This thesis has been submitted in fulfilment of the requirements for a postgraduate degree

(e.g. PhD, MPhil, DClinPsychol) at the University of Edinburgh. Please note the following
terms and conditions of use:

This work is protected by copyright and other intellectual property rights, which are
retained by the thesis author, unless otherwise stated.
A copy can be downloaded for personal non-commercial research or study, without
prior permission or charge.
This thesis cannot be reproduced or quoted extensively from without first obtaining
permission in writing from the author.
The content must not be changed in any way or sold commercially in any format or
medium without the formal permission of the author.
When referring to this work, full bibliographic details including the author, title,
awarding institution and date of the thesis must be given.
Processing of novel grammatical features during real-time

second language production and comprehension

Qingyuan Liu Gardner

PhD Psychology

University of Edinburgh

2020
Abstract

Second language (L2) learners often have difficulties acquiring grammatical

features which do not exist in their first language (e.g. inflectional morphology,
number agreement etc.), and exhibit real-time production and comprehension errors
when these features are involved. What are the causes behind such errors? Moreover,
what do they tell us about second language processing in general?

The primary aim of this thesis was to examine the nature of second language
production errors and to scrutinise them with reference to each stage of accepted
models of language production, specifically, whether there are consistent error
patterns which reveal the source(s) of erroneous production in L2 learners. The
second aim of this thesis was to examine the comprehension of novel grammatical
features in the second language, more specifically, whether L2 learners could acquire
the ability to consistently apply L2 grammatical knowledge relating to newly
acquired grammatical features in real-time. The third aim of this thesis was to
examine whether L2 learners have fundamental perceptual deficiencies or biases
concerning selective L2 phonological features as a result of first language
experience, and whether this would affect the perception of specific grammatical
features in the L2.

This thesis addressed these questions in seven experiments by examining the

acquisition of inflectional morphology among L1 Mandarin speakers of L2 English.
Experiments 1, 2 and 3 investigated spoken and written production of inflectional
morphology via an elicited production paradigm. The results argued against
representational accounts and supported activation processing accounts of erroneous
second language production. Information complexity and production modality were
also found to contribute to low production accuracy. Experiments 4 and 5
investigated L2 learners’ auditory and visual sensitivity to inflectional omissions
using self-paced listening and self-paced reading paradigms, where auditory cues
were found to facilitate the detection of inflectional omissions. Experiments 6 and 7
investigated perceptual sensitivity to phonologically variable inflectional morphemes
in an auditory discrimination paradigm. Second language learners exhibited no
perceptual deficiency to novel phonological features, and no consistent perceptual
biases favouring L1 phonological features relative to native speakers.

Taking these findings as whole, the results favoured a processing account of

errors in language production and comprehension, whilst recognising the role of
phonological constraints in both processes. Importantly, these conclusions are drawn
from a broad analysis of multiple aspects of language processing, recognising the
role of conceptual distinctions, grammatical representations, lexical forms, and
phonological factors in second language production and comprehension.
Additionally, this thesis recognises the value of both psycholinguistic models of
language processing and linguistic theories of second language processing.
Lay Summary

Second language (L2) learners often have difficulties acquiring elements of

grammar which do not exist in their first language (L1), and make grammatical
errors when these features are involved. What are the causes behind such errors?
Moreover, what do they tell us about second language processing in general?

The primary aim of this thesis was to examine the why L2 learners make
grammatical errors during L2 speech and to scrutinise them with theoretical outlines
of human speech production, specifically whether there are consistent error patterns
which reveal the source(s) of error in speech or in writing. The second aim of this
thesis was to examine L2 learners’ understanding of L2 grammatical features, more
specifically, whether they could acquire the ability to consistently apply grammatical
knowledge in real-time. The third aim of this thesis was to examine whether L2
learners have fundamental perceptual biases about sounds in the L2 as a result of first
language experience, and whether this would affect L2 learners’ ability to perceive
specific grammatical features in the L2.

This thesis addressed these questions in seven experiments by examining the

acquisition of inflectional morphology (3rd person singular -s and past -ed) among L1
Mandarin speakers of L2 English. Experiments 1, 2 and 3 investigated how L2
learners produce L2 inflectional morphology via a scene description task. The results
support the idea of processing errors to be the main cause of L2 production errors.
The complexity of information marked by L2 grammatical features and modality
(spoken or written) were also found to affect production accuracy. Experiments 4 and
5 investigated L2 learners’ sensitivity to inflectional omissions across auditory and
visual modalities, where auditory cues were found to facilitate the detection of
inflectional omissions. Experiments 6 and 7 investigated L2 learners’ perceptual
sensitivity to inflectional morphemes with different phonological features. Overall,
the findings showed that L2 learners did not perceive L2 phonological features
differently relative to native speakers, and that L2 learners do not have consistent
perceptual biases towards L1 phonological features.
 Taking these findings as whole, the results favoured processing errors over
representational errors in producing and understanding L2 grammatical features,
whilst recognising the role of phonological constraints in both processes.
Importantly, these conclusions are drawn from multiple aspects of language
processing, recognising the role of conceptual distinctions, grammatical
representations, lexical forms, and phonological factors in second language
production and comprehension. Additionally, this thesis recognises the value of both
psycholinguistic models of language processing and linguistic theories of second
language processing.
Acknowledgement

First and foremost, my sincere gratitude goes to Holly, for being the cornerstone
of my PhD. She was not only a source of academic wisdom, but also a kind, caring
mentor who provided personal support during the more challenging times of my PhD
journey. My deepest thanks also go to Vicky, for providing detailed support and
feedback for my experimental design and write up. Her professional advice was
invaluable. I was incredibly lucky to have supervisors who not only recognised the
value of my ideas, but also helped me to develop them into good quality research.

A massive thank you to past and present members of Holly and Martin’s
psycholinguistic lab: Laura and Ruth for being my ‘stats-wiki’; Yue, Yangzi and
Fang for their expertise as native Mandarin psycholinguists; Ellise, Michela and
everyone else for their invaluable support during the last four and a half years. My
appreciation also goes to members of the psycholinguistic and bilingual research
community at the University of Edinburgh, for their invaluable encouragement and
feedback. It has been a pleasure knowing you all.

I would like to thank Daniel, Sam, and Michael for lending me their voices, and
patiently adjusting the speed, volume and pronunciation of hundreds of tedious
sentences during my recording sessions. My thanks also go to more than 500 native
Mandarin and native English research participants. Without their presence and
encouragement, I would not have been able to complete my experimental research on
time.

A big thank you to Dr. Caroline Young, for encouraging me to pursue

Psychology when I was 17.

I would also like to thank my grandmother and my late grandfather, who don’t
speak (much) English and knew very little about my speciality, but nevertheless
firmly believed that I could do a thing called a ‘PhD’. They always valued my
independent thinking, and taught me the value of perseverance from a very young
age. Last but not least, I would like to thank both my parents, Dave and Chun, for
supporting my pursuit of academia, both mentally and financially. All of this would
not have been possible without you.
Table of Contents

Chapter 1.............................................................................................................................. 1
Introduction .............................................................................................................. 1

Aims and structure of thesis ..................................................................................... 4

1.1. L2 grammatical processing and factors affecting L2 grammatical
attainment.............................................................................................................. 6
1.1.1. Critical period and late bilingualism ........................................................... 8
1.1.2. L1-specific effects ...................................................................................... 11
1.1.3. Phonological acquisition and L1 influence................................................ 14
1.2. The influence of processing modality ...................................................... 16

Interim summary ............................................................................................................ 18

1.3. L2 language production: accounts and issues .......................................... 19
1.3.1. Monolingual language production frameworks ......................................... 19
1.3.2. Bilingual adaptations of language production models .............................. 22
1.3.3. Sources of L2 inflectional errors ............................................................... 24
1.4. L2 language comprehension: auditory and visual accounts .................... 27
1.4.1. Models of auditory and visual comprehension .......................................... 28
1.4.2. Sources of L2 comprehension errors ......................................................... 29

Interim summary ............................................................................................................ 33

1.5. Real-time L2 sentence processing in L2 learners .................................... 34
1.5.1. Incremental sentence processing in L2 learners ........................................ 34
1.5.2. Extent of L1 effects on L2 sentence processing .......................................... 36
1.5.3. Cognitive capacity and task demands ........................................................ 37
1.5.4. Temporal information processing in the L2 ............................................... 39
1.6. Linguistic properties of Mandarin Chinese and English ......................... 41
1.6.1. Expressions of temporality ........................................................................ 42
1.6.2. Phonological properties of morphemes ..................................................... 45

Summary of literature review and research questions ............................................ 47

Chapter 2................................................................................................................... 49
2.1. Introduction ............................................................................................. 50
2.1.1. Theories of L2 inflectional production errors ............................................ 52
 2.1.2. Morphological processing in language production .................................. 54
2.1.3. The current study ...................................................................................... 57

2.2. Experiment 1 ............................................................................................ 60

2.2.1. Methods .......................................................................................................... 60
Participants .............................................................................................................. 60
Materials .................................................................................................................. 61
Design ..................................................................................................................... 62
Procedure ................................................................................................................. 62
Coding and Scoring.................................................................................................. 64
2.2.2. Results ............................................................................................................ 65
Overall Inflectional Accuracy .................................................................................. 67
Inflectional Type ...................................................................................................... 68
3rd Person Singular -s (3SG -s) responses .............................................................. 68
Past -ed responses ................................................................................................. 69
Inflectional Omission responses ............................................................................... 70
2.2.3. Interim discussion ....................................................................................... 75

2.3. Experiment 2 ............................................................................................... 76

2.3.1. Methods .......................................................................................................... 76
Participants .............................................................................................................. 76
Materials .................................................................................................................. 76
Design ..................................................................................................................... 76
Procedure ................................................................................................................. 77
Coding and Scoring.................................................................................................. 77
2.3.2. Results ............................................................................................................ 78
Overall Inflectional Accuracy .................................................................................. 78
Inflectional Type ...................................................................................................... 79
3rd person singular -s (3SG -s) responses ............................................................. 79
Past –ed responses ............................................................................................... 80
Inflectional Omission responses ............................................................................... 81
2.3.3. Interim discussion ....................................................................................... 81

2.4. Experiment 3 ............................................................................................... 83

2.4.1. Methods .......................................................................................................... 83
Participants .............................................................................................................. 83
Materials .................................................................................................................. 83
Procedure ................................................................................................................. 83
Coding and Scoring.................................................................................................. 84
2.4.2. Results ............................................................................................................ 84
 Overall Inflectional Accuracy .................................................................................. 84
Inflectional Type ...................................................................................................... 85
3rd Person Singular -s responses .......................................................................... 85
Past –ed responses ............................................................................................... 86
Inflectional Omission responses ............................................................................... 87
Between Experiment Comparisons ........................................................................... 87
2.4.3. Interim discussion ....................................................................................... 96

2.5. General Discussion ...................................................................................... 96

2.5.1. The Locus of L2 Inflectional Errors ............................................................ 97
2.5.2. Limitations and Future Directions ............................................................ 100

Chapter 3................................................................................................................. 103

3.1.1. Fundamental differences between L1 and L2 morphosyntactic processing 106
3.1.2. Theories of L2 sentence processing .......................................................... 109
3.1.3. The effect of comprehension modality on grammatical sensitivity ............ 113
3.1.4. The current study ...................................................................................... 114

3.2. Experiment 4 ............................................................................................. 117

3.2.1. Method.......................................................................................................... 117
Participants ............................................................................................................ 117
Materials ................................................................................................................ 118
Design ................................................................................................................... 119
Procedure ............................................................................................................... 120
Coding and scoring ................................................................................................ 121
3.2.2. Results ................................................................................................ 121
Self-paced Listening (SPL) Task ............................................................................ 121
3.2.3. Interim Discussion .................................................................................... 124

3.3. Experiment 5 ............................................................................................. 125

3.3.1. Methods ........................................................................................................ 125
Participants ............................................................................................................ 125
Materials ................................................................................................................ 125
Design ................................................................................................................... 126
Procedure ............................................................................................................... 126
Coding and Scoring................................................................................................ 126
3.3.2. Results .......................................................................................................... 127
Self-Paced Reading (SPR) Task ............................................................................. 127
Between-experiment (Modality) Comparisons ....................................................... 130
 3.3.3. Interim discussion ..................................................................................... 130

3.4. General Discussion .................................................................................... 131

3.4.1. L2 comprehension mechanism and application of implicit knowledge .... 132
3.4.2. Modality effects in L2 comprehension ..................................................... 135
3.4.3. Limitations and remaining issues ............................................................ 136

Chapter 4................................................................................................................. 138

4.1. Introduction ............................................................................................... 139
4.1.1. L1 and L2 phonological development ....................................................... 142
4.1.2. Accounts of L1 phonological influence ..................................................... 146
4.1.3. The current study ...................................................................................... 149

4.2. Experiment 6 ............................................................................................. 151

4.2.1. Methods ........................................................................................................ 151
Participants ............................................................................................................ 151
Materials ................................................................................................................ 151
Design ................................................................................................................... 152
Procedure ............................................................................................................... 152
Coding and Scoring................................................................................................ 154
Phonological Discrimination (PD) Task ............................................................... 154
Phoneme Elision (PE) Task .................................................................................. 154
4.2.2. Results .......................................................................................................... 155
Phonological Discrimination (PD) Task ................................................................. 155
Phonological Ending Effects ................................................................................. 156
Temporal Context Effects...................................................................................... 157
Phoneme Elision (PE) Task....................................................................................... 159
4.2.3. Interim Discussion .................................................................................... 161

4.3. Experiment 7 ............................................................................................. 162

4.3.1. Methods ........................................................................................................ 162
Participants ............................................................................................................ 162
Materials ................................................................................................................ 162
Design ................................................................................................................... 162
Procedure ............................................................................................................... 162
Coding and Scoring................................................................................................ 162
4.3.2. Results ............................................................................................................... 163
Phonological Discrimination (PD) Task ................................................................. 163
Phoneme Elision (PE) Task .................................................................................... 165
4.3.3. Interim Discussion ...................................................................................... 166
 4.4. General Discussion .................................................................................... 167
4.4.1. L2 perceptual biases and comprehension.................................................... 168
4.4.2. Extent of L1 phonological constraints on L2 production............................. 169
4.4.3. Limitations and remaining issues ................................................................ 171

Chapter 5................................................................................................................. 173

5.1. Inflectional errors in L2 production ......................................................... 174
5.1.1. Theoretical motivations and findings (Chapter 2) ....................................... 174
5.1.2. Theoretical implications.............................................................................. 175
5.1.3. Limitations and future directions ................................................................ 176

5.2. L2 morphosyntactic processing in real time.............................................. 177

5.2.1. Theoretical motivations and findings (Chapter 3) ....................................... 177
5.2.2. Theoretical implications.............................................................................. 179
5.2.3. Limitations and future directions ................................................................ 180

5.3. L1 phonological effects on L2 comprehension and production ................ 182

5.3.1. Theoretical motivations and findings (Chapter 4) ....................................... 182
5.3.2. Theoretical implications.............................................................................. 183
5.3.3. Limitations and future directions ................................................................ 184

5.4. Further theoretical and methodological considerations ............................. 185

5.4.1. Theoretical considerations .......................................................................... 185
5.4.2. Methodological considerations ................................................................... 187

5.5. Further research questions ......................................................................... 188

5.5.1. L2 grammatical features in L1-L2 mixed-speech ........................................ 188
5.5.2. Syllabification of L2 grammatical features ................................................. 189

5.6. Summary and conclusions ........................................................................ 190

Appendices .............................................................................................................. 193

Appendix A. Experiment 1, 2 and 3: L1 Mandarin group language background information . 193

Appendix B. Experiments 1, 2 and 3: Experimental and filler verbs for the scene description task
........................................................................................................................................ 193

Appendix C. Experiments 1, 2 and 3: Experimental and filler images for the scene description
task .................................................................................................................................. 194
 Appendix D. Experiments 1, 2 and 3: Experimental legend and vocabulary list for the
scene description task ..................................................................................................... 200

Appendix E. Experiment 1: Multiple-choice section of Oxford Placement Test (Allan,

1992) with answers ......................................................................................................... 201

Appendix F. Experiment 1: Descriptive and inferential statistics (t-test) on the Oxford

Placement Test ................................................................................................................ 206

Appendix G. Experiments 4 and 5: L1 Mandarin group language background

information...................................................................................................................... 206

Appendix H. Experiments 4 and 5: Experimental sentences with comprehension questions

for the self-paced listening and self-paced reading tasks ................................................ 207

Appendix I. Experiments 4 and 5: Filler sentences with comprehension questions for the
self-paced listening and self-paced reading tasks ........................................................... 214

Appendix J. Experiment 4 and 5: Morphological Proficiency Test (with answers)......... 216

Appendix K. Experiments 4 and 5: Descriptive and inferential statistics on the

Morphological Proficiency Test ...................................................................................... 219

Appendix L. Experiments 6 and 7: L1 Mandarin group language background information

........................................................................................................................................ 219

Appendix M. Experiments 6 and 7: Experimental (Sets A, B and C) and filler verbs for the
Phonological Discrimination Task .................................................................................. 220

Appendix N. Experiments 6 and 7: Practice and test items from the Phoneme Elision Task
........................................................................................................................................ 222

Appendix O. Experiments 6 and 7: Sample transcription of trials from the Phoneme

Elision Task..................................................................................................................... 223

Appendix P. Experiments 6 and 7: Average response accuracy for individual items from
the Phoneme Elision Task ............................................................................................... 225

References ............................................................................................................... 226

 Chapter 1

Chapter 1

Introduction

In a modern, multilingual society, it is very common, and often necessary for

people to learn new language(s) in addition to their own native language during
adolescence and adulthood. However, learning a new (second) language can be a
notoriously difficult and lengthy process, and people do it with varying degrees of
success. To achieve high second language (L2) proficiency, a learner must have a
good command of L2 vocabulary, as well as a mastery of L2 grammar. One of the
most common obstacles for L2 learners is learning new L2 grammatical features
which do not exist in their native language (L1), for example, inflectional
morphology, tense agreement and case marking. As such, one may frequently
observe L2 learners make grammatical errors when producing L2 speech in everyday
situations, even when they ‘know’ the correct grammatical forms. What are the
causes behind such errors? And what can these error patterns tell us about second
language acquisition and processing?

What might be the first steps to learning a new language? First, a beginner might
start with learning new words: their sounds and their meanings, how they are written,
and how they are pronounced. At the same time, he / she learns how words have
different functions, and how they connect to express a message. For adult learners,
the rules of how words should be constructed organised can be straightforward to
learn in the form of metalinguistic knowledge (e.g. a subject is always followed by a
verb, or -ed must be attached to the verb when the action took place in the past), but
it could be much more difficult for them to apply these types of knowledge
spontaneously in everyday situations e.g. in a conversation. As such, learners often
‘know’ the correct grammatical rules, but often make mistakes which violate these
rules when speaking or writing (e.g. placing the verb before the subject, or omitting -
ed for action verbs in a past context). Why do they make such errors? Is it because
their language systems do not ‘understand’ or ‘access’ these L2 grammatical
features, especially when they don’t exist in the learner’s L1? Or is it because they

1
 Chapter 1

cannot consistently ‘remember’ the forms which L2 grammatical features take? Or is

it because they cannot generate or process the sounds representing these grammatical
features? All these are plausible explanations for L2 grammatical errors in L2
production, but few attempts have been made to tease these explanations apart in
psycholinguistic research. In order to address this question, the first aim of this thesis
is to examine the nature of L2 grammatical errors with reference to psycholinguistic
frameworks of language production.

Another way to examine whether L2 learners apply L2 grammatical knowledge

is to investigate whether they are sensitive when these grammatical rules are violated
during real-time L2 comprehension (e.g. listening or reading L2 input where past -ed
is missing in a context which requires it). In order to fully comprehend a message, L2
learners must monitor all relevant grammatical features which may affect the
interpretation of the sentence. Selective attention to grammatical features which
provide crucial information, or the lack of integration between different linguistic
cues could result in misinterpretation of the message. For example, ‘yesterday the
girls play in the park’ may be perfectly acceptable to a Mandarin learner of English
who does not use -ed as a cue for temporal information, but a native English speaker
may find the sentence confusing as play, indicating present tense, does not fit with
the temporal context indicated by yesterday. In other words, in order to achieve
native-like understanding of language, an L2 learner must go beyond knowing L2
features as metalinguistic knowledge, but also acquire the ability to use these
grammatical features in understanding the L2 in a sentential context. The second aim
of this thesis is to examine whether L2 learners make use of L2 grammatical features
when attempting to understand the L2. Moreover, depending on how L2 grammatical
features are learnt, people may be more sensitive to how they look in text form rather
than how they sound in the auditory form. In fact, we often see L2 learners
experience more difficulties when listening to L2 speech compared with reading L2
text. If people do in fact find L2 speech more difficult to understand, then L2 learners
may be less sensitive to grammatical violations when listening to L2 speech than
reading L2 text. Thus, the third aim of the thesis is to examine whether L2 learners
have different levels of sensitivity to grammatical violations in auditory and visual

2
 Chapter 1

comprehension, specifically, whether learners are less sensitive ‘hearing’ than

‘seeing’ L2 grammatical violations.

As we acquire our native language, we become sensitive to how minute

distinctions in sounds may denote different meaning in the L1. Consequently,
distinctions which are irrelevant for differentiating meanings in the L1 are gradually
lost (e.g. length of vowel, tonal distinctions). In speech, L2 learners may find some
L2 words difficult to ‘hear’, as minute differences between sounds can denote
different meanings in the L2 but do not in the L1. For example, a Mandarin learner
of English may mistake sheep for ship as the length of the vowel can denote different
meanings in English but does not in Mandarin. Things become more complicated
when new sounds represent new grammatical features. For example, consonants (e.g.
/s/) may come at the end of an English word to mark an inflection (e.g. 3 rd person
singular -s) which results in a consonant cluster (e.g. /ts/ in shouts). However, if such
consonants clusters are not permitted (or rare) at the end of Mandarin words,
Mandarin learners of English may fail to ‘hear’ this sound, or do not interpret it as an
inflectional marking. To complement previous questions on whether L2 grammatical
features are harder to ‘hear’ in the auditory form compared with in the visual form,
the fourth aim of the thesis is to examine L2 learners’ perceptual sensitivity to
sounds created by L2 grammatical features, and whether L2 learners are more
sensitive to sounds which are permitted (or frequent) in their L1 than those which are
not (or rare).

3
 Chapter 1

Aims and structure of thesis

The main aim of this thesis is to examine how L2 learners use and apply L2
grammatical features during real-time L2 production and comprehension. In Chapter
2, I will examine L2 grammatical errors in production and scrutinise them with
reference to each stage of accepted models of language production. Particularly, I
consider whether there are consistent error patterns during production of inflectional
markings which could reveal the source(s) of erroneous production. I contemplate
two main possibilities: Are grammatical errors in L2 production caused by
representational deficits, or breakdowns in processing? Moreover, to what extent
does the modality of production (spoken or written) affect grammatical accuracy?

In Chapter 3, I will examine how L2 learners process L2 grammatical features

(those absent in the L1) during real-time L2 comprehension. Particularly, I consider
whether L2 learners could acquire the ability to consistently assimilate and integrate
semantic and syntactic information from L2 grammatical features in a native-like
manner. Do L2 learners experience comprehension difficulties if the L2 contains
grammatical violations? Moreover, does modality of comprehension (auditory or
visual) affect L2 learners’ sensitivity to these grammatical violations?

In Chapter 4, I will examine how L1 phonological properties influence the L2

learner’s ability to perceive and produce L2 speech sounds. Particularly, I consider
whether L2 learners have fundamental perceptual biases concerning selective
phonological features as a result of L1 acquisition, and whether this in turn would
affect the comprehension and production of specific L2 grammatical features. Do L2
learners find L2 phonological features shared by the L1 easier to detect? If so, does
the overlap of phonological features between L1 and L2 help L2 learners to become
more sensitive to the relevant L2 grammatical features in comprehension?

For the rest of this chapter, I will present a selection of relevant empirical
research on factors affecting L2 grammatical attainment and processing, as well as
psycholinguistic models of production and comprehension.

4
 Chapter 1

First, I will examine existing research on L2 acquisition, specifically those

concerning the acquisition of inflectional morphology. I will also evaluate key
factors influencing L2 grammatical attainment, including age, L1 background and L1
phonological influence. Subsequently, I will review current accepted
psycholinguistic models of language production, with specific focus on monolingual
and bilingual grammatical processing. For comparison, I will also examine second
language learning theories which provide explanations for L2 inflectional errors
during production. Then, I will review monolingual models of auditory and visual
comprehension with emphasis on the effect of modality. Then, I will focus on how
bilinguals process information from L2 inflectional morphology during real-time
processing. In the last part of Chapter 1, I will introduce the different ways in which
Mandarin and English express temporal information, and key differences in their
phonological properties. These details will be useful when we introduce the
motivations for each set of experiments in Chapters 2, 3 and 4.

To summarise, this thesis investigates sources of error for L2 language

production and integration of information in L2 comprehension, taking multiple
aspects of language acquisition (conceptual distinctions, grammatical
representations, lexical forms, phonological factors) into account. Simultaneously, I
examine non-native like L2 production and comprehension using both traditional
psycholinguistic frameworks and second language learning theories, with a view to
reconcile different approaches to second language processing and to highlight
common grounds. Detailed discussions concerning production and comprehension of
inflectional morphology in L2 learners will follow in Chapters 2, 3 and 4 as I discuss
the motivations for each set of experiments.

5
 Chapter 1

Literature Review

1.1. L2 grammatical processing and factors affecting L2 grammatical

attainment

What does it mean to acquire a second language (L2)? One may argue that like
acquiring the native language (L1), successful second language acquisition is not
simply learning a new set of words and rules of about how they connect together, but
also learning to linguistically interact in the L2 in a spontaneous manner. Highly
proficient L2 learners not only can accurately understand L2 auditory and written
input, but are also able to produce utterances that are both pragmatically appropriate
and adhere to L2 grammatical rules. During this process, successful acquisition of L2
grammatical features is fundamental to both real-time L2 comprehension and
production.

Acquiring L2 grammatical features (e.g. number agreement, inflectional

morphology etc.) is a complex task which requires the L2 learner to learn new
associations between words and their functions. As such, researchers past and present
have used acquisition of L2 grammar as an indicator of L2 attainment. To what
extent is L2 grammatical processing in L2 learners similar to L1 grammatical
processing in native-L1 speakers? Are we presented with qualitative or quantitative
differences in L1 and L2 grammatical processing?

Past research has shown fundamental qualitative differences between L1 and L2

grammatical processing using both neurological and behavioural measures. Event-
related potential (ERP) studies have revealed disparities in P600 responses between
L1 and L2 learners when encountering grammatical violations during comprehension
(Chen, Shu, Liu, Zhao & Li, 2007; Frenck‐Mestre, Foucart, Carrasco, &
Herschensohn, 2009; Frenck‐Mestre, Osterhout, McLaughlin, & Foucart, 2008;
Jiang, 2004; 2007; Osterhout & Mobley, 1995). Among other cases, L1 Chinese1

1
The terms Chinese and Mandarin Chinese are used interchangeably according to the descriptions in individual
studies. Though the Chinese language encompasses many dialects, they are assumed to have the same underlying
grammatical properties (excluding Cantonese, Hokkien, Hakka etc. which are grammatically distinct).

