ThesisPDF Available

Age effects in a minimal input setting on the acquisition of English morpho-syntactic and semantic properties by L1 speakers of Arabic

Authors:

Abstract

This thesis reports a study of the effects on long-term proficiency of starting to learn English in a minimal input setting (the classroom) at different ages. A total of 132 Saudi Arabian college students participated in the study, 50 of whom started learning English in elementary school (3-11 years) and 82 in middle school (12-13 years), along with a control group of 11 native speakers. Previous research has generally found no significant starting age effect on long-term proficiency in minimal input settings (although Larson-Hall (2008) is an exception). The focus of the present study was on five linguistic properties that differ between English and Saudi Arabic: vP ellipsis, the use of ‘resumptive’ pronouns, adverb placement, the contrast in meaning between progressive/habitual forms, and the contrast in meaning between preterite/present perfect forms. Data relating to knowledge of these properties and relevant individual factors were collected through an Acceptability Judgment Task, Aspectual Interpretation Task, Gap-Filling Task, Cloze test, and background questionnaire. Results showed no statistically significant effect of starting age on second language (L2) performance, but clear effects of property type. Across varying ages of L2 learning, vP ellipsis, the use of resumptives and the interpretation of the preterite/present perfect contrast emerge as the most problematic, whereas adverb placement and the interpretation of the progressive/habitual contrast appear as the least problematic for L2 learners. The implications of these findings for claims about a possible ‘critical period’ for language acquisition, for the causes of persistent L2 speaker divergence from native speakers, and for the best age to introduce foreign language learning in a classroom setting are discussed.
Age Effects in a Minimal Input Setting on the Acquisition of
English Morpho-Syntactic and Semantic Properties by
L1 Speakers of Arabic
Kholoud A. Al-Thubaiti
A thesis submitted for the degree of Doctor of Philosophy in Applied Linguistics
Department of Language and Linguistics
University of Essex
May 2010
ii
© 2010 by Kholoud A. Al-Thubaiti
All rights reserved.
iii
To my parents,
Abdullah Al-Thubaiti & Khadija Al-Qurashi
iv
Acknowledgements
This thesis, certainly, would not have been possible without the contribution of so many people.
My deepest gratitude goes first to my supervisor, Prof. Roger Hawkins, for his enduring support,
guidance, encouragement, and especially his confidence in my work. In spite of all his
commitments and busy schedule, he always provided meticulous and timely feedback on drafts
of the thesis, papers, conference abstracts, applications, and even e-mail communications. His
critical comments and valuable discussions have made a great impact on the growth of my ideas,
and indeed inspired research undertaken in this thesis. It was a pleasure to be a student of Roger
Hawkins, not just for his influential spirit in the field of second language acquisition, but also for
his characteristic of humbleness and openness. No word is enough to say ‘Thank You’.
I am also greatly indebted to Sheikh Al-Nahayan Fellowship Award granted by The International
Research Foundation for English Language Education (TIRF). Special thanks extend to TIRF
Board of Trustees for choosing me for this award.
I am greatly thankful to the examiners of the thesis: Prof. Bob Borsley and Prof. María del Pilar
García Mayo for agreeing to take part of this work, and for the time they spent in reading and
commenting on the thesis. Their comments and discussions have indeed refined the quality of
this work, and helped a lot in exploring my ideas for future research.
I also wish to express my appreciation to members of my supervisory board, Prof. Andrew
Radford (Chair) and Dr. Karen Roehr (Advisor) for all their effort, support, and guidance
throughout my PH.D in Essex. My sincere thanks extend to the linguistics staff at Essex (in
alphabetical order): Prof. Andrew Radford, Prof. Bob Borsley, Dr. Enam Al-Wer, Dr. Ingrid
Leung, Dr. Karen Roehr, Prof. Louisa Sadler, Prof. Martin Atkinson, Dr. Phil Scholfield, and
Acknowledgments
v
Prof. Roger Hawkins. All of whom, I sincerely thank for the valuable tuition I received in their
courses, discussion groups, or workshops. I owe special thanks to Dr. Phil Scholfield for being
so helpful in statistical consultations both in person and via e-mail. I also owe Prof. Louisa
Sadler special thanks for discussing with me her comments on my paper on vP ellipsis. Special
warm thanks go to Dr. Enam Al-Wer for all her moral support, sincere advice, and help with the
transliteration of the Arabic examples.
My sincere gratitude is also due to my home university, Umm Al-Qura in Saudi Arabia, for
granting me a scholarship to pursue my PH.D in the UK. Special thanks are due to the English
Language Department there for their incalculable support and belief in me. I wish to express my
genuine appreciation to Dr. Shadyah Ann Cole, Dr. Ali Abu-Reesh, and Ms. Dima Abduljabbar
for all their assistance and support during my field trip. This leads me to extend my great
appreciation to all the participants who took part in this study for their time and enthusiasm.
I am indeed indebted to all scholars who sent me list of references or their papers (in alphabetical
order): Prof. Abdelkader Fassi-Fehri, Prof. Alain Rouveret, Dr. Alison Gabriele, Prof. Ianthi-
Maria Tsimpli, Dr. Jason Rothman, Dr. Sarah Liszka, and Dr. Theresa Biberauer. I also extend
my sincere appreciation to the people I met and had insightful discussions with in the following
conferences, GASLA-10 (University of Illinois, 2009), EuroSLA-17 (University of Newcastle,
2007), Cambridge-Essex Workshop (Looking at Language Acquisition, 2009; 2010), CamLing
Postgraduate Conference (2007), Newcastle Postgraduate Conference in Theoretical and Applied
linguistics (2007). Special thanks are due to members of the SLAG at Essex (past and present)
for their discussions and valuable feedback on my presentations in the group.
I also wish to express my sincere gratitude to Adele Beck, Claire Batterham, Francis Eaves-
Walton, Heather Grainger, and Kaili Clackson, for generously sharing their comments and ideas
when designing the test instruments. Special thanks are due for Adele Beck, for her time and
Acknowledgments
vi
effort spent in recording the test material. I cannot forget thanking my fellows, Carol Jaensch and
Ghisseh Sarko, for the good company and friendship throughout our journey. Special thanks are
also due to Najla Al-Ghamdi for the sincere care she showed during the writing up period. My
thanks also go to Hanan Kabli for her friendship and moral support.
Last but indeed not least, I owe special warm thanks to my beloved parents for their infinite love
and care. Their influence on the development of my heart and mind is incalculable. I cannot
thank enough my father and first mentor, Prof. Abdullah Al-Thubaiti, whose impact on my
education is immense. I definitely owe him my accomplishments. No word is enough to say
thank you to my mother, Khadija Al-Qurashi, for her enduring love, care, and sincere interest in
my research. I owe many thanks as well to my lovely sisters, Samah, Wafa, Naseem, Saja, and
our brother, Nizar, for believing in me. I admit that my stay in the UK would not be a successful
experience without their immeasurable support, which kept me going and eased my pressure.
vii
Abstract
This thesis reports a study of the effects on long-term proficiency of starting to learn English in a
minimal input setting (the classroom) at different ages. A total of 132 Saudi Arabian college
students participated in the study, 50 of whom started learning English in elementary school (3-
11 years) and 82 in middle school (12-13 years), along with a control group of 11 native
speakers. Previous research has generally found no significant starting age effect on long-term
proficiency in minimal input settings (although Larson-Hall (2008) is an exception). The focus
of the present study was on five linguistic properties that differ between English and Saudi
Arabic: vP ellipsis, the use of ‘resumptive’ pronouns, adverb placement, the contrast in meaning
between progressive/habitual forms, and the contrast in meaning between preterite/present
perfect forms. Data relating to knowledge of these properties and relevant individual factors
were collected through an Acceptability Judgment Task, Aspectual Interpretation Task, Gap-
Filling Task, Cloze test, and background questionnaire.
Results showed no statistically significant effect of starting age on second language (L2)
performance, but clear effects of property type. Across varying ages of L2 learning, vP ellipsis,
the use of resumptives and the interpretation of the preterite/present perfect contrast emerge as
the most problematic, whereas adverb placement and the interpretation of the
progressive/habitual contrast appear as the least problematic for L2 learners.
The implications of these findings for claims about a possible ‘critical period’ for language
acquisition, for the causes of persistent L2 speaker divergence from native speakers, and for the
best age to introduce foreign language learning in a classroom setting are discussed.
viii
Table of Contents
Acknowledgements .................................................................................................................. iv
Abstract .................................................................................................................................. vii
List of Tables ......................................................................................................................... xiv
List of Figures ....................................................................................................................... xvii
Transliteration Symbols ...................................................................................................... xviii
Chapter 1 Introduction and theoretical background .............................................................. 1
1.1.
Introduction .................................................................................................................... 1
1.2.
Selective divergence in L2 grammars .............................................................................. 1
1.2.1.
A maturational-based account: No/Partial Access ..................................................... 3
1.2.1.1.
No access position ............................................................................................. 3
1.2.1.2.
Partial access position ........................................................................................ 4
1.2.2.
A non-maturational-based account: Full Access ........................................................ 6
1.2.2.1.
Insufficient input ............................................................................................... 7
1.2.2.2.
Output limitations .............................................................................................. 7
1.3.
