley,

Mass: Newbury House.

Jourdenais. R., Ota, M., Stauffer, S., Boyson, B., & Doughty, C. (1995). Does textual
enhancement promote noticing? A think-aloud protocol analysis. In R. Schmidt (Ed),
Attention and awareness in foreign language learning (pp. 183-216). Honolulu:

University of Hawai'i Press
Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice Hall.

Kellerman, E. (1979). Transfer and non-transfer: Where are we now. Studies in Second

Language Acquisition, 2, 37-57.

Kellerman, E., & Jordens, P. (1977). Towards a characterization of the strategy of transfer
in second-language Ieaming. Interlanguage Studies Bulletin, 2, 58-145.

Kellogg, R. T., & Dare, R. S. (1989). Explicit memory for unattended information.
Bulletin of the Psychonomic Society, 27, 409-12.

263

Kennedy, G (1990). Collocations: Where grammar and vocabulary teaching meet. In S.
Anivan (Ed), Language teaching methodology for the nineties (pp.215-229).
Singapore: SEAMEO Regional Language Centre.

Kennedy, G (2003). Amplifier collocations in the British National Corpus: Implications
for English language teaching. TESOL Quarterly, 37, 467-487.

Knight, S. (1994). Dictionary: The tool of last resort in foreign language reading? A new
perspective. Modern Language Journal, 78, 285-299.

Knowlton, B. J. & Squire, L. R. (1996). Aritificial grammar Ieaming depends on implicit
acquisition of both abstract and exemplar-specific information. Journal of

Experimental Psychology: Learning. Memory, and Cognition, 22, 169 -181.

Krashen, S. (1982). Principles and practice in second language acquisition. Oxford:

Pergamon.
Krashen, S. (1985). The input hypothesis. London: Longman.

Krashen. S. (1989). We acquire vocabulary and spelling by reading: Additional evidence
for the Input Hypothesis. Modern Language Journal, 73, 440-464.

Krashen, S. (1992). Formal grammar instruction: Another educator comments. TESOL
Quarterly, 26, 409-411 .

Krashen, S. (1994) The input hypothesis and its rivals. In N. Ellis (Ed), Implicit and
explicit learning of language (pp. 45-77). London: Academic Press.

Krashen, S., & R. Scarcella. (1978). On routines and patterns in second language
acquisition and performance. Language Learning, 28, 293-300.

Laufer, B. (2000). Electronic dictionaries and incidental vocabulary acquisition: does
technology make a difference? In U. Heid, S. Evert, E. Lehmann, & C. Rohrer
(Eds), EURALEX (pp. 849-854). Stuttgart University.

Laufer, B. (2003). Vocabulary acquisition in a second language: Do learners really

264

acquire most vocabulary by reading? Some empirical evidence. The Canadian
Modern Language Review, 59, 567-587.

Laufer, B. ,& Hulstijn, J. (2001). Incidental vocabulary acquisition in a second language:
The construct of task-induced involvement. Applied Linguistics, 22, 1-26.

Laufer, B., & Goldstein, Z. (2004). Testing vocabulary knowledge: size, strength, and
computer adaptiveness. Language Learning, 54, 399-436.

Leeman, J ., Arteagoitia, 1., Fridman, B., & Doughty, C. (1995). Integrating attention to
form with meaning: Focus on form in content-based Spanish instruction. In R.
Schmidt (Ed), Attention and awareness in foreign language learning (pp. 217-258).
Honolulu: University of Hawai'i Press.

Lennon, P. (1996). Getting “easy” verbs wrong at the advanced level. IRAL, 34, 23-36.

Leow, R. P. (1997). Attention, awareness. and foreign language behavior. Language
Learning, 47, 467-505.

Levelt, W. (1989). Speaking: From Intention to Articulation. Cambridge: MIT Press.

Lewis, M. (1993). The lexical approach. Hove and London, England: Language Teaching
Publications.

Lewis, M. (Ed) (2000). Teaching collocation. Further development in the lexical
approach. Hove: LTP.

Li ghtbown, P. (1991). What have we here? Some observations on the influence of
instruction on L2 Ieaming. In R. Phillipson, E. Kellerman, L. Selinker, M. Sharwood
Smith & M. Swain (Eds), F oreign/ second language pedagogy research
(pp. 197-212). C levedon, Avon: Multilingual Matters.