6
 Chapter 1

learners of English were found to lack native-like P600 responses evident in L1-
English speakers when processing subject-verb agreement violations (e.g. the price
of the car were too high), a grammatical feature which is absent in Mandarin Chinese
(Chen et al., 2007). Importantly, Chen et al. (2007) found that whilst L1 Chinese
learners of English exhibited non-native-like neurological responses during on-line
comprehension, they performed as well as native English speakers on explicit
grammaticality judgement tasks, indicating that non-native-like grammatical
processing in L1 Chinese learners was restricted to real-time comprehension.

Looking at behavioural studies which examined sensitivity to syntactic

agreement during sentence comprehension, there is further evidence for fundamental
differences in L1 and L2 grammatical processing. In reading studies which focus on
subject-verb agreement and plural marking, L2 learners must integrate local syntactic
information as well as processing suffixes. Jiang (2004, 2007) showed that L1
Chinese learners of English were not sensitive to English subject-verb agreement
(e.g. The bridges to the island was about ten miles away*2) and plural marking (e.g.
The child was watching some of the rabbit in the room*) violations during real-time
reading comprehension, even though they exhibited offline knowledge of both types
of L2 agreement. Insensitivity to inflectional morphemes found in these studies could
be interpreted as the result of a lack of real-time syntactic integration between the
relevant linguistic cues (e.g. plurality of bridges and singular form of be).

On a morphological level, one prominent study by Silva and Clahsen (2008)

showed that L2 English learners do not decompose English morphological structure
to the same depth as L1 English speakers. Four priming experiments tested L1
Chinese, German and Japanese speakers of English in their processing of inflected
and derivational word forms in L2 English. The findings showed that unlike L1
English speakers who consistently showed priming effects across both derivational
and inflected forms (i.e. humid – humidity produced faster reaction times than humid
– loud; and boil – boiled produced faster reaction times than boil - jump), L2 learners
showed only partial priming effects for derivational but none for inflected forms (i.e.
humid – humidity produced faster reaction times than humid – loud, but boil – boiled

2
* denotes ungrammaticality throughout this thesis.

7
 Chapter 1

did not produce significantly faster reaction time compared to boil - jump ). This
indicated that L2 learners store derivational forms as semantically related words but
inflected forms as uninflected wholes rather than as stem + suffix combinations
(Ullman, 2004; 2005). The authors also discussed these findings in relation to
previous research which suggested that L2 English learners could fail to acquire L2
functional categories for the past tense grammatical feature at a syntactic level, or
they could have problems with feature specifications for inflections3. In other words,
non-native like processing of inflected forms could reflect one of two underlying
problems: 1) missing representation for a grammatical feature, or 2) failure to map
feature to form. Overall, current findings indicate that L2 learners do not decompose
morphologically complex words in a native-like manner (also see Clahsen, Felser,
Neubauer, Sato & Silva, 2010, and Clahsen & Felser, 2018, for review and
discussion).

To summarise, a number of studies has shown that L2 learners have generalised

difficulties processing and integrating L2 grammatical features in a native-like way.
What are the key factors affecting L2 grammatical attainment? In the next section, I
review a selection of studies investigating the extent of influence imposed by age
(critical period), L1 background and phonological development.

1.1.1. Critical period and late bilingualism

Over the years, researchers have recognised the different ways in which
individuals could become fully functioning bilinguals with a good command of the
L2. Whilst some learners successfully acquire both languages simultaneously after
birth, others acquire one language after the other either during childhood or later in
life. With a view to focus on a specific group of late L2 learners, we must recognise
the effect of age and maturation on the L2 population. That is, the effect of a critical
period on L2 learners who have already developed their L1 since birth.

3
See 1.3.3 for details of these accounts.

8
 Chapter 1

Second language researchers have long argued for an effect of age and
maturation on level of language attainment (Lenneberg, 1967). Particularly,
researchers proposed the concept of a critical period, where a significant difference
in L2 attainment could be observed for L2 learners who began acquiring the L2
within and outside the critical period (see Birdsong, 2005). Second language studies
have shown clear differences in proficiency between L2 learners who acquired the
L2 at different ages. For example, Johnson and Newport (1989) tested L2
grammatical proficiency in L2 English learners from L1 Mandarin or L1 Korean
backgrounds who lived in the US, with their age of arrival (3 to 39) used as a proxy
for age of acquisition. The findings most notably showed that L2 learners who
arrived before reaching the end of puberty bore a significant advantage for
grammatical proficiency over those who arrived after puberty. Among early arrivals
(before the end of puberty, age 17), the correlation between age and performance
declined after age seven. Moreover, grammatical proficiency as measured by
grammatical judgement tasks was significantly more variable with no clear trend
between individuals who arrived after puberty (after age 17). Such patterns were
found across different types of L2 grammatical features, including the focus on the
current thesis - English inflectional morphology.

The critical period assumption is not without controversy. Studies have also
shown that near-native levels of grammatical attainment are possible for learners
who acquired the L2 after puberty. Previous brain imaging studies have shown that
native-level grammatical processing is achievable, if the learner’s L1 shares
considerable similarity with the L2 (Tolentino & Tokowicz, 2011). Using
behavioural measures, Hopp (2010) showed that L2 learners (from Russian, Dutch
and English L1s) who acquired the second language after puberty were able to
achieve near-native level performance in offline (explicit) grammaticality judgement
and (implicit) self-paced reading tasks relating in German number and case marking.
He hypothesized that if L2 learners were as sensitive as native-L1 speakers to L2
grammatical violations, they should show similar accuracy in grammaticality
judgement tasks, and similar increased processing effort (in reaction time) in reading
tasks. His findings showed that it was possible for the most proficient L2 learners to

9
 Chapter 1

exhibit near-native sensitivity (in grammaticality judgment) and native-like

processing (in self-paced reading) to L2 grammatical violations.

What was most interesting about Hopp’s (2010) findings is how L2 groups
performed against the native-L1 group in tasks with high processing load. Data from
the speeded grammaticality judgement task in these studies showed that L2 learner’s
performance declined in this task compared with previous offline and real-time
language processing tasks. But crucially, the native-L1 group also showed a decline
in performance in the condition with the highest processing load, despite showing a
robust advantage over L2 groups at lower processing loads. This demonstrated that
native-L1 speakers were not immune to making inaccurate grammatical responses,
rather that native-L1 speakers had a higher threshold for making processing errors
compared with L2 learners. Hopp claimed that these findings demonstrated that the
differences observed between L1 and L2 performance are more likely a quantitative
difference in processing efficiency, rather than a qualitative difference in
grammatical representation (see also Kilborn, 1992). I argue that these two sets of
evidence from Johnson and Newport (1989) and Hopp (2010) are not strictly
opposing in nature, and that apparently conflicting assumptions about critical period
constraints and near-native L2 acquisition could be compatible at least in some cases.
In other words, the critical period constraint on L2 grammatical proficiency is not a
binary concept. Instead, the extent of the constraint could be lessened or exacerbated,
depending on individual cognitive abilities. In fact, even for theories which support a
qualitative difference in grammatical processing between L1 and L2 learners (e.g.
The Fundamental Differences Hypothesis, Bley-Vroman, 1988, 2009), proponents
recognise individual differences (e.g. role of verbal-analytical skills) in helping L2
learners to achieve near-native levels of grammatical proficiency (DeKeyser, 2000).
Such cognitive skills could benefit L2 learning on an individual basis, irrespective of
whether a critical period fundamentally constrains L2 grammatical attainment.

To summarise, research studies have found significant differences in

performance between L2 learners who acquired the L2 before and after puberty.
Such evidence has been argued in support of a critical period for language
acquisition. However, other research has also found near-native levels of

10
 Chapter 1

grammatical attainment in some late L2 learners, with native-L1 speakers also

producing similar errors in cognitively demanding situations.

Further questions regarding the role of L1 (in L2 learners) remain unresolved:

Do all L2 learners find aspects of L2 acquisition equally difficult, or does the degree
of similarity between L1 and L2 contribute to the ultimate attainment of L2
grammar? In other words, could L2 learners experience acquisitional difficulties as a
result of L1-specific constraints?

1.1.2. L1-specific effects

Do all L2 learners find L2 acquisition equally difficult, or do similarities

between L1 and L2 grammar facilitate L2 acquisition in some L2 learners? One may
speculate that L2 learners could find aspects L2 grammar particularly difficult to
process in real-time if the learner’s L1 does not have an equivalent grammatical
feature which serves a similar function (e.g. using inflectional morphology to convey
temporal information). Researchers have argued that the extent of differences
between L1 and L2 grammar can influence the nativelikeness of L2 grammatical
processing (see Tolentino & Tokowicz, 2011, for review). That is, L2 learners would
perform less well if their L1 did not share similar grammatical features than if it did.
In fact, Hopp (2010) found that L1 Russian learners of German, with a
morphologically rich L1, were more native-like in their real-time L2 sentence
processing than L1 Dutch and L1 English learners of German, indicating an L1-
based advantage in the L2 group. In a similar way, Liu, Bates and Li (1992) have
found strong L1 influence on both Chinese-English and English-Chinese late
bilinguals on their use of animacy-based and word order strategies during L2
sentence processing. However, previous findings from a number of studies on on-line
syntactic processing seemed to argue against this assumption (see Marinis, Roberts,
Felser & Clahsen, 2005; Felser, Clahsen, Roberts & Gross, 2003 and Papadopoulou
& Clahsen, 2003 for evidence against L1-induced parsing biases). Hence, there are
mixed interpretations of whether similarities between the learner’s L1 and L2 would
pose significant advantages in L2 grammatical attainment, especially concerning L2

11
 Chapter 1

learners’ ability to apply L2 grammatical knowledge in real-time (this topic will be

expanded further in 1.5.).

Turning to the target population of the current thesis, a good number of second
language studies have examined L1 Chinese learners of English on their acquisition
of English grammar, due to the fact that many English grammatical features are
absent in Mandarin Chinese, e.g. inflectional morphology, subject-verb agreement.
One notable case is the study of one native Mandarin / Hokkien speaking individual
(named Patty), who had been living in the US for over ten years at the time of
investigation. Lardiere (1998a; 1998b; 2000; 2003) studied her oral and written
production data over a period of eight years. Notably, Patty’s oral production data
contained low accuracy in regular past tense inflection (5.8%) even after prolonged
L2 immersion. This stands in contrast with her written production data, which was
substantially more accurate for regular past tense marking (78%). Similar production
error patterns have been found for L1 Mandarin learners of L2 English across
different age groups, showing that this is not simply an individual case of recurrent
production errors, but rather a common trend across L2 English learners from L1
Mandarin backgrounds (Paradis, Tulpar & Arppe, 2016; Hsieh, 2008). Is L2 English
morphology more difficult to acquire for L1 Mandarin learners of English compared
with L2 learners from other L1s?

Looking back at the study by Silva and Clahsen (2008), among the three groups
of L2 English learners from L1 Chinese, German and Japanese backgrounds, the L1
Chinese group exhibited less priming effects, producing significantly higher error
rates and longer primed reaction times than the L1 German group. Crucially
however, all three groups showed similar response patterns, exhibiting priming
effects for derivational forms, but not inflected forms, which were significantly
different from native-L1 speakers. Therefore, the authors concluded that despite
performance was better in L2 learners from L1s which share similar linguistic
properties to L2 English (i.e. rich morphology of German and past tense suffix in
Japanese), they did not make L2 learners’ acquisition more native-like than L2
learners from L1s without shared linguistic properties (i.e. Mandarin Chinese).

12
 Chapter 1

In L2 production, L1 (Mandarin) Chinese learners of English have been found to

exhibit higher error rates than learners from other L1 backgrounds. For example,
Hawkins and Lizska (2003) contrasted L1 Chinese, German and Japanese learners of
English on the spontaneous oral production of past tense markings. Given that all
participants were advanced L2 learners as measured by offline grammar tests, L2
learners from L1 Chinese backgrounds were significantly less accurate in spoken
production than L2 learners from L1 German and L1 Japanese backgrounds on both
regular past tense -ed forms (e.g. talked) and irregular past tense forms (e.g. ran).
These findings have been attributed to the absence of past tense feature in the L1
Chinese instead of other L1 factors (e.g. phonological properties). In another study,
Amaro, Campos-Dintrans and Rothman (2018) contrasted past -ed production in L1
Spanish, Japanese and Mandarin learners of English based on the phonological
properties of their L1s. Specifically, all three L1s restrict consonant clusters but only
Japanese has a prosodic structure4 which allows adjunction of the past -ed
morpheme. In spoken production, all three L2 groups exhibited significantly lower
accuracy in past -ed than native-L1 controls. Consistent with previous research, L1
Mandarin learners again showed the lowest accuracy of past -ed production across all
three L2 groups. Across groups, L1 Mandarin learners were significantly less
accurate than L1 Japanese learners, but were not significantly less accurate than L1
Spanish learners of English in their spoken production of past -ed. Importantly, L1
Spanish learners of English were not significantly less accurate than L1 Japanese
learners of English either. The authors argued that given only Japanese uses prosodic
structure which allows for the adjunction of the past tense morpheme, they should
outperform both L1 Spanish and L1 Mandarin groups if prosodic structure was the
determining factor for past -ed production. However, this was not the case, and
therefore L1 phonological properties, especially prosodic structure could not solely
account for errors in past -ed production.

To summarise, studies which contrasted production accuracy of L2 learners from

multiple L1 backgrounds have not shown a distinct and deterministic effect of L1-
specific properties on production of L2 inflectional morphology. Instead, researchers

4
see 1.3.3 and Chapter 4 for detailed explanations of prosodic structure.

13
 Chapter 1

found a generalised disadvantage in L2 learners, but could not yet agree on the
primary cause of L2 inflectional errors.

1.1.3. Phonological acquisition and L1 influence

One of the key challenges potentially facing L2 learners is L2 phonological

acquisition, which is critical for L2 perception and production. Due to the language-
specific nature of phonology, L2 learners must learn to perceive and identify
phonological units (e.g. phonemes and syllables) in L2 speech as well as rules
governing how these units could be combined (i.e. phonotactics; Stockwell, 1954; cf.
Hill, 1958) during L2 acquisition. Unlike L2 grammatical acquisition, which can be
facilitated by explicit learning, L2 phonological processing relies primarily on
implicit learning, which could be fundamentally constrained by exposure to the
learner’s L1. Here, I review a selection of studies examining the extent of L1
phonological constraints on L2 phonological acquisition, as non-native like L2
phonological processing can affect how L2 speech is perceived as well as how it is
produced.

Different from more explicit forms of learning, phonological development is

driven by implicitly-learned perceptual sensitivity to speech sounds, which could
begin with minimal language exposure. L1 learners develop increased perceptual
sensitivity to phonological distinctions in their native language starting from a very
young age (see Best, 1994; Maurer & Werker, 2014, for review). For example, this
can relate to phoneme categories, stress or pitch, depending on the phonological
features which mark semantic distinctions in the L1. As perceptual sensitivity to
these phonological distinctions become more prominent, sensitivity to distinctions
irrelevant in the L1 are gradually lost (Speech Learning Model; Flege, 1995).
Consequently, during the initial stages of L2 acquisition, L2 learners may interpret
L2 phonological features in terms of articulatory similarity to the L1 (Perceptual
Assimilation Model, Best, 1995). This perceptual bias favouring L1 phonological
distinctions can be problematic for L2 speech perception and production if selective
L2 phonological distinctions are irrelevant in the L1. For example, phonological

14
 Chapter 1

distinctions between the phonemes /r/ and /l/ are important for an L1-English speaker
as they mark semantic distinctions in English phonology (e.g. it distinguishes the
word race from the word lace), but this distinction do not denote different meanings
in Japanese. Consequently, L1 Japanese learners of English are known to be
insensitive to distinctions between the phonemes /r/ and /l/, and exhibit difficulties in
both perception of words involving these phonemes (Goto, 1971). If L2 learners are
unable to perceive selective L2 phonological distinctions, such insensitivities may
also contribute to the erroneous production of these distinctions, though studies have
shown that initial deficiencies in perception and production can be improved through
exposure and training (Strange & Dittmann, 1984; Bradlow, Pisoni, Akahane-
Yamada & Tohkura, 1997). If we take these findings in the context of acquisition of
L2 grammar, selective insensitivity or perceptual bias to L2 phonemes which mark
important L2 grammatical features (e.g. inflectional morphology) can influence how
these grammatical features are perceived and produced in real-time.

Similar to learning phonological distinctions, infants also implicitly learn about

the likelihood of speech sounds cooccurring in their native language (phonotactics)
from a very young age. Such statistical regularities pose constraints on speech
perception in a way that facilitates the identification of words and word boundaries
(see Romberg & Saffran, 2010, for review). Empirical evidence has shown that
infants begin to assimilate phonotactic information from a very young age (Jusczyk.,
Luce & Charles-Luce, 1994; Friederici & Wessel, 1993), with effects persisting
through adulthood (Vitevitch, Luce, Charles-Luce, & Kemmerer, 1997). If we
assume that the constraints imposed by phonotactic information apply in a language-
specific way, then all L2 learners are faced with the task of ‘relearning’ new
phonotactic regularities during L2 acquisition. Again, this could be especially
problematic when L2 phoneme combinations which violate L1 phonotactic rules
mark important grammatical features in the L2, and thus affecting how an L2
message should be interpreted. For example, an L2 English learner must recognise
all possible phonological cues for past -ed marking (i.e. [t], [d], [ɪd]) in combination
with the relevant verbs in order to successfully comprehend past events, though this
may not always be permissible in the learner’s L1. In fact, according to McQueen
and Cutler (1998) and Cutler and Clifton (1999), although listeners do not

15
 Chapter 1

obligatorily decompose morphologically complex words in their phonological form

for spoken word recognition, higher-level processing requires the extraction of
information from phonological cues. Failure to establish the link between
phonological cues and constituent morphemes would theoretically be detrimental to
the correct interpretation of spoken sentences.

1.2. The influence of processing modality

Whilst most researchers recognise the production-comprehension distinction in

language acquisition (e.g. Flynn, 1986), modality distinctions within L2 production
and comprehension have not received as much attention in psycholinguistic research
(Meyer, Huettig & Levelt, 2016). These distinctions have both methodological and
theoretical implications with respect to the validity of data and the conclusions we
draw from them. To put it another way, it is important that we do not solely rely on
data from one modality (auditory or visual) for generalised conclusions of language
acquisition and processing. As will be discussed in 1.3 and 1.4, language production
and comprehension have distinct stages of processing depending on the modality of
output and input. However, relatively few research studies on second language
acquisition have directly contrasted production and comprehension data across
auditory and visual modalities in a controlled manner.

Returning to Johnson and Newport’s (1989) study on the effect of age on L2

English grammatical attainment, Johnson (1992) noted that as the stimuli in the
grammaticality judgement task was presented only in the auditory modality, age
effects could have been confounded by errors in L2 speech segmentation (also see
Anderson, 1980). Consequently, Johnson replicated the study presenting the same
stimuli to the same participants but in the visual (reading) modality. Johnson’s
findings showed that grammaticality judgement accuracy in L1 Mandarin and
Korean learners improved substantially for late arrivals, with over twice as many
errors in the auditory versus the visual modality. Johnson attributed the results to the
visual nature of the stimuli, and that participants were much more likely to perform
well if they could review previous segments of the sentence and did not have time
restrictions, whereas this would not be possible in the auditory modality. Murphy

16
 Chapter 1

(1997) replicated the auditory disadvantage on grammaticality judgement accuracy

using new stimuli, with reaction time data from a grammaticality judgement task also
supporting an auditory disadvantage on processing speed. Specifically, participants
produced longer reaction times for auditory compared with visual grammaticality
judgements, supporting the claim that auditory speech adds processing load during
L2 comprehension. Such findings raise questions of modality biases in research
methods, such that auditory measures of L2 grammatical attainment must account for
the influence of L2 speech processing before making generalised conclusions.

In L2 production research, Lardiere (1998a; 1998b; 2000) collected both oral

and written production data from Patty, the L1 Mandarin / Hokkien L2 English
learner who had been living in the US for over 10 years. The data were analysed for
the accuracy of regular and irregular English verb production. Most notably,
Lardiere’s findings showed that Patty’s production accuracy on regular English verbs
(requiring past -ed) was substantially more accurate in the written production
compared to oral production (78% vs. 5.8%). This pointed towards an effect of
articulation on increased errors during oral production. Of course, one may argue that
as most production data are spontaneous, and researchers do not have control over
the content and syntactic context that production takes, direct comparisons between
data across production modalities are not conclusive in this way. Therefore, to
improve the validity of production data comparisons, it is essential that both
production content and modality are controlled for in future production studies. One
good example which implemented this comparison is a study by Amaro et al. (2018),
which contrasted past -ed production accuracy in spoken and written modalities
across L2 English learners from L1 Mandarin, L1 Japanese and L1 Spanish
backgrounds. Not only did they investigate past -ed production in both spoken and
written modalities, they also restricted participants’ responses by supplying the bare
verb for sentence completion tasks. Their findings again showed higher accuracy in
written compared with spoken production, such that L2 learners from L1 Mandarin
and L1 Spanish backgrounds were as accurate as L1 English controls in written
accuracy for past -ed.

17
 Chapter 1

To summarise, past studies which focused on L2 modality effects have found

that modality affects both grammatical sensitivity in L2 comprehension and
grammatical accuracy in L2 production. However, potential confounding factors and
the implications of existing evidence require further examination.

Interim summary

In section 1.1., I have discussed a selection of research studies examining the

neurological evidence for L1 and L2 grammatical processing. The majority of ERP
studies demonstrated that L2 processing use different mechanisms compared to
native-L1 learners on a non-behavioural level, especially with regard to L2
grammatical / syntactic violations. This effect is mediated by factors such as L2
modality, L2 proficiency, L1-L2 similarity etc. Subsequently, I reviewed key
research concerning three major influencing factors over L2 grammatical attainment:
age (critical period), L1 specific effects, and phonological constraints. First, past
research suggested late L2 learners exhibit low performance accuracy compared with
early L2 learners, pointing towards a constraining effect of age and maturation
(Johnson & Newport, 1989). However, more recent research suggested that native-
level attainment is possible in some cases (Hopp, 2010). Research on production of
L2 grammar suggested that L2 learners from some L1 backgrounds can experience
more difficulties if L2 grammatical features are absent in the L1 (Hawkins & Liszka,
2003), though a L1-specific disadvantage is still unclear with regard to L2
comprehension. Lastly, research studies suggested a role of phonological
development in L2 production and comprehension. With maturational changes
during L1 acquisition affecting perceptual sensitivity to phonological distinctions and
phonotactic learning, I argue that L2 learners are likely encounter difficulties during
perception of L2 grammatical features if the relevant phonological and phonotactic
constraints are not permitted in the L1. This may in turn affect production of these
features as well.

In 1.2., I examined existing research on the effect of modality on L2

comprehension and production. Previous studies show significant differences in L2
performance depending on the modalities of comprehension and production. Such

18
 Chapter 1

findings highlight the modality-specific nature of L2 comprehension and production

and raise the methodological importance of conducting studies across multiple
modalities. The precise contribution of speech segmentation and articulation to L2
comprehension and production will be discussed in 1.3.

1.3. L2 language production: accounts and issues

When we speak, we must first think about the message we wish to convey,
retrieve the concepts and words we wish to use, organise them in a way that adheres
to the grammar of the language spoken (L1 or L2), retrieve corresponding sounds,
before articulating the message in the form of an utterance. This is a complex, multi-
stage process which occurs at great pace during language production. Given this, it is
predictable that L2 learners, given the non-native nature of their acquisition, could
make grammatical errors when producing L2 speech. Psycholinguistic models of
language production have formalised this process for monolingual speakers and have
adapted versions for bilingual speakers. Here, I will briefly outline of the basic
components of language production models before discussing possible underlying
causes of L2 grammatical errors in speech production. I will focus on spoken
production errors first, before moving on to discuss the implications for written
production.

1.3.1. Monolingual language production frameworks

Current psycholinguistic models commonly recognise a modular structure where

each stage of the language production process is responsible for a different aspect of
language processing (see Figure 1.2). First, the speaker must conceptualise the
message (conceptualisation), and decide on the communicative intention, perspective
and semantic relationships of the message (macroplanning and microplanning). At
this stage, the preverbal message is not yet language-specific. However, according to
Levelt (1989), the speaker takes the grammatical properties of the intended language
spoken into account, and selectively processes information necessary for the overt
production (also see Schlesinger, 1977 and Slobin, 1987).

19
 Chapter 1

Second, the preverbal message undergoes grammatical encoding. At this stage,

lexical concepts or lexical representations concerning the message are also selected,
which in turn activate the syntactic structure of each concept, otherwise known as the
lemma. Lemmas are part of the mental lexicon, which contain syntactic information
(such as syntactic category, e.g. noun, verb), as well as diacritic features (e.g. tense,
aspect, number) of the concepts. The activation of the lemma produces the surface
structures necessary for subsequent processing. Note that in Bock and Levelt’s
(1994) account of grammatical encoding (Figure 1.1), function assignment is added
to grammatical encoding as a separate step following lemma activation. Function
assignment allows syntactic information from the lemma to be used to assign
syntactic relations and grammatical functions. Important for our discussion, Bock
and Levelt also claimed the generation of fine-grained details of words (e.g.
inflectional morphemes) are encoded as part of positional processing as the speaker
assembles the order of words (including position of morphemes) in the message.

Figure 1.1. Stages of grammatical and phonological encoding in language production. Taken from
Bock and Levelt (1994).

20
 Chapter 1

Figure 1.2. A theoretical outline of language production, containing 6 main steps from
conceptualisation to articulation. Taken from Levelt, Roelofs and Meyer (1999) and annotated.

Following this, the message undergoes (morpho) phonological encoding, where

the phonological codes of each word (lexemes) are retrieved from the mental lexicon.
In this process, phonemes and phonological properties (stress, pitch) of morphemes
become available to the speaker and syllabification takes place to create the
phonological word of the message. During syllabification, default canonical forms
are ‘resyllabified’ to accommodate the phonological context of speech. For example,
the speaker must allow for the fact that inflectional morphemes might add additional
syllables to words (e.g. predict with -ed inflection must be syllabified as pre-dic-ted).
Subsequently, phonetic encoding activates the articulatory gestures and converts
them into articulatory score. Finally, the message is converted into overt speech by
executing the articulatory score (articulation). It should be noted that in written

21
 Chapter 1

production, it is assumed that orthographical and motor gestural information are

activated instead (see van Galen, 1991, for detailed descriptions).

Under Levelts’ assumptions, this production process relies on spreading

activation principles which work in a feed-forward manner without between-level
interactions. This means that node representations with the highest activation at each
level are selected for further processing (Dell, 1986), and activation at each stage of
production is unidirectional without feedback from later stages.

1.3.2. Bilingual adaptations of language production models

Bilingual adaptations of Levelt’s (1989) original model suggested several

important changes. For example, in de Bot’s (1992) adaptation, bilingual speakers
select the choice of language during conceptualisation and language information is
contained within the preverbal message. In addition, bilinguals have a single mental
lexicon which contain subsets of items (lemmas and lexemes) for each language.
With regard to how the two languages interact, L1 and L2 lexical items have been
suggested to share conceptual representations but have separate lexical
representations for each language (Kroll, 1993). Moreover, bilingual speakers have a
single articulator for the overt execution of articulatory score.

Figure 2.1. Selection of language-specific lemma from conceptual level information. Taken from
Poulisse and Bongaerts (1994).