Empirical evidence in the literature ................................................................................. 8
1.3.1.
Evidence for the maturational-based account ............................................................ 9
1.3.2.
Evidence for the non-maturational-based account ................................................... 19
1.4.
Age effects in a minimal input setting ........................................................................... 26
1.5.
Statement of the research problem ................................................................................ 31
1.6.
Thesis organisation ....................................................................................................... 33
Table of Contents
ix
Chapter 2 The Morpho-syntactic and semantic properties investigated in the present study
................................................................................................................................................. 34
2.1.
Introduction .................................................................................................................. 34
2.2.
The identity problem in English vP ellipsis ................................................................... 34
2.2.1.
Elision of thematic verb and copula be in partial identity conditions ...................... 37
2.2.2.
be and have stranding in partial identity conditions ................................................. 40
2.3.
The impermissibility of ‘resumptive’ pronouns in English ............................................ 42
2.3.1.
Subject and object complex wh-questions ............................................................... 43
2.3.2.
Null-operator structures .......................................................................................... 51
2.4.
The (im)-permissible conditions of English adverb placement ....................................... 53
2.5.
The aspectual contrast between V-s and be+V-ing ........................................................ 56
2.6.
The temporal contrast between V-ed and have+V-en .................................................... 58
2.7.
Conclusion .................................................................................................................... 61
Chapter 3 Empirical design ................................................................................................... 62
3.1.
Introduction .................................................................................................................. 62
3.2.
Study sample and context .............................................................................................. 62
3.3.
Study instruments ......................................................................................................... 64
3.3.1.
Background questionnaire ....................................................................................... 65
3.3.2.
Proficiency Cloze test ............................................................................................. 67
3.3.3.
Acceptability Judgment Task (AJT) ........................................................................ 68
3.3.3.1.
vP ellipsis ........................................................................................................ 70
3.3.3.2.
Resumptive pronouns ...................................................................................... 72
3.3.3.3.
Adverb placement ............................................................................................ 74
Table of Contents
x
3.3.3.4.
Test material validity ....................................................................................... 77
3.3.3.5.
Scoring accuracy judgments ............................................................................ 78
3.3.4.
Aspectual Interpretation Task (AIT) ....................................................................... 79
3.3.5.
Gap-Filling Task (GFT) .......................................................................................... 83
3.4.
Procedures of test administration .................................................................................. 86
3.5.
Procedures of data analysis ........................................................................................... 86
Chapter 4 L2 speakers’ knowledge of syntactic constraints on vP ellipsis in English ......... 88
4.1.
Introduction .................................................................................................................. 88
4.2.
vP ellipsis ..................................................................................................................... 88
4.2.1.
Research questions and predictions ......................................................................... 88
4.2.2.
Results .................................................................................................................... 91
4.2.2.1.
Elision of thematic verb and copula be ............................................................. 91
4.2.2.2.
Progressive and perfect auxiliary stranding .................................................... 100
4.2.3.
Discussion ............................................................................................................ 106
4.3.
Conclusion .................................................................................................................. 115
Chapter 5 Unlearning resumptive pronouns in L2 English ................................................ 116
5.1.
Introduction ................................................................................................................ 116
5.2.
Resumptive pronouns .................................................................................................. 116
5.2.1.
Research questions and predictions ....................................................................... 116
5.2.2.
Results .................................................................................................................. 118
5.2.2.1.
Subject and object wh-questions (Finite) ........................................................ 119
5.2.2.2.
Null-operator structures (Non-finite) .............................................................. 131
5.2.3.
Discussion ............................................................................................................ 137
Table of Contents
xi
5.3.
Conclusion .................................................................................................................. 144
Chapter 6 L2 speakers’ knowledge of (im)-permissible adverb placement in English ...... 145
6.1.
Introduction ................................................................................................................ 145
6.2.
Adverb placement ....................................................................................................... 145
6.2.1.
Research questions and predictions ....................................................................... 145
6.2.2.
Results .................................................................................................................. 148
6.2.2.1.
Adverb placement in transitive structures ....................................................... 148
6.2.2.2.
Adverb placement in intransitive structures .................................................... 156
6.2.3.
Discussion ............................................................................................................ 163
6.3.
Conclusion .................................................................................................................. 171
Chapter 7 L2 speakers’ knowledge of the English aspectual contrast: progressive/habitual
............................................................................................................................................... 172
7.1.
Introduction ................................................................................................................ 172
7.2.
Aspectual contrast: progressive vs. habitual ................................................................ 172
7.2.1.
Research questions and predictions ....................................................................... 172
7.2.2.
Results .................................................................................................................. 174
7.2.2.1.
Habitual and progressive contexts .................................................................. 174
7.2.2.2.
Progressive contexts by predicate type ........................................................... 182
7.2.2.3.
Habitual contexts by predicate type ................................................................ 188
7.2.3.
Discussion ............................................................................................................ 191
7.3.
Conclusion .................................................................................................................. 200
Chapter 8 L2 speakers’ knowledge of the English temporal contrast: preterite/perfect ... 201
8.1.
Introduction ................................................................................................................ 201
Table of Contents
xii
8.2.
Temporal contrast: present perfect vs. preterite ........................................................... 201
8.2.1.
Research questions and predictions ....................................................................... 201
8.2.2.
Results .................................................................................................................. 203
8.2.3.
Discussion ............................................................................................................ 217
8.3.
Conclusion .................................................................................................................. 224
Chapter 9 Evaluating the long-term benefits of early EFL study on L2 performance ...... 226
9.1.
Introduction ................................................................................................................ 226
9.2.
Age effects in a minimal input setting ......................................................................... 226
9.2.1.
Research questions ................................................................................................ 226
9.2.2.
A note on scoring .................................................................................................. 227
9.2.3.
Results .................................................................................................................. 228
9.2.3.1.
Starting age of EFL learning and general task performance ............................ 228
9.2.3.2.
Starting age of EFL learning and other confounding factors ........................... 234
9.2.3.3.
Effects of starting age of EFL learning on later error rates across properties .. 242
9.2.3.4.
The relative long-term benefits of starting EFL at elementary or middle school
245
9.2.4.
Discussion ............................................................................................................ 247
9.3.
Conclusion .................................................................................................................. 253
Chapter 10 Summary and implications of findings ............................................................ 254
10.1.
Introduction ................................................................................................................ 254
10.2.
Overall summary ........................................................................................................ 254
10.3.
Implications of findings .............................................................................................. 257
10.3.1.
Selective divergence in a minimal input setting .............................................. 257
Table of Contents
xiii
10.3.2.
Early EFL instruction..................................................................................... 261
10.4.
