republic of turkey Çukurova university …bu nokta, çalışmanın yarı-deneysel bölümünün...
TRANSCRIPT
REPUBLIC OF TURKEY
ÇUKUROVA UNIVERSITY
THE INSTITUTE OF SOCIAL SCIENCES
ENGLISH LANGUAGE TEACHING DEPARTMENT
A COMPARATIVE INVESTIGATION OF LEXICAL NETWORKS OF
TURKISH LEARNERS OF ENGLISH AS A FOREIGN LANGUAGE:
A CORPUS BASED STUDY
İhsan ÜNALDI
A Ph.D. DISSERTATION
ADANA, 2011
REPUBLIC OF TURKEY
ÇUKUROVA UNIVERSITY
THE INSTITUTE OF SOCIAL SCIENCES
ENGLISH LANGUAGE TEACHING DEPARTMENT
A COMPARATIVE INVESTIGATION OF LEXICAL NETWORKS OF
TURKISH LEARNERS OF ENGLISH AS A FOREIGN LANGUAGE:
A CORPUS BASED STUDY
İhsan ÜNALDI
Advisor: Assoc. Prof. Dr. Yasemin KIRKGÖZ
A Ph.D. DISSERTATION
ADANA, 2011
To the Directorship of the Institute of Social Sciences, Çukurova University
We certify that this dissertation is satisfactory for the award of degree of Doctor
of Philosophy in the subject matter of English Language Teaching
Supervisor: Assoc. Prof. Dr. Yasemin KIRKGÖZ
Member of Examining Committee: Prof. Dr. Aytekin İŞMAN
Member of Examining Committee: Prof. Dr. Hatice SOFU
Member of Examining Committee: Assoc. Prof. Dr. Ahmet DOĞANAY
Member of Examining Committee: Asst. Prof. Dr. Abdurrahman KİLİMCİ
I certify that this dissertation confirms to the formal standards of the Institute of Social
Sciences
…/……/……
Prof. Dr. Azmi YALÇIN
Director of the Institute
PS. The uncited usage of the reports, charts, figures, and tables in this dissertation,
whether original or quotes for mother sources, is subject to the Law of Works of Art and
Thought No: 5846
Not: Bu tezde kullanılan, özgün ve/veya başka kaynaktan yapılan rapor, çizelge, şekil
ve tabloların kaynak gösterilmeden kullanımı 5846 sayılı Fikir ve Sanat Eserleri
Kanunu’ndaki hükümlere tabidir.
iii
ÖZET
İNGİLİZCEYİ YABANCI DİL OLARAK ÖĞRENEN TÜRK ÖĞRENCİLERİN
KELİME AĞLARININ KARŞILAŞTIRILMALI ARAŞTIRMASI:
DERLEM TABANLI BİR ÇALIŞMA
İhsan ÜNALDI
Doktora Tezi, İngiliz Dili Eğitimi Anabilim Dalı
Danışman: Doç. Dr. Yasemin KIRKGÖZ
Aralık 2011, 174 sayfa
Bu çalışma, betimsel ve yarı-deneysel olarak iki ana bölümden oluşmaktadır.
Çalışmanın ilk amacı, İngilizceyi yabancı dil olarak öğrenen Türk öğrenciler tarafından
yazılan metinlerdeki sözcük özellikleriyle, anadili İngilizce olan öğrenciler tarafından
yazılmış olan metinlerdeki sözcük özelliklerinin farklılıklarını belirlemektir. Çalışmanın
diğer amacı ise, derlem tabanlı dil öğrenme aktiviteleri kullanarak Türk öğrencilerin
sözcük bilgisi ile ilgili sorunları çözmelerine yardımcı olmaktır.
Çalışmanın betimsel aşamasında, 49 Türk ve 100 Amerikalı-İngiliz öğrenci
tarafından yazılmış metinler karşılaştırılmıştır. Metinler, çevrimiçi bir veri tabanı olan
Coh-metrix kullanılarak işlenmiştir. Türk öğrencilerinin metinlerindeki sözcük
bağlantılarını, anadili İngilizce olan grupların metinlerindeki sözcük bağlantılarıyla
karşılaştırırken, Coh-metrix’teki indekslerden biri olan Gizli Anlam Analizi (GAA)
kullanılmıştır. İstatistiksel analizler, Türk öğrencileri tarafından yazılmış metinlerin,
diğer grup tarafından yazılmış olan metinlerden atıf, anlam, okunabilirlik ve sözdizimi
indekslerinde anlamlı farklılıklar gösterdiğini ortaya çıkarmıştır. Odak grup
görüşmeleri, bu farklılıkların çoğunlukla sözcük bilgisi yetersizliğinden kaynaklandığını
göstermiştir. Bu nokta, çalışmanın yarı-deneysel bölümünün gerekçesini
oluşturmaktadır.
Çalışmanın yarı-deneysel bölümünde, orta ve orta-üstü İngilizce yeterliliği olan
37 Türk öğrenci yer almıştır. Deney grubu 18, kontrol grubu ise 19 öğrenciden
oluşturulmuştur. Deneysel uygulamadan önce, her iki gruba da kelime tanıma ve
kompozisyon yazma testleri verilmiştir.
iv
Çalışmanın yarı-deneysel süreci, derlem tabanlı bağlamlı-dizin uygulamaları
içermiş ve yaklaşık 10 hafta sürmüştür. Uygulamanın sonunda, gruplara uygulamanın
başında verilmiş olan sözcük tanıma ve kompozisyon testleri tekrar verilmiştir.
Uygulamanın gecikme etkilerini görebilmek için aynı kompozisyon testi yaklaşık iki
hafta sonra tekrar uygulanmıştır.
Analiz sonuçları, sözcük tanıma testinde, deney grubunun kontrol grubuna göre
anlamlı bir şekilde daha başarılı olduğunu göstermiştir. Deney grubu ayrıca, sözcük
üretimi açısından da kontrol grubunu anlamlı bir şekilde geride bırakmıştır. Öte yandan,
bağlamlı-dizin çalışmalarının, GAA tekniği kullanılarak ölçülen sözcük bağlantıları
üzerinde anlamlı bir etkisinin olmadığı da saptanmıştır.
Çalışmanın son aşamasında deneysel gruptaki öğrencilerle bağlamlı-dizin
aktivitelerinin dil öğrenimi bağlamında kullanılmasıyla ilgili yarı-yapılandırılmış
görüşmeler yapılmıştır. Sonuçlar, öğrencilerin çoğunluğunun bu aktivitelere karşı
olumlu tutumlar geliştirdiklerini ve bu aktiviteleri faydalı bulduklarını göstermiştir.
Anahtar Kelimeler: Derlem, Öğrenci derlemi, Sözcük ağları, Sözcüksel uyum, Uyum-
dizini
v
ABSTRACT
A COMPARATIVE INVESTIGATION OF LEXICAL NETWORKS OF
TURKISH LEARNERS OF ENGLISH AS A FOREIGN LANGUAGE:
A CORPUS BASED STUDY
İhsan ÜNALDI
Ph. D. Dissertation, English Language Teaching Department
Supervisor: Assoc. Prof. Dr. Yasemin KIRKGÖZ
December 2011, 174 pages
The current study is composed of two main parts as descriptive and
experimental. The aims are, first to determine lexical differences between texts written
by Turkish EFL learners and native speakers of English, and then to make use of
corpus-based language learning activities to help the learners overcome lexicon related
problems.
In the descriptive phase of the study, essays written by 49 Turkish EFL learners
and 100 native speakers of English were compared. These essays were processed in
Coh-metrix which is an online database. One of the indices in Coh-metrix, Latent
Semantic Analysis (LSA), was used to compare lexical cohesion in learners’ texts to the
native ones. Statistical analyses revealed that learner essays were significantly different
from native ones in referential, semantic, reading comprehensibility and syntactic
aspects. Focus group interviews which were carried out with the same learner group
brought to light that these differences mostly stemmed from lexicon related problems.
This point was the main rationale behind the quasi-experimental phase of the study.
In the quasi-experimental phase, 37 intermediate and upper-intermediate level
Turkish EFL learners participated in the study. The experimental group was composed
of 18 learners, and there were 19 learners in the control group. Before the treatment,
both groups were given vocabulary recognition and essay writing tests.
The experimental treatment included corpus-based concordancing activities.
This treatment lasted for about 10 weeks, and at the end, both groups were given the
same vocabulary recognition and writing test again. In order to test the delayed effects
of the treatment, the same written test was given to both groups after about two weeks.
vi
Analysis results revealed that the experimental group obtained significantly
better scores compared to the control group in terms of vocabulary recognition and
production. However, concordancing activities did not have any significant effects on
lexical cohesion in control group’s texts.
As the last stage of the study, the participants were interviewed about the use of
concordancing activities in language instruction. The results indicated that the majority
of the participants improved positive attitudes towards these activities and found them
beneficial.
Keywords: Corpus, learner corpora, lexical networks, lexical cohesion, concordancing
vii
ACKNOWLEDGEMENTS
I came across many people during my study and now I realize that it is, in fact, a
good thing not to know who to thank; there are so many of them.
First of all, I would like to thank to Associate Professor Dr. Yasemin KIRKGÖZ
for all her help and patience throughout the study.
Special thanks go to Professor Dr. Hatice SOFU, and Professor Dr. Aytekin
İŞMAN, Associate Professor Dr. Erdoğan BADA and Associate Professor Dr. Ahmet
DOĞANAY. Assistant Professor Dr. Abdurrahman KİLİMCİ, Assistant Professor Dr.
Fehmi Can SENDAN and Assistant Professor Dr. Rana YILDIRIM deserve my
gratitude for their supports.
Dr. Naime Feyza (Altınkamış) TÜRKAY, whose help was always in time and
extremely valuable, also deserves special thanks.
I would also like to thank to my colleagues Assistant Professor Dr. Birsen
BAĞÇECİ and Mehmet ALTAY, for their encouragement and support. I am also
grateful to D. Celeste SHOPE for kindly and wisely proofreading the whole study.
I also feel grateful to Sylviane GRANGER and Danielle MCNAMARA for
generously sharing their valuable data and experience without any sign of hesitation.
To my family,
Words in any language expressing gratitude
are blind and deaf at the same time,
and they are no substitute.
And to those whose names I forgot unknowingly,
“I can no other answer make, but, thanks, and thanks.”
viii
TABLE OF CONTENTS
Page
ÖZET………………………………………………………...………………………...
ABSTRACT…………………………………..……..………………………………...
AKNOWLEDGEMENTS……………………..……………..……………………....
LIST OF ABBREVIATIONS…………………..……………………..……..….…...
LIST OF TABLES………………………………..………………………..………....
LIST OF FIGURES………………………………..………………………..………..
LIST OF APPENDICES……………………………..………………..….……….....
CHAPTER 1
INTRODUCTION
1.0. Introduction ………………………………………………………….…………….
1.1. Background of the Study……………………………………………..……………
1.2. Statement of the Problem…………………………………………..……………....
1.3. Aim and Research Questions……………………………………..…………….….
1.4. Significance of the Study………………………………………..………………...
1.5. Assumptions and Limitations……………………………………..……………….
1.6. Operational Definitions and Key Terms...………………………..…………….….
CHAPTER 2
REVIEW OF LITERATURE
2.0. Introduction………………………………………………………..……..……..…
2.1. Theoretical Framework……………………………………..……………...……....
2.1.1. Interlanguage theory…………………………………..…………………...
2.1.2. Differences between L1 and L2 writing……………..………………….…
2.2. What is Coh-metrix?.…………………………………………....………………….
2.2.1. Readability………………………………………….…..………………….
2.2.2. General Word and Text Information……………….…..………………….
2.2.3. Syntax …………………………………………….…..…………………...
iii
v
vii
xii
xiii
xvi
xvii
1
1
3
4
6
6
7
9
9
10
11
16
16
17
17
ix
2.2.4. Referential and Semantic Aspects……………………..………….……….
2.2.5. What is Latent Semantic Analysis?……….….…..…………….………….
2.2.6. Situation Model Dimensions…………………………..………….……….
2.3. Interlanguage Lexicon………………………………………..……………………
2.3.1. Breadth and Depth of Vocabulary Knowledge…………………………....
2.3.2. Lexical Networks……………………………………………………….....
2.3.3. Lexical Cohesion…………………………………………………….…….
2.4. Corpus Linguistics………………………………………………….…………..…..
2.4.1. Learner Corpora: A Revolution……………………………………...…….
2.4.2. Corpus Linguistics and Language Pedagogy………………………..…….
2.4.3. Corpus Linguistics and Vocabulary Teaching/Learning……………....…...
2.5. Data-driven learning…………………………………………………………...…...
2.5.1. Criticisms and Downsides of DDL……………………………….…….…
2.6. Using Concordancers in Language Teaching/Learning…………………………....
2.6.1. Learners’ Attitudes towards Corpora as a Language Learning Tool………
2.7. Summary………………………………………………………………….......……
CHAPTER 3
METHODOLOGY
3.0. Introduction……………………………………………………..…………………
3.1. The Design of the Study…………………………………………………..……….
3.2. Sampling and the Participants…………………………….………………..………
3.3. Context of the students……………………………………………………..……...
3.4. Data Collection Tools and Procedures………………………………………..…...
3.4.1. Descriptive Procedures………………………………………….................
3.4.2. The Selection of Corpora………………………………..……..………….
3.4.3. Semi-Structured Focus Group Interviews……………….……...................
3.4.4. Experimental Procedures…………………………………..……...……….
3.4.5. Piloting of the Multiple Choice Test for the Target Vocabulary…….……
3.4.6. Pre-test for Written Production…………….……..……………..………...
3.4.7. Vocabulary Teaching Material…………………..…………...…................
3.4.8. Classroom Procedures………………………….…...……………………..
18
18
21
22
23
25
27
32
34
37
41
42
43
45
50
51
53
53
55
57
58
58
58
62
63
63
65
66
68
x
3.4.9. LSA Scores………………………………...……..………………………..
3.4.10. Target Vocabulary Density Scores………………..………………...……..
3.4.11. Semi-structured Interviews………………………..……..………………..
3.5. Overview of the Descriptive and Experimental Procedures…..…………………...
3.6. Overview of the Statistical Techniques Used in the Study…..……………………
CHAPTER 4
RESULTS
4.0. Introduction……………………………………………………..…………………
4.1. Results of the Descriptive Phase………………………………..………………….
4.1.1. Normality and homogeneity of the data………………..……………..…….
4.1.2. Referential and Semantic Indices……………………..……...……………...
4.1.3. Readability Indices…………………………………..…………..………….
4.1.4. Syntax Indices……………………………………..………..………………
4.1.5. Results of the Focus Group Interviews……………..………..……...……...
4.2. Results of the Experimental Phase…………………………..……………………..
4.2.1. Experimental Results for Recognition…………………..…..……................
4.2.2. Confirmation of the Descriptive Results for the Experimental Phase……....
4.2.3. Target Vocabulary Density Scores…………..…………...………………….
4.2.4. Experimental Results for Lexical Cohesion……..…………..……………...
4.2.5. Semi-structured Interview Results……………..………...…………….……
4.3. Overview of the Descriptive Results……………………..………………………..
4.4. Overview of the Experimental Results…………………….……………................
4.5. Summary………………………………………………..………………………….
CHAPTER 5
DISCUSSION AND CONCLUSIONS
5.0. Introduction……………………………………………..……………………..…..
5.1. General Summary of the Study………………………..…………………………..
5.2. Review of the Findings in Relation to the Research Questions…..………………..
5.3. Implications of the Study………………………………………..………................
70
70
72
72
73
75
75
76
79
93
96
103
106
106
109
110
114
121
125
128
128
131
131
134
139
xi
5.4. Recommendations for Further Research……………………..…………………….
5.5. Personal Reflections and Criticism of the Study……………..……………………
REFERENCES………………………………..……………...…………….…………
APPENDICES ……………………………………………………..………………….
CURRICULUM VITAE ………………………………………..……………………
140
141
142
150
172
xii
TABLE OF CONTENTS
Page
ÖZET………………………………………………………...………………………...
ABSTRACT…………………………………..……..………………………………...
AKNOWLEDGEMENTS……………………..……………..……………………....
LIST OF ABBREVIATIONS…………………..……………………..……..….…...
LIST OF TABLES………………………………..………………………..………....
LIST OF FIGURES………………………………..………………………..………..
LIST OF APPENDICES……………………………..………………..….……….....
CHAPTER 1
INTRODUCTION
1.0. Introduction ………………………………………………………….…………….
1.1. Background of the Study……………………………………………..……………
1.2. Statement of the Problem…………………………………………..……………....
1.3. Aim and Research Questions……………………………………..…………….….
1.4. Significance of the Study………………………………………..………………...
1.5. Assumptions and Limitations……………………………………..……………….
1.6. Operational Definitions and Key Terms...………………………..…………….….
CHAPTER 2
REVIEW OF LITERATURE
2.0. Introduction………………………………………………………..……..……..…
2.1. Theoretical Framework……………………………………..……………...……....
2.1.1. Interlanguage theory…………………………………..…………………...
2.1.2. Differences between L1 and L2 writing……………..………………….…
2.2. What is Coh-metrix?.…………………………………………....………………….
2.2.1. Readability………………………………………….…..………………….
2.2.2. General Word and Text Information……………….…..………………….
2.2.3. Syntax …………………………………………….…..…………………...
iii
v
vii
xii
xiii
xvi
xvii
1
1
3
4
6
6
7
9
9
10
11
16
16
17
17
xiii
2.2.4. Referential and Semantic Aspects……………………..………….……….
2.2.5. What is Latent Semantic Analysis?……….….…..…………….………….
2.2.6. Situation Model Dimensions…………………………..………….……….
2.3. Interlanguage Lexicon………………………………………..……………………
2.3.1. Breadth and Depth of Vocabulary Knowledge…………………………....
2.3.2. Lexical Networks……………………………………………………….....
2.3.3. Lexical Cohesion…………………………………………………….…….
2.4. Corpus Linguistics………………………………………………….…………..…..
2.4.1. Learner Corpora: A Revolution……………………………………...…….
2.4.2. Corpus Linguistics and Language Pedagogy………………………..…….
2.4.3. Corpus Linguistics and Vocabulary Teaching/Learning……………....…...
2.5. Data-driven learning…………………………………………………………...…...
2.5.1. Criticisms and Downsides of DDL……………………………….…….…
2.6. Using Concordancers in Language Teaching/Learning…………………………....
2.6.1. Learners’ Attitudes towards Corpora as a Language Learning Tool………
2.7. Summary………………………………………………………………….......……
CHAPTER 3
METHODOLOGY
3.0. Introduction……………………………………………………..…………………
3.1. The Design of the Study…………………………………………………..……….
3.2. Sampling and the Participants…………………………….………………..………
3.3. Context of the students……………………………………………………..……...
3.4. Data Collection Tools and Procedures………………………………………..…...
3.4.1. Descriptive Procedures………………………………………….................
3.4.2. The Selection of Corpora………………………………..……..………….
3.4.3. Semi-Structured Focus Group Interviews……………….……...................
3.4.4. Experimental Procedures…………………………………..……...……….
3.4.5. Piloting of the Multiple Choice Test for the Target Vocabulary…….……
3.4.6. Pre-test for Written Production…………….……..……………..………...
3.4.7. Vocabulary Teaching Material…………………..…………...…................
3.4.8. Classroom Procedures………………………….…...……………………..
18
19
21
22
23
25
27
32
34
37
41
42
43
45
50
51
53
53
55
57
58
58
58
62
63
63
65
66
68
xiv
3.4.9. LSA Scores………………………………...……..………………………..
3.4.10. Target Vocabulary Density Scores………………..………………...……..
3.4.11. Semi-structured Interviews………………………..……..………………..
3.5. Overview of the Descriptive and Experimental Procedures…..…………………...
3.6. Overview of the Statistical Techniques Used in the Study…..……………………
CHAPTER 4
RESULTS
4.0. Introduction……………………………………………………..…………………
4.1. Results of the Descriptive Phase………………………………..………………….
4.1.1. Normality and homogeneity of the data………………..……………..…….
4.1.2. Referential and Semantic Indices……………………..……...……………...
4.1.3. Readability Indices…………………………………..…………..………….
4.1.4. Syntax Indices……………………………………..………..………………
4.1.5. Results of the Focus Group Interviews……………..………..……...……...
4.2. Results of the Experimental Phase…………………………..……………………..
4.2.1. Experimental Results for Recognition…………………..…..……................
4.2.2. Confirmation of the Descriptive Results for the Experimental Phase……....
4.2.3. Target Vocabulary Density Scores…………..…………...………………….
4.2.4. Experimental Results for Lexical Cohesion……..…………..……………...
4.2.5. Semi-structured Interview Results……………..………...…………….……
4.3. Overview of the Descriptive Results……………………..………………………..
4.4. Overview of the Experimental Results…………………….……………................
4.5. Summary………………………………………………..………………………….
CHAPTER 5
DISCUSSION AND CONCLUSIONS
5.0. Introduction……………………………………………..……………………..…..
5.1. General Summary of the Study………………………..…………………………..
5.2. Review of the Findings in Relation to the Research Questions…..………………..
5.3. Implications of the Study………………………………………..………................
70
70
72
72
73
75
75
76
79
93
96
103
106
106
109
110
114
121
125
128
128
131
131
134
139
xv
5.4. Recommendations for Further Research……………………..…………………….
5.5. Personal Reflections and Criticism of the Study……………..……………………
REFERENCES………………………………..……………...…………….…………
APPENDICES ……………………………………………………..………………….
CURRICULUM VITAE ………………………………………..……………………
140
141
142
150
172
xvi
LIST OF ABBREVIATIONS
AWL : Academic Word list
DDL : Data Driven Learning
EFL : English as a Foreign Language
ELT : English Language Teaching
ESP : English for Specific Purposes
L1 : Native Language
L2 : Foreign Language
LC : Learner Corpora
LOCNESS : Louvain Corpus of Native English Essays
LSA : Latent Semantic Analysis
NC : Native Corpora
SLA : Second Language Acquisition
SVD : Singular Value Decomposition
TL : Target Language
xvii
LIST OF TABLES
Table 1: A Sample Term-document Matrix……………………...…………………......
Table 2: Research Paradigm and Processes for Each Stage of the Study…....................
Table 3: Demographic Data about the Descriptive Phase of the Study…...……………
Table 4: Demographic Data about the Experimental Phase of the Study…....................
Table 5: Comparison of Three Corpora Used in the First Phase of the Study………....
Table 6: Coh-metrix Indices Used in the Current Study………………...…....................
Table 7: Item Analyses for the Target Vocabulary Test Used as the Pre-test and the
Post-test………………….………………......................................................
Table 8: Results of the Pilot Study for the Target Vocabulary Test………..…..……....
Table 9: Summary of the Procedures Followed throughout the Study......………….....
Table 10: Normality and Homogeneity Results for the Referential & Semantic and
Readability Indices…...……………………………………………………….
Table 11: Normality and Homogeneity Results for the Syntax Indices….......................
Table 12: Descriptive Results for Anaphor Reference for Adjacent Sentences….........
Table 13: Kruskal Wallis Test Results for Anaphor Reference for Adjacent Sentences.
Table 14: Descriptive Results for All-distance Anaphor References.…………………..
Table 15: Kruskal Wallis Test Results for Anaphor References…………………….….
Table 16: Descriptive Results for Argument Overlap for Adjacent Sentences………..
Table 17: Kruskal Wallis Test Results for Argument Overlap for Adjacent Sentences...
Table 18: Descriptive Results for All-distance Argument Overlap…..……....................
Table 19: Kruskal Wallis Test Results for All Argument Overlap……...…....................
Table 20: Descriptive Results for Adjacent Stem Overlap……...…...………………….
Table 21: Kruskal Wallis Test Results For Adjacent Stem Overlap…..……...................
Table 22: Descriptive Results for All-distance Stem Overlap…………..…....................
Table 23: Kruskal Wallis Test Results for All-Distances Stem Overlap……..................
Table 24: Descriptive Results for LSA Scores for Adjacent Sentences.…......................
Table 25: One-way ANOVA Results for Adjacent LSA Scores……...………………...
Table 26: Descriptive Results for All-distance LSA Scores……...…………..................
Table 27: One-way ANOVA and Scheffe Test Results for All-distance LSA Scores.....
Table 28: Descriptive Results for Flesch Reading Ease Scores…..……………………..
Table 29: Kruskal Wallis Test Results for the Flesch Reading Ease Scores……............
Page
20
54
56
57
59
60
64
65
74
77
79
79
80
80
81
82
82
83
83
85
86
87
88
89
89
90
90
94
94
xviii
Table 30: Descriptive Results for the Flesch-Kincaid Reading Ease Scores…….....…..
Table 31: Kruskal Wallis Test Results for Flesch Kincaid Reading Ease Scores……...
Table 32: Descriptive Results for the Personal Pronoun Incidence Scores……………..
Table 33: Mann Whitney U-test results for the Personal Pronoun Scores..….................
Table 34: Descriptive Results for the Incidence of All Connectives Scores ..................
Table 35: Mann Whitney U-test results for All Connectives Scores….……………….
Table 36: Descriptive Results for Type-token Ratio Scores………...…...……...………
Table 37: Mann Whitney U-test for Type-token Ratio Scores …..…………......……....
Table 38: Descriptive Results for the Number of Words before the Main Verb..……...
Table 39: Mann Whitney U-test for the Number of Words Before the Main Word…….
Table 40: Thematic Chart for the Focus Group Interviews and Theme Frequencies…...
Table 41: Normality and Homogeneity Test Results for AWL Pre-test and Post-test
Scores……………………...…….……..……………………………………..
Table 42: Mean and Corrected Mean Scores of the Experimental and Control Groups
for the Vocabulary Recognition Test…………...……………………………..
Table 43: ANCOVA Results for Pre-test & Post-test……………...…………………....
Table 44: T-test Results of the Learner and Native Groups for Adjacent Sentences LSA
Scores…..……………..…………………………………………..…………...
Table 45: T-test Results of the Learner and the Native Group for All-distance LSA
Scores…………………………………………..…………………..................
Table 46: Post-test Mean and Corrected Mean Scores of the Experimental and Control
Groups for Target Vocabulary Density Scores…………......………………...
Table 47: ANCOVA Test Results Comparing the Experimental and Control Groups for
Target Vocabulary Density Scores (Post-test)…..………………….………...
Table 48: Mean and Corrected Mean Scores for Target Vocabulary Density for the
Experimental and Control Groups (Delayed Post-test)….......……………….
Table 49: ANCOVA Test Results Comparing the Experimental and Control Groups
for Target Vocabulary Density Scores (Delayed Post-test)…………..............
Table 50: LSA Mean and Corrected Mean Scores for Adjacent Sentences for the
Experimental and Control Groups (Post-test)………….……...…...................
Table 51: ANCOVA Test Results for LSA scores for Adjacent Sentences of the
Experimental and Control Groups (Post-test)………………………………...
95
95
97
97
98
98
99
100
100
101
104
106
108
108
109
109
112
112
113
113
115
116
xix
Table 52: Adjacent Sentence LSA Mean scores and Corrected Means for the
Experimental and Control Groups (Delayed Post-test)……...……………….
Table 53: ANCOVA Test Results for LSA scores for Adjacent Sentences of the
Experimental and Control Groups (Delayed Post-test)………….…………...
Table 54: All-distance LSA Mean and Corrected Mean Scores of the Experimental and
Control Groups (Post-test)…………………………………………………….
Table 55: ANCOVA Test Results for All-distance LSA scores of the Experimental and
Control Groups (Post-test)……………………..……..……………………….
Table 56: All-distance LSA Mean and Corrected Mean Scores for the Experimental and
Control Groups (Delayed-test)……………………………………………….
Table 57: ANCOVA Test Results for All-distance LSA Scores of the Experimental and
Control Groups (Delayed-test)…………………………..……………………
Table 58: Semi-structured Interview Thematic Chart and Theme Frequencies……..….
Table 59: Descriptive Results for the Interview Questions about the Effects of
Concordancing Activities on Vocabulary Recognition and Production...........
116
117
118
119
120
120
121
124
xx
LIST OF FIGURES
Page
Figure 1: A General Outline of Literature Review……………….……………….…….
Figure 2: A Toy Network Constructed with Four Sentences….……...……....................
Figure 3: Parameters to be added to a Learner Corpus……….…..…………..................
Figure 4: Text Coverage in a 10 Million-word Corpus of Spoken
and Written English……………………………………….……......................
Figure 5: A Concordancing Screenshot for the Word ‘example’…..……………………
Figure 6: A Sample Concordancing Activity to Teach Vocabulary…...……..................
Figure 7: Coh-metrix Input Screen………………………………..….………………….
Figure 8: Sample Coh-metrix Results for a Text Input…………………………………
Figure 9: Input Screen for the Target Vocabulary Items……………..…………………
Figure 10: Concordancing Screen for the Target Vocabulary Item (Not Gapped)….....
Figure 11: Concordancing Screen for the Target Vocabulary Item (Gapped)….............
Figure 12: Word Ranges for Both Sides of the Target Vocabulary Item…….................
Figure 13: Immediate and All Potential Collocations for the Target Vocabulary Item…
Figure 14: The AWL Highlighter Input Screen…………………….…...........................
Figure 15: The Output Screen for the AWL Count…………………………..................
Figure 16: Pre-test & Post-test Mean Scores for Vocabulary Recognition for the
Experimental and Control Groups…………………………………………...
Figure 17: AWL Density Scores Measured as Number of Types per 1,000 Words…….
Figure 18: The Change in AWL Density Scores of the Experimental and Control
Groups from Pre-test to Delayed Post-test……..............................................
Figure 19: Pre-test, Post-test and Delayed Post-test LSA (Adjacent Sentences) Mean
Scores for the Experimental and Control Groups in Comparison with the
Native Group…………………………………………..…………………….
Figure 20: Pre-test, Post-test and Delayed Post-test LSA (All-distance) Mean Scores for
the Experimental and Control Group in Comparison with the Native Group.
9
26
35
41
45
50
61
61
66
66
67
67
68
71
71
107
110
111
114
118
xxi
LIST OF APPENDICES
Page
APPENDIX 1. Indices of Coh-Metrix ........................................................................ 150
APPENDIX 2. Academic Vocabulary Scan Test ....................................................... 152
APPENDIX 3. Argumentative Essay Topics for the Written Tests ........................... 155
APPENDIX 4. A Sample Concordancing Activity .................................................... 156
APPENDIX 5. Sample Learner Essays ...................................................................... 159
APPENDIX 6. Experimental Group Interview Questions .......................................... 161
APPENDIX 7. Academic Vocabulary Load Comparisons ......................................... 162
CHAPTER 1
INTRODUCTION
1.0. Introduction
Theoretically speaking, with our current understanding of human mind it is
nearly impossible to know what exactly is going on in language learners’ minds while
they are trying to deal with the language they are learning. With the help of
neuroimaging techniques, it might of course be possible to get a glimpse of the real
processes taking place during language learning. When technical and ethical points are
taken into consideration, however, it is not as easy as it may sound.
There are much practical ways to get involved in the enigmatic language
learning process, and one of them is by trying to analyze learner language. Learners’
written or spoken productions are valuable data as to their cognitive issues concerning
both their native and target languages. These productions are not monolithic but rather
highly variable (Ellis & Barkhuizen, 2005, p. 4); and through their analyses,
information flow in discourse, syntactic, lexical and rhetoric features in L2 texts
(Hinkel, 2005, p. 615) could be better understood. One of the most effective ways to
understand them is to compare L1 and L2 written productions.
Among the variables mentioned above, lexical features of L2 texts is an area
which has been traditionally overlooked (Meara, 2002). Furthermore, in the related
literature it has been claimed that lexical features alone could not be used to
discriminate L1 and L2 texts (Connor, 1984; Reynolds, 1995, cited in Crossley and
McNamara, 2009). However, with the advance of computational linguistics, analysis of
EFL learner’s lexicon is within reach.
1.1. Background of the Study
Second language teaching and learning (SLA/SLL) is an expanding field with
newly emerging sub-fields. This domain is, in fact, a multi-disciplinary one which
gleans insights and methods from a range of disciplines such as linguistics, sociology,
sociolinguistics, psychology, psycholinguistics and education (Ellis & Barkhuizen,
2
2005, p. 3). Naturally, being multidisciplinary comes with rapid developments, and
these developments are, most of the time, on a par with new technology.
In theory, research possibilities in SLA are vast; however, much of SLA
research has traditionally focused on describing learner language or learners’
interlanguage; their sequences and patterns of development have been the focal point in
these studies (Pica, 2005, p. 265). Coined by Selinker in 1972 the term interlanguage, or
learner language, could be defined as the interim stage between a learner’s native
language and the target language s/he is trying to learn.
The characteristics of learner language have been researched from numerous
aspects. With contrastive analysis as a paradigm, this stage has been analyzed for lexical
and grammatical errors since 1960’s, and to some the assumption was that these errors
stemmed from an interference of L1 in L2 acquisition process. Contrastive rhetoric,
whereby discourse features of L2 is examined, has also been among the research topics
in SLA. The outcomes of these studies have been discussed, analyzed, confirmed or
denied by researchers; however, some aspects of learner language have been ignored.
Among these aspects, lexical cohesion in learner language is a potentially fruitful one.
In language teaching-learning context, trying to deal with cohesion in learners’ texts is
like sailing into uncharted waters; traditionally, it lacks attention (Cook, 1989, p. 127;
Flowerdew, 2009, p. 85). This lack of attention seems to be noteworthy, as the use of
lexical cohesive ties has been reported to be a significant differentiating factor between
native and non-native speaker writing (Connor, 1984). Although trying to deal with
cohesion in EFL context is both a relatively rare and a definitely problematic issue,
there are methods for analyzing cohesion in learner language by making use of corpus
linguistics.
With ever-growing technological facilities at hand, corpus linguistics has been
practically supplying language teachers and learners with substantial amount of
authentic language samples for over forty years now. In language teaching, the focus
has traditionally been on native speaker corpora (Leech 1998; McEnery &Wilson 1997).
Another tradition is that the practical use of these corpora in language classrooms has
had two aspects. The first application concerns the native-corpus centered dictionaries
or applications, a typical of which is the COBUILD project. Through this project,
learners of English have had the chance of encountering the authentic use of language.
3
The second application is integrating these corpora directly into teaching materials by
language teachers or by students themselves through comparisons and analyses.
Discussions about the applicability of corpus linguistics in EFL have been
modified since the systematic collection of learner corpora was introduced to the field
through an international project by Sylviane Granger during 1990’s. By using strict
criteria, her team started collecting data from high intermediate and advanced learners
of English. The outcomes gleaned from these corpora have sometimes been used to
produce software to be used in ELT classes (Milton, 1998). The underlying rationale
has been to provide the learners with negative evidence and show them typical mistakes
the learners make. Obviously, the active paradigm has been the focus on learners’
mistakes. This paradigm might still be useful in many ways, but making learners aware
of their mistakes does not seem to be enough.
1.2. Statement of the Problem
The problem that the current study is trying to deal with relates to freshman
engineering students learning English as a foreign language at the Higher School of
Foreign Languages at University of Gaziantep. Written productions of EFL learners is a
topic of concern for the instructors in this department. It is a common view that there is
something missing in their essays other than grammar mistakes or errors.
In this institution, throughout years, language teaching has been modified,
modernized and eventually relatively improved. However, lexical cohesion in learners’
written productions is nowhere near adequate. Although discussions concerning the
issue go on continuously, let alone trying to come up with feasible solutions, the problems
have not been named yet.
The common view among the teaching staff at this institution is that engineering
students do better in grammar subjects, but when it comes to learning and retaining new
vocabulary items and using them appropriately the teaching/learning process falters.
This topic is an ongoing one in teachers’ rooms. When the learners are asked to talk
about their problems they encounter while learning a second language, the very same
topic surfaces. When the written productions of these learners are examined, which is
done officially during mid-term and final exams, teachers’ observations concerning
lexical cohesion are confirmed.
4
As a practitioner, trying to teach vocabulary and help learners retain what they
learn has been among my concerns. Learners’ constant complaints as to the fuzziness
and shakiness of vocabulary items being learned have directed me to find ways to deal
with the problem. When encountered with new words to learn, learners try to take notes
where these words have a couple of simplistic matching right next to them. When the
learners see these words in new contexts with different meanings, the problems begin.
Applied corpus linguistics was one of my resorts to help students tackle this problem.
The rationale behind concordancing, which is to expose lexical items in their own
surroundings in different contexts and in relation with other words, seemed to facilitate
learning new vocabulary items and retaining them. Because with the help of this
exposure, learners would be able to analyze the new words they are trying to learn in
several contexts and from the very beginning; and when it comes to writing, they would
be able to produce lexically more cohesive essays. As a matter of fact these
interpretations have stemmed from observations and intuitions. In order to validate
these interpretations and intuitions, examining learners’ noticeable inadequacy in lexical
cohesion in their written productions through a systematical approach will be the main
concern of the current study.
1.3. Aim and Research Questions
The primary aim of this study is to determine quantifiable differences between
the written productions of EFL learners and native speakers of English. These
quantifiable aspects include readability of these productions and semantic, referential
and syntactic issues present in them. The focal point will be on determining lexical and
cohesive differences between texts written by Turkish EFL learners and texts written by
native speakers of English. These differences are expected to shed light onto the lexical
and cohesive flaws in learners’ texts.
Another aim of the current study is to help students overcome these flaws by
making use of concordancing activities. The expectation is that through these activities
learners will gain insights as regards to the ties that exist among lexical items in a
context. These insights are expected to help learners internalize and better recall the
lexical items being learned, and write more cohesive essays.
This quasi-experimental study has two aspects in this respect; one being
theoretical and the other being practical. From theoretical point of view, this study could
5
be regarded as an attempt to answer certain questions and raise new ones concerning
written productions of Turkish EFL learners bringing the interlangual lexicon and
cohesion in their texts into the foreground. These questions are, however, context-bound
i.e. they are limited to a certain teaching/learning context. The rationale behind this
paradigm is that every learner, every teacher and every teaching/learning context is
unique (Brown, 2007, p. 18); so the problems surfacing in any context need to be
handled by taking into account the parameters in the same context. From a practical
point of view, trying to make use of what corpus linguistics has to offer to SLA will
also fall within the scope of this study.
Taking the related literature into account the following research questions are the
main concerns of this study:
1. Regardless of prompt or average number of words used in the texts, to what
extent do texts written by Turkish EFL learners deviate from texts written
by native speakers of English in terms of referential and semantic aspects?
2. Regardless of prompt or average number of words used in the texts, to what
extent do texts written by Turkish EFL learners deviate from texts written
by native speakers of English in terms of:
a) Flesch Reading Ease Score
b) Flesch-Kincaid Reading Ease Score
3. With average number of words per text being similar, to what extent do
texts written by Turkish EFL learners deviate from texts written by native
speakers of English in terms of syntactic features?
4. What are learners’ perceptions and feelings about the differences between
their written productions and that of native ones?
5. Can concordancing activities induce vocabulary recognition?
6. Can concordancing activities induce vocabulary production?
7. Can concordancing activities induce lexical cohesion in EFL learners’
written productions?
8. What are learners’ perceptions and feelings about the use of concordancing
activities to learn vocabulary?
6
Among these questions, the first four concern the descriptive phase of the study;
while the last four are related to the experimental phase.
1.4. Significance of the Study
This study could be regarded as an important attempt to determine the
differences between L1 and L2 writings at different levels. It is important because
Turkish EFL learners’ written productions have not been subject to comparisons such as
readability, referential, semantic and syntactic aspects. Among these aspects, lexical
errors of language learners are regarded as global errors (Ellis, 1995; Gass & Selinker,
2008, p. 449), which means that these errors cause communication breakdowns. The
main importance of the current study is that it is an attempt to systematically determine
and solve problems concerning Turkish EFL learners’ lexical networks; as a matter of
fact, this alone is an end itself.