22
 Chapter 1

Figure 2.2. A blueprint of L1 and L2 diacritic features connected to lemma level nodes. Solid lines
indicate complete acquisition of features; Dotted lines indicate partially acquired, or missing features.
Adapted based on Levelt et al., (1999).

Important for our discussion, is how the L2 subset within the mental lexicon is
created and activated. If we take the structure of lemmas from the monolingual
production model, we expect lemmas to contain diacritic features which specify the
various syntactic properties of the corresponding concepts. In this way, as languages
are different in their syntactic properties, lemmas are in essence language-specific.
Indeed, this is the stance taken by later adaptations of Levelt’s (1989) model
(Poulisse & Bongaerts, 1994). As seen in Figure 2.1 and 2.2., lemmas contain
language-specification as well semantic and syntactic information regarding the
concepts. The appropriate lemmas are activated based on the concepts and the choice
of language from conceptual level information. This is compatible with findings by
Poulisse and Bongaerts (1994) which suggested morphological encoding is also pre-
determined by the choice of language (see also Myers-Scotton, 1992). The authors
claimed that L2 lemmas would not activate L1 inflectional morphemes. In other
words, L2 learners would not mix L1 inflections with L2 words during L2 speech.
For example, an L1 Dutch speaker of English would not use L1 Dutch morphemes
with L2 English words if the language spoken is intended in the L2. This has
implications for discussions of L2 inflectional errors in 1.2.3.

With respect to L2 phonological encoding, de Bot’s (1992) bilingual production

model assumes L1 and L2 use the same large set of sounds (basic units of speech
production). L2 learners use L1 sounds as much as possible during L2 acquisition,
and novel L2 sounds are only developed if necessary. This is consistent with Flege’s

23
 Chapter 1

(1995) Speech Learning Model where he claimed that L2 learners equate L2

phonemes with L1 phonemes with similar sounds at the very beginning of the
acquisition process, and only develop new sound categories after prolonged L2
exposure.

More relevant to our discussion, is how sequences of speech sounds are prepared
during syllabification for phonetic encoding. Based on the principles of WEAVER
(Word Encoding by Activation and VERification; Roelofs, 1997), speakers construct
phonological representations of words in an incremental fashion where default
canonical forms are ‘resyllabified’ to accommodate the phonological context of
speech. However, there is currently little consensus over how phonological encoding
accounts for sounds from different languages during syllabification, especially when
L1 and L2 have different rules regarding the formation of syllables (but see Roelofs,
2015 for how WEAVER++ can be applied to Mandarin and Japanese). If we assume
that some L1 phonological representations are ‘borrowed’ for L2 acquisition, then is
quite possible that L2 learners would encounter problems when L1 phonological
representations, together with L1 rules for syllabification are used in L2 phonological
contexts, leading to errors in forming the phonological word. For example, an L1
Mandarin learner of English may fail pronounce the word ‘walked’ due to L1
syllabification rules which forbids the phonemes /k/ and /t/ in the word-final position.
This may result in Mandarin learners’ inability to pronounce these consonants
successively at the end of words. In this case, a Mandarin speaker may add additional
vowels at the end of ‘walked’ to facilitate pronunciation of the -ed morpheme, or
omit the allophone [t] (for -ed) altogether.

1.3.3. Sources of L2 inflectional errors

As reviewed in 1.1.2, L2 learners frequently make inflectional errors during L2

production. How could we explain L2 inflectional errors using the psycholinguistic
framework? Through this framework, we can attempt to tease apart the different
possibilities for the loci of L2 inflectional errors.

24
 Chapter 1

The first one concerns representation of L2 conceptual distinctions and selective

information processing during the formation of the preverbal message. If Levelt’s
(1989) assumptions stand regarding selective information processing during
conceptualisation, then the L2 learner might not routinely encode information that is
not grammatically marked in the L1. As a result, even when the preverbal message is
intended to be articulated in the L2, key information (e.g. tense) might be missing for
L2 inflectional production.

The second possibility concerns the representation and activation of diacritic

features at the lemma level for the L2 (e.g. tense, aspect, number etc.; see Figure
2.2.). If L2 lemmas do not contain the diacritic features necessary for inflectional
morphemes (e.g. due to incomplete acquisition), then the lemma would not be able to
provide the correct syntactic structure for the retrieval and production of inflectional
morphemes. Alternatively, diacritic features at the lemma level could be present but
not consistently activated during production.

The third possibility lies with the retrieval of morphological forms. If we assume
diacritic features necessary for inflectional production are present for an L2 learner,
then based on the spreading activation principle (Dell, 1986; outlined in 1.3.1.), it is
still possible that morpheme retrieval might fail due to inappropriate levels of
activation, leading to errors in production of L2 inflectional morphemes.

Another possibility lies within the final stages of phonological / phonetic

encoding preceding articulation. It is possible that L2 learners apply rules of L1
phonology to L2 forms during syllabification. Specifically, if the L1 does not permit
certain phonological structures in the L2, then the L2 learner would have difficulty
generating the appropriate phonological word for subsequent phonetic encoding and
articulation.

Turning to linguistic theories concerning second language production,

researchers have specifically addressed the underlying causes of L2 inflectional
errors. Whilst some have argued that L2 learners have fundamental representational
deficits with regard to inflectional morphology (Hawkins & Chan, 1997; Hawkins &
Liszka, 2003), others lean towards an account based on processing breakdowns
(Haznedar & Schwartz, 1997; Prevost & White, 2000).

25
 Chapter 1

Representational accounts of L2 inflectional errors assume a fundamental deficit

with acquisition of ‘functional categories’ or ‘features’ beyond the critical period
(Hawkins & Chan, 1997). Proponents claim that any L2 features which are absent in
the L1 cannot be acquired after the end of the critical period. Therefore, failures of
L2 inflectional production in L2 learners are caused by the absence of grammatical
features (e.g. tense marking) at an abstract level, resulting in total (L2) inflectional
omission. This is comparable with missing diacritic features at the lemma level
presented in psycholinguistic language production models. Without appropriate
diacritic features in L2 lemmas, the L2 learner would not have the correct
morphological structure for lexical retrieval. However, this account of absolute
omission would not explain ‘optional’ production as often seen in L2 production,
systematically consisting of some inflected forms (which require lemma-level
representations) but not others.

In contrast, processing based accounts of L2 inflectional errors claim that L2

learners can acquire the underlying L2 functional features, but have problems with
realising the corresponding surface inflectional forms (Prevost & White, 2000). In
other words, L2 learners might have knowledge of tense and agreement, but cannot
consistently access the appropriate inflections for production. This is comparable to
problematic retrieval of inflectional morphemes from the mental lexicon.

Examining L1 phonological influences, Goad, White and Steele (2003) claimed

that L1 prosodic structure may fundamentally constrain their ability to attach
inflection morphemes to the prosodic word in the L2. In other words, if L1
phonology follows a structure which does not permit attaching inflectional
morphemes to words as suffixes, then such constructions in the L2 would be difficult
for learners to produce. This could be compared with how phonological encoding
accommodates L1 and L2 phonological rules to produce the phonological word in
psycholinguistic models. Syllabification might be problematic if L1 phonological
rules constrain the learner’s ability to form a wider range of sound combinations in
the L2.

26
 Chapter 1

There are other theories which provide other explanations concerning the cause
of L2 grammatical errors. For example, Hawkins (2007) attributed some cases of
inflectional errors to the complexity of grammatical features (e.g. including multiple
features, such as tense, subject number etc.). Empirical evidence from Chondrogianni
and Marinis (2012) confirmed this by showing production asymmetry between 3SG -
s and past -ed inflectional morphemes, with more production errors found for
contexts requiring 3SG -s (contains subject number and tense features) than past -ed
(contains only tense feature). Lardiere (2008) proposed the Feature Reassembly
Hypothesis where she claimed that L2 grammatical acquisition involves the
reconfiguration of lexical features in the learner’s L1. However, this theory cannot
account for all grammatical errors in production. As White (2009) pointed out,
reorganising existing L1 grammatical features is sometimes not enough, and errors
would still occur if specific L2 grammatical features are absent in the L1 and cannot
be acquired through ‘re-organisation’ alone. In the context of the current thesis, it
does not sufficiently explain how L2 learners acquire L1-absent features. These
accounts will be discussed in more detail Chapter 2 as I expand further on theories
explaining L2 inflectional errors in production.

To summarise, L2 language production involves complex stages of processing

where both representational and processing errors could occur. By reviewing stages
of production in the monolingual and bilingual psycholinguistic production models,
we can begin to systematically examine where errors may occur for L2 learners. This
is complemented by existing second language learning theories regarding L2
acquisition and production errors. Evidently, they share commonalities with respect
to their explanations for L2 inflectional errors even though they take different
perspectives. I will further examine these possibilities and their assumptions in
Chapter 2 with the first set of experiments (Experiments 1, 2 & 3).

1.4. L2 language comprehension: auditory and visual accounts

Language comprehension is a demanding task, which not only requires the

learner to have sufficient explicit knowledge of vocabulary and grammar, but also

27
 Chapter 1

the ability to perceive and integrate linguistic cues. Auditory and visual linguistic
cues demand different processing mechanisms for lexical activation but do not differ
in their end goal, which is to comprehend a message in the form of an utterance or
text. Compared with L1 speakers, L2 learners are faced with the challenging task of
processing and integrating auditory and visual linguistic cues which may not match
the linguistic properties of their L1. At the same time, L2 learners may face
competition from items with similar auditory or visual cues from their L1. In order to
discuss L2 language comprehension, we must start with how basic units of speech
and visual text are perceived and recognised. In 1.4.1, I will briefly discuss
connectionist models of monolingual language comprehension in auditory and visual
modalities. Then, in 1.4.2., I will examine the implication of these models in a
bilingual scenario where I will discuss potential problems L2 learners could
experience during L2 comprehension.

1.4.1. Models of auditory and visual comprehension

When a listener hears a word, he /she must first extract the basic information
from the acoustic speech signal, map them onto larger units of speech (i.e.
phonemes), before matching these speech sounds onto word representations.

Connectionist accounts of speech perception assume interactive activation

between levels of detector units (e.g. TRACE, McClelland & Elman, 1986). For
example, the TRACE model comprises three layers of detectors, each responsible for
a type of speech signal (feature, phoneme and word levels) with competing activation
at each level. Specifically, nodes with the highest level of activation at each level
trigger nodes at the next level. At the same time, existing knowledge regarding the
context and language-specific properties of the message influences word recognition
in a ‘top-down’ fashion. As one set of nodes is activated, competing nodes are
inhibited. For example, when a listener encounters the word ‘boat’, acoustic
properties of speech will activate node detectors at the feature level. Features of
speech (e.g. acuteness, nasality, voicing) will compete to activate the phonemes /b/-
/o/- /a/- /t/. Once activated, based on the sequence information of phonemes, nodes at

28
 Chapter 1

the phoneme level will activate the word ‘boat’ at the word level. Moreover, if the
listener knows that the message is spoken in English, then this top-down information
would facilitate the activation of the English word ‘boat’ from the lexicon, instead of
words with similar acoustic properties in other languages.

In a similar way, when a person encounters written text, visual features of

symbols must map onto larger units of symbols (e.g. letters) before activating word
representations. Earlier accounts of visual word recognition provide explanations
similar to TRACE for activation of words in the written form (e.g. the Interactive
Activation Model, Rumelhart & McClelland, 1982). Instead of detecting features
from acoustic signal, the reader must detect visual features (i.e. parts of letters) at the
feature level in order to activate larger units of symbol (i.e. letters). At the letter
level, positional information is taken into account to activate words. Again, at each
level, the detector nodes compete for activation levels, and the nodes with the highest
activation levels trigger activation at the next level.

These accounts of auditory and visual word comprehension will serve as the
basis for bilingual adaptation of auditory and visual comprehension discussed in the
next section.

1.4.2. Sources of L2 comprehension errors

How do L2 learners understand L2 speech? There are three stages: First, L2

learners must be able to perceive speech sounds correctly and identify the relevant
language. Second, as mentioned in 1.4.1. the L2 learner must be able to perceive and
map speech sounds onto the correct words. Then, at the sentence level, L2 learners
must be able to integrate lexical and grammatical information to understand the
message.

The first issue relates to the categorical perception of phonemes. Relevant

empirical research studies were discussed in 1.1.3, and will not be expanded further
here. The second issue relates to lexical access in bilingual speech perception.
Specifically, how phonological and grammatical properties of the L1 constrain the

29
 Chapter 1

perception of phonemes and how they form meaningful words in the L2. If we use
the architectural framework from McClelland and Elman (1986), then bilingual
models of speech perception must include mechanisms which distinguishes L2 from
L1 input. According to the Bilingual Interactive Model of Lexical Access (BIMOLA,
Grosjean, 1997; Léwy & Grosjean, 2008; Figure 3), feature level nodes are non-
language-specific, but bilinguals have subsets of language-specific nodes at phoneme
and word levels, which accounts for language-specific (at the phoneme level) and
lexical (at the word level) information from L2 speech input (see Figure 2.1).
Moreover, much like the model for monolingual auditory recognition (McClelland &
Elman, 1986), activations between phoneme and word levels are bidirectional. In
other words, just as phonotactic information can facilitate the identification of
phoneme and word subsets, choice of language can also influence the activation of
word subsets in a top-down manner, triggering lower level phoneme subsets, hence
the interactive nature of this model. Similar to the model for monolingual visual
word recognition (Rumelhart & McClelland, 1982), the Bilingual Interactive
Activation model contains visual feature nodes at the lower level instead of acoustic
features, and letter level nodes instead of phoneme nodes at higher levels (Dijkstra &
Van Heuven, 1998). Importantly, the model claims that letter level activation is not
language selective, and language is only determined at the word level.

This is an interesting theoretical point regarding the orthographic nature of the

speaker’s two languages. What if the L2 learner has an L1 with non-alphabetic
orthography (e.g. Mandarin Chinese)? In which case, do L2 learners acquire distinct
sets of ‘letter’ level nodes during L2 comprehension? To put it another way, if letter
recognition is only relevant for the learner’s L2 but not the L1, do we still expect
‘letter’ level nodes not to be language selective? Alternatively, do letter level nodes
implicate a wider range of features for a bilingual? This prompts an interesting
theoretical discussion, but it will not be the focus of this thesis. For the purpose of
our discussion, we assume that top-down activation is strong enough so that only one
set of languages is activated during L2 comprehension, with no interference from the
learner’s L1.

30
 Chapter 1

Figure 3. Bilingual Interactive Model of Lexical Access, with feature, phoneme and word level
representations for L1 and L2. Taken from Léwy and Grosjean (2008).

The last, and most important issue in the context of this thesis relates to how L2
learners integrate linguistic cues from multiple sources within an utterance or text.
Languages differ in their grammatical properties and have different rules governing
the construction of sentences. Similar to how phonotactic knowledge facilitates the
phonological perception and identification of words, knowledge of grammatical
properties also facilitates the understanding of sentences or discourse during real-
time comprehension. In essence, by knowing the grammatical rules of a language,
the listener or reader would also know which cues are important for understanding
messages in that language. Problems arise when L1 and L2 differ in their
grammatical properties, and the L2 requires obligatory agreements between words in

31
 Chapter 1

a given context, which may either different or absent in the learner’s L1 (e.g. the use
of temporal inflections as indicated by temporal adverbials).

Connectionist theories have offered explanations for how L2 learners acquire

linguistic cues (e.g. case marking, inflectional morphology etc.) for L2
comprehension (The Competition Model; Bates & MacWhinney, 1989). The basic
assumption of this model is that learners (of any language) rely on the reliability of
cues to meaning (or form-function mappings) during language acquisition. This
means that language learners acquire the most reliable cues to meaning first, and less
reliable cues later. For example, if inflectional morphology is obligatory for
conveying temporal information, then after exposure and acquisition, learners would
consider inflectional morphemes an effective and reliable cue for understanding the
temporal context of the message. Using the terminology from this theory, inflectional
morphology would have high cue validity in the context of temporal information. On
the other hand, if temporal adverbials (e.g. every day; last week) provide additional,
but not essential temporal information, then they would have low cue validity in this
language.

In order to accurately understand L2 speech or text, L2 learners must learn to

prioritise highly valid linguistic cues in the L2, even if they are not highly valid in
the L1. Problems would potentially arise when L1 cue validity is applied to L2
comprehension. Take the example above, if inflectional morphology is obligatory in
the expression of temporal context (as determined by L2 grammar), then L2 learners
must prioritise information from inflectional morphemes when listening or reading
L2 input. Failure to prioritise this information could result in insensitivity to
grammatical violations concerning inflectional morphology (e.g. failing to detect past
-ed omissions with yesterday as a temporal adverbial).

Usage-based approaches have characterised the processing of L2 linguistic cues

in a similar way. Specifically, under the Associative Learning assumptions, L2
learners have been found to focus on certain aspects of linguistic cues but not others,
depending on the saliency and frequency of cues (i.e. learned attention). An example
given in Ellis and Wulff (2008) is the acquisition of tense inflections alongside
temporal adverbials in French. Their claim was that it is often unnecessary for L2

32
 Chapter 1

French learners to interpret both temporal adverbials and tense inflections for
meaning. Given the low saliency of inflectional markings, they could be considered
redundant in the overall interpretation of the sentence compared to temporal
adverbials during initial acquisition. This is consistent with previous studies
documenting slow acquisition of inflectional markings among L2 learners (see
Bardovi-Harlig, 2000).

Interim summary

In sections 1.3. and 1.4., the basic stages of bilingual language production and
comprehension have been outlined. Given the architecture of these models, I
discussed possible sources of inflectional errors during L2 production, including
conceptualisation errors, representational deficits at the lemma level, and processing
breakdowns during activation of representations and retrieval of inflectional forms,
and also phonological processing for articulation. In L2 comprehension, I discussed
ways in which L2 speech signals and text could be recognised and processed, and
reasons for lack of L2 grammatical mapping and integration at a more abstract level.
These theoretical accounts will serve as the basis for the experiments in Chapters 2, 3
and 4, where I will examine the nature of inflectional errors during L2 production,
and integration of information (from inflectional features) during L2 comprehension.

33
 Chapter 1

1.5. Real-time L2 sentence processing in L2 learners

Successful real-time sentence comprehension requires the listener or reader to

assimilate and integrate each part of linguistic input as it becomes available. This is a
fast-paced event which requires the listener or reader to apply their knowledge of
lexical and morphosyntactic cues in an incremental fashion. Real-time sentence
comprehension could be particularly demanding for non-native L2 learners as they
must not only acquire the relevant L2 metalinguistic knowledge, but also efficiently
apply this knowledge in real time. In the context of the current thesis, existing
literature on how L2 learners process different types of L2 constructions in
comparison to native-L1 speakers will be informative in several ways. Firstly, it
could reveal whether L2 learners could, at least in some circumstances, adopt a
native-like incremental processing strategy during L2 sentence comprehension.
Secondly, existing research could demonstrate to what extent L2 learners’ ability to
parse and interpret sentences is affected by the grammatical properties of their L1.
Thirdly, existing research could reveal to what extent individual cognitive capacity
plays a part in L2 sentence processing. Lastly, past research studies could also reveal
how learners under different task demands could exhibit different processing
behaviour during real-time comprehension. Taking these factors into consideration,
one can begin to hypothesise how L2 learners may learn to process sentences with
temporal markings that are substantially different to their L1, and more importantly,
whether L2 learners could be sensitive to potential mismatches between lexical and
morphosyntactic temporal cues in the linguistic input in real time.

1.5.1. Incremental sentence processing in L2 learners

A major discussion surrounding real-time L2 sentence processing concerns the

notion of incrementality. Incremental parsing allows the listener or reader to interpret
linguistic input as each part of sentence becomes available, instead of interpreting all
parts of a sentence together after the sentence is complete. Past research has
famously shown via the ‘garden-path’ phenomenon that native speakers assimilate
words of a sentence incrementally by interpreting each word as it is encountered

34
 Chapter 1

based on the simplest syntactic structure possible (see Minimal Attachment and Late
Closure, Frazier, 1978) - that is, until the sentence becomes syntactically ambiguous
or semantically implausible, prompting a reanalysis of the sentence for another
interpretation (e.g. the horse raced past the barn fell). It is therefore of interest
whether L2 speakers could also parse L2 sentences incrementally in a native-like
way. Such investigations would provide insight into whether L2 learners could apply
L2 linguistic knowledge in real time by showing processing difficulties when
mismatches occur between lexical and morphosyntactic cues.

Using a self-paced reading paradigm with an online grammaticality judgement

task, Juffs and Harrington (1995) examined how Chinese learners of L2 English
process ‘garden-path’ sentences with pre-posed adjuncts (e.g. After Bill drank the
water proved to be poisoned) and with complement clauses (e.g. Sam warned the
student cheated on the exam) compared to non-garden-path sentences (e.g. After Sam
arrived the guests began to eat). Participants viewed each sentence on a word-by-
word basis and made a grammaticality judgement after each word is revealed. The
findings showed that Chinese learners of English were significantly slower to make
grammaticality judgements (longer reaction times) after viewing the disambiguating
verb (proved / cheated) than with the unambiguous intransitive verb (arrived),
indicating that they interpreted L2 sentences incrementally and encountering
processing difficulties at the disambiguating verb. However, although L2 learners
experience processing difficulties with garden-path sentences like native-L1
speakers, their attempt at reanalysing the sentence is not always successful. As
demonstrated by Juffs and Harrington (1996), Chinese learners of L2 English did not
always respond accurately to comprehension questions following trial sentences,
indicating that they have not successfully recovered from garden-path difficulties and
correctly reinterpreted these sentences. As a result, though there was clear evidence
that L2 learners did experience processing difficulties at the point of disambiguation
in garden-path sentences, they do not behave like native-L1 speakers in resolving
these difficulties (also see Felser et al., 2003). By showing significant processing
difficulties when the initial interpretation became implausible, existing findings
support the idea that L2 learners do in fact apply their knowledge of L2 linguistic
properties incrementally during sentence comprehension. However, non-native-like

35
 Chapter 1

processing was still evident in L2 learners from their overall interpretation of

syntactically ambiguous (or garden-path) sentences.

1.5.2. Extent of L1 effects on L2 sentence processing

The second point relevant to the current thesis relates to how L1 grammatical
properties affect incremental parsing of L2 sentences, and in turn affect interpretation
of these sentences in real-time. The key idea is that if a specific L2 syntactic
construction is shared by the L2 learners’ L1, then these L2 learners should be more
native-like in their parsing of L2 sentences than L2 learners whose L1 do not share
similar constructions.

Again using a self-paced reading task, Marinis, Roberts, Clahsen and Felser
(2005) examined how adult L2 English learners from German, Greek, Chinese and
Japanese L1s process long distance wh-dependencies in L2 English sentences. In
particular, whether they make use of intermediate gaps when processing sentences
with fronted wh-phrases (e.g. The nurse who the doctor argued that the rude patient
had angered is refusing to work late). Though proficient L2 English learners in this
study successfully comprehended these L2 sentences (unlike Juffs & Harrington,
1996), they did not show native-like sensitivity to syntactic cues (i.e. postulate
intermediate gaps) during real-time processing. Critically, L2 English learners from
German and Greek backgrounds did not exhibit native-like reading times despite the
presence of subjacency constraint in their L1, indicating a lack of L1 transfer effects.
Instead, their reading times were similar to that shown by L2 English learners from
Chinese and Japanese backgrounds with no subjacency constraint in their L1.

Similarly, in an earlier study by Williams, Möbius and Kim (2001) examining

the processing of filler-gap dependencies in English wh-questions, L2 English
learners from L1 Korean, Japanese and Chinese backgrounds also behaved similarly
in their online plausibility judgement to garden-path questions (e.g. Which shop did
the criminal kill in the city yesterday evening?) despite differences in wh-
construction in their L1. However, although both native-L1 and L2 learners
postulated gaps at the first position consistent with L2 grammar, native-L1 speakers

36
 Chapter 1

reanalysed the sentence more quickly than L2 learners (as indicated by shorter
reading times in the segment following an implausible verb e.g. kill). Overall, these
findings argue against the notion of L1 transfer effects, and are partially in favour of
fundamental differences between L1 and L2 real-time sentence processing.

The relationship between L1 effects and L1-L2 fundamental difference is far from
clear-cut. Returning to Hopp (2010), where Russian, Dutch and English learners of
German were tested on their offline knowledge and online processing of German
number and case marking, the experiments showed a between-group effect between
L2 learners from different L1s. Specifically, though it was clear that all L2 learners
could apply their offline knowledge of German number and case marking (as
observed through sensitivity to sentences with ungrammatical or dispreferred word
ordering), advanced Russian learners of L2 German (with a morphologically
complex L1) were closer to the native-L1 group than Dutch and English learners in
their reaction times in critical segments. Such evidence has been used to argue in
favour of L1-specific effects.

1.5.3. Cognitive capacity and task demands

The third point of discussion concerns how individual differences in cognitive

capacity affect how L2 learners parse and interpret L2 sentences in real-time. Take
the processing of English filler-gap dependencies in wh-questions, successful real-
time comprehension of such sentences requires L2 learners to store words or
segments of a sentence in working memory to allow for additional syntactic
processing or reanalysis after the initial interpretation becomes implausible. Using
the example from Williams et al. (2001), the implausibility of kill is only evident if
the participant (native-L1 speakers or L2 learners) stores all previous segments of the
sentence (i.e. Which shop did the criminal). In a later study by Williams (2006),
among advanced Korean, Chinese and German learners of L2 English, only those
with high working memory capacity (as measured by a memory probe task) showed
native-like online plausibility judgements after each segment when comprehending
wh-questions (e.g. Which girl / river did the man push a bike into late last night?).

37
 Chapter 1

Similar effect of working memory capacity was also shown in grammaticality

judgement tasks in Chinese speakers of English (Dussias & Piñar, 2010), and across
different L1 backgrounds (McDonald, 2006).

The final point of concern relates to how task demands affect the nativelikeness of
L2 processing. Previous studies mentioned in 1.5.1 and 1.5.2 have used a mixture of
tasks examining online processing (i.e. self-paced reading) and offline knowledge,
some relate to experimental items (e.g. plausibility, grammaticality judgement),
others relate to measures of general L2 proficiency. Some may argue that the
differences found in the degree of nativelikeness and L1 effects were driven by the
nature of the offline task. In other words, it is possible that real-time L2 processing
behaviour could depend on whether L2 learners were asked to specifically monitor
for grammatical violations or read for meaning.

Jackson and colleagues conducted a series of studies examining the effect of task
demands again on the processing of wh-constructions. Using identical stimuli, the
contrasting data from Jackson and Bobb (2009) and Jackson and Dussias (2009)
showed that L2 learners could show native-like processing when explicitly required
to make grammaticality judgements. However, when the experimental task also
probes into the L2 learners’ understanding of experimental items, L2 learners do not
show the same native-like recovery or reanalysis of the sentence. This indicates that
despite showing native-like online processing difficulties initially, L2 speakers do
not always recover from processing difficulties in a native-like manner. Instead, as
noted by Roberts, Mackey and Marsden (2016), L2 learners could be more likely to
carry out delayed parsing decisions, meaning that real-time processing could be less
incremental following the critical ungrammatical / disambiguating segment.