Future research ........................................................................................................... 263
Appendix A Acceptability Judgment Task ........................................................................ 264
Appendix B Aspectual Interpretation Task....................................................................... 275
Appendix C Gap-Filling Task ............................................................................................ 292
Appendix D Cloze Test ....................................................................................................... 294
Appendix E Background Questionnaire ............................................................................ 296
References ............................................................................................................................. 301
xiv
List of Tables
Table 3.1 Experiential Characteristics of the Child and Teen Starters’ Groups (N=132) ........... 63
Table 3.2 Hours of English Instruction in Public and Private Schools ...................................... 66
Table 3.3 Groups’ Scores on the Cloze Test ............................................................................. 68
Table 3.4 Accuracy Scale Coding Key .................................................................................... 78
Table 3.5 Predicates used Across Progressive and Habitual Story Contexts .............................. 81
Table 3.6 Distribution of Verbs by Predicate Type in Tense Contexts ...................................... 84
Table 4.1 Mean Accuracy on Thematic Verbs and Copula be: Between-groups Comparisons .. 92
Table 4.2 Mean Accuracy on Thematic Verbs and Copula be by Proficiency: Within-groups
Comparisons ............................................................................................................................ 95
Table 4.3 Mean Accuracy on Thematic Verb Elision in Partial Identity Conditions .................. 99
Table 4.4 Mean Accuracy on Two Types of Auxiliary Stranding: Between-groups Comparisons
............................................................................................................................................... 100
Table 4.5 Mean Accuracy on Auxiliary Stranding by Proficiency: Within-groups Comparisons
............................................................................................................................................... 104
Table 5.1 Mean Accuracy on Subject wh-questions: Between-groups Comparisons ............... 119
Table 5.2 Mean Accuracy on Object wh-questions: Between-groups Comparisons ................ 121
Table 5.3 Mean Accuracy on (Un)grammatical Subject/Object wh-questions (-that) by
Proficiency: Within-groups Comparisons ............................................................................... 124
Table 5.4 Mean Accuracy on (Un)grammatical Subject/Object wh-questions (+that) by
Proficiency: Within-groups Comparisons ............................................................................... 126
Table 5.5 Mean Accuracy on Resumptive Pronouns by Proficiency: Within-groups Comparisons
............................................................................................................................................... 128
List of Tables
xv
Table 5.6 Mean Accuracy on Null-Operator Structures: Between-groups Comparisons .......... 131
Table 5.7 Mean Accuracy on Null-Operator Structures by Proficiency: Within-groups
Comparisons .......................................................................................................................... 134
Table 5.8 Mean Accuracy on Resumptives in (Non-)Finite Clauses by Advanced Groups ...... 136
Table 6.1 Mean Accuracy on SAVO and *SVAO by Adverb Type: Between-groups
Comparisons .......................................................................................................................... 148
Table 6.2 Mean Accuracy on SAVO and *SVAO by Proficiency: Within-groups Comparisons
............................................................................................................................................... 151
Table 6.3 Mean Accuracy on SAVO and *SVAO by Verb Form: Advanced Groups ............. 154
Table 6.4 Mean Accuracy on SAVPP and SVAPP by Adverb Type: Between-groups
Comparisons .......................................................................................................................... 158
Table 6.5 Mean Accuracy on *SVAO and SVAPP by Proficiency: Paired t-test Comparisons 161
Table 7.1 Accuracy Mean (in Percentage) for Sentences with be+V-ing and V-s Forms in
Progressive and Habitual Contexts: Between-groups Comparisons ......................................... 175
Table 7.2 Accuracy Mean (in Percentage) for Sentences with be+V-ing and V-s Forms in
Progressive and Habitual Contexts by Proficiency .................................................................. 178
Table 7.3 Mean Percentage of Error Rates for Incompatible Verb Forms in Progressive Contexts
by Predicate Type ................................................................................................................... 183
Table 7.4 Mean Percentage of Error Rates for Incompatible Verb Forms in Progressive Contexts
by Predicate Type across Proficiency...................................................................................... 185
Table 7.5 Accuracy Mean (in Percentage) for Sentences with be+V-ing and V-s Forms in
Progressive Contexts by Predicate Type across Proficiency groups ......................................... 186
Table 7.6 Mean Percentage of Error Rates for Incompatible Verb Forms in Habitual Contexts by
Predicate Type ........................................................................................................................ 188
List of Tables
xvi
Table 7.7 Mean Percentage of Error Rates for Incompatible Verb Forms in Habitual Contexts by
Predicate Type across Proficiency .......................................................................................... 189
Table 7.8 Accuracy Mean (in Percentage) for Sentences with be+V-ing and V-s Forms in
Habitual Contexts by Predicate Type across Proficiency ......................................................... 190
Table 8.1 Mean Suppliance (in Percentage) of Different Tense Forms in Obligatory Preterite
Contexts ................................................................................................................................. 203
Table 8.2 Mean Suppliance (in Percentage) of Different Tense Forms in Obligatory Present
Perfect Contexts ..................................................................................................................... 205
Table 8.3 Mean Suppliance (in Percentage) of Three Tense Forms in Obligatory Contexts .... 207
Table 8.4 Mean Suppliance (in Percentage) of Three Tense Forms in Obligatory Contexts by
Proficiency ............................................................................................................................. 209
Table 8.5 Mean Suppliance (in Percentage) of Tense Forms in Present Perfect Contexts by
Lexical Aspect and Proficiency .............................................................................................. 213
Table 9.1 Pearson Correlation of Starting Age with Test Scores over Disaggregated Subgroups
of Early Starters ...................................................................................................................... 231
Table 9.2 Correlation Coefficients Among the Explanatory Variables (N=132) ...................... 236
Table 9.3 Correlation Coefficients Between the Explanatory Variables and the Test Scores
(N=132) .................................................................................................................................. 237
Table 9.4 Summary of Hierarchical Regression Analysis for Variables Predicting Saudi Arabic
EFL Learners’ Scores in an AJT (N=132) ............................................................................... 238
Table 9.5 Summary of Hierarchical Regression Analysis for Variables Predicting Saudi Arabic
EFL Learners’ Scores in an AIT (N=132) ............................................................................... 240
Table 9.6 Summary of Hierarchical Regression Analysis for Variables Predicting Saudi Arabic
EFL Learners’ Scores in a GFT (N=132) ................................................................................ 240
xvii
List of Figures
Figure 9.1 Scatter-plots Showing the Relationship Between Starting Age (N=132) and Scores on
Four Tests: (A) PCT (out of 40), (B) AJT (out of 136), (C) AIT (out of 24), and (D) GFT (out of
20). ......................................................................................................................................... 229
Figure 9.2 Scatter-plots for Scores on the Proficiency Cloze Test as a Function of Starting Age
for (A) Whole Sample (N=132), (B) 3-6yrs (n=26), (C) 3-7yrs (n=43), and (D) 3-9yrs (n=48).232
Figure 9.3 Percent of Incorrect Responses on Five Properties by Three Starting Age Groups . 244
Figure 9.4 Percent of Incorrect Responses on Five properties by Elementary and Middle School
Starters ................................................................................................................................... 246
xviii
Transliteration Symbols
Arabic
letter Symbol
Description Arabic
letter Symbol
Description
ب b voiced bilabial stop ط
ŧ
voiceless emphatic stop
ت t voiceless alveolar stop ظ
voiced (emphatic)
interdental stop
ث voiceless interdental
fricative ع ¿ voiced pharyngeal fricative
ج
voiced palatal affricate غ
©
voiced velar fricative
ح ° voiceless pharyngeal
fricative ف f voiceless labiodentals
fricative
خ
x
voiceless velar fricative ق
q
voiceless pharyngeal stop
د
d
voiced alveolar stop ك
k
voiceless uvular stop
ذ
voiced interdental fricative ل
l
alveolar laterals
ر
r
alveolar flap م
m
labial nasal
ز
z
voiced alveolar fricative ن
n
interdental nasal
س
s
voiceless alveolar fricative
h
voiceless glottal fricative
ش
ß
voiceless palatal fricative و
labial approximant
ص S voiceless emphatic
fricative ي y palatal approximant
ض D voiced emphatic stop ء
÷
voiceless glottal stop
In this guide, IPA symbols are mainly used except for the following emphatic sounds (S,
D, & Đ), and palatal approximant (y).
Long vowels are marked by double vowels (e.g., aa)
Chapter 1 Introduction and theoretical background
1.1. Introduction
The central goal of this thesis is to examine long-term age effects on the acquisition of English
morpho-syntactic and semantic properties by Saudi Arabic learners in a minimal input setting. In
a generative approach to second language (L2) acquisition, the thesis aims to examine the
interaction between the age factor and context of exposure. This chapter aims to situate the age
inquiry of the thesis within the broad context of generative L2 research. It presents the
assumptions and arguments for maturational and non-maturational-based accounts of selective
divergence in L2 grammars (1.2). Then, it overviews empirical evidence for both accounts from
research conducted in naturalistic settings (1.3), followed by classroom settings (1.4). Given all
the background, the statement of the research problem is then presented (1.5). The chapter
finishes by outlining the thesis organisation (1.6).
1.2. Selective divergence in L2 grammars
One of the major goals of generative theories of second language acquisition (SLA) has been to
determine the causes of ‘selective’ divergence in the development of interlanguage grammars.
Age at which an L2 learner is first exposed to the target language has been considered an
important interacting factor. The potential explanatory role of starting age is at the heart of an
on-going debate as to whether the effects of a putative critical period observed in first language
(L1) acquisition do extend to L2 acquisition. The critical period for L1 development is
hypothesised by Lenneberg (1967) to extend from early infancy up to puberty. If L1 speakers
have not been exposed to language stimuli during the hypothesised critical period, language
development is impaired if not impossible. While there is suggestive evidence for the existence
of a critical period in L1 development (for a review see Singleton & Ryan, 2004), the issue is yet
Chapter
1
:
Introduction and theoretical background
2
far from settled in L2 acquisition.
1
In generative L2 research, this issue concerns whether
assumed (universal and parameterised) properties of Universal Grammar (UG) are subject to
maturational constraints. In other words, whether innate linguistic knowledge becomes
unavailable in its entirety or partially ceases to be available for post-childhood L2 learners. A
related issue in this inquiry is L1 influence on interlanguage development. A wealth of studies
has linked divergence from the grammars of native speakers of the target language to persistent
transfer of L1 properties, which are incompatible with the L2 grammar (e.g., Hawkins et al.,
2008; Schwartz & Sprouse, 1994, 1996; Tsimpli & Dimitrakopoulou, 2007; White, 1985b,
1989). While there is almost a consensus on the effects of the L1 grammar
2
, the question remains
as to why L1 effects persist in some domains of the L2 grammar, but seem to fade away in other
domains (Hawkins, 2000). This is the phenomenon of persistent selective divergence regardless
of length of exposure to the target language. It is important to realise that L1 influence is just one
factor involved in producing divergence and not the main cause of it. The debate about the
causes of divergence in post-childhood L2 grammars is yet unresolved, especially the
phenomenon of selective divergence. This debate can be summarised around three proposed
interacting factors: (a) maturational constraints, (b) lack of triggering evidence in the L2 input,
and (c) output limitations. In a UG-based approach, these factors have been expressed in a
number of UG hypotheses (no access, partial access, and full access). They differ in their view
of whether UG is subject to maturational constraints, and how they envisage the interaction of
1
Central to the critical period debate is whether a clear cut-off point is attested in acquisitional data, or whether a
continuous deterioration in performance is observed with increasing age. The age inquiry in this thesis does not
intend to verify either trend, but rather looks for any advantages for those who started at a younger age compared to
older learners or substantial differences between the two groups.