As it was mentioned previously, generally speaking lexical cohesion in learner
language is not a no man’s land completely, but more studies are needed about the
issue. When we look at the issue with Turkish EFL learners in mind, this time we
definitely have a no man’s land in front of us. Bearing in mind the lack of studies
concerning cohesion in Turkish EFL learners, this study could be regarded as the first
attempt to deal with the problem.
The study at issue is expected to be quite lucrative for the researcher and maybe
for the colleagues working in the same department in terms of professional/personal
development.
1.5. Assumptions and Limitations
First of all, it is assumed that individual differences among the subjects
participated in the current study, such as socio-economic and cultural backgrounds will
not have significant effects on the statistical outcomes.
The second assumption relates to the computational processes that were carried
out in Coh-metrix. Computerized analyses of L2 essays in large scale assessments like
the Test of English are reported to have misidentified L2 textual features with an error
ratio of 21 % (Frase, Faletti, Ginther, & Grant, 1999). The related assumption is that the
online tool, Coh-metrix, which was used in both the descriptive and the experimental
parts of this study, yields reliable measurements concerning both L1 and L2 corpora.
7
Another limitation concerning Coh-metrix is that all the calculations concerning
referential & semantic, readability and syntactic aspects are limited with the indices
which are present in Coh-metrix. That is to say, syntactic issues, which are far more
complex than just mathematical calculations, will be dealt with to the extent that Coh-
metrix allows.
In this study, a collection of texts written by native speakers of English is used
as the reference point for comparison with Turkish EFL learners. The assumption, albeit
a strong one, concerning this point, is that the reference corpus collected from native
speakers is flawless in terms of lexicon and grammar, which would mean that the more
native-like a learner text is, the more coherent it is.
As for the limitations, both descriptive and experimental results of the current
study are limited to Turkish EFL learners whose proficiency levels vary from
intermediate to upper-intermediate. Furthermore, the number of the learners is too
limited for broader generalizations.
Another limitation of the study concerns the delayed effects of the experimental
phase. Although the learners were tested after two weeks from the last activity of the
treatment, there might be certain delayed effects of the implementations, which is
beyond the limitations of the present study.
1.6. Operational Definitions and Key Terms
Corpus/corpora: The term corpus (pl. corpora) refers to bodies of systematically
collected and digitalized texts created from written or spoken language.
Coh-metrix: Coh-metrix is an online tool which can make textual calculations at
multiple levels by applying numerous measures, and it is freely available at
cohmetrix.memphis.edu. See Appendix 1 for the indices used in this database.
Concordance: Basically, it is a list of words taken from the same context. Lexical items
in a language rarely occur randomly in a context; certain words tend to company other
certain words, which creates a kind of harmony among them. As Sinclair (1991, p. 32)
puts it:
A concordance is a collection of the occurrences of a word-form, each in its own textual environment. In its simplest form it is an index. Each word-form is indexed and a reference is given to the place of occurrence in a text.
8
Flesch-Kincaid & Flesch Reading Ease Scores: These reading ease scores are results
of mathematical calculations and they are indications of how easy it is to understand a
text written in English. Scores vary between 0-12 in Flesch-Kincaid score calculations,
and the texts obtain scores ranging from 0 to 100 in Flesch score calculations. A higher
Flesch-Kinkaid score means that the text is hard to read, while lower Flesch reading
scores indicate difficulty.
Latent Semantic Analysis (LSA): LSA is a mathematical technique used for
evaluation of textual cohesion. It uses Singular Value Decomposition (SVD), a
mathematical matrix decomposition technique which is used to reduce thousands of
dimensions and relationships between words to a more manageable number.
Lexical networks: The itemized picture of vocabulary items in human mind is hard to
justify; words in our long-term memories are not stored alphabetically or in a
dictionary-like manner. Words are connected to each other in small clusters, and these
clusters form bigger ones and lexicons are formed based on shared connections and this
system is called lexical networks (Crossley & McNamara, 2009).
Referential and semantic aspects of texts: Referential and semantic aspects of texts
are used together in the current study as they are directly related to textual cohesion. A
number of referential structures such as anaphoric references, stem or argument
overlaps are taken into account, and semantic analysis will be performed in terms of
Latent Semantic Analysis.
Syntactic features of texts: In the context of the current study syntactic features of a
given text are limited to syntactic parameters present in Coh-metrix. These parameters
include personal pronoun counts, pronoun ratios, type-token ratios, syntactic structure
similarities for both adjacent and all-distance sentences, connectives (additive,
temporal, causal and logical connectives), logic and conditional operators, negations,
noun phrase incidence scores, modifiers per noun phrase, higher level constituents and
the mean number of words before the main verb of main clause in sentences.
2.0
an
dif
int
co
lite
ins
2.1
ide
als
me
illu
Fig
I
0. Introd
In this
nd then issu
fferentiating
terlanguage
rpora in lan
erature con
struction is
1. Theor
The cu
entify probl
so relates to
ethodology
ustrated in F
gure 1. A g
Interlangua
DiffeL2 w
Lexic
betwe
duction
section, fir
ues related
g features
e are discus
nguage teac
ncerning EF
summarized
retical Fram
urrent study
lematic aspe
o corpus lin
(Leech, 19
Figure 1.
general outli
ge Theory
erences betweewriting
cal differences
een L1 & L2 w
REVIE
rst the theor
d to interla
of L2 and
ssed. Some
ching/learnin
FL learners’
d.
mework
y directly r
ects in writ
nguistics as
992, p. 115
ine of literat
en L1 &
s
writing
Theoret
CHAPTE
EW OF LIT
retical fram
anguage the
d L1 writin
insights as
ng contexts
’ attitudes
relates to th
tten product
s it is the m
5). The out
ture review
Corpus
tical Framew
ER 2
TERATUR
mework of t
eory are b
ng are det
s to data-dr
s are given n
towards the
he interlang
tions of Tur
main parad
tline of thi
Linguistics
Comparison corpora
Use of concolanguage tea
work
RE
he current
riefly over
ailed, the
riven learni
next. As the
e use of co
guage theory
rkish EFL s
digm of the
s study’s li
s
of L1 & L2
ordancing in aching
study is pre
rviewed. A
lexical asp
ing and the
e last point,
orpora in la
y in that it
students. Th
study in t
iterature re
9
esented,
fter the
pects of
e use of
, related
anguage
tries to
he study
erms of
eview is
10
2.1.1. Interlanguage theory
Coined by Larry Selinker in 1972 (Ellis & Barkhuizen, 2005, p. 54), the term
interlanguage has been used to refer to an interim stage where language learners are
close to the target language and not far away from their native one, thus making it look
like both and neither at the same time. This approximation stage is, most of the time,
appears as a system. Ellis & Barkhuizen (ibid., p. 54 ) characterize this stage as follows:
A learner’s interlanguage
knowledge constitutes a system;
consists primarily of implicit linguistic knowledge;
is permeable;
is transitional;
is variable;
is the product of multiple interacting factors;
and may fossilize.
This interim stage appears as a system because the errors at this stage are not
random but systematic; this stage has its own structures even for the errors. It contains
certain elements that neither the target language (TL) nor the native language (NL) has.
Learners’ accurate productions (written or spoken) as well as their errors are analyzed to
make sense of this stage. For example, the overuse or underuse of certain structures by
the learners can be attributed to neither TL nor NL.
Language learners’ linguistic repertoire is mostly based on implicit knowledge.
As the learners are exposed to TL, the sequences are stored as chunks to be used
automatically when needed. Sometimes this situation appears as a formulaic utterance
(e.g. Nice to meet you.), or a collocational one (e.g. quit + Ving, I want to quit smoking.).
Although language learners can sometimes use explicit or declarative forms of language
knowledge, the knowledge orienting their competence is generally accepted to be
implicit.
Learner language is also permeable. That is, there is no stability or a unity in the
system being built by the learner. This stage is vulnerable to new linguistic forms and
rules, and these can be achieved either internally, by referring to L1, or externally,
11
through exposure to TL. This permeability also brings a transitional characteristic to this
stage, which means that this stage is constantly revised and modified.
At any stage, the learners might use different forms for the same grammatical
structures. There is an extent of variability that learners can employ, but these
possibilities are still, most of the time, easy to predict.
During this interim stage, learners might also make use of general learning or
communication strategies. For example, they might overgeneralize or oversimplify
certain structures as a general learning strategy, or they might employ general
communication strategies like paraphrasing while trying to ask a question.
Learner language can also be subject to fossilization. In other words, learners
may stop advancing further before achieving a native like grammar. Why and how it
happens is still a discussion topic (see Long, 2003 for an overview), however, the
process appears to be in line with the sensitive age hypothesis which states that around
puberty human beings lose their capacity to acquire languages.
Although it might sometimes sound and appear to be solid and simple, this stage
is actually full of unresolved issues such as stabilizing, fossilization, backsliding (Long,
2003), language transfer, interlanguage pragmatics (Mitchell & Myles, 1998) and
interlanguage lexicon (Nation, 1990). As it was mentioned before, among these issues
interlanguage lexicon has been traditionally overlooked (Meara, 2002).
2.1.2. Differences between L1 and L2 writing
Hinkel (2005, p. 617) distinguishes between L1 and L2 writing in terms of micro
and macro features. She refers to macro features as global aspects of texts such as
discourse construction, arrangements of ideas, cohesion, and coherence. A description
of micro features is given as textual features that have the function of marking discourse
organization and aiding in the development of cohesive and coherent prose.
Understanding both macro and micro features of L2 texts is important because if
significant differences exist between L1 and L2 written productions, ESL practitioners
need to have a clear understanding of these differences (Silva, 1993) in order to be able
to form intelligent contrasts between adopting and/or adapting L1 practices. In his
seminal meta-analysis, Silva (1993) screened and analyzed 72 empirical reports
involving a direct comparison of L1 and L2 written productions in order to develop a
12
clear understanding of the nature of L2 writing. The subjects in his study came from
different language backgrounds including Arabic, Chinese, Japanese and Spanish as
L1s. They were predominantly undergraduate students in their late teens or early
twenties with fairly advanced English proficiency levels. The reports involving these
subjects were compared in terms of fluency, accuracy, quality, structure,
morphosyntactic/stylistic and lexicosemantic features. The results suggested that, in
general, adult L2 writing is distinct from and less effective than L1 writing. L2
composing appears to be more constrained, more difficult and less effective. L2 writers
appeared to be doing less planning and having problems with setting goals, as well as
generating and organizing materials. Their transcribing was more laborious, less fluent,
and less productive. Reviewing, rereading and reflecting were less common, but they
revised more. Naturally, they were less fluent and less accurate. In terms of lower-level
linguistic concerns, L2 writers’ texts were stylistically distinct and simpler in structure.
Their sentences included more but shorter t-units, fewer but longer clauses, more
coordination, less subordination, less noun modification, and less passivization. One
important point regarded the use of cohesive devices. They used more conjunctive and
fewer lexical ties, and exhibited less lexical control, variety and sophistication.
Another similar and important study was carried out by Ferris (1994). A corpus
of 160 ESL texts was analyzed. There were 40 texts each by students from four L1
groups: Arabic, Chinese, Japanese and Spanish. The papers were from a university
placement exam in which they were asked to write about culture shock. 62 quantitative,
lexical and syntactic features of the text were identified and counted in the corpus. For
the purpose of statistical analysis, some of these features were either dropped or
combined leaving 28. Some of these features were; number of words, impersonal
pronouns, modals, negation, coordination, coherence features and repetition. The groups
were divided into two; one of the groups consisted of learners at a lower level of
proficiency in English and the other group consisted of advanced learners of English. A
discriminant analysis was performed to see how the mentioned variables would contrast
the two groups. The results revealed that students at higher levels of L2 proficiency
used a variety of lexical choices, syntactic constructions, and cohesive devices, and their
texts received higher holistic scores. The study also showed that micro-level attention
and instruction might be of more significance than many practitioners realized.
13
In her comprehensive study, Hinkel (2001b) compared native English speakers
with speakers of Chinese, Japanese, Korean, Vietnamese, and Indonesian in terms of the
frequency rates of overt exemplification markers in essay texts, listed in full, (as) an
example, for example, for instance, in (my/our/his/her/their) example, like, mainly,
namely, such as ..., that is (to say). 1,087 students’ essays were analyzed via non-
parametric statistical techniques. The analysis of the data from student essays showed
that NNSs employed far more example markers (conjunctions), first person pronouns,
and past tense verbs in their academic texts than NSs did. The overuse of personal
pronouns at this point is noteworthy; Biber (1995), after the analyses of large English-
language corpora, points out that first person pronouns serve as markers of interpersonal
discourse and direct involvement of the writer, and they are usually more characteristic
of spoken rather than written registers. From a pragmatic point of view, according to
Hvitfeldt (1992), the idea of truth results from everyday experience, and personal
examples can be just as valid as the information obtained from literary sources, which
may be why EFL learners make use of personal pronouns more than necessary to
consolidate their truths. In line with this insight Hinkel (ibid.) concludes that speakers
of Chinese, Japanese, Korean, Vietnamese, and Indonesian who have completed their
training in ESL and writing courses rely on accounts of personal experiences and stories
as a means of thesis support in formal essays significantly more frequently than NS
students do.
Hinkel (2002) carried out another large scale empirical analysis of 68 lexical,
syntactic and rhetorical features of L2 text. The corpus included texts written by
advanced learners of English from six different languages: Arabic, Chinese, Indonesian,
Japanese, Korean and Vietnamese. According to Hinkel, even after years of study in
English, the learners still lack some aspects that native speakers have. The results of her
study indicate that L2 writers have a severely limited lexical and syntactic repertoire.
This led the learners to produce simplistic texts which are rooted in conversational
discourse in English language. The results reveal that there appears to be a big gap
between L1 and L2 texts in terms of basic academic writing. Bridging the gap will
require alternative pedagogical methodologies for teaching writing.
Regarding connectives, Schleppegrell (1996) analyzed ESL writers use
strategies for conjunctions that are typical of spoken English, the focal example of her
study being the conjunction because. The students were mainly Asian immigrants who
14
had lived in the US for different lengths of time and who were considered to be at an
advanced level of English proficiency. She analyzed the essays written by these students
to identify clauses introduced by because and the functions of those clauses. She
discovered that the ESL learners in her study used the conjunction because more
frequently than native speakers of English do. Furthermore, a parallelism between uses
of because clauses in spoken English and ESL writing was detected, which was
interpreted as an indication of how ESL writers draw on spoken registers
inappropriately in constructing their academic essays.
The studies mentioned up to this point are all related to non-western Asian
languages predominantly Asia. In his study, Tankó (2004) built an original corpus to
analyze learner writing consisting of 93 argumentative essays written in an examination
environment by second and third year students attending a five-year degree course in
English. The participants were all native speakers of Hungarian with ages between 20
and 24 years old. In the exam, they were expected to write formal English texts of
approximately 500 words. Their essays were compared with a reference native corpus,
and the results of the analysis showed a comparative overuse of adverbial connectors in
their essays. The explanation for this was that the Hungarian language does not require
the overt marking of relations between linguistic units of the text. This difference is
thought to have influenced the teachers of academic English writing to put more
emphasis on the explicit teaching of adverbial connectors to Hungarian students.
In another attempt to analyze connectives in EFL learners’, Altenberg & Tapper
(1998) carried out a study to investigate how advanced Swedish EFL learners use
connectives in argumentative essays in comparison with American university students’
usages. They collected data from the International Corpus of Learner English (ICLE):
the Swedish sub-corpus and the control corpus of American university student essays.
The aim of their study was to examine the use of three types of connectives: adverbial
conjuncts (e.g. therefore, in particular); certain style and content disjuncts (e.g.
actually, indeed); and some lexical discourse markers (e.g. result, compare). The results
of their study revealed that advanced Swedish EFL learners tended to overuse adverbial
connectives compared to their American counterparts, and appeared to be using slightly
more types of connectives than the American students; Swedish learners varied their use
of connectives more than the American students did. The results of their study also
15
revealed that a high frequency of connectives was not an indicator of good writing
quality for either group of student writers.
By taking the related literature into account, in a recent study Hinkel (2011)
summarizes research findings concerning the differences between L1 and L2 up to
present as follows;
Micro Features (Grammar and Vocabulary) of L2 Writing Compared to L1 prose, L2 texts
• exhibit less lexical variety and sophistication;
• contain significantly fewer idiomatic and collocational expressions;
• have smaller lexical density and lexical specificity, and more frequent
vocabulary misuses;
• rely on shorter sentences and clauses (aka T-units) with fewer words per
clause and fewer words (e.g., nouns and modifiers) per verb;
• involve high rates of incomplete or inaccurate sentences (e.g., missing
sentence subjects or verbs, incomplete verb phrases, sentence fragments);
• repeat content words more often (i.e., nouns, verbs, adjectives, and
adverbs);
• provide twice as many simple paraphrases or avoid paraphrasing
altogether with a preponderance of referential pronouns (e.g., this, that, it);
• use shorter words (fewer words with two or more syllables), more
conversational and high frequency words (e.g., good, bad, ask, talk);
• incorporate fewer modifying and descriptive prepositional phrases, as well
as a higher rate of misused prepositions;
• employ less subordination and two to three times more coordination.
L2 texts also employ
• fewer passive constructions;
• fewer lexical (e.g., adjectives and adverbs) and syntactic modifiers (e.g.,
subordinate clauses) of sentences, nouns, and verbs;
• inconsistent uses of verb tenses;
• more emotive and private verbs (e.g., believe, feel, think);
16
• significantly higher rates of personal pronouns (e.g., I, we, he) and lower
rates of impersonal/referential pronouns (e.g., it, this, one);
• markedly fewer of abstract and interpretive nouns, and nominalizations
(e.g., rotation, cognition, analysis);
• fewer adverbial modifiers and adverbial clauses;
• fewer epistemic and possibility hedges (e.g., apparently, perhaps) and
more conversational hedges (e.g., sort of, in a way);
• more conversational intensifiers, emphatics, exaggeratives, and
overstatements (e.g., totally, always, huge, for sure);
• fewer downtoners (e.g., almost, hardly);
• more lexical softening devices (e.g., maybe).
The micro level differences between L1 and L2 texts have been investigated in
their different aspects by using different tools. One of these tools, Coh-metrix, is an
online database introduced by Graesser, McNamara, Louwerse, & Cai (2004), which
can analyze texts written in English at multiple levels. This tool is also employed in the
current study as one of the main data collection devices.
2.2. What is Coh-metrix?
Coh-metrix is an online database which can assess texts in English at multiple
levels. While making calculations about texts, it takes into account five indices:
readability scores, general word and text information, syntax, referential and semantic
aspects and situation model dimensions. Each of these indices is composed of several
sub-indices. In this respect, although some counting is done, Coh-metrix is not a word
counter in classical terms. It is highly analytical, and singles out each aspect of a text
from the others, yielding precise numerical values.
2.2.1. Readability
Flesch Reading Ease: This index scores the texts with a scale of 0-100. Higher
scores indicate that the text at hand is easier to read. The following formula is used to
determine a score.
READFRE = 206.835 - (1.015 x ASL) - (84.6 x ASW) where;
17
ASL = average sentence length = the number of words divided by the number of
sentences.
ASW (comes from CELEX database) = average number of syllables per word = the
number of syllables divided by the number of words.
Flesch-Kincaid Grade Level: This reading level formula, more common than the
Flesch Readability formula, calculates the readability scores by converting them to a US
grade-school level. The scores range between 0 and 12, 12 being the hardest level to
read. To obtain a score using this formula, the text at hand should have more than 200
words.
2.2.2. General Word and Text Information
The general word and text information index sets include incidence scores on
word and text units. It also includes the mean values of characteristics of content
words, such as frequency of usage in the English language and concreteness. This index
includes ‘shallow’ parameters like number of words, sentences, paragraphs, and
syllables per word, words per sentence, and sentences per paragraph. Additionally, there
are deeper concepts taken into account like hypernymy and concreteness.
Hypernymy refers to a kind of hierarchy among lexical items i.e. being
subordinates or superordinates. The values are obtained from WordNet, an online
database. For example, the word automobile, compared to vehicle, has a higher
hypernymy value, which makes vehicle more abstract than automobile.
Concreteness is measured by attributing values to lexical items based on their
levels of concreteness. The values range from 100 to 700, and are retrieved from MRC
Psycholinguistics Database (Coltheart, 1981). If a lexical item scores high in this
parameter, it means that the item is concrete rather than abstract.
2.2.3. Syntax
Syntactic aspects include a number of parameters, assessing syntactic
complexity, syntactic composition, and the frequency of particular syntactic classes or
constituents in a text. The syntactic analyses are based on the Charniak syntactic parser.
This index also computes the number of noun-phrases, number of verb-phrases and
constituents per 1000 words. Mean number of words before the main verb, scores
18
concerning connectives and logical operators are other parts of this index. Moreover,
calculations concerning pronouns and type-tokens ratios are made.
Additionally, one of the complex parameters -sentence syntax similarity
parameter- scores sentences based on the syntactic tree structures of adjacent sentences
and across paragraphs.
2.2.4. Referential and Semantic Aspects
Referential and semantic index sets focus on referential cohesion i.e.
coreference. Referential cohesion is generally a matter of the overlapping of
constituents within a text. Argument overlaps and stem overlaps between adjacent
sentences are taken into account in this index.
Argument overlap is a proportion ratio score which calculates the ratio of
sentence pairs sharing one or more arguments (nouns, pronouns etc.). Stem overlap
refers to the proportion of adjacent sentences sharing common word stems. For
example, in the following sentence;
The students prepared their presentations meticulously, that’s why the
preparations took weeks.
Words prepared and preparations share the same stem and are therefore is
called a stem overlap in the database.
Coreference (noun or stem overlaps between adjacent sentences) is one way that
Coh-metrix employs to determine the similarity within a text.
Another way through which Coh-metrix determines similarity is Latent
Semantic Analysis (henceforth LSA). LSA (also known as Latent Semantic Indexing) is
a mathematical, statistical technique for representing world knowledge, based on a large
corpus of texts. It makes use of Singular Value Decomposition (SVD) technique which
could be seen as a type of factor analysis, reducing large corpora of texts to many fewer
dimensions. LSA values from Coh-metrix are taken from the college level TASA
(Touchstone Applied Science Associates, Inc.) corpus (Crossley, Salsbury, McCarthy,
& McNamara 2008).
19
2.2.5. What is Latent Semantic Analysis?
Latent Semantic Analysis is a linguistic theory and method which has been
utilized in natural language processing to determine semantic relationships in large
bodies of corpora. This mathematical technique is mostly used in popular search
engines like Google to categorize and organize large bodies of texts.
LSA makes use of Singular Value Decomposition (henceforth SVD), a
mathematical matrix decomposition technique which is used to reduce thousands of
dimensions and relationships between words to a more manageable number. In this
respect, SVD is akin to factor analysis (Landauer, Foltz, & Laham, 1998, p. 262 ).
Basically, it converts words in a sentence, paragraph or passage into numerical values
by making use of a mathematical technique.
LSA can be applied to sentences, small paragraphs or large bodies of digitalized
texts. As the first step in the process, function words (stop words in terms of
computational linguistics) are eliminated from the text. These are high frequency words
like am, is, are, and, in etc., and to a very large extent, they do not change or relate to
the content of the text at hand. Proper names (words beginning with uppercase) and
abbreviations are also eliminated. For example; when the function words and
abbreviations in the following paragraph are eliminated,
Latent Semantic Analysis (henceforth LSA) is a linguistic theory and
method. It has been used in natural language processing to determine
semantic relationships in large bodies of corpora.
we have only the following lexical items left:
latent semantic analysis linguistic theory method used natural language
processing determine semantic relationships large bodies corpora
In the next step, the lexical items at hand are stemmed. It is a simple process in which
words are reduced to their root forms, called lexemes. The above items will look like
this after stemming:
latent semantic analysis linguistic theory method use nature language
process determine relationship large body corpora
20
A matrix system is constructed with these lexemes by putting them into rows.
This matrix is called a term-document matrix. Each row represents a unique word, and
each column represents the context from which the words are extracted. The context
could be a sentence, paragraph or a whole passage. A sample term-document matrix is
given in Table 2.1.
Table 1
A Sample Term-document Matrix Document 1 Document 2 Document 3 Document n
Lexeme 1 1 2 0 n
Lexeme 2 1 0 3 n
Lexeme 3 1 1 0 n
Lexeme 4 1 0 0 n
Lexeme 5 0 0 4 n
Lexeme 6 1 1 0 n
Lexeme 7 1 0 0 n
Lexeme 8 0 2 1 n
Lexeme 9 1 1 0 n
Lexeme n n n n n
In Table 1, each row stands for a stemmed lexeme (Lexeme 1, Lexeme 2 etc…),
and each column represents the context, i.e. the passage or the text. The numerical
values in each cell shows how many times a lexeme occurs in a certain document. For
example, Lexeme 1 occurs once in Document 1 and twice in Document 2; however it
does not occur in Document 3 hence a null value is assigned.
The next step in applying LSA is term weighting. In order to determine which
words occur more than the others, a local weighting factor is calculated. The process is
rather simple: words appearing many times in a text are given greater weights than
words that appear only once.
A global weighing factor is also calculated to determine items that occur many
times across the document sets. Like local weighting factor, it is a simplistic word-count
process. These two calculations are common, but they are not the only techniques in
LSA (see Landauer et al., 1998 for details of this calculation step).
21
After term weighting, only the nonzero values change in the matrix. However,
these values hardly mean anything, as they are still raw data at this point. Imagine this
raw data as a big group of people standing in rows directly in front of you, and you want
to know the dominant color(s) these people are wearing. Looking from a direct angle it
is impossible to see the people in the back rows, so you have to change your point of
view and find the optimum angle to catch the whole crowd. The rationale is similar with
the last step of LSA which is Singular Value Decomposition (SVD), which is used to
reduce the semantic dimensions to smaller manageable units. Again, there are many
techniques that could be employed at this stage and SDV is one of the options. The
outcome is generally a score between 0 and 1, and contexts having close scores close to
1 are interpreted as having more cohesion than contexts scoring closer to 0 (see
Appendix 5 for LSA scores obtained from two learner essays) . However, it cannot be
deduced that there is a perfect correlation between these values and cohesion.
An interesting note at this stage is that the absence of certain lexical items is just
as important as what is present in the context. In other words, what is important is not
the direct relationships among lexemes but how irrelevant one group of lexemes in a
text is to the other(s). This principle is claimed to have shed light onto children’s
inexplicably fast vocabulary acquisition in their first languages. That is, a child’s
knowledge about vocabulary when reading is claimed to come from the words that are
not in the text rather than what is available. For example, a typical American seventh
grader learns 10-15 words a day. Landauer et al. (1998, p. 274) suggest that about three-
fourths of the gain in total comprehension vocabulary that results from reading a
paragraph is indirectly inferred knowledge about words which are not in the paragraph
at all (see Landauer & Dumais, 1997, pp. 211-240 for a detailed discussion).
2.2.6. Situation Model Dimensions
In this index, there are four situational dimensions namely causation,
intentionality, time and space. These sub-indices make calculations about lexical items
based on an online database called WordNet, where nouns, verbs, adjectives and
adverbs are grouped into sets of cognitive synonyms (synsets) by human raters. Casual
verbs taken from WordNet are counted, and if the scores are high the text is assumed to
convey causality. Intentional verbs, again taken from WordNet, are calculated; high
scores mean the text at hand has a goal-driven content. Spatial content consists of
22
location nouns and prepositions, as well as motion actions and prepositions and is based
on the same database. Temporal dimension of a text is, in fact, about how events and
actions are articulated. In a text, articulation of events and actions is possible through
different tenses, and the tense repetition scores are taken into account in this index. This
is measured by means of repetition scores of tenses.
2.3. Interlanguage Lexicon
It has now become clearer that second language learners’ lexicon is more than
just a bunch of words; the issue goes far beyond that. Regardless of what theoreticians
claim, the lexicon is obviously a key component of language. However, it has often
been a subordinate in the mainstream of second language acquisition research.
Reflection on the issues is likely to reveal that the lexicon is central to the whole system
of language as it involves not only semantic, but also phonological and morphological
information. In addition, syntax is not divorced from lexicon.
Defining what is meant by “knowing a word” is also a controversial issue.
Nation (2001, p. 27) suggests the following word knowledge types that are necessary if
one claims to have complete knowledge of a word.
Form
• Spoken (What does it sound like? Eight sounds like [eit])
• Written (spelling)
Meaning
• Form and meaning (What is the meaning of a particular form?)
• Concept and referents (What concepts are included?)
• Associations (What words do we think of when we hear this form?)
Use
• Grammatical functions (the patterns the word occurs in)
• Collocations (What words can occur with the word—for example, with vacation,
one says take)
• Constraints on use (e.g., registers—in what contexts do we expect to hear this
word?)
23
There is also a distinction between the lexicon that language learners possess
and the lexicon which they use to produce written or spoken language. The former is
often referred to as receptive vocabulary, and the latter is called productive vocabulary.
In the receptive/productive distinction, Gass & Selinker (2008, p. 451) define receptive
knowledge of vocabulary items as follows:
Receptive vocabulary includes
• recognizing the word in writing or orally
• knowing the general meaning
• knowing the specific meaning in a specific context of use
• knowing that it is made up of the component parts—over, extend, -ed
• knowing that it has a possible negative connotation (as opposed to overqualify,
which may or may not have a negative connotation)
• knowing that it generally occurs with himself, herself, oneself, themselves,
ourselves, yourself
• knowing that the opposite is underextended.
They go on to explain productive knowledge which involves greater specificity
and includes:
• knowing how to accurately pronounce a word or correctly spell it
• knowing the precise meaning in a variety of contexts
• knowing that She overextended herself is OK, but that She overextended her
chair is probably not OK in the absence of a highly specific context
• knowing the precise context of use.
After taking the above points into account, it would be safe to assume that
language learners’ receptive vocabulary is greater than their productive vocabulary.
2.3.1. Breadth and Depth of Vocabulary Knowledge
Nation (2001) divides learners’ lexicon into two aspects as breadth and depth.
The former refers to the number of words learners know, whereas the latter refers not
only to word meanings, but also other parameters such as semantic relationships,
collocations and syntactic patterning which is related to parts of speech of lexical items.
Cobb (1999) further develops this division stating that breadth of vocabulary knowledge
comes from explicit learning of words on lists, while depth of vocabulary knowledge
24
comes from implicit learning of words through extensive reading. To attempt bridge this
division, Cobb (ibid.) carried out an experimental study in which he made the
participants create their own dictionaries of words to be learned through use of
concordance and database software. The participants in this study were provided with a
corpus assembled from the reading materials they were supposed to study. They were
assigned 200 words a week for 12 weeks. Control groups used a wordlist and
dictionary; experimental groups made their own dictionaries. Participants’ definitional
knowledge and transfer of knowledge were tested through pre-test, post-test and weekly
quizzes. The results revealed that control and experimental groups both made
substantial gains in terms of definitional knowledge, while only the concordance-
lexicography groups made significant gains on the novel text measure. In addition, the
control groups’ definitional knowledge did not last long, and delayed retention tests
consistently revealed that control groups did not retain their definitional knowledge,
while the concordance groups increased their definitional knowledge over time.
Learners of a second language are most likely to encounter certain lexical
problems. For example, native speakers of English know that the verb walk can take a
direct object as in the sentence ‘She is walking her dog.’, and that the verb give requires
two objects as in the following sentence ‘He gave his phone number to the man
unwillingly.’. It’s quite obvious from this aspect that knowing the meaning of a word is
not enough, and for EFL learners, being able to make such distinctions will take
considerable amount of time.
Knowing a word also involves the skill in making sound decisions about the
formation of lexical items by taking into account their positions in sentences. In her
study, Olshtain (1987) focused on the acquisition of new word formation devices in the
target language as an indication of near-native competence at the advanced level of
second language acquisition. The subjects were native speakers and two learner groups
of Hebrew. In the study, data was collected through written questionnaires composed of
production, evaluation and interpretation tasks. The results suggested that the
acquisition of word formation devices in the target language is as a gradual process.
In a longitudinal study, David (2008) investigated lexical diversity development
of British learners of French over a five-year period of time through a semi-guided,
picture-based oral task. The participants in the study were 80 learners of French whose
development was analyzed between the ages of 9 and 13. The data for the study was
25
gathered from the French Learner Language oral corpora (FLLOC). The main concern
of the study was to see if there were any differences between verbs and nouns in terms
of developmental rates. The results revealed that, compared to native speakers’
composition, learners’ spoken lexicon included more noun types than verb types
throughout the period of study.
Colloquial, idiomatic or collocational knowledge of words are also factors
affecting the quality of lexical knowledge. For example, the phrase big deal might not
make any sense to an EFL learner even with a considerable amount of contextual clues.
These kinds of lexical difficulties that EFL learners encounter, and related literature
support the idea that L1 and L2 lexicons have inherently different structures.
On the other hand, there are studies claiming a similarity between L1 and L2
lexicons. In a relatively recent study, Wolter (2001) casts doubt on the notion that
structures of L1 and L2 mental lexicons are different from each other. 13 Japanese
speakers of English as a second language and 9 native speakers of English participated
in the study. They were given a word association test using the aural-oral method. The
responses of the participants were classified into three groups, paradigmatic,
syntagmatic and clang-other, then native and non-native responses were compared. The
results revealed that like L1 lexicon, L2 lexicon is not randomly and loosely structured
as the past research claims. Phonological connections do take precedence over semantic
connections for moderately well-known words; however, the situation changes as the
learners gain greater understanding of individual words, at which point syntagmatic
connections become dominant. That is, as the second language speakers of English gain
depth of knowledge, their mental lexicon begins to resemble that of native speakers.
2.3.2. Lexical Networks
In the human mind, words are not stored in isolation; on the contrary, knowledge
and word acquisition is a process that involves creating connections between related
words (Haastrup & Henriksen, 2000). Through these connections, words form clusters
or groups, and the retention of a new word in memory depends on whether it is
connected to one of these clusters or not. These clusters or groups of related words are
also connected to each other, and this system is called as a lexical network (Crossley &
McNamara, 2009). As new words enter the network, the entire structure becomes
26
stronger. Figure 2, taken from Ferrer i Cancho & Sole (2001), illustrates a lexical cluster
using only a limited number of sentences.
Figure. 2. A toy network constructed with four sentences (Taken from Ferrer i Cancho & Sole, 2001)
In Figure 2, the toy network was constructed with the following sentences:
1. John is tall.
2. John drinks water.
3. Mary is blonde.
4. Mary drinks wine.
In this illustration, the available words are presented in part (a,) and their
relations are illustrated in part (b). In the figure, black dots represent a commonality and
the white dots represent rare words, and the words are linked if they co-occur
significantly.
Word associations and networks of native speakers of a language differ from
language learners in certain aspects. Meara (1978) investigated the lexical associations
made by learners of French and compared them to those of native speakers of French. It
was found that learners tended to relate words in a more simplistic way than the native
speakers. Native speakers appeared to make associations based on pragmatic or
syntagmatic factors. For example, the native speakers of French who participated in the
27
study associated the word man with the words woman, dog, boy or child; they
associated the word brush with the word teeth. On the other hand, the learners who
participated in the study tended to associate the words based on phonological
similarities. For example, they associated the French word plafond (ceiling) and the
word professeur with the word profond (deep). The results of his study were interpreted
as the lack of a network in the learners’ L2 lexicon.
Another related study is by Zareva (2007). Her rationale for the study was that
little research had been done on the role of language proficiency in the associative
patterning of L2 learners’ lexical knowledge, especially the way it affects the
quantitative and qualitative patterns of connection. Moreover, she claims that no
attention is devoted to the strength of the relationship between these patterns. In her
study she tried to determine differences in the organization of lexical knowledge
between L2 speakers and NSs. 87 university students participated in the study; 29 of
them were native speakers of English, and 58 of them were L2 learners of English. The
L2 group was divided into two groups according to their proficiency level. The
participants were given a vocabulary test which had been compiled from a dictionary by
random selection. The test involved stimulus word (SW) and word association (WA)
questions. In the WA section the subjects were required to fill in the blanks in sentences
like ‘I associate this word with …………………’. After the implementation, a list of the
associations generated by each participant was compiled and lemmatized. Eventually,
there were three separate lists of lemmatized words. The analysis of these lists revealed
that lexical differences were quantitative rather than qualitative. The quantitative
differences were most noticeable in the intermediate learner group. Adult L2 learners,
like native speakers, showed a preference for a greater proportion of paradigmatic rather
than syntagmatic connections for familiar words.
Meara (2002) in a recent review of four books about second language lexical
acquisition highlights two important issues. Firstly, he claims that unlike syntax and
morphology, lexical development in L2 has been sidelined since the 1950s. He goes on
to say that L2 lexicon is currently enjoying a rediscovery period insofar as studies
which were once overlooked are now better understood. This situation, in fact, raises
additional issues, and it is beyond time that we fill the gaps to create a sound L2 lexicon
theory.
28
2.3.3. Lexical Cohesion
When EFL teachers are asked about cohesion in writing, it is most likely that a
discussion about grammatical cohesion will start; however, lexical cohesion has an
equally crucial place in determining the overall quality of a text. According to both
Silva (1993) and Ferris (1994), cohesion plays an important part in the lexical
development of L2 writers, and coherence is often a connotation for the term cohesion.
There is a slight difference between these two concepts. Louwerse (2004, cited in
Crossley & McNamara, 2009) makes the distinction between these two concepts as
follows:
Coherence refers to the representational relationships of a text in the mind of a reader whereas cohesion refers to the textual indications that coherent texts are built upon. In essence, then, cohesion consists of the elements of the text, while coherence refers to the consistency of the elements as a mental representation. The more cohesive devices in a text, the more coherent it will be and the easier it will be to understand.
It is clear from the above explanation that coherence is the mental representative
for cohesion; that is, coherence is a mental process, while cohesion is a textual quality.
Connor (1996, p. 83) defines cohesion as "the use of explicit linguistic devices to signal
relations between sentences and parts of texts." By making use of these cohesive
devices, phrases or words, the reader associates adjacent statements to make sense out
of the text at hand.
Theoretical background concerning cohesion is generally attributed to Halliday
& Hasan (1976). In their seminal work, they use the word text to refer to any passage,
spoken or written, of any length, that forms a unified whole. It is treated as a semantic
unit which, unlike what is generally accepted, does not consist of sentences, but is
realized by sentences. Cohesion is therefore a schematic concept; it refers to
relationships or ties that exist within the text and define it as a text. When talking about
ties, Halliday & Hasan refer to single instances of cohesion versus cohesively related
item, and it is at this point that two facets of cohesion come in two focus: grammatical
and lexical. Haliiday & Hasan predominantly deal with grammatical cohesion, and
entire chapters are designated for grammatical issues such as substitution, ellipsis and
conjunctions.
Lexical cohesion, however, is given in only one chapter, in which they classify
lexical cohesion as follows (Halliday & Hasan, 1976, p. 279)
29
1. Reiteration
a) same word repetition
b) synonym or near synonym
c) superordinate
d) general word
2. Collocation
The following examples are given to illustrate reiteration.
I turned to the ascent of the peak.
The ascent (repetition)
The climb (synonym)
The task (superordinate)
The thing (general word)
is perfectly easy.
Reiteration to form cohesion can be realized through repetition of certain
structures. However, redundant repetitions (McGee, 2009) are potential problems in
EFL learners’ texts because too many repetitions might disorient the reader. Repetition
is, in fact, one of the characteristic features of the spoken genre, and EFL learners tend
to transfer these features into the written genre (Cobb, 2003).
Making use of synonyms might also help create cohesion in a text. However,
making use of appropriate synonyms while writing is also quite a challenging task for
EFL learners. Harvey and Yuill (1997) provide a detailed account of the role Collins
COBUILD English Language Dictionary played in the completion of written tasks by
EFL learners. The learners were required to identify and distinguish various types of
information about a word they could look up in the dictionary. Synonym searching was
among the most frequent activities performed by the learners. However, 36.1% of
synonym searches were reported to be unsuccessful. So, even if the learners can access
synonyms of the words they want to use, it is not that easy for them to pick out the
appropriate one for the context.