Lightbown, P. M. (1998). The importance of timing in focus on form. In C. Doughty & J.

Williams (Eds), Focus on form in classroom second language acquisition (pp.
177-196). Cambridge: Cambridge University Press.

265

Lightbown, P., & Spada, N. (1990). Focus-on-fonn and corrective feedback in
communicative language teaching: Effects on second language Ieaming. Studies in
Second Language Acquisition, 12, 429-448.

Lococo, V. (1976). A comparison of three methods for the collection of L2 data: free

composition, translation and picture description. Working Papers on Bilingualism, 8,
59-86

Logan, G (1988). Toward an instance theory of automatization. Psychological Review, 95,
492-527.

Long, M. (1988). Instructed interlanguage development. In L. Beebe (Ed), Issues in
second language acquisition: Multiple perspectives (pp. 115-141). Rowley, Mass:
Newbury House.

Long, M. (1991). Focus on form: A design feature in language teaching methodology. In
K. de Bot, R. Ginsberg, & C. Kramsch (Eds), Foreign language research in

cross-cultural perspective (pp. 39-52). Amsterdam: John Benjamins.

Long, M. (1996). The role of the linguistic environment in second language acquisition.
In W. Ritchie and T. Bhatia (Eds), Handbook of second language acquisition (pp.
413-468). San Diego, CA: Academic Press

Longman Dictionary of Contemporary English. (1996). Pearson P T R.

Lorenz, G (1999). Adjective Intensification — Learners Versus NSs. A Corpus Studv of
Argumentative Writing. Amsterdam: Rodopi.

Mackey, A. Philp, J. Fujii, A., Egi , T., & Tatsumi, T. (2002). Individual differences in I
working memory, noticing of interactional feedback and L2 development. In P.
Robinson & P. Skehan (Eds), Individual difi‘erences in L2 learning (pp. 181-208).
John Benjamins.

Mackey, A., & Gass, S. (2005). Second language research: methodology and design.
Lawrence Erlbaum Associates.

266

Mackey, A., & Philp, J. (1998). Recasts, interaction and interlanguage development: Are
responses red herrings? Modern Language Journal, 82, 338-356.

Mackey, A., Gass, S., & McDonough, K. (2000). How do learners perceive interactional
feedback? Studies in Second Language Acquisition, 22, 471-497.

Mackin, R. (1978). On collocations: Words shall be known by'the company they keep. In
P. Strevens (Ed) In Honour of A. S. Hornby, (pp. 149-165). Oxford: Oxford
University Press.

Macrory, G, & Stone, V. (2000). Pupil progress in the acquisition of the perfect tense in
French: The relationship between knowledge and use. Language Teaching Research,
4, 55-82.

MacWhinney, B. (1997). Implicit and explicit processes: commentary. Studies in Second

Language Acquisition, 19, 277-281.

Marcel, A. J. (1983). Conscious and unconscious perception: An approach to the relations
between phenomenal experience and perceptual processes. Cognitive Psychologv, 15,
238-302.

Marton, W. (1977). Foreign vocabulary Ieaming as problem No. 1 oflanguage teaching at
the advanced level. Interlanguage Studies Bulletin, 2, 33-57.

Master, P. (1994). The effect of systematic instruction on Ieaming the English article
system. In T. Odlin (Ed), Perspective on pedagogical grammar (pp. 229-252).
Cambridge: Cambridge University Press.

Mathews, R. C., Buss, R.R., Stanley, W.B., Blanchard-Fields, F., Cho, J. R., & Druhan, B.
(1989). Role of implicit and explicit processes in Ieaming from examples: A

synergistic effect. Journal of Experimental Psychology: Learning, Memory and
Cognition, 15, 1083-1100.

McAndrews, M. P., & Moscovitch, M. (1985). Rule-based and exemplar-based
classification in artificial grammar Ieaming. Memory & Cognition, 13, 469-475.