To summarise, studies on real-time sentence processing in L2 learners have

demonstrated that L2 learners can, in some cases, process syntactic constructions
incrementally in a native-like manner. This has been reflected in their sensitivity to
syntactic ambiguities (garden-path sentences). However, the degree to which real-
time L2 sentence processing is affected by the grammatical properties of their L1
remains somewhat debatable. Whilst some have shown that L2 learners from specific
L1 can be more sensitive to specific grammatical features (e.g. number and case

38
 Chapter 1

marking in Hopp, 2010), others have failed to find significant differences between
different L2 groups with significantly different typology in their L1 (e.g. Marinis et
al., 2005; Felser et al., 2003; Papadopoulou & Clahsen, 2006). The nativelikeness of
L2 real-time sentence processing is also likely to be influenced by individual
cognitive capacity, as well as the nature of experimental tasks, though neither is
unrefuted.

1.5.4. Temporal information processing in the L2

Turning to the topic of the current thesis, how can one falsify whether L2
learners of English process L2 inflectional morphology in a native-like way during
real-time sentence comprehension? Presuming the presence of explicit grammatical
knowledge of L2 inflectional morphology (e.g. that regular English verbs should
have a past -ed suffix with a past temporal adverbial), sensitivity to mismatches
between lexical and morphosyntactic should reveal whether L2 learners could apply
their knowledge of inflectional use in real-time. Moreover, one can show the extent
of L1 influence by contrasting real-time sentence processing in L2 English learners
from multiple L1 backgrounds (like Marinis et al., 2005 and Williams et al., 2001)
by testing whether L2 learners with different temporal marking properties in the L1
would respond differently to temporal mismatches in the L2.

Roberts and Liszka (2013) examined whether L2 English learners from French
and German L1 backgrounds would exhibit sensitivity (or processing difficulties) to
temporal mismatches in L2 sentences, and whether their responses would be
significantly different to those in L1-English speakers. L1 French and L1 German
learners of English in this study, despite demonstrating proficient offline L2
grammatical knowledge to temporal markings, responded differently to L1-English
speakers and to each other when encountering past simple and present perfect
temporal mismatches (e.g. When / Since she first started her job, Emma loved / has
loved the work very much). L1 French learners of English experienced significant
processing difficulties (observed via longer reaction times) to temporal markings in
both present perfect and past simple contexts, whilst L1 German learners of English

39
 Chapter 1

did not exhibit such processing difficulties in either context. Unlike previous studies
like Marinis et al. (2005) and Williams et al. (2001) which found no obvious
processing differences between L2 English learners from typologically different L1
backgrounds, Roberts and Liszka found significant differences between how L2
English learners from French and German backgrounds responded to temporal
mismatches in L2 English. According to Roberts and Liszka, sensitivity to past and
present perfect temporal mismatches was contingent on whether their L1 uses overt
aspect markings, thus attributing real-time sensitivity to L2 temporal mismatches to
L1 transfer effects. What was more interesting, was that L1 English speakers in this
study did not exhibit a behaviourally observable processing cost when the temporal
mismatch occurred in a past simple context (e.g. Since* she first started her job,
loved the work very much), contrary to ERP evidence in monolingual English
speakers from an earlier study (Steinhauer & Ullman, 2002). This seemed to indicate
that L1 English participants found a present perfect adverbial (e.g. Since) with a past
simple verb form (e.g. loved) more grammatically acceptable than a past temporal
adverbial (e.g. When) with a present perfect verb form (e.g. has loved). In the
authors’ words, there appeared to be ‘different degrees of ungrammaticality’ for the
two types of temporal mismatches, thus the former case did not cause observable
processing difficulties for L1 English participants.

Findings from Roberts and Liszka (2013) has important implications for the
current thesis. First, if real-time comprehension of temporal information in L2
English is dependent on L2 learners’ ability to incrementally process segments of a
sentence for temporal information, then any mismatch between the temporal
adverbial and the verb form (e.g. where -ed inflection is omitted when the temporal
adverbial unambiguously indicates a past tense context) should cause processing
difficulties in the form of slower reading times at or just after the verb segment.
Second, if L1 grammatical properties distinctly affects the degree of sensitivity to
temporal mismatches in the L2, then one would also expect L2 learners from L1s
with few or no temporal (inflectional) markings to exhibit little or no sensitivity to
such mismatches (e.g. L1 Mandarin learners of L2 English). These possibilities will
be discussed further in Chapter 3.

40
 Chapter 1

Interim summary

In section 1.5., a range of literature on real-time L2 sentence processing has been

reviewed in relation to several themes: the incremental nature of L2 sentence
processing, the extent of L1 effects on L2 sentence processing and the effect of
cognitive capacity and task demands. This provides the backdrop to our discussion to
how L2 learners might in principle process temporal information in sentential
contexts during L2 comprehension. More importantly, this discussion has provided
insight to whether L2 learners can process temporal information from grammatical
features which are absent in their L1. This in turn informs us about how L1 Mandarin
learners of L2 English, with no inflectional morphology in their L1, can process
temporal information from inflectional markings during real-time English sentence
comprehension.

1.6. Linguistic properties of Mandarin Chinese and English

As we examine L1 Mandarin learners of L2 English, it is essential that the basic

linguistic properties of Mandarin Chinese are detailed, particularly, the way
Mandarin conveys temporal information, as well as Mandarin phonology with regard
to temporal markings. In 1.6.1, I give a brief description of how temporal
information is expressed in Mandarin and in English, and where the main
commonalities and differences lie. These distinctions will be relevant in Chapter 2,
where I examine the conceptualisation of temporal information during L2
production. Moreover, I explain the basic properties of Mandarin phonology with
regard to temporal markings and how they contrast with English. These distinctions
will be relevant in Chapter 4, where I examine possible perceptual biases in L1
Mandarin learners of L2 English.

41
 Chapter 1

1.6.1. Expressions of temporality

Mandarin Chinese and English are grammatically distinct languages in the way
they express temporal information (temporality), especially with regard to
expressions of tense (event-external) and aspect (event-internal; Comrie, 1976;
1985). Though there are differing opinions regarding whether Mandarin Chinese is
categorically tense-free (see J. Lin, 2010; T. Lin, 2015 for arguments for and
against), it is commonly accepted that Mandarin does not have overt morphology to
mark temporal information (Li & Thompson, 1981; Smith, 1991). Instead, Mandarin
uses aspectual markers and adverbials which mark temporal contexts to denote
event-internal properties of an event (Smith, 1994; Tang, 2016). In contrast, English
together with many other Indo-European languages use inflectional morphology to
convey tense and aspect information of an event (Booij, 2005).

Mandarin has no overt morphology to indicate tense, and the temporal status of
an event is conveyed mainly through perfective and imperfective aspectual markers
attached to a Mandarin verb (Li & Thompson, 1981). There are two recognised
perfective aspectual markers in Mandarin: le and guo. Le, the most common
perfective aspectual marker, usually indicating that the action denoted by the verb is
complete and still holds at the time of speaking. In the example a) , the perfective
property of go is expressed by placing the aspectual marker le after the verb qu. Le
can also be detached from the verb and come at the end of the sentence, though with
a slightly different interpretation depending on its context. This is sometimes known
as the imperfective le, where the action denoted by the verb may still be ongoing
(Chan, 1980; Li & Thompson, 1981; Chen, 2009). Given the example below, the
action of shui (sleep) in (b) can be interpreted as an event which has begun but not
yet completed. The aspectual marker guo after the verb also gives the verb a
perfective meaning. In example (c), it is unambiguous that the endpoint of the event
no longer holds at the time of speaking. This distinct property is what differentiates
guo from le. As they denote different temporal properties of the event, these two
markers are not mutually exclusive.

42
 Chapter 1

a) ta1 qu4 le0 shang1dian4

she go [PERFECTIVE shop

ASPECT]

She has gone / went to the shop.

b) ta1 shui4 le0

she sleep [IMPERFECTIVE

ASPECT]

She is sleeping.

c) ta1 chi1 guo4 le0

She eat [PERFECTIVE [PERFECTIVE

ASPECT] ASPECT]

She has eaten.

d) ta1 chi1 zhe0 fan4 ne0

She eat [IMPERFECTIVE meal

ASPECT]

She is eating a meal.

e) ta1 zai4 chi1 fan4 ne0

She [IMPERFECTIVE eat meal

ASPECT]

* bold text indicates aspectual markers and corresponding features.

Apart from le which can have an imperfective interpretation, there are two other
imperfective aspect markers in Mandarin, namely zai and zhe. zai usually comes
before a verb and expresses progressive aspect (imperfective). In the example in d),
the act of eating (chi1) is still in progress at the time of speaking. On the other hand,
zhe, attached after the verb, indicates the ongoing state of the situation with a view
on the result (see e). Both imperfective aspectual markers can be used in combination
with temporal adverbials (e.g. mei3tian1, shang4zhou1) to refer to periods of time in

43
 Chapter 1

the present as well as the past in Mandarin. It should be noted that temporal
adverbials can only occur at sentence-initial position in Mandarin.

Compared with Mandarin, which expresses temporality using aspectual markers

without tense markings, English require inflectional morphology to mark tense and
aspectual information. Inflectional morphology by definition creates derivative forms
by conjugating the verb (Booij, 2005). In f), the past tense inflection -ed is attached
as a suffix to the end of the verb to indicate the act of walking is complete at the time
of speech. Aspectual information in English can be much more ambiguous, as it is
not always marked by a distinct morpheme. The habitual aspect is one such example.
In morphological terms, 3rd person singular -s (3SG -s) contains tense information
(present) but also the number of subjects taking part in the action (singular).
However, if 3SG -s is used in conjunction with a temporal adverbial (e.g. every day),
which may occur in a variety of positions in an English sentence, then the temporal
property of the action takes on a habitual aspect (see g). Other aspectual markings
include the progressive -ing, which denotes the ongoing nature of an action, are also
used in conjunction with temporal adverbials. However, they go beyond to the scope
of this thesis and will not be discussed further.

f) Last week the girl walk -ed a mile

[TEMPORAL [PAST]
ADVERBIAL]

g) Every day the girl walk -s a mile

[TEMPORAL [PRESENT]
ADVERBIAL]
[SINGULAR]

* bold text indicates inflectional morphemes and corresponding features

To summarise, Mandarin Chinese conveys temporal information through

aspectual markers (le and guo - perfective; zai and zhe - imperfective) without tense
marking, whereas English conveys both tense and aspectual temporal information
through inflectional morphology. Moreover, both Mandarin and English share the
use of temporal adverbials (e.g. mei3tian1 - every day, shang4zhou1 - last week) in

44
 Chapter 1

temporal expression. Moreover, English marks 3rd person singular with -s in the
present tense.

1.6.2. Phonological properties of morphemes

Mandarin Chinese and English are also distinct in their phonological properties.
Aside from its tonal nature, Mandarin is also much less varied in its phonological
structure compared to English. In this section, I will briefly discuss some key
phonological distinctions between Mandarin and English and how different
phonological features can be formed by Mandarin and English morphemes.

Mandarin is traditionally classified as a tonal language, which marks semantic

distinctions with lexical tones. For example, the sound shu can have four distinct and
unrelated semantic interpretations depending its tone (Table 1).

Table 1.

Examples of tone-based semantic distinctions in Mandarin Chinese with Pinyin

transcriptions and English translations.

Chinese 书 熟 鼠 树
character

Pinyin shu1 / shū shu2 / shú shu3 / shǔ shu4 / shù

Tone flat rising falling and falling

rising

Translation book familiar mouse tree

As seen above, one of the most distinct features of Mandarin is that it consists
mainly of monosyllabic morphemes (Smith, 1991). Generally speaking, single
Mandarin syllables have a CGVX structure (Duanmu, 2000): C (consonant), G
(glide), V (vowel) and X (consonant or extension of long vowel). Although there is
some debate over whether Mandarin consists of consonant clusters at all, it is
generally accepted that multiple consonants rarely occur together in the word-final

45
 Chapter 1

position. In contrast, English allows consonant clusters in a variety of positions

within a morpheme or word (e.g. word-initial: flower; word-final: last; mid-word:
citron).

Table 2.
Pinyin transcriptions and English translations of Mandarin bimorphemic words.

Morpheme A Morpheme B Bimorphemic word (A+B)

Pinyin hua1 / huā pen2 / pén hua1pen2 / huā pén

Translation flower pot flowerpot

Pinyin hua1 / huā bao1 / bāo hua1bao1 / huā bāo

Translation flower bud flower bud

Relevant to our discussion, are the phonological properties of Mandarin and

English in the context of adjacent morphemes: That is, distinct phonological features
when multiple morphemes are placed together in a single word. Similar to English,
monosyllabic morphemes in Mandarin can be placed together to form a bimorphemic
word (see Table 2). Importantly, the Mandarin morphemes would retain their syllabic
properties and would not create new phonological features when placed together in a
new word. For example, the basic phonetic realisation for hua1 (flower) would not
change if it was placed with ‘pen2’ (flowerpot), or with bao2 (flower bud). The same
rule applies to aspectual markers following verbs (e.g. le after qu4 or chi1). This
stands in contrast with English inflectional morphology, which may have different
phonetic realisations depending on its phonological context. Take example f): the
past tense morpheme -ed is realised as [t] in walked, but could be realised as [d] in
yelled and [ɪd] in shouted. Such phonological variability could be problematic for L2
English learners during comprehension, especially if the learner’s first language (e.g.
Mandarin Chinese) does not allow for context-dependent phonological variability for
morphemes. Implications of this issue will be further discussed in Chapter 4.

46
 Chapter 1

Summary of literature review and research questions

Existing research suggests that L2 learners are prone to grammatical errors

during production and comprehension, which might be exacerbated if L2
grammatical features are absent in the learner’s L1. This difficulty is mediated by the
way in which L2 attainment is being measured. Specifically, modalities which
necessarily impose time restrictions on the L2 learner also require more demanding
cognitive processing, resulting in lower performance accuracy than those without
time restrictions. Given these findings, studies which examine L2 production and
comprehension errors across multiple modalities would be valuable, especially for
L2 learners acquiring grammatical features which are absent in the L1.

As previous research studies have demonstrated, L1 Mandarin learners of L2

English frequently exhibit difficulties producing L2 English inflections. However,
systematic investigations controlling for temporal context and production modality
are currently lacking. In Chapter 2, I will present data from three experiments
(Experiment 1, 2 and 3) where L1 Mandarin speakers5 of L2 English produced
inflectional markings in controlled temporal contexts in spoken and written
modalities. The main research questions are: 1) Whether the reoccurrence of
inflectional errors in L2 production is a representational and/or processing problem;
2) Whether complexity of information contained in inflectional markings affects
inflectional accuracy during L2 production; and 3) Whether oral articulation induces
more inflectional errors in spoken than in written production.

As previous research studies have also shown, L1 Mandarin learners of L2

English do not process English subject-verb agreement and morphology in a native-
like way (Jiang, 2007), whilst exhibiting proficient grammatical knowledge (Chen et
al., 2007). It is possible that whilst L2 learners have explicitly learnt the grammatical
rules, they do not have a comprehension mechanism to apply them appropriately in
real-time. In Chapter 3, I present data from two experiments (Experiment 4 and 5)

5
L1 Mandarin learners of L2 English will be referred to as ‘L1 Mandarin speakers’ in the context of L2
production in Chapter 2. For the rest of the thesis, L2 English learners from Mandarin backgrounds will be
referred to as ‘L1 Mandarin learners of L2 English’ in the context of L2 comprehension and in the context of
phonological processing.

47
 Chapter 1

where L1 Mandarin learners of L2 English comprehended L2 English sentences with

grammatical violations (inflectional omissions) in auditory and visual modalities.
The main research questions are: 1) Whether L2 learners can integrate semantic and
syntactic linguistic cues for L2 temporal comprehension (i.e. temporal adverbials and
inflectional morphology); 2) Whether auditory stimuli would make L2
comprehension more difficult for the L2 learner compared with visual stimuli.

Given known phonological constraints imposed by L1 phonological

development, L2 learners may find sounds or sound combinations shared by their L1
to be easier to detect and process, and those which are novel in the L2 to be more
difficult to detect and process. Such perceptual biases may result in comprehension
difficulties. In Chapter 4, I present two additional experiments (Experiments 6 and 7)
examining the effect of phonological features on perception of L2 grammatical
features and the extent of L1 phonological influence on L2 production. Given the
syllabic nature of Mandarin morphemes, L1 Mandarin learners of L2 English may
find syllabic endings more salient in perception compared with consonant clusters
(which are rare in Mandarin). The main research questions are: 1) whether L1
Mandarin learners of L2 English differ from L1 English learners in the perception of
syllabic and consonant cluster inflectional endings; 2) whether saliency of
phonological features affect the processing of temporal information in inflectional
markings; and 3) whether production difficulties in phoneme adjunction are
restricted to inflectional morphemes.

In Chapter 5, I will sum up the key experimental findings presented in this thesis
and discuss their implications for the field. Discussions for Experiments 1 to 3 will
focus on the likely causes affecting inflectional accuracy during L2 production, and
how my findings fit in with psycholinguistic models of language production and
existing production studies in the field. Discussions for Experiment 4 and 5 will
focus on the application of grammatical knowledge in real-time comprehension, and
how comprehension modality affects this process. Discussions for Experiment 6 and
7 will focus on the effect of phonological saliency in speech perception and
comprehension, and the extent of L1 phonological influence on L2 production.
methodological considerations will be evaluated, including both positive aspects and
potential limitations, and how these points will be valuable going forward.

48
 Chapter 2

Chapter 2

Spoken and written production of L2 temporal

inflections in L1 Mandarin speakers of L2 English

Second language (L2) speakers from different native language (L1) backgrounds
have been widely observed to make errors when producing morphological inflections
in their L2. I report three experiments that investigated how such inconsistency might
arise within the language production system, focusing on L2 English speakers whose
L1 does not mark tense grammatically and does not use a morphological system to
indicate temporal properties of events. L1 Mandarin and L1 English (control)
participants produced spoken (Experiments 1 and 2) or written (Experiment 3)
descriptions of events involving different temporal contexts. In all three experiments,
L1 Mandarin participants showed sensitivity to L2 temporal cues when producing
present and past morphemes. These results indicate that L2 speakers may acquire and
process features that do not occur in their L1 but cannot always activate and retrieve
these features accurately during spoken and written production. Critically, L1
Mandarin speakers found the featurally complex inflection (3 rd person singular -s)
more difficult to produce accurately than the featurally less complex inflection (past -
ed), indicating that the complexity of inflectional morphemes also affects accuracy of
production. Finally, given that similar patterns of inflectional errors were found not
only in spoken but also written production, the loci of erroneous inflectional
production could not be solely attributed to articulatory problems.

49
 Chapter 2

2.1. Introduction

Second language (L2) speakers often make errors when producing inflectional
markings in their L2. For example, L2 English speakers frequently omit the past
tense inflection -ed when the grammatical context demands it, e.g., Yesterday the
chef shout at the waiter in the restaurant*. Although there is abundant evidence for
erroneous inflectional production by L2 speakers from different L1 backgrounds,
there is little agreement over the causes of such inconsistencies (Goad, White, &
Steele, 2003; Hawkins & Chan, 1997; Lardiere, 2008; Prévost & White, 2000), and,
in particular, little consideration of how morphological errors might be accounted for
within psycholinguistic models of language production. For instance, do they reflect
L2 speakers’ failure to acquire conceptual distinctions that are absent in their L1,
inability to represent and activate grammatical features, inconsistent retrieval of
inflectional forms, or difficulties in articulating inflectional markings? In this
chapter, I focus on the spoken and written production of English tense inflections
(i.e., 3rd person singular –s and past tense –ed, as in walks and walked) in L2 learners
whose L1 (Mandarin) does not overtly mark for tense morphology, in order to
investigate the locus of erroneous morphological inflections in L2 language
production.

Many previous studies on inflectional production have found that L1 Mandarin

speakers are particularly prone to inflectional errors in L2 English production,
especially in comparison with L2 English speakers with tense marking in their L1s.
In a series of longitudinal studies, Lardiere (1998a; 1998b; 2000; 2003) found that a
native Mandarin-Hokkien speaking adult, Patty, who had been living in the US for
more than 10 years, showed only 5.8% regular past tense marking in her spoken
production even after prolonged L2 immersion. Similarly, a picture-description study
with L2 upper-intermediate to advanced adult speakers of English from L1 Mandarin
backgrounds also showed past inflectional production at chance level or below after
6 months of L2 immersion (Goad et al., 2003). Converging evidence from different
production tasks also revealed that L2 English users from L1 Mandarin backgrounds
have a consistent tendency to omit past tense marking especially compared with L2

50
 Chapter 2

English speakers from other L1 backgrounds (Bayley, 1996; Hawkins & Liszka,
2003).

One important factor which might play a role in L1 Mandarin speakers’ poorer
performance on English inflectional production is differences in the temporal
properties across languages: Whilst English uses a combination of tense (event
external) and aspectual morphemes (event internal) to express temporal information,
Mandarin is a non-inflectional language that does not overtly mark for tense on the
verb and uses aspectual marker with temporal adverbials which mark contexts
(Smith, 1991). In Mandarin, the perfective aspectual marker le is used to indicate the
completed status of events, with additional temporal adverbial marking that the event
is in the past (see h), whereas in English, tense is an obligatory feature and is marked
by an inflection when producing a verb phrase (see i).

h) zuo2 tian1 ta1 kan4 le0 wang3qiu2 bi3sai4

yesterday she watch [PERFECTIVE tennis match

ASPECT]

i) yesterday she watch -ed (a) tennis match

yesterday she watch [PAST] tennis match

[PERFECTIVE
ASPECT]

‘Yesterday she watched a tennis match’

Therefore, for an L1 Mandarin speaker learning English to show high accuracy

in temporal inflectional production, they must not only conceptualise tense
distinctions (e.g., present vs. past) and represent the appropriate grammatical forms
which mark these distinctions (e.g., 3rd person singular -s vs. past –ed), but must also
then use these distinctions and produce the correct inflectional markings in the
appropriate contexts during processing. That is, successful inflectional production
involves factors relating to both conceptualisation, representation and processing of
tense distinctions during inflectional production.

51
 Chapter 2

2.1.1. Theories of L2 inflectional production errors

What factors might underlie L1 Mandarin speakers’ poor inflectional accuracy

in English? Previous research cited within theoretical linguistic frameworks has
proposed several possible sources for L2 inflectional errors, implicating
representational deficits, morphological processing failures, and/or prosodic or
articulatory failures in language production. Given that morphemes can be broken
down into smaller units or features (e.g. number, person, tense etc.), some theories
locate difficulties in inflectional production in specific representational deficits on
these features. For example, L2 speakers are unable to acquire new L2 featural
representations after the critical period if they are absent in the speaker’s L1 (Failed
Functional Feature Hypothesis; Hawkins & Chan, 1997). As Mandarin speakers do
not have tense features, they would not be able to form new functional categories for
tense inflections during L2 acquisition. As a consequence, this theory would likely
predict that L1 Mandarin speakers would never produce correct inflections.
However, another possibility would be optional inflectional production without
necessarily having specific representation of their temporal features.

Other representational accounts proposed that prosodic constraints affect

inflectional production, and specifically that the use of L1 prosodic features in L2
production is solely responsible for L2 inflectional omission (Prosodic Transfer
Hypothesis; Goad et al., 2003). If the speaker’s L1 does not permit the use of certain
prosodic structures (e.g. adjunction of inflection to the phonological word), L2
speakers are likely to fail to process the corresponding inflectional markings. For
instance, the study by Goad and White (2006) showed adjunction to the prosodic
word (e.g. attaching [ɪd] to [ˈʃaʊtɪd]), an phonological operation not permitted in
Mandarin, was more difficult to acquire [drʌŋk] compared with phonological
operations inside the prosodic word (e.g. [drɪŋk] becoming [drʌŋk]). Since -ed
adjunctions are essential to English past tense marking, the inability to perform this
phonological operation due to L1 restrictions would significantly hinder the accurate
production of -ed in required contexts. In another study, a Mandarin-Hokkien
speaker Patty showed consistent difficulty with word-final consonant clusters on
English regular verbs, a pattern that was plausibly linked to L1 prosodic constraints

52
 Chapter 2

and difficulties with articulating specific phonemes (Lardiere, 2003). This hypothesis
was further supported by evidence that omission of -t/-d phonemes in Mandarin
speakers occurred in other non-tense contexts as well (Bayley, 1996; Hawkins &
Liszka, 2003). Given that phonological representations are most strongly implicated
in spoken production, prosodic constraints should result in inflectional errors
primarily in the spoken modality (Goad et al., 2003).

Other accounts postulate that L2 inflectional errors are not the result of
representational deficits, rather inconsistent retrieval of L2 inflectional forms
(Missing Surface Inflection Hypothesis, MSIH; Prevost & White, 2000). This account
is thus more in keeping with existing data of ‘inconsistent production’ instead of
‘absolute omission’. Cross-linguistically, inconsistent inflectional retrieval has been
linked to the complexity of information an inflection contains (Featural Complexity
Theory; Hawkins, 2007): Inflections that contain more complex features are more
difficult for L2 speakers to produce accurately. For example, the featurally complex
3rd person singular -s (3SG -s), which codes for person, subject number and tense
whereas to past -ed which codes only for tense. Empirical research has found that
Turkish-English sequential bilingual children (L2 English) had particularly high error
rates for the featurally complex 3SG -s, compared with past -ed. Critically, although
production was variable, they were sensitive to inflectional omissions as
ungrammatical constructions, indicating intact L2 syntactic representations rather
than deficits in syntactic representations (Chondrogianni & Marinis, 2012).

These linguistic accounts provide plausible proposals for why L2 speakers might
produce inflectional errors, but they are not embedded within psycholinguistic
models of processing. Therefore, these accounts do not elucidate the specific
representational and processing deficits that lead to inflectional errors in L2
speakers’ language production in the first place.

53
 Chapter 2

2.1.2. Morphological processing in language production

Our concern in this study was to consider how L2 speakers’ erroneous

inflectional production can be explained within psycholinguistic models of language
production. Current modular models standardly assume that L1 production involves
stages of constructing a preverbal message (conceptualization); activating lexical
representations, assigning grammatical functions/syntactic structure, retrieving word
forms, activating phonological representation, forming phonological words and
associated phonetic plans (grammatical and phonological encoding); and finally
executing phonetic articulatory gestures (articulation; Bock, 1982; Bock & Levelt,
1994; Garrett, 1975; Levelt, 1989, 2001; Levelt, Roelofs, & Meyer, 1999).
Adaptations of this model, assuming the same basic architecture, have been proposed
for bilingual language production (de Bot, 1992; 2003; de Bot & Schreuder, 1993).
In these adapted models, L1 and L2 share conceptualization and articulatory
processes but have separate subsystems for grammatical encoding and lexical access.

To exemplify the relevant processes, consider an L1 English speaker describing

an event in which a chef is shouting at a waiter in a restaurant. During
conceptualization, the speaker constructs a preverbal message that contains not only
concepts such as chef, waiter, restaurant and shout, but also semantic relations such
as the concept of in, and crucially, temporal properties of the event. This message is
assumed not to be language-specific, but nevertheless encodes only information
strictly relevant to the language of the intended utterance (microplanning; Levelt,
1989). Hence the L1 English speaker would code information about the event that
included tense and aspect.