Superordinates are general words that refer to a class like the words animal or
vehicle. Hyponyms, on the other hand, refers to specific members of a class, dog and cat
or car and truck. Making use of such words interchangeably might reinforce cohesion
30
in a text. The following example (Salkie, 1995, p. 28) demonstrates the cohesive link
established by making use of a superordinate.
Brazil, with her two-crop economy, was even more severely hit by the Depression than other Latin American states and the country was on the verge of complete collapse.
In the example above, there is cohesion between the words Brazil and country,
and the word country is a superordinate, while Brazil is a hyponym because it is more
specific than country.
Collocation, on the other hand, covers all types of lexical relations that do not
need referential identity and cannot be described as a type of reiteration (see Halliday &
Hasan, 1976 for details). After analyzing use of collocation in GeCLE (German Corpus
of Learners of English) Nesselhauf (2005, p.71) reports that out of the 207 learners
represented in GeCLE, 183 produced incorrect or questionable collocations.
Nesselhauf's analysis supports the previously existing conclusion that there is no
correlation between proficiency and use of collocation. The data actually shows a
negative correlation of years spent learning English and relative collocation use.
Furthermore, Fan (2007) also reports that the results of the study indicate that the
performance of the Hong Kong students in collocational use might be adversely affected
by their L1 and L2, as well as their inadequacy in the lexis and grammar of the target
language. These analyses clearly indicate that EFL learners experience serious problems
concerning collocation use regardless of their level of proficiency.
As can be interpreted from the discussion thus far, cohesion is usually regarded
as both a grammatical and a lexical issue. This dichotomous view of cohesion, however,
has been subject to criticism. Mahlberg claims that cohesion is a fundamentally lexical
phenomenon (in Flowerdew & Mahlberg, 2009, p. 118). Pronouns, articles and other
function words might, to some extent, help cohesion, but the real ties or cohesion lie in
lexical relationships. This point of view goes back to Lewis (1993, p. 91) who states
that “language consists of grammaticalised lexis, not lexicalized grammar.”
Since these concepts have been discussed for years and their haziness has made
them difficult for language teachers to teach, studies were carried out to clarify this
issue in terms of EFL learners’ learning strategies.
31
Palmer (1999), for example, observed the way that students make use of
cohesion and coherence in their written assignments in English classes. 89 intermediate
level learners participated in the study. Their written pieces of discourse were produced
in a test environment. The subjects were divided into two groups, A and B. The subjects
in group A were lectured on cohesion and coherence, and several exercises concerning
the issue were carried out. The lectures mainly consisted of topics like overall length of
text, use of paragraphs to divide information in a coherent way, lexical reiteration and
pronominalisation as a cohesive device. The topics that the students were asked to write
about were similar ones. The results revealed no significant difference between the two
groups in terms of text length and paragraph division. However, group B, which was not
lectured about lexical reiteration, used more reiteration compared to group A, which
was aware of other possibilities that could be used to create coherent texts. This led to
the conclusion that lexical reiteration is a predominant technique among ESL learners.
Palmer also gives some remarks about how to improve cohesion in learners’ texts.
A new and digitally-oriented method for determining lexical cohesion in texts
has been under way since the development of Coh-metrix (see Section 2.2 for details).
The online database Coh-metrix was first introduced by a team of researchers (Graesser
et al., 2004). One of the indices in Coh-metrix, LSA, was tested to explore how it can be
used as a method to examine the lexical development of L2 speakers. The aim of the
study was to see if LSA measures of semantic co-referentiality increases as learners
study an L2, and whether a common measurement of lexical proficiency demonstrates
growth. A group of L2 English learners who were enrolled in an intensive language
learning program at a state university in the United States were involved in the study.
Their lexical growth was tracked using of LSA scores over a long period of time. The
participants were at the lowest proficiency level at the beginning. A spoken corpus was
formed through interviews over one year and the data collected in the 2nd, 4th, 16th, 32nd,
50th and 52nd weeks was recorded. Statistical analysis revealed that the values computed
in the last meeting (52nd week) were statistically significant from those of the first
meeting. It was concluded that over time, subjects’ proficiency levels increased in terms
of lexical relations in their utterances.
Using the indices in Coh-metrix, a recent and comprehensive study by Crossley
and McNamara (2009) explored how lexical differences related cohesion and lexical
networks can be used to distinguish between texts produced by native English speakers
32
and ESL learners. Two corpora were used; one was from LOCNESS (Louvain Corpus
of Native English Essays), and the other one comprised essays written by Spanish
learners of English and was taken from the International Corpus of Learners of English
(ICLE). The learners’ ages and their learning contexts were similar: they were all
university students in their twenties. The native corpus comprised 208 texts (151,046
words in total) and the learner corpus was comprised of 195 essays (124,176 words in
total). Both corpora included argumentative essays, whose topics were also taken from
ICLE. A discriminant function analysis was conducted, and in the process, Coh-metrix
indices that measure lexical features related to cohesion and lexical networks were
selected. The texts were compared in terms of word hypernymy, word polysemy,
argument overlap, motion verbs, CELEX written frequency, age of acquisition,
locational nouns, LSA givenness, word meaningfulness, and incidence of casual verbs.
The results demonstrated that deeper-level lexical indices related to cohesion and
network models in Coh-metrix tool can significantly distinguish between L1 and L2
texts. The importance of this study is that, as a contrast to the related literature (Connor,
1984; Reynolds, 1995, cited in Crossley and McNamara, 2009), it is the first study to
distinguish L1 and L2 texts solely based on lexical features.
Up to this point, literature concerning the descriptive phase of the current study
has been discussed. The next section of the literature review will involve practical issues
concerning the use of corpus linguistics in language pedagogy.
2.4. Corpus Linguistics
A corpus is a collection of texts stored on a computer or a digital device. The
first and most important feature of these texts is that they are collected systematically
and in accordance with certain principles. The contexts provided by corpus software are
rich because they consist of millions of words (British National Corpus for example),
and they are natural because they are collected as they are without adjustments are
made. The authenticity of corpora is not a topic for discussion; however, their richness
is.
Corpus linguistics came into being when it was first considered that language is
something observable. This point of view has been around for a relatively long period of
time, but there have been some changes in its paradigm. One of these changes took
place after the distinction between parole and langue was made by Saussure (Finch,
33
2005, p. 24). Parole was regarded as the language as it is being used and it is, in fact, an
outgrowth of a much more complex system called langue.
Similar to this Saussurean perspective, in his cognitive-algorithmic model,
Noam Chomsky tries to attribute humans’ knowledge and ability concerning language
to an unconscious system which could enable utterances in any language; this model is
based on human competence. On the other side there are supporters of a performance
model which takes into consideration the actual psychological and physical processes
involved during language production.
In the competence-performance dispute it was stated that “… information about
the speaker-hearer’s competence … is neither presented for direct observation nor
extractable from data by inductive procedures of any known sort.” (Chomsky, 1965, p.
18). From this point of view, in an interview Chomsky would implicitly refer to corpus
data as junk (Aarts, 2001, p. 6).
To corpus linguists’ defense, corpus linguistics embraces the idea that “a word in
or on itself does not carry any meaning, but the meaning is often made through several
words in a sequence” (Sinclair, 1991); and “[t]he aim is not to study idiosyncratic
details of performance which are, by chance, recorded in a corpus. On the contrary, a
corpus reveals what frequently recurs, sometimes hundreds or thousands of times, and
cannot possibly be due to chance” (Stubbs, 2004, p. 111).
As these discussions were going on, thanks to technological developments,
digital data storage was becoming easier and compiled language data was becoming
more and more accessible. In 1960’s the first electronic corpus, Brown Corpus, was
compiled. It contained digitalized documents comprising of a million words and it is
still in use today.
When the technological advances made scanners available, it was now time to
deal with a tremendous amount of data which was beyond the capabilities of any
methodology previously available. The emergence of the internet in the same era also
contributed to the accumulation of language data. Overwhelmingly large language
samples started flowing online through emails, websites and personal blogs. This
situation could be counted as an advantage as it is in line with the concept that a corpus
is supposed to be representative, which means that if it is to meet a related need, it must
represent certain aspects of a language through common features of the data it contains.
34
On the other hand, this abundance of collection brought about the issue of accurate
interpretation. That is, as the language data became larger and larger in amount, it was
also getting harder and harder to interpret them accurately.
Applications from different areas of study derived from corpus linguistics;
among them the most noteworthy ones are lexicography, translation, stylistics,
grammar, gender studies, forensic linguistics, computational linguistics and language
teaching (Tognini Bonelli, 2001, p. 1). There are numerous free/commercial software
packages and online tools for corpus analyses such as Microconcord, WordSmith Tools,
Coh-metrix, AntConc and Statistica.
2.4.1. Learner Corpora: A Revolution
Learner corpus (henceforth LC) research is a relatively recent branch of corpus
linguistics. In our usage, LC refers to the collections of digitalized texts of written or
spoken genres in English produced by EFL learners. The origin of LC is often attributed
to Sylviane Granger and her team’s studies in the early 1990’s (see Granger, 1993,
1994, 1998, 1999, 2003; Granger and Tyson, 1996).
Granger (2008) summarizes the outcomes of these studies and establishes certain
points that need clarification. First of all, she distinguishes learner corpora from that of
native ones in terms of their sources; that is, the data is gathered from English language
learners from different L1 backgrounds like Spanish, Swedish or Chinese. In order to be
able to call a corpus LC, the native or the first language(s) of the learners must be non-
English. That is to say, written or spoken productions of EFL learners coming from
countries where English language has a somehow official status, such as India, do not
count as valid data.
The context from which learner productions are taken from is another issue for
Granger (ibid.). When the issue is native English data, compiling a corpus is
straightforward, albeit hard. However, learners rarely produce written or spoken data in
their target language. Tasks like reading aloud, picture descriptions or informal
interviews are contexts where language learners are expected to produce tangible data.
When she compares written and spoken learner data, for practical reasons
spoken data obviously outnumbers by written data by a ratio of up to 10 to 1 (ibid.).
Written productions of learners are easy to collect and digitalize, since digital
environments like the internet and email are fairly widely available. For example, Belz’s
35
(2004) study which includes details about the Telecollaboration project is a good
example in this respect.
Size is another consideration in the collection of LC. In general, when the size is
considered in corpus linguistics, the classical paradigm is “small is not beautiful, it is
simply a limitation” (Sinclair, 2004, p. 189). However, as Granger puts it, in certain
cases even a collection of only one learner’s essays could be valuable. Supporting this
view, Sinclair (2005) states that small scale corpora are particularly useful in Language
for Specific Purposes (ESP) while teaching and learning vocabulary.
When the design issues of LC are considered, the critical point is the
representativeness of the collected learner data. To overcome potential
representativeness problems, Granger (2008) adds the following parameters to learner
corpus design:
Figure 3. Parameters to be added to a learner corpus (Taken from Granger, 2008)
Such a parametric approach to design enables researchers to focus on the issues
they want to reveal in an LC. For example, if a researcher wants to analyze lexical
issues in advanced proficiency level language learners, s/he will have to design or use a
corpus from advanced learners (learner variable), and manipulate the task type (task
variables) in the process.
According to Granger (2008, p. 259), the use of LC in language teaching/learning
is supposed to have two important functions:
36
Learner corpora, which can be roughly defined as electronic collections of texts produced by language learners, have been used to fulfill two distinct, though related, functions: they can contribute to Second Language Acquisition theory by providing a better description of interlanguage (i. e. transitional language produced by second or foreign language learners) and a better understanding of the factors that influence it; and they can be used to develop pedagogical tools and methods that more accurately target the needs of language learners.
Since the systematic collection of learner corpora and the beginning of
contrastive corpus analysis through digital technologies, the relevant studies have been
mostly Europe-centric. To remedy this oversight, Cobb (2003) replicated three studies
carried out in Europe to see if the same results could be found among the North
American counterparts of European EFL learners. The original studies were carried out
by Ringbom, Cock, Granger, Leech, & McEnery and Petch-Tyson all in the same year,
1998.
The first replication was of Ringbom’s study, which was a comparison of L1 and
L2 texts in terms of word frequency. The participants came from seven different L1
backgrounds. The results revealed that they consistently used 100 very high frequency
words in their writings, about 4-5 % more than native speakers. These were not the
highest frequency words but were rather included slightly less common content words
along with some common content words like people, new, many from the top 30-100
range.
The second replicated study was from Cock et al. (1998, cited in Cobb 2003),
and it explored prefabricated structure in L2 oral productions compared to L1. For this
purpose, they made use of a matched set of 25 transcribed learner and NS interviews.
The results showed that learners tend to use as many prefabricated chunks in their oral
productions as native speakers of English do, which was against expectations. However,
when the variety of these chunks were analyzed, there appeared a significant difference
between the two groups in terms of variety, which meant that the native speakers
prefabricated structures had more variety than those of the learners.
The third study Cobb replicated was from Petch-Tyson who compared written
and spoken productions of advanced learners. The impetus for this study was the non-
native like productions of advanced EFL learners. In general, it is supposed that spoken
genre tends to be more context, space and time-oriented because of the presence of
37
receivers; while written genre ends to be free of time and space because the reader is
never present. The question was whether the written productions of advanced learners
were talk written down. The results showed that the advanced learners who participated
in the study employed from two to four times the number of spoken language features
than equivalent American NSs did. Especially personal pronouns in learner texts
statistically outnumbered personal pronouns in native speakers’ texts.
The three studies mentioned above were replicated by making use of Quebec
learner corpus, which consists of over 250,000 words and was divided into two main
sections, advanced and intermediate, by taking the proficiency levels of the learners into
account. The statistical analyses were performed on WordSmith Tools (Mike Scott) and
the results showed that:
1) In relation to Ringbom’s study, like their European counterparts, North
American advanced EFL learners showed a parallel overuse of most frequent
vocabulary items. However, overuse of the top 30-100 range vocabulary items
was significantly higher than European advanced EFL learners.
2) In relation to Cock’s study, North American EFL advanced learners do use
precasts, albeit with much lower variety. Therefore, the results of the two studies
confirm each other as to the usage of prefabricated structures. However, as is
mentioned by Cobb, at this point there is a proviso that the corpora used in this
replication differ in genre one being spoken and the other being written.
3) In relation to the Petch-Tyson study, once again the results overlap. There is a
significant difference between L1 and L2 written productions in terms of
pronominal references. The results indicate that in both learner corpora, learners’
interpersonal involvement carry most of the signaling load, which would mean
that while writing they try to communicate as if they were in a personal
conversation.
2.4.2. Corpus Linguistics and Language Pedagogy
Fligelstone (1993) describes three aims of corpus-based linguistics in teaching:
teaching about the principles behind corpora, teaching the learners how to exploit
corpora and exploiting corpora to teach. The first point refers to revealing the rationale
behind corpus linguistics which is the rich reality of the target language. The second
38
point is about showing the learners how to make use of corpora, giving a path to
autonomy. The last idea relates to language teachers, and draws attention to how to
make use of corpora to obtain authentic teaching material which is like having a native
speaker present in their classrooms.
Conrad (1999), in order to highlight the fact that corpus-based studies have
generally focused on only certain points through automated analysis, carried out a study
to show how important corpus data is for language classrooms and how it can be used
without computer programs. In the study, language teachers were provided with
information on how useful it would be to use concordancing software and data gleaned
from this software while teaching. The development of cloze tests and improvements in
language testing in general were also attributed to concordancing tools and corpus
linguistics. Conrad acknowledges that frequency studies alone could provide teachers
with very valuable information which would be overlooked otherwise. The problem was
that language teachers were getting a false one-dimensional impression of corpus
linguistics. Focusing too much on lexical items alone or their connections with
grammatical structures was in fact meant ignoring more complex grammatical analyses.
She goes on to report the three main important characteristics of corpus-based
research. The first characteristic feature is that it is systematic and natural. A principled
collection is regarded as a must for a corpus since it might be employed in different
contexts for a variety of purposes; that is to say, not all corpora are the same.
The second characteristic feature is that it is digital, i.e. it involves computers in
the analysis process. It would be literally impossible for researchers to count elements
or analyze large bodies of texts.
The final feature of corpus-based research is its inclusion of both quantitative
and functional interpretations for analyses. This is important in that numbers obtained
from digital software would not be enough for functional interpretations concerning
communication. Since some of this software provide researchers with both of them,
descriptions from both aspects are possible.
To illustrate these points, Conrad (ibid.) carried out a corpus-based study
concerning linking adverbials. In the analysis process, after describing problematic
areas of these connectors, she used Longman Spoken and Written English Corpus which
is a grammatically tagged corpus. In the process, she focused on frequency, semantic
39
analysis, and grammatical analysis of these adverbials. Through these analyses she
found out that these adverbials were most common in academic prose. Their semantic
breakdown revealed that in academic prose result/inference adverbials account for the
largest proportion of linking adverbials. The comparison of written and spoken genre
revealed that these adverbials were similarly higher in frequency in academic prose but
their variety was strikingly different. Spoken academic prose contained small number of
adverbial items used repeatedly.
From a holistic point of view, Conrad’s study reveals that in corpus datasets
there is more than meets the eye; a variety lurks behind concordancing lines waiting to
be analyzed. Furthermore, the difference between written and spoken genres determined
in her study is also worth mentioning. Conrad concludes her remarks by stating that
corpus-based research might be out of reach for language learners, but for language
teachers it surely is valuable.
In an argument in the late 1990s concerning language teaching and corpora, two
questions arose: (1) to what extent do corpora provide descriptions of real language, and
(2) whether such English is what foreign learners need (Cook, 1998). In an attempt to
take a position in the argument, Gavioli and Aston (2001) tried to summarize the
highlights of corpora in language pedagogy.
The first point they make is regarding corpora as a means to test intuitions
concerning language use. Language teachers are never unaware of this point as they
occasionally ponder which exact thing a native speaker would utter in a certain context.
Sometimes they resolve the issue by merging intuition and dictionary entries. At this
point, corpora can be used to refute or confirm this fuzziness through evidence.
The second point is that corpora can provide clarification when teaching certain
grammatical or lexical issues. Imagine a language teacher caught up in a teaching
context where s/he has to facilitate learners’ comprehension of verb tenses. A tagged
native corpus would provide the teacher with valuable data as to the frequencies of verb
tenses in English. S/he would easily prioritize the subjects, and explain the rationale
behind it, save time and energy as a result. Gavioli and Aston’s (ibid.) insight about this
point is “…that while corpora do not tell us what we should teach, they can help us
make better-informed decisions, and oblige us to motivate those decisions more
carefully.”
40
All in all, they regard corpora as a valuable resource for language learners both
inside and outside the classroom. Learners can problematize language, explore texts,
and by merging their own experience with the reality present in corpora they can
authenticate the discourse they want to create. Teachers also benefit from corpora to
make better-informed decisions.
Gavioli and Aston conclude their discussions with three main requisites for
encouraging learners to use corpora, thereby increasing learner autonomy. The first one
is direct access of the learners to a variety of corpora which include different genres,
such as spoken and written language. Secondly, more user-friendly corpus software is
needed so that the learners do not waste much time learning how to make use of them.
The last point concerns corpus-based activities. It is claimed that more research is
needed to determine what language learners should be exposed to, and corpora should
be used for different language proficiency levels. Answers are expected to unfold as
learners become more involved in the process.
Braun (2005) mentions direct and indirect influence of corpora on language
pedagogy. Regarding indirect-use he claims that corpus-based analysis of English has
influenced syllabus design, the methods and materials for language teaching and
learning, test design, feedback and evaluation, references and the contents of reference
works, and grammar. According to Römer (2005, p. 266), ELT textbook accounts of
language use are often decontextualized and lack empirical basis. On the other hand,
direct influence occurs when learners make use of corpora to get a rich and realistic
picture of the target language.
In a recent study, Vannestål and Lindquist (2007) attempted to increase their
students’ motivation by showing them that English grammar is just more than a set of
rules in a book and by enabling them to assume more responsibility for their own
learning. The underlying idea was to introduce the real use of language into the
curriculum as a complement to grammar textbooks. The subjects worked with problem
solving assignments that involved formulating their own grammar rules by using the
examples taken from the corpus. Students’ work was evaluated by means of
questionnaires and interviews. One important conclusion was that using corpora with
students requires a large amount of introduction and support. Some participants
appreciated working with corpora while others, especially weak students, found it
difficult and boring. Vannestål and Lindquist claim that, through the qualitative
ev
ex
co
inf
co
pe
2.4
be
En
co
%
wo
Fig(T
valuation of
xperience of
rpus-based
formation f
rpora.
The p
dagogy can
First, observteachecorpuswhich analysgramm
4.3. Corpu
The di
en overloo
nglish dictio
ntexts. Nati
of the writt
ords (see Fi
gure 4. TexTaken from O
f the corp
f their own
instruction
found in la
points made
n be summar
learner autve languageer. With thes promotes interlangu
sis activitiematical cons
us Linguist
istinction be
oked. Amon
onaries, som
ion (2001, p
ten and/or s
gure 4) whi
xt coverage O’Keeffe, 2
pus work,
n learning s
n to be us
anguage co
e up to this
rized as foll
tonomy is e and make e observatio
hypothesisuage is thes are eassciousness-r
tics and Vo
etween high
ng hundred
me words o
p. 11) make
spoken cont
ich is known
in a 10 mill2007, p. 32)
they gaine
situation. A
sed in teac
ourse books
s point con
lows (Conra
increased generalizat
ons and gens formation
hought to sily designraising.
ocabulary T
h and low v
ds of thous
occur more
es this situa
texts in Eng
n as the Gen
lion-word c
ed explicit
Another con
cher trainin
s with the
ncerning cor
ad, 2005, p.
as studentstions, ratherneralizationn and testiprogress. F
ned to pro
Teaching/L
ocabulary it
sands of le
than the ot
ation more m
glish langua
neral Servic
orpus of sp
knowledg
nclusion wa
ng, for ins
authentic l
rpus linguis
. 402):
s are taughr than depens, the analying, the paFurthermoromote noti
earning
tems is anot
exical entri
thers in spo
manageable
age are mad
ce List (GSL
oken and w
ge about st
as the poten
stance, to
language fo
stics and la
ht how to nding on a ysis of the athway by re, corpus icing and
ther aspect
ies in well
oken and/or
e by stating
de up of abo
L).
written Engli
41
tudents’
ntial for
contrast
ound in
anguage
that has
l-known
written
g that 80
out 2000
ish
42
It is quite obvious from Figure 4 that about 80 % of the most frequent 10 million
words in spoken and written English consists of 2000 words. This point of view could
be a valuable paradigm in dealing with what vocabulary to teach to beginner level
learners, and it could save considerable amount of time and energy.
Next to this concept comes Academic Word List (henceforth AWL) which was
compiled by Coxhead (2000). AWL is considered a good example of the real language
that is used in academic texts. This list was extracted from academic texts from four
faculty areas (arts, commerce, science and law) by making use of computational tools,
and has been used to develop teaching/learning materials for some time now. The AWL
covers 10 % of the vocabulary coverage of academic texts (Nation, 2005, p. 583). The
idea stemming from this statistics is that trying to focus on high-frequency words rather
than trying to learn every lexical item we encounter is time-efficient in general. When
this point of view is considered from learners’ angle, it might be helpful for them in
focusing on what is more important to learn. From this respect, words do not appear to
be equal especially in teaching/learning context.
Considering the differences among learning contexts, it is difficult to decide
what type of corpus is suitable for teaching academic vocabulary. Different versions of
EAP (English for Academic Purposes) corpora are available for such concerns
(Coxhead, 2010). For example, if the teaching/learning context focuses on written
productions of the learners, it will be a wise decision to make use of a written corpus
like British National Corpus (henceforth BNC); if the context is more specific like law,
then it will make sense to utilize a more specific corpus like BNC Law corpus to
develop vocabulary materials.
Corpus size is another point of concern at this point, which is directly related to
the issue of representativeness of a corpus (see Biber, 1993 for details). When the aim is
teaching grammar a relatively small-sized corpus may help the language teacher reveal
grammatical points in a language; however, when the aim is to deal with vocabulary
with relatively low frequency, corpora with greater size will be required (O’Keefe et al.,
2007, p. 55)
2.5. Data-driven Learning
With ever-expanding technological improvements, large scale corpora are easily
available to language teachers and learners. Anyone involved in the language
43
teaching/learning process can now access large databases and make analyses according
to their needs. This paradigm, in fact, overlaps with the current view in educational
research about ideal learners and learning environments, which are defined as motivated
and autonomous learners with unique interests and needs which can be met via the
internet, anywhere, any time.
Data Driven Learning (henceforth DDL) is an approach and a technique which is
also known as discovery learning. Tim Johns (1991) used the term and the technique
itself for the first time. The goal was to put the learners in a pro-active role in their
learning process by making them analyze sets of concordances extracted from a corpus.
The advantage of such an approach is that it could both/either be teacher-initiated and/or
learner-oriented. That is, teachers can make use of concordance while preparing
teaching materials, or after some training the learners could develop autonomy to learn.
Johns used concordancing printouts in his class, which was appreciated by his
students who found this technique more helpful than simple gap-filling preposition
exercises. Not only the students but also Johns himself benefitted from these activities.
Through concordancing activities, he was able to notice important lexical or
grammatical points that he had overlooked before (Johns, 1986).
It could be said to constitute a form of comprehensible input (Krashen 1988), particularly when the content of the corpus is carefully chosen to be familiar to the learners (Allan 2009). It does, however, differ from Krashen’s scenario in one important way. The simplified language or caretaker talk which Krashen describes as helpful to the learner is absent here. Although the content may be familiar, the language in a native speaker monolingual corpus consists of attested examples of actual language use. The multiple contexts do, however, enable the learner to observe patterns. Thus, in the concordance lines with ‘end up’ listed above, the learner can observe that this phrasal verb is followed by the -ing form of the verb (‘end up hating’, ‘end up living’), by a preposition (‘end up in the army’), by an adjective (‘end up homeless and uneducated’), or by a noun (‘end up some kind of fat separatist’). It cannot be guaranteed, of course, that a learner looking at that concordance will learn all these uses. As no one advocates data-driven learning as the main component in an approach to language learning, but rather as an enhancement of text-based work, the learner may be using the concordance to check if one particular use is correct, and the concordance could confirm that and reinforce the learning process.
In addition to the ideas presented above Bernardini (2004, p. 22) also suggests
that data-driven language learning supports exploratory and discovery learning because
44
of the richness of the environment and the endless possibilities that corpus software
offer.
2.5.1. Criticisms and Downsides of DDL
With free and commercial software like Microconcord and Wordsmith Tools as
well as online corpora like BNC being more available, Johns’ endeavors found some
followers. These software and online datasets started serving language teachers and
learners. However, before long, criticism of the fad emerged. Cheng, Warren, & Xun–
Feng, (2003) listed some downsides of these analyses and activities. They mention a
confusion among students, stemming from “lack of knowledge and skills in choosing
and using corpora and in using computer software like concordancers, and lack of
enough data in the ICE-GB to enable confirming or refuting some hypotheses” (ibid.)
and the analyses and related activities were reported to be time-consuming, laborious
and tedious. Yoon and Hirvela’s (2004) findings, despite a considerable amount of
positive attitudes among learners, also reveal the time-consuming aspect of DDL
activities.
In addition to these studies, Chambers (2005) reported some other negative
aspects of DDL activities. Learners who participated in her study stated that these
activities could not be a substitute for classical grammar books; they were also aware of
the limitations of the corpus that was used in the study. In parallel with Yoon and
Hirvela’s findings, Chambers’ (2005) subjects also found the activities laborious and
time consuming. Another criticism coming from the learners was that they were lacking
certain skills to make use of corpus data; they felt untrained, thus unskilled. Some of the
learners went even further and stated that such activities would be beneficial only for
advanced learners. The last criticism from the learners was related to the availability of
the corpus, which brought up questions concerning the learning environments of the
subjects.
The second and holistic impetus for criticism was the issue of context. It has been
claimed that concordance activities decontextualize lexical items (Widdowson, 2000).
He goes on to say:
To point out these rather obvious limitations is not to undervalue corpus analyses but to define more clearly where its value lies. What it can do is reveal the properties of text, and that is impressive enough. But it is
45
necessarily only a partial account of real language. For there are certain aspects of linguistic reality that it cannot reveal at all. In this respect, the linguistics of the attested is just as partial as linguistics possible.
These ideas actually make sense when a concordance screen with the cut-off
edges is visualized (see Figure 5 for an example). However, considering this point with
lexical ties in mind would also make sense, because one of the main rationales behind
concordancing activities is to raise awareness of these lexical ties in limited contexts,
and learners do not have to understand everything in a concordancing line. As Gavioli
and Aston (2001) puts it “[a] concordance does not make sense in itself: sense has to be
attributed to it by the reader, who must infer patterns which will as far as possible
account for the data.”
2.6. Using Concordancers in Language Teaching/Learning
A concordancer is a section of occurrences of a target word presented in multiple
contexts. Figure 5 illustrates a real concordancing screenshot taken from
http://www.lextutor.ca for the word ‘example’.
Figure 5. A concordancing screenshot for the word ‘example’
As the figure illustrates, the word ‘example’ is presented in an authentic and rich
environment. Most of the possible collocations are present in the concordance outcome
screen. Collocation in this context refers to the restrictions on how words can be used
together, for example which prepositions are used with particular verbs, or which verbs
and nouns are used together (Richards & Schmidt, 2002). In the example provided in
46
Figure 5, the preposition ‘for’ and the adjective ‘excellent’ can be counted as two of the
strongest collocations of the word ‘example’, because they appear to be occurring with
more frequency than other collocations.
Sinclair (2003, pp. xvi-xvii) tries to outline a set of steps that should uncover the
mysteries of concordances with the foresight that these procedures will, one day, be
carried out by computers, which are needlessly laborious to a human. The steps to be
taken are outlined as follows:
Step 1. Initiate. Look at the words that occur immediately to the right of the NODE. Note any that are repeated. Do the same with the words immediately to the left of the node. Decide on the "strongest" pattern and start there. Deciding which the strongest pattern is depends on the circumstances, and with small numbers of instances is to some extent a matter of judgment. If one particular word form occurs in the same position in more than half the instances then it is pretty dominant, and is likely to be the best place to start; if there is no single word that stands out, but a grammatical word class is apparent in most of the lines, then start there. If there is nothing obvious at first sight, count which side has the largest number of repeated words; this is an indication of the coverage of repetitions, and a reliable place to start. Where you have strong patterns on both sides of the node it is safe to start on either side, since the retrieval of patterns is a cyclical procedure, and you will retrieve neglected patterns at Step 5. Step 2. Interpret. Look at the repeated words, and try to form a hypothesis that may link them or most of them. For example, they may be from the same word class, or they may all have similar meanings. Step 3. Consolidate. Assuming that Step 2 has been successful, now look for other evidence that can support the hypothesis - for example, single occurrences that come close to the criterion that you have set up, or structures that are different ways of expressing a similar meaning. Also you should look beyond the word position that you have started with, because there can be variations that separate elements of a pattern; look at the adjoining words and even some more distant ones, and in some cases also consider words on the other side of the node. Use always the criterion of how close they are to coming under the hypothesis that you have set up, and be prepared to revise and loosen up the hypothesis a little if by doing so you can include several more instances. Examples of the sort of variants that occur are as follows. A pattern like "his N" can be stretched to include "Bill's N", or even "the N of the village". It can be developed into "his own N", pushing the two words apart, and even "his funny old N". The choice of active versus passive voice in grammar can alter the positions of words relative to each other, e.g. "they drove away in a bus" versus "the bus was driven away".
47
Step 4. Report. When you have exhausted the patterns you can observe, and have revised your hypothesis so that it is as flexible as it needs to be and as strong as it can be, write it out so that you have an explicit, testable version for the future. You will be surprised how often you may need to return to this and rephrase it without fundamentally altering the classification. Step 5. Recycle. Now start with the next most important pattern in the vicinity of the node - probably on the other side from the first initiation. Go through the same steps as before, and after that look for the strongest pattern remaining on either side. Continue until you are not finding any repeated patterns, and then look at the remainder. If there are any instances that have not been cited as evidence for at least one hypothesis, examine them to see if they are unusual, or if there is something that this selection is not emphasizing enough. If there are signs of an underlying pattern that has not been brought out by this selection, make a tentative note of it. Step 6. Result. Make a final list of hypotheses and link them in a final report on the node that you started with. Step 7. Repeat. Now gather a new selection from the corpus and start by applying your report to this new data. Go through the same steps, and confirm, extend or revise your hypotheses as you go along.
Cobb (1997) tried to identify a specific learning effect which could be attributed
to the use of concordance software in language learning. He started with a research
question which was an extension of a previous one; will the superiority of concordance
information over a single sentence prevail, if (a) the information appears on a computer
screen instead of on a paper, and (b) the task is not to recall known words but to learn
new ones. The rationale behind this approach was that learning a new vocabulary item
from several contexts tends to produce rich transferable knowledge (Mezynski, 1983
cited in Cobb, 1997). The subjects involved in the study were first year Arabic-speaking
university students in an intensive English learning program. In the process, a suite of
vocabulary learning activities named PET-200 was designed and tested with about 100
learners for a year. A 10,000 word corpus was assembled from learners’ reading
materials. About 250 words that the learners were unlikely to know were chosen for the
study. The five activities included;
1) choosing a definition
2) finding words
3) spelling words
4) choosing words for new texts
48
5) writing words for new texts
In the experimental version of it PET-200 learners were exposed to
concordancing activities through PET-200; in the control version, however,
concordancing exposure was missing. These two versions were run with the subjects on
alternate weeks for 12 weeks in total. That is, one week the students were exposed to
concordancing activities, the next week they were not. Several measures for word
knowledge such as vocabulary level tests and weekly quizzes were used, and a
questionnaire was given to the subjects at the end of the implementation. Eventually,
only 11 out of 100 subjects were chosen randomly for statistical analysis of the results.
The results revealed some steady gains throughout the experiment. The results of
vocabulary level tests, weekly quizzes and interviews also confirmed these gains. When
compared to the results of the previous study, the facilitation of transferable word
knowledge through concordancing activities, be it on paper or on a computer screen,
was confirmed to be advantageous.
In a project, Thurstun & Candlin (1998) used a concordance software,
Microconcord, to introduce students who were unfamiliar with academic discourse to
the most frequent and important aspects of academic English vocabulary. With the
motive to focus on a restricted set of vocabulary items and to provide learners with
intensive exposure to the use of these items, they developed materials which would help
both native speakers of English and non-native ones. As mentioned above, the richness
of variety that the learners would experience was one of the focuses of this project. This
richness was supposed to come from exposure to multiple examples of the same
vocabulary item in context, which would eventually raise awareness as to the
meaningful collocational relationships possible among the vocabulary items.
The Microconcord corpus of academic texts contained 1,016,000 words from a
variety of academic texts, and it was used to establish the frequency of use of particular
items in the corpus. Concordances for 100 items were extracted from the database for
the student research activities and 400 items for learning activities. The important point
here is that the subjects were not expected to understand everything word or concept
presented in the concordancing lines. Instead, it was a familiarization process whereby
the learners got the chance to develop insights concerning the lexical and grammatical
associations among words.
49
The materials developed through Microconcord were put into a series of
vocabulary activities, namely:
Look – Screening for the key words to learn and the other words surrounding it Familiarize – Referring to the concordances to familiarization with the target
word Practice – Trying to remember the target word without referring to the
concordances Create – Trying to create a piece of writing
The results revealed that the both teachers and learners who participated in the
study experienced was a very different and innovative approach to vocabulary learning.
The vocabulary items chosen for the study were also confirmed to have overlapped with
the needs of the subjects. Another important conclusion was that some students might
benefit more from the material if it was presented in a teacher-mediated environment,
while independent study could be an advantage for others. They stressed the idea that
concordancing activities provide learners with multiple examples of the same
vocabulary items in context. Another important point mentioned in the study was that
these activities raised awareness about collocational relationships among these words.
Nation (2001) also stresses the same point and adds that this kind of exposure of
academic vocabulary will enable students to use collocational relations while writing.
In a small-scale project carried out by Weber (2001), the aim was to teach
undergraduate law students to write legal essay by raising the students’ awareness of the
generic and structural features of legal essays. To do this, the researchers collected
samples of legal essays from the University of London LLB Examinations which were
written by native speakers. The students were given access to the essays through
Longman Mini-Concordances and WordSmith Tools and encouraged to investigate
various aspects of these essays. The activities were carried out individually or in small
groups. The task was challenging from the very beginning but the students managed to
identify certain elements of legal essays. With these elements in mind, the students were
asked to read the essays again, in order to identify certain lexical items which correlated
with the generic structures of the texts. From time to time, the students were given the
chance to work on non-legal corpora. At the end of the project, the students were
presented with a number of case studies and were asked to write short essays about
these cases, taking into consideration the generic and structural features which they had
identified before. As a result, they were able to write acceptable essays from linguistic
an
dis
co
aca
rel
lea
ne
vo
Sc
Fig(T
(e.
giv
ma
ten
(C
im
on
2.6
tha
rel
ge
Re
nd legal poi
scussions. W
ncordance
ademic wri
lationship b
arner to kno
ed to raise s
In ord
ocabulary te
chmitt, 2005
gure 6. A saTaken from S
In the
.g. brief an
ven below
akes sense
ndencies tha
Conrad, 200
mprove a mo
ne.
6.1. Learn
The u
at corpus-ba
lated to rea
t a clear pi
eppen (2002
ints of view
With these
and genre w
iting is an
between col
ow the coll
students’ aw
der to visua
eaching, the
5, p. 196).
ample concSchmitt and
activity pre
nd channel),
the target
in two way
at words de
5, p. 397).
ore sophisti
ners’ Attitu
se of corpo
ased studies
al language
icture of wh
2, p. 201) a
w. These e
positive ou
work shoul
issue. Mo
llocational
ocational b
wareness co
alize what k
following
ordancing ad Schmitt, 2
esented in F
, then the l
vocabulary
ys: first of
monstrate.
Secondly, t
cated under
udes toward
ora in EFL
s provide st
as it is use
hat structur
cknowledge
essays were
utcomes, th
ld be an ess
oreover, Wa
and technic
behaviors of
oncerning fr
kinds of co
activity can
activity to te2005, p. 196
Figure 6, the
learner is a
are most l
all, it is a
This pheno
through this
rstanding o
ds Corpora
classes has
tudents with
ed. Through
res are more
e that altho
e peer revie
he researche
sential elem
ard (2007),
cality, state
f words. An
requency an
oncordancin
n be checke
each vocabu6)
e target voc
asked to gue
likely to oc
good demo
omenon is re
s activity, l
f English le
a as a Lang
been detai
h a consider
h corpus-ba
e frequent i
ugh frequen
ewed and s
ers were fu
ment in all E
, who claim
es that it is
nd in his stu
nd collocatio
ng activities
ed as an exa
ulary
abulary is p
ess which o
ccur with t
onstration o
eferred to as
anguage lea
exicon rathe
guage Learn
iled so far.
rable amoun
sed studies
in the real
ncy informa
subjected to
ully convinc
EFL classe
ms that the
s important
udy he stre
ons.
s could be
ample (Schm
presented at
of the collo
them. This
of the collo
s semantic p
arners are l
er than a sim
ning Tool
It is quite
nt of authen
, EFL learn
language. B
ation canno
50
o group
ced that
s where
ere is a
for the
sses the
used in
mitt and
t the top
ocations
activity
cational
prosody
likely to
mplistic
obvious
ntic data
ners can
Biber &
ot be the
51
only factor in material design, this approach provides a more solid basis than relying
only on intuitions and accepted practice. A reliable corpus database available in the
classroom also has the potential to provide more than a native speaker can in terms of
language use. Additional points could be added to this list of benefits. However, the
crucial consideration here is how EFL learners feel about corpus related activities in
language learning. As a matter of fact, EFL learners’ positive attitude towards corpus
use in language classes is reported in the relevant literature (Thurstun & Candlin, 1998
see the previous section; Sun, 2000 and Yoon & Hirvela, 2004).