267

McLaughlin, B. (1990). The relationship between first and second languages: language
proficiency and language aptitude. In B. Harley, P. Allen, J. Cummins, & M. Swain
(Eds) The Development of Second Language Proficiency. Cambridge: Cambridge

University Press

McLaughlin, B., Rossman, T., & McLeod, B. (1983). Second language Ieaming: An
information-processing perspective. Language Learning, 33, 135-158.

McLeod, B. & McLaughlin, B. (1986) Restructuring or automaticity? Reading in a
second language. Language Learning, 36, 109-123.

Melton, A. W. (1963). Implications of short-terrn memory for a general theory of memory.

Journal of Verbal Learning and Verbal Behavior, 2, 1-21.

Meulemans, T., & Van der Linden, M. (1997). Associative chunk strength in artificial
grammar Ieaming. Journal of Experimental Psychology: Learning, Memory and
Cognition, 23, 1007-1028.

Miller, G. A. (1956) The magical number seven, plus or minus two: some limits on our

capacity for processing information. Psychological Review, 63. 81-97.

Moon, R. (1997). Vocabulary connections: multi-word items in English. In N. Schmitt &
M. McCarthy (Eds), Vocabulary: description. acquisition and pedagogy (pp. 40-63).
Cambridge: Cambridge University Press.

Nation, 1. (2001). Learning Vocabulary in Another Language. Cambridge: Cambridge

University Press.

Nation, R., & Mclaughlin, B. (1986). Experts and novices: an information-processing
approach to the "good language learner" problem. Applied Psycholinguistics, 7,
41-56.

Nattinger, J., & DeCarrico, J. (1992). Lexical phrases and language teaching. Oxford:
Ox ford University Press.

268

Nayak, N., Hansen, N., Krueger, N., & Mclaughlin, B. (1990). Language learning
strategies in monolingual and multilingual adults. Language Learning, 40, 221-244.

Nelson, K. (1973). Structure and strategy in Ieaming to talk. Monographs of the Society
for Research in Child Development, 38.

Nesselhauf, N. (2003). The use of collocations by advanced learners of English and some

implications for teaching. Applied Linguistics, 24: 223-242.
New Oxford Dictionary of English. (1998). Oxford: Clarendon Press.

Newell, A. (1990). Unified theories of cognition. Cambridge, MA: Harvard University

Press.

Newton, J. (1995). Task-based interaction and incidental vocabulary Ieaming: a case

study. Second Language Research, 11, 159-1 77.

Norris, J. M., & Ortega, L. (2000). Effectiveness of L2 instruction: A research synthesis
and quantitative meta-analysis. Language Learning. 50, 417-528.

Oliver, R. (1995). Negative feedback in child NS-NNS conversation. Studies in Second
Language Acquisition, 17, 459-482.

Oxford Dictionary of Current Idiomatic English. (1983). London: Oxford University

Press.
Palmer, H. E. (1933) Second Interim Report on English C ollocations. Tokyo: Kaitakusha.
Paribakht, T., & Wesche, M. (1997). Vocabulary enhancement activities and reading for
meaning in second language vocabulary acquisition. In J. Coady & T. Huckin (Eds),
Second language vocabulary acquisition: A rationale for pedagogy (pp. 174-200).

Cambridge: Cambridge University Press.

Pawley, A., & Syder, F. (1983). Two puzzles for linguistic theory: nativelike selection and
nativelike fluency. In J. Richards & R. Schmidt (Eds), Language and

269

communication (pp 191-225). London: Longman.

Peters, S. (1998). Evaluating language Ieaming technology from a linguist's perspective.
Paper presented at the Invitational Symposium on Advancing Technology Options in
Language Learning, Honolulu, February.

Peters. A. (1983). The units of language acquisition. Cambridge: Cambridge University

Press.

Pica, T. (1988). Interlanguage adjustments as an outcome of NS-NNS negotiated
interaction. Language Learning, 38, 45-73.

Pica, T., Holliday, L.,-Lewis, N., & Morgenthaler, L. (1989). Comprehensible output as
an outcome of linguistic demands on the learner. Studies in Second Language

Acquisition, l I, 63-90.

Pienemann, M. & Johnston, M. (1986). An acquisition-based procedure for language

assessment. Australian Review of Applied Linguistics, 9, 92 - 112.

Pienemann, M. (1987). Determining the influence ofinstruction on L2 speech processing.