In the following stage, the speaker activates the relevant lexical representations
(lemmas; e.g. syntactic structure relating to chef / waiter / shout / restaurant) with
the associated diacritic features such as number, tense etc. She also determines
relevant grammatical functions or syntactic relations, e.g., subject number, by
consulting the preverbal message. Activation of features at the lemma level underlies
subsequent morphological processing of the relevant inflections at the form level
(e.g., activation of the perfective aspect and present tense features associated with the
verb lemma, together with third person and singular features associated with the

54
 Chapter 2

subject lemma, underlie subsequent processing of –s). The speaker subsequently

retrieves relevant phonological representations, including phonemes, syllable and
stress information, and carries out syllabification to form the phonological word. The
phonological word then undergoes phonetic encoding, where articulatory gestures
are planned. Finally, during articulation, the speaker executes the relevant phonetic
articulatory gestures to form the sounds for chef, waiter, restaurant etc.

Typically, activation flows smoothly from one stage to another, resulting in

successful production of The chef shouts at the waiter in the restaurant, but on rare
occasions breakdowns in transmitting activation between levels can result in a speech
error, e.g., *The chef shout at the waiter in the restaurant (see Dell, 1986; Dell,
Chang, & Griffin, 1999; Dell, Schwartz, Martin, Saffran, & Gagnon, 1997; Foygel &
Dell, 2000). Note that the nature of activation is not binary, but rather relies on
activation from one level to another reaching a sufficient threshold for transmission
to progress onto the next level.

Within this model, we can identify a range of ways in which inflectional errors
might in principle arise during L1 Mandarin speakers’ production of L2 English.
First, errors might arise from representational or processing deficits during
conceptualization. If L2 speakers are unable to represent conceptual distinctions that
do not exist in their L1, L1 Mandarin speakers would fail to encode event-external
information (in our example, how the act of shouting as a whole relates to the time of
speech) in the preverbal message, because Mandarin does not grammaticalize such
information. As the preverbal message representation drives subsequent linguistic
formulation, and event-external information is critical to determining tense, speakers
would fail to produce (i.e., would always omit) tense inflections. Equally, if L2
speakers are able to represent conceptual distinctions that do not exist in their L1, but
experience difficulty in processing conceptual distinctions that do not exist in their
L1, this would result in a tendency to produce tense inflections inconsistently
(showing optionality; i.e., sometimes correctly but sometimes incorrectly).

Alternatively, errors might arise during formulation. At the lemma level, there
could be a representational deficit for the relevant diacritic features (consistent with
Hawkins and Chan’s account). If L2 speakers can make relevant conceptual

55
 Chapter 2

distinctions but do not have corresponding diacritic feature representations (as

Mandarin does not encode these features), they would not be able to encode the
temporal features necessary for subsequent morphological encoding (in our example,
the present tense feature underlying subsequent processing of –s) and so would omit
inflections. If L2 speakers represent these diacritic features but experience difficulty
in activating them appropriately, this would lead to inconsistent production.

A further deficit at the lemma level might lie in the association

between syntactic functions (who-did-what-to-whom) and properties of
morphological features (e.g. 3SG -s inflection following a third person singular
subject). If L2 speakers do not have knowledge of these associations (i.e., a
representational deficit), they would never activate the appropriate feature
representations. If they had a processing deficit, they might do so inconsistently;
in our example, failure to appropriately assign the subject syntactic function to ‘chef’
and process its number information would result in the inconsistent production of
3SG -s.

At the word form level, L2 speakers might have a processing deficit in

transmitting activation from morphological feature representations and to
corresponding word (inflectional) forms sufficiently for successful retrieval. This
would lead L2 speakers to produce inflections inconsistently (consistent with Prevost
and White’s (2000) Missing Surface Inflection Hypothesis, which claimed
difficulties in realising surface form). In our example, an L2 speaker might have the
conceptual distinction of tense and the relevant diacritic feature representations but
still fail to produce the correct inflectional morphology on some occasions because
she could not effectively activate and retrieve the –3SG –s inflection.

Lastly, there could be representational or processing deficits at the

phonological level. If L2 speakers do not represent specific phonemes or their use is
restricted by L1 phonological constrained operations (i.e. syllabification), they may
fail to activate L2 phoneme patterns or perform phonological operations that blend
sounds together. For example, [ɪd] (-ed) in [ʃaʊtɪd] (shouted) is not a plausible
phoneme combination in Mandarin. So, in order to produce a word like [ʃaʊtɪd]
(shouted), L2 speakers would need both the relevant phonemic representations and

56
 Chapter 2

the ability to assemble them to form the appropriate inflection. In this case, the
phonological operation that creates a consonant cluster like [tɪd] also has
morphological correspondents (i.e. t-ed). Therefore, the absence of such
phonological operations would give rise to omission of specific phonemes in
inflections. This is partially consistent with Goad et al.’s (2003) account on L1
prosodic transfer which claimed that speakers have fundamental difficulties
performing phonological adjunctions which are illegal in the L1 during L2
inflectional production. In the previous example, it is possible that even though the
L2 speaker recognises the temporal context of ‘shouting’, the syntactic structure
which indicates the person who performed the act of shouting and the inflectional
forms past -ed or 3SG -s, generating the phonological structure required for
processing (and producing) shouted could still be difficult if the adjunction of [ɪd] to
[t] is not permitted in the speaker’s L1.

Finally, errors might have an articulatory source: L2 speakers’ articulatory

gestures may be limited to permitted phoneme combinations of their L1, so that they
do not acquire additional articulatory gestures for L2 phonemes. This would give rise
to consistent omission of specific phonemes in the spoken modality (e.g., failure to
articulate 3SG –s in our example, since /ts/ is not a permissible combination in
Mandarin). This would be consistent with Lardiere’s (2003) finding of a discrepancy
between Patty’s written versus oral accuracy in past tense inflection (78% vs. 5.8%).

To summarize, a psycholinguistic model of language production offers several

potential loci for L2 inflectional errors in production: Such errors might in principle
occur because of representational or processing deficits, and these deficits might be
associated with conceptualization, formulation (at the lemma, morphological, and/or
phonological level), and/or articulation.

2.1.3. The current study

To investigate whether L2 inflectional errors in production might arise from

representational versus processing deficits, and to identify the level(s) at which such
deficits might occur, I now report three experiments that investigated the production

57
 Chapter 2

of L2 temporal morphology in adult L1 Mandarin speakers of L2 English (and a

control group of L1 English speakers). Spoken (Experiment 1 & 2) and written
(Experiment 3) responses were elicited using a description paradigm, in which
participants produced (under a time limit) descriptions of action scenes, using
temporal cues (calendar pictures indicating either Present Habitual or Past temporal
contexts), regular verbs (e.g. shout) and pictures of people, objects and locations
(e.g., Every day the chef shouts at the waiter in the restaurant). I analysed
participants’ production of inflections (3SG -s & past -ed) with respect to inflectional
accuracy, inflectional type (3SG -s and past -ed) and inflectional omission.

I investigated two broad possibilities for why L1 Mandarin speakers of L2

English might make errors when producing (temporal) inflectional morphology.
Specifically, it is unclear whether the source of error lies with representational
deficits or processing breakdowns at different stages of language production.

Representational deficits can be implicated at multiple stages of language

production. First, if L2 speakers do not conceptualize information relevant to L2
morphological production when the relevant conceptual distinction does not exist in
their L1, they should not produce the relevant morphological inflections under any
circumstances. This account predicts that L1 Mandarin speakers of L2 English
should show significantly poorer performance in inflectional production (i.e., lower
accuracy) than L1 English speakers across the board, and specifically that they would
fail to produce 3SG -s and past -ed inflections entirely (i.e., absolute omission), in
any temporal context (i.e., whether in a Present Habitual context or a Past context).
Second, if L2 speakers do not have associations between syntactic functions (who-
did-what-to-whom), they would never activate the appropriate diacritic features
under the correct syntactic contexts (e.g., activating the values 3rd and SINGULAR for
a verb lemma’s PERSON and NUMBER diacritic features respectively, following a 3rd
person singular subject). This account does not rule out L2 inflectional production
altogether but predicts that L1 Mandarin speakers of L2 English would produce
inflections randomly without accounting for temporal or syntactic contexts. Third, if
L2 speakers do not have representations for relevant diacritic features at the lemma
level, they should not systematically produce inflections associated with those
features. This account again predicts that L1 Mandarin speakers should show

58
 Chapter 2

significantly poorer performance than L1 English speakers across the board (i.e.,
they would fail to produce both 3SG -s and past -ed inflections in the appropriate
contexts), and that they would do so to the same extent for both inflection types.

Processing breakdowns can also occur at multiple stages during language

production. First, L2 speakers may represent the relevant diacritic features at the
lemma level, but cannot activate and integrate them consistently in relation to the
verb. This account predicts inconsistent inflectional production, but as the number of
diacritic features differs for different inflections, it would further predict
asymmetrical patterns for inflections with different numbers (or complexity) of
features. For example, L1 Mandarin speakers of L2 English should make more errors
for inflections involving both subject number and tense information (i.e., 3SG -s)
than inflections involving only tense information (i.e., past -ed). Second, L2 speakers
may represent the relevant associations between syntactic functions, but cannot
activate them consistently under the correct syntactic contexts. This account predicts
inflectional production that is unsystematic and insensitive to temporal and syntactic
contexts. For example, L1 Mandarin speakers of L2 English should produce 3SG -s
inconsistently, but at the same time, they should not be more likely to produce 3SG -
s in Singular Subject conditions than in Plural Subject conditions. Equally, they
should not be more likely to produce past -ed in Past contexts than in Present
Habitual contexts. Third, L2 speakers, given the appropriate conceptual and lemma
level representations and activations, may still experience difficulties in activating
and retrieving the relevant inflectional forms during morphological encoding.
Specifically, L2 speakers may find inflections which require more than one feature
connection from the lemma level to the morphemic level more difficult to retrieve
consistently than inflections which require a single feature. This account predicts that
L1 Mandarin speakers should show poorer performance than L1 English speakers,
sometimes producing the correct inflection and sometimes omitting it (i.e.,
displaying optionality), but crucially they would do so in a way that was sensitive to
the temporal context. That is, they would be more likely to produce 3SG -s
inflections in Present Habitual contexts than in Past contexts, and more likely to
produce past -ed inflections in Past contexts than in Present Habitual contexts.

59
 Chapter 2

Finally, if articulation difficulties contribute to L2 inflectional errors, L1

Mandarin speakers’ performance should be significantly better in written production
(which does not involve overt articulation) compared with spoken production (which
does involve overt articulation). The specific pattern of performance across the two
modalities would be informative about the extent to which articulation underlies
inflectional errors. If articulation is the primary source of such errors (i.e., speakers
do not have other representational and processing difficulties during earlier stages of
production), L1 Mandarin speakers would produce errors in spoken production but
not in written production. If, however, articulatory difficulties exacerbate other
representational and processing sources of error at earlier stages, then L1 Mandarin
speakers would produce similar patterns of error in both spoken production and
written production but the error rate would be higher in spoken production than in
written production.

2.2. Experiment 1

2.2.1. Methods

Participants

16 native Mandarin (L1 Mandarin) speakers of English aged 19-25 (M= 22.6,
SD=1.3) and 18 monolingual native English (L1 English) speakers aged 21-33 (M=
25.1, SD=3.0) from the University of Edinburgh participated in Experiment 1. The
L1 Mandarin group (i.e., L2 English) consisted of late learners of English who only
had regular exposure to English after the age of five. The monolingual English
control group (L1 English) consisted of native English speakers who were not
exposed to any other languages before the age of five. The L1 Mandarin participants
had achieved an overall score of at least 6.5 on the International English Language
Testing System (IELTS, assessing speaking, listening, reading and writing) within
the last two years, indicating intermediate to advanced L2 proficiency; all L1
Mandarin participants were within 24 months of their first arrival in the UK6.

6 See Appendix A for additional information on Mandarin participant’s language background.

60
 Chapter 2

Materials

For the scene description task, nine transitive experimental verbs with alveolar
consonant endings were chosen, eliciting phonologically salient inflectional endings
in the past temporal context (see Appendix B). In addition, 36 scenes depicting these
transitive actions (four per verb) were created as PNG image files for display on a
1024 x 768 pixel computer screen (see Figure 4 for example). Each scene contained
four clip-art items: a calendar image depicting the temporal context of the action
(every day, yesterday), and three images depicting the entities taking part in the
action (an agent, a patient and an instrument or location). The calendar was placed
top-centre and the three action images were placed below from left to right,
congruent with the direction of reading. Nine additional transitive and intransitive
filler verbs were chosen and 36 additional filler scenes were created (Appendix B).
96 entities (people, objects, animals, location etc.) were used multiple times to create
72 action scenes (Appendix C for the full collection of stimuli). Singular and plural
subjects were counterbalanced across both temporal contexts for each verb. A
vocabulary list and a pictorial legend were also prepared to familiarise participants
with items the scene description task (Appendix D).

The Oxford Placement Test (Allan, 1992) was prepared on paper for participants
to complete by hand (see Appendix E).

Figure 4. Experiment 1: Example of trial image from the scene description task, including a
temporal cue (calendar image) and entities in the action (chef, waiter, restaurant).

61
 Chapter 2

Design

This experiment used a 2 x 2 x 2 mixed design with subject number (singular vs.
plural), temporal context (present habitual vs. past) as within-subject variables, and
group (L1 Mandarin vs. L1 English) as a between-subject variable. The experimental
design was identical in Experiment 2 and 3.

Procedure

At the beginning of the session, all participants provided demographic details.

The Mandarin group also provided information about the history of their L2
acquisition, L2 proficiency, and current L2 usage. Subsequently all participants
completed the scene description task.

Figure 5. Experiment 1: Three-step trial procedure for the scene description task, including
the presentation of fixation (1000ms), target verb (2000ms), and trial image (7000ms).

Before completing scene description task, the experimenter explained the

interpretation of the calendars, i.e., that a multi-coloured calendar represented that

62
 Chapter 2

the action in the scene took place habitually (every day) and a red-yellow calendar
represented that the action in the scene was completed once in the past (yesterday).
Participants were given further examples of the trial procedure on paper, in which
temporal adverbials (every day or yesterday) appeared at the beginning of each
sentence. However, participants were not told explicitly that temporal adverbials
were obligatory in their description. Participants then studied the vocabulary list. If
they did not understand any concepts, the concepts were explained first in English,
and then – if still unclear – in Mandarin.

Participants then completed the scene description task on a computer. Scenes

were presented using E-Prime (Version 2.0; Schneider, Eschman & Zuccolotto,
2002). A headset with microphone was prepared to record participants’ responses.

On each trial, a fixation point was presented for 1000 milliseconds (ms)
followed by the target verb (see Figure 5). The verb was presented on-screen for
2000 ms. This was followed by the action scene, which was presented for 7000 ms.
Participants described the action scene aloud using the given verb and all items on
the screen within the given time; responses were recorded via a microphone. Each
trial was immediately succeeded by the next trial. Participants had five practice trials
before the main experiment began. All participants provided descriptions for all 72
action scenes in two blocks of 36 (18 verbs repeated across singular and plural
subjects, in both Present Habitual and Past temporal contexts), separated by a self-
paced break. Presentation order was randomised for each participant. (Participants
subsequently repeated this procedure with the same 72 items in a different
randomised order, but these data are not discussed further here).
Participant subsequently completed a 100-item English grammar test7.

The experiment lasted approximately 30 minutes, and participants were either

offered cash (£5) or course credit for their participation.

7
The L2 English grammatical proficiency measure (Oxford Placement Test; Allan, 1992) did not
predict production responses across analyses in Experiment 1 (but see Appendix F for a
descriptive summary with analysis).

63
 Chapter 2

Coding and Scoring

All trials were recorded as 7000 ms audio files via E-Prime, and were
transcribed exactly as spoken. Only the first response attempt was coded, as
participants’ initial production response was most comparable across trials; any
corrections were ignored (some correction attempts occurred after time-limit and
therefore were not reliably recorded). 8% of responses with non-target verbs were
excluded. The subject of the sentence was coded for number (singular or plural);
trials where the subject number in participants’ response was incongruent with the
image (e.g. the speaker used ‘duck’ instead of ‘ducks’ for an image of more than one
duck) were excluded. Target verb inflections were coded into three categories: zero
inflection (e.g. wait), past tense -ed (e.g. waited) and 3rd person singular –s (3SG -s,
e.g. waits); five responses containing other verb inflections (e.g. progressive -ing) or
auxiliary verbs (e.g. have been waiting) were excluded. Non-target past -ed
responses in Present Habitual contexts (Past Habitual response) were included as
they were necessary for past -ed likelihood analyses.

Verbs with zero inflection, past tense -ed and 3SG –s responses were then scored
for inflectional accuracy (1 or 0) based on the temporal context and subject number
(see Table 3). For trials scored as incorrect, error type and subject number were
coded as: omission error when an obligatory inflection was omitted, e.g. missing
3SG -s for singular subject in Present Habitual context or commission error when an
incorrect inflection was produced, e.g. 3SG -s in Past contexts. Only omission
responses will be presented from here on.

64
 Chapter 2
Table 3.

Coding and scoring criteria for inflectional production responses in Experiments 1, 2 and 3.

Temporal Subject Verb Inflection Accuracy Error Type

Context Number
Present Singular Past -ed (Shout-ed) 0 Commission
Habitual (the Chef) Shout 3SG -s (Shout-s) 1 --
(Every Day) Zero Infl. (Shout) 0 Omission
Plural Past -ed (Wait-ed) 0 --
(the Ducks) Wait 3SG -s (Wait-s) 0 Commission
Zero infl. (Wait) 1 --
Past Singular Past -ed (Applaud-ed) 1 --
(Yesterday) (the Teacher) Applaud 3SG -s (Applaud-s) 0 Commission
Zero Infl. (Applaud) 0 Omission
Plural Past -ed (Paint-ed) 1 --
(the children) Paint 3SG (Paint-s) 0 Commission
Zero Infl. (Paint) 0 Omission

2.2.2. Results

Outcome variables (response accuracy, responses of different inflectional types,

and of inflectional errors) from Experiment 1 were analysed using logistic mixed
effects regression models (LMEs). a forward model building strategy was used with
a maximal random effects structure (Barr, Levy, Scheepers & Tily, 2013). Predictor
variables (Group, Temporal Context and Subject Number) were contrast-coded
before being included as fixed effects predictors. Participant was included as a
random intercept. Item and Temporal Context were included as random slopes if they
significantly improved model fit. Log-likelihood ratio tests (chi-squared test) were
used to compare alternative logit regression models in order to decide whether the
new model with additional fixed and random effect variables significantly improved
the goodness-of-fit.

Three sets of analyses were carried out on the data from the scene description
task. The first and second set focused on the overall accuracy of inflection depending
on temporal context and number regardless of morpheme (with accurate responses
coded as 1, and inaccurate responses coded as 0) and the likelihood of production for

65
 Chapter 2

each type of morpheme (3SG -s and past -ed). For both sets of analyses, a logistic
mixed effects regression model was built with Group, Temporal Context and Subject
Number as fixed effect predictors. Participant was then included as a random
intercept. Item was included as a random intercept or slope if the log-likelihood chi-
squared model comparison showed it significantly improved the fit of the model.
Separate subgroup analyses were also conducted for L1 Mandarin (L2 English) and
L1 English groups, deducting group as a fixed effects predictor but keeping all other
variables the same.

The third set of analyses focused on responses involving inflectional omissions8.

For these analyses, Bayesian logistic mixed effects models (BLME) were built to
address the problem of partial separations (Rainey, 2016; Zorn, 2005). This was
caused by the missing response category of plural subject omission in the present
habitual temporal context, i.e. participants cannot make omission errors if the
condition requires zero inflection, leading to consistent scores of 0 in this error
category. This in turn led to the maximum likelihood estimate (Wald’s Test) of a
non-Bayesian logistic regression model tending towards infinity for the outcome
variable (See Hauck & Donner, 1977). Consequently, a BLME model was used to
impose a fixed prior to the fixed effect parameters, improving parameter estimates
for inflectional error. Note that although Subject Number was not included as a
predictor for omission error analyses (for reasons stated above), few numbers of
errors across conditions and groups still qualified the use of a Bayesian model. As
error patterns is expected to differ substantially across groups, subgroup analyses
were conducted for L1 Mandarin and L1 English groups. The BLME model
consisted of temporal context as the only fixed effects predictor, and participant as
random intercept and item as random slope. Subject number was not included as
fixed effects predictor due to missing response category. Other instances of model
non-convergence were dealt with using the ‘bobyqa’ algorithm for constrained
optimisation by increasing the number of iterations to 10000.

Our presentation focuses on key main effects and interactions; see Tables 5-8 for
complete inferential statistics for each model.

8
Only omission errors were analysed as the numbers of commission error responses were very low across
conditions.

66
 Chapter 2

Overall Inflectional Accuracy

Response accuracy in each temporal context and subject condition was first
analysed (i.e., 3SG -s responses in the Present Habitual Singular Subject condition;
zero-inflection responses in the Present Habitual Plural Subject conditions; and past -
ed responses in the Past Singular / Plural Subject conditions; Figure 6).

Figure 6. Experiment 1: Average proportion of accurate inflectional responses in Present

Habitual and Past temporal contexts in Scene description task for L1 Mandarin and L1
English groups (N=16;18). Error bars denote +/- 1 SE.

There was a significant main effect of Group. Inflectional accuracy was more
variable across conditions in the L1 Mandarin group than in the L1 English group,
with the L1 Mandarin group producing most accurate responses in the Present
Habitual Plural Subject condition (which did not require any inflection) and fewest
accurate responses in the Present Habitual Singular Subject condition (which
required the 3SG -s inflection; M=0.86 vs. M=0.26; L1 English: M=0.79 vs.
M=0.89). There was a significant three-way interaction between Group, Temporal
Context and Subject Number (Table 5). Subgroup analyses revealed that in the L1
Mandarin group, Temporal Context interacted with Subject Number; in the L1
English group, there was no such interaction.

67
 Chapter 2

Inflectional Type

3rd Person Singular -s (3SG -s) responses

Further analyses were conducted 3SG -s responses in each condition (Figure 7).
Note that a 3SG -s response was a grammatically correct response in the Present
Habitual Singular Subject condition, but an error (i.e., production of an incorrect
inflection [commission error]) in all other conditions.

Figure 7. Experiment 1: Average proportion of 3SG -s inflectional production across

Present Habitual and Past temporal conditions for L1 Mandarin and L1 English groups
(N=16;18). Error bars denote +/- 1 SE.

There was a significant main effect of Temporal Context and of Subject

Number, with a significant three-way interaction between Group, Temporal Context
and Subject Number (Table 6): Although the L1 Mandarin group produced more
3SG -s inflections in the Present Habitual Singular Subject context (M=0.26) than in
other contexts, they did so to a lesser extent than the L1 English group (M=0.89).

Subgroup analyses confirmed the effect of Temporal Context in each group.

They also revealed that in the L1 Mandarin group, there was a significant effect of
Subject Number, with participants significantly more likely to produce 3SG -s
inflections following a singular subject than a plural subject (M=0.16 vs. M=0.08).
But critically, there was not a significant interaction between Subject Number and

68
 Chapter 2

Temporal Context: Participants did not produce significantly more 3SG -s inflections
in the Present Habitual Singular Subject condition than in other conditions.

In the L1 English group, there was a significant effect of Subject Number, with
participants more likely to produce 3SG -s inflections following a singular subject
than a plural subject. There was also a significant interaction between Subject
Number and Temporal Context: Participants produced more 3SG -s inflections in the
Present Habitual Singular Subject condition than in other conditions.

Past -ed responses

Figure 8. Experiment 1: Average proportion of past -ed inflectional production across

Present Habitual and Past temporal contexts for L1 Mandarin and L1 English groups
(N=16;18). Error bars denote +/- 1 SE.

Further analyses examined past -ed responses in each condition (Figure 8). Note
that a past -ed response constituted a grammatically correct response in the Past
conditions, but an error in the Present Habitual conditions. Further analyses
examined past -ed responses in each condition.

There was a significant main effect of Group and of Temporal Context, with a
significant two-way interaction between Group and Temporal Context (Table 7):
Although the L1 Mandarin group produced more past -ed inflections in the Past
conditions (M=0.48) than in the Present Habitual conditions (M=0.04), they did so to
a lesser extent than the L1 English group (M=0.95 vs. M=0.09). Subgroup analyses
revealed that in the L1 Mandarin group, there was a significant effect of Temporal

69
 Chapter 2

Context but no other significant effects, and that in the L1 English group, there was
similarly a significant effect of Temporal Context but no other significant effects.

Inflectional Omission responses

Table 4.

Experiment 1: Number of inflectional omission responses out of all inflectional errors (in
each condition) for L1 Mandarin and L1 English groups across Present Habitual and Past
temporal contexts.
L1 Mandarin L1 English
Present Habitual Singular Subject 63/66 (95%) 3/17 (18%)
Present Habitual Plural Subject 0/11 (0%) 0/30 (0%)
Past Singular Subject 62/68 (91%) 1/10 (10%)
Past Plural Subject 44/47 (94%) 8/11 (73%)
Total 169 / 192 (88%) 12 / 68 (18%)

Inflectional omission responses were analysed across the Present Habitual

Singular Subject and Past Singular / Plural conditions (Table 4). Note that
participants could not make omission errors in the Present Habitual Plural Subject
condition, where a zero inflection would be grammatical (e.g., Every day the chefs
shout). A BLME model was therefore used to analyse the likelihood of inflectional
omission out of all inflectional errors using Group and Temporal Context as
predictors; Subject Number was not included as a predictor due to the missing
response category for the Present Habitual Plural Subject condition.

There was a significant main effect of Group (Table 8): The L1 Mandarin group
was significantly more likely to produce inflection omission responses than the L1
English group. There was also a main effect of Temporal Context: On average, there
were more inflection omission responses in the Past contexts than in the Present
Habitual contexts. However, there was no significant interaction between Group and
Temporal Context, indicating that the effect of Temporal Context did not differ
significantly between the L1 Mandarin and L1 English groups. Subgroup analyses
revealed that there was no significant effect of Temporal Context in the L1 Mandarin
group but there was in the L1 English group.

70
 Chapter 2
Table 5.

Experiments 1, 2 and 3: Logistic mixed-effects statistics for inflectional accuracy for L1 Mandarin and L1 English groups.

Experiment 1 Experiment 2 Experiment 3

(N=16;18) (N=37;36) (N=48;46)
B (SE) p B (SE) p B (SE) p
Main Model
Intercept 1.92 (0.31) <.001 1.55 (0.16) <.001 1.81 (0.14) <.001
Group (L1 Mandarin vs. L1 English) 2.73 (0.60) <.001 2.56 (0.29) <.001 2.07 (0.27) <.001
Temporal Context (Present Habitual vs. Past) 0.38 (0.24) .109 -0.22 (0.17) .191 -0.13 (0.16) .401
Subject Number (Singular vs. Plural) 0.25 (0.24) .307 0.43 (0.17) .010 0.30 (0.16) .060
Group ✕ Temporal Context 2.41 (0.46) <.001 -0.07 (0.26) .780 -1.55 (0.31) <.001
Group ✕ Subject Number -2.99 (0.46) <.001 -0.83 (0.26) .001 -0.24 (0.31) .432
Temporal Context ✕ Subject Number -0.81 (0.46) .082 -1.16 (0.34) <.001 -1.30 (0.32) <.001
Group ✕ Temporal Context ✕ Subject Number 4.03 (0.89) <.001 2.45 (0.52) <.001 -0.35 (0.63) .573
L1 Mandarin
Intercept 0.27 (0.39) .478 0.24 (0.17) .156 0.77 (0.16) <.001
Temporal Context -1.05 (0.30) <.001 -0.18 (0.19) .331 0.67 (0.21) <.001
Subject Number 2.03 (0.29) <.001 0.84 (0.19) <.001 0.42 (0.21) .043
Temporal Context ✕ Subject Number -3.20 (0.58) <.001 2.38 (0.38) <.001 -1.05 (0.42) .012
L1 English
Intercept 3.06 (0.47) <.001 2.91 (0.29) <.001 3.03 (0.30) <.001
Temporal Context 1.36 (0.35) <.001 -0.28 (0.23) .211 -0.94 (0.27) <.001
Subject Number -0.96 (0.36) .006 0.04 (0.22) .873 0.17 (0.27) .544
Temporal Context ✕ Subject Number 0.80 (0.67) .233 0.05 (0.45) .918 -1.52 (0.55) .006

71
 Chapter 2
Table 6.
Experiments 1, 2 and 3: Logistic mixed-effects statistics for 3SG -s responses for L1 Mandarin and L1 English groups.