While developing an internet-based concordance approach to language learning
Sun (2000) investigated Taiwanese EFL students' attitudes toward this learning tool. He
designed a 3-week on-line corpus lesson and implemented it with a sample of 37
college students at a Taiwanese university. He then administered a questionnaire survey
to solicit student's feedback on the web-based concordance. The results of his analyses
indicated that students in the study showed positive attitudes towards using online
concordancing in EFL lessons.
In their survey study, Yoon and Hirvela (2004) tried to determine EFL learners’
attitudes towards corpus use in L2 writing. They investigated the ways in which corpus
use is beneficial for learning L2 writing, difficulties that the learners experience when
using a corpus, how they feel about using corpora in writing instruction and their overall
evaluations of corpus use in L2 academic writing. The participants were intermediate
and advanced learners of English in a writing course in an American university. The
intermediate and advanced groups were involved in corpus-based writing activities in
separate classes. They both worked with the free version of Collins COBUILD Corpus.
The overall approach in the activities was to lead the students through explanation,
demonstration and then production. They were given explicit explanations as to how to
conduct concordance searches and how to interpret results from such searches. Later the
learners were instructed to form prototype strings (a kind of synthesis of concordance
outputs) and present these to the class with discussion. By the end of the term, the
learners had put together a large collection of these strings. Although there were certain
differences between the intermediate and the advanced group, in general the learners
agreed that these activities were beneficial especially in terms of vocabulary and phrase
use and in improving their writing skills. The difficulties that the participants
experienced were mostly related to the fact that these activities were time-consuming
52
and laborious. Most of the students found creating prototype strings helpful in writing
instruction. The last finding revealed that the learners were generally satisfied with the
activities. The follow up interviews concerning the corpus-based activities confirmed
these findings; however, it was revealed that the learners did not make use of prototype
strings in their own writing. The results of these interviews also showed that there was
an increase in learners’ confidence in L2 writing.
2.7. Summary
Seminal works (Silva, 1993; Ferris, 1994, Hinkel 2001b) suggest structural
differences between L1 and L2 written productions at micro and macro levels. These
differences have been well-established with EFL learners from different L1
backgrounds like Chinese, Japanese, Korean, Vietnamese and Indonesian (Hinkel,
2001b); Arabic, Chinese, Japanese and Spanish (Ferris, 1994) and Hungarian Tankó
(2004). However, related literature lacks data concerning such differences in Turkish
EFL learners’ writing.
Although there are contrasting views concerning the differences between L1 and
L2 lexicon (Wolter, 2001), EFL learners’ lexical features still need examining (Hinkel,
2005, p. 615). Moreover, related literature also lacks data concerning Turkish EFL
learners’ lexicon.
Trying to deal with lexicon related problems in EFL learners, Cobb (1997),
Thurstun & Candlin (1998) and Sinclair (2003) have suggested certain models for
concordancing activities, and reported positive results. The collocation activities
suggested by Schmitt and Schmitt (2005, p. 196) are in line with the notions of lexical
networks and cohesion. Furthermore, Thurstun & Candlin (1998), Sun (2000) and Yoon
& Hirvela (2004) have suggested EFL learners’ positive attitudes towards the use of
concordancers in language classes. These points together make up the rationale behind
both the descriptive and experimental phases of the current study.
53
CHAPTER 3
METHODOLOGY
3.0. Introduction
In this chapter, first of all the design of the current study will be introduced; the
rationale behind choosing a mixed research method will be made clear. Then,
procedures concerning the first phase of the study will be detailed. The first phase of the
study comprises descriptive results of the comparisons of learner and native essays.
In the next part, methodological details concerning the quasi-experimental phase
of the study will be given. Data collection tools and data analysis techniques for both
phases will be explained.
3.1. The Design of the Study
In this study, mixed methods were used to collect and analyze data. The
rationale behind this choice was that, as it is stated in Miles and Huberman (1994), it
has now become obvious that there is not much point in polarization of research
paradigms. Qualitative and quantitative paradigms do not have to be dichotomies, they
could as well be supporters of each other and they could be used together to reinforce
research findings. This point of view is also referred to as triangulation, which means
“the generation of multiple perspectives on a phenomenon by using a variety of data
sources, investigators, theories, or research methods with the purpose of corroborating
an overall interpretation” (Denzin, 1978, p. 301) With these notions in mind, both
qualitative and quantitative data collection, analysis tools and techniques were used. In
Table 2, the stages and processes involved in the study and the research paradigm for
each stage are detailed.
54
Table 2
Research Paradigm and Processes for Each Stage of the Study
Stages
Process
Research Paradigm
Stage 1 (May, 2010)
Learners’ essays from the first phase were digitalized and processed in Coh-metrix, and then the results were compared with native groups.
Quantitative
Stage 2 (May, 2010)
Focus group interviews were carried out to find out about the perceptions of the learners concerning the problematic areas in their essays.
Qualitative
Stage 3 (February 2011)
Learners’ essays from the second phase were digitalized, and processed in Coh-metrix for confirmation of the results of the first phase.
Quantitative
Stage 4 (June, 2011)
Experimental group and the control group were compared for recognition and production through pre-tests, post-tests and delayed post-tests.
Quantitative
Stage 5 (June, 2011)
Learners’ perceptions and feelings concerning concordancing activities were analyzed through semi-structured interviews.
Qualitative
The first phase of the study involves both qualitative and quantitative data. At
this stage quantitative data is gathered from the written productions of the participants
and these were processed through Coh-metrix. The results of these comparisons were
discussed with the participants through focus-group interviews and the results of these
interviews compose the qualitative aspect of the first stage.
The second phase of the study also involves qualitative and quantitative data.
Quantitative data in this phase comes from the quasi-experimental process which
involves the statistical analyses of learners’ written texts and test results for vocabulary
recognition. Quasi-experimental design is similar to true experimental design except
that in the latter the participants are selected randomly. Because of practical constraints,
working with unequal groups has become an accepted methodology where random
assignment of the participants is impossible (Dörnyei, 2007, p. 117). According to
Cohen et al. (2007, p. 287) quasi-experiments come in several forms, for example:
55
- Pre-experimental designs: the one group pretest-post-test design; the one group
posttests only design; the post-tests only nonequivalent design.
- Pretest-post-test non-equivalent group design.
- One-group time series.
In the first category, pre-experimental designs, the one group pretest-post-test
design involves only one group and a treatment, and the effects of the treatment are
tested through pre-tests and post-tests. In the one group posttest only design, the effects
of a treatment on one group is tested through only one post-test. In the post-test only
nonequivalent design, two groups with unequal participants are tested through only one
post-test.
The second category includes pretest-post-test non-equivalent group design in
which the groups are not randomized for the number of the participants. The control
group included in the design and testing the effects of a treatment through pre-tests and
post-tests make this design more preferable as it is the most similar one to the true
experimental design.
In the one-group time series, the group is the experimental group, and it is given
more than one pretest and more than one post-test. Several tests and observations are
used both before and after the treatment. Through this design, evaluation of a treatment
can be performed at multiple levels, and it is supposed to increase reliability.
In this study, a pretest-post-test design with two unequal groups in size was
employed. The sampling procedure and descriptive information about the participants
are discussed in the next part.
As for the qualitative data gathered in the second phase of the study, 10 of the
participants from the experimental group were asked to share their thoughts and feelings
about the concordancing activities in semi-structured interviews.
3.2. Sampling and the Participants
In the descriptive phase of the study (2009-2010), initially, the participants were
850 freshman engineering students at University of Gaziantep. Their ages varied from
19 to 23, and most of the participants were male. In order to meet the requirements for a
learner corpus (Granger, 2003), subjects’ proficiency levels were determined using a
valid and reliable placement test (Allen, 1992). The results were checked to see if their
56
levels were homogeneous as would be expected. However, the results of the test
showed that the subjects’ levels varied from A2 (elementary) to B2 (upper-intermediate)
level, which, in our case, demanded adjustments concerning homogeneity. Therefore,
only intermediate and upper-intermediate level learners (49) were involved in the study.
The assumption was that the learners who have proficiency levels lower than
intermediate level would still be dealing with some basic grammar and lexical issues,
which was likely to affect the results negatively. Demographic data concerning the
descriptive phase of the study is presented in Table 3.
Table 3
Demographic Data about the Descriptive Phase of the Study
Group IntermediateUpper-
intermediate Total
Turkish EFL Learners
N % N %
49 37 76 12 24
Native Speakers of English - 100
As is clear from the above table, 37 intermediate and 12 upper-intermediate level
EFL learners (49 in total) participated in the study, and most of them (76%) were at
intermediate level of proficiency; as for the native speakers, 100 essays taken from
LOCNESS (Louvain Corpus of Native English Essays) were used in the comparison
process.
The quasi-experimental phase of the study was carried out during 2010-2011
academic year. After the proficiency test mentioned above was given to three different
groups of learners, two of these groups were selected according to their proficiency
levels. The group which was excluded from the study was at a low-intermediate
proficiency level on average; therefore with the same assumption concerning
homogeneity, this group was excluded from the study. Parallel to the descriptive stage,
learners with proficiency levels lower than intermediate were excluded from the
analysis process, although they were involved in all of the activities carried out during
the experimental stage. Table 4 provides demographic data concerning the experimental
phase of the study.
57
Table 4
Demographic Data about the Experimental Phase of the Study
Group Intermediate Upper-intermediate Group Total Total
Experimental N % N %
18 37
14 78 4 22
Control N % N %
19 17 89 2 11
From the data provided in Table 4, it can be seen that the experimental group
consisted of 18 learners in total and 19 learners participated in the study as the control
group. It is obvious that most of the learners who participated in the experimental phase
of the study (78 % in the experimental and 89 % in the control group) were at an
intermediate level of proficiency, and the rest were at an upper-intermediate level (22 %
in the experimental group and 11 % in the control group).
Both in the descriptive and the quasi-experimental phase of the study including
the interviews, purposive sampling was employed as it is more viable when description
rather than generalization is the goal (Dawson, 2002, p. 49).
3.3. Context of the students
The learners mentioned above were freshman engineering students who had
taken an intensive English preparation class for a year. The preparation program
provides training in four different language skills (listening, reading, writing and
speaking) with a strong emphasis on reading skill. Although all freshman students come
from the same program, sometimes there are significant discrepancies among these
learners in terms of language skills and attitudes. From time to time, exam-smart
learners become freshman students without written or spoken production, as the overall
score obtained from the related tests are insignificant in the grading process.
As an expectation, at the end of this intense program, the students become
freshman with an upper-intermediate level of proficiency at least. Once they enroll in
freshman classes, they have an ESP (English for Specific Purposes) program for four
hours a week. In this program, ESP is carried out with reading and listening skills as the
focal points. The reason for this is to assist the students with their major courses in
58
which they have to deal with a considerable amount of reading from their textbooks and
listening to their lecturers. During the ESP program, the students also study essay
writing, and its purpose is to help the students with their project reports in their major
lessons.
3.4. Data Collection Tools and Procedures
Data needed for quantitative and qualitative analyses was collected through a
number of tools and procedures. There were certain commonalities and differences
between the descriptive and the experimental phases of the study concerning these tools
and procedures.
3.4.1. Descriptive Procedures
In the descriptive phase, after the proficiency levels of the participants were
determined, 49 essays written in a 40-minute midterm exam were collected. In terms of
genre, the collected essays were all argumentative essays, as other genres were likely to
affect both the macro- and micro-linguistic characteristics of the samples (Ellis &
Barkhuizen, 2005, p. 29). The participants were not allowed to use dictionaries or
reference tools. The rationale was that the participants did not have any training about
how to use them effectively. One of the observations at this point was that Turkish EFL
learners embraced rather simplistic views of dictionary use. Generally, when they need
an English equivalent of a Turkish word, they tend to look it up in the dictionary and
accept the first word appearing in the dictionary as the exact match. This situation was
most likely to negatively affect the lexical processes carried out in Coh-metrix, by
yielding inaccurate results.
The topics for the participants essays were taken from ICLE (see Appendix 3).
These essays were digitalized, and during the digitalization process only the spelling
mistakes of the participants were corrected. The participants were also involved in the
process when the correct spelling of a misspelled item could not be determined.
3.4.2. The Selection of Corpora
The selection of corpora was one of the most demanding parts of the current
study. The standard view would be the comparison and then interpretation of two
different sets of corpora. In their studies, Crossley & McNamara (2009) compared L1
and L2 sets of corpora. They concluded that the online database (Coh-metrix) they used
59
in their study is able to distinguish between L1 and L2 texts. In the present study,
another dimension was added to the equation: a third set of L1 texts. Table 5 is a
description of three corpora used in this study.
Table 5
Comparison of Three Corpora Used in the First Phase of the Study
Name of the Corpus
N Total
Number of Words
Average Words
per EssayEssay Type Prompt
Learner corpora 49 16,334 333 Argumentative Exam/Timed
Native corpora 1 54 21,605 400 Argumentative Exam/Timed
Native corpora 2 46 54,397 1182 Argumentative Untimed
The average number of words for LC texts was 333, and for the first L1 corpus it
was 400 words per text; however, the second native corpus had an average of 1182
words per text. All three corpora were composed of argumentative essays, and all essay
topics were taken from Granger (1997). The rationale for adding another native corpus
was that if the mentioned database or software is capable of distinguishing between L1
and L2, then it should not be able to distinguish between the two L1 sets of corpora.
Parametric and non-parametric statistical techniques were conducted with an
expectation that the analysis of variance would differentiate the L2 text sets from L1s.
The corpora mentioned in Table 5 were processed in the online database Coh-
metrix (see Chapter 2 for details). Among the indices present in Coh-metrix, only the
ones directly related to the concerns of the current study were chosen. Indices that were
used in the comparison process are listed in Table 6.
60
Table 6
Coh-metrix Indices Used in the Current Study
Referential and Semantic Indices
Readability Indices Syntax Indices
-Anaphor reference (adjacent) -Anaphor reference (all) -Argument overlap (adjacent) -Argument overlap (all) -Stem overlap (adjacent) -Stem overlap (all) -LSA sentence (adjacent) -LSA sentence (all)
-Flesch Reading Ease Score -Flesch-Kincaid Reading Ease Score
-Personal pronouns -Type-token ratio -All connectives -Number of words before the main verb
Including syntactic and readability indices in a study related to lexical networks
of EFL learners might sound like a deviation in scope; however, it has now been well
established that lexicon is not free from syntax (Mahlberg, 2009, p. 108). Syntactic
issues should be covered to some extent if sound analyses are needed concerning L2
lexicon.
Among these three groups of indices, the syntax indices are the ones most
sensitive to total words used in a text, as they mostly include incidence scores. This is
why the comparisons concerning syntax indices were conducted between the learner
group and the native group with relatively similar average words per text (LC=333 and
NC=400, see Table 5). When the digitalizing process was over, these texts were
processed via the online database, Coh-metrix. This process took about two months,
from April to May 2010.
This database consists of inquiry segments requiring proper input, such as Title,
Genre, Source, Job Code, LSA Space and the text entry space. The Title, Source and
Job Code bars are used for categorizing the outputs. In the Genre drop-down selection
bar there are three options, namely Science, Narrative and Informational. This section is
used to determine the discourse settings of a text. The last drop-down selection bar is
61
used to determine the LSA space of the texts like College Level, Narrative,
Encyclopedia and Physics. The Coh-metrix input screen is demonstrated in Figure 7.
Figure 7. Coh-metrix input screen The Coh-metrix database was used to process learners’ written productions in
both descriptive and experimental stages, “Informational” was chosen for the Genre,
while “College Level” was chosen for the LSA space as these two were the most
appropriate options for the participants.
After submission, depending on the length of the text input, in a couple of
minutes an outcome screen pops up and the results for each of the indices are exhibited
(see Figure 8). These outcomes are converted into MS Office Excel sheets and then to
SPSS datasets where the values are ready for statistical analyses.
Figure 8. Sample coh-metrix results for a text input
62
The same procedure was carried out for the L1 texts taken from LOCNESS.
Once the Coh-metrix outcomes of the participants’ data were compared with L1 texts,
the results were shared with the students via focus group interviews.
3.4.3. Semi-Structured Focus Group Interviews
Contrasted with traditional interview, focus group interview involves a group
format with 6-10 participants. It is neither a one-on-one interview, nor a group
discussion, but rather a sort of idea-sharing or brainstorming session about issues that
are relevant to the entire group. The interviewer acts as the moderator, asks a number of
questions to be discussed and takes notes or tape records the discussion in order to
systematically analyze it. The interview generally takes about one hour and is
performed with at least four groups (Dörnyei, 2007, p. 144).
The data analysis process is, in fact, a painstaking procedure. Ritchie & Spencer
(1994) developed a technique called ‘framework analysis’, which could be used for both
individual and focus-group interviews. This analysis technique comprises five main
steps: familiarization, identifying a thematic framework, indexing, charting, mapping
and interpretation.
The familiarization step involves reviewing the points made during the interview
a couple of times. At this stage, certain patterns begin to emerge. In the next step,
categories and a thematic framework are determined; ideas, concepts and commonalities
are watched for. At the indexing stage, these ideas, concepts and commonalities are
sifted. At the charting stage, quotations are separated from the context and the rest of
the data is rearranged to form a unified theme. This theme is mapped with the research
questions in mind; in other words, the researcher tries to relate the thematic framework
with the research questions at hand. At the last stage, the data is interpreted.
In the descriptive phase of the study, five different groups composed of 30
students in total were involved in the interviewing process. The participants were asked
semi-structured questions, and were asked to reflect and comment on the questions.
They were clearly informed that they were not supposed to solve the problems which
came to surface but rather create a consensus concerning the problems. The interview
was digitalized and analyzed through the framework analysis mentioned above.
63
3.4.4. Experimental Procedures
Two groups of intermediate and upper-intermediate learners, 37 in total,
participated in the experimental part of the study. In this phase of the study, four
research questions mentioned in the introduction were resolved. In order to test the
effectiveness of concordancing activities in vocabulary recognition, a multiple choice
test was constructed.
3.4.5. Piloting of the Multiple Choice Test for the Target Vocabulary
Initially, the test was made up of about 65 multiple choice questions, and it
contained vocabulary items to be taught in that semester. It was piloted with 54
intermediate and upper-intermediate EFL learners from other groups which were not
involved in the study. Their proficiency levels were determined with the same test that
was used in the first phase of the study (Allen, 1992). After the pilot study, the test was
analyzed by three EFL instructors working at the same department. Some questions
were extracted from the test and 50 items were left (see Appendix 2). The reason for the
extraction was that some of the items were found to be too easy or too difficult by the
instructors and the item analysis results confirmed some of these opinions. The results
of this item analysis are presented in Table 7.
64
Table 7
Item Analyses for the Target Vocabulary Test Used as the Pre-test and the Post-test
Item no. Discrimination Difficulty
(%) Item no. Discrimination Difficulty
(%) Item 1 .44 44 Item 26 .22 64 Item 2 .22 57 Item 27 .39 55 Item 3 .22 40 Item 28 .22 62 Item 4 .28 37 Item 29 .22 68 Item 5 .22 37 Item 30 .33 24 Item 6 .33 62 Item 31 .22 18 Item 7 .22 29 Item 32 .22 50 Item 8 .33 48 Item 33 .39 40 Item 9 .22 27 Item 34 .39 48 Item 10 .28 35 Item 35 .28 46 Item 11 .22 51 Item 36 .67 53 Item 12 .56 75 Item 37 .28 64 Item 13 .72 64 Item 38 .28 18 Item 14 .28 16 Item 39 .56 61 Item 15 .44 35 Item 40 .39 29 Item 16 .33 31 Item 41 .22 27 Item 17 .39 46 Item 42 .33 48 Item 18 .67 61 Item 43 .33 51 Item 19 .39 64 Item 44 .22 55 Item 20 .22 25 Item 45 .28 46 Item 21 .56 38 Item 46 .56 48 Item 22 .44 48 Item 47 .22 37 Item 23 .39 40 Item 48 .22 18 Item 24 .44 40 Item 49 .22 44 Item 25 .28 38 Item 50 .56 68
Item analysis for the target vocabulary test, which was used as the pre-test and the
post-test, is given in Table 7. Items with difficulty levels below 15 and above 90, and
items with discrimination levels below .20 were removed from the test. When the
vocabulary items being tested are taken into consideration, these items were assumed to
be new to the learners, this situation might have caused some of the items’
discrimination and difficulty levels to be around an almost acceptable level.
For further analyses, scores concerning mean, standard deviation, mode, median,
and item facility were calculated. In addition, in order to check the inter-item
correlations for reliability of the test (Mackey & Gass, 2005, p. 130), Kuder-Richardson
formula 21 was calculated by using the formula presented below.
65
(k) (1- X (k-X)) R = ___________________ k-1 kS where R= test reliability k = number of items on the test X = mean of raw scores from the total test S = variance from the raw scores of the total test
Related results concerning the pilot study for the target vocabulary test are
presented in Table 8.
Table 8
Results of the Pilot Study for the Target Vocabulary Test
Items N x ̄ sd Mode Median Item Facility Mean (%) KR 21
50 54 22.41 7.66 26 22.50 44.60 .81
It is clear from Table 8 that the results of the 50-item target vocabulary test,
which was given to 54 subjects, yielded a mean score of 22.41 (out of 50). The mean for
item facility appears to be 44.60, and the Kuder-Richardson score is .81 which could be
regarded as quite acceptable (ibid., p. 130). After the flaws in the test were fixed, it was
given to both the control and the experimental group in the second week of February
2011 as the pre-test.
3.4.6. Pre-test for Written Production
When the pre-test for recognition of the target vocabulary items was completed,
the participants in both groups were required to write argumentative essays about the
same topics used in the first phase of the study. These essays were digitalized by the
students themselves, and they were allowed to make spelling corrections in their essays.
After spelling corrections, the essays were processed in Coh-metrix to compare the
results of this group with the previous one in the descriptive part of the study. The
reason for such comparison was to confirm the prominent differences between L1 and
L2 written productions that were determined in the first phase.
3.4
ma
co
in
nu
Fig
ke
sta
sel
tar
Ye
Fig
4.7. Vocabu
In the
aking use
rpora like B
total. The
umber of lin
gure 9. Inpu
First, t
y words are
age in line w
lected (two
rget item, th
es and No as
gure 10. Co
ulary Teach
experimen
of online
BNC and B
input scre
nes, gapped,
ut screen fo
the lexical i
e the items
with the lea
o million w
he option ne
s the option
oncordancin
hing Mater
ntal process
corpora at
rown’s spok
een of the
scan for an
or the target
item must b
to be taugh
arners’ cont
words in tota
ext to the G
).
ng screen fo
rial
, vocabular
http://www
ken and wr
database in
ny recurring
vocabulary
be entered in
ht. Next, a c
text, only w
al). To cre
Gapped? is s
or the target
ry teaching
w.lextutor.c
ritten corpus
ncludes var
g word and l
y items
nto the Key
corpus is ch
written corp
ate concord
set to Yes (s
vocabulary
materials w
a. This we
s with abou
riables like
line width (
yword(s) bar
osen from t
pora of BNC
dance lines
see Figures
y item (not g
were develo
ebsite inclu
ut 15 million
keyword,
(see Figure 9
ar; in our ca
the options.
C and Brow
with gaps
10 and 11 f
gapped)
66
oped by
udes 19
n words
corpus,
9).
se these
. At this
wn were
for the
for both
67
In Figure 10, the word example is displayed in concordancing lines. To make a
learning activity out of these lines, the target vocabulary item is left blank for the
students to fill. An example is given in Figure 11.
Figure 11. Concordancing screen for the target vocabulary item (gapped)
The next important option to be considered at this stage is the Scan for any
recurring word option (see Figure 9). In this option, rather than the immediate
collocations, the potential collocations of a given word in a certain range between 5 and
15 are determined. This option determines the range of the potential collocations for the
keyword input. In the same variable, there is also a frequency option with a value
between 4 and 10. This option determines the frequency of the potential collocations for
the given keyword. For example, if the user sets the range of occurrence as 5 words,
and the occurrence frequency to 4 (these are the default values) the database makes
inquires within these values. Figure 12 illustrates an example for these options.
Figure 12. Word ranges for both sides of the target vocabulary item
Along with the concordancing lines, other inquiry results are presented at the
bottom of the screen. A set of potential collocations for the word example is shown in
Figure 13.
68
Figure 13. Immediate and all potential collocations for the target vocabulary item
In Figure 13, the first group of words titled as ‘left immediate collocations for
example’ reveals the immediate surroundings of this word. The values next to each item
represent word frequency, i.e. how many times it occurs next to the key word. In this
example, the word for occurs for 437 times next to the word example on the left. The
option left in this example is again a default value which can be set as right or either
sides alternatively. The words are lined up from greater to smaller according to their
frequency values.
In the second group of words entitled ‘all potential collocations for example’
you can find potential collocations of the word example within a range of five words
from either side. Again, next to each item there are frequency values lined up from
greater to smaller.
Classroom teaching materials were constructed by making use of potential
collocations of target words within a certain range (not immediate collocations) and
gapped concordancing lines. Function words in the potential collocations part of a target
word were eliminated as they carry no or too little meaning related to cohesion or
coherence. Proper nouns and abbreviations were also eliminated from the list. The
content words left from these extractions were used in the preparation of worksheets.
3.4.8. Classroom Procedures
The experimental part of the study took about 10 weeks in total. Both groups had
four hours of instruction per week. In the process, the control and experimental groups
had the same teaching materials involving the pre-determined vocabulary items to be
focused on. These items were present in meaningful contexts like reading passages. For
the experimental group, these items were taught in three stages in accordance with the
principles proposed by Sinclair (2003) and Thurstun & Candlin (1998). In line with
common vocabulary teaching/learning procedures, pre, while and post activities
69
Thornbury, 2002, p. 75) were also present in the process. Guessing the meanings of
words from the context which they are used in is one of these procedures (ibid., p. 148).
The positive results of negotiation among the learners of the meaning of the
vocabulary to be learnt (Nation, 2002, p. 269) was also taken into account and it was
realized at the end of each activity as it will be explained later in this chapter.
In the first stage, the items were emailed to the students before class as a simple
matching activity. In the activity, the target vocabulary items were present as a column
on the left with their immediate collocations jumbled in a column on the right (see
Appendix 4 for a sample used in this study; see Schmitt & Schmitt, 2005, p. 196 for
other samples). The participants in the experimental group were expected to match up
the target words with their potential collocation, and they were allowed to use
dictionaries at this stage. This is, in fact, a derivative of guessing word meaning from
the context, which was explained previously.
In the second stage of vocabulary learning activities, the matching items that the
participants were assigned as homework were discussed at the beginning of the lesson
during which the items were exposed to the participants in meaningful contexts
(Thornbury, 2002, p. 76). During these discussions, the participants stated their ideas as
to the connections between the vocabulary items and their collocations. These
discussions took about ten minutes and were also useful as warm-up activities.
During the two-hour lessons, participants analyzed the vocabulary items mostly
via reading passages. Sometimes, these items were presented through listening
activities. The participants were encouraged to analyze and discuss the connections
between the target words and the lexical items surrounding them. Generally, there was
an overlap between the matching activity carried out at the beginning of the lesson and
the contexts in which the items were presented. That is to say, it was possible to see a
target vocabulary item along with the same collocations in a reading passage as in the
matching activity.
Another point which was discussed during these sessions was the overlapping
structures in concordancing lines. For example, the overlapping structures between
sentence pairs in these lines were emphasized, and the ties that would be created using
such structures were mentioned.
70
As the final wrap-up activity, the participants were given worksheets involving
the same vocabulary items, but as a concordancing activity this time. This last activity
(see Appendix 4 for a sample) involved about five concordancing lines per item with
the target vocabulary items left blank. The reason for limiting the concordancing lines to
five is that over-exposure might somehow tire the learners if the activities are solely
based upon deduction from concordance lines (Thurstun, & Candlin, 1998). These lines
were selected by the researcher from hundreds of concordancing lines so as to best
match the participants’ technical contexts.
3.4.9. LSA Scores
The first argumentative essays written by the experimental and the control group
were processed in Coh-metrix and the results were recorded as the pre-test scores. At
the end of the experimental procedures which took about 10 weeks, both groups were
given the same topics on which to write their second argumentative essays; again their
essays were processed in Coh-metrix and these results were recorded as the post-test
scores. As the last step, about two weeks after the last activity, a final writing exam was
given to the participants and they wrote about the same topics. The essays were
processed in the same way as the pre-test and the post-test, and the results were
recorded as the delayed post-test scores.
3.4.10. Target Vocabulary Density Scores
In an attempt to determine whether concordancing activities induced production
of target vocabulary items in writing, learners’ essays were checked for these items.
Each essay produced by the learners from both groups was checked for the target
vocabulary items. These items, measured as number of types per 1,000 words, were
quantified for density scores. In the measuring process, the AWL counter in the
following website was used.
71
Figure 14. The AWL highlighter input screen (Taken from http://www.nottingham.ac.uk/~alzsh3/acvocab/awlhighlighter.htm)
The AWL highlighter illustrated in Figure 14 is a digital database that was
designed to count academic vocabulary load in a given text. The user can paste or write
manually in the input section provided, and any text up to 2400 characters is processed
for AWL. Since there are 10 sub-lists in the main list, the user is also provided with the
sub-list option. In order to cover all the items in the list, option 10 was set to default for
all computations. After the submission, a screen pops up and provides the user with a
screen where the words from AWL are highlighted.
Figure 15. The output screen for the AWL count
The highlighted words are counted as types per 1000 words. That is, each word
from the AWL list is counted as one, and multiple incidences of the same word are not
taken into account. Some of these items were not in the items to be taught in the
experimental phase; therefore, after the counting, the irrelevant items were detected and
removed from the list. This assessment of learners’ essays was before the beginning of
concordancing activities (pre-test), about a week after the last concordancing activity
72
(post-test), and about two weeks after the post-test to examine the delayed effects of the
concordancing activities. Since 37 learners in total participated in the study, 111 (37 x
3) essays were processed in the same way.
3.4.11. Semi-structured Interviews
In the final stage of the study, participants’ perceptions and feelings concerning
the experimental stage were analyzed through semi-structured interviews. The aim of
these interviews was to determine whether or not concordancing activities had any
effects on the experimental groups’ perceptions on vocabulary learning. 10 participants
were chosen randomly from the experimental group, and were asked semi-structured
questions (see Appendix 6 for the English translations of the questions). Each session
took about 10-15 minutes. During these sessions, the participants were asked to review
what had been done during lessons concerning concordancing activities. Once the
participants clearly remembered details about these activities, further questions were
asked and the participants evaluated these activities and shared their feelings and
perceptions. The sessions were recorded and transcribed right after the interviews so as
to make interpretations about participants’ comments more accurate. As mentioned
earlier, the interviews were analyzed through the framework analysis developed by
Ritchie & Spencer (1994).
3.5. Overview of the Descriptive and Experimental Procedures
The current study started in 2010 by giving a placement test to the learners who
participated in the descriptive phase. Then the argumentative essays written by these
learners in an exam were collected, digitalized, and processed in Coh-metrix. Following
the analysis of these essays, focus group interviews were carried out with the same
learners in five different groups.
The experimental phase of the study started in March, 2011 again with a
placement test. By taking the results of this test into account, two groups were
determined randomly as the experimental and the control group. These groups were
given a vocabulary recognition test and the results were recorded as the pre-test for
vocabulary recognition. At the same time, the participants were given a written test and
the essays were collected and digitalized. They were then processed in Coh-metrix for
confirmation of the results from the descriptive phase and recorded as the pre-test for
production. After these processes, concordancing activities started with the
73
experimental group, and they lasted for 10 weeks, and 15 activities were carried out in
total. At the end of the activities, both groups were given the same vocabulary
recognition test as the post-test. They were also given a written post-test, and the essays
written in this exam were digitalized and processed in Coh-metrix. Since the production
of newly learnt vocabulary items are supposed to take some time, the participants were
given the same written test as the delayed post-test after about two weeks. Once the
concordancing exercises were over, 10 students who were chosen randomly from the
experimental group were interviewed.
3.6. Overview of the Statistical Techniques Used in the Study
Both the first phase (descriptive) and the second phase (experimental) of the
current study involve quantitative analyses of the data gathered. In the descriptive
phase, while comparing mean scores of two groups, t-test was used if the concerning
scores had normal and homogeneous distributions (Field, 2009, pp. 325-329); Mann
Whitney U-test was used if these two requirements were violated (ibid., p. 540). While
trying to compare three groups, one-way ANOVA was used if the scores had normal
and homogeneous distributions (ibid., p. 359); if the situation was otherwise, Kruskal
Wallis test was used (ibid., pp. 559-560). In the experimental phase, where a pre-
test/post-test and pre-test/delayed post-test design was used, ANCOVA test was
employed. The rationale behind this choice was that ANCOVA is one of the most
effective ways to control other variables that are likely to affect the dependent variable
(ibid., p. 396).
As a demonstrative summary, in order to help the reader with a quick reference,
all of the procedures that were carried out for about two years are provided in Table 9.
74
Table 9
Summary of the Procedures Followed throughout the Study
Descriptive Phase Procedures
March, 2010 -Placement tests were given to 850 learners.
April, 2010 -Argumentative essays written by intermediate and upper-
intermediate learners (49) were collected.
May, 2010
-The essays were digitalized and processed in Coh-metrix.
-Native essays taken from LOCNESS were also processed in
Coh-metrix.
-Results gathered from learner and native essays were compared.
-Focus group interviews were carried out with 30 learners.
Experimental Phase
February, 2011
-Placement tests were given to three groups of learners (80 in
total).
-Experimental and control groups were determined (18 in the
experimental, 19 in the control group).
-Pre-test for vocabulary recognition was given to both groups.
-The groups were given a written test and the essays were
collected and digitalized as the pre-test for production.
-The essays were processed in Coh-metrix for confirmation with
the results from the descriptive phase.
-Concordancing activities started with the experimental group.
May, 2011
-Post-test for vocabulary recognition was given to both groups.
-The groups were given a written test as the post-test and they
were digitalized and processed in Coh-metrix.
June, 2011
-The groups were given a written test as the delayed post-test and
they were digitalized and processed in Coh-metrix.
-Semi structured interviews were carried out with 10 participants
from the experimental group.
In this section, methodological issues related to the current study were
discussed; the procedures and processes that were followed for about two years were
also made clear. In the next section, the results gleaned from the procedures and
processes that were demonstrated in Table 9 will be analyzed and discussed.
75
CHAPTER 4
RESULTS
4.0. Introduction
This chapter presents the results of the descriptive and experimental phases of
the study. First, descriptive results comparing Turkish EFL learners and two groups of
native speakers of English in terms of referential and semantic aspects are presented,
and then readability scores are discussed. Next, the syntax-related results comparing
Turkish EFL learners with the similar native group in terms of average words per essay
are detailed. Later, focus-group interview results concerning these comparisons are
illustrated.
In the next part, in order to check whether concordancing activities had any
effects on vocabulary recognition, results concerning the experimental phase of the
study are discussed. The experimental and control groups are compared in terms of
recognition of target vocabulary items. Then, target vocabulary density mean scores of
the groups are analyzed. In addition, to see the effects of concordancing activities on
lexical cohesion, the groups are compared in terms of LSA scores. Next, learners’
perceptions and thoughts about the concordancing activities in which they participated
for about ten weeks are discussed. Lastly, the results are summarized by taking both
quantitative and qualitative aspects into consideration.
4.1. Results of the Descriptive Phase
In the analysis process, referential and semantic comparisons were carried out
among three groups (Learner, Native 1 and Native 2). However, the indices in the
syntax index set are highly sensitive to the number of average words used in a text.
Therefore, comparisons concerning syntax were performed between two groups with
similar number of words per text (Learner and Native 1, see Table 5 for details).
The results of statistical analysis revealed that there are certain significant
differences between learner and native texts. These differences included the following
indices: Referential and Semantic (anaphor reference, stem overlap and LSA scores),
Readability Scores (Flesch Reading Ease Score and Flesch-Kincaid Reading Ease
76
Score), and Syntax (personal pronoun incidence score, connectives, and type-token ratio
and the number of words before the main verb).
4.1.1. Normality and Homogeneity of the Data
Since the study contains multiple groups, group scores were tested to see if they
are suitable for parametric comparisons before data analysis began. It is common
knowledge that in order to be able to make use of parametric tests and make inferences
regarding their results, the scores gathered from the subjects must be normally
distributed. With very large populations normality is generally not a concern. However,
if the population is not that large, the assumption that the subjects’ scores are distributed
equally has to be tested. The next concern about parametric comparisons is the
homogeneity of variance of the scores; it is required that the variances should be the
same throughout the data. Generally, the data to be used in any study comes from
different populations; if the score variances of these groups are homogeneous, then
these groups are suitable for parametric comparisons.
In order to test normality and homogeneity of the participants’ scores, frequency
measures and Levene’s test of homogeneity of variance were performed. The results
concerning the three groups (Learner, Native 1 and Native 2) are exhibited in Table 10.
77
Table 10
Normality and Homogeneity Results for the Referential & Semantic and Readability
Indices
Name of the index
SD
Skewness
Standard
Error
z
Levene
test Referential and semantic aspects
Adjacent anaphor reference .138 .370 .196 1.887 .028**
Anaphor reference .080 .935 .196 4.770* .008**
Adjacent argument overlap .145 -.412 .196 2.102* .497
All-distance argument overlap
.133 .516 .196 2.632* .013**
Adjacent stem overlap .170 -.131 .196 .668 .032**
All-distance stem overlap .160 .372 .196 1.897 .032**
Adjacent LSA .057 -.228 .196 1.163 .985
All-distance LSA .060 .216 .196 1.102 .487
Readability
Flesch reading ease score 9.762 -.237 .196 1.209 .003**
Flesch-Kincaid reading ease score
1.98 -.943 .196 4.811* .057
*Values greater than 1.96 are significant at .05 level. **Significant at .05 level
In Table 10, the first column represents the name of the index from Coh-metrix.
The second column indicates the standard deviation values. In the third column,
Skewness values are indications of how much the score distributions are skewed
compared to a perfectly distributed ideal. The z value is the result obtained from the
division of Skewness value by the standard error value and is presented in the fifth
column. If the score obtained from this division is greater than 1.96, which is taken
from the normal distribution table, it means that the scores are not normally distributed.
The last parameter to be checked, Levene’s test of homogeneity, is given in the final
column. This test checks whether the variance of the scores of a given population are
78
homogeneous or not. If these values for a certain index are significant (p<.05), then it
can be claimed that the groups are not suitable for parametric comparisons.
A quick glance at Table 10 will make it clear that nearly all of the indices violate
either the normality or homogeneity assumption; or in some cases both assumptions are
violated (e.g. all-distance anaphor overlap). However, LSA scores, both adjacent and
all-distances, appear to have normal distributions and homogenous variances.
Taking all of the above analyses into account, as the group scores are neither
normally distributed nor homogeneous, the Kruskal Wallis test, a non-parametric test
for multiple groups, will be employed in comparison of the three groups. Since the total
participants in this part of the study is relatively large (NTotal=149), the Monte-Carlo
method will be employed to determine the exact significance values for each of the
comparison done by using the Kruskal Wallis test. Since there are three groups to
compare, a post-hoc test will be necessary to see which of them will be excluded from
the group. At this point, carrying out a post-hoc test is not as easy as it would be in
parametric tests although it is not impossible. Among the options for a post-hoc test for
non-parametric analyses, carrying out binary Mann Whitney U-tests for each of the
groups is an option. That is, the first, second and third group will be compared in pairs:
1-2, 1-3, and 2-3. The differences will be examined to see if any of the group scores are
significantly different from the others. The catch at this point is the liability to Type 1
error, which is to believe that there is a genuine effect in our population when, in fact,
there is not. To overcome this issue, Bonferroni correction is performed (Field, 2009, p.