2
It is important to point out that there are at least two positions that do not support the view of L1 influence in L2
development. First, the ‘No Transfer’ position that is associated with the Strong Continuity Hypothesis of UG
(Epstein, Flynn, & Martohardjono, 1996). Second, the ‘Partial Transfer’ position realised as the Minimal Trees
Hypothesis (Vainikka & Young-Scholten, 1996a, 1996b) and the Valueless Features Hypothesis (Eubank, 1993,
1996). These positions are not discussed in the thesis because L1 transfer is not the focus of inquiry. The thesis, in
fact, assumes a Full Transfer position.
Chapter
1
:
Introduction and theoretical background
3
UG with the L1 grammar and L2 input. They are described in the following sections ((1.2.1) and
(1.2.2)).
1.2.1. A maturational-based account: No/Partial Access
In this account, there is a critical period for UG availability in L2 development. UG knowledge
altogether or some components of it are subject to maturational constraints. If L2 learning took
place after puberty, ultimate competence is impaired. This has been expressed in the ‘No Access’
(1.2.1.1) and ‘Partial Access’ (1.2.1.2) positions to UG.
1.2.1.1. No access position
This position has often been linked with Bley-Vroman’s Fundamental Difference Hypothesis
(FDH) (1989, 1990).
3
Adopting Lenneberg’s (1967) view of a critical period for L1
development, Bley-Vroman argues that child L1 acquisition and adult L2 learning are
fundamentally different. In the FDH, child L1 acquisition depends on UG knowledge and
proceeds via domain specific learning procedures, whereas adult L2 learning relies on L1
knowledge and develops via general cognitive strategies that can be used in general ‘skill
learning’. In this position, UG in its entirety ceases to be available for post-childhood L2
learners. Therefore, a cognitive theory instead of a linguistic theory is required for L2
development (Clahsen & Muysken, 1986). This is a position that is not adopted in the inquiry
undertaken in this thesis because it is assumed that a linguistic theory is central to the
explanation of L2 development, and therefore that language acquisition involves domain-specific
knowledge.
3
For a very recent discussion on the FDH, see a special issue (31) of the journal Studies in Second Language
Acquisition (2009).
Chapter
1
:
Introduction and theoretical background
4
1.2.1.2. Partial access position
Unlike the ‘No Access’ position, this position holds the view that UG is partially affected by the
critical period. In particular, parametric properties not instantiated in the L1 grammar are only
accessible within some critical period, whereas universal properties remain accessible throughout
life. According to Eubank and Gregg (1999), parametric properties require specific linguistic
experience to be set, and it is only during the critical period “the human brain is capable of
adjusting to varying input” (p.91). On the other hand, universal properties require “some
minimal amount of linguistic stimuli” (p.91), and therefore, their accessibility is not time-
constrained. Throughout development in linguistic theory, there have been different incarnations
of this position from the No-Parameter Resetting Hypothesis (NPRH) (Tsimpli & Roussou,
1991), Failed Functional Features Hypothesis (FFFH) (Hawkins & Chan, 1997), and most
recently the Interpretability Hypothesis (IH) (Hawkins & Hattori, 2006; Tsimpli &
Dimitrakopoulou, 2007).
4
These hypotheses have been associated with the Representational
Deficit Hypothesis (RDH) (Hawkins, 2003, 2005). The central argument is that target-like
performance could over-represent the underlying competence, and ‘despite appearances’ in L2
performance, ‘features are missing’ in the interlanguage of post-childhood L2 learners.
5
Of particular relevance to the central question of the thesis is the partial accessibility claim, as
spelled out in the IH. In this hypothesis, Tsimpli and Dimitrakopoulou (2007) argue there is a
critical period for accessing uninterpretable features, but not interpretable, at the LF interface.
Interpretable features are semantically relevant, and thus, need to be visible at LF; on the other
4
In these hypotheses, parameterised properties assumed to be subject to maturational constraints are expressed in
light of the predominant linguistic theory of its time (parameters, functional categories, and recently uninterpretable
features).
5
The interface phenomena, whether grammar-internal or grammar-external, are now a current trend in investigation.
The general claim is that while grammar-internal interfaces (syntax-semantics) are less challenging, grammar-
external interfaces (syntax-discourse) prove more problematic (White, 2009). It is yet not clear for me whether the
interface hypothesis claims a maturational account for potential divergence in this domain. For extensive work on
this, see Sorace and colleagues (Belletti, Bennati, & Sorace, 2007; Sorace & Filiaci, 2006; Tsimpli & Sorace, 2006).
Chapter
1
:
Introduction and theoretical background
5
hand, uninterpretable features are semantically irrelevant, and therefore, need to be invisible at
LF to avoid derivation crash. In cases where uninterpretable features are grammaticalised, they
are spelled out at PF (e.g., resumptive pronouns in Greek and Arabic). The claim of the IH is that
properties associated with uninterpretable features not already activated in the L1 grammar will
pose a learning problem for post-childhood L2 learners because they are inaccessible beyond a
critical period. On the other hand, properties associated with interpretable features are acquirable
even if they are not part of their L1 grammar because they remain accessible throughout life.
Therefore, any linguistic properties associated with uninterpretable features, but not
interpretable, are prime candidates for fossilisation. The L1 grammar is assumed to have a
persistent effect on post-childhood L2 learners’ grammars. Success in acquiring linguistic
properties associated with uninterpretable features is ‘apparent’. With the view that L2 grammars
are UG-constrained, it is assumed that L2 learners will make use of available resources such as
the interpretable features (assumed accessible throughout life) to approximate target-like
performance, but not competence. In such cases, interpretable features compensate for the
unavailability of uninterpretable features. Interpretable features can be accessed through L1
grammar or directly from the UG inventory. In this respect, the IH argues for a representational
deficit in the interlanguage of post-childhood L2 learners. Uninterpretable features are missing
despite their ‘apparent’ effects in L2 performance.
Since the IH offers testable predictions for potential problematic domains for L2 learners, the
thesis aims to test these predictions across a set of morpho-syntactic and semantic properties in
the L2 grammars of child and teen starters exposed to English in a minimal input setting (as will
be described in section 1.5).
It is important to understand why only uninterpretable features are affected by the critical period.
As explained by Tsimpli and Dimitrakopoulou (2007, p.224), uninterpretable features are only
Chapter
1
:
Introduction and theoretical background
6
present in the language system to implement syntactic operations (such as Move), but
interpretable features are represented at the conceptual and linguistic levels in the mental
lexicon. Therefore, uninterpretable features become invisible at LF once their role is
accomplished in the narrow syntax to avoid derivation crash, whereas interpretable features
remain visible at LF to convey the semantic content. By being present at the LF interface, which
links linguistic and conceptual representations, interpretable features can be accessed either top-
down from the mental lexicon to the LF interface or bottom-up from language to cognition. In
contrast, uninterpretable features cease to be accessible once they have been deleted before
convergence at LF. Hawkins and Hattori (2006) offer another explanation that links the
differences to “functional usefulness” (p.271). While native speakers of a language can regularly
generate ‘new’ lexical items, it is normally unusual to generate ‘new’ rules in one’s grammar.
Based on this principle, there are advantages for interpretable features to remain accessible
throughout life, as they are required to generate “new open class lexical items” (p.271). On the
other hand, it could be disadvantageous to have uninterpretable features available throughout life
because grammars are rule-constrained, and these features are associated with functional
categories, which form a closed class of items.
1.2.2. A non-maturational-based account: Full Access
From an alternative perspective, the ‘Full Access’ position offers a non-maturational-based
account. It contends there is no critical period for accessing any part of UG. Divergence in post-
childhood L2 grammars is the effect of other factors, such as ‘insufficient input’ (1.2.2.1) or
‘output limitations’ (1.2.2.2). On this view, factors extrinsic to UG restrict accessibility to innate
linguistic knowledge.
Chapter
1
:
Introduction and theoretical background
7
1.2.2.1. Insufficient input
This is the view of the Full Transfer/Full Access (FT/FA) hypothesis as proposed by Schwartz
and Sprouse (1994, 1996), and originally expressed in the work of White (e.g., 1985a, 1985b,
1989). In the FT/FA hypothesis, the L1 final state constitutes the L2 initial state by which L2
learners analyse L2 input based on their L1 grammar. L2 development is potentially fully
successful given that L2 input provides unambiguous triggering evidence. Schwartz (1999)
argues that restructuring in L2 development is a failure-driven process. L2 grammar develops
gradually and conservatively depending on how informative and sufficient the evidence is in the
L2 input. The process of restructuring may involve (a) unlearning an L1 property that is not part
of L2 grammar, (b) assigning a ‘new’ property from scratch to accommodate L2 grammar, or (c)
resetting the value of an L1 property to be compatible with L2 grammar. Some cases of L2
learning might be very demanding, and thus, require more time and evidence to take place, and
vice versa. Among the most problematic cases of restructuring is unlearning an L1 property.
This, particularly, exemplifies a case of fossilisation because there is no evidence in the L2 input
that could force unlearning an L1 property that does not exist in L2 grammar. Even if negative
evidence is provided, it is argued as ineffective to cause grammar restructuring, and thus, cannot
engage UG (Schwartz, 1993; Schwartz & Sprouse, 1996). In short, although convergence on
target-like grammar is potentially possible, it is not guaranteed because it depends on how the L1
grammar interacts with properties of L2 input and UG (White, 1996).