Experiment 1 Experiment 2 Experiment 3

(N=16;18) (N=37;36) (N=48;46)
B (SE) p B (SE) p B (SE) p
Main Model
Intercept -2.89 (0.34) <.001 -2.70 (0.36) <.001 -2.82 (0.48) <.001
Group (L1 Mandarin vs. L1 English) 0.05 (0.55) .920 1.06 (0.32) .001 0.81 (0.28) .004
Temporal Context (Present Habitual vs. Past) -4.09 (0.55) <.001 -2.39 (0.63) <.001 -2.65 (0.87) .002
Subject Number (Singular vs. Plural) -1.75 (0.56) .002 - - - -
Group ✕ Temporal Context -3.76 (0.87) <.001 -2.07 (0.34) <.001 -1.90 (0.39) <.001
Group ✕ Subject Number -1.35 (0.91) .135 - - - -
Temporal Context ✕ Subject Number 3.27 (1.08) .002 - - - -
Group ✕ Temporal Context ✕ Subject Number 4.19 (1.74) .016 - - - -
L1 Mandarin
Intercept -2.76 (0.36) <.001 -2.61 (0.28) <.001 -2.49 (0.24) <.001
Temporal Context -1.77 (0.44) <.001 -1.80 (0.26) <.001 -2.49 (0.37) <.001
Subject Number -0.88 (0.45) .040 -1.15 (0.26) <.001 -2.22 (0.38) <.001
Temporal Context ✕ Subject Number 0.60 (0.87) .486 1.44 (0.51) .005 0.53 (0.76) .483
L1 English
Intercept -3.03 (0.51) <.001 -3.57 (0.80) <.001 -1.69 (0.13) <.001
Temporal Context -5.70 (0.88) <.001 -5.04 (1.42) <.001 -3.11 (0.26) <.001
Subject Number -2.43 (0.88) .006 - - - -
Temporal Context ✕ Subject Number 5.29 (1.70) .002 - - - -

72
 Chapter 2
Table 7.

Experiments 1, 2 and 3: Logistic mixed-effects statistics for past -ed responses for L1 Mandarin and L1 English groups.
Experiment 1 Experiment 2 Experiment 3
(N=16;18) (N=37;36) (N=48;46)
B (SE) p B (SE) p B (SE) p

Main Model
Intercept -0.32 (0.30) .297 -0.89 (0.27) .001 -0.49 (0.18) .008
Group (L1 Mandarin vs. L1 English) 2.76 (0.60) <.001 1.24 (0.52) .016 -0.12 (0.36) .739
Temporal Context (Present Habitual vs. Past) 5.64 (0.39) <.001 5.48 (0.31) <.001 5.05 (0.25) <.001
Subject Number (Singular vs. Plural) -0.35 (0.34) .301 0.00 (0.23) .985 -0.24 (0.18) .174
Group ✕ Temporal Context 3.42 (0.73) <.001 4.75 (0.50) <.001 3.26 (0.45) <.001
Group ✕ Subject Number -1.11 (0.65) .089 0.42 (0.34) .219 -0.53 (0.35) .132
Temporal Context ✕ Subject Number -0.76 (0.67) .253 -0.37 (0.46) .419 -0.34 (0.36) .333
Group ✕ Temporal Context ✕ Subject Number -2.05 (1.31) .117 0.07 (0.69) .914 0.13 (0.71) .851
L1 Mandarin
Intercept -2.05 (0.50) <.001 -1.51 (0.35) <.001 -0.43 (0.26) .097
Temporal Context 3.64 (0.50) <.001 3.08 (0.27) <.001 3.53 (0.30) <.001
Subject Number 0.32 (0.44) .467 -0.20 (0.26) .431 0.07 (0.26) .779
Temporal Context ✕ Subject Number 0.49 (0.93) .599 -0.43 (0.51) .400 -0.40 (0.51) .438
L1 English
Intercept 0.78 (0.36) .028 -0.29 (0.39) .450 -0.55 (0.27) .046
Temporal Context 6.80 (0.56) <.001 7.71 (0.51) <.001 6.66 (0.46) <.001
Subject Number -0.80 (0.47) .094 0.22 (0.29) .442 -0.50 (0.30) .091
Temporal Context ✕ Subject Number -1.57 (0.92) .087 -0.35 (0.58) .546 -0.28 (0.60) .645

73
 Chapter 2
Table 8.
Experiments 1, 2 and 3: Bayesian logistic mixed-effects statistics for inflectional omission for L1 Mandarin and L1 English groups.

Experiment 1 Experiment 2 Experiment 3

(N=16;18) (N=37;36) (N=48;46)
B (SE) p B (SE) p B (SE) p
Main Model
Intercept 1.28 (0.43) .003 0.59 (0.40) .145 0.03 (0.62) .958
Group (L1 Mandarin vs. L1 English) -2.64 (0.67) <.001 -1.61 (0.43) <.001 -0.63 (0.59) .285
Temporal Context (Present Habitual vs. Past) 1.44 (0.61) .019 -2.04 (0.63) .001 1.52 (0.87) .078
Subject Number (Singular vs. Plural) - - - - - -
Group ✕ Temporal Context 0.12 (0.15) .868 -0.08 (0.61) .900 0.23 (0.75) .762
Group ✕ Subject Number - - - - - -
Temporal Context ✕ Subject Number - - - - - -
Group ✕ Temporal Context ✕ Subject Number - - - - - -
L1 Mandarin
Intercept 2.34 (0.57) <.001 0.94 (0.41) .022 0.12 (0.58) .842
Temporal Context -1.12 (0.79) .157 -2.02 (0.62) .001 1.50 (0.80) .062
Subject Number - - - - - -
Temporal Context ✕ Subject Number - - - - - -
L1 English
Intercept -1.01 (0.54) .061 -0.76 (0.76) .315 -0.01 (0.78) .993
Temporal Context -1.38 (0.64) .030 1.13 (0.93) .228 -0.97 (0.97) .313
Subject Number - - - - - -
Temporal Context ✕ Subject Number - - - - - -

74
 Chapter 2

2.2.3. Interim discussion

Experiment 1 showed that although L1 Mandarin speakers made errors when

producing temporal inflections, with an overall accuracy rate of only 52% across
conditions, they nevertheless showed sensitivity to temporal context. That is, they
were more likely to produce 3SG –s and -ed inflections in a temporal context that
was appropriate for those inflections than in a temporal context that was
inappropriate for those inflections. However, their performance was not uniform
across temporal inflections. Notably, they showed particularly low accuracy in the
Present Habitual Singular Subject condition (requiring 3SG -s), relative to L1
English speakers’ performance in the same condition, and relative to their own
performance in the Past Tense conditions (requiring past –ed). These results suggest
that L1 Mandarin speakers are able to conceptualise and linguistically encode
relevant tense distinctions, but that they are not able to produce them consistently,
with one inflection type being more susceptible to error than another.

In Experiment 2, I sought to replicate these findings, with a larger sample and a

more robust experimental paradigm. In Experiment 1, which used speeded
presentation, participants mis-recalled the verb on 8% of trials. Consequently,
Experiment 2 used self-paced verb presentation to increase the proportion of valid
responses.

75
 Chapter 2

2.3. Experiment 2

2.3.1. Methods

Participants

37 L1 Mandarin speakers of L2 English aged 20-29 (M=23.4; SD=1.8) and 36

L1 English speakers aged 19-46 (M=24.3; SD=5.5) took part in Experiment 2. In
addition to the participant recruitment criteria for the L1 Mandarin group (L2
English) in Experiment 1, Mandarin participants were also required to have at least a
score of 5.5 on the spoken component of the IELTS exam. Recruitment criteria for
the L1 English group was identical to those in Experiment 1.

Materials

I used the images, verbs, vocabulary list and legend aid from Experiment 1, with
minor adjustments to remove ambiguity in some items. Each combination of
experimental verb and scene was presented only once (hence, the total number of
trials was halved to 72 trials). Trial presentation was self-paced to allow participants
more time to remember the verb.

Design

The experimental design was identical to Experiment 1.

76
 Chapter 2

Procedure

Figure 9. Experiment 2: Three-step self-paced trial procedure for the scene description task,
including presentations of fixation (1000ms), target verb (self-paced) and target image
(7000ms).

The experimenter followed the same protocol as Experiment 1 but emphasised

the self-paced element of the scene description task (Figure 9). Before the start of the
experiment, participants were reminded that they should remember the verb before
viewing the action scene, use the objects in the scene from top to bottom and left to
right, and avoid using auxiliary verbs in descriptions.

The trial procedure was identical to Experiment 1, except that participants

controlled the progress of each trial by pressing the [SPACE BAR] after reading the
target verb. Participants had five practice trials before commencing the 72
experimental trials. Participants were paid £5 in cash or given course credit for their
time.

Coding and Scoring

The coding and scoring procedures for the scene description task were identical
to Experiment 1.

77
 Chapter 2

2.3.2. Results

Analyses were conducted as in Experiment 1, except where otherwise stated.

Overall Inflectional Accuracy

Figure 10. Experiment 2: Average proportion of accurate inflectional responses in Present

Habitual and Past temporal contexts for scene description task across L1 Mandarin and L1
English groups (N=37;36). Error bars denote +/- 1 SE.

There was a significant main effect of Group and of Temporal Context, with a
significant three-way interaction between Group, Temporal Context and Subject
Number (Table 5). Inflectional accuracy was more variable across conditions in the
L1 Mandarin group than in the L1 English group, with the L1 Mandarin group
producing most accurate responses in the Present Habitual Plural Subject condition
and fewest accurate responses in the Present Habitual Singular Subject condition
(M= 0.77 vs. M=0.37; L1 English: M=0.92 vs. M=0.92; Figure 10).

Subgroup analyses revealed that in the L1 Mandarin group, there was a

significant interaction between Temporal Context and Subject Number; in the L1
English group, there was no such interaction.

78
 Chapter 2

Inflectional Type

3rd person singular -s (3SG -s) responses

Group and Temporal Context were used as fixed effects predictors for the 3SG -
s analysis BLME model. Subject Number was dropped due to a missing category
problem in the L1 English group (no response for Past Plural Subject condition;
Figure 12).

Figure 11. Experiment 2: Average proportion of 3SG -s inflectional production across

Present Habitual and Past temporal contexts for the scene description task across L1
Mandarin and L1 English groups (N=37;36). Error bars denote +/- 1 SE.

There was a significant main effect of Group and of Subject Number, with a
significant interaction between Group and Temporal Context (Table 6): Although the
L1 Mandarin group produced more 3SG -s inflections in the Present Habitual
conditions (M=0.24) than in the Past conditions (M=0.06), they did so to a lesser
extent than the L1 English group (M=0.45 vs. M=0.05). Subgroup analyses revealed
that in the L1 Mandarin group, Subject Number was a significant predictor, with
participants being more likely to produce 3SG -s inflections following a singular
subject than a plural subject. In the L1 English group, Subject Number was also a
significant predictor.

79
 Chapter 2

Past –ed responses

Figure 12. Experiment 2: Average proportion of past -ed inflectional production across
Present Habitual and Past temporal contexts for the scene description task across L1
Mandarin and L1 English groups (N=48;46). Error bars denote +/- 1 SE.

There was a significant main effect of Group and of Temporal Context, with a
significant two-way interaction between Group and Temporal Context (Table 7):
Although the L1 Mandarin group produced more past -ed inflections in the Past
conditions (M=0.54) than in the Present Habitual conditions (M=0.12), they did so to
a lesser extent than the L1 English group (M=0.90 vs. M=0.05; Figure 12). Subgroup
analyses revealed that in the L1 Mandarin group, there was a significant effect of
Temporal Context but no other significant effects; likewise, in the L1 English group,
there was a significant effect of Temporal Context but no other significant effects.

80
 Chapter 2

Inflectional Omission responses

Table 9.
Experiment 2: Number of inflectional omission responses out of all inflectional errors (in
each condition) for L1 Mandarin and L1 English groups across Present Habitual and Past
temporal contexts.

L1 Mandarin L1 English
Present Habitual Singular Subject 151/183 (83%) 11/23 (48%)
Present Habitual Plural Subject 0/55 (0%) 0/25 (0%)
Past Singular Subject 103/123 (84%) 3/28 (11%)
Past Plural Subject 129/142 (91%) 30/30 (100%)
Total 383/503 (76%) 44/106 (42%)

There was a significant main effect of Group (Table 6): The L1 Mandarin group
was significantly more likely to produce inflection omission responses than the L1
English group (Table 9). There was no main effect of Temporal Context: Participants
across L1 Mandarin and L1 English groups were no more likely to make omission
errors in the Present Habitual contexts than in the Past contexts. There was no
significant interaction between Group and Temporal Context, indicating that the
effect of Temporal Context did not differ significantly between the L1 Mandarin and
L1 English groups. Subgroup analyses revealed that there was no significant effect of
Temporal Context in neither the L1 Mandarin nor the L1 English group.

2.3.3. Interim discussion

Experiment 2 replicated the finding that although L1 Mandarin participants

made errors when producing temporal inflections, they nevertheless showed
sensitivity to temporal context. As in Experiment 1, they showed particularly low
accuracy in the Present Habitual Singular Subject condition (requiring 3SG -s),
relative to L1 English speakers’ performance in the same condition, and relative to
their own performance in the Past Tense conditions (requiring –ed).

These results provide further evidence that L1 Mandarin speakers are able to
conceptualise and linguistically encode relevant tense distinctions, but are not able to

81
 Chapter 2

produce them consistently. They also replicate the pattern whereby the (featurally
complex) 3SG –s inflection is more susceptible to error than the (featurally less
complex) Past –ed inflection. However, whether this difference was due to
inconsistent retrieval of inflectional forms or errors in oral articulation could not be
determined.

In Experiment 3, I therefore examined whether these patterns of inflectional

error would remain when participants did not orally articulate their responses. To do
this, I used the same task as Experiment 2, but asked participants to produce typed
responses on a computer keyboard instead. If L1 Mandarin participants’ inflectional
errors in Experiments 1 and 2 resulted from oral articulatory failures alone, we
should see significantly higher inflectional accuracy in Experiment 3 compared to
Experiment 2. On the other hand, if retrieval failures accounted for the inflectional
errors previously observed, similar patterns of inflectional production should remain
even with the articulatory component of the task removed.

82
 Chapter 2

2.4. Experiment 3

2.4.1. Methods

Participants

48 L1 Mandarin speakers of L2 English aged 18-31 (M=22.88; SD=2.26) and 46

monolingual L1 English speakers aged 17-20 (M=18.20; SD=0.74) participated in
Experiment 3. Participants were recruited based on identical criteria as Experiment 2
with no additional requirements.

Materials

24 experimental items and 24 filler items were used from Experiment 2

(retaining 12 of the 18 verbs: six experimental + six filler verbs) in order to shorten
the experimental session (as participants were considerably slower to produce typed
than spoken responses). The trial structure, experimental legend and illustrative trial
examples were identical to Experiment 2. Individual images of animal, place, food,
activity etc. were labelled with their corresponding names (e.g. dog, park etc.) were
created as JPGs as part of a vocabulary training session (Appendix D).

Procedure

Prior to the formal experimental session, participants passively viewed images

with their corresponding names as part of a vocabulary training session in order to
familiarise themselves with the trial images. The instructions for the scene
description task were identical to Experiment 2 except that participants were told that
during each action scene presentation, they must type out their descriptions on a
keyboard, with no option to edit their responses, i.e., written scene description task.
The duration of each action scene was increased to 15000 ms to reflect the response
modality. Afterwards, participants were paid £5 for their time.

83
 Chapter 2

Coding and Scoring

The coding and scoring procedures for the written scene description task were
identical to Experiment 1 and 2.

2.4.2. Results

Overall Inflectional Accuracy

Figure 13. Experiment 3: Average proportion of accurate inflectional responses in Present

Habitual and Past temporal contexts for the scene description task across L1 Mandarin and
L1 English groups (N=48;46). Error bars denote +/- 1 SE.

There was a significant main effect of Group, with a significant interaction

between Group and Temporal Context (Table 5). Inflectional accuracy was more
variable across temporal conditions in the L1 Mandarin group than in the L1 English
group, with the L1 Mandarin group showing a greater disparity between performance
in the Past and Present Habitual conditions than the L1 English group (L1 Mandarin
M= 0.72 vs. M=0.60; L1 English: M=0.88 vs. M= 0.93; Figure 13). Notably,
however, there was not a significant three-way interaction between Group, Temporal
Context, and Subject Number.

84
 Chapter 2

Subgroup analyses revealed that in the L1 Mandarin group, there were

significant effects of Temporal Context and Subject Number, and a significant
interaction between Temporal Context and Subject Number. In the L1 English group,
there was a significant effect of Temporal Context and a significant interaction
between Temporal Context and Subject Number interaction; however, there was not
a significant effect of Subject Number,

Inflectional Type

3rd Person Singular -s responses

Figure 14. Experiment 3: Average proportional production of 3SG -s inflection across

Present Habitual and Past temporal contexts for the scene description task across L1
Mandarin and L1 English groups. (N=48;46) Error bars denote +/- 1 SE.

There was a significant main effect of Group and of Temporal Context, with a
significant interaction between Group and Temporal Context (Table 6): Although the
L1 Mandarin group produced more 3SG -s inflections in the Present Habitual
conditions (M=0.30) than in the Past conditions (M=0.05), they did so to a lesser
extent than the L1 English group (M=0.46 vs. M=0.04; Figure 14).

Subgroup analyses revealed that in the L1 Mandarin group, there was a

significant effect of Subject Number, with participants more likely to produce 3SG -s
inflections following a singular subject than a plural subject; there was also a

85
 Chapter 2

significant interaction between Temporal Context and Subject Number, with

participants most likely to produce 3SG -s inflections in the Present Habitual
Singular Subject condition. In the L1 English group, there was a significant effect of
Temporal Context.

Past –ed responses

Figure 15. Experiment 3: Average proportional production of past -ed inflection across
Present Habitual and Past temporal contexts for the scene description task across L1
Mandarin and L1 English groups. (N=48;46) Error bars denote +/- 1 SE.

There was a significant main effect of Temporal Context, with a significant two-
way interaction between Group and Temporal Context (Table 7): Although the L1
Mandarin group produced more past -ed inflections in the Past conditions (M=0.72)
than in the Present Habitual conditions (M=0.19), they did so to a lesser extent than
the L1 English group (M=0.88 vs. M=0.05; Figure 15). Subgroup analyses revealed
that in the L1 Mandarin group, there was a significant effect of Temporal Context but
no other significant effects; likewise, in the L1 English group, there was a significant
effect of Temporal Context but no other significant effects.

86
 Chapter 2

Inflectional Omission responses

Table 10.
Experiment 3: Number of inflectional omission responses out of all inflectional errors (in
each condition) for L1 Mandarin and L1 English groups across Present Habitual and Past
temporal contexts.
L1 Mandarin L1 English
Present Habitual Singular Subject 83/124 (67%) 7/19 (37%)
Present Habitual Plural Subject 0/53 (0%) 0/8 (0%)
Past Singular Subject 46/67 (69%) 6/23 (26%)
Past Plural Subject 63/66 (95%) 29/29 (100%)
Total 192/410 (47%) 42/79 (53%)

There were no significant effects of Group but there was a marginal effect of
Temporal Context in the main analysis (Table 8). There was a marginal effect of
Temporal Context in the L1 Mandarin group but not the L1 English group in the
subgroup analysis (Table 10).

Between Experiment Comparisons

Three sets of analyses were conducted concerning the effect of production

modality (spoken vs. written) on inflectional accuracy, inflectional type (3SG -s and
past -ed), and inflectional omission (including existing predictors Group, Temporal
Context and Subject Number). Data from Experiment 2 and Experiment 3 were
combined, including only verbs which were common across the two experiments
(verbs which were used in Experiment 2 but not Experiment 3 were excluded). For
full statistics, see Tables 11-14.

There was no significant main effect of Modality on inflectional accuracy

overall. Participants overall were not more likely to produce an accurate response in
the written modality (M=0.79) compared with the spoken modality (M=0.73).
Interestingly, there was a two-way interaction between Group and Subject Number
irrespective of Modality. There were greater differences between singular and plural
subjects in the L1 Mandarin group than the L1 English group (L1 Mandarin: M=0.57

87
 Chapter 2

vs. 0.67; L1 English: M= 0.91 vs. 0.91). There was a three-way interaction between
Group, Temporal Context and Modality and a four-way interaction also including
Subject Number (Table 11). Subgroup analyses reveal a similar picture. Different
from within-experiment analyses previously, there were significant effects of
Temporal Context for both L1 Mandarin and L1 English groups. However, two
groups show different effects of temporal context. Whilst L1 Mandarin speakers
were less likely to produce accurate inflections in the Present Habitual than the Past
Context irrespective of production modality (M = 0.59 vs. M = 0.63), L1 English
were more likely to produce accurate responses in the Present Habitual Context than
the Past Context (L1 English: M= 0.93 vs. M = 0.89). Critically, a significant main
effect of Modality was not found in neither the L1 Mandarin nor the L1 English
group. Interestingly, whilst there was a two-way interaction between Temporal
Context and Modality, and a three-way interaction between Temporal Context,
Subject Number and Modality for the L1 Mandarin group, there were no such
interactions for the L1 English group.

For 3SG -s production, there was no significant main effect of Modality overall.
Participants were not more likely to produce 3SG -s in the written modality
(M=0.21) compared with the spoken modality (M=0.21). Modality did not interact
with any other predictors (Group, Temporal Context and Subject Number). Subgroup
analyses did not reveal any significant effects of Modality, nor any interactions
(Table 12).

For past -ed production, there was no significant main effect of Modality.
Similar to 3SG -s, participants were not more likely to produce past -ed in the written
modality (M=0.46) compared with the spoken modality (M=0.40). The interaction
between Group and Modality was close to significance (Table 13). Subgroup
analyses revealed a significant effect of Modality for the L1 Mandarin group for past
-ed production, but not for the L1 English group. No other interactions were found
involving Modality in either group.

For inflectional omissions, there was a marginal main effect of Modality overall
(Table 14): Numerically, participants omitted fewer inflections in the written
modality than the spoken modality (Tables 5 and 11). There was also a marginal

88
 Chapter 2

interaction between Group and Modality. Subgroup analyses revealed a significant

main effect of Modality on inflectional omission for the L1 Mandarin group, but not
the L1 English Group: L1 Mandarin participants were more likely to make omission
errors in the spoken modality than in the written modality, but the L1 English
participants made very few errors overall and were close to ceiling levels in terms of
accuracy in both modalities.

89
 Chapter 2
Table 11.
Experiments 2 and 3: Between-experiment analyses on the effect of production modality on inflectional accuracy.

B (SE) p
Main Model
Intercept 1.73 (0.11) < .001
Group (L1 Mandarin vs. L1 English) -2.27 (0.21) <.001
Temporal Context (Present Habitual vs Past) 0.23 (0.11) .039
Subject Number (Singular vs Plural) 0.33 (0.11) .004
Modality (Spoken vs Written) 0.24 (0.21) .246
Group ✕ Temporal Context -1.09 (0.23) <.001
Group ✕ Subject Number -0.44 (0.23) .048
Group ✕ Modality -0.40 (0.41) .324
Temporal Context ✕ Subject Number -1.22 (0.23) <.001
Temporal Context ✕ Modality 0.18 (0.23) .422
Subject Number ✕ Modality -0.07 (0.23) .749
Group ✕ Temporal Context ✕ Subject Number 0.74 (0.46) .104
Group ✕ Temporal Context ✕ Modality -1.06 (0.46) .020
Group ✕ Subject Number ✕ Modality 0.43 (0.46) .341
Temporal Context ✕ Modality ✕ Subject Number -0.20 (0.46) .652
Group ✕ Temporal Context ✕ Subject Number ✕ Modality -2.43 (0.92) .008

90
 Chapter 2
(Table 11 continued)

L1 Mandarin
Intercept 0.57 (0.11) <.001
Temporal Context 0.31 (0.12) .008
Subject Number 0.54 (0.12) <.001
Modality 0.43 (0.23) .056
Temporal Context ✕ Subject Number -1.56 (0.23) <.001
Temporal Context ✕ Modality 0.71 (0.23) .002
Subject Number ✕ Modality -0.28 (0.23) .226
Temporal Context ✕ Subject Number ✕ Modality 1.01 (0.47) .031
L1 English
Intercept 3.03 (0.22) <.001
Temporal Context -0.80 (0.20) <.001
Subject Number 0.11 (0.20) .582
Modality 0.05 (0.39) .901
Temporal Context ✕ Subject Number -0.87 (0.40) .029
Temporal Context ✕ Modality -0.35 (0.40) .380
Subject Number ✕ Modality 0.13 (0.40) .752
Temporal Context ✕ Subject Number ✕ Modality -1.45 (0.80) .070

91
 Chapter 2
Table 12.

Experiments 2 and 3: Between-experiment analyses on the effect of production modality on 3SG -s production.

B (SE) p
Main Model
Intercept -1.83 (0.08) <.001
Group (L1 Mandarin vs. L1 English) 0.24 0(.15) .122
Temporal Context (Present Habitual vs. Past) -2.64 (0.13) <.001
Modality (Spoken vs. Written) -0.01 (0.15) .941
Group ✕ Temporal Context -1.01 (0.26) <.001
Group ✕ Modality -0.05 (0.31) .882
Temporal Context ✕ Modality -0.25 (0.26) .342
Group ✕ Temporal Context ✕ Modality 0.20 (0.52) .695
L1 Mandarin Intercept -2.11 (0.14) <.001
Temporal Context -2.26 (0.18) <.001
Modality 0.06 (0.27) .817
Temporal Context ✕ Modality -0.31 (0.35) .373
L1 English Intercept -1.67 (0.10) <.001
Temporal Context -3.05 (0.19) <.001
Modality -0.06 (0.19) .752
Temporal Context ✕ Modality -0.14 (0.38) .705

92
 Chapter 2
Table 13.

Experiments 2 and 3: Between-experiment analyses on the effect of production modality on past -ed production.