565). This correction method is basically a restriction of the critical value to avoid Type
1 error. This is done by simply dividing the critical alpha value (.05) by the number of
the groups involved in the study. For instance, when there are three groups at hand and
they are to be compared by non-parametric tests, the critical value changes from .05 to
.0167 (.05/3=.0167). The interpretations as to the significance of the outcomes are
performed taking .0167 as the critical alpha value instead of the standard .05.
Regarding the comparison of the two groups (Learner & Native 1) with similar
number of words per text, the scores of the two groups were, again, checked for
normality and homogeneity with the same rationale mentioned before. The results of
these tests are given in Table 11.
79
Table 11
Normality and Homogeneity Results for the Syntax Indices
Syntax Indices
SD
Skewness
Standard
Error
z
Levene Test
Personal pronouns 35.599 .59 .238 2.478* .004**
Type-token ratio .1063 -.966 .239 4.041* .179
All connectives 19.081 .714 .238 3* .003**
Number of words before the main verb
1.433 .910 .238 3.823* .151
*Values greater than 1.96 are significant at .05 level. **Significant at .05 level
The results of these tests, exhibited in Table 11, show that none of the indices
from Coh-metrix are suitable for parametric comparisons of the two groups. The scores
either violate the normality of distribution or the homogeneity of variance assumption.
These violations require a non-parametric approach for comparison. Since there are two
groups to be compared by syntax index set, the Mann Whitney U-test, a non-parametric
test for the comparison of two groups, will be employed.
4.1.2. Referential and Semantic Indices
Among the indices mentioned before, the first index from the referential and
semantic index set to demonstrate significant difference between L1 and L2 texts was
the anaphor reference index for adjacent sentences. This index calculates the references
occurring in sentences next to each other. Descriptive results for this index are provided
in Table 12.
Table 12
Descriptive Results for Anaphor Reference for Adjacent Sentences
Group N x̄ sd
Learner (L) 49 .401 .140
Native 1 (N1) 54 .312 .143
Native 2 (N2) 46 .294 .101
80
Descriptive results provided in Table 12 clearly show that the learner group (L)
scored higher (x̄L=.401) than the two native groups (x̄N1= .312, x ̄N2= .294). To check if
this difference is statistically significant, a Kruskal Wallis test was conducted. The
results are revealed in Table 13.
Table 13
Kruskal Wallis Test Results for Anaphor Reference for Adjacent Sentences
Group N Mean Rank df x2 p Post-hoc
Learner 49 97.47 2 19.841 .000 L>N1&N2
Native 1 54 64.83
Native 2 46 63
The results of the Kruskal Wallis test for adjacent anaphor reference for the three
groups are displayed in Table 13. The difference among the groups appears to be
statistically significant [x2 (2) = 19.841, p< .01]. The post-hoc test result reveals that this
difference is between the learner and the native groups (L>N1&N2). This means that
the learner group makes use of referential tools much more than the native groups,
regardless of the number of words used in the texts.
The next referential index is all-distance anaphor reference. This index takes
into account the references in a given text for all distances, which means that it counts
referential incidences across the entire text, be it the second sentence or the last one.
Table 14 provides descriptive results for this index.
Table 14
Descriptive Results for All-distance Anaphor References
Group N x̄ sd
Learner (L) 49 .196 .086
Native 1 (N1) 54 .122 .066
Native 2 (N2) 46 .145 .056
81
Descriptive results for all-distance anaphor reference are displayed in Table 14.
It is clear from the table that the learner group scored higher in this index (x̄L=.196) than
the native groups (x̄N1=.122, x̄N2=.145). In order to determine if this difference is
statistically significant, a Kruskal Wallis test was performed and the results are provided
in Table 15.
Table 15
Kruskal Wallis Test Results for Anaphor References
Group N Mean Rank df x2 p Post-hoc
Learner 49 102.40 2 29.551 .000 L>N1&N2
Native 1 54 62.94
Native 2 46 59.98
The results of the Kruskal Wallis test for the three groups (one learner, two
native), in terms of all-distance anaphor reference, are presented in Table 15. The
analysis of the results reveals that there is a statistically significant difference among
groups [x2 (2) = 29.551, p< .01]. The result of post-hoc test indicates that the there is a
statistically significant difference between the learner group and the two native ones.
When we refer back to the mean scores displayed in Table 14, it is obvious that the
learner group scored higher than both of the native groups (x̄L=.196, x̄N1= .122, x̄N2=
.145). This outcome indicates that in learner texts there is a plethora of references, even
when compared to texts which were written by native speakers of English and which
have significantly higher numbers of words.
The next index in Coh-metrix related to referential aspects is the argument
overlap for adjacent sentences. This index calculates the overlapping arguments (nouns,
verb etc.) in a given text. There are two indices which calculate argument overlaps: one
does adjacent overlaps and the other does all overlaps across the texts. This is a
proportion score, so the adjacent overlap index yields ratio scores of argument overlaps
between adjacent sentences. Descriptive results concerning argument overlaps between
adjacent sentences are displayed in Table 16.
82
Table 16
Descriptive Results for Argument Overlap for Adjacent Sentences
Table 16 reveals descriptive results for the adjacent argument overlap scores.
The means of the three groups appear to be similar (x̄L=.543, x̄N1= .562, x ̄N2=.547). A
Kruskal Wallis test was performed to determine the statistical difference among the
groups and the results are shown in Table 17.
Table 17
Kruskal Wallis Test Results for Argument Overlap for Adjacent Sentences
Group N Mean Rank df x2 p
Learner 49 71.94 2 1.003 .613
Native 1 54 79.68
Native 2 46 72.77
Results concerning the adjacent argument overlap are shown in Table 17. The
analysis of the results indicate that there is, statistically, no significant difference among
the groups [x2 (2) = 1.003, p> .01]. This could mean that there is the same amount of
adjacent argument overlaps in the texts of the learners and the native groups. These
overlaps are, in fact, related to repetitions of nouns, verbs, noun phrases etc. There is
nothing surprising about this same amount of repetition appearing in adjacent sentences
of all three groups as they are all trying to relate to the essay topic provided. The next
index, all-distances argument overlap, calculates these repetitions across the entire text.
Descriptive results are revealed in Table 18.
Group N x̄ sd
Learner (L) 49 .543 .147
Native 1 (N1) 54 .562 .156
Native 2 (N2) 46 .547 .122
83
Table 18
Descriptive Results for All-distance Argument Overlap
Group N x̄ sd
Learner (L) 49 .446 .134
Native 1 (N1) 54 .449 .148
Native 2 (N2) 46 .437 .096
Descriptive results exhibited in Table 18 indicate that all-distance argument
overlap scores for the three groups are quite similar (x̄L=.446, x ̄N1= .449, x ̄N2= .437). To
verify this similarity Table 19 should be checked for the results of Kruskal Wallis test.
Table 19
Kruskal Wallis Test Results for All Argument Overlap
Group N Mean Rank df x2 p
Learner 49 73.700 2 .125 .941
Native 1 54 76.600
Native 2 46 75.500
The Kruskal Wallis test scores concerning all-distance argument overlap for the
three groups are displayed in Table 19. Again, as in the adjacent overlap scores, there
seems to be no significant difference among groups [x2 (2) = .125, p> .01]. The
similarities among the groups in terms of adjacent and all-distance scores could be
regarded as quite normal since the subjects were given certain topics to write about and
they were required to stick to them. This restriction is likely to be the cause of lexical
repetitions and therefore overlaps appear between adjacent and distant sentences. The
following excerpts taken from both the learner and the native corpora are likely to
demonstrate this point.
84
An excerpt from an essay about computers written by B. İ. A. from the learner
group:
Although the first scientific computers were only used for calculations or encryption, these computers made life easier in some ways. From that time on, scientists were able to make hard calculations in seconds. These days we use computers more often or always for scientific reasons and in our life for daily works even it is ordering a pizza for lunch. We do not need to go out if we have a computer and internet connection. We can make all of our work by using technology.
An excerpt from an essay about computers which was written by another native
speaker of English:
Native excerpt 1:
The impact of computers on the world has been great. They have changed the way people do business and have radically altered the way data and information is dealt with. In short, the productivity of people has increased ten-fold. As we move into the 2st century this fact will become more important. The amount of information that is available in the world will require the use of computers to organize and extract that which is of interest. The personal computer has, and will continue to play a major role in our lives.
An excerpt from an essay about wars written by A. E. T. K. from the learner
group:
Some of people believe that the number of soldiers is the most important factor of a victory. Yes, number is important but not the most important one. In ancient world yes it was the most important one but today’s world; wars don’t starts at an isolated area, wars are all in countries. And soldiers don’t shoot each other. Tanks, airplanes, helicopters make the war. So if you have your well-developed war industry you can get the victory!
An excerpt from an essay about wars which was written by a native speaker of
English:
Native excerpt 2:
The twentieth century has seen more wars than any other previous bloc of time. Though advances in communication, transportation, and information sharing, the world as a body of people living in close proximity, has rapidly shrunk. In the second of the world wars, the race was on to create the most devastating, most powerful, and most frightening weapon our people had ever known. In my opinion, the discovery and harness of atom and its energy and the corresponding invention of nuclear weapon
85
have been the most significant factors of change in our lifetimes, if not, perhaps, in several lifetimes.
The excerpts presented above are composed of five sentences each. From these
excerpts, it can be clearly seen that both the learners and the native speakers of English
use nearly same amount of argument repetitions. In the excerpt about computers written
by a learner the words scientific, calculations and we are repeated throughout the
paragraph. In the next paragraph about the same topic written by a native speaker the
words computers, people and information are repeated. In the excerpt about wars
written by another learner, the words number, soldiers, important, war, wars and victory
are repeated; and in the native paragraph written about the same topic the words wars,
world, people and weapon are repeated. These repetitions are regarded as argument
overlaps. As was mentioned before, these overlaps seem to be occurring in nearly the
same amount in both groups’ texts, and this is confirmed with the statistical analyses
presented previously in this chapter.
The next index in Coh-metrix is related to stem overlaps between adjacent
sentences. In this index, the part of speech aspect of lexical items enters the scene. It
means that overlapping lexical items with common word roots are taken into account.
For example, one sentence might include the word lose and the next sentence might
include the words losing or lost. This incidence is counted as an adjacent stem overlap.
Descriptive results concerning this index are reported in Table 20.
Table 20
Descriptive Results for Adjacent Stem Overlap
Group N x ̄ sd
Learner (L) 49 .438 .176
Native 1 (N1) 54 .560 .165
Native 2 (N2) 46 .530 .135
Descriptive results for the three groups concerning adjacent stem overlap are
displayed in Table 20. According to these results, the learner group scored particularly
lower than the native ones (x̄L=.438), whereas the native group scores appear to be quite
86
similar (x̄N1= .560, x ̄N2= .530). This exclusion could be confirmed with results presented
in the following table.
Table 21
Kruskal Wallis Test Results for Adjacent Stem Overlap
Group N Mean Rank
df x2 p Post-hoc
Learner 49 57.28 2 13.407 .001 L<N1&N2
Native 1 54 87.84
Native 2 46 78.80
The comparison of adjacent stem overlap scores for the three groups is presented
in Table 21. The results of the comparison clearly indicate that there is a statistically
significant difference among the groups [x2 (2) = .125, p< .01]. The results of Mann
Whitney U-test make it clear that this difference is between the learner and the native
groups (L<N1&N2). This difference might be an indication of learners’ lack of
proficiency in modifying lexical items according to their syntactic requirements. It
could also mean that learners’ knowledge concerning L2 vocabulary lacks depth
disregarding the parts of speech of the lexical items at their disposal. The following
three different excerpts taken from three different native essays will illustrate that native
speakers of English who participated in this study compose cohesion by making use of
stem overlaps.
Native excerpt 3:
The teaching of New Age ideas raises an ethical issue due to its different values, and because it is not yet established. When attempting to establish a certain case with a group of people, there must be unquestionable authority so the people will buy into the argument.
Native excerpt 4:
School integration has been a hot topic across the United States since 1954 when the Supreme Court decided that, <*> and racial segregation in schools was declared unconstitutional. After this decision was made, the number of black students attending schools with white students increased slowly but surely.
87
Native excerpt 5:
The discovery of penicillin as an antibiotic was one of the greatest advances in medicinal chemistry. As a natural antibiotic, penicillin was discovered to have many uses in stopping and preventing the spread of infections.
The underlined words in the above excerpts are called stem overlaps; they share
common lexical roots. At this point, very few examples of stem overlap from the learner
essays could be determined. The following excerpts are from essays written by two of
the learners who participated in the descriptive part of the study. Following is an excerpt
from the essay written by E. A. from the learner group:
And it will not be by itself, it will be done by imagination and
dreaming about the future. Of course we had more space to imagine about something, because there was so much more need than now.
Below is another excerpt from the essay written by E. S. from the learner group:
Science is developing all the time. Scientists find out and discover new things.
These two rare excerpts taken from the essays written by two of the learners also
demonstrate examples of stem overlaps for adjacent sentences. However, as the
statistically analyses that were carried out previously suggest, stem overlaps are very
rare in learner texts compared to native ones.
All-distance stem overlap is another index used in Coh-metrix. In this index,
stem overlaps are calculated by taking into consideration the whole text, not just
adjacent sentences. Descriptive results about this index are given in Table 22.
Table 22
Descriptive Results for All-distance Stem Overlap
Group N x̄ sd
Learner (L) 49 .341 .155
Native 1 (N1) 54 .472 .164
Native 2 (N2) 46 .431 .110
88
Table 22 exhibits descriptive results for the all-distance stem overlap index.
Although the native groups appear to have similar mean scores (x̄N1= .472, x ̄N2= .431),
the learner group stands out with a relatively low mean score (x̄L=.341). The following
table verifies that this difference between the learner and the native groups is
statistically significant.
Table 23
Kruskal Wallis Test Results for All-Distances Stem Overlap
Group N Mean Rank df x2 p Post-hoc
Learner 49 54.89 2 16.931 .000 L<N1&N2
Native 1 54 88.98
Native 2 46 80.01
The Kruskal Wallis and post-hoc test results for all-distance stem overlap for the
three groups can be checked in Table 23. The results exhibit a definite and statistically
significant difference among the three groups [x2 (2) = 16.931, p< .01]. When this
difference is checked through a Mann Whitney U-test to determine which group was
statistically excluded from the others, the scores of the learner group appear to be
significantly lower than the scores of the native groups (L<N1&N2). As was mentioned
before, this index calculates the stem overlaps across a given text. Since the native
groups scored significantly higher than the learner group in both adjacent and all-
distance stem overlap indices, it would not be an assumption to say that the learner
group lacks the ability and flexibility to make use of the different types of speech of
lexical items. This index alone could be regarded as an indication of a lack of
connection among sentences written by the learner group.
Another index through which Coh-metrix measures lexical cohesion is LSA,
which is a statistical technique akin to factor analysis. This index has three parameters,
adjacent sentences, all-distance and paragraph LSA measures, LSA scores concerning
paragraphs were not taken into consideration as certain problems were observed in
learners’ writing about paragraph forming. In order for this parameter to yield accurate
calculations at this point, the paragraphs have to be well-constructed. In our case, there
were essays with only one paragraph consisting of 300 words, or paragraphs with too
89
many and unnecessary paragraphs. For this reason, only adjacent sentences and all-
distance LSA scores were taken into consideration. Both of these indices yielded
statistically significant differences among the three groups that participated in the study.
Descriptive results for adjacent LSA scores are detailed in Table 24.
Table 24
Descriptive Results for LSA Scores for Adjacent Sentences
Group N x̄ sd
Learner (L) 49 .185 .056
Native 1 (N1) 54 .230 .053
Native 2 (N2) 46 .233 .049
Descriptive data concerning LSA scores for adjacent sentences of the three
groups are displayed in Table 24. The data revealed in the table clearly indicates that the
learner group (x ̄L=.185) has a lower mean than both of the native groups (x ̄N1= .230,
x ̄N2= .233).
As was mentioned before, since the normality of distribution and the
homogeneity of variance among LSA group scores were at acceptable levels (see Table
10) a parametric test, one-way ANOVA, was employed to see if the observed mean
difference among the three groups was statistically significant. One-way ANOVA and
post-hoc (Scheffe) test results for adjacent LSA scores are presented in Table 25.
Table 25
One-way ANOVA Results for Adjacent LSA Scores
Sum of Squares df
Mean Square F p Scheffe
Between Groups .072 2 .036 13.009 .000 L<N1&N2
Within Groups .406 146 .003
Total .479 148
As can be observed from Table 25, adjacent LSA scores for the groups differ
significantly (F(2-146)=13.009, p< .05). In order to determine the nature of this significant
difference a post-hoc test (Scheffe) was performed. The results of this post-hoc test
90
clearly indicate that the learner group scored significantly lower than the native groups
(L<N1&N2). The next parameter in LSA index is all-distance scores, which are
calculated by taking into account LSA outcomes throughout an entire text. Descriptive
results concerning all-distance LSA scores are presented in Table 26.
Table 26
Descriptive Results for All-distance LSA Scores
Group N x ̄ sd
Learner (L) 49 .166 .053
Native 1 (N1) 54 .214 .062
Native 2 (N2) 46 .209 .052
A quick glance at Table 26 makes it clear that the learner group scored lower
than the native groups (x ̄L=.166, x ̄N1= .214, x ̄N2= .209). The significance of this
difference is calculated by means of one-way ANOVA and Scheffe test, and the results
are presented in Table 27. Table 27
One-way ANOVA and Scheffe Test Results for All-distance LSA Scores
Sum of Squares df
Mean Square F p Scheffe
Between Groups .07 2 .03 11.067 .00 L<N1&N2
Within Groups .46 146 .03
Total .53 148
Analysis result of all-distance LSA scores presented in Table 27 indicates that
the difference among groups concerning all-distance LSA index scores are statistically
significant [F(2-146)=11.067, p<.05]. Furthermore, the Scheffe test result reveals that this
difference statistically excludes the learner group from the native ones (L<N1&N2).
Parametric analyses of adjacent and all-distance LSA scores revealed that the
learner group scored significantly lower than both of the native groups. The noteworthy
aspect of these outcomes is that LSA scores, both adjacent and all-distance, were not
91
influenced by the average number of words in the texts produced by the three groups. In
other words, no matter what the text lengths of the native writers were, the learner group
scored significantly lower LSA scores. When the relation of LSA scores with lexical
cohesion in texts is taken into consideration, written productions of the learner group
can be claimed to be less cohesive compared to both of the native groups’ productions.
The significant differences among groups concerning referential and semantic
indices emerged in the anaphor reference index, the stem overlap index and the LSA
index. All of these indices, in fact, yield conclusions as to the unity and cohesion in a
text. In these indices, learners scored significantly lower than native groups. The learner
group scored higher in the anaphor reference index, and this is likely to be a result of
pronoun overuse. Furthermore, the learner group scored lower in the stem overlap index
and the LSA index. This is a strong indication that learners’ sentences are disconnected.
The following excerpts from four different essays written by learners who participated
in the first phase of the study reveal the possible causes of the significant differences
between the learner and the native text sets in terms of lexical cohesion.
An excerpt from an essay about the media written by F. K. from the learner group:
There are a lot of brochures, newspapers at the newspaper markets. Every brochure cannot be good, because sometimes there are magazines brochures between them. The brochures usually are read by young people.
Another excerpt from an essay about college education written by S. D. from the
learner group:
A real or theoretical world in university! How do we grow up? Actually, a number of students always think that they are different from other people. Most of universities just give information as theoretic. Thus, I believe that real world is more important than theoretical world for realizing prospective of things that they will use.
An excerpt from an essay about wars written by A. K. from the learner group:
Gun power is getting so important all over the world where the world smells like a battle area. Almost every single technology is involved in the war industry. New technology war toys, they are not called as guns anymore, can cause catastrophic results. However, they are nothing without a thinking brain. Hence, the question is that, should the kids who are going to play with these toys be professionals or anyone?
92
Below is another excerpt from an essay about the media written by S. S. from the
learner group:
If you wanted to harm a nation, it would be enough that you should take their media. TV, radio, newspapers are the most important things, and the easiest way for reaching all the people. If the media is healthy, the nation is healthy too, and in Turkey, we are not healthy.
In the excerpts above, written by four different learners, cohesion is undermined
by a lack of ties between sentences, which makes the sentences seem to be floating
around in a disconnected manner. The following excerpts are presented to demonstrate
the lexical connections present in native texts.
Native Excerpt 6:
This topic came to me on Friday night when "The State" called to solicit a subscription to their newspaper. I refused to take a subscription to their paper. Several years ago I considered myself to be an avid reader. However lately, I feel the quality of their news has begun to go down. I remember last year when I was- interviewed by a representative of "The State", on the patio of the Russell House. I told them at that time that they had lost the quality of their news, and the price was just too high for what they had in their paper.
Native Excerpt 7:
In all of the colleges and universities across the United States, administrators are trying to increase enrollment. This is entirely ethical and in most cases necessary to continue to have many of the programs and facilities that the universities offer. To accomplish this, many universities add courses and provide new things such as research labs and new cutting edge equipment. Many universities even try to accommodate the changing needs of the students by providing services such as free tutors and bigger staffs.
Native Excerpt 8:
The twentieth century has seen more wars than any other previous bloc of time. Though advances in communication, transportation, and information sharing, the world as a body of people living in close proximity, has rapidly shrunk. In the second of the world wars, the race was on to create the most devastating, most powerful, and most frightening weapon our people had ever known. In my opinion, the discovery and harness of atom and its energy and the corresponding invention of nuclear weapons have been the most significant factors of change in our lifetime, if not, perhaps, in several lifetimes.
93
Native Excerpt 9:
An invention of the 20th century that I think has significantly changed people's lives is television. TV allows the instantaneous communication of ideas, images, news, etc. Now, if something is happening in any part of the world- from across town all the way to Beijing -- every person in the world can watch this event as it is happening. While previously people had to wait days, or even weeks for news (which consisted of second-or third-hand accounts, which always involved much interpretation & distortion) now they can see it happening, for themselves, as it is happening, this allows them to make their own, informed judgment about the event; and if a reaction is necessary- such as the gathering of opposition to a certain law -- it can be made immediately, while the emotions are still strong, and before the action is finalized.
Apparently, the four native excerpts presented above are much more coherent than
the learner essays presented previously. As was mentioned in Chapter 3, cohesion in a
text is realized through the use of certain structures such as reiterations and collocations.
In these native excerpts, reiterations are rare, if any. However, cohesion is present due
to collocational ties among sentences. That is, in the sixth native excerpt about
newspapers the words like subscription, interviewed, paper, reader, news and
newspaper; in the seventh native excerpt about college education the words like
colleges, labs, enrollment, programs, facilities, courses, research and universities; in
the eighth native excerpt about wars the words like wars, weapon, powerful,
frightening, devastating, nuclear, atom and weapon; and in the ninth native excerpt
concerning the media the words communication, images, news and watch collocate
strongly with each other. However, when the learner excerpts are analyzed in the same
way, the words brochure and newspapers in the first excerpt; theoretical university,
information and students in the second excerpt; gun, power, battle, war and
catastrophic in the third excerpt; and newspapers, TV, radio and media in the last
excerpt can be regarded as collocations for each other, but these collocations do not
seem to be enough to compose a native-like textual cohesion, and this observable
difference in cohesion in learner essays can be confirmed with the LSA values
presented previously.
4.1.3. Readability Indices
There are two indices in the readability index: Flesch reading ease score index
and Flesch-Kincaid reading ease score index. The scores for the former range from 0
to100, and the scores for the latter are from 0 to 12. The higher the Flesch reading ease
94
score is, the easier it is to read the text. For the Flesch-Kincaid reading ease score, lower
scores mean the text at hand is easier to read. Table 28 gives an overall description of
the number of subjects in the groups, their mean scores and the standard deviations.
Table 28
Descriptive Results for Flesch Reading Ease Scores
Group N x̄ sd
Learner (L) 49 62.612 8.451
Native 1 (N1) 54 54.673 10.974
Native 2 (N2) 46 53.171 6.593
It is clear from Table 28 that the learner group (L) scored higher than both of the
native groups (x ̄L=62.612, x ̄N1= 54.673, x ̄N2= 53.171). The mean scores of the native
groups are quite close. To see if the difference is statistically significant, Kruskal Wallis
and Mann Whitney U-tests were conducted, and the results of these tests are detailed in
Table 29.
Table 29
Kruskal Wallis Test Results for the Flesch Reading Ease Scores
Group n Mean Rank
df x2 p Post-hoc
Learner 49 100.78 2 28.528 .000 L>N1&N2
Native 1 54 68.65
Native 2 46 55
Kruskal Wallis test results for the Flesch reading ease scores and post-hoc
results are presented in Table 29. It is clear from the table that there is a statistically
significant difference among the groups [x2 (2) = 28.528, p< .01]. The binary
comparisons of the groups through Mann Whitney U-tests reveal that the difference is
between the learner and the native groups (L>N1&N2).
95
The next reading ease score that Coh-metrix makes use of is Flesch-Kincaid
reading ease score. This reading score is quite similar to Flesch reading ease score, but
yields different numerical values. Descriptive results for related scores are introduced in
Table 30.
Table 30
Descriptive Results for the Flesch-Kincaid Reading Ease Scores
Group N x̄ sd
Learner (L) 49 8.011 2.042
Native 1 (N1) 54 10.500 1.343
Native 2 (N2) 46 10.591 1.293
Mean scores and the standard deviations of the scores are shown in Table 30.
Again, the learner group (L) stands out from the native groups in terms of the second
reading ease score. The mean score for the learner group (x̄L=8.011) is lower than the
mean scores of the native groups (x ̄N1= 10.500, x ̄N2= 10.591). The statistical
significance of this difference is presented in Table 31.
Table 31
Kruskal Wallis Test Results for Flesch Kincaid Reading Ease Scores
Group N Mean Rank
df x2 p Post-hoc
Learner 49 39.44 2 49.738 .000 L<N1&N2
Native 1 54 92.12
Native 2 46 92.78
Kruskal Wallis test results and post-hoc test results are presented in Table 31. It
is clear from the table that the mean differences among the three groups are statistically
significant [x2 (2) = 49.738, p< .01]. The post-hoc test results i.e. binary comparison of
the three groups through the Mann Whitney U-test, indicate that the learner group
scores are significantly lower than those of the native groups (L<N1&N2).
96
As was mentioned previously, reading ease scores are indications of the
comprehensibility of any given text in English. For the Flesch reading ease score, as the
score goes down the texts becomes harder to understand; on the other hand, in the
Flesch-Kincaid reading ease score, as the score decreases the document becomes easier
to read. Now that it is clear that the learner group’s reading ease scores are significantly
different from native group scores, it might be claimed that the learner group produces
simpler and less sophisticated sentences compared to their native counterparts. The
following excerpts taken from the learner and native essays will verify the statistical
differences determined among the groups in terms of reading ease scores.
Below is an excerpt about technology written by E. A. from the learner group:
Technology is vital for us, isn’t it? Yes, most people think, in the same way. We depend on not only technology, but also science and industrialization. In the imperial age people used to have a bit technology. But that was not enough. They had to grown up with new inventions. Now, however somebody can think that when we invent new things, our imagination value decreases. Let’s have a look at this situation.
Native Excerpt 10:
Science and technology have allowed me to travel to this part of the world, they have provided a living for my family, they have cured my grandmother of cancer, and they provide horizons of hope and knowledge in the fields of medicine, science, engineering, and even the less overt corners of our lives. The use of the means humans have developed is wherein the greatest problems lie. Alone, they do not threaten us, but when they become welded with certain aspects of and certain people in our societies, they become potentially the most dangerous things that we hold in our hands.
The first essays written by one of the learners obtained 64.793 from Flesch
Reading ease calculation and 6.52 from Flesch-Kincaid ease score. On the other hand,
the tenth excerpt presented above which was written by a native speaker obtained
44.494 and 12 respectively. When observed, the paragraph written by the learner is
composed of less sophisticated words with simpler grammar structures, and this
noticeable simplicity could be counted as the confirmation of the statistical analysis
mentioned previously.
4.1.4. Syntax Indices
The next index set in Coh-metrix to yield results differentiating L1 and L2 texts
is the syntax index set. Because of their inherent dependence on text size, the average
97
number of the words used in the text, the two text sets with similar word count averages
(x̄L=333 words, x̄N1=400 words; see introduction for details) were compared. Four of
the syntax indices (scores concerning personal pronoun incidence, incidence of all
connectives, type-token ratio and the number of words before the main word)
demonstrated statistically significant differences between the learner group and the
native group with similar averages for the number of words per text.
The first index to demonstrate statistically significant difference is the personal
pronoun incidence score index. Descriptive results concerning the personal pronoun
incidence scores are detailed in Table 32.
Table 32
Descriptive Results for the Personal Pronoun Incidence Scores
Group N x̄ sd
Learner (L) 49 91.362 30.833
Native 1 (N1) 54 42.291 20.284
It is clear from Table 32 that the learner group (L) scores noticeably higher in
this index (x̄L=91.362). In order to verify that the difference is statistically significant, a
Mann Whitney U-test was conducted; relevant data is provided in Table 33.
Table 33
Mann Whitney U-test Results for the Personal Pronoun Scores
Group N Mean Rank Sum of Ranks U p
Learner 49 74.14 3633 238 .000
Native 1 54 31.91 1723
The results of the Mann Whitney U-test revealing the differences between the
groups in terms of personal pronoun usage are presented in Table 33. It is quite clear
that there is a statistically significant difference between the groups (U= 238, p< .05).
This result clearly indicates that, in their written productions, the learner group makes
use of significantly more pronouns than their native counterparts. The situation becomes
98
more remarkable when the average words per text for each group is taken into
consideration (x ̄L=333 words, x ̄N1=400 words). The learner group, with relatively lower
average words per text, scores higher than the native group in terms of pronoun counts.
One of the indices in the syntax index set that needs to be highlighted is the
incidence of all connectives index. In this index positive connectives (and, after,
because), negative connectives (but, until, although), additive connectives (also,
moreover, however, but), causal connectives (because, so, consequently, although,
nevertheless), logical connectives (or, actually, if), and temporal connectives (after,
before, when, until) are taken into account. Generally, these connectives are analyzed
or compared separately as each of these groups of connectives are related to different
contextual aspects. However, since the focal point in the current study is lexical
cohesion, with a holistic approach to the issue, all the connectives were analyzed by
taking their total density scores in the texts and their mean scores were calculated over
1000 as presented in Table 34.
Table 34
Descriptive Results for Incidence of All Connectives Scores
Group N x̄ sd
Learner (L) 49 93.253 21.444
Native 1 (N1) 54 81.812 14.873
One can see from Table 34 that learners’ mean scores concerning all connectives
are higher than the native text set (x ̄L=93.253, x ̄n1=81.812). Statistical significance of
this difference is displayed in Table 35.
Table 35
Mann Whitney U-test for Incidence of All Connectives Scores7
Group N Mean Rank Sum of Ranks U p
Learner 49 59.781 2929 942 .008
Native 1 54 44.943 2427
99
Table 35 reveals the Mann Whitney U-test for incidence of all connectives for
the learner and native groups. The results indicate a significant difference between the
two groups (U= 238, p< .05). As in personal pronoun count, it is noteworthy that with a
lower average number of words per text (x̄L=333 words), the learner group again
significantly outnumbered the native group (x̄N1=400 words) in terms of connectives
count.
Connective and pronoun counts yielded significant difference between the
groups even with an unequal average number of words per text. It is obvious that, in
their written productions, the learner group makes use of pronouns and connectives
more than the native group does. This plethora of connectives and pronouns could be
regarded as one of the obvious characteristics of the EFL group that participated in the
present study.
The next index in the syntax index set to yield a statistically significant
difference among the groups was the type-token ratio index. Descriptive results
concerning this index are displayed in Table 36.
Table 36
Descriptive Results for Type-token Ratio Scores
Group N x̄ sd
Learner (L) 49 .633 .082
Native 1 (N1) 54 .710 .060
Descriptive results concerning type-token ratio suggests that the learner group
has a lower mean score (x ̄L=.633) than the native group (x ̄N1=.710). To check if this
observed difference is significant in any way, a Mann Whitney U-test was conducted,
and the results are presented in Table 37.
100
Table 37
Mann Whitney U-test for Type-token Ratio Scores
Group N Mean Rank Sum of Ranks U p
Learner 49 36.08 1768 543 .000
Native 1 54 66.44 3588
The Mann Whitney U-test results for the type-token ratio scores of the two
groups are displayed in Table 37. The results clearly show that there is a significant
difference between the learner and native group (U= 543, p< .05). As was explained in
Chapter 3, type-token ratio is an indication of the variation in a written text. This ratio
could be used to make deductions regarding the lexical variety of any given text. In our
case, it is obvious that the learner group lacks the lexical variety that the native group
has in their written productions.
The next index in the syntax index set to yield a significant difference between
the groups is the number of words before the main verb index. This index is a simple
count of the number of words that appear before the main verb of the main clause in the
sentences of a text. Descriptive results concerning learner and native data are presented
in Table 38.
Table 38
Descriptive Results for the Number of Words before the Main Verb
Group N x̄ sd
Learner (L) 49 3.723 1.202
Native 1 (N1) 54 4.660 1.490
Table 38 demonstrates the mean results and related standard deviations
concerning the number of the words before the main verb for both the learner and native
groups. The results obviously show that the native group (x̄N1=4.660) scored higher than
the learner group (x̄L= 3.723). To check the significance of this difference, a Mann
Whitney U-test was conducted and the results are displayed in Table 39.
101
Table 39
Mann Whitney U-Test for the Number of Words before the Main Word
Group N Mean Rank Sum of Ranks U p
Learner 49 41.39 2028 803 .000
Native 1 54 61.63 3328
Table 39 reveals the results of the Mann Whitney U-test, and the difference
between the learner and native groups appears to be statistically significant (U= 803, p<
.05). As was mentioned before, this index is related to the number of words before the
main verbs of the sentences in texts. The number of words before the main verb is an
indication of working memory load. Therefore, the related significant difference
between the native and learner groups could be interpreted as the naturally weaker
mental activity of the learner group in the target language (English in our case)
compared to that of the native group.
All of these differences between the learner and native texts in terms of syntax
can be analyzed in the following two excerpts. The first excerpt is an essay written by a
learner who participated in the descriptive phase of the study, and the next excerpt is
about the same topic and it was written by a native speaker of English. The first essay
was written by E. A. from the learner group:
I do not agree with that some people say about that in our modern world, dominated by science, technology and industrialization, there is no longer a place for dreaming and imagination. Science, technology and industrialization are things which the human being created with dreaming and imagination itself. We have come to this day with another expression to our modern world with keeping dreaming about the future. There is no end for imagination and dreaming.
We created so many things, cars, buildings, roads to each city of the world, airplanes, ships, houses etc. First thing caused that inventions was the feeling for the need all of those things, cars, houses, roads especially major things that we need to have to have a good life. We could not come this far without imagination and dreaming. And we still don’t know where the science is standing in the history told, there may be more invention and I believe there will be. And it will not be by itself, it will be done by imagination and dreaming about the future. Of course we had more space to imagine about something, because there was so much more need than now. And I believe that human being will always want more, will always need some things more special, and it can’t be done without imagination. For
102
example, let we say, we created a car first, but it was moving so slowly be we could go somewhere we wanted using that car. And then we wanted more, we improved the car to go faster. And there are still cars coming out to market which are going faster and faster. There is no end for it. We did not finish it with inventing a car to transport ourselves we invented planes which much faster than cars to minimize to distance and time. People will always want better.
There is no end for dreaming or imagination when the people keep living. There will always need for something.
The above text is composed of 328 words in total. The density score for personal
pronouns is 84.848/1000 and 106.061/1000 for connectives. Type-token ratio was
calculated as 0.597 and the mean number of words before the main verb appeared to be
2.842. Next is a sample essay about the same topic and it was written by a native
speaker of English.
Native Excerpt 11
Much has occured in the 20th century that has changed the way people live. The invention that comes to mind as being the most influential is the computer. This invention of the late 20th century has forever changed the way people live and work.
The first computers were expensive monsters that filled an entire room and could perform only a few calculations a second. With the advancement in technology and the move toward miniaturization, the desktop PC has evolved. These small, relatively inexpensive machines can do everything their predecessors did and much more. Every day new and exciting improvements are being made to enhance the performance of these tools.
The impact of computers on the world has been great. They have changed the way people do business and have radically altered the way data and information is dealt with. In short, the productivity of people has increased ten-fold. As we move into the 2st century this fact will become more important. The amount of information that is available in the world will require the use of computers to organize and extract that which is of interest. The personal computer has, and will continue to play a major role in our lives. The day is just around the corner when every human being in the developed world will have their own PC linked by cellular modem to the databases of the world, other PCs, and small digital assistants that will replace the printed page. We have only just begun to realize the impact of this 20th century miracle. We are limited only by our imaginations and the future of this invention looks "bright" indeed.
The text above is composed of 273 words in total. In this text, personal pronoun
density score was calculated as 29.304/1000 and the outcome for connectives density
103
score was 54.945/1000. Type-token ratio appeared to be 0.787, and average number of
words before the main verb was 5.412.
When the analysis results of these texts are compared, the statistically significant
differences concerning syntactic differences between the learner and native texts are
confirmed. The learner text presented above contains nearly three times as many
personal pronouns and twice as many connectives. The differences between these two
texts in terms of type-token ratio and the average number of words before the main verb
are also noteworthy. Altogether these aspects form the prominent syntactic differences
between the learner and native texts that were analyzed throughout this study.
4.1.5. Results of the Focus Group Interviews
As was mentioned in Chapter 3, the descriptive results of the first phase of the
study were discussed with five groups in five focus group interview sessions. The
learners who participated in these interviews were asked clear and simple questions
regarding the differences between their essays and the essays written by native speakers
of English. During these five sessions, no technicalities from the results were
mentioned; only the following points were discussed with the learners:
The essays that you wrote during this semester were compared to the essays
written by native speakers of English, and it was observed that compared to
these native speakers,
a) you use more pronouns and conjunctions than these native speakers
b) you write much simpler sentences
c) your sentences are less connected
What do you think the reason for these flaws could be?
The learners were only asked to share their opinions concerning the clear
deviations of their essays from native ones. Their responses were transcribed and
evaluated through framework analysis. In the last stage of the evaluation, three themes
emerged. A thematic chart was created and it is presented in Table 40.
104
Table 40
Thematic Chart for the Focus Group Interviews and Theme Frequencies
Theme f %
Inadequacy in Vocabulary 5 100
Confusion about what vocabulary to learn 3 60
Compensation for inadequacy in vocabulary 4 80
From Table 40, one can see that the interviews yielded three major themes,
namely inadequacy in vocabulary, confusion about what to learn, and compensation.
The first theme, inadequacy in vocabulary, is the most pervasive one as it emerged in all
of the group sessions (f: 5; 100%). The learners attribute their significant deviation from
the native speakers of English to the flaws in their vocabulary. The following sample
excerpts make this point clearer.
Theme 1: Inadequacy in Vocabulary
“I always feel that I am inadequate about learning new vocabulary. I can’t
remember any of them. In addition, I cannot use the words I learn in the
lessons in speaking or writing classes.” A learner from Group 1
“My biggest problem is about remembering the words I learn in English.”
A Learner from Group 3
“I am really bad at vocabulary learning.” A learner from Group 4
“I can’t learn or remember new words no matter how hard I try. I write, I
read, nothing happens.” A learner from Group 5
The sample excerpts presented above verify that the learners in all of the groups
suffer from an inability to retain English words that they encounter. This situation is one
of the commonalities detected in all of the groups.
The next theme that emerged during the interviews is the confusion that the
learners experience when trying to learn new words. This confusion results from the
overwhelming number of new words to be internalized by the learners and the desire to
105
cope with each and every one of them. This theme appeared in three out of five group
sessions (f: 3; 60%). The following sample excerpts taken from the interviews suggest a
perplexity about vocabulary items to learn among the learners.