Australian Review of Applied Linguistics, 10, 83-113.

Pienemann. M. (1989). Is language teachable? Applied Linguistics, 10, 52-79.

Posner, M. 1., & Petersen, S. E. (1990). The attention system of the human brain. Annual
Review of Neuroscience, 13, 25-42

Ramsay, R. M. G (1980). Language-Ieaming approach styles of adult multilinguals and
successful language learners. Annals of the New York Academy of Sciences, 75,

73-96.

Reber, A. (1993). Implicit learning and tacit knowledge. Oxford, England: Clarendon
Press.

Reber, A. S. (1967). Implicit Ieaming of artificial grammars. Journal of Verbal Learning
and Verbal Behavior; 77, 317-327.

270

Reber, A. S. (1976), Implicit Ieaming of synthetic languages: The role of instructional set.
Journal of Experimental Psychology: Human Learning and Memory, 2, 88-94.

Reber, A. S. (1989). Implicit Ieaming and tacit knowledge. Journal of Experimental
Psychology: General, 118, 219-235.

Reber, A. S., & Allen, R. (1978). Analogy and abstraction strategies in synthetic grammar
Ieaming: A functionalist interpretation. Cognition, 6, 189-221.

Reber, A. S., Kassin, S. M., Lewis, S., & Cantor, G W. (1980). On the relationship
between implicit and explicit modes in the Ieaming of a complex rule structure.

Journal of Experimental Psychology: Human Learning and Memory, 6, 492-502.

Robb, T., Ross, S., & Shortreed, I. (1986). Salience of feedback on error and its effects on
EFL writing quality. TESOL Quarterly, 20, 83-95.

Robinett, B., & Schachter, J. (Eds) (1983). Second language learning: C ontrastive
analysis, error analysis, and related aspects. Ann Arbor: The University of

Michigan Press.

Robinson, P. (1995a). Aptitude, awareness, and the fundamental similarity of implicit and
explicit second language learning. In Richard Schmidt (Ed) Attention and
awareness in foreign language learning (Technical Report #9, pp. 303-357).

Honolulu: Hawai'i: University of Hawai'i.

Robinson, P. (1995b). Attention, memory and the “noticing” hypothesis. Language
Learning, 45, 283-331.

Robinson, P. (1996). Learning simple and complex second language rules under implicit,
incidental, rule-search and instructed conditions. Studies in Second Language

Acquisition, 18, 27-67.

Robinson, P. (1997a). Individual differences and the fundamental similarity of implicit
and explicit adult second language learning. Language Learning, 47, 45-99

271

Robinson, P. (1997b). Generalizability and automaticity of second language Ieaming
under implicit, incidental, enhanced, and instructed conditions. Studies in Second
Language Acquisition, 19, 223-247.

Robinson, P. (2003). Attention and memory during SLA. In C. Doughty & M. Long
(Eds), The handbook of second language acquisition (pp. 631-677). Malden, MA:
Blackwell.

Robinson, P. J ., & Ha, M. A. (1993). Instance theory and second language rule Ieaming
under explicit conditions. Studies in Second Language Acquisition, 15, 412-438.

Rosa, E., & O'Neill, M. (1999). Explicitness, intake and the issue of awareness. Studies in
Second Language Acquisition, 21, 511-556.

Salica, C. (1981). Testing a model of corrective discourse. Unpublished masters in TESL

thesis, University of California, Los Angeles.

Sanders, A. (1998). Elements of human performance. Mahwah, NJ: Lawrence Erlbaum

Associates.

Sanz. C. & Morgan-Short, K. (2004). Positive evidence versus explicit rule presentation
and explicit negative feedback: A computer-assisted study. Language Learning. 54,
35-78.

Schachter, J. (1991). Corrective feedback in historical perspective. Second Language
Research, 7, 89-102.

Schmidt, M. (1980). Coordinate structures and language universals in interlanguage.

Language Learning, 30, 397-416.

Schmidt, R. & Frota, S. (1986). Developing basic conversational ability in a second
language: A case study of an adult learner of Portuguese. In R. Day (Ed), Talking to
learn: Conversation in second language acquisition (pp. 237-322). Rowley, MA:

Newbury House.