1.2.2.2. Output limitations
To account for persistent variability’ in oral production, advocates of the FT/FA hypothesis
have proposed a number of hypotheses like the Missing Surface Inflection Hypothesis (MSIH)
(Prévost & White, 2000) and the Prosodic Transfer Hypothesis (PTH) (Goad & White, 2006). In
these hypotheses, variability is the effect of ‘output limitations’ rather than ‘representational
Chapter
1
:
Introduction and theoretical background
8
deficits’. Contra the RDH, it is argued that performance under-represents competence.
According to the MSIH, omission of inflectional morphology in oral production is a
computational problem; under communication pressure, the morphological/phonetic form fails to
be mapped to the relevant syntactic category.
6
Most recently, the PTH argues that morphological
variability is the effect of transferring L1 prosodic constraints. This transfer restricts generating
possible representations for L2 grammar (see also Goad, White, & Steele, 2003).
7
The claims of both the ‘Full Access’ and ‘Partial Access’ positions are considered in this thesis
in relation to results obtained from learners in a minimal input setting.
Having now presented the theoretical assumptions and arguments for maturationaland non-
maturational accounts of divergence in post-childhood L2 grammars, the following section
(1.3) presents empirical evidence from the literature supporting both accounts. The literature on
this topic is undoubtedly vast and cannot be covered in this thesis in detail.
8
Selective samples of
empirical studies related to the tested morpho-syntactic and semantic properties in the thesis are
presented from UG and Non-UG-based studies.
1.3. Empirical evidence in the literature
Empirical evidence for either account is gained from two groups of studies. The first group
examines the relationship between age of arrival and the L2 proficiency of immigrants who were
immersed in a naturalistic setting at different ages (referred to as ‘immigrant studies’ hereafter)
(e.g., Bialystok & Miller, 1999; Birdsong & Molis, 2001; DeKeyser, 2000; Johnson & Newport,
6
Lardiere adopts this position to explain variability in past tense marking in the oral production of Patty, a native
speaker of Mandarin/Hokkien with long immersion in English. For more discussion on this see Lardiere (2000,
2007).
7
The MSIH and PTH are not tested in any of the studies conducted in this thesis because the focus here is not to
examine ‘variability’ in oral production data with respect to inflectional morphology. Rather, the primary focus of
the thesis is to examine selective divergence across a range of properties and age groups.
8
For comprehensive overviews see (Birdsong, 2009; DeKeyser & Larson-Hall, 2005; Long, 2007; Singleton &
Ryan, 2004, among others)
Chapter
1
:
Introduction and theoretical background
9
1989, 1991; Patkowski, 1980). The primary aim of these studies is to confirm or disconfirm the
existence of a putative critical period. Proponents of a critical period seek to find qualitative
differences between a pre-critical and post-critical point of closure. Therefore, no age function
should emerge after a putative critical period (as will be discussed in (1.3.1)). If, otherwise, a
continuous deterioration in performance is observed across increasing age, that would falsify the
hypothesis of a putative critical period. It would rather suggest that deterioration is an effect of
age-related differences (length of exposure, attitudinal and motivational factors, etc). This is one
argument used by advocates of a ‘no critical period’ hypothesis (as will be discussed in (1.3.2)).
The second group of studies is UG-based. These studies investigate post-childhood L2 learners
of very-advanced proficiency levels (or near-natives) without comparing them with child L2
learners (or child starters). This research focuses on L2 knowledge of some subtle linguistic
properties in the ultimate grammars of post-childhood L2 learners (e.g., Hawkins et al., 2008;
Hawkins & Chan, 1997; Hawkins & Franceschina, 2004; Montrul & Slabakova, 2003; Tsimpli
& Dimitrakopoulou, 2007; White & Genesee, 1996). In the UG-based studies, research seems to
be motivated by Long’s (2005) interpretation of the critical period in which it is not assumed that
“all early starters will achieve native-like abilities, but that only early starters can, while late
starters cannot(p.310). Hence, proponents of the critical period seek to support the hypothesis
by finding ‘No’ exceptional adult L2 learners who can show native-like norms (as will be
illustrated in (1.3.1)). On the other hand, proponents of the opposite camp seek to disprove the
critical period, by providing evidence that there are exceptional near-native speakers who fully
converge on native-like norms (as will be illustrated in (1.3.2)).
1.3.1. Evidence for the maturational-based account
The findings of Johnson and Newport (1989, 1991) appear to provide evidence to support a
critical period for L2 acquisition, and have been among the most influential and often-cited in
Chapter
1
:
Introduction and theoretical background
10
the literature. In their first study (1989), they tested 46 Chinese and Korean immigrants on 12
morphological and syntactic rules of English. These rules cover a wide range of language-
specific properties (e.g., word order, determiners, movement constraints, number and tense
markings, and pronouns). The immigrants’ age of arrival in the USA ranged from 3 to 39 years,
with an average length of almost 10 years residence in the States. In a timed aural grammaticality
judgment test, they had 276 English sentences with 136 grammatical and 140 ungrammatical.
Results revealed a negative linear relationship between age of arrival and test performance (r=
-.77). This was further attested even after partialling out the effects of experiential and attitudinal
variables. A linear relation was also shown between age of arrival and each of the 12 rule types,
irrespective of the observed variation across the different structures. Results also proved
statistically that the scores of the Chinese and Koreans who were first immersed before the age
of 7 were not significantly distinguishable from the test scores of the natives’ group. On the other
hand, the scores of those who were immersed after the age of 7 were significantly lower from the
natives. This led Johnson and Newport to conclude that the age of 7 marks the terminus of the
critical period, and despite long exposure beyond this point, deterioration in ultimate
performance is inevitable. Of particular interest to the inquiry of this thesis is the variation
observed in the error rates across the 12 tested rules. The present progressive and word order
show the lowest error rates (10%), whereas determiners and plural markings on nouns achieved
the highest error rates (70%). In a linguistic account, it could be argued that because the
functional category (Determiner) and features of Number are not instantiated in Chinese and
Korean, determiners and plural markings are predictable candidates for divergence in their L2
grammars. However, Johnson and Newport do not support the view of an L1 effect on L2
performance. They rather argue that the pattern they obtained with the different rule types is
similar to the pattern of difficulty obtained in other L2 studies with different L1 groups. They
also argue that it is consistent with the pattern of difficulty shown by Genie, who was first
Chapter
1
:
Introduction and theoretical background
11
exposed to L1 after puberty. Overall, they assert that their results provide strong evidence for a
critical period in L2 acquisition across all rule types of language-specific properties.
In another related study, Johnson and Newport (1991) sought to find out whether UG properties
are also subject to a critical period in the same way language-specific properties are affected.
Therefore, they tested L2 knowledge of Subjacency constraints with a group of adult Chinese
speakers of English. Subjacency is a UG principle that governs how far a constituent can move
from its extraction site (described as ‘locality conditions’ in Minimalism). Contrary to English,
Chinese wh-questions are formed without movement, which means that Subjacency is not
activated in the Chinese grammar. Chinese speakers of English have no way to acquire
knowledge about such a property unless they have full access to UG, especially when
Subjacency is not part of their L1 grammar. Similar to their previous study (1989), Johnson and
Newport adopted length of residence as a criterion for selecting L2 speakers deemed to be near-
natives. They conducted two related studies employing the same tasks and procedures. They first
tested a group of 23 Chinese adult starters who had a minimum of 5 years of continued residence
in the States. Their age of arrival had to be after 17. Their age at first exposure via formal
classroom instruction was ignored because in their previous study (Johnson & Newport, 1989), it
did not significantly correlate with proficiency. In an aural grammaticality judgment task with
156 core items, they tested four structure types (grammatical declarative sentences and wh-
questions, and ungrammatical wh-questions with no subject-verb inversion and another set with
Subjacency violations). Results on Subjacency violations showed that the adult starters
performed significantly below the native control group. More specifically, they rejected only 22
out of 36 items with Subjacency violations, whereas the natives rejected almost all of them
(35/36). They consequently concluded that the adult Chinese speakers accept ‘many’ items with
Subjacency violations as grammatical. To verify their Subjacency measure, they gave a subset of
the participants a comprehension task. It contained paired sets of a declarative sentence as (The
Chapter
1
:
Introduction and theoretical background
12
policeman knew where the thief hid the jewels.), and a related question as in (What did the
policeman know where the thief hid?) (Johnson & Newport, 1991, p.241). For each declarative
sentence, the participants had to answer the corresponding question in light of the content of that
sentence. They assumed that if they responded correctly (jewels would be the correct answer to
the exemplified question), this would be an indication that the wh-word had moved from an
embedded clause as intended. If they gave a response in the form of “someone/something not
mentioned” that indicated the wh-word had moved from a different site other than the embedded
clause; perhaps from a permissible site that did not violate Subjacency. If they gave a ‘don’t
know’ response that would indicate that the wh-word had moved neither out of the matrix nor
embedded clause. They found that the adult Chinese participants parsed questions with
Subjacency violations to a large degree appropriately, and according to what was intended. This
suggests when they accepted ungrammatical wh-questions with Subjacency violations, they were
aware of the extraction site of the wh-word. In their follow-up study, they tested whether L2
knowledge of Subjacency could vary over different age groups of Chinese L1 speakers (4-7 yrs,
8-13 yrs, 14-16 yrs). They compared the performance of the adult Chinese speakers (from the
first study) with the performance of the three age groups, along with a control group of natives.