B (SE) p
Main Model
Intercept -0.68 (0.15) <.001
Group (L1 Mandarin vs. L1 English) 0.35 (0.30) .251
Temporal Context (Present Habitual vs Past) 5.29 (0.20) <.001
Subject Number (Singular vs Plural) -0.12 (0.14) .393
Modality (Spoken vs Written) 0.39 (0.30) .199
Group ✕ Temporal Context 3.76 (0.37) <.001
Group ✕ Subject Number -0.17 (0.27) .543
Group ✕ Modality -1.05 (0.61) .098
Temporal Context ✕ Subject Number -0.46 (0.28) .086
Temporal Context ✕ Modality -0.24 (0.36) .495
Subject Number ✕ Modality -0.29 (0.28) .296
Group ✕ Temporal Context ✕ Subject Number 0.12 (0.55) .822
Group ✕ Temporal Context ✕ Modality -0.82 (0.71) .248
Group ✕ Subject Number ✕ Modality 0.84 (0.55) .130
Temporal Context ✕ Subject Number ✕ Modality 0.22 (0.56) .689
Group ✕ Temporal Context ✕ Subject Number ✕ Modality -0.01 (1.11) .993

93
 Chapter 2
(Table 13 continued)

L1 Mandarin
Intercept -0.85 (0.20) <.001
Temporal Context 3.38 (0.18) <.001
Subject Number -0.03 (0.15 .824
Modality 0.91 (0.39) .021
Temporal Context ✕ Subject Number -0.52 (0.30) .080
Temporal Context ✕ Modality 0.17 (0.34) .621
Subject Number ✕ Modality 0.13 (0.30) .660
Temporal Context ✕ Subject Number ✕ Modality 0.23 (0.60) .706
L1 English
Intercept -0.51 (0.23) .029
Temporal Context 7.26 (0.39) <.001
Subject Number -0.20 (0.23) .383
Modality -0.13 (0.47) .774
Temporal Context ✕ Subject Number -0.40 (0.47) .389
Temporal Context ✕ Modality -0.68 (0.64) .289
Subject Number ✕ Modality -0.71 (0.47) .129
Temporal Context ✕ Subject Number ✕Modality 0.21 (0.94) .820

94
 Chapter 2
Table 14.

Experiments 2 and 3: Between-experiment analyses on the effect of production modality on inflectional omissions.

B (SE) p
Main Model
Intercept 0.80 (0.17) <.001
Group (L1 Mandarin vs. L1 English) -1.33 (0.36) <.001
Temporal Context (Present Habitual vs. Past) 1.53 (0.22) <.001
Modality (Spoken vs. Written) -0.63 (0.33) .060
Group ✕ Temporal Context 0.74 (0.59) .208
Group ✕ Modality 1.18 (0.73) .104
Temporal Context ✕ Modality 0.19 (0.44) .673
Group ✕ Temporal Context ✕ Modality 0.28 (1.17) .811
L1 Mandarin Intercept 1.07 (0.21) <.001
Temporal Context 1.40 (0.25) <.001
Modality -0.88 (0.41) .035
Temporal Context ✕ Modality 0.13 (0.50) .791
L1 English Intercept -0.24 (0.25) .330
Temporal Context 1.85 (0.49) <.001
Modality 0.24 (0.50) .633
Temporal Context ✕ Modality 0.38 (0.91) .679

95
 Chapter 2

2.4.3. Interim discussion

Experiment 3 replicated the key findings of Experiments 1 and 2 in the written

modality (which did not require articulation): L1 Mandarin speakers showed
sensitivity to temporal context, but exhibited particularly low accuracy in the Present
Habitual Singular Subject condition (requiring 3SG -s), relative to L1 English
speakers’ performance in the same condition, and relative to their own performance
in the Past Tense conditions (requiring –ed). Critically, although L1 Mandarin
speakers were not statistically less likely to be accurate in the spoken modality
(Experiment 2) than in the written modality (Experiment 3), they were more likely to
make omission errors in the spoken modality than in the written modality. L1
English speakers did not show such effects. These findings are consistent with a
processing account of inflectional errors, based on inconsistent retrieval of
inflectional forms; they are not compatible with an account that attributes inflectional
errors purely to articulatory difficulties.

2.5. General Discussion

Previous research has established that L2 speakers frequently produce

inflectional errors but has not reached consensus over the loci of such errors.
Specifically, it is unclear whether erroneous inflectional production arises from
deficits in representation or processing, and where within the production system such
deficits might be located.

In Experiments 1 and 2, both L1 Mandarin and L1 English participants produced

spoken descriptions, and in Experiment 3, they produced written descriptions.
Predictably, in all three experiments, L1 Mandarin speakers produced less accurate
inflections than L1 English speakers in both Present Habitual and Past temporal
contexts. However, like L1 English speakers, L1 Mandarin speakers were also more
likely to produce the correct inflectional markings under appropriate contexts than
inappropriate contexts. This pattern held for both 3SG -s and past -ed, with a higher
proportion of past -ed produced across all three experiments. Particularly, this pattern

96
 Chapter 2

also held across spoken and written production, with higher inflectional accuracy for
written production, though this was not statistically significant.

2.5.1. The Locus of L2 Inflectional Errors

These results are informative about the nature of L2 speakers’ erroneous

inflectional production within a processing model of language production and cast
light on previous accounts that have been proposed from a theoretical linguistic
perspective. Taking these results as a whole, our spoken and written L2 production
data provide compelling evidence that errors in L2 inflectional production more
likely reflect processing breakdowns rather than representational deficits. At the start
of this paper, I outlined possible sources of error in L2 inflectional production within
current models of language production (Levelt, 1989; Levelt et al., 1999), together
with linguistic accounts of L2 inflectional errors (Hawkins & Chan,1997; Prevost &
White, 2000; Goad et al., 2003). Representational deficits and processing
breakdowns are implicated at multiple stages of language production. These include
conceptualization failures, missing or inconsistent activation of lemma level diacritic
features, missing or inconsistent association between syntactic functions, inconsistent
lexical retrieval and articulation failures. Current findings are consistent with some
but not all of these accounts.

We first consider possible representational deficits at multiple stages of language

production and evaluate the implications of our data. First, we consider the
possibility of conceptualization failures as a potential source of error during
inflectional production. Under this account, if an L1 Mandarin speaker did not have
conceptual distinctions necessary for inflectional morphology, they would omit L2
inflections across the board regardless of modality. Our findings contradict this
claim. L1 Mandarin speakers produced inflections with 55% (Exp. 2) and 66% (Exp.
3) accuracy in spoken and written modality across conditions, and were more likely
to produce both 3SG –s and past -ed inflections in appropriate temporal contexts than
in inappropriate temporal contexts. This indicates that L1 Mandarin speakers were
sensitive to L2 temporal distinctions at the level of conceptualization, refuting the
idea of representational and processing deficits at the conceptual level.

97
 Chapter 2

Second, we turn to the possibility that inflectional errors are the consequence of
missing diacritic features (e.g. tense) at the lemma level. If this were true, speakers
without relevant diacritic representations at the lemma level in their L1 would be
unable to make the syntactic distinctions for producing inflectional morphology
entirely. This would again predict that L1 Mandarin speakers of English would omit
inflections across the board. Current findings suggest otherwise: Consistent with
previous accounts of ‘optional’ inflectional production in second language
acquisition research (i.e., sometimes producing and sometimes omitting the
appropriate inflection), our participants’ inflectional production was systematically
variable. L1 Mandarin speakers of English systematically produced both 3SG -s and
past -ed consistent with temporal context, indicating that they had not only acquired
the underlying temporal distinctions, but also the syntactic distinctions for
subsequent retrieval of inflectional forms.

Third, an alternative possibility would be that representational deficits at the

lemma level could lead to random production of inflections, whereby L2 speakers
would fail to associate syntactic functions with appropriate diacritic features (e.g.,
activating the values 3rd and SINGULAR for a verb lemma’s PERSON and NUMBER
diacritic features respectively, following a 3rd person singular subject). This account
would predict no significant differences in the production of inflections across
temporal contexts and subject number. Again, our data clearly refute this assumption.
It is also clear from the appropriate use of 3SG -s in particular that L1 Mandarin
participants were able to carry out appropriate syntactic assignment for subject
number, even though 3SG -s production was poorer overall.

Hence our data clearly demonstrate that erroneous inflectional production was
not the result of failure to acquire relevant diacritic representations or syntactic
associations. As such, they argue against Hawkins and Chan’s (1997) Failed
Functional Feature Hypothesis, which claimed that it was not possible for L2
speakers to acquire grammatical features which do not exist in the speaker’s L1. Our
findings demonstrate that L2 speakers of English whose L1 does not use inflectional
morphology were capable of acquiring L2 temporal distinctions and inflectional
features after the critical period (all our participants acquired L2 English after age 5).

98
 Chapter 2

Instead, our results are in line with accounts which attribute inflectional error to
processing breakdowns. Within a psycholinguistic model of production, inflectional
errors could be caused by a processing breakdown which failed to activate the
relevant diacritic features at the lemma level and the appropriate syntactic
associations, which in turn would lead to failure to activate the corresponding
inflectional forms (in the same way as speech errors can arise in L1 production; Dell
et al., 1997). This account would predict that speakers’ production of production of
specific inflectional forms would be sensitive to temporal context, but that it would
be susceptible to error especially under processing load (e.g., time constraints), and
would show an effect of featural complexity. Particularly, considering the number of
links between lemma level representations and inflectional forms, inflections
involving more complex features (e.g., distinctions based on both subject number
and tense) would rely on accurate activation of multiple feature nodes, making
successful retrieval less likely. This stands in contrast with inflection markings
involving singular or less complex features (e.g. tense only), which only require
activation from one feature node, making successful retrieval more likely.

Our findings are compatible with this account. Our L2 speakers were sensitive to
temporal context, but nevertheless produced errors in terms of sometimes omitting to
produce inflections required (in linguistic terms, optionality). Moreover, 3SG -s,
requiring both subject number and tense information, was more frequently omitted
than past -ed. These findings therefore support Hawkins’ (2007) account of featural
complexity, and are consistent with data from Chondrogianni and Marinis (2012),
where 3SG -s was found to be more difficult to produce accurately than past -ed.
This finding can be viewed in conjunction with Dell et al.’s (1997) theory of L1
speech errors, where speech production errors in aphasic patients can be explained by
inappropriate weights between connections during transmission of activation. If the
same principle applies in the case of L2 production, the speaker may have
inappropriately weighted connections between feature nodes for activating the
correct inflectional form where context requires it. Such inappropriate weights
between node connections might be the result of L1 transfer.

Our findings are also compatible with Prevost and White’s (2000) Missing
Surface Inflection Hypothesis. Under this account, inaccurate or optional production

99
 Chapter 2

of inflections was attributed to difficulties in morphological encoding and activating

the relevant inflectional forms rather than representational deficits. By implication,
this indicates fundamentally a processing difficulty where inflections containing
more complex features (i.e., 3SG -s) should be more susceptible to this difficulty. In
our study, L1 Mandarin participants still processed L2 temporal and subject number
information during L2 inflectional production (more 3SG -s responses than any other
context). This suggested difficulties in activation and integration during retrieval of
inflectional forms rather than the lack of representations. Note that the current data
cannot tease apart activation of features and retrieval of inflectional forms.

Turning to an account locating L2 speakers’ inflectional errors in articulation

failures, production data across spoken (Exp. 2) and written (Exp. 3) modalities were
especially revealing on the role of articulation in inflectional production. Despite the
overall increase in inflectional accuracy by L1 Mandarin speakers of English in
written compared with spoken production, inflectional error patterns from spoken
production persisted in written production even when no overt articulation was
involved. This clearly indicated that articulation difficulties alone were not the
primary cause of inflectional errors and cannot solely account for erroneous
inflectional production. The source of inflectional error must primarily occur earlier
on in the production process. However, this does not preclude potential phonological
processing difficulties in the L2 that might contribute to higher incidence of errors in
spoken production. Additionally, the higher inflectional accuracy in the written
modality (i.e. ‘writing’ on a keyboard) might be linked to the increased response time
permitted in Experiment 3 (Experiments 1 and 2: 7000ms; Experiment 3: 15000ms),
which gave L1 Mandarin speakers more time to activate morphological
representations and retrieve corresponding lexical forms.

2.5.2. Limitations and Future Directions

Experiments 1, 2 and 3 provide convincing evidence that patterns of inflectional

errors are likely to be down to processing breakdowns, but is unable to tease apart
different types of processing breakdowns (e.g. activation of diacritic features and

100
 Chapter 2

retrieval failures). Moreover, current evidence cannot tease apart processing and
articulatory difficulties in the current data.

Assuming that phonological mediation occurs in both spoken and written

production (Chenoweth & Hayes, 2003; Friederici, Schoenle, & Goodglass, 1981;
Zhang & Damian, 2010), it remains plausible that discrepancies in L1 and L2
phonological properties constrained the speaker’s ability to encode relevant
phonemes correctly. In Levelt’s terms, difficulties in phonological and phonetic
processing (perhaps due to L1 phonotactics) could restrict native-like formation of
the phonetic plan (or inner speech), resulting in difficulties in the execution of
articulatory or orthographic motor gestures. In other words, even if the message is
not intended for oral articulation, the generation of the phonological word and
phonetic plan would still be necessary for written production. In the context of our
study, neither 3SG -s nor past -ed in the word final position are plausible phoneme
combinations in Mandarin Chinese. One may speculate whether all L2 phoneme
combinations, especially those which are not permitted, can be effectively planned
during phonological and phonetic encoding, and in turn executed during inflectional
production.

As L1 Mandarin speakers of English in the current study have all acquired

English after the first critical period (AoA > 5 years), one additional consideration
could be given to the role of explicit or metalinguistic knowledge in real-time
production. In Levelt’s terms, explicit knowledge could be a way for post critical
period learners to establish lemma level representations (diacritic features) and their
associated morphological forms. However, this does not necessarily mean L2
learners can activate these representations consistently during real-time production.
In other words, explicit knowledge enables ‘competence’ at a representational level,
but does not necessarily enable proficient ‘performance’ at a processing level.
Though explicit knowledge may not be critical for the accurate inflectional
production of 3SG -s and past -ed for all learners, but given the age of L2 acquisition
of our participants, the application of explicit metalinguistic knowledge during
production is likely to be crucial, especially for less proficient L2 speakers.

101
 Chapter 2

One last question concerns how L2 knowledge could be acquired for late L2
learners. One may speculate whether L1 Mandarin speakers initially viewed tense
inflections as functionally equivalents to Mandarin aspectual markers. Specifically,
L1 Mandarin speakers might map past -ed onto the Mandarin aspectual marker le for
functional use. This would be consistent with the core principle of the Feature
Reassembly Hypothesis (Lardiere, 2008), which could explain the high proportion of
past -ed responses to past perfective aspect events.

Taking these results as a whole, they provide compelling evidence through both
spoken and written L2 production, that erroneous L2 inflectional production is more
likely a processing problem, rather than a representational one. More importantly,
these findings have been interpreted in terms of both psycholinguistic framework of
language production and linguistic theories of L2 inflectional error. What is most
valuable in the current context is the attempt to reconcile these two perspectives in
their theoretical assumptions and predictions, highlighting areas where the two
perspectives complement each other, as well as areas where the two sets of theories
fall short.

102
 Chapter 3

Chapter 3

The effect of comprehension modality on L2 inflectional

processing in L1 Mandarin learners of L2 English

Second language (L2) comprehension is difficult for L2 learners, which may be

exacerbated when the learner’s L1 does not draw information from the same
semantic or morphosyntactic cues for meaning as the L2. Additionally, L2 learners
may find comprehension more difficult under greater cognitive load. Two self-paced
comprehension experiments investigated whether L1 Mandarin learners of L2
English could show sensitivity to semantic and morphosyntactic mismatches from
temporal adverbials and inflectional morphemes (which do not exist in Mandarin)
during real-time sentence processing, and whether this sensitivity is affected by
comprehension modality. Advanced L1 Mandarin learners of L2 English and L1
English controls either listened to (Experiment 4) or read (Experiment 5) English
sentences in a self-paced moving-window paradigm. Results showed that L1
Mandarin learners of L2 English exhibited non-native like sensitivity to omission of
inflectional morphemes compared with L1 English speakers, suggesting that L2
learners are sensitive to lexical and morphosyntactic mismatches during L2
comprehension, even if similar morphosyntactic features do not exist in the learners’
L1. Critically, whilst L1 Mandarin learners of L2 English showed sensitivity to
omission of inflectional morphemes during auditory (listening) comprehension, they
did not during visual (reading) comprehension. These results suggest that auditory
comprehension is not invariably more difficult for L2 learners; rather, sensitivity to
grammatical violations (i.e. omission of inflections) can in some contexts be
facilitated by perceptually salient auditory cues.

103
 Chapter 3

3.1. Introduction

Second language (L2) comprehension is a cognitively demanding task for L2

learners, especially for those who acquired the L2 during or after puberty (Johnson &
Newport, 1989; Segalowitz & Segalowitz, 1993). Not only does L2 comprehension
require the learner to have broad L2 lexical and grammatical knowledge, but it also
requires the learner to use L2 lexical and grammatical cues to understand L2
sentences in real-time. Whilst L2 vocabulary and grammatical knowledge can be
explicitly learnt in the classroom, native-like comprehension mechanisms - which
allows the learner to implicitly assimilate information from linguistic cues in real-
time - can be much more difficult for L2 learners to acquire. This is especially the
case when the L2 contains features that do not exist in the speaker’s first language
(L1). For example, the acquisition and processing of English inflectional morphology
for L1 Mandarin speakers of L2 English.

Inflectional morphology marks both semantic and syntactic information (e.g.,

person, number, tense, aspect etc.), and its use is reflective of the syntactic structure
and semantic information contained in the surrounding sentence or discourse
(Marslen-Wilson, 2007). In order to understand morphosyntactic information
represented by inflectional markings, the listener or reader must not only have a good
command of grammatical knowledge, but must also be able to readily assimilate
information within the grammatical context. For example, in a sentential context,
native-like processing of inflectional morphology requires the listener or reader to
know the conditions upon which the use of an inflectional marking (e.g. past -ed) is
appropriate, as well as the ability to assimilate lexical (e.g. subject, temporal
adverbial) with morphosyntactic cues (e.g. inflectional morphology) incrementally in
real-time. If the listener or reader is able to do this, then any mismatches between
these cues which violates L2 learners’ grammatical knowledge will result in
processing difficulties.

104
 Chapter 3

How do learners use L2 linguistic features for L2 comprehension when similar

features do not exist in the learner’s L1? Two possibilities could be considered. On
the one hand, learners might use their existing L1 comprehension mechanisms (or L1
implicit knowledge), so that they extract limited information from L2 input. On the
other hand, advanced L2 learners might develop a new set of comprehension
mechanisms (or L2 implicit knowledge) as part of the L2 acquisition process, so that
they become sensitive to lexical and morphosyntactic mismatches from L2 input in a
native-like manner, though they may be substantially slower than L1 speakers.
Whilst considerable L2 processing research has focused on the parsing of ambiguous
sentences, syntactic dependencies and agreement mismatches for evidence of native-
like sentence processing (e.g. Jiang, 2004; 2007; Juffs & Harrington, 1995; 1996;
Marinis, Roberts, Felser, & Clahsen, 2005; Papadopoulou & Clahsen, 2003; Roberts
& Felser, 2011 etc.), not a great deal of attention has been paid to L2 learners’ real-
time sensitivity to mismatches between lexical and morphosyntactic cues relating to
temporal information in sentential contexts (but see Roberts & Liszka, 2013).
Moreover, direct comparisons focusing on how comprehension across auditory and
visual modalities affect this process has not been made.

In this chapter, I focus on how L1 Mandarin learners of L2 English comprehend

English inflectional morphology (3rd person -s or past -ed) across auditory
(listening) and visual (reading) modalities. I test whether L1 Mandarin speakers are
sensitive to mismatches involving these inflectional morphemes during L2 English
comprehension, given that the tense feature underlying inflectional morphology is
absent (and therefore not grammaticalized) in Mandarin but obligatory in English.
Previous research has shown that inflections containing more complex information
(e.g. tense and number for 3rd person singular -s) are more difficult for L2 learners to
produce consistently than inflections containing less complex information (e.g. tense
for past -ed; Chondrogianni & Marinis, 2012), but it is not yet clear whether
information complexity would affect the use of morphological cues during
comprehension. Hence, I investigate whether L1 Mandarin learners experience
different levels of processing difficulties depending the amount of information
contained within the inflectional morpheme. Moreover, previous research has shown
that L2 auditory comprehension is disadvantaged compared with L2 visual

105
 Chapter 3

comprehension in grammaticality judgement tasks (Haig, 1991; Johnson, 1992;

Murphy, 1997), as it involves additional speech segmentation (Anderson, 1980) and
time pressure (Johnson, 1992). It is unclear if this disadvantage would persist if
speech is readily segmented. It is also unclear whether this disadvantage would still
apply during ‘normal’ comprehension, i.e., with the goal of determining a semantic
interpretation without explicit grammaticality judgement (as previously seen in
Jackson & Bobb, 2009; Jackson & Dussias, 2009). Therefore, I also seek to examine
cross-modality differences during morphological processing in L1 and L2 English
groups in a semantic-oriented comprehension task.

To summarise, the current study investigates whether L1 Mandarin speakers are

sensitive to lexical and morphosyntactic mismatches between temporal adverbials
and inflectional morphology during real-time L2 comprehension, and whether they
exhibit different levels of sensitivity to inflectional morphemes contingent on the
complexity of morphosyntactic features. Most significantly, I examine whether
processing modality (auditory or visual) modulates L2 learners’ sensitivity to
inflectional omissions.

3.1.1. Fundamental differences between L1 and L2 morphosyntactic

processing

A key prerequisite to discussing the nature of real-time L2 morphosyntactic

processing is whether there exists a qualitatively or quantitatively difference between
how native L1 speakers and non-native L2 learners process morphosyntactic cues in
real time. In keeping with the Shallow Structure Hypothesis (Clahsen & Felser,
2006), numerous empirical studies have demonstrated clear differences between
native L1 speakers and L2 learners in their neurological and behavioural responses to
grammatical violations during real-time comprehension.

Past ERP studies have revealed dedicated brain regions for the processing of
specific language features in L1 speakers. Comparatively, L2 learners exhibit
different sensitivities towards grammatical violations in these regions during
comprehension. Whilst proficient L2 learners show L1-like event-related potentials

106
 Chapter 3

(N400 responses) to semantic violations in L2 speech, they do not show the same
level of sensitivity towards L2 syntactic violations (P600 responses). This pattern has
been found across L2 learners from different L1 backgrounds (Hahne, 2001; Hahne
& Federici, 2001), and across different comprehension modalities (Weber-Fox &
Neville, 1996). Focusing on Chinese learners of English specifically, Chen et al.
(2007) found that unlike L1 English speakers, who exhibited sensitivity to subject-
verb agreement violations in the form of ERP responses, L1 Chinese learners of
English did not show such responses, despite showing highly accurate offline
grammatical knowledge. These findings suggest that L2 learners’ morphosyntactic
processing differs qualitatively from native L1 users, in ways that are also mediated
by L2 proficiency (VanPatten, Keating & Leeser, 2012).

Using a self-paced reading paradigm, Jiang (2004) found that even proficient L1
Mandarin learners of L2 English showed little sensitivity towards plural marking
violations. Jiang found that unlike L1 English speakers who showed significant
differences in reaction time between sentence segments with grammatical and
ungrammatical number marking (e.g. The child was watching some of the rabbit(s) in
the room*), L1 Mandarin learners of L2 English did not show any significant
differences between the two conditions (i.e. rabbit or rabbits), even though they
exhibited offline knowledge of appropriate number marking. Similar findings were
observed for subject-verb agreement violations among L1 Mandarin learners of L2
English (e.g. The bridges to the island was about ten miles away*), where they did
not show native-like sensitivity towards the ungrammatical segment (i.e. was) during
online-comprehension (Jiang, 2007).

Examining the processing of temporal agreement specifically, Roberts and

Liszka (2013) found that L2 English learners from L1 French and German
backgrounds, whilst demonstrating proficient offline L2 grammatical knowledge,
responded differently to L1-English speakers when encountering past simple and
present perfect temporal mismatches (e.g. When / Since she first started her job,
Emma loved / has loved the work very much). Whilst L1 English speakers
experienced selective processing difficulty with temporal mismatches between a
fronted temporal adverbial (at the beginning of the sentence) and an inflected verb in
the present perfect condition (e.g. Last year, James has gone swimming every day.

107
 Chapter 3

Now he’s getting bored of it.), they did not in the past simple condition (e.g. Since
last week, James went swimming every day. Now he’s getting bored of it.). In
contrast, L1 French learners of English experienced significant processing
difficulties to temporal in both present perfect and past simple conditions, whilst L1
German learners of English did not exhibit such processing difficulties in either
condition.

These findings are significant in several ways. Consistent with previous

research, Roberts and Liszka (2013) demonstrated non-native-like morphosyntactic
processing in L2 learners from different L1 backgrounds. At the same time, the
authors pointed towards language specific L1-effects in real-time L2 temporal
processing. Specifically, French and German both encode tense, but only French
grammaticalizes aspect whilst German lacks any overt aspectual morphemes
(Comrie, 1976; Schilder, 1997). This stands in contrast with English, where both
tense and aspect are grammaticalized using either an inflected verb (e.g. wanted) or
an auxiliary with a perfect form (e.g. has wanted). Thus, the apparent absence of
processing difficulty to mismatches among L1 German learners could be attributable
to the lack of overt aspectual markers in their L1. This has important implications in
the context of the current study, as Mandarin lacks overt markings for tense, which
could affect sensitivity to L2 mismatches between the temporal adverbial and the
inflected verb in L1 Mandarin learners of L2 English.

To summarise, behavioural and neurological evidence has repeatedly shown via

explicit and implicit measures that L2 morphosyntactic processing is fundamentally
different to L1 morphosyntactic processing. Importantly, despite having proficient
offline L2 grammatical knowledge, evidence has shown that L2 learners are not
sensitive to morphosyntactic violations in a native-like way during real-time sentence
processing. In 3.1.2. and 3.1.3, real-time processing of temporal information from
morphosyntactic cues will be discussed in relation to theories of monolingual and
bilingual sentence processing. This is particularly important to our discussion as we
focus on the role of the L1 on L2 learners’ ability to process temporal information
from inflectional morphology and detect potential mismatches between lexical and
morphosyntactic cues.

108
 Chapter 3

3.1.2. Theories of L2 sentence processing

Real-time sentence processing is a fast-paced event which requires individuals to

retrieve the semantics of lexical words as well as incrementally establishing the
grammatical structure of sentences (Rueschemeyer & Gaskell, 2007). Under time
pressure, the task of assimilating and interpreting lexical and morphosyntactic cues
can require significant cognitive resources in L2 learners, affecting automaticity of
L2 processing (Segalowitz & Segalowitz, 1993). The theoretical discussion
surrounding the processing of L2 morphosyntactic mismatches in sentential contexts
in this study centre around two main questions: 1) how do L2 learners handle two
sets of grammatical knowledge and acquire the ability to assimilate information from
L2 grammatical features when such features (or cues) are absent in the L1, and 2)
how L2 learners handle these differences to comprehend sentences in real-time.