Schmidt, R. (1990). The role of consciousness in second language Ieaming. Applied

272

Linguistics, 11, 129-158.

Schmidt, R. (1994). Implicit Ieaming and the cognitive unconscious. In N. Ellis (Ed),
Implicit and explicit learning of languages (pp. 165-209). London: Academic Press.

Schmidt, R. (1995). Consciousness and foreign language Ieaming: A tutorial on the role
of attention and awareness in Ieaming. In R. Schmidt (Ed) Attention and Awareness

in Foreign Language Learning (pp. 1-63). Honolulu: University of Hawaii Press.

Schmidt, R. (2001). Attention. In P. Robinson (Ed), Cognition and Second Language
Instruction (pp. 3-32). Cambridge, England: Cambridge University Press.

Schmitt, N. (2000). Vocabulary in Language Teaching. Cambridge: Cambridge

University Press.

Schmitt, N., & McCarthy, M. (Eds) (1997). Vocabulary: Description, Acquisition and
Pedagogy. Cambridge: Cambridge University Press. '

Schneider, W. & Detweiler, M. (1988). A connectionist control architecture for working
memory. In G H. Bower (Ed) The Psychology of Learning and Motivation, 21,

54-119. New York: Academic Press.

Schneider, W. & Shiffrin, R. M. (1977). Controlled and automatic information processing

1: Detection, search and attention. Psychological Review. 84, 1-66.

Schwartz, B. (1993). On explicit and negative data effecting and affecting competence
’ and linguistic behavior. Studies in Second Language Acquisition, 15, 147-163.

Schwartz, 8., & GubaIa-Ryzak, M. (1992). Leamability and grammar reorganization in
L2A: Against negative evidence causing the unleaming of verb movement. Second

Language Research, 8, 1-38.

Scott, V. (1989). An empirical study of explicit and implicit teaching strategies in French.
Modern Language Review, 72, 1, 14-22.

Seliger, H. W. (1979). On the nature and function of language rules in language teaching.

273

TESOL Quarterly, 13, 359-369.
Semke, H. (1984). The effects of the red pen. Foreign Language Annals. 17, 195-202.

Sergeant, J . (1996). A theory of attention: an information processing perspective. In G
Reid Lyon & N. A. Krasnegor (Eds), Attention and executive function (pp. 57-69).
Baltimore, MD: Paul H. Brookes.

Shanks, D. R., & St. John, M. F. (1994). Characteristics of dissociable human Ieaming
systems. Behavioral and Brain Sciences, 17, 367-395.

Shiffrin, R.M., & Schneider, W. (1977). Controlled and automatic human information
processing: 11. Perceptual Ieaming, automatic attending, and a general theory.
Psychological Review, 84, 127-190.

Shook, D. (1994). FL/L2 reading, grammatical information, and the input-to-intake
phenomenon. Applied Language Learning, 5, 57-93.

Sinclair, J ., & Fox, G. (1990). Collins C OBUILD English Grammar. London: Collins.

Sinclair, J. M. (1987). Collocation: a progress report. In R. Steele & T. Threadgold (Eds),
Language topics: essays in honor of Michael Halliday (Vol. II, pp. 319-331). John

Benjamins, Amsterdam.

Sinclair, J. M. (1991). Corpus, concordance, collocation. Oxford: Oxford University

Press.

Skehan, P. (1989). Individual differences in second-language learning. London: Edward
Arnold.

Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford
University Press.

Slimani, A. (1992). Evaluation of classroom interaction. In J. Alderson & A. Beretta

(Eds), Evaluating Second Language Education (pp. 197-220). Cambridge, UK:
Cambridge University Press.

274

Smith, B. (2004). Computer-mediated negotiated interaction and lexical acquisition.

Studies in Second Language Acquisition, 26, 365-398.

Spada, N. (1987). The relationship between instructional differences and Ieaming
outcomes: A process-product study of communicative language teaching. Applied
Linguistics, 8, 137-155.

Spada, N., & Lightbown, P. (1993). Instruction and the development of questions in L2

classrooms. Studies in Second Language Acquisition, 15, 205-224.