Theme 2: Confusion about what to learn
“I have big problems about learning vocabulary. I am confused about new
words. I can’t tell which word is important and which ones are not
important, and there are too many words to learn.” A learner from Group 3
“When I learn a new word in English, I write down the Turkish meaning of
it. Later when I see the same word with a different meaning in another
passage, I don’t know what to do.” A learner from Group 4
“I memorize the words that I learn. I can memorize well. I have a good
memory. Even so, I am not sure about what I am doing, because most of the
words I memorize don’t appear in the exams.” A learner from Group 5
The last theme, compensation, is related to a sort of strategy which the learners
employ in order to deal with certain problems in written exams and it appeared in four
out of five sessions (f: 4; 80%). That is, since the learners who participated in the
current study lack proficiency in vocabulary, they try to make up for this deficiency by
overusing certain words or phrases. The following excerpts are examples of this
strategy.
Theme 3: Compensation
“I don’t know enough vocabulary, so when I am writing an essay I use the
words I already know or repeat the same words, because in the exams you
want us to write minimum 400 or 500 words. I feel forced to write
something to fill my paper.” A learner from Group 1
“I don’t know many words in English, so I try to write simple sentences, I
use simple words. I repeat the same words because of word limitations in
the writing exams.” A learner from Group 2
106
“We try to fulfill the tasks in the exams and write as many as required
number of words in writing exams. We have to produce a certain number of
words in writing exams. Maybe that’s why we use so many conjunctions.”
A learner from Group 4
The above points made by the participants are clear indications that the learners
make use of an overproduction strategy in order to cope with the problems in their
vocabulary. When the context is an exam, the issue becomes more serious and they
overuse certain structures like pronouns and conjunctions.
Other interesting points worth mentioning emerged during focus group
interviews carried out in the first phase of the study. For instance, some of the
participants from different interview groups mentioned that most of the time they
avoided writing complex sentences in order not to make mistakes. Some others stated
that they were forced by their instructors to use more conjunctions in their essays. One
of the participants even stated that he received very low grades because his instructor
thought that he had not used enough conjunctions in one of the written exams.
4.2. Results of the Experimental Phase
4.2.1. Experimental Results for Recognition
As was detailed in Chapter 3, the learners who participated in the second phase
of the current study were given a 50-item pre-test composed of randomly selected
vocabulary items to be studied during the semester. Before the comparisons of the
group scores were carried out, the scores were tested for normality and homogeneity of
variances for the same reasons explained previously in this section. The results are
presented in Table 41.
Table 41
Normality and Homogeneity Test Results for AWL Pre-test and Post-test Scores
SD
Skewness
Standard Error
z
Levene test
Pre-test 6.465 .069 .388 1.778 .361
Post-test 6.874 -.038 ,388 .098 .187
bo
tha
of
sui
va
the
ExpCon
Figex
Fig
Th
of
AN
pre
1
1
2
2
3
3
The te
oth tests are
an 1.96. Wh
f a Levene t
itable for p
alue .05. Be
eir compari
perimental Groupntrol Group
gure 16. Pxperimental
The gr
gure 16 ind
here seems t
f the post-tes
NCOVA te
esented in T
0
5
10
15
20
25
30
35
P
est results pr
e actually su
hen it come
test of varia
arametric c
efore these
son can be c
p : pre-test= 24; :pre-test= 23; p
Pre-test & and control
roup mean
dicate that th
to be a diffe
st results (x̄E
est. Descrip
Table 42.
Pre-test
resented in
uitable for
s to the req
ance indica
comparisons
comparison
checked in
post-test= 31 post-test= 26
post-test l groups
scores for p
he pre-test
erence betw̄E=31; x̄C=2
ptive result
Pos
Table 41 m
parametric
quirement of
ate that both
s, as the ob
ns, to get a
Figure 16.
mean scor
pre-test and
results are
ween the exp
26). These o
ts concerni
st-test
make it clear
comparison
f the homog
h the pre-te
tained score
better view
res for voc
d post-test r
very close t
perimental a
observed dif
ing group
r that the sc
ns as the z
geneity of v
est and the
es are great
w of the gro
cabulary re
results whic
to each oth
and the con
fferences we
means for
Experimental
Control
cores obtain
scores are
variance, the
post-test ar
ter than the
oups’ mean
ecognition
ch are displ
her (x̄E=24;
ntrol group i
ere analyze
r the post-t
107
ed from
smaller
e results
re again
e critical
n scores,
for the
layed in
x̄C=23).
in terms
d via an
test are
108
Table 42
Mean and Corrected Mean Scores of the Experimental and the Control Group for the
Vocabulary Recognition Test
Group N x ̄ Corrected Means
Experimental 18 30.833 30.268
Control 19 26.105 26.641
A quick check with Table 42 will make it clear that the post-test mean scores of
the two groups are 30.833 for the experimental group and 26.105 for the control group.
When these scores were corrected by taking the pre-test score means as the covariant,
the means were calculated as 30.268 for the experimental group and 26.641 for the
control group. In order to check if this difference is statistically significant, Table 43 can
be analyzed.
Table 43
ANCOVA Results for Pre-test & Post-test
Source Sum of Squares
df Mean
Square F p
Pre-test 967.494 1 967.494 62.443 .000
Group 120.193 1 120.193 7.757 .009
Error 526.795 34 15.494
Total 31555 37
ANCOVA results for the post-test results are presented in Table 43. These
results were calculated taking the pre-test results as the covariant. It is obvious from the
table that there is a significant difference between the two groups in terms of corrected
post-test results [F(1-34) = 7.757, p< .05]. Referring back to Table 43, we can see that the
experimental group did significantly better than the control group (x ̄E=30.268;
x ̄C=26.641). As a result, the answer to the fifth research question, trying to determine
whether concordance activities induce vocabulary recognition, appears to be
affirmative. That is, concordance activities carried out with the experimental group
seem to have had positive effects in terms of vocabulary recognition. Since the point
about vocabulary recognition has been established at post-test level, delayed effects of
these activities were not analyzed.
109
4.2.2. Confirmation of the Descriptive Results for the Experimental Phase
Before the comparison of the two groups in terms of LSA results could be
completed, the outcomes of the descriptive phase of the current study had to be
confirmed. In the descriptive phase of the study, it was established that the learners who
participated in the study differed significantly from the native group in terms of LSA
scores (see Tables 25 and 27). In the experimental phase of the study, there were
different participants (NTotal=37; NExperimental=18, NControl=19); therefore, their LSA
scores needed to be checked to confirm that they also deviated from native LSA scores.
To confirm this, t-tests were performed for both adjacent and all-distance LSA scores.
The results of the t-test for adjacent sentences are presented in Table 44.
Table 44
T-test Results of the Learner and Native Groups for Adjacent Sentences LSA Scores
Group Levene’s Test
N x̄ sd df t p F Sig.
Learner .095 .759
37 .172 .055 89 -5.072 .000
Native 54 .230 .053
Table 44 exhibits t-test results for the learner and native groups in terms of
adjacent sentences LSA scores. The results of Levene’s test for homogeneity of
variance indicate that the variance in group scores is acceptable for a parametric
comparison (p=.759 >.05). Group means appear to be different (x̄L=.172; x ̄N= 230), and
this difference appears to be statistically significant [t(89) = -5.072, p< .05].
The second confirmation of the parallelism between the descriptive phase and
the experimental phase was the all-distance LSA scores of the learner and native groups.
T-test results for this comparison are exhibited in Table 45.
Table 45
T-test Results of the Learner and the Native Group for All-distance LSA Scores
Group Levene’s Test
N x ̄ sd df t p F Sig.
Learner .361 .549
37 .136 .055 89 -6.172 .000
Native 54 .215 .063
pre
sui
an
sco
do
co
ex
lea
W
an
4.2
vo
sco
Con
Exp
Fig
1
1
2
2
T-test
esented in T
itable for p
nd native gro
The re
ores confirm
ouble-check
hesion is on
xperimental
arners throu
With this no
nalyzed in th
2.3. Target
As w
ocabulary ite
ore means f
ntrol Group
perimental Group
gure 17. A
0,00
5,00
10,00
15,00
20,00
25,00
results of
Table 45. T
parametric c
oups appear
esults of th
m the desc
for LSA s
ne of the fla
phase of th
ugh the imp
tion in min
he next secti
Vocabular
as explaine
ems was me
for the expe
: pre-test= 6.00
p : pre-test= 5.96
AWL density
Pre
the learner
The results o
comparison
rs to be stati
e t-tests for
riptive resu
cores of the
aws of learn
his study. I
plementatio
nd, the exp
ion of this c
ry Density
ed in the
easured as t
erimental an
03; post-test= 12
61; post-test= 17
y scores me
Post
and native
of Levene’s
s (p=.549>
istically sig
r LSA for
ults from th
e learners n
ners’ writte
In other wo
on of any te
perimental
chapter.
Scores
methodolo
the number
nd control gr
2.823; delayed=13
7.068; delayed=22
asured as nu
Delaye
e groups for
s test of var
.05). The d
gnificant [t(8
adjacent se
he first pha
not only str
en productio
ords, any in
echnique co
outcomes o
gy section
r of types pe
roups are de
3.929
2.296
umber of ty
ed
r all-distanc
riance show
difference b
9) = -6.172,
ntences and
ase of the c
rengthens th
ons, but it a
ncrease in L
ould be reg
of concord
, participan
er 1,000 wo
emonstrated
ypes per 1,0
Experime
Control
ce LSA sco
w that the sc
between the
p< .05].
d all-distan
current stud
he idea that
also rational
LSA scores
garded as v
dance activi
nts’ use of
ords. These
d in Figure
000 words
ental
110
ores are
ores are
learner
nce LSA
dy. This
t lexical
lizes the
s of the
valuable.
ities are
f target
density
17.
111
Density score means for the experimental and control groups are shown in
Figure 17. In their first essays (pre-test), the experimental and the control group made
use of nearly the same number of target vocabulary items (x̄E= 5.961, x ̄C = 6.003). In
their second essays (post-test), this similarity faltered as the experimental group scored
a higher mean than the control group (x ̄E=17.068, x ̄C = 12.823). Furthermore, in their
third essays (delayed post-test), it became clear that the experimental group’s mean
score for the use of target vocabulary continued to increase, whereas there seemed to be
little change in the control group’s mean score (x̄E=22.296, x ̄C=13.929). In order to
better visualize this difference in target vocabulary use between the experimental and
control groups, Figure 18 should be checked.
Figure 18. The change in awl density scores of the experimental and control groups
from pre-test to delayed post-test
In Figure 18, the progress of the experimental and the control group in target
vocabulary use can be identified easily. Both groups start with similar mean scores, a
difference appears in the post-test stage, and a greater difference between the group
means can be observed in the delayed post-test stage. However, this observable
difference between the groups does not mean much without statistical comparisons. In
order to be able to perform such comparisons, the equality of variance in group scores
was checked through Levene’s test, and the results showed that the group variances
were equal (p=.552>.05). The pre-test density scores for target vocabulary use were
0,00
5,00
10,00
15,00
20,00
25,00
Pre Post Delayed
Experimental
Control
112
taken as the covariant; mean and corrected mean scores for the experimental and control
groups are exhibited in Table 46.
Table 46
Post-test Mean and Corrected Mean Scores of the Experimental and the Control
Groups for Target Vocabulary Density Scores
Group N x̄ Corrected Means
Experimental 18 17.069 17.081
Control 19 12.822 12.811
Corrected means for the experimental and control groups for vocabulary density
scores are exhibited in Table 46. It is clear from the data presented in the table that the
experimental group scored a higher mean score compared to the control group
(x̄E=17.081, x̄C=12.811). In order to determine the significance of this difference, an
ANCOVA test was conducted. The results of this comparison are presented in Table 47.
Table 47
ANCOVA Test Results Comparing the Experimental and Control Groups for Target
Vocabulary Density Scores (Post-test)
Source Sum of Squares
df Mean
Square F p
Pre-test 586.285 1 586.285 8.898 .005
Group 168.563 1 168.563 2.558 .119
Error 2240.365 34 65.893
Total 11194.522 37
Target vocabulary density scores for the experimental and control groups are
analyzed in Table 47. When the data presented is examined, one can conclude that the
difference between the two groups concerning target vocabulary density scores is
statistically insignificant [F(1-34) = 2.558, p> .05]. This result shows that although there is
a certain amount of difference between the two groups’ mean scores, this difference
does not suggest that concordancing activities induced target vocabulary production for
the experimental group. However, when the issue is vocabulary production, delayed
113
effects of the learning process have to be considered. With this concept in mind, the
learners were given another writing exam (delayed post-test), and target vocabulary
density scores were calculated as before. Delayed post-test mean and corrected mean
scores for target vocabulary density for the experimental and control groups are
presented in Table 48.
Table 48
Mean and Corrected Mean Scores for Target Vocabulary Density for the Experimental
and Control Groups (Delayed Post-test)
Group N x̄ Corrected Means
Experimental 18 22.297 22.302
Control 19 13.928 13.923
Again, Levene test for equality of variance was carried out and the result came
out negative (p=.643>.05), which showed that the scores were suitable for parametric
comparisons. After variance equality was determined, an ANCOVA test was carried out
with the corrected delayed test scores by taking the pre-test scores as the covariant. The
results are presented in Table 49.
Table 49
ANCOVA Test Results Comparing the Experimental and Control Groups for Target
Vocabulary Density Scores (Delayed Post-test)
Source Sum of Squares
df Mean
Square F p
Pre-test 105.665 1 105.665 .936 .340
Group 648.907 1 648.907 5.750 .022
Error 3837.058 34 112.855
Total 16577.421 37
ANCOVA test results comparing the experimental and control groups for
target vocabulary density scores gathered from the delayed post-test clearly indicate that
there is a statistically significant difference between the groups [F(1-34) = 5.750, p< .05].
Therefore, it could be claimed that concordancing activities had positive effects on
lea
cer
4.2
Co
sco
gro
ob
po
in
Be
ch
an
ConExpNat
Fig
sco
co
0
0
0
0
0
0
arners’ voc
rtain amoun
2.4. Experi
Writte
oh-metrix fo
ores obtain
oup only) w
btained from
ost-test score
an exam en
The fi
efore the sta
hecked for a
nd native gro
ntrol Group perimental Grouptive
gure 19. P
ores for the
In Fig
mpared. Gr
0,000
0,050
0,100
0,150
0,200
0,250
P
abulary pro
nt of time.
mental Res
en productio
or comparis
ed from the
were record
m their essa
es. After a p
nvironment
irst analysi
atistical ana
a better visu
oups.
: pre-test= .175p : pre-test= .170 : .231
Pre-test, po
experiment
gure 19, th
roup means
Pre P
oduction, bu
sults of Lex
ons of the e
son of LSA
e essays wr
ded as the
ays written
period of tim
as the delay
s carried o
alyses of th
ualization o
5; post-test= .2030; post-test= .209
ost-test and
tal and cont
he control,
both before
ost Del
but that this
xical Cohes
experiment
A scores for
ritten befor
pre-test sc
at the end
me, the lear
yed post-tes
out was of
he related re
of the differ
3; delayed= .1929; delayed= .189
d delayed p
trol groups
experimen
e and after t
layed Na
s effect see
sion
al and cont
adjacent an
re concorda
ores; in the
of these ac
rners were
st.
the LSA s
esults are di
rences betw
post-test LS
in comparis
ntal and na
the period o
ative
ems to have
trol groups
nd all-distan
ncing activ
e same way
ctivities we
required to
cores for a
iscussed, Fi
ween the exp
A (adjacen
son with the
ative group
of treatment
Control
Experime
Native
e appeared
were proce
nce sentenc
vities (exper
y, the LSA
ere recorded
write a thir
adjacent sen
igure 19 sh
perimental,
nt sentences
e native gro
p mean sco
t are exhibit
ental
114
after a
essed in
ces. The
rimental
A scores
d as the
rd essay
ntences.
hould be
control
s) mean
oup
ores are
ted. The
115
mean score of the native group, however, is given as a reference point for comparison.
As can be seen in Figure 19, the control group scored a mean of .170 in their first
essays, which were regarded as the pre-test. In their post-test, the control group scored a
mean of .203, and in the delayed post-test the control group scored .192. The
experimental group also scored similar means; .175 for the pre-test, .209 for the post-
test and .189 for the delayed post-test. Post-test and delayed post-test results show
similarity, and all learner scores are lower than the native group’s mean score (x ̄N=
.231).
The first comparison, to see if concordancing activities had any effects on LSA
scores, was carried out with the post-test results. When the mean scores of the
experimental and control groups are corrected by using the pre-test results as the
covariant, the mean scores change slightly. Group means and their corrections are
displayed in Table 50.
Table 50
LSA Mean and Corrected Mean Scores for Adjacent Sentences for the Experimental and
Control Groups (Post-test)
Group N x̄ Corrected Means
Experimental 18 .209 .210
Control 19 .203 .203
In Table 50, LSA mean scores for adjacent sentences and their corrections for
the experimental and control groups are displayed for comparison. The mean score for
the post-test for the experimental group is .209, and it is .210 when corrected. On the
other hand, the mean score for the post-test for the control group appears to be .203, and
its correction yields .203. This difference between the group mean scores does not seem
to be significant, so to analyze this difference ANCOVA test results are displayed in
Table 51.
116
Table 51
ANCOVA Test Results for LSA scores for Adjacent Sentences of the Experimental and
Control Groups (Post-test)
Source Sum of Squares
df Mean
Square F p
Pre-test .010 1 .010 3.395 .074
Group .001 1 .001 .192 .664
Error .097 34 .003
Total 1.681 37
The ANCOVA test results for LSA scores for adjacent sentences of the
experimental and control groups are displayed in Table 51. Before carrying out this
ANCOVA test, the variances in group scores were tested for equality by using the
Levene’s test; the significance level for variance differences was determined to be .888
which suggested an equal variance of the scores.
When Table 51 is analyzed, it is clear that the slight difference between the
experimental and the control group scores is, in fact, statistically insignificant [F(1-34) =
.192, p> .05]. This outcome indicates that concordance activities did not have any
significant effect on the experimental group in terms of lexical cohesion for adjacent
sentences.
In order to see if concordancing activities had any delayed effect on lexical
cohesion, a second post-test was given to the groups. LSA mean scores and the
corrected means of this delayed post-test for the experimental and control groups are
presented in Table 52.
Table 52
Adjacent Sentence LSA Mean scores and Corrected Means for the Experimental and
Control Groups (Delayed Post-test)
Group N x̄ Corrected Means
Experimental 18 .189 .191
Control 19 .191 .190
117
As is presented in Table 52, mean score for the experimental group is .189, and
the mean score for the control group is .191. When mean scores are corrected, the mean
score for the experimental group becomes .191, and the control group’s mean becomes
.190. The difference between the corrected means seems to be too small to be
statistically significant. To determine if this difference is statistically significant or not,
an ANCOVA test was conducted; mean scores were checked for equality of variance by
Levene’s test beforehand, and the result was negative (p=.167>.05), which means that
the group means are suitable for a parametric comparison. The results of this ANCOVA
test are exhibited in Table 53.
Table 53
ANCOVA Test Results for LSA scores for Adjacent Sentences of the Experimental and
Control Groups (Delayed Post-test)
Source Sum of Squares
df Mean
Square F p
Pre-test .054 1 .054 12.308 .001
Group 7.590 1 7.590 .002 .967
Error .150 34 .004
Total 1.546 37
A quick check with Table 53 will reveal that the difference between the
experimental and control groups in terms of LSA adjacent scores is statistically
insignificant [F(1-34) = .002, p> .05].
The next index calculates all-distance LSA scores in texts. The mean scores for
all-distance LSA of the experimental and control groups are demonstrated in Figure 20.
ConExp
Nativ
Fig
the
ex
Fig
an
me
de
fro
gro
gro
an
Ta
All
Gr
G
E
C
0
0
0
0
0
0
ntrol Group perimental Groupve
gure 20. Pr
e experimen
The p
xperimental
gure 20. Th
nd .155 in th
ean score o
layed post-t
om these sc
oup in both
oups are tak
nd corrected
able 54
l-distance L
roups (Post-
Group
Experiment
Control
0,000
0,050
0,100
0,150
0,200
0,250
P
: pre-test= .137p : pre-test= .135 : .215
re-test, post
ntal and con
pre-test an
and contro
he control g
he delayed
f .135 in th
test. The re
cores that t
h of the post
ken into acc
d mean score
LSA Mean a
-test)
N
tal 18
19
Pre P
7; post-test= .1585; post-test= .186
t-test and d
ntrol group i
nd post-tes
l groups in
group scored
post-test. O
he pre-test, a
ference nati
the experim
t-tests. Whe
count as the
es of the tw
and Correct
x ̄ .185
.158
ost Del
8; delayed=.1556; delayed=.166
delayed post
in comparis
st results
n compariso
d a mean of
On the othe
a mean scor
ive group’s
mental grou
en the pre-te
e covariant,
wo groups ar
ted Mean Sc
Correc
.
.
layed Na
t-test LSA
son with the
for all-dis
on with the
f .137 in the
er hand, the
re of .186 i
mean score
up scored h
est results o
these mean
re presented
cores of the
cted Means
185
157
ative
(all-distance
e native grou
tance LSA
native grou
e pre-test, .1
e experimen
n the post-t
e appears to
igher mean
f the experi
ns need to b
d in Table 54
Experimen
Control
Experime
Native
e) mean sc
up
A scores
up are prese
158 in the p
ntal group s
test and .16
o be .215. It
ns than the
imental and
be corrected
4.
ntal and Con
ntal
118
ores for
for the
ented in
post-test
scored a
66 in the
t is clear
control
d control
d. Mean
ntrol
119
From the data presented in Table 54, corrected means of the experimental and
control groups appear to be different (x̄E= .185, x ̄C = .157). Before carrying out an
ANCOVA test, the variances in group scores were tested for equality by using the
Levene’s test, and the significance level for variance differences turned out to be .532,
which suggested an equal variance of the scores. In order to check the significance level
of the difference between the two groups, an ANCOVA test was conducted and the
results are presented in Table 55.
Table 55
ANCOVA Test Results for All-distance LSA scores of the Experimental and Control
Groups (Post-test)
Source Sum of Squares
df Mean
Square F p
Pre-test .008 1 .008 3.386 .074
Group .007 1 .007 2.856 .100
Error .083 34 .002
Total 1.177 37
The ANCOVA test results for all-distance LSA scores of the experimental and
control groups are exhibited in Table 55. The observed difference between the control
and experimental group does not seem to be statistically significant [F(1-34) = 2.856, p>
.05]. However, when the data presented in Figure 20 is checked, it will be noticed that
there has been an increase in the mean scores of both groups, and the increase in the
mean score of the experimental group is greater than that of the control group.
Therefore, it could be stated that there has been a greater increase, albeit statistically
insignificant, in the experimental group’s all-distance LSA scores.
In order to check if there are any delayed effects of concordancing activities on
all-distance LSA scores, learners’ third written productions the delayed post-test were
checked for all-distance LSA scores. All-distance LSA mean and corrected mean scores
for the experimental and control groups are presented in Table 56.
120
Table 56
All-distance LSA Mean and Corrected Mean Scores for the Experimental and Control
Groups (Delayed-test)
Group N x̄ Corrected Means
Experimental 18 .166 .166
Control 19 .155 .154 From the data presented in Table 56, one can see that there is a difference
between the groups in corrected mean scores (x̄E= .166, x ̄C = .154). The correction was
made by taking the participants’ pre-test scores as the covariant. In order to see if the
observed difference is statistically significant, an ANCOVA test was again conducted
following the Levene’s test, whose results for mean score variance appeared to be
negative (p=.320>.05). The results of the ANCOVA test are presented in Table 57.
Table 57
ANCOVA Test Results for All-distance LSA Scores of the Experimental and Control
Groups (Delayed-test)
Source Sum of Squares
df Mean
Square F p
Pre-test .030 1 .030 6.092 .019
Group .001 1 .001 .258 .615
Error .167 34 .005
Total 1.148 37
Table 57 exhibits the groups’ ANCOVA test results for all-distance LSA
scores gathered from the delayed post-test. The results clearly show that the observed
difference between group mean scores is, in fact, statistically insignificant [F(1-34) = .258,
p> .05].
To sum up the results revealed thus far, LSA scores of the experimental group,
which are indications of lexical cohesion, did not seem to improve significantly
compared to the control group. Even though certain fluctuations were present in the
experimental group’s LSA scores between the pre-test and delayed post-test, statistical
121
analyses revealed no significant improvement in the experimental group in terms of
both adjacent and all-distance LSA scores.
4.2.5. Semi-structured Interview Results
As the last stage of the study, 10 of the participants in the second phase of the
study were interviewed. The aim of this interview was to determine participants’ ideas
about concordancing activities and their effects on vocabulary recognition and
production. The interview was composed of five clear questions (see Appendix 6)
seeking to determine how participants felt and what they thought about the activities.
The data analysis which was conducted was similar to the focus group interview
carried out in the first phase. Framework analysis was employed for the data gathered
from 10 participants, and three common themes emerged. These themes are presented in
Table 58.
Table 58
Semi-structured Interview Thematic Chart and Theme Frequencies
Theme f %
Vocabulary Recognition Speed 7 70
Lasting Vocabulary Retention 7 70
Relationships among Words 6 60
Table 58 makes it clear that nearly all of the participants dwelled on three major
themes: vocabulary recognition speed, lasting vocabulary retention and relationships
among words. The first theme, vocabulary recognition speed (f: 7; 70%), was the first
one to appear and it is related to the effects of concordancing activities on learners’
vocabulary retrieval time. The following excerpts from the interviews are examples of
the learners’ conception of these activities.
Theme 1: Recognition speed
“I remember those words faster.” Learner 1
122
“While my friends in other classes spend too much time to remember those
words in the exams, I can remember them much faster, and go on to deal
with other parts of the exam freely. This is what heard from them.”
Learner 2
“In the first midterm I was so fast in the vocabulary part.” Learner 3
“When I see the words I learned from these activities, I can remember their
meanings better.” Learner 4
“By looking around the words I can eliminate the irrelevant options
faster.” Learner 5 (Talking about the vocabulary parts of the exams where
synonyms are tested)
“In the exams when I see a word, I see its meaning just like that, very
fast.” Learner 10
The sample excerpts above clearly indicate that the participants share the
opinion that concordancing activities helped them remember the words that they had
learned faster and better.
The next theme to appear during the interviews was vocabulary retention (f: 7;
70%). This theme is related to the effects of concordancing activities on how long the
new words are kept in learners’ vocabulary. The following excerpts from the interviews
are related to this process.
Theme 2: Lasting Vocabulary Retention
“I can remember the words we learned during these activities better.”
Learner 1
“I think the main purpose of these activities was to maintain vocabulary
retention.” Learner 6
“Our aim was to internalize the words that we were trying to learn.”
Learner 7
“Our ultimate aim was to make the words we learned more lasting.”
Learner 8
123
“In the exams I felt that the words we were trying to learn were more
lasting than the others.” Learner 9
As can be deduced from the excerpts, the participants have a common
understanding as to the effectiveness of concordancing activities on vocabulary
retention. They obviously share the idea that, through concordancing activities the
words that they had learned became more lasting.
The next theme determined through framework analysis concerns the
relationships among words (f: 6; 60%). This common theme relates to natural ties
among words occurring in the same contexts. Participants’ comments on the issue are
presented below.
Theme 3: Relationships among Words
“I used to think words in English as a column and their Turkish
equivalents as another column. But now I can see clearly that certain
words are used with other certain words.” Learner 2
“There is a kind of fellowship among words.” Learner 5
“When I see a word in a reading passage, I look around it and try to make
connections and then try to understand the paragraph.” Learner 5
“Before these activities, I used to look up a word in a dictionary, write its
Turkish meaning or synonym across it and that was all. But now I try to
think about the words that can be used with the word I am trying to
learn.” Learner 6
“When I see a word in the exam I try to think about the words it is used
with.” Learner 7
“There are many words with similar meanings; through these activities I
can now understand the relationships among words.” Learner 10
The above excerpts highlight a common understanding among the
participants. Even though the interviews were carried out in different sessions,
they refer to similar concepts. The participants constantly refer to the relationships
124
among words; they regard lexical items not as isolated units but rather units acting
together to form integrated meanings. Learner 2 refers to this situation as “certain
words used with other certain words”, and Learner 5 tries to make connections
among words before trying to make judgments about them.
As the discussion point at this stage, descriptive results for the fourth and
fifth interview questions regarding the effects of concordancing activities on
vocabulary recognition and production are presented in Table 59.
Table 59
Descriptive Results for the Interview Questions about the Effects of Concordancing
Activities on Vocabulary Recognition and Production
Interview Questions YES NO NO IDEA
N % N % N %
Recognition Do you think that concordancing activities have had any positive effects on your recognition of the new words you have learned?
10 100 0 0 0 0
Production Do you think that concordancing activities have had any positive effects on your writings?
2 20 5 50 3 30
Descriptive results presented in Table 59 suggest that the participants share the
idea that concordancing activities had positive effects on recognition of the vocabulary
items they learned. When asked whether these activities had any positive effects on
vocabulary recognition (interview question 4), all of the participants (f: 10; 100 %)
responded positively. On the other hand, when the participants were asked if
concordancing activities had any positive effects on their writings (interview question
5), only two of them answered positively (f: 2; 20 %), half of them responded
negatively (f: 5; 50 %), and three of them (f: 3; 30 %) stated that they had no opinion
about the matter.
In general terms, the results of the semi-structured interviews indicate that there
are certain changes in the participants’ perceptions concerning English vocabulary. In
addition, nearly all of the participants think that concordancing activities affected their
125
vocabulary recognition in a positive way. However, these results should not induce high
hopes, as the participants appeared to be over-motivated both during the interviews and
the activities. Below are two examples from the interviews:
“Freshman English was the lesson I attended most, because I felt I was
learning something.” Learner 10
“It was like we were creating a new method for learning.” Learner 6
Nevertheless, when the pre-test/post-test results for vocabulary recognition are
taken into consideration (see Table 43), we can see that the interview results support the
test results. The participants did practically better after concordancing activities and
verified the improvements themselves during semi-structured interviews. However, the
situation is a bit different when the issue is production. Although the LSA scores
gathered from both the experimental and the control group yielded observable changes
in the participants’ essays, these differences were statistically insignificant. The results
of the interview confirmed this result as more than half of the participants stated either
that they did not feel or see any improvements in their writings or that they were not
aware of such improvements. These outcomes make it almost perfectly clear that
concordancing activities could be an asset in inducing vocabulary recognition, but they
are hardly useful in vocabulary production if they are implemented without integration
into writing classes. That is, a sort of transition from recognition to production must be
developed if we want our learners to utilize the words that they are learning.
4.3. Overview of the Descriptive Results
Regarding the first research question, the results concerning referential and
semantic aspects indicate that there is also a significant difference between the learner
and the native groups. This outcome is important in that no matter how many words are
used, or whatever the prompt is, the learners appear to have something in common in
their writings in terms of semantics and the use of referential tools. The results showed
that the learner group was making use of referential tools much more than the native
groups, regardless of the number of words used in the texts. Moreover, in the learner
texts there was a plethora of references even when compared to texts with a higher
average of words written by the native group.
126
Another result related to referential aspects showed that the native groups scored
significantly higher than the learner group in both adjacent and all-distance stem overlap
indices. It could be claimed that the learner group lacks the ability and flexibility to
make use of the different types of speech of lexical items, which creates a disconnection
among sentences written by the learner group. When the parametric comparison results
of LSA scores are taken into account, this disconnection is validated. The learner group
obtained significantly lower scores than the native group both in adjacent and all-
distance LSA scores. There was also a significant difference between the learner and the
native groups in terms of anaphor references, which is directly related to the pronoun
counts in the essays. The learner group used significantly more pronouns than both of
the native groups.
With regard to the second research question, regardless of the average number of
words in the texts and writing prompts, there appeared to be significant differences
between the native and learners’ text sets in terms of readability. As was mentioned
earlier, readability scores indicate how readable and comprehensible a text is. In the
case of Flesch reading ease score if the score is higher, it is easier to read the text at
hand; however, if the score of a text is higher in the Flesch-Kincaid reading index,
reading becomes more difficult. As was detailed in Chapter 3, readability scores
obtained from three groups, one learner and two native were analyzed. Compared to
both of the native groups, the learners scored significantly higher in Flesch reading ease
index and they scored significantly lower in Flesch-Kincaid reading index. This result
suggests that the learner group produces significantly simpler sentences compared to the
native groups, and the average number of words used in the texts does not seem to
affect the results. Both of the reading ease scores suggest that the learner group
produced simplistic sentences presumably because of their limited lexical and syntactic
repertoires (Silva, 1993; Hinkel, 2002).
With regard to the third research question, a significant difference between the
learner and native groups was present in terms of syntactic features. For the reasons
explained before, the learner group was compared with the native group which had a
similar number of words per text. Syntax indices including the personal pronoun
incidence score, the incidence of all connectives, the type-token ratio, and the number of
words before the main verb yielded significant differences between the two groups. The
overuse of personal pronouns and connectives could be interpreted as a compensation
127
for the learners’ inability to produce proper lexical units. The overuse of connectives is
in line with the related literature (Silva, 1993; Tankó, 2004). The outcome related to
pronoun overuse is particularly significant when we consider that the learners’ native
language is Turkish, which is a pro-drop language, meaning that pronouns are generally
omitted because of pragmatically ready inferences making the subject of an action clear.
A significant difference was also present between the learner and the native
group in the type-token ratio index. This difference is also an indication of a
significantly different lexical density between the groups. This could be viewed as less
lexical variety in learners’ text sets compared to native ones. This outcome is also in
accordance with the related literature (Hinkel, 2002).
As for the fourth research question, concerning the participants’ perceptions and
feelings about the differences between their written productions and the native ones,
focus group interview results and, the related literature overlap at this point. Focus
group interview results show that students try to remain simplistic or keep low profiles
because of the fear of making mistakes or lexical inadequacy; therefore they produce
basic lexical items and syntactic structures even if more complex structures might be
expected from their proficiency levels. As Hinkel (2005) would suggest “...other crucial
factors that confound L2 writing and text have to do with shortfalls of writers' language
proficiencies and restricted linguistic repertoire that significantly undermine L2 writers'
ability to produce high quality texts.” Moreover, while writing learners try to
compensate for the lack of proper vocabulary by repeatedly using the same items from
their lexicon; the case worsens when there is a minimum word count in their exams.
This limitation, as the interview results would suggest, might cause learners to repeat
similar lexical and syntactic structures. Another important outcome of the interview is
that the lexical insufficiency of the subjects appears to stem from an inability to
differentiate between important (high-frequency) and less important (low-frequency)
vocabulary items. Furthermore, subjects’ responses revealed that they are, in a way,
forced by some of the instructors to use as many conjunctions as possible to make their
writings better. Thus, the learners are convinced that “the more conjunctions, the
better”.
128
4.4. Overview of the Experimental Results
In the experimental phase of the study, the experimental and control groups were
given a 50-item vocabulary exam as a pre-test and post-test. The results of statistical
analysis revealed that there was a significant difference between the experimental and
control groups in terms of recognition of the target vocabulary.
When the two groups were compared in terms of target vocabulary density
scores, the experimental group did clearly better in the post-test comparison, but
statistical analysis revealed no significance at this stage. However, statistical analysis
results showed that there was, in fact, a statistically significant difference in the delayed
post-test comparison. The experimental group scored significantly better than the
control group in terms of target vocabulary density scores.
When it comes to the effects of concordancing activities on lexical cohesion in
learners’ texts, the results were not as expected. Although there was a slight difference
between the learner and native groups, statistical analysis of the post-test and the
delayed post-test LSA scores revealed no significant difference between the
experimental and the control groups.
Semi-structured interviews were carried out to determine the participants’
perception and feelings about the concordancing activities implemented during the
semester. The analysis of these interviews yielded some common themes. The
participants thought that these activities enabled them to remember target vocabulary
items better and faster. They also thought that these activities caused the vocabulary
items to last longer in their memory. Lastly, they shared the idea that the words they
were trying to learn had connections with other words. One important point revealed at
this point was that the participants shared a common understanding about the effects of
concordancing activities on their writing skills. Most of them either found no effects of
concordancing activities on their writing skills or they stated that they had no idea on
this point.
4.5. Summary
Results concerning the descriptive phase of the study yielded answers to the
related research questions. First of all, significant differences between EFL learners and
native essays were present in terms of referential and semantic aspects. The learners
obtained significantly different scores from native speakers in anaphor reference, stem
129
overlap and lexical cohesion scores (research question 1). These scores indicate an
overall lack of cohesion in Turkish EFL learners’ essays.
With regard to the readability scores, the learner group obtained significantly
higher scores in Flesch reading ease score, and significantly lower scores in Flesch-
Kincaid reading ease score. Both of these scores indicate the same conclusion that, in
general, Turkish EFL learners produce significantly simpler sentences even at
intermediate or upper-intermediate levels (research question 2).
As for the syntax indices (research question 3), the learner group appeared to be
using significantly more personal pronouns and connectives in their writing. There was
also a significant difference between the learner and native groups in type-token ratio
scores. Besides being an indication of certain syntactic features, this outcome is also a
strong indication of lexical density. Turkish EFL learners who participated in the
current study produced lexically less dense essays. The learner group also scored
significantly lower scores in terms of the in number of words before the main verb. This
outcome is thought to be related to low working memory load of foreign language
learners in the target language (McNamara et al. 2010).
These significant differences were discussed with the participants in 5 different
groups through focus group interview sessions. The results made it clear that the
participants had lexical problems in the target language (research question 4). The
outcomes of these interviews also suggested that the participants were applying a
strategy to compensate for their inadequate vocabulary by making use of an excessive
amount of conjunctions and pronominal structures. Furthermore, the participants
reported that past instructors over-encouraged the use of conjunctions whenever
possible.
Results concerning the quasi-experimental phase of the study also yielded
answers to the respective research questions. First of all, positive effects of
concordancing activities on vocabulary recognition were made clear (research question
5). The analysis of pre-test and post-test scores of the experimental and control groups
revealed that concordancing activities promoted vocabulary recognition. In terms of
vocabulary production (research question 6), pre-test and post-test comparison of the
experimental and control groups indicated minor differences; however, the analysis of
delayed post-test for production revealed significant differences between the groups.
130
The experimental group obtained significantly higher scores than the control group in
the delayed post-test for production. At this point, it could be discussed that the learners
who participated in the treatment could have forced themselves to use of the vocabulary
that they had learned during the concordancing activities, and the quantitative outcomes
could be captious. In order to parry such doubts, randomly chosen essays from both the
experimental and control groups were added to Appendix 7. In these sample essays, the
target vocabulary items were boldfaced for cohesion analysis. The analysis will make it
clear that the target lexical items are suitable for the contexts which they are used in.
The only difference is that, compared to the control group, the experimental group
appears to be making significantly more of these items.
With regard to the effects of concordancing activities on lexical cohesion on
learner texts (research question 7), LSA scores of the two groups were compared
through pre-test, post-test and delayed post-test. The comparisons were performed by
taking into account adjacent and all-distance sentences in learners’ texts. There was an
obvious difference between the experimental and the control group both in the post-test
and the delayed post-test scores. However, the analyses of these test results revealed no
statistically significant changes in LSA scores in favor of the experimental group.
As for the participants’ feelings and perceptions about the concordancing
activities that were carried out during a semester (research question 8), the participants
found these activities useful and effective in terms of vocabulary recognition; however,
they were either hesitant or negative about the effects of these activities on their written
production.
131
CHAPTER 5
DISCUSSION AND CONCLUSIONS
5.0. Introduction
In this chapter, a general and a brief summary of the study will be presented.
Then, the findings concerning each of the research questions will be summarized.
Following this summary, some recommendations for further studies will be made.
Personal reflections and criticism of the current study will be the last part of both this
chapter and the study.
5.1. General Summary of the Study
Taking the related literature into account, the current study dealt with the
following research questions:
1. Regardless of prompt or average number of words used in the texts, to what
extent do texts written by Turkish EFL learners deviate from texts written
by native speakers of English in terms of readability?
2. Regardless of prompt or average number of words used in the texts, to what
extent do texts written by Turkish EFL learners deviate from texts written
by native speakers of English in terms of referential and semantic aspects?