How does L2 learners’ comprehension mechanism deal with two sets of

grammatical knowledge in their language systems? According to The Competition
Model (Bates & MacWhinney, 1989), comprehension is led by a series of interactive
activations based on form-meaning probabilistic mappings. Grammatical cues
compete with each other during language processing, with the strongest activations
converging on the most fitting interpretation of the sentence. Two factors are
important to the competition of grammatical cues, cue validity (the value of cues
extracted from linguistic input) and cue strength (priority or preference assigned to
cues as determined by reliability). The Competition Model utilises the connectionist
assumption which states that all mental processes (including language) share the
same cognitive structures and principles (MacWhinney, 1987). As such, L1 and L2
would share the same set of processing mechanisms. This means that the
mechanisms which had previously been adapted to process L1 grammar would
inevitably transfer onto the L2, resulting in interference. As the two sets of grammar
compete, the amount of interference would depend on the degree of overlap between
L1 and L2 grammatical properties. If the two grammatical systems share a number of
similar properties, the value of grammatical cues (cue validity) would be transferred
and strengthened from L1 to L2, and receive little interference from the L1.

109
 Chapter 3

However, if the two sets of grammar are dissimilar and have different valid cues, L1
cue validity could (inappropriately) transfer from L1 to L2, causing interference.

Returning to the first aim of the current study, how could L2 learners acquire the
ability for processing temporal agreement between temporal adverbials and
inflectional morphology when such features are absent in the learners’ L1? The
Competition Model would argue that L2 learners must adjust their processing
priorities contingent on the validity and reliability of L2 grammatical features (e.g.
inflectional morphemes) as cues to meaning. As Mandarin does not have inflectional
morphology, we cannot assume inflectional morphemes carry high cue validity for
L1 Mandarin learners of English by default, and therefore it is highly probable that
they do not initially assign hight processing priority to inflectional morphemes in
order to interpret the temporal context of the sentence. However, if L1 Mandarin
learners of English were to process inflectional morphemes consistently (evident in
the form of sensitivity to adverbial-inflection mismatches) much like L1 English
speakers do, this would be evidence for L1 Mandarin learners of English adopting L2
cue validity during real-time L2 comprehension.

The second aim of the study concerns whether L2 learners make processing
distinctions between inflections containing different numbers of features. In other
words, inflectional markings can require agreement with more than one cue
depending on context, which necessitates the L2 learner to carry out context-specific
cue processing. For example, (temporal) inflectional morphology requires obligatory
agreement with temporal adverbials if present. However, in some cases, inflectional
morphology must also account for subject number (e.g. 3SG -s) if presented in a
specific temporal context (e.g. present habitual). These two scenarios require
different priorities over linguistic cue processing. For example, L2 English learners
would only need to refer to temporal cues in a past tense context during
morphological processing, but would also need to refer to subject number
information if a singular subject is involved in a present tense (habitual) context,
which could be extremely challenging in real time. If L2 learners exhibit non-native-
like processing behaviour, and do not refer to subject number information, then they
would not be sensitive if subject number agreement is violated (e.g. if 3SG -s was
omitted).

110
 Chapter 3

Although the Competition Model outlines the principles behind acquiring new
processing priorities for comprehending L2 input, as Bates and MacWhinney (1989)
pointed out, its sole purpose is not to account for real-time processing of sentences in
bilinguals. Importantly, it does not make a comprehensive distinction between
language competence and language performance. In other words, whether L2
learners could in principle assimilate information from inflectional morphology
based on explicit grammatical knowledge and be sensitive to mismatches between
lexical and morphosyntactic cues does not necessarily mean that they will
consistently do so in real-time. Hence, further theories which account for this crucial
distinction are necessary for this discussion.

Theories in second language research make distinctions between explicit and

implicit knowledge in language acquisition (R. Ellis, 2005; 2006), which parallels
the competence vs. performance distinction. Specifically, explicit knowledge reflects
conscious, metalinguistic knowledge of grammar which can be accessed and
measured via language tests, whereas implicit knowledge reflects intuitive,
automatised processes which apply grammatical knowledge during real-time L2
comprehension. Though the two sets of knowledge are theoretically related, the
former does not necessarily predict the latter. In other words, acquiring explicit
grammatical knowledge does not necessarily mean this knowledge will be applied in
real-time during L2 comprehension.

In the context of the current study, assuming that L2 English learners have
acquired the grammatical (or explicit) knowledge for morphosyntactic agreement
between temporal adverbials and inflectional morphology despite its absence in the
L1, native-like processing would also require learners to have the implicit knowledge
of how inflectional morphemes should be consistent with the temporal adverbial in
real-time. If they do, then L2 learners should be sensitive to any potential
mismatches between the temporal adverbials and inflectional morphology (Yesterday
she walk a mile*). If they do not, then we would assume implicit knowledge is
missing or not fully acquired. This would be consistent with previous studies where
L2 learners exhibited non-native-like grammatical processing despite having
proficient offline knowledge (Jiang, 2004; 2007; Marinis, Roberts, Felser & Clahsen,
2005; Roberts & Liszka, 2013).

111
 Chapter 3

Accounts of second language acquisition also make claims about cue saliency
and redundancy during L2 acquisition. Under the usage-based approach, speakers
will prefer to direct their attention to the most salient and effective cue during L2
comprehension (learned attention; Ellis & Wulff, 2008). If a grammatical feature is
not an effective and unique cue to the overall semantic interpretation of the message,
it will often be considered redundant by the speaker. For example, as temporal
morphology is most frequently used alongside temporal adverbials (e.g. every
morning), inflectional morpheme on the verb can often be considered redundant in
understanding the overall meaning of the sentence. In fact, research evidence has
shown that L2 learners are faster to comprehend the temporality of events when
sentences included both temporal adverbials and verb morphology than verb
morphology alone (e.g. Lee, Cardierno, Glass & VanPatten, 1997; Boatwright,
1999), indicating that temporal adverbials are more powerful during processing of
temporal information. This difficulty is compounded by knowledge of existing
mappings between words and functions, which overshadows the acquisition of
additional cues (i.e., the blocking phenomenon; N. Ellis, 2006). For example,
knowing that temporal adverbials indicate temporal properties of events (e.g.
yesterday) could make it harder for L2 learner to acquire another cue which also
indicates temporal properties (e.g. inflectional morphology such as past -ed).
However, for languages where inflectional morphology is obligatory, they have to be
consistent with other temporal cues such as temporal adverbials. Therefore, even
though inflectional morphology may not be critical for the overall interpretation of
the message, if an L2 speaker has an adopted native-L1 like processing mechanism,
agreement processing of both temporal adverbial and inflectional morphology would
be an essential part of successful L2 comprehension.

To summarise, theories in L2 sentence processing have in their own terms

explained: 1) how L2 learners may in principle acquire a comprehension mechanism
which adjusts processing priorities depending on the validity of cues to meaning in
the L2, 2) how the ability to apply grammatical knowledge in real-time is necessary
for L2 native-like sentence comprehension, and 3) why the ability to process
inflectional morphology as temporal marking may be hard to acquire. As noted by
Slabakova (2015), it is possible for L2 learners to acquire temporal meaning without

112
 Chapter 3

morphology. Therefore, the key question we ask in this chapter is not whether L2
learners of English can acquire temporal concepts in the L2, nor whether they
possess the relevant grammatical knowledge, but rather whether L2 learners can
acquire a native-like comprehension mechanism to apply their explicit grammatical
knowledge in real-time. Sensitivity to mismatches between linguistic cues is
therefore indicative of whether L2 learners have a native-like comprehension
mechanism or implicit knowledge for real-time L2 comprehension.

3.1.3. The effect of comprehension modality on grammatical sensitivity

The final key point of consideration for the current study concerns the effect
comprehension modality on L2 learners’ ability to detect potential
ungrammaticalities, specifically mismatches between temporal adverbials and
inflectional morphology during auditory (listening) and visual (reading)
comprehension.

Past research has revealed an effect of comprehension modality on grammatical

sensitivity by presenting identical experimental stimuli to participants in auditory and
visual forms. Using identical stimuli from Johnson and Newport (1989), Johnson
(1992) showed that visual presentations of stimuli revealed significantly greater
levels of accuracy and sensitivity than auditory presentations on English
grammaticality judgement tasks in L2 learners. This was shown for English
morphemes (Johnson, 1992) as well as subjacency violations (Haig, 1991; Murphy,
1997). Aside from methodological differences which may have confounded
experimental findings (longer time given to visual (text) presentations), Anderson
(1980) also pointed out natural auditory stimuli require L2 learners to actively
segment continuous streams of speech, making it more cognitively demanding for L2
learners to process compared with L2 text. L2 listeners would also have to carry out
more complex processing in the auditory modality (sound-to-form-to-meaning) than
in the visual modality (form-to-meaning), potentially resulting in slower and less
accurate reaction on grammaticality judgement.

113
 Chapter 3

Interaction between task-specific demands and comprehension modality also has

important implications for real-time L2 processing. Specifically, grammaticality
judgement tasks could induce the L2 learner to strategically and intentionally
monitor grammatical violations, for which visual stimuli could be more efficiently
processed, even if these grammatical violations do not hinder the correct
interpretation of the message. In contrast, meaning-oriented comprehension without
grammaticality judgement is closer to natural L2 comprehension, where grammatical
violations are only critical if they hinder the interpretation of the message. As shown
by Jackson and Bobb (2009) and Jackson and Dussias (2009), L2 learners tended to
show greater resemblance to native-like processing in grammaticality judgement
tasks than in tasks which assess L2 learners’ understanding of the experimental
sentences. These findings point towards the notion that L2 learners’ real-time
processing of L2 sentences could be significantly affected by the task they are asked
to perform.

3.1.4. The current study

Current review of existing literature and models indicates that: 1) Successful L2

comprehension requires the L2 learner to assimilate information from multiple
linguistic cues. However, it is not clear whether L2 learners could acquire the
comprehension mechanism (or implicit knowledge) to consistently and incrementally
assimilate the relevant cues in real-time L2 sentence processing, especially when the
relevant grammatical features (i.e. inflectional morphology) are absent in the L1; 2)
The comprehension of L2 temporal morphology requires context-specific processing
with different cues (e.g. for 3SG -s vs. past -ed), but it is not clear whether L2
learners would make such distinctions in real-time L2 processing; 3) Auditory
comprehension is significantly less accurate for L2 learners as it requires additional
speech segmentation. However, it is not clear whether auditory processing would still
be disadvantaged compared with visual processing if L2 learners no longer have to
segment continuous speech, and with the intention of carrying out meaning-oriented
comprehension instead of monitoring for grammatical violations.

114
 Chapter 3

This chapter presents two comprehension experiments examining the processing

of English temporal morphology 3SG -s (e.g., walks) and past -ed (e.g., walked)
using self-paced listening and self-paced reading paradigms (Ferreira, Henderson,
Anes, Weeks & McFarlane, 1996; Just, Carpenter & Woolley, 1982). In each
paradigm, participants were presented with grammatical and ungrammatical
sentences where ungrammatical constructions were associated with missing
inflectional morphemes in obligatory contexts. L2 learners’ reaction time (RT) at
each segment of the sentence was compared against L1 English controls, where
longer RTs at ungrammatical verb segments indicated processing difficulty during
corresponding grammatical verb segments.

In two experiments, L1 Mandarin and L1 English participants either listened to

(Experiment 4) or read (Experiment 5) English sentences involving different
temporal contexts. Experimental sentences uniformly contained a temporal adverbial,
a singular subject, a transitive verb, an object and a prepositional phrase in this order.
Temporal context (Present Habitual, Past) was manipulated via the temporal
adverbial (e.g. every weekend, yesterday), and Grammaticality was manipulated via
the omission of inflectional morphology (3SG -s or past -ed). Participant reaction
time (RT) was recorded for each segment of the sentence, focusing on the critical
verb segment (where inflectional omission takes place).

Given current accounts of L2 sentence processing, there are two possibilities. If

L1 Mandarin learners of L2 English acquired the grammatical knowledge regarding
inflectional morphology as a cue to temporal information, and have the ability to
apply this knowledge in real-time in native-like way, they would be sensitive to
mismatches between temporal adverbial and the omission of corresponding 3SG -s
and past -ed inflections (e.g. every day / yesterday the girl paint sunflowers in the
park). This would result in significantly slower RTs in critical verb segments with
omitted morphology compared with those without (i.e. a significant effect of
Grammaticality on verb segment RTs). On the other hand, if L1 Mandarin learners of
L2 English have acquired the explicit grammatical knowledge regarding inflectional
morphology as a cue to temporal information, but have not acquired the implicit
knowledge for applying grammatical knowledge in real time, one would not expect
significant RT differences between critical verb segments with and without 3SG -s

115
 Chapter 3

and past -ed omissions (i.e. no significant effect of Grammaticality on verb segment
RTs). Moreover, it remains unclear to what extent L1 Mandarin learners of L2
English would resemble native-like sensitivity to inflectional omissions. If they do,
then there would not be a significant effect of Group on verb segment RTs. If they
exhibit weaker levels of sensitivity to inflectional omissions, then one would expect a
significant effect of Group on verb segment RTs.

Additionally, I examine whether there are significant differences between L1

English speakers and L2 English learners’ sensitivity to the omission of 3SG -s and
past -ed inflectional morphemes. As stated previously, these two temporal markings
require agreement to different cues depending on context. Whereas 3SG -s denotes
agreement with both tense and subject number, past -ed denotes agreement only with
tense. It is noted that as subject number and tense are both marked using 3SG -s in
English, teasing the use of subject number cue from tense cues in L2 learners in
isolation is difficult. However, by comparing response RTs to past -ed omissions
against 3SG -s omissions, one could examine whether the addition of subject number
cue in 3SG -s facilitates or hinders the detection of inflectional omissions in L2
learners. In other words, one could observe whether having two agreement features
(tense and subject number) would make the detection of missing inflections easier or
more difficult for L2 learners compared with having only one agreement feature. If
additional features affect participants’ sensitivity towards 3SG -s compared with past
-ed omissions, then one would see a significant effect of Temporal Context (i.e. 3SG
-s vs. past -ed inflectional marking) with a Temporal Context ✕ Grammaticality
interaction.

Finally, as noted by Johnson (1992), inaccurate auditory perception of stimuli

could significantly confound the result of grammaticality judgements. Therefore, I
carried out experiments examining both auditory and visual comprehension, placing
significant emphasis on cross modality variations during L2 comprehension. If
auditory modality significantly disadvantages comprehension due to inaccurate or
slower perception of auditory stimuli compared with visual stimuli, then L1
Mandarin learners of L2 English should exhibit weaker sensitivity to inflectional
omissions in the auditory than in the visual modality. However, if auditory cues
facilitate comprehension in the auditory modality, then L1 Mandarin learners of L2

116
 Chapter 3

English should exhibit stronger sensitivity in the auditory compared with the visual
modality. In statistical terms, one would expect differences in effect sizes between
auditory and visual statistical models if comprehension modality does in fact affect
sensitivity to inflectional omissions.

To sum up, the current study examines: 1) whether L2 learners, with no

inflection in their L1, could acquire a native-like comprehension mechanism (or
implicit knowledge) to apply explicit grammatical knowledge relating to inflectional
use during real-time L2 comprehension; 2) whether L2 learners’ sensitivity to
inflectional omissions is facilitated or reduced by additional agreement features (i.e.
subject number in 3SG -s) during L2 comprehension; 3) whether L2 learners exhibit
weaker sensitivity to inflectional omissions during auditory than in visual L2
comprehension.

3.2. Experiment 4

3.2.1. Method

Participants

61 L1 Mandarin learners of L2 English aged 19-34 (M=23.79, SD=2.48) and 56

L1 English speakers aged 17-31 (M=20.48, SD=5.11) participated in this experiment.
All L1 Mandarin participants were late learners of English (AoA > 5 years) and had
an IELTS (International English Language Testing System) score of 6.5 or above
with 6 or above in the listening component of the IELTS. Their length of stay (in
months) and daily exposure to L2 English (in hours) were also recorded (see
Appendix G). All L1 English speakers had no exposure to any other languages
before the age of five. An additional measure of morphological proficiency was used
as part of the experiment for both L1 Mandarin and L1 English groups to show that
they have the appropriate offline grammatical knowledge of inflectional morphology
under different temporal contexts9.

9
See Appendix J for a sample copy of the Morphological Proficiency Test, and Appendix K for descriptive and
inferential statistics on this test across L1 Mandarin and L1 English groups.

117
 Chapter 3

Materials

Nine regular (experimental) and nine irregular (filler) transitive verbs were
chosen for the listening experiment. Four different sentences were created around
each verb, with each sentence having four versions corresponding to the four
experimental conditions (see Table 15 for examples). All 288 sentences10 included a
temporal adverbial, a subject, a verb, an object and a prepositional phrase.

Table 15.

Experiment 4: Example of a stimuli sentence across temporal contexts and grammatiaclity.

Forward slashes (/) denote segment boundaries.

Temp. Context Grammaticality temp adv. / subject / verb / object / prep. phrase
Present Habitual Grammatical Every Saturday / the girl / paints / sunflowers / in the park.
Present Habitual Ungrammatical Every Saturday / the girl / paint / sunflowers / in the park.*
Past Grammatical Yesterday / the girl / painted / sunflowers / in the park.
Past Ungrammatical Yesterday / the girl / paint / sunflowers / in the park.*
Question: Do / did the girl paint sunflowers in the gallery?

* indicates ungrammaticality.

For experimental sentences, I used singular subjects and manipulated temporal

context (Present Habitual vs. Past) and grammaticality of the inflectional form
(grammatical: without omission vs. ungrammatical: with omission). Two
grammatical versions of each sentence containing verbs with 3SG -s and past -ed
inflectional endings were generated and two ungrammatical versions with these
inflections omitted. Temporal context was also indicated by temporal adverbials for
Present or Past temporal contexts e.g. Every weekend (Present Habitual) vs.
Yesterday (Past). For filler sentences, location prepositional phrases were used
instead of temporal adverbials, and non-inflectional grammatical errors, such as the
incorrect use of articles, determiners and prepositions were used instead of
inflectional omissions. Different versions of the same sentence were assigned across

10 See Appendix H and I for the full list of experimental and filler sentences used in Experiments 4 and 5.

118
 Chapter 3

four different participant groups using a Latin square design, such that each
participant would hear each sentence under only one experimental condition.
Frequency of nouns was tallied to ensure no word was overly repetitive across
sentences. Closed comprehension questions were created for a quarter of the
sentences based on the non-verb content of each sentence (see Table 15). The tense
of the auxiliary verb ‘do’ was always consistent with the temporal context of the
sentence. This was to avoid participants’ attention being drawn to intentionally
monitor the temporal contexts of sentences and corresponding verb inflections.

All sentences were recorded in the University of Edinburgh PPLS recording

studio with a male RP English speaker with a sampling rate of 44.1 kHz (16-bit)
in .wav format. The speaker took natural pauses between phrases to avoid co-
articulation. Every iteration of each sentence was recorded separately and edited the
recordings by dividing each sentence into five audio segments (see Table 12). All
audio files were programmed by trial and condition in E-Studio 2.0 for stimulus
presentation. Subsequently, a list of audio file durations (in milliseconds) was
compiled using Praat (version 6.0, Boersma & Weenink, 2015) and scripts provided
by the UCLA Phonetics Lab.

An original Morphological Proficiency Test was used to assess participants’

offline knowledge of temporal inflectional morphemes. The test is targeted at L2
learners of English with intermediate to advanced proficiency (see Appendix K). The
test contains two sections, consisting of 30 multiple choice questions and 20 gap-
filling exercises.

Design

This experiment used a 2 x 2 x 2 mixed design. The between-subject variable

was Participant Group (L1 Mandarin vs. L1 English), and the within-subject
variables were Temporal Context (Present Habitual vs. Past) and Grammaticality
(Grammatical vs. Ungrammatical).

119
 Chapter 3

Procedure

The experiments took place in a quiet lab with a computer and a pair of stereo
headphones. Before commencing, participants provided demographic details
including language background information and gave consent for their data to be
used. Each participant was introduced to the experimental set-up by reading written
instructions in English. The experimenter repeated the instructions in Mandarin if
participants did not fully understand the tasks.

For the self-paced listening task, each participant listened to 72 sentences (36
experimental + 36 filler) presented via E-Studio 2.0 and a pair of headphones. For
each trial, participants first viewed the phrase [READY?] on a standard-sized 1920 x
1080 computer screen. Then, participants initiated the presentation of each sentence
by pressing the [SPACE BAR]. Each segment of the sentence was then played
sequentially at each press. At the end of each trial, participants either answered a
comprehension question by pressing one of two number keys (1 for YES; 2 for NO)
based on the sentence they just heard, or pressed the [SPACE BAR] to continue with
the next trial. Participants completed five practice trials containing sentences with
and without comprehension questions (to reflect selective presentation of
comprehension questions in the task) before starting the experiment. Presentation
order was randomised to prevent trial order effect. Participants had an optional break
after every 12 trials to prevent fatigue. They were reminded that they should progress
through the sentences at a steady pace, and not press the [ SPACE BAR] before each
segment ended.

Following the self-paced listening task, each participant completed the

Morphological Proficiency Test. They were asked complete it as quickly as possible
and hand it back to the experimenter. Participants received 0.5 hours course credit or
£5 cash for their time.

120
 Chapter 3

Coding and scoring

The self-paced listening task produced cumulative reaction times (RT) from the
start of every audio segment to the point of response. RTs for each segment was
calculated by deducting the duration of audio files (calculated using Praat) from the
cumulative RTs (recorded via E-Prime), and responses to comprehension questions
were scored as binary data (0 or 1). 7% of trial data were excluded based on the
following criteria: 1) temporal adverbial and verb segments with negative raw RTs,
where participants responded before the end of the segment, 2) extreme raw RTs
outside +/- 2 SD, and 3) trials with incorrect comprehension question responses.

3.2.2. Results

Self-paced Listening (SPL) Task

For SPL reaction time analyses, a forward model building strategy was used
with a maximal random effects structure (Barr et al., 2013). If the addition of a
predictor significantly improved model fit, then it was kept as part of the final model.
In order to analyse L1 Mandarin speakers’ auditory sensitivity to L2 English
inflectional omissions on the critical verb in a given temporal context, a general
linear mixed-effects models (GLMMs) was used with Group (between-subject),
Temporal Context and Grammaticality (within-subject) as main effect predictors, and
Participant, Item and Trial Order as random intercepts if they significantly improved
model fit. Fixed-effects predictors were contrast-coded, and the outcome variable
(RT) was centred before being added to the model.

For the purposes of this study, descriptive figures for all five segments across
both temporal contexts are presented, but GLMM models are only reported for RT
data from the critical verb segment.

121
 Chapter 3

Figure 16. Mean RTs for temporal adverbial, subject, verb, object and prepositional phrase
segments for L1 Mandarin and L1 English groups (N=61;56). Error bars denote +/- 1 SE.

Figure 16 shows the mean RTs for each of the five segments across Present
Habitual (3SG -s) and Past (-ed) temporal contexts for L1 Mandarin and L1 English
groups. Highlighted regions contain the pre-critical, critical and post-critical
segments.

Overall, there was no significant main effect of Group. The L1 Mandarin group
did not have significantly longer RTs in the critical segment compared with the L1
English group (M=605.44 vs. M= 605.06). However, participants showed a
significant main effect of Grammaticality irrespective of their L1: Both L1 Mandarin
and L1 English groups produced shorter RTs in the grammatical condition compared
with the ungrammatical condition (L1 Mandarin: M= 591.86 vs. M= 618.47; L1
English: M= 552.82 vs. M= 656.95). There were significant interactions between
Group and Temporal Context and Group and Grammaticality respectively, indicating
significant differences between how L1 Mandarin and L1 English speakers
responded to inflectional omissions as a whole, and to 3SG -s and past -ed omissions
separately (Table 16). Interestingly, there was also a two-way interaction between
Temporal Context and Grammaticality, indicating sensitivity to inflectional omission
was different for 3SG -s and past -ed.

122
 Chapter 3

Subgroup analyses confirmed the significant main effect of Grammaticality for

both L1 Mandarin and L1 English groups, indicating sensitivity to inflectional
omissions across groups. Temporal Context however, did not produce a significant
main effect for the L1 Mandarin group (Present Habitual 3SG -s: M = 606.18 vs.
Past -ed: M = 603.99), but did for the L1 English group (Present Habitual 3SG -s: M
= 594.48 vs. Past -ed: M = 615.29). Similarly, the interaction between Temporal
Context and Grammaticality was also not significant for the L1 Mandarin group but
was for the L1 English group. These results indicated that the L1 Mandarin group did
not process 3SG -s and past -ed omissions differently, but the L1 English group did.

Table 16.

Experiment 4: Generalised linear mixed-effects model (GLMM) statistics for RT on critical

verb segment in self-paced listening task for L1 Mandarin and L1 English groups
(N=61;56).

B (SE) p
Intercept 609.60 (29.27) <.001
Group (L1 Mandarin vs. L1 English) 16.28 (43.07) .706
Temporal Context (Present Habitual vs. Past) 11.77 (7.32) .108
Grammaticality (Grammatical vs. Ungrammatical) 69.85 (7.33) <.001
Group ✕ Temporal Context 30.16 (14.66) .040
Group ✕ Grammaticality 67.13 (14.64) <.001
Temporal Context ✕ Grammaticality -39.15 (14.65) .008
Group ✕ Temporal Context ✕ Grammaticality -16.83 (29.31) .566

L1 Mandarin
Intercept 602.00 (34.52) <.001
Temporal Context -3.09 (10.90) .777
Grammaticality 37.30 (10.91) .001
Temporal Context ✕ Grammaticality -29.54 (21.85) .176
L1 English
Intercept 617.21 (38.66) <.001
Temporal Context 25.99 (9.73) .008
Grammaticality 102.95 (9.74) <.001
Temporal Context ✕ Grammaticality -46.18 (19.46) .018

123
 Chapter 3

3.2.3. Interim Discussion

In Experiment 4, I examined L2 temporal processing during auditory

comprehension in L1 Mandarin learners of L2 English and L1 English controls via a
self-paced listening task. There were several findings of interest. First, L1 Mandarin
participants showed significantly slower RTs on ungrammatical trials (where
inflections were omitted) compared with grammatical trials (where inflections were
attached), much like L1 English participants. This critically showed that L1
Mandarin participants were sensitive to the omission of inflectional markings where
contexts required them. However, it was also clear that L1 Mandarin and L1 English
participants responded differently to inflectional omissions depending on temporal
context. Whereas L1 Mandarin participants did not show differential sensitivity
towards 3SG -s and past -ed omissions, L1 English exhibited superior sensitivity to
3SG -s omissions than past -ed omissions, which indicated integral differences in
3SG -s and past -ed processing in native-like comprehension.

As noted in Johnson (1992) and Murphy (1997), auditory presentation of stimuli

can impose problems in comprehension tasks, resulting in low performance
accuracy. Specifically, erroneous phonological processing can confound measures of
higher-level processing (i.e. grammaticality judgement). In this case, one cannot be
sure whether the phonological properties of 3SG -s or past -ed would be harder to
‘hear’ for a L1 Mandarin participant on a perceptual level. Therefore, in the next
experiment, the current experiment was administered in a non-auditory modality, i.e.
self-paced reading, eliminating the element of auditory processing from sentence
comprehension.

124
 Chapter 3

3.3. Experiment 5

3.3.1. Methods

Participants

Two new groups of participants, including 61 L1 Mandarin (L2 English) aged

21-28 (M=22.72, SD=1.32) and 57 L1 English speakers aged 18-43 (M= 22.14,
SD=4.13) participated in this experiment. L1 Mandarin speakers were required to
have a score of 6.5 overall as well as a score of 6 on the reading section of the IELTS
exam. All other recruitment criteria were identical to Experiment 4. The same
morphological proficiency test from Experiment 4 was used in Experiment 5.

Materials

Identical sentences from