Stubbs. M. (1995). Corpus evidence for norms of lexical collocation. In G Cook & B.
Seidlhofer (Eds) Principles and practice in applied linguistics (pp. 245-256).
Oxford University Press.

Tarallo, F., & Myhill, J. (1983). Interference and natural language in second language
acquisition. Language Learning, 33, 55-76.

Tarone, E. (1982). Systematicity and attention in interlanguage. Language Learning. 32,
69-82.

Terrell, T. (1991). The role of grammar instruction in a communicative approach. Modern

Language Journal, 75, 52-63.

Tomlin, R., & Villa, V. (1994). Attention in cognitive science and SLA. Studies in Second
Language Acquisition, 16, 183-204.

Trahey, M., & White, L. (1993). Positive evidence and preemption in the second
language classroom. Studies in Second Language Acquisition, 15, 181-204.

Truscott, J. (1998). Noticing in second language acquisition: a critical review. Second

Language Research, 14, 103-135.

VanPatten, B. (1990). Attending to form and content in the input: An experiment in
consciousness. Studies in Second Language Acquisition, 12, 287-301.

275

VanPatten, B. (1994). Evaluating the role of consciousness in second language
acquisition: Terms, linguistic features and research methodology. AILA Review, 11,
27-36.

VanPatten, B., & Oikkenon, S. (1996). Explanation versus structured input in processing
instruction. Studies in Second Language Acquisition, 18, 495-510.

Vokey, J. R. & Brooks, L. R. (1992). Salience of item knowledge in Ieaming artificial

grammars. Journal of Experimental Psychology: Learning, Memory and Cognition,
18, 328-344.

Watanabe, Y. (1997). Input, intake, and retention: Effects of increased processing on
incidental Ieaming of foreign language vocabulary. Studies in Second Language
Acquisition, 19, 287-307.

Westney, P. (1994). Rules and pedagogical grammar. In T. Odlin (Ed), Perspectives on
pedagogical grammar (pp. 72-96). Cambridge: Cambridge University Press.

White, L. (1991). Adverb placement in second language acquisition: Some effects of
positive and negative evidence in the classroom. Second Language Research, 7,

133-161.

White, L. Spada, N. Lightbown, P., & Ranta, L. (1991). Input enhancement and L2
question formation. Applied Linguistics, 12, 416-432.

Wickens, C. (1984). Processing resources in attention. In R. Parasuraman & D. Davies
(Eds), Varieties of attention (pp 63-102). New York: Academic Press.

Wickens, C. (1989). Attention and skilled performance. In D. Holding (Ed), Human
skills (pp. 71-105). New York: John Wiley.

Wickens, C. (1992). Engineering psychology and human performance: 2nd edition. New
York: Harper Collins

276

Williams, J. (1999). Memory, attention, and inductive Ieaming. Studies in Second

Language Acquisition, 21, 1-48.

Williams, J., & Evans, J. (1998). What kind of focus and on which forms? In C. Doughty
& J. Williams (Eds), Focus on form in classroom second language acquisition (pp.
139-155). Cambridge: Cambridge University Press.

Winter, W., & Reber, A. S. (1994). Implicit Ieaming and natural languages. In N. Ellis

(Ed), Implicit and explicit learning of languages (pp. 115-146). London: Academic

Press.

Wolfe-Quintero, K. (1992). Leamability and the acquisition of extraction in relative

clauses and Wh-questions. Studies in Second Language Acquisition, 14, 39-70.

Wong-Fillmore, L. (1976). The second time around: cognitive and social strategies in

second language acquisition. Unpublished PhD thesis, Stanford University.

Wren, D. (1982). A case study of the treatment of oral errors. Selected papers in TESOL,
1, 90-103.

Yoon. H. (2004). Collocational competence and L2 writing development. Paper presented

at Second Language Research Forum 2004, Pennsylvania.

Zemick, U., & Dyer, M. (1987). The self-extending phrasal lexicon. Computational
Linguistics, 13, 308-27.

Zobl, H. ( I995). Converging evidence for the acquisition/Ieaming distinction. Applied
Linguistics, 16, 35-57.

277

 

  

     

IIIIIIIIII

llllllllllll ] 1]]

i RRRRRRRRR
3 1293

       

 

llllllllllllll

36 7576