3. With average number of words per text being similar, to what extent do
texts written by Turkish EFL learners deviate from texts written by native
speakers of English in terms of syntactic features?
4. What are learners’ perceptions and feelings about the differences between
their written productions and that of native ones?
5. Can concordancing activities induce vocabulary recognition?
6. Can concordancing activities induce vocabulary production?
7. Can concordancing activities induce lexical cohesion in EFL learners’
written productions?
8. What are learners’ perceptions and feelings about the use of concordancing
activities to learn vocabulary?
132
As is clear from the research questions presented above, this study was an attempt
to determine lexical issues in Turkish EFL learners’ written productions. These issues
were identified through the analysis of semantic, referential, syntactic and cohesive
features of these texts. Argumentative essays written by these learners (NLearner=49)
were compared to two sets of argumentative essays written by two groups of native
speakers of English (NNative1=54, NNative2=46). The comparisons were performed through
Coh-metrix which is an online database for text analysis. The results of these
comparisons revealed that the learner essays were significantly different from native
essays in terms of semantic, referential, syntactic and cohesive aspects. The learner
group produced significantly simpler sentences compared to the native groups. They
also used too many pronominal structures and connectives in their writing. Cohesion,
which was measured by using LSA scores, was also weakened because of lexically
disjoint sentences in learners’ essays.
Focus-group interviews were carried out to find out the learners’ perceptions and
feelings about the flaws that surfaced in their essays. The results of these interviews
suggested that the main reason for such differences was mostly because of participants’
lexical inadequacy. Almost all of the groups participated in the interviews confirmed
this inadequacy.
Since the related literature suggests positive outcomes concerning the use of
concordancing activities on vocabulary learning (Cobb, 1997, 1999, and Thurstun &
Candlin, 1998), two groups of intermediate and upper-intermediate EFL learners -one
experimental and the other control group- were chosen for the concordancing treatment,
which would likely to solve the problems determined previously. As a confirmation,
essays written by these groups were also processed through Coh-metrix and the
previous findings concerning cohesion were verified.
The experimental (NExperimental=18) and the control group (NControl=19) were quite
similar in terms of the learning process except that the experimental group used
concordancing activities in the process. At the beginning, both of the groups were given
a pre-test for vocabulary recognition. This test included the vocabulary items to be
taught during that semester. In order to test production and cohesion, the groups were
asked to write argumentative essays. These essays were processed in Coh-metrix and
recorded as the pre-test for production. The concordancing treatment lasted for about 10
weeks. In the process, the experimental group was given pre-activities (see Appendix 4,
133
Activity 1 for a sample activity) about the vocabulary items to be taught before each
session; these activities were constructed by using online corpora (www.lextutor.ca)
composed of about 15 million words. In order to match the activities with the technical
context of the participants, concordancing lines from technical contexts were gleaned.
After each session, another concordancing activity (see Appendix 4, Activity 2 for a
sample activity) was given to the experimental group. The sessions were equally timed
and the control group received classical treatments such as cyclic repetitions for the
same vocabulary items.
At the end of the treatment, both of the groups were given the vocabulary
recognition test that was given before the treatment as the post-test. The groups were
also asked to write about the same topics used at the beginning of the study. These
essays were processed in Coh-metrix and the scores were recorded as the post-test for
production and cohesion. In order to check the delayed effects of the treatment, the
groups were given the same writing task after about two weeks, and the results gathered
from Coh-metrix were recorded as the delayed post-test scores for production and
cohesion. Statistical analysis of these results revealed that, in terms of vocabulary
recognition, the experimental group obtained statistically higher scores than the control
group. Vocabulary production and cohesion scores were analyzed by using the delayed
post-test scores; although the experimental group appeared to have obtained higher
scores than the control group, this difference was not statistically significant. When the
delayed post-test results for vocabulary production and cohesion were analyzed, the
difference between the experimental and the control group appeared to be statistically
different in favor of the experimental group in terms of vocabulary production.
Furthermore, LSA scores of the experimental group appeared to be higher than the
control group. However, statistical analysis of LSA scores once again revealed no
significant difference between the experimental and the control group.
As the last step in the study, semi-structured interviews were performed with 10
of the participants chosen randomly from the experimental group. The results of this
interview revealed positive attitudes towards the use of concordancing activities in
vocabulary instruction. All of the participants thought that these activities affected
vocabulary recognition positively. However, the students also agreed that these
activities had no obvious effects on their writings.
134
5.2. Review of the Findings of the Research Questions
Research Question 1
Regardless of prompt or average number of words used in the texts, to what
extent do texts written by Turkish EFL learners deviate from texts written by native
speakers of English in terms of referential and semantic aspects?
The results concerning the first research question revealed significant differences
between the learner and native groups in terms of referential and semantic aspects of
their written productions. The learner group significantly deviated from the native group
with regard to stem overlap and anaphor references. The learner group obtained
significantly lower scores in stem overlap analysis, and significantly higher scores in
anaphor references. Lower scores concerning stem overlap indicate that the learners
lack depth of vocabulary knowledge (Nation, 2001). Because, unlike native speakers of
English, EFL learners are inadequate in using different parts of speeches of a word;
therefore, learners’ texts include significantly smaller numbers of stem overlaps.
In addition, higher scores in anaphor reference index reveal an overuse of
pronominal references. In the related literature, Biber (1995) states that first person
pronouns serve as markers of interpersonal discourse and direct involvement of the
writer, and they are usually more characteristic of spoken rather than written registers.
This could be the main reason why EFL learners make use of personal pronouns more
than necessary.
Furthermore, parametric comparisons of the learner and native groups in terms
of LSA scores also revealed statistically significant differences between the groups. The
groups were compared through two indices: LSA adjacent and all-distance scores. The
learner group obtained significantly lower scores than both of the native groups in both
indices. As was mentioned before, LSA is a new mathematical technique which is used
to make calculations concerning lexical cohesion. These results could also be regarded
as a validation of a digitally-oriented method for lexical analysis.
These results clearly indicate that the learner group produces lexically
disconnected sentences (Connor, 1984; Silva, 1993; Hinkel, 2001a) with too many
anaphoric references.
135
Research Question 2
Regardless of prompt or average number of words used in the texts, to what
extent do texts written by Turkish EFL learners deviate from texts written by native
speakers of English in terms of readability?
The results of statistical analysis revealed that essays written by Turkish EFL
learners differed significantly from essays written by native speakers of English in
terms of readability. Two readability scores were used for analysis; these were Flesch
and Flesch-Kincaid reading scores. Flesch reading scores vary between 0 and 100 and
the lower the obtained score is, the harder it becomes for the reader to comprehend it.
On the other hand, Flesch-Kincaid reading scores vary between 0 and 12, and this test
rates texts on a U.S. school grade level. For example, a score of 8.0 for a document
means that it can be understood by an eighth grader. The analysis of both scores
revealed that the essays written by the learners were significantly simpler than the
native essays. This outcome confirms the research findings claiming that L2 writing in
English is simpler in structure (Silva, 1993; Hinkel, 2011).
Research Question 3
With average number of words per text being similar, to what extent do texts
written by Turkish EFL learners deviate from texts written by native speakers of English
in terms of syntactic features?
Another finding of the current study concerning syntactic aspects of L2 writing
was that compared to the native group with similar number of words per text, the
learner group used significantly more personal pronouns in their essays. This finding is
in line with the related literature (Granger & Rayson, 1998; Petch-Tyson, 1998; Cobb,
2003; Hinkel, 2001b, 2011); and the common idea emerged from these studies brings
the involved nature of speech into L2 writing. That is, EFL learners are making use of
spoken registers in their writing.
The next result concerning the third research question was that Turkish EFL
learners who participated in the first phase of the current study overused connectives
when compared to the native speakers of English. This outcome is again in line with the
related literature (Silva, 1993; Schleppegrell, 1996; Hinkel, 2001b, 2011; Altenberg &
Tapper, 1998 and Tankó, 2004).
136
At this point, it is quite clear that Turkish EFL learners use too many
connectives and pronominal structures in their writing. In the related literature, there are
references to language-specific conditions for such flaws. For instance, Tankó (2004)
discovered an overuse of adverbial connectors in Hungarian EFL learners’ texts. His
explanation was that the Hungarian language does not require the overt marking of
relations between linguistic units of the text. It is claimed that this difference influenced
the teachers of academic English writing to put more emphasis on the explicit teaching
of adverbial connectors. In addition, the related literature also attributes the overuse of
pronominal references to pragmatic notions; that is, the overuse of personal pronouns
should be counted as the learners’ way of trying to consolidate their ideas that they
present in their essays (Hvitfeld, 1992).
Another result concerning syntactic differences between the learner and the
native group is related to type-token ratio. The results revealed that there is a
statistically significant difference between the learner and the native group concerning
type-token ratio scores. As was mentioned before type-token is related to the lexical
density or the lexical variety in a text; therefore, written productions of the learners who
participated in the current study appear to have less lexical variety, which is supported
by Hinkel’s (2011) overview of the related literature.
There was also a significant difference between the learner and the native group
in terms of the number of the words before the main verb in sentences. McNamara et al.
(2010) relates the number of words before the main verb to syntactic complexity of a
text; and more words before the main verb means that the text is more taxing on
working memory of both the writer and the reader.
Research Question 4
What are learners’ perceptions and feelings about the differences between their
written productions and that of native ones?
One of the common ideas that emerged in focus group interviews was that
instructors put too much emphasis on connectors, and this result is in line with Tankó’s
(2004) findings which were mentioned previously. In our case, the participants in the
focus group interviews stated that they were in a way forced by the instructors to use
more conjunctions in their writing. However, Turkish EFL learners who participated in
the current study shared the idea that they were compensating their lexical inadequacy
137
by using too many referential and connective structures. Nevertheless, learner language
might be operating both ways; that is, while writing, EFL learners might be struggling
to overcome pragmatic concerns along with their lexical inadequacy. EFL learners’
severely limited lexical and syntactic repertoires Hinkel (2002) appear to be playing an
important role at this point. On the other hand, concerning the overuse of certain
structures, some studies in the related literature bring pragmatic issues into the
foreground (e.g. Hvitfeld, 1992). Again, from a broader point of view, it may not be a
matter of either/or, but rather a matter of both/and. In other words, EFL learners might
be overusing connectives and personal pronouns for pragmatic reasons, and by
overusing these structures, they might be compensating for their lexical inadequacy.
Research Question 5
Can concordancing activities induce vocabulary recognition?
The results related to the fifth research question showed that concordancing
activities significantly affected vocabulary recognition. The experimental and control
groups were given an originally constructed multiple-choice vocabulary test as the pre
and post-test. The results of comparisons of the pre-test and post-test scores revealed a
statistically significant difference between the experimental and the control group. The
comparison revealed that the experimental group obtained significantly higher scores in
the post-test given for vocabulary recognition. Semi-structured interviews that were
carried out following the experimental phase confirmed that the learners who
participated in these activities also found concordancing activities beneficial in terms of
vocabulary recognition. This finding has also been confirmed in the related literature
(Cobb, 1997, 1999; Thurstun & Candlin, 1998).
Research Question 6
Can concordancing activities induce vocabulary production?
In the current study, the learners’ writings were analyzed for the target
vocabulary items through pre-test, post-test and delayed post-test; they were computed
as types per 1000 words. The experimental and control groups were compared by using
the results of these computations. The results revealed that the experimental group
scored higher than the control group in the post-test, but the difference was not
statistically significant. However, when the groups were compared by taking the
138
delayed post-test scores into account, a statistically significant difference was
determined. Therefore, the answer to the research question about whether
concordancing activities have any effects on vocabulary production appears to be
affirmative.
Research Question 7
Can concordancing activities induce lexical cohesion in EFL learners’ written
productions?
As was previously mentioned in Chapter 2, there are conventional methods for
cohesion analysis. In the process, reiterations and the use of collocations are analyzed
(Halliday & Hasan, 1976, p. 279). A relatively more recent technique, LSA, was used to
gather data as to the lexical cohesion in Turkish EFL learners’ essays. Firstly, LSA
scores for adjacent sentences were computed and the experimental and the control group
were compared in this respect. Then, all-distance LSA scores were computed and
statistical comparisons were made. These comparisons were performed by taking into
account the pre-test, post-test and delayed post-test LSA scores of the participants.
Compared to pre-test results, there appeared to be slight differences in the post and
delayed evaluations of the experimental group’s LSA scores. However, these
differences were not statistically significant. As an answer to this research question,
concordancing activities did not have any significant effects on lexical cohesion in the
participants’ writing. As the use of LSA scores to measure lexical cohesion in L2
writing lacks proper and reliable literature, no comparison with the relevant literature
can be done about these outcomes.
Research Question 8
What are learners’ perceptions and feelings about the use of concordancing
activities to learn vocabulary?
At the end of the concordancing treatment, 10 participants who were chosen
randomly from the experimental group were interviewed to get a picture of these
activities from their points of view. The results of these semi-structured interviews
revealed positive attitudes among learners towards concordance use to learn new
vocabulary. This outcome is again in line with the related literature indicating EFL
learners’ positive attitudes towards corpus use in language classes (Thurstun & Candlin,
139
1998; Sun, 2000; Yoon & Hirvela, 2004). However, the majority of the participants felt
that these activities did not have any positive impact on their written production.
Another problematic issue at this point is that from their responses or their
enthusiasm during the interviews it was observed that the learners participated in the
current study might have overstated their feelings about these activities. However, when
the results of the statistical analysis for vocabulary recognition and production are taken
into account, participants’ positive attitudes towards these activities do not seem to be
an overestimation because parallel to what they stated in the interviews the participants
did significantly better in both of the vocabulary tests for recognition and production.
5.3. Implications of the Study
The study revealed some problematic areas in Turkish EFL learners’ texts while
some other aspects were reconfirmed. Turkish EFL learners who participated in the
study produced structurally simpler texts, they used too many conjunctions and
pronominal references, and their sentences lacked cohesion. These problems appeared
to have stemmed from participants’ lexical inadequacy. However, without checking it
with our subjects’ native language writing skills, it is hard to say that these outcomes
are universally valid for language learners. That is to say, our subjects might already be
unskilled writers in their native language trying to survive a foreign language by making
do with whatever linguistic repertoire they have at their disposal.
This study could also be regarded as a validation of some of the indices in Coh-
metrix to evaluate English texts, both native and learner, at multiple levels. Some
indices in this database managed to differentiate between native and learner text sets.
Moreover, the same indices were unable distinguish two different native corpora. This
is important in that it paves the way to the possibility of grading learners’ texts digitally,
which would solve the problem of subjectivity in the process.
The use of concordancing activities in vocabulary instruction was the practical
aspect of the study. After a 10-week concordancing treatment significant changes were
recorded in learners’ vocabulary recognition and production. Moreover, the participants
developed positive attitudes towards these activities. Over time, it seems like the phrase
teaching vocabulary will make less sense, and we will hear more of language teachers
building lexical networks together with their students by making use of the real
140
language out there through authentic corpora. The process could start by focusing more
on raising learners’ awareness about connections among words in any text.
Another implication of this study could be the introduction of some textual
parameters in Coh-metrix in EFL reading instruction. As the results of the current study
suggest, the learner groups obtained significantly lower scores in reading ease, type-
token ratio and the number of words before the main verb parameters. These parameters
are of valuable use to language teachers. In practice, reading materials are either readily
available or they are compiled by teachers themselves. In both cases, intuition plays a
big role in deciding the comprehensibility of these materials. These three parameters
can easily be used to determine the difficulty levels of reading materials, which would
provide language teachers with tangible data to verify their intuitions about the
appropriateness of these materials for a desired proficiency level.
5.4. Recommendations for Further Research
One of the results of the current study was that although significant effects were
established concerning vocabulary recognition, concordancing activities did not have
any direct effects on participants’ writing skills; there was no improvement in lexical
cohesion in their writings. This outcome was confirmed through both qualitative and
quantitative analysis. LSA scores of the participants did not seem to change after about
10 weeks of concordancing activities. Integration of concordancing activities directly
into writing classes and analyzing their effects in terms of lexical cohesion could be a
subsequent step to be taken. During the integration process, learners could be trained
about using concordancing to improve their writing skills, which is consistent with the
literature that has argued for the need to train learners for successful use of corpora
(Cobb, 1997; Flowerdew, 1996; Kennedy & Miceli, 2001; Yoon & Hirvela, 2004).
Another potentially fruitful research topic could be the use of LSA in Turkish
EFL learners’ spoken production. Comparison of learners’ spoken and written
productions by making use of LSA might help practitioners to find out the similarities
and differences between the two registers in terms of lexical cohesion. By comparing
the outcomes with native corpora, writing and speaking instructions in EFL classes
could be modified and improved.
141
5.5. Personal Reflections and Criticism of the Study
This study was an attempt to determine micro-level differences between Turkish
EFL learners’ and native texts, and the focal point was the lexical differences between
the essays written by the groups. In this study, these differences were confirmed twice,
with 49 learners in the first phase and with 37 learners in the second. In order to help
students deal with lexicon-related problems, corpus-based vocabulary learning materials
were developed. Both post-test and semi-structured interview results revealed that these
materials were helpful to the participants to some extent. However, corpus-based
learning activities did not have significant effects on lexical cohesion in learners’
essays. The main reason for this was the lack of integration of these activities directly
into writing sessions. This integration was unfeasible because of time constraints; the
ESP program that the learners took was only four hours a week, and the program
included several skills such as listening, reading and writing.
Another point worth mentioning here is that the participants in the experimental
group were confused about the concordancing activities at the beginning of the
treatment. This confusion lasted for about three weeks after which most of the
participants eventually started to make sense out of these activities. This showed that in
order to obtain effective results from corpus-based activities, training the participants
beforehand is more than necessary.
Looking back at this two-year experience, I can unhesitatingly state that having
access to corpora in any way in a language learning environment is like having many
native speakers in the classroom, which can be a confidence booster for language
teachers.
All in all, if I were asked to summarize the current study with a couple of words,
knowing that nothing decent can be summarized, I would have a quick but incomplete
answer: This study includes clear answers to clear questions.
142
REFERENCES
Aarts, B. (2001). Corpus linguistics, Chomsky and fuzzy tree fragments. In C. Mair &
M. Hundt (Eds.), Corpus linguistics and linguistic theory (pp. 5–13). Amsterdam:
Rodopi.
Altenberg, B. & Tapper, M. (1998). The use of adverbial connectors in advanced
Swedish learners' written English. In S. Granger (Ed.), Learner English on
computer (pp. 80–93). London: Addison Wesley Longman.
Allen, D. (1992). Oxford placement test 2 (New edition). Oxford: Oxford University
Press.
Belz, J. A. (2004). Learner corpus analysis and the development of foreign language
proficiency. System, 32(4), 577–591.
Bernardini, S. (2004). Corpora in the classroom: An overview and some reflections on
future developments. In J. Sinclair (Ed.), How to use corpora in language
teaching (pp. 15–36). Amsterdam and Philadelphia, PA: John Benjamins.
Biber, D. (1993). Representativeness in corpus design. Literary and linguistic
computing, 8, 1–15.
Biber, D. (1995). Dimensions of register variation. Cambridge: Cambridge University
Press.
Biber, D. & Reppen, R. (2002). What does frequency have to do with grammar
teaching? Studies in Acquisition, 24, 199–208.
Braun, S. (2005). From pedagogically relevant corpora to authentic language learning
contents. ReCALL, 17(1), 47–64.
Brown, D. H. (2007). Principles of language learning & teaching. New York: Longman
Chambers, A. (2005). Integrating corpus consultation in language studies. Language
Learning & Technology. 9(2), 111–125.
Cheng, W., Warren, M., Xun–Feng, X. (2003). The language learner as language
researcher: putting corpus linguistics on the timetable. System 31(2), 173–186.
Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press.
Cobb, T. (1997). Is there any measurable learning from hands–on concordancing?
System, 25(3), 301–315.
Cobb, T. (1999). Breadth and depth of vocabulary acquisition with hands–on
concordancing. Computer Assisted Language Learning, 12, 345–360.
143
Cobb, T. (2003). Analyzing late interlanguage with learner corpora: Québec replications
of three European studies. The Canadian Modern Language Review/La Revue
canadienne des langues vivantes, 59(3), 393–423.
Cohen, L., Manion, L. & Morrison, K. (2007). Research methods in education. New
York: Routledge.
Coltheart, M. (1981). The MRC psycholinguistic database. Quarterly Journal of
Experimental Psychology, 33, 497–505.
Connor, U. (1984). A Study of cohesion and coherence in English as a second language
students’ writing. Papers in Linguistics, 17(1–4), 301–316.
Connor, M. (1996). Contrastive rhetoric: Cross–cultural aspects of second–language
writing. Cambridge, UK: Cambridge University Press.
Conrad, S. (1999). The importance of corpus–based research for language teachers.
System, 27(1), 1–18.
Conrad, S. (2005). Corpus linguistic and L2 teaching. In E. Hinkel (Ed.), Handbook of
research in second language teaching and learning (pp. 393– 409). Mahwah, NJ:
Lawrence Erlbaum.
Cook, G. (1989). Discourse. Oxford: Oxford University Press.
Cook, G. (1998). The uses of reality: A reply to Ronald Carter. ELT Journal, 52(1),
57–63.
Coxhead, A. (2000). A new academic word list. TESOL Quarterly, 34(2), 213–38.
Coxhead, A. (2010). What can corpora tell us about English for Academic Purposes? In
A. O'Keefe & M. McCarthy (Eds.), The Routledge handbook of corpus linguistics,
(pp. 458–470). Oxford: Routledge.
Crossley, S. A., Salsbury, T. McCarthy, P. M., & McNamara, D. S. (2008). Using
Latent Semantic Analysis to explore second language lexical development. In D.
Wilson & G. Sutcliffe (Eds.), Proceedings of the 21st international Florida
artificial intelligence research society (pp. 136–141). Menlo Park, California:
AAAI Press.
Crossley, S. A., & McNamara, D. S. (2009). Computationally assessing lexical
differences in second language writing. Journal of Second Language Writing, 17
(2), 119–135.
David, A. (2008). A developmental perspective on productive lexical knowledge in L2
oral interlanguage. Journal of French Language Studies, 18(3), 315–331.
Dawson, C. (2002). Practical research methods. Wiltshire: Cromwell Press.
144
Denzin, N. K. (1978). The Research Act: A Theoretical Introduction to Sociological
Methods. Englewood Cliffs, N. J. :Prentice Hall.
Dörnyei, Z. (2007). Research methods in applied linguistics. Oxford: Oxford University
Press.
Ellis, R. (1995). Modified oral input and the acquisition of word meanings. Applied
Linguistics, 16, 409–435.
Ellis, R. & Barkhuizen G. (2005). Analyzing learner language. Oxford: Oxford
University Press.
Fan, M. (2007). An exploratory study of collocational use by ESL students –
A task based approach. System, 37(1), 110–123.
Ferrer i Cancho, R., & Sole´, R. V. (2001). The small world of human language.
Proceedings of the Royal Society of London, B, 268, 2261–2265.
Ferris, D. R. (1994). Lexical and syntactic features of ESL writing by students at
different levels of L2 proficiency. TESOL Quarterly, 28, 414–420.
Field, A. (2009). Discovering statistics using SPSS. London: Sage Publications
Finch, G. (2005). Key concepts in language and linguistics. New York: Palgrave
Macmillan.
Fligelstone, S. (1993). Some reflections on teaching, from a corpus linguistics
perspective. ICAME Journal, 17, 97–109.
Flowerdew, J. (1996). Concordancing in language learning. In M. Pennington, The
Power of CALL (pp. 97–113). Houston/Texas: Athelstan.
Flowerdew, J. (2009). Use of signaling nouns in a learner corpus. In J. Flowerdew & M.
Mahlberg, (Eds.), Lexical cohesion and corpus linguistics, (pp. 85-102).
Amsterdam & Philadelphia: John Benjamins.
Frase, L., Faletti, J., Ginther, A., & Grant, L. (1999). Computer analysis of the TOEFL
test of written English (Research report 64). Princeton, NJ: Educational Testing
Service.
Gass, S., Selinker, L. (2008). Second language acquisition: An introductory course.
New York, NY: Routledge.
Gavioli, L. & Aston, G. (2001). Enriching reality: language corpora in language
pedagogy. ELT Journal, 55(3), 238–246.
Graesser, A. C., McNamara, D. S., Louwerse, M. M., & Cai, Z. (2004). Coh–Metrix:
Analysis of text on cohesion and language. Behavioral Research Methods,
Instruments, and Computers. 36, 193–202.
145
Granger, S. (1993). The international corpus of learner English. In J. Aarts, P. de Haan,
& N. Oostdijk (Eds.), English language corpora: Design, analysis and
exploitation (pp. 57–69). Amsterdam: Rodopi.
Granger, S. (1994). The learner corpus: A revolution in applied linguistics. English
Today, 39 (10/3), 25–29.
Granger, S. & Tyson, S. (1996). Connector usage in the English essay writing of native
and non–native EFL speakers of English. World Englishes, 15, 19–29.
Granger, S. (Ed.). (1998). Learner English on computer. London: Longman.
Granger, S., Rayson, P. (1998). Automatic lexical profiling of learner texts. In S.
Granger (Ed.), Learner English on Computer (pp. 119–131). London & New
York: Addison Wesley Longman.
Granger, S. (1999). Use of tenses by advanced EFL learners: Evidence from an error–
tagged computer corpus. In H. Hasselgård and S.Oksefjell (Eds.), Out of Corpora
– Studies in honour of Stig Johansson (pp. 191–202). Amsterdam & Atlanta:
Rodopi.
Granger, S. (2003). The international corpus of learner English: A new resource for
foreign language learning and teaching and second language acquisition research.
TESOL Quarterly, 37(3), 538–545.
Granger, S. (2008). Learner corpora. In A. Lüdeling, M. Kytö, (Eds.), Corpus
linguistics: An international handbook (pp. 259–275). Berlin and New York:
Mouton De Gruyter.
Haastrup, K., & Henriksen, B. (2000). Vocabulary acquisition: Acquiring depth of
knowledge through network building. International Journal of Applied
Linguistics, 10(2), 221–240.
Halliday, M. A. K., & Hasan, R. (1976). Cohesion in English. London: Longman.
Harvey, K., & Yuill, D. (1997). A study of the use of a monolingual pedagogical
dictionary by learners of English engaged in writing. Applied Linguistics, 18(3),
253 – 278.
Hinkel, E. (2001a). Matters of cohesion in L2 academic texts. Applied Language
Learning, 12, 111–132.
Hinkel, E. (2001b). Giving examples and telling stories in academic essays. Issues in
Applied Linguistics, 12, 149–170.
Hinkel, E. (2002). Second language writers’ text. Mawah, NJ: Lawrence Erlbaum
Associates.
146
Hinkel, E. (2005). Analyses of second language texts and what can be learned from
them. In E. Hinkel (Ed.), Handbook of research in second language teaching and
learning (pp. 615–628). Mahwah, NJ: Lawrence Erlbaum.
Hinkel, E. (2005). Introduction. In E. Hinkel (Ed.), Handbook of research in second
language teaching and learning (pp. xvii–xxii). Mahwah, NJ: Lawrence Erlbaum.
Hinkel, E. (2011). What research on second language writing tells us and what it
doesn't. In E. Hinkel (Ed.), Handbook of research in second language teaching
and learning, Volume 2 [uncorrected proof], (pp 523-538). New York: Routledge.
Retrieved November, 10.2010 from http://www.elihinkel.org/downloads.htm.
Hvitfeld, C. (1992). Oral orientations in ESL academic writing. College ESL, 2(1), 29–
39.
Johns, T. (1986). Micro–Concord: A language learner’s research tool. System, 14(2),
151–62.
Johns, T. (1991). Should you be persuaded: two samples of data-driven learning
materials. In T. Johns, P. King (Eds.), Classroom Concordancing (pp. 1–16).
ELR: Birmingham.
Kennedy, C. & Miceli T. (2001). An evaluation of intermediate students’ approaches to
corpus investigation. Language Learning & Technology, 5 (3), 77–90.
Landauer, T. K., and Dumais, S. T. (1997). A solution to Plato's problem: The Latent
Semantic Analysis theory of the acquisition, induction, and representation of
knowledge. Psychological Review, 104, 211–240.
Landauer, T. K., Foltz, P. W, & Laham, D. (1998). Introduction to Latent Semantic
Analysis. Discourse Processes, 25, 259–284.
Leech, G. (1992). Corpora and theories of linguistic performance. In J. Svartvik (Ed.),
Directions in Corpus Linguistics (pp 105–122). Berlin: Mouton de Gruyter.
Leech, G. (1998). Learner corpora: what they are and what can be done with them. In S.
Granger (Ed.), Learner English on computer (pp. xiv–xx). London: Longman.
Lewis, M. (1993). The lexical approach: The state of ELT and a way forward. Hove:
Language Teaching Publications.
Long, M. H. (2003). Stabilization and fossilization in interlanguage development. In M.
H. Long & C. Doughty (Eds.), Handbook of second language acquisition (487–
535). Oxford: Blackwell.
Louwerse, M. M. (2004). A concise model of cohesion in text and coherence in
comprehension. Revista Signos, 37, 41–58.
147
Mackey, A. & Gass, S.M. (2005). Second language research: Methodology and design.
Mahwah, NJ: Lawrence Erlbaum.
Mahlberg, M. & Flowerdew J. (2009). Lexical cohesion and corpus linguistics.
Amsterdam: John Benjamins.
McEnery, A. M., Wilson, A. (1997). Teaching and language corpora. ReCALL, 9(1): 5–
14.
McGee, I. (2009). Traversing the lexical cohesion minefield. ELT Journal , 63 (3), 212–
220.
McNamara, D.S., Crossley, S.A. & McCarthy, P.M. (2010). Linguistic features of
writing quality. Written Communication, 27, 57–86.
Meara, P. (1978). Learners’ word associations in French. Interlanguage Studies
Bulletin-Utrecht, 3(2), 192–211.
Meara, P. (2002). The discovery of vocabulary. Second Language Research, 18(4),
393–407.
Miles, M. B. & A. M. Huberman. (1994). Qualitative data analysis. Thousand Oaks,
California : Sage.
Milton, J. (1998). Exploiting L1 and interlanguage corpora in the design of an electronic
language learning and production environment. In S. Granger (Ed.), Learner
English on computer (pp. 186–198). London & New York: Addison Wesley
Longman.
Mitchell, R., & Myles F. (1998). Second language learning theories. London: Arnold.
Nation, I. S. P. (1990). Teaching and learning vocabulary. New York: Newbury House.
Nation, I. S. P. (2001). Learning vocabulary in another language. Cambridge:
Cambridge University Press.
Nation, I. S. P. (2002). Best Practice in Vocabulary Teaching and Learning. In C.
Richards Jack C. & A. Willy Renandya (Eds.), Methodology in language
teaching: An anthology of current practice (pp. 267–272). Cambridge: Cambridge
University Press.
Nation, I. S. P. (2005). Teaching and learning vocabulary. In E. Hinkel (Ed.), Handbook
of research on second language teaching and learning (581–596). Mahwah, New
Jersey: Lawrence Erlbaum Associates.
Nesselhauf, N. (2005). Collocations in a learner corpus. Amsterdam: John Benjamins.
O’Keeffe, A., McCarthy, M. J. and Carter, R. A. (2007). From corpus to classroom:
Language use and language teaching. Cambridge: Cambridge University Press.
148
Olshtain, E. (1987). The acquisition of new word formation processes in second
language acquisition. Studies in Second Language Acquisition, 9(2), 221–231.
Palmer, J. C. (1999). Coherence and cohesion in the English language classroom: The
use of lexical reiteration and pronominalisation. RELC Journal, 30, 61–85.
Pica, T. (2005). Second language acquisition research and applied linguistics. In E.
Hinkel (Ed.), Handbook of research in second language teaching and learning
(263–280). New Jersey: Lawrence Erlbaum Associates, Inc.
Petch–Tyson, S. (1998). Writer/reader visibility in EFL written discourse. In S. Granger
(Ed.), Learner English on computer (107–118). London: Longman.
Richards, J. C. & Schmidt, R. (Eds.) (2002). Longman dictionary of language teaching
and applied linguistics. New York: Longman.
Ritchie, J. & Spencer, L. (1994). Qualitative data analysis for applied policy research. In
A. Bryman & B. Burgess (Eds.), Analyzing qualitative data (pp. 173–194).
London: Routledge.
Römer, U. (2005). Progressives, patterns, pedagogy: A corpus–driven approach to
English progressive forms, functions, contexts, and didactics. Amsterdam: John
Benjamins.
Salkie, R. (1995). Text and discourse analysis. Florence, KY, USA: Routledge
Schleppegrell, M. (1996). Conjunction in spoken English and ESL writing. Applied
Linguistics, 17(3), 271–285.
Schmitt, D. & Schmitt, N. (2005). Focus on vocabulary: Mastering the academic word
list. White Plains, NY: Pearson.
Silva, T. (1993). Toward an understanding of the distinct nature of L2 writing: The ESL
research and its implications. TESOL Quarterly, 27(4), 657–77.
Sinclair, J. M. (1991). Corpus, concordance and collocation. Oxford: Oxford
University Press.
Sinclair, J. M. (2003). Reading concordances. London: Longman.
Sinclair, J. M. (2004). Trust the text: Language, corpus and discourse. London:
Routledge.
Sinclair, J. M. (2005). Corpus and text: Basic principles. In M. Wynne (Ed.),
Developing linguistic corpora: A guide to good practice (1–16). Oxford: Oxbow
Books.
Stubbs, M. (2004). Language corpora. In A. Davies & C. Elder (Eds.), The handbook of
applied linguistics (106–132). London: Blackwell Publishing.
149
Sun, Y. C. (2000). Using on–line corpus to facilitate language learning. Paper presented
at the annual meeting of the teachers of English to speakers of other languages.
British Columbia: Canada.
Tankó, G. (2004). The Use of Adverbial Connectors in Hungarian University Students’
Argumentative Essays. In J. Sinclair (Ed.), How to Use Corpora in Language
Teaching (157–181). Amsterdam: John Benjamins.
Thornbury, S. (2002). How to Teach Vocabulary. London: Longman.
Thurstun, J. & Candlin, C. (1998). Concordancing and the teaching of the vocabulary of
academic English. English for Specific Purposes, 17, 267–280.
Tognini Bonelli, E. (2001). Corpus linguistics at work. Amsterdam: John Benjamins.
Vannestål, E. & Lindquist, M. H. (2007). Learning English grammar with a corpus :
Experimenting with concordancing in a university grammar course. ReCall, 19,
329–350.
Ward, J. (2007). Collocation and Technicality in EAP Engineering. Journal of English
for Academic Purposes, 6, 18–35.
Weber, J. (2001). A concordance– and genre–informed approach to ESL essay writing.
ELT Journal, 55, 14–20.
Widdowson, H. G. (2000). On the limitations of linguistics applied. Applied Linguistics,
21(1), 3–5.
Wolter, B. (2001). Comparing the L1 and L2 Mental Lexicon. Studies in Second
Language Acquisition, 23, 41–69.
Yoon, H. & Hirvela, A. (2004). ESL Student attitudes towards corpus use in L2 writing.
Journal of Second Language Writing, 13, 257–83.
Zareva, A. (2007). Structure of the second language mental lexicon: how does it
compare to native speakers’ lexical organization? Second Language Research,
23 (2), 123–153.
150
APPENDICES APPENDIX 1 - Indices of Coh-Metrix Text Information Index Index Name
1 Title 2 Genre 3 Source 4 Job Code 5 LSA Space 6 Date
Referential and Semantic Aspects Index Index Name
7 Adjacent anaphor reference 8 Anaphor reference 9 Adjacent argument overlap
10 Argument overlap 11 Adjacent stem overlap 12 Stem overlap 13 Content word overlap 14 LSA sentence adjacent 15 LSA sentence all 16 LSA paragraph
Situation model dimensions Index Index Name
20 Causal content 21 Causal cohesion 22 Intentional content 23 Intentional cohesion 27 Temporal cohesion 28 Spatial cohesion
Syntax Index Index Name
17 Personal pronouns 18 Pronoun ratio 19 Type-token ratio 24 Syntactic structure similarity adjacent 25 Syntactic structure similarity all 26 Syntactic structure similarity, sentence all 29 All connectives 30 Conditional operators 31 Pos. additive connectives 32 Pos. temporal connectives 33 Pos. causal connectives 34 Pos. logical connectives 35 Neg. additive connectives 36 Neg. temporal connectives 37 Neg. causal connectives 38 Neg.logical connectives 39 Logic operators 48 Negations 49 NP incidence 50 Modifiers per NP 51 Higher level constituents 52 Mean number of words before the main verb of main clause in sentences
151
General Word and Text Information Index Index Name
40 Raw freq. content words 41 Log freq. content words 42 Min. raw freq. content words 43 Log min. freq. content words 44 Concreteness content words 45 Min. concreteness content words 46 Noun hypernym 47 Verb hypernym 53 No. of words 54 No. of sentences 55 No. of paragraphs 56 Syllables per word 57 Words per sentence 58 Sentences per paragraph
Readability Index Index Name
59 Flesch Reading Ease Score 60 Flesch-Kincaid Reading Ease Score
152
APPENDIX 2 - Academic Vocabulary Scan Test Name/Surname ………………………………………………… 1) The television report stated that many people had died, but didn't _________ the exact number. a)occur b)specify c)achieve d)survey 2) A biopsy is needed to _________whether she actually has cancer or not. a)confirm b)assume c)response d)design 3) The most recent survey _________ a high level of dissatisfaction with the current government. a)pursues b)exceeds c)indicates d)displays 4) Scientists estimate that hydrogen _________ around 90 to 99 percent of all matter in the universe. a)imposes b)removes c)comprises d)aids 5) Online communications can _________ learning in many different ways. a)deny b)impact c)exclude d)require 6) I think that the _________ of students that have to repeat a level in this program is usually about 25% or less. a)precision b)percentage c)contrast d)section 7) It is _________ that you get regular exercise if you want to lose weight. a)crucial b)inherent c)prohibited d)constant 8) Chemical compounds which are not _________ can be very dangerous. a)schematic b)stable c)internal d)precise 9) The _________ of your language skills will include a speaking test and an essay assignment. a)phase b)assessment c)reinforcement d)structure 10) The _________ thing to do if you're feeling sick is to go to bed. a)preliminary b)ultimate c)phenomenal d)obvious 11) The temperature inside the aquarium needs to _________ constant throughout the year; otherwise, the fish will die. a)suspend b)trigger c)remain d)detect 12) The most _________ minerals to the human body are: salt for maintaining water levels, iron for red blood cells, and calcium for bones. a)minor b)physical c)dynamic d)essential 13) The supervisor has announced that our first staff meeting will last for two hours, and _________ meetings for only one hour. a)subsequent b)capable c)precise d)accurate 14) Visitors' parking can be found _________ to the main entrance to the apartment complex. a)confined b)specified c)adjacent d)alternative 15) The day is _________ when computers will be a part of every person's daily life everywhere in the country. a)regulating b)ensuring c)channeling d)approaching 16) Too many children in this country are leaving home in the morning without having eaten a/an _________ meal. a)adequate b)major c)individual d)preliminary
153
17) Some diseases are _________ at birth, and can be dealt with right away. a)structural b)detectable c)automated d)modified 18) The victims were put in _________ to keep the other patients from being exposed to the disease. a)isolation b)layer c)process d)source 19) In 1876, a spokesman for Western Union suggested that the telephone was a/an _________ of no value. a)design b)factor c)device d)process 20) It can be difficult for young children to _________ to a new school. a)revise b)adjust c)coordinate d)schedule 21) The _________ of the media is just one of many influences that shape a child's attitudes and behaviors. a)revision b)circumstance c)aspect d)impact 22) Nearly 50% of deaths among children and teenagers aged 10 to 19 are due to _________ causes, usually car accidents. a)subsequent b)appropriate c)external d)initial 23) The earthquake occurred at _________ 9:02 this morning. a)subsequently b)precisely c)obviously d)considerably 24) In our legal system, the _________ is that you are innocent until proven guilty. a)presumption b)component c)consistency d)capacity 25) Many people were _________ to radiation after the accident at the nuclear power plant. a)isolated b)restricted c)triggered d)exposed 26) The primary function of the space shuttle is to carry personnel and _________ into space. a)equipment b)components c)sections d)dimensions 27) It is difficult to _________ the effectiveness of the medication after such a short time. a)insert b)evaluate c)structure d)attach 28) Some people prefer to use natural herbs as a/an _________ to prescription drugs. a)suspension b)alternative c)reinforcement d)summary 29) Bill Gates' continued efforts to _________ and change have helped to keep Microsoft at the top of the computer world. a)clarify b)innovate c)assume d)submit 30) The city usually puts _________ on water use in the summertime in order to prevent shortages. a)panels b)vehicles c)restrictions d)exposures 31) You can expect minor _________ in your weight during the time you are doing heavy exercise. a)consumptions b)fluctuations c)routes d)estimations 32) The various components are manufactured by different suppliers around the world, and _________ in our factory in Mexico. a)assembled b)affected c)attained d)exploited 33) The patient is bleeding _________, and needs to be operated on immediately. a)professionally b)positively c)dynamically d)internally 34) Children with learning disabilities _________ a lot of assistance at school. a)demonstrate b)require c)inhibit d)reinforce
154
35) Our response to the controversial social _________ of our time will determine our future. a)issues b)authorities c)sites d)motives 36) Steel is 100 percent recyclable, and can be _________ almost indefinitely. a)detected b)concluded c)reprocessed d)graded 37) A space _________ must move at a rate of at least 15 miles per second to escape Earth's gravitational pull. a)complement b)vehicle c)element d)layer 38) The tests we are giving you have been carefully _________ to identify your weak points in English. a)located b)constructed c)clarified d)indicated 39) The leader of the party is trying to _________ his position by surrounding himself with his most loyal supporters. a)attach b)reinforce c)resolve d)detect 40) Protein synthesis is a/an _________ process in which DNA is transformed into protein. a)sequential b)obvious c)consistent d)ultimate 41) There is a noticeable lack of _________ in his work. One day he does really well, and then the next day he can't seem to get anything done. a)access b)location c)issue d)consistency 42) The _________ economy of the U. S. is the envy of the world. a)random b)dynamic c)eventual d)visual 43) The newest video games will be on _________ at the fair. a)display b)route c)alert d)schedule 44) Many car drivers in Sweden turn on their headlights during the day in order to be more _________ to other drivers. a)sufficient b)manual c)alert d)visible 45) The witness gave a/an _________ description of the criminal - she even knew his eye color! a)alternative b)preliminary c)accurate d)minor 46) The two programs _________, so students in the first program had to miss the first few days of classes in the second program because of final exams. a)overlapped b)evolved c)involved d)contracted 47) Eating plenty of fruits, vegetables, protein and dairy products will _________ your body gets the minerals it needs. a)equip b)ensure c)regulate d)channel 48) Current _________ forbid the use of company cars for personal reasons. a)factors b)suspension c)regulations d)areas 49) He is an expert in the _________ of the effects of pollution on ocean plankton. a)vehicle b)area c)section d)layer 50) The doctor _________ a tube in the patient's nose for oxygen. a)inserted b)restricted c)authorized d)isolated
155
APPENDIX 3- Argumentative Essay Topics for the Written Tests Name/Surname …………………………………………….. Choose one of the topics and write a well-developed three-paragraph essay. Try to use 400 words in total.