Results showed that the performance of the 4-7 age group was not significantly different from
the natives, whereas the performance of each of the other two age groups (8-13 and 14-16)
significantly lagged behind the natives. They also found that the adult Chinese group
significantly lagged behind the other three age groups. Thus, Johnson and Newport argue that
there is a gradual decline in performance with increasing age. Again, to verify their findings,
they also administered a comprehension task to a small sample of the three age groups. They
found that the Chinese speakers were parsing the statements with Subjacency violations as
intended. Based on results obtained from both studies, they concluded that similar to their
findings on the acquisition of language-specific properties (i.e., parametric knowledge), access to
Chapter
1
:
Introduction and theoretical background
13
Subjacency is also subject to a maturational decline. Hence, they assert that no part of UG
possesses an inherent immunity to maturational effects.
Moving now to UG-based studies, L2 acquisition of linguistic properties associated with
uninterpretable features has been the focus of recent research. Two recent studies that test the
assumptions of the IH as described in section (1.2.1.2.) are presented. Tsimpli and
Dimitrakopoulou (2007) present a case for the failure of L2 learners to unlearn the effects of
uninterpretable features (as with resumptive pronouns), and Hawkins et al. (2008) look at the L2
acquisition of the semantic effects of uninterpretable features (as in the aspectual contrast
between progressive and habitual interpretations).
To start with, Tsimpli and Dimitrakopoulou (2007) sought to examine whether proficient Greek
L1 speakers would be able to abandon using resumptive pronouns in English wh-questions.
Unlike English, Greek allows structures like, Who did you say that they insulted (*him) without a
reason? (p.220). Resumptive pronouns represent a case of an L1 property that is incompatible
with L2 grammar. In Greek, subject-verb agreement and object clitics are resumptively used in
the positions of the extracted subject and object, respectively. Their role is to spell out
uninterpretable agreement features. While subject-verb agreement is obligatorily required,
resumptive object clitics are optional. English, by contrast, exclusively employs the gap strategy
in subject and object extraction sites. Based on such parametric differences, they predicted that
proficient Greek speakers would persistently allow resumptive pronouns in the gap position of
extracted wh-words in English as would be possible in their L1. They conducted a study with 21
intermediate and 27 advanced Greek speakers. They tested them via a timed bimodal
acceptability judgment task. The test consisted of 30 (un)grammatical test items. The
grammatical constructions had a gap in subject and object extraction sites (e.g., Who do you
think that Jane likes ec?) (p.227). The ungrammatical constructions had a resumptive pronoun in
Chapter
1
:
Introduction and theoretical background
14
the subject and object extraction sites (e.g., Who do you think that Jane likes *him?) (p.227). The
ungrammatical constructions are grammatical in Greek, and thus, are designed to measure the
degree of L1 effects. To examine the compensatory role of interpretable features like ‘Animacy’
and ‘(D)iscourse-linking’, their test items included animate and inanimate subject/object
pronouns (he/him, she/her vs. it) and wh-words (who vs. what). The test items also had D-linked
wh-words (wh-NP as in Which student/book) and non-D-linked wh-words (who and what).
9
Consistent with the IH, resumptive pronouns showed a persistent L1 effect even at advanced
levels of development. Results revealed high acceptability ratings of resumptive pronouns where
a gap is expected. The rate of acceptance was affected by the extraction site with subject
pronouns accepted to a higher rate in wh-questions than object pronouns. Tsimpli and
Dimitrakopoulou (2007) argue that the observed pattern of acceptance is an effect of L1
grammar. Greek speakers mis-analyse subject pronouns as equivalent to subject-verbal
agreement markers as in their L1. Since subject-verbal agreement is obligatorily required in
Greek, subject resumptives turn out to be more problematic than object clitics, which are
optional in their L1.
10
Regarding the role of interpretable features (animacy and D-linking), there was a significant
interaction between these features and the acceptance rate of subject/object resumptive pronouns
by the advanced group. Similar effects were not observed in the performance of the English
control group. Even though Greek does not mark animacy on wh-elements and pronouns, the
Greek speakers successfully acquired the animacy distinction marked on English pronouns from
early stages. Tsimpli and Dimitrakopoulou (2007) use this as evidence for the possibility of
9
The notion of Discourse-linked elements was first introduced in the work of Pesetsky (1987).
10
A similar phenomenon has been also observed in the English L2 grammars of bilingual (Basque/Spanish) children
(García Mayo, Lázaro Ibarrola, & Liceras, 2005). This is based on cases where a subject pronoun is used alongside
another subject (e.g., The wolf he opened the door) (p.447). The doubled subject pronoun he is described as a
placeholder, and is analysed as an agreement marker. For a detailed analysis of this phenomenon, the reader is
referred to the cited reference.
Chapter
1
:
Introduction and theoretical background
15
acquiring interpretable features not already instantiated in the L1 as they are assumed accessible
throughout life. They also argue that interpretable features of animacy play a compensatory role
in the use of resumptives, by which they constrain optionality in L2 grammars. They found that
inanimate resumptive pronouns (it) are favoured over animate ones (he/she). In addition, they
argue that interpretable features of D-linking constrain optionality in the use of object
resumptives. The acceptability rate of object pronouns increases with D-linked wh-questions, as
would be possible in L1 Greek. Based on these effects, they argue that interpretable features are
helpful in approximating target-like performance, by constraining optionality in L2 advanced
grammars. The Greek speakers, nevertheless, have an underlying representation that diverges
from the native’s grammar. In sum, the study provides evidence that post-childhood L2 learners
at advanced proficiency levels cannot pre-empt an L1 effect associated with uninterpretable
features.
Turning to the domain of semantics, Hawkins et al. (2008) examined L2 knowledge of the
semantic consequences of verb-raising. They focused on the aspectual contrast between
progressive verbal forms be+V-ing and habitual V-s in English. They assume, following a
proposal by Déchaine and Manfredi (2000, November) that there is a parameter of variation
involving the interpretation of null/simple tense. Null tense is absence of any form of overt tense
marking in languages like Fongbe and Igbo; simple tense is illustrated by the simple present of
English and Italian. Where languages lack subject-verb agreement, as in the case of Fongbe and
Igbo, the inherent aspectual properties of the vP determine tense interpretation, in the absence of
any contextual determinant like a temporal adverb. For example, telic vPs like eat the bread
produce past tense interpretations. In languages that have subject-verb agreement, like English
and Italian, T has an interpretation that is independent of the aspectual properties of the vP. In
particular, the simple present tense has a generic/habitual interpretation whatever the vP.
Furthermore, where a language has a strong uninterpretable feature that forces main verbs to
Chapter
1
:
Introduction and theoretical background
16
raise to T, this produces a quantificational existential/progressive reading for T. Thus in Italian,
which has main verb raising, the simple present tense can be interpreted both as habitual and as
progressive. Because English main verbs do not raise to T, the simple present only has a
generic/habitual interpretation. Following Adger (2003), Hawkins et al. assume that Progressive
is an independent aspectual head with an interpretable [progressive] feature, and a strong
uninterpretable [uInfl] feature that forces the head to raise to T, which is then realised as a form
of be, and is only interpreted as existential/progressive.
Since the means by which aspectual interpretation is assigned vary cross-linguistically, Hawkins
et al. sought to examine how post-childhood L2 learners from typologically distinct L1s come to
acquire the aspectual contrast in English between simple V-s and progressive be+V-ing. They
compared three L1 groups of advanced proficiency in L2 English: Chinese, Japanese, and a
group of verb-raising languages including Arabic, French, German, and Spanish speakers. The
argument is that if the effects of uninterpretable features are acquirable beyond a critical period,
L2 learners at advanced levels of development, provided they have had sufficient input, should
perform in a target-like way, suggesting that they have target-like grammars, regardless of the
L1. If, however, the L1 groups perform differently, this would be consistent with transfer of L1
properties and failure to acquire new properties in the L2. The cross-linguistic differences among
the tested L1 groups can be summarised as follows: Although verb-raising languages instantiate
uninterpretable features as in English, they differ in the strength of these features. In verb-raising
languages, all verbs carry a strong uninterpretable feature that requires local valuation through
movement, and thus, raising verbs convey habitual and progressive interpretations. Conversely,
Chinese and Japanese differ from English in that they have no agreement morphology, and thus,
they lack verbal uninterpretable features. Aspectual interpretation is assigned by the use of an
adverbial modifier (zai in Chinese and -te in Japanese) that interacts with predicate type in a
systematic fashion. Based on these differences, the learning task for speakers of verb-raising
Chapter
1
:
Introduction and theoretical background
17
languages is to recognise that feature strength associated with uninterpretable features is
different from their L1. They need to realise that a strong uninterpretable feature is not a property
of thematic verbs as in their L1, but only a property of be (and also have). On the other hand, the
Chinese and Japanese speakers need to establish uninterpretable features on English verbs from
scratch because they are not present in their L1.
Testing knowledge of aspectual interpretation, Hawkins et al. designed a timed bimodal
acceptability judgment task. The L2 learners had to rate the appropriateness of a pair of well-
formed sentences as potential continuations for a given context as in this example (2008, p.340):
(1) Whenever Mary and Alan meet...
a. they talk about Linguistics until late -2 -1 0 +1 +2
b. they are talking about Linguistics until late -2 -1 0 +1 +2
The test had a balanced set of contexts for progressive and habitual readings, and a balanced
distribution of predicate type (activities, achievements and statives). Results showed that the
Chinese and Japanese speakers were less likely to accept a progressive reading for be+V-ing
when the predicate involved is of the achievement kind, whereas speakers of verb-raising
languages were less likely to reject a habitual interpretation for be+V-ing forms across predicate
types. For Hawkins et al., these results suggest that the Chinese and Japanese speakers under-
generalised the use of be+V-ing to convey a progressive reading when the verb involved was an
achievement. On the other hand, speakers of the verb-raising languages tended to allow be+V-
ing to have generic/habitual interpretation, regardless of predicate type.
Based on these results, they argue that the underlying representation of all three groups is non-
target-like, and that each manifests a persistent L1 effect. In the case of the Chinese and Japanese
speakers, they failed to establish the strong uninterpretable features on the Prog syntactic head,
Chapter
1
:
Introduction and theoretical background
18
which forces a progressive reading across all predicate types. This constitutes evidence for
failing to establish uninterpretable features not instantiated in their L1 as predicted by the IH. In
such cases, L2 learners resort to other UG resources through their L1 grammar to analyse L2
input. They, hence, assumed that the Chinese and Japanese groups were treating be+V-ing forms
as an aspectual modifier (zai in Chinese and -te in Japanese) which can only precede an activity
verb type as in their L1. In the case of speakers of verb-raising languages, since uninterpretable
features are available in their L1 (but with a different strength), their problem is different from
the Chinese and Japanese groups. They rather argue that Prog has not been established as an
independent syntactic category, and thus, speculate that be is possibly recognised as a light
raising verb “with the same interpretive consequences as thematic verb raising” (2008, p.348). In
fact, Hawkins et al. expected an alternative pattern of over-generalisation to occur with V-s
forms as would be possible in a verb-raising language. Overall, Hawkins et al. maintain that
semantic effects of uninterpretable features are another area of persistent difficulty for post-
childhood L2 learners.
To summarise, the immigrant studies’ as in the work of Johnson and Newport (1989, 1991)
provide strong evidence for a critical period in the L2 acquisition of language-specific properties
and universal properties such as Subjacency. Additionally, ‘UG-based studies’ in the domain of
morpho-syntax (Tsimpli & Dimitrakopoulou, 2007) and semantics (Hawkins et al., 2008)
provide evidence that post-childhood L2 learners at very advanced levels of development still
diverge in performance from native speakers. The source of divergence is associated with
uninterpretable features as hypothesised by the IH. In the next section, empirical studies that
argue against evidence in these studies are discussed.
Chapter
1
:
Introduction and theoretical background
19
1.3.2. Evidence for the non-maturational-based account
Opponents of a maturational account of language acquisition accept that age of onset has an
effect on general L2 proficiency, but deny that this is the result of the closure of a critical period
during which the resources of an innate language faculty (UG) are fully available. They interpret
differences in the outcome between younger and older L2 learners in terms of differences in
accessing input, differences in the way that affective factors are involved, and differences in
performance on language tests, but not in terms of biological constraints. From this alternative
perspective the findings of Johnson and Newport’s (1989) study have been subjected to intense
scrutiny. A number of studies have replicated the original study with different adaptations to
methodology and have argued that findings can be interpreted as the effect of age-related
differences and not the effect of a critical period (e.g., Bialystok & Hakuta, 1999; Bialystok &
Miller, 1999; Birdsong & Molis, 2001; Flege, Yeni-Komshian, & Liu, 1999).
Bialystok and Miller (1999), assume that if there is a critical period for language acquisition
there should be qualitative differences in outcome for learners who acquire a language within the
critical period and those who acquire a language beyond the critical period. The effects of L1,
structure type, and modality of testing should only emerge in the L2 performance of post-critical-
period groups. Younger learners with different L1 backgrounds should perform in a target-like
way across structures and modes of testing. With this view of the critical period, they examined
the L2 acquisition of English morpho-syntactic properties by Chinese and Spanish native-
speakers. As in Johnson and Newport (1989), the dividing age point for younger and older
learners was 15 years of age. The mean arrival age for the younger Chinese and Spanish groups
was 10 years, and for the older Chinese 23 years and Spanish 21 years. To test the extent to
which the younger or older learners perform in a target-like way, they also included a control
group of English native speakers. Their test methodology included a language questionnaire, L1
Chapter
1
:
Introduction and theoretical background
20
proficiency test, and a grammaticality judgment task in oral and written form. The
grammaticality judgment task focused on five structures (plurals, determiners, future tense,
present progressive, and collocation restriction). The test items were chosen to test linguistic
differences and similarities between Chinese, Spanish, and English. They were balanced in terms
of (un)grammatical pairs and word and syllable length.
Overall results yield different patterns of performance determined by the L1, Chinese or Spanish,
but few differences due to age of acquisition between the younger and the older learners. In
particular, results for both L1 groups reveal a continuity effect for age of arrival on performance
on the oral grammaticality judgment task. This would appear to be problematic for the critical
period hypothesis.
Nevertheless, examining the performance of four age subgroups (1-8 yrs, 9-15 yrs, 16-22 yrs and
23-33 yrs), results show that the Chinese and Spanish speakers up to 8 years of age perform in a
target-like way on the oral judgment task. After that point, only the Spanish groups show a
continuous decline with increasing age. Contrary to Bialystok and Miller’s (1999) claim, this is
suggestive of a closure point of a critical period because it marks a qualitative difference
between those who were exposed to English before age 8 and afterwards. This is similar to the
closure point at 7 years of age identified by Johnson and Newport (1989). However, Bialystok
and Miller do not accept such an interpretation and, in fact, warn against over-interpreting this
finding because the closure point of the critical period is generally hypothesised to be around
puberty.
Results also reveal effects for similarity between grammatical properties in the L1 and L2 in the
younger and older Chinese learners, but no effects for the Spanish groups across age of
acquisition. Since the effects were observed both with the younger and older Chinese learners,
Bialystok and Miller (1999) take this as evidence against a critical period. However, they seem
Chapter
1
:
Introduction and theoretical background
21
to emphasise the effects of similarity over difference, where the latter is the litmus test. The
acquisition of syntactic properties not present in one’s L1 grammar by older learners would be a
stronger argument against a critical period. Interestingly, their findings by property were
consistent with Johnson and Newport’s (1989) findings. The present progressive appeared the
easiest for both L1 groups, and plural marking proved difficult for both groups as well. Yet, they
dismiss an L1 effect account and do not even seem to consider the possibility of a universal
pattern of L2 development as in the case of Johnson and Newport. They rather explain this in
terms of the “inherent difficulty of the sentences” (1999, p.141) without going into detail. This is
a central issue in the age debate and requires a linguistic account. In their study, the cause of the
difficulty posed by the determiner and plural marking across age of onset of acquisition merits
closer examination.
As for the effects of test type, Bialystok and Miller (1999) report an effect for modality (oral
versus written) on accuracy and speed of judgments for both the Chinese and Spanish groups
across age of acquisition. The oral form of the grammaticality judgment task proved more
difficult than the written form. This is unexpected under a biological account for the younger
learners, provided the native speakers show no difference in accuracy in both testing modes.
Taken together, all three types of evidence (age of arrival, L1 effect, and test modality) appear to
refute the hypothesis of a critical period. Bialystok and Miller, nevertheless, accept the general
trend in associating success with acquisition at a younger age. For them, this could be the effect
of other factors, such as the educational experience of language learning in Canada as young
children.
Moving now to the UG-based studies, White and Genesee (1996) is a study set up to challenge
Johnson and Newport’s (1991) findings about Subjacency. They designed a study to test near-
native speakers’ knowledge of Subjacency constraints in English. They were sceptical about the
Chapter
1
:
Introduction and theoretical background
22
criteria adopted in existing age studies for the selection of ‘end-state’ L2 speakers such as length
of residence and amount of language use. Therefore, they devised a stringent set of criteria to
identify very proficient L2 learners who are at the end-state. By interviewing their participants,
they elicited spontaneous samples of English using a Thematic Apperception Test. Two English
native-speaker judges rated randomly selected portions of the data from natives and non-natives
according to an 18-point rating scale ranging from non-native to native-like. They evaluated their
nativeness over a wide range of linguistic ability, including vocabulary, phonology, syntax,
morphology, pronunciation, and even fluency. This resulted in two groups with 45 near-natives
and 44 non-natives. They had different L1 backgrounds, with the majority of them native-
speakers of French (58 French out of a total 89). The L1 involved has consequences for the
interpretation of the findings, as will be discussed. According to age at first significant exposure
to English, their participants were divided into four age groups (0-7 yrs, 8-11 yrs, 12-15 yrs, and
16+ yrs). Similar to Johnson and Newport, they consider prior formal instruction as irrelevant to
age of onset of acquisition, and rely on age when first immersed in the L2 context as onset of
acquisition. They designed an online grammaticality judgment task and a production task. The
grammaticality judgment task was used to measure accuracy and speed judgments. It consisted
of a balanced set of grammatical and ungrammatical wh-questions. The ungrammatical items test
knowledge of Subjacency violations in different conditions: (a) wh-extractions from a noun
complement (e.g., What did you hear the announcement that Ann had received?), (b) extraction
from relative clauses (e.g., Who does Tom love the woman who married?), (c) extraction from
adjunct islands (e.g., Who did you meet Tom after you saw?), (d) extraction from subject islands
(e.g., What was a dish of cooked by Ann?), and (e) that-trace (e.g., Who did Sam say that likes
Jane?)