1. Money is the root of all evil. 2. In his novel Animal Farm, George Orwell wrote "All men are equal: but some are more equal
than others". How true is this today? 3. Feminists have done more harm to the cause of women than good. 4. Most university degrees are theoretical and do not prepare students for the real world. They are
therefore of very little value. 5. The prison system is outdated. No civilized society should punish its criminals: it should
rehabilitate them.
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
....................................................................................
156
APPENDIX 4 - A Sample Concordancing Activity (Teachers Copy) Activity 1: Match the following vocabulary items with their concordances*.
1) inherent ………………….
a) first, data, peace, search, time, stage, parallel, possible, system, number, food, form, during, environment, single, social, development, manufacturing, industrial (3)
2) precise ………………….
b) extraordinary, specific, obtained, year, known, result, amount, state, profit, tax, certain, first, search, used, particular, disclose, every shown user (12)
3) process …………………. c) greater, only, such, time, new, world, defined,
possible, best, company, knowledge, like, need, recall, use (2)
4) plus …………………. d) length, width, depth, allowance, equal, cost,
enough, extra, measurements, number, per, (4)
5) contrast …………………. e) dangers, meanings, due, resolution, system (1)
6) accurate …………………. f) high, used, fairly, made, measurements, system,
time (6)
7) specify …………………. g) color, sharply (5)
8) automate …………………. h) library, new, office, (8)
9) vehicle ………………….
i) free, number, input, process, different, values, form, random, rate, temperature, widely, individual, age, degrees, prices, considerable, costs, data, example (10)
10) variation …………………. j) number, form, appendix, completed, entries,
actions, format, represents, section (7)
11) element …………………. k) state, motor, management, cars, military,
purchase (9)
12) item …………………. l) fixed, experience, sophisticated, air, single,
submarine (11)
*Concordance: In a text, certain words match other words; they occur in similar contexts. These words are called concordances.
157
Activity 2: Find the words from activity 1that match the blanks. Write it in the blanks provided above each box.
1. ………………………………… (process)
is in pork introduces another factor which must be dealt with in food _______ing. To permit the storage of food for long periods of time, a mal management problems of current and next-generation high-power micro_______ors, which increasingly exhibit the properties of furnaces like researchers in other subjects, make use of computers for word- _______ing, for cataloguing the books in their departmental libraries, searches as well, over- simplifying what is in fact a very complex _______. In order to get a response in subject searching, the user has need to be investigated as an integral part of the information-seeking _______. Users in the past have shown a preference for the direct shelf
2. ………………………………… (specify)
searching, the user has had not only to specify his needs but also to _______ them in a way that " matches" the system. Unlike users of some topic or situation and that in general, the user is unable to _______ precisely what is needed to resolve that anomaly." An IR system orbits in the system can be partially described by an integer n _______ing the number of times that they wind around the z-axis. Such plied them from a mobile lithograph press. Orders of the day began to _______ the standard map for the movement. Sherman proved that a mentioned house trailers, and two others referred to trailers without _______ing the type. In two cases, airplanes only were indicated. It is
3. ………………………………… (plus)
Lining To the finished width measurement by length measurement ____ an extra 7cm (2 ¾in) for each lath channel positioned at t Blind fabric and lining to the width and length measurements, ____ an additional 3cm (1 ¼in) on width and length for turnings builder, is convinced that the total cost of all the heating systems ____ the oil distribution system is no greater than would be gas heating As in the United States, there is a flat fee-per-day rental charge ____ a few cents per kilometer driven, and the per-day rate drops if t effectiveness of control on all of the area now under treatment ____ the additional acres so that after the initial period only maintenance
4. ………………………………… (accurate)
100 times more sensitive and yields numerical results which can be ________ly repeated at will over a period of time. If a wedge-shaped co (or interstage coupler) is utilized, while a magnetic system requires ________ adjustment of the solenoid, which is heavy and bulky. As it w to Spirito. He drew toward the composite design from his meticulously ________ memory, without need to consult his sketches. Soon he was asurements in between these. The more measurements you take, the more ________ly you can plot the curve (fig. 49). Take width measurement were to be reduced as far as possible and they should be such that an ________ heat balance can be made. In order to reduce the number of
5. ………………………………… (automate)
jobs than the production industries which are becoming more and more ________d. Typical recent examples must be the many importers of Tunnel linings are manufactured at a specially constructed and highly ________d plant on the Isle of Grain in the Thames Estuary and in error? In many cases, mobile equipment of this kind is completely ________d and not designed for carrying people. Even in the back: the key to interaction In attempting to help users in searching ________d catalogues, it is perhaps too easy to assume that the hours a day. In the last few years the telephone company has managed to ________ many areas of their service. It has not been any great
6. ………………………………… (inherent)
more detailed nature. It is recognized that a mail questionnaire has ________ limitations. There is the danger that the questions will mean a thing of the past. International events have shown that there is an ________ contradiction between a one- party state and mass democracy. particularly through keyword access, seems to have encouraged an ________ need for subject searching. Thus the ability to manipulate the unit would eliminate the differences and complications that are ________ in a system of 39 different and independent assessing units". demands. This weakness is not unique to labor surplus areas, for it is ________ in the system of local school districts in this country. Plan
7. ………………………………… (vehicle)
due to the effect of wind gusts, engine noise, turbulence around the _______, etc. One of the greatest problems associated with automatic l Company, a medium-sized firm which manufactured four- wheel-drive _______s and other off-road equipment, had recently constructed an State funding of the provision of mass- produced easily accessible _______s which can be adapted for disabled use will give real mobility odium and ran off. The main thing was that the parade by the military _______s did not take place. Mr. Roshka said that several members design. The horseless carriages of the first decade had made way for _______s designed as motor cars in their own right. The great British
8. ………………………………… (variation)
on the ground), though it is certainly that as well. In fact all the _________s are dominated by the intervals of the ground, and what is a German electoral law was amended so as to impose stricter limits on _________s in constituency magnitudes. This resulted in the loss of the or as the upper section of Table 8.2 shows, there can be considerable _________s between different estimates, although the data do provide b same rate either for the daily rate or rate per kilometer driven. The _________s are not too great. Rates for American cars are somewhat high 11 inches long. What data there are on growth indicate considerable _________ in rate; unfortunately, no one has kept complete records of
9. ………………………………… (item)
determine which combination of procedures is practical for any specific ____ in order to evaluate the dimensional changes of textile fabrics have to be undertaken to take special care with potentially flammable ____s and those which may be chemically/physically unstable when management or to quantity overall use of library, materials apart from ____s recorded in circulation statistics. Browsing: a definition Vario use of a tool, such as a library catalogue or bibliography, to locate ____s on a specific topic. General and general purposive browsing diff phase of the state planning program. This phase consists of four ____s: urban land use, rural land use, physical features and public
158
10. ………………………………… (precise)
variables of pressure, volume, temperature, stress and strain have been _______ly formulated. Simple elongation has been treated in detail. occam constructors work. We see how algebraic laws allow us to give a _______ and succinct description of each operator. The laws given spiral at once into C+; this is how one determines experimentally the _______ parameter value at which the strange set becomes attracting. is sufficient. Finally, do not let the following calculations and _______ measurements deter you. It is possible to juggle the pleating only one sharpening. As soon as the time came for re- sharpening, the _______ form of the gear tooth was lost and a new cutter had to be mad
11. ………………………………… (contrast)
reported a profits slide of ten per cent and saw its shares slip. In ________, construction and minerals group English China Clays jumped d The product uses an 8-line super twist LCD screen with adjustable ________ and gives excellent legibility. The user interface looks like xenon and helium- xenon lasers have given to laser instabilities. In ________ to the Lorenz model of a homogeneously broadened laser discus more muted and darker on a rough, softer surface. Texture provides ________ in a room, and adds further interest and another dimension. P nor a collection of isolated and neutral sensory qualities. In ________ to all this, primary data are data of a self involved in envi
12. ………………………………… (elements) divisions, the hours of night. Thus there was a confusing number of _______s on earth, above it and below it which contributed to the after the greatest importance. Consequently, air, surface, and submarine _______s overshadow the mine, fixed installations, and intelligence. on projected potentialities. Then the enthusiasm and energy of all _______s can be channeled to produce cumulative progress toward a be 6 percent and 8 percent. Thus, the combined efficiency of the _______s replaced by the two fiber plates (with a combined efficiency analysis for parameter values near point X is much harder, involving _______s of all the other analyses mentioned so far, but it is at leas
159
APPENDIX 5 – Sample Learner Essays A Sample Essay from the Experiment Group Written as the Pre-test. LSA, Adjacent sentences : 0,097 LSA, All combinations : 0,054 Everyone has rights which are obtained from coming to the world. There rights are structure of living happily and independence. However, all the time there are people want more and more. They envy others and they never see themselves. In that point the people I mean, ruin balance or try to ruin. It's certain that every one of us knows that inequality exists. George Orwell wrote 'Animal Farm' many decades ago. He was discussing about Russian management system. He found weak points on their big rules. Not only based on Russia, also a lot of countries live problems about strength of equality. I strongly believe that there isn't equality. It's normal. We should think about how we can close the difference of high and low. It can be only with understanding and conscious people. Now it’s time of education. It's a simple thing to talk much about problems. Creating solution is vital thing. Furthermore, humans have the power of mind and they can do it best. In early of our lives we can't choose our parents, our language, our country and our fate. Inequality begins here. But inequality I mean isn't same as inequality I explained. Subsequently, we should turn our sights to the education for tomorrow's sake. In the novel George Orwell says "All animals are equal; pigs are more equal. We just aim not to grow people who has characteristic of pigs.”
160
A Second Sample Essay from the Experiment Group Written as the Pre-test. LSA, Adjacent sentences : 0,3 LSA, All combinations : 0,258
As the time passes social construction based on equality is unfortunately leaving its place to category equality. But, what is category based equality? Although people seem to be treated equally by the law, this equality sometimes changes due to appearance or/and financial, social and intelligence status. You can understand the fact that life is not fair; once you are not categorized as beautiful, handsome, rich or strong.
To give a simple example; if you are not clever enough to attract your teacher
between your peers, you sit on the back desks of the classroom throughout your education life. Or vise versa, if you attract your teacher with your intelligence; certainly, you will be one step further than your peers. From another angle; if you are not pretty enough you can easily realize the waiter’s interest, which you lack of although it is a café you frequently visit, to a new-comer beautiful lady.
And if you are not in the rich or strong classes, the situation is not that different.
A traffic policeman can treat you according to your car’s number plate. If your car is Audi he would say: “Sir, could I see your driving license?” And if your car is Tofas he would say: “ Driving license?”. Or if a politician and a butcher are in a common place the buttons of the jacket would definitely buttoned up for the politician.
To summarize, yea all the people are equal but the richer, stronger, smarter and
more beautiful people are more equal.
161
APPENDIX 6 - Experimental Group Interview Questions (English translation of the original Turkish one.)
1) Can you summarize vocabulary activities we have carried out this semester?
2) What do you think of the ultimate aims of these activities?
3) Have these activities changed your perspectives about English vocabulary?
4) Do you think that concordancing activities have had any positive effects on your
recognition of the new words you have learned?
5) Do you think that concordancing activities have had any positive effects on your writings?
162
APPENDIX 7 – Academic Vocabulary Load Comparisons (Target vocabulary items are boldfaced) Random Essay Samples from the Experimental Group The essay written by E. K. before the treatment. Money is very important for all men. In old days, life of people changed with money. People depend on money for everything but money is the root of all evil so if the importance of money increases, evil increases. In those days, people apply a lot of bad ways for money. They dedicate themselves to money, so they can do every kinds of evil for money. The best of friendships, partnerships can finish because of money. It has people's minds so people don't do anything without money. Illegal works, unsuitable jobs started to be found in the world. For example; smuggling, kidnapping, robbery are done by people because of money. Money causes a lot of bad events. I think, money is a devilish power. It provides making many mistakes. It directs all evil and it affects people's status. Discrimination of status among people happens because of money. Money is the most important thing for all human. Therefore, it is the root of evil. People can kill each other for money because they don't want to live without money. To earn a lot of money, they accept the risk of losing their self-esteems. Money can remove from being a good person. We don't give very importance to money. We don't permit it to become the owner us. There is equality of the right side and the left side in mathematics. That means, there's not equality even in theoretical life. Therefore, as George Orwell says "All man are equal; but some are more equal than others." The essay written by E. K. two weeks after the treatment. Money is the most important thing for some people, so people can do everything for it. All the bad and good things occur because of money. However, money often leads to bad events. Namely, money is the root of all evil. Money is a source to survive, so people always need money. Most of the people believe that money brings happiness. Sometimes, this thought is true but these days, money is used for evil. For example; people are killing each other for the money. To earn a lot of money, they apply bad ways, for instance; smuggling, kidnapping, theft, etc. People can do everything for it. Obviously, money destroys humanity. Most of the people are losing their humanity. In addition, it will cause that money blind the eyes of humanity. Money is increasing illegal events. For example; historical smuggling began for it. People began to kidnap children to get ransom. In other words, money affects people's lives. How does it do it? For example; in our country, doctors give importance to money therefore poor people cannot go to a good hospital and cannot be examined. Namely, money can affect people's health. In addition, there are some jobs in this kind of event. For instance; poor quality construction materials are used by civil engineers and they do this work with less money. Pharmacists can sell a lot of different drugs to earn a lot of money. In our country, people give up their rights for money, and sell their votes. Money still continues to gain importance and bad events are increasing rapidly day by day because of money. If you continue these bad events, the world will become uninhabitable. People may not connect everything the money. People should understand that money leads us to all evil.
163
The essay written by E. E. before the treatment. Money is needed by people but much money is harmful for people's life. Money is root of all evil. I don't agree with this idea. Much money is the root of all evil. I agree with this idea. Because if people don't have money, they do anything. Money is very important for their life. To me money is good, much money is very harmful. Money is the most important all over the world for people. Because people need it for everything. Example, for study, healthy life etc...But much money is changed people or people's life. I want to give an example. I have a friend. I was thinking he is very good. He had got money for his life. One day he had got much money. After this day, he became very much. He didn't meet old friends. I was surprised him. But I couldn't do something. When people have got much money, they can do harmful thing. However much money is very good. People can do everything they want. But much money is root of all evil. Much money are harmful for people's life. It can be done harmful thing. I think people don't want much money. They should want enough money, happiness and a healthy life. The essay written by E. E. two weeks after the treatment. The ocean of the world occupies over 70 %25 of the earth's surface. Ocean is the most important thing all over the world because oceans look like blood of humans' body. Because they are very important things. However, people don't protect the oceans. Instead of protecting the ocean they destroy the oceans. Because of ocean pollution a lot of results happen.
Oceans have got a lot of fish. Fish live in the ocean and fish is very important for people. Fish is essential for people's life. Because fish includes B12, and it is a very important vitamin for people. But nowadays human activities destroy the oceans, because of fish deaths. Fish deaths affect people's life.
The oceans are very important things all over the world. Because they are harmless on human's life. However, people ultimately destroy the ocean. They don't know they destroy themselves. I think we must protect the ocean. We don't pollute the ocean. Because we must live healthy.
164
The essay written by F. K. before the treatment. Money is necessary for all people. Many people find money essential and couldn't manage without it. We use money in everyday life too much. Some people use money for good aims or bad aims. If a person has so much money, he can't control himself about how to spend his money.
All people work to earn money. In this world, if you don't have money, so you are hungry. But some people earn money much more than other people. This system is very bad. People are very greedy. Every time they want to have more and more. They earn money and they want to have more. People have two big eyes. In these eyes have big holes. That can never fill these holes. Every time they want to have everything. However, people don't think other people's life. Money is necessary I know but every people must be equal. I want to burn all money in the world. Maybe, evils lose over the world. People don't know how to use money; they usually use money in bad ways. These people, they think they are lord of the world. Money ambition is very dangerous for all people.
In conclusion, I believe that people are very selfish. They always want to have more. They can't think one day every people will die. I understand these people they have money and don't want to lose this money. But they must be careful about money ambition.
The essay written by F. K. two weeks after the treatment. In our life, money is important to make something. Nowadays, we use money our everyday life and we use it everywhere. Many people find them essential and couldn't manage without them. Lydia which civilization lived Anatolia in the past found money. When they found money this system has started to be evil. There are many impact money on our daily life. These are economics and manage system of world.
The impact money on our economic life is very important. We use money everywhere. For instance, make shopping, buy a car, buy a house, eat something sometimes go to school etc. If you don't have any money, you can't do these. Every people want to money to live easily and may be to live luxury. In our economic life, everything is money, if you have money, you can make everything. All people are greedy in our world. Every time they want to gain money more more more. This greedy will change the control of the world.
The impact money on manage system of world has been great. In this world, every governments want to be powerful. If governments want to be powerful, they should have a lot of money. Many governments begin the war some poor countries to gain money. So this powerful governments control our world. Nowadays petrol is very important and expensive. Powerful governments exploit to petrol to gain money to keep hands the power.
In conclusion, money impact our everyday life many ways, sometimes economics, sometimes politics. We know a thing "money is a evil ". every people want to gain money but shouldn't kill someone.
165
The essay written by İ. U. D. before the treatment. As the time passes social construction based on equality is unfortunately leaving its place to category equality. But, what is category based equality? Although people seem to be treated equally by the law, this equality sometimes changes due to appearance or/and financial, social and intelligence status. You can understand the fact that life is not fair; once you are not categorized as beautiful, handsome, rich or strong.
To give a simple example; if you are not clever enough to attract your teacher between your peers, you sit on the back desks of the classroom throughout your education life. Or vise versa, if you attract your teacher with your intelligence; certainly, you will be one step further than your peers. From another angle; if you are not pretty enough you can easily realize the waiter's interest, which you lack of although it is a café you frequently visit, to a new-comer beautiful lady.
And if you are not in the rich or strong classes, the situation is not that different. A traffic policeman can treat you according to your car's number plate. If your car is Audi he would say: "Sir, could I see your driving license?" And if your car is Tofas he would say: " Driving license?". Or if a politician and a butcher are in a common place the buttons of the jacket would definitely buttoned up for the politician.
To summarize, yea all the people are equal but the richer, stronger, smarter and more beautiful people are more equal.
The essay written by İ. U. D. two weeks after the treatment. Successful universities give theoretical lesson for a short time. They believe that they had experienced. When students learn something, they sometimes make mistake, but experienced is learned real life. Scientists say "Experienced students are more successful than the students who take theoretical education. Some universities are aware of this method. They change education system. Most university degrees are theoretical and do not prepare students for the business life.
The best education is fifty fifty theoretical and experience. For instance; you know the issue completely, but you don't apply your knowledge the business life so you get failed. Some universities reach agreement with industry so students' success increases and students have experience. When we contrast academic person with experienced person, academic person makes something quickly like solving problems. Some universities send students to industrial areas for lesson. Students learn real life. For instance; doctors' education is six years in total, but they learn theoretical lesson for five years. They attain experience one year at hospitals. This method results in success. In Germany, engineer education is six years. Theoretical lessons are given three years. They work three years in industrial areas, so they are ready for the real world. Theoretical lesson is good for academic person. When you keep an appointment, first of all, they ask you "Have you got experience?". There is an important distinction between vocational education and liberal arts education. Universities are not vocational schools. I think the coursework gave me a basis for understanding what I do; however, no class can completely prepare you for what you will encounter in the workplace.
In conclusion, experience has a positive effect from theoretical lesson. If you want to have a good job, you should have experience. Students spend lots of time studying theoretical things because of that they couldn't improve for themselves.
166
The essay written by M. A. D. before the treatment. Everyone has rights which are obtained from coming to the world. There rights are structure of living happily and independence. However, all the time there are people want more and more. They envy others and they never see themselves. In that point the people I mean, ruin balance or try to ruin.
It's certain that every one of us knows that inequality exists. George Orwell wrote 'Animal Farm' many decades ago. He was discussing about Russian management system. He found weak points on their big rules. Not only based on Russia, also a lot of countries live problems about strength of equality. I strongly believe that there isn't equality. It's normal. We should think about how we can close the difference of high and low. It can be only with understanding and conscious people. Now it's time of education. It's a simple thing to talk much about problems. Creating solution is vital thing. Furthermore, humans have the power of mind and they can do it best.
In early of our lives we can't choose our parents, our language, our country and our fate. Inequality begins here. But inequality I mean isn't same as inequality I explained. Subsequently, we should turn our sights to the education for tomorrow's sake. In the novel George Orwell says "All animals are equal; pigs are more equal. We just aim not to grow people who has characteristic of pigs."
The essay written by M. A. D. two weeks after the treatment. In prehistoric times, life was hard. There was no technology, no regular language. Communication, in this point had a major part of rule for living. They improved by passing time. They planted, trained by themselves against dangers. When these people brought their land to a good point, the conflicting started. People used to exchange their productions with each other; however exchanging became insufficient. They needed a solution that will open every door: money. Although it's a powerful thing, it will change everything and nothing will be the same without it. Money was our technology, our politics, our religion, our solid language from that point on. Furthermore, this power has side effects.
Consuming makes people happy; however only happy people can consume. What's the root of happiness? If consuming is resulted by happiness, money indicates conditions. Money is earned by working people. Do you want more money? You can earn more money by working more? In this section, we can discuss this subject around ethics and human rights. Everybody works and earns money. In contrast there are a few people at the top of prism. These elite people force big crowds to confined spaces. The big part of prism is full of people who fight for live and live with discrimination, moreover in inequality, dependency etc. Money created a scale it has two places for weight. A few people steal the freedom of other sides'. There is a fact that they're getting stronger. To prevent this power, a new power is created called evil. As you see, there is a circulation and no one gains. We can see money's harms in many aspects. For instance, people kill each other, some people die with starvation and children grow without love.
Money is a system that grows up but never ends its growing. It creates evil, constrains our freedom and also eats poor people and never full. I hope, people will understand there are more valuable things than money. I want to finish my essay with a Native American Indian quote, only when the last tree has died and the last river has been poisoned and the last fish has been caught will we realize we cannot eat money.
167
Random Essay Samples from the Control Group The essay written by E. K. before the treatment with the experimental group Who doesn't want to graduate from university? The education is very important for all of our life. Everybody who I ever know studies to join a good university which gives him/her good ability to survive in a big world.
There are lots of universities in our country. Each city of the country has it in early days. That is a good thing to our people but the university is not only buildings. It isn't like a high school to build anywhere in the country. Building universities anywhere is good but the system of education wouldn't be only theoretical. The teachers of lessons must be quality in their department. The laboratories must have equipments of the lessons. I want to give an example about this. Our university is famous for its engineering department. It can give the students good ability about their field. For example, my department is mechanical engineering. In the real world there are some difficulties waiting for me. My department's teachers know that. Because of that reason, they give internships and orientation to us in summer. When I go there, I will be learning welding, shearing, grinding, etc. A mechanical engineer must know how to shear, bash or cast metal. There are lots of machines in our laboratory which are same with the factories in the world.
The way passes through quality universities, and the aim is to survive in a big real world.
The essay written by E. K. two weeks after the treatment with the experimental group Do you ever know a university which has laboratories of each department in my country? Each city of my country has university, but do they have a good quality to prepare students for the real world?
When I imagine the university life I only thought that ''I will be far away my family and I can do what I want. I didn't think what the qualities of my university were. How it gives education. I heard something about universities where my friends were. My friends only have theoretical lessons. They will be engineers in the future. One of my best friend doesn't know how to weld and drill something. When I accessed the University of Gaziantep, I first heard that there is an internship at the summer of first class. We will be a mechanical engineer and we must know welding, screwing, drilling etc. Most universities don't have laboratories to give practical education. My university has lots of mechanical devices which like factories used. There is no difference between each other. This is the most important reason to select this university. Universities aren't like high schools. It doesn't have to be at each city. Governments don't have to build it in every city of our country. If they build, they must give the university good equipments, qualified personal. I thanked to my teachers because they help us every time. We will face with some problems at our future life about our work. We fall the problems off about our job before we graduate to school. This gives us an experience about problems and we can know how to cope with them.
Problems have never been stopped. But when we face them, we have to be ready to solve and manage them. Life is a long way to run and the university is the starting point of this long way.
168
The essay written by H. İ. G. before the treatment with the experimental group . Money is very important for our lives. We must earn and spend money to take care for ourselves, maybe our family. Nobody can take same salary and some people think this is right but if you ask me: If you earn too much money, this situation can bring you evil.
I can give some examples about this topic. For example, I am a fan of some singer, actor, actress. But when I watch the magazine programs or read a magazine about famous people's life, I see a lot of bad news. For example, he/she uses cocaine, heroin and various drugs. And I think, they have money, so they can buy anything they want and they buy drugs because they want to feel some different emotions. It can be reasonable for them. But it is a giant mistake. Another example about tough guys: We can meet them on news. You can find them especially in poor neighborhoods. The tough guys are richer than other people who live in the same neighborhood. And he wants poor people to obey him, do whatever they want, pick money. In addition, he can punish them when they don't obey his rules. But nobody should be powerful because of their money.
All in all, money is the roof of all evil. If you don't realize that, money can control your life. After when you look at your life, it may be a chaos. Don't be slave to your money.
The essay written by H. İ. G. two weeks after the treatment. We have to do something for live. Eating, drinking, sleeping and earning money are some of them. Money is more different than others. Because it can change your life, may be your personality. If you do not know have you can use the money, it brings you pure evil and you going to be slave of cash.
I can give two example about this situation. First example; if you have so much money, you can buy what you want like expensive holidays and luxurious cars. But you enjoy from this life until a grade. You will want to do something unusual. You can watch them on the TV. Especially, famous people act weird after to become rich. They are using some drugs, to be addicted plays (jackpot, poker, blackjack etc.) or alcohol, cheating their boy / girl friend / wife / husband. These are common things in our lives. Famous people do not be alone about this. Regular-rich people can do all this stuff. If you do not abandon from this harmful habits, you will expose to money's evil face.
Second example is about my personal life. I had a friend in the high school. He was handsome, smart and has so much money. He was a good guy but he has a major issue. The issue is he use up the girls. He impressed the girls with his specialty. I and my friends see that problem and we decide to talk him. We warn him and said: "This is a giant trouble. You cannot use the girls up with your money. You can destroy your future." And he said: "I have money and power. I can do what I want. It is not your business." These words are enough for us and we banished him our friend group. After this case, he was alone and he lost all his friends.
You must be careful about the money, because it can exploit your life. If you aren't regardful, you can lose everything.
169
The essay written by H. K. before the treatment. System of the most university is not practical; it means the system does not prepare students for the real world. Especially, lesson should be practical for engineering faculty. Some lessons are not necessary for this faculty; for example chemistry. On the other hand, theoretical knowledge is required to an extent, but experiment is more important than some knowledge. Preparing a project can be a lesson in engineering faculty. And also it should be an important lesson. Manufacturing is a useful thing and it should be a purpose for a student. It is only possible at practical lesson. If a faculty is four years, practical lessons should have three years of the four years. When a student makes a new machine and something useful for people, press researches student's university, as a result student's university will known all over the world. In conclusion, experiment should be gained at university for students. When an inexperienced student starts working, it takes a lot of his/her years. University must show different ways for student's projects. If a university wants to be known all over the world, it must educate quality students. The essay written by H. K. two weeks after the treatment. People make some major faults in their lives. There are a lot of causes for them. But causes don't let them make criminals. For instance, if you ask a murder "Why did you kill the person?", the murderer can response in many different ways. And his or her opinion, he%5Cshe is right, but if he or she does not regret, they should punish him/her. But the purpose must be to rehabilitate them. There are different ways to gain them. The prison system has the most important task. If states use correct system, rehabilitation will be easier. In fact in my opinion, before a crime occurs, precautions should be researched and evaluated.
Correct prison system is not just punishing. Punishment is only one way. Some criminals are major for example killing a person, but some of them are minor faults. Criminals should be assessed in different categories. If you punish all people, after years all people make the same crime. States should not forget that murders or thieves are human. And they can rehabilitate them in correct ways.
One other hand, experts know some people regret. Managers of prisons can forgive people who regret their faults but of course, forgiving is only for minor criminals. System of the prison can be suitable for criminals. For example murderers should read beneficial books especially about religion, and must learn the fault that is done is too big a crime. Most of the experts decided that teaching religion is very effective on murders or thieves. Eventually, each person can make a mistake, but the thing that is important is to regret.
170
The essay written by H. E. before the treatment. In our world, there are lots of different people and we call some people bad. How we can decide it? Maybe that person is very intelligent and he needs only education. We cannot know this completely but our punishment system is not checking this. It is only giving most of times prison punishment to the guilty people. In my opinion, prison punishment is out of date and we need to make some changes about this.
What is the prison? It is only four walls and you are isolated. Maybe sometimes you can see the garden. I do not know what they are doing there all days. What can they do? Nothing, because they are criminals. But every people can make some mistakes. Maybe it was very small mistake but after entering a prison they are not adding some skills to themselves and they are becoming very bad people. I think it is not the right way for punishment. Because if you do not educate them, they will do same mistakes again. They need rehabilitations. They do not need dungeons like old time punishments. We have to protect the other people from them. It is correct but the prison is not needed.
To sum up, every person can make mistakes but if we do not educate and rehabilitate them they will not be better and I think we do not need prisons, we need rehabilitation centers now. If we leave them alone, they will be bad people every time.
The essay written by H. E. two weeks after the treatment. University is the last gate to have a professional degree in our life. When we attend a university it means after this long term course we will have a job. On the other hand all before educational courses prepare us for a university. Because of this reasons, university is a very important effective part in our life. So our expectation from a university is very much from practical and academically aspects. But most of universities are giving too much theoretical courses and then they do not give us enough time to practices.
In real life, if we have only and only a theoretical knowledge, there is no way to become an engineer for me. We need to see how the machines work or how we can fix a problem on real life. Maybe the directors can think if we have lots of academically knowledge it would be better, but if we know a lot of thing then you need on a process, we would be confused. In my opinion, we can learn how much we need and then we have to make a lot of practices. The experience is a very important thing for our life , so we have to learn lots of things from seniors. If we do not have much more practice courses, we will not be able to adopt the real life and it means very much lost time.
In conclusion, if we do not want to lose too much time for adaptation to the real life, we have to have much more practical courses in our universities then now. Most of university degrees are academically and theoretical so it does not prepare us for the real people life. We have to make something for this big problem immediately.
171
The essay written by M. G. before the treatment. Money is the root of all evil but inside that money is very useful to do favors. Initially Money may change the people. It may make a bad guy a person. Because it is very desirable for almost all people. It makes an important person a person. Nowadays everything depends on the Money. If you moneyless ; you can be homeless, hungry, ignorant etc%85. The point in here the money is the root of all evil in addition that being broke is also root of all evil.
Firstly a broke guy may do all kinds of evil may be expected everything, because he is helpless. His child can be sick and he needs money, his family can be hungry, his son can be in the university and he needs money. In these conditions a person can do everything to provide his and his family's main needs. He can be a thief, he can be murderer and he can kill people for the money. In summary he can be very harmful for another people. On the other hand; a rich person wants richer than he is, because the money makes people's eyes blind. People can also do everything to be richer. Because people's needs are infinite. Especially nowadays that is in the technology age, the people needs increase and increase. This condition needs more money and they make more poor people.
In summary both of them, money and being broke are the roots of all evil. They can change the people. They can be killers, thieves anything else. Therefore if we can do, we should help poor people provide their main needs, if we don't want them to be bad guys.
The essay written by M. G. two weeks after the treatment. Money is the root of all evil. Is this true? Money causes so much evil. A man can do so many things for money. It can be legal or illegal. In China, so many people work very hard 14 hours a day, but they earn little money. Because of the fact that they are poor people, they have to work. But this money is not enough for their needs. Therefore they can apply illegal method. They can be murderer, thieves anything else. Beside that they can start gambling for money. Generally those men lose everything that belongs to them, their car, house, money. Those men have a family, their wife, and children. May be they are sick. But because of the fact that they lose everything they cannot meet these needs, and they can get divorced, also they will be unhappy. These men apply everything for these results don't happen.
On the other hand, money is the root of all favors, because you cannot do good things if you are moneyless. You cannot help poor people, helpless, old people these favor need a lot of money. If you are a rich person, you can help poor people. In our days everything needs money. For instance; science researches, festivals... etc. If you are a very rich man; you can employ people. You can meet their needs, you can help them, and you can feed hungry people. In addition, doing these, you would have contributed to your country, because you decrease unemployment. Due to these reason money is also the root of all favors.
In conclusion, both money is the root of all favors, and money is the root of all evil. It may be evil and favor. But we should always prefer favor. Presumably we are poor people, we are helpless people. We cannot help ourselves and our family. We cannot meet their needs. Then we have two options. Either we will do bad things such as killing somebody for money or being thief, or another good man will help us and we will not apply these bad things.
172
CURRICULUM VITAE
KİŞİSEL BİLGİLER
Adı / Soyadı : İhsan ÜNALDI
Doğum yeri : Ankara
Doğum tarihi : 20/06/1975
EĞİTİM DURUMU
Doktora :Çukurova Üniversitesi İngiliz Dili Eğitimi (2011)
A Comparative Investigation of Lexical Networks of Turkish
Learners of English as a Foreign Language: A Corpus Based
Study
Yüksek Lisans :Gaziantep Üniversitesi İngiliz Dili Eğitimi (2003)
Faculty Attitudes towards Instructional Use of Computers at
University of Gaziantep
Lisans :Çukurova Üniversitesi İngilizce Öğretmenliği (1997)
KİŞİSEL YETERLİLİKLER
Bildiği Yabancı Diller:
İngilizce : ÜDS (100)
Fransızca : KPDS (74)
Diğer Yeterlilikler:
İleri seviyede bilgisayar okur-yazarlığı, optik sınav değerlendirme sistemleri
Çalıştığı Kurumlar / Görevleri
2009- : Gaziantep Üniversitesi,
Yabancı Diller Yüksekokulu - Okutman
2007-2009 :Yabancı Diller Yüksekokulu
Müdür Yardımcılığı
2005-2007 :Yabancı Diller Yüksekokulu
173
Bilgisayar Destekli Dil Eğitimi Koordinatörlüğü
2001-2002 :Gaziantep Üniversitesi, Yabancı Diller Bölümü –
Ölçme Değerlendirme Birimi
2000-2001 :Gaziantep Üniversitesi, Yabancı Diller Bölümü
Okutman
1997-2000 :Milli Eğitim Bakanlığı
İngilizce Öğretmenliği
YAYINLAR, SUNUMLAR VE KATILIMLAR
Yayınlar
Kitap Bölümleri
Ünaldı, İ. (2011). Birleştirilmiş Sınıflarda Öğretim Bağlamında Çocuklarda Yabancı
Dil Öğrenimi. In T. Dilci (Ed.), Teoriden Pratiğe Birleştirilmiş Sınıf Uygulamaları
(257-270). İstanbul: İdeal Kültür Yayıncılık.
Dergi Yayınları
Kırkgöz, Y., Ünaldı, İ. (basımda). Coh-metrix: introduction and validation of an online
tool for text analysis. Cukurova University Faculty of Education Journal.
Tırfarlıoğlu, F., Ünaldı, İ. (2006). Faculty attitudes towards computer assisted
instruction at the University of Gaziantep. Journal of Language and Linguistic
Studies, 2 (1), 43-55.
Ünaldı, İ., Kırkgöz, Y. (2011). Latent semantic analysis: An analytical tool for second
language writing assessment. Mustafa Kemal Üniversitesi Sosyal Bilimler
Enstitüsü Dergisi, 8 (16), 487-498.
Sunumlar
Bağçeci, B., Cinkara, E., Ünaldı, İ. (2008). Hazırlık Eğitimi Alan İkinci Öğretim
Öğrencilerin Karşılaştıkları Problemler. Yabancı Dil Bölümleri Ve Yüksekokullarının
Yabancı Dil Öğretimindeki Sorunları. Muğla Üniversitesi.
Kırkgöz, Y., Ünaldı, İ. (2011). İngilizceyi Yabancı Dil Olarak Öğrenen Türk
Öğrencilerin Sözcük Ağlarının Karşılaştırmalı Araştırması. XXV. Ulusal Dilbilim
Kurultayı. Çukurova Üniversitesi Yay., s. 90-91
174
Katılımlar
Yurtiçi Katılımlar
2004 - Akademik Bilişim, Trabzon
2005 - Akademik Bilişim, Gaziantep
Yurtdışı Katılımlar
2010 - Erasmus Eğitim Alma Hareketliliği - Silesian University of Technology
(Foreign Language Teaching Centre), Gliwice/Polonya