11
(p.265). They included grammatical wh-questions to ascertain whether the non-native
11
Test items on that-trace violations were removed from analysis because all groups including the native speakers
scored worse on them (White & Genesee, 1996).
Chapter
1
:
Introduction and theoretical background
23
speakers of English would realise that wh-extraction out of embedded clauses (i.e., long
movement) is permissible, and thus, accept such structures (e.g., What does Ann believe that
Jane bought for her birthday?) (p.265). The production task was in the form of a Question
Formation Test. It was used to verify whether the non-native speakers would produce any of the
wh-movement violations. Results showed that the group of near-natives performed in a target-
like way on the grammatical and ungrammatical structures in terms of accuracy and speed. There
were also no substantial effects due to age of exposure. The groups of near-natives who were
first exposed to English under the age of 7 and those after the age of 16 showed indistinguishable
target-like competence. This was attested in production, judgment, and speed data. Thus, they
concluded that attainment of native-like competence is possible beyond the critical period, and
UG knowledge remains accessible throughout life. Yet, given that the majority of the L1
speakers are French, which has wh-movement as in English, their results are suggestive of
positive transfer from the L1 (Long, 2005, 2007). The authors themselves acknowledge this
interpretation, and identify their L1-L2 pairing as a limitation in their study. This, nevertheless,
highlights an advantage for Johnson and Newport’s (1991) study in that they examined
Subjacency as a property that is differently instantiated in L1 and L2. In fact, results of White
and Genesee’s (1996) study are not inconsistent with the claims of the ‘Partial Access’
hypothesis to UG. Recall that it predicts divergence when the L2 property is not instantiated in
L1, but not when the property is present in L1. On the contrary, it predicts convergence in the
latter case. This is apparently compatible with results from the French speakers in White and
Genesee (1996).
Turning to the domain of semantics, Montrul and Slabakova (2003) argue against a critical
period for the L2 acquisition of aspect. In their study, they intended to challenge Coppieters’
(1987) generalisation that aspectual contrasts in an L2 that are not present in the L1, as in the
French imparfait and passé composé contrast that does not exist in English, are un-acquirable
Chapter
1
:
Introduction and theoretical background
24
even in the L2 grammars of near-native speakers deemed to have reached an end-state. In the L2
acquisition of Spanish, they studied the aspectual contrast between the preterite and imperfect in
the L2 grammars of very advanced English non-native speakers. This contrast illustrates the
semantic differences between a perfective and an imperfect reading. The perfective reading
describes a bounded event that has a starting and an end point as in (2), whereas the imperfect
reading describes an event that is unbounded with no starting or ending point as in (3). Examples
from Montrul and Slabakova (2003, p.357).
(2) Paula pintó un cuadro. ‘Paula painted a picture’
(3) Paula pintaba un cuadro. ‘Paula was painting a picture’
The preterite/imperfect distinction is morphologically marked on Spanish past tense verbs, but is
not morphologically marked on English verbs. As indicated from the English gloss, the
equivalent of the Spanish imperfect marker (-aba) is the progressive (-ing) in English. Since
English does not morphologically realise the imperfect, Montrul and Slabakova assume that
English lacks the [-perfective] feature. They tested this case over a range of morphological and
semantic consequences against the assumptions of the FFFH (Hawkins & Chan, 1997). Recall
this hypothesis assumes a critical period for acquiring abstract features not instantiated in the L2
speakers’ L1, and that even at very advanced levels, L2 speakers will be unable to represent
these features in a target-like way. Yet, they can make use of the available UG-resources to
approximate L2 grammar, and thus, ‘appear’ native-like. This is often overlooked in debates
against the hypothesis, however. Montrul and Slabakova conducted a study with three groups of
proficient English learners of Spanish: 24 advanced learners, 23 superior, and 17 near-natives.
They were classified according to their combined score on a proficiency test and ratings of
nativeness. They used the same procedures to that of White and Genesee (1996). They designed
three instruments: a morphological recognition task, a sentence conjunction judgment task, and a
Chapter
1
:
Introduction and theoretical background
25
truth value judgment task. Through these tasks, they studied the effect of [+/-perfective] over a
range of semantic consequences: (a) generic vs. specific interpretation of impersonal
constructions with preterite and imperfect forms, (b) habitual vs. one-time readings of the
preterite and imperfect, and the (c) morphological contrast between preterite and imperfect.
Overall results showed that the group of near-natives performed in a target-like way on all three
tasks and conditions, with 70% of them indistinguishable from the natives at the individual level.
They take this as evidence against the FFFH. Although the feature [-perfective] is not
instantiated in L1 English, their Spanish L2 groups managed to realise the semantic distinction
between the preterite and imperfect morphology. They assert that it is possible to acquire even
among those who are not fully immersed in the target language.
However, if one holds the view that [+/-perfective] is an interpretable feature as Montrul and
Slabakova do, then their results should be compatible with the current version of the partial
accessibility to UG hypothesis, the IH. Recall, interpretable features are not subject to
maturational constraints, and that fossilisation is not predictable in this area. This is a position
that the present thesis adopts for the semantics domain.
Overall, the samples of studies discussed in this section raise many concerns for the critical
period debate. They highlight the impact of a number of interrelated factors such as, L1
grammar, linguistic domain, structure type, test modality, and context of exposure. In one view,
the different findings obtained about Subjacency by White and Genesee (1996), and Johnson and
Newport (1991) can be interpreted in terms of L1 effects rather than absence of age effects.
Another issue is the effect of age across different linguistic domains. Together results from
studies like Hawkins et al. (2008), Tsimpli and Dimitrakopoulou (2007), and Montrul and
Slabakova (2003) illustrate how far morpho-syntactic and semantic properties could be
differently affected by the age factor (cf. resumptives vs. aspect). Even at a micro-level, different
Chapter
1
:
Introduction and theoretical background
26
structure types appear more vulnerable to an age effect than others (cf. progressive vs.
determiners). This has been consistently shown in the findings of Johnson and Newport (1989),
and Bialystok and Miller (1999). Most importantly, the role of context (cf. naturalistic vs. formal
instruction) in the age inquiry is an important factor because it determines the nature of input
(i.e., significant vs. insignificant exposure). From all factors mentioned, the context of exposure
is a central issue to the inquiry of age in this thesis. Therefore, age studies conducted in a formal
classroom setting merit a closer look. This is the topic of section 1.4.
1.4. Age effects in a minimal input setting
In the debate about whether there is a critical period, it is unresolved whether the younger is
better in the long runphenomenon observed in naturalistic situations applies in minimal input
(foreign language classroom) situations as well. According to Muñoz (2006a), the effect of
context (i.e., instructed vs. naturalistic) has been ignored in this debate, and age-differences
observed in naturalistic situations (e.g., Johnson & Newport, 1989, 1991) are wrongly
generalised to instructed situations with minimal input. Therefore, the optimal age to start
learning a foreign language remains a central topic that requires critical examination. Indirect
evidence that suggests an effect of context of exposure can be traced back to Johnson and
Newport’s (1989) seminal work. In their study, age of arrival in the USA (i.e., naturalistic
setting) correlated highly with test scores of the Chinese and Korean L2 learners (r= -.77),
whereas age at first formal instruction (i.e., instructed setting) showed a lower correlation with
the test scores (r= -.33). From these results, they concluded that starting age makes little
difference to outcome in an instructed setting, but can be advantageous in a naturalistic setting.
Thereafter, in most of the critical period studies (e.g., Birdsong, 1992; Hawkins & Chan, 1997;
Johnson & Newport, 1989, 1991; White & Genesee, 1996), age at first classroom instruction in a
Chapter
1
:
Introduction and theoretical background
27
foreign language was ignored because it was deemed to be insignificant, with age effects only of
significance where immersion is involved.
However, more direct evidence from studies of instructed learners in classroom setting is
required before the effects of starting age in these environments can be fully evaluated. Two
older, but well-known, studies conducted in the seventies are Oller and Nagato (1974), and
Burstall, Jamieson, Cohen, and Hargreaves (1974). Both studies suggest little advantage for early
instruction. In the context of Japan, Oller and Nagato compared a group of Japanese students
who had first started learning English in elementary school with another group who began at
intermediate school level, using a Cloze test. Results showed that although the elementary
starters outperformed the intermediate starters at grades 7 and 9, their performance became
indistinguishable from the intermediate starters at the 11
th
grade. These findings imply that in 5
years teen starters can learn what child starters have learned in 11 years (Singleton & Ryan,
2004). Similar findings were obtained in the context of Britain. Burstall et al. report on a pilot
scheme that aimed to evaluate the advantages of teaching French as a foreign language to
English students at the age of 8 years old in elementary school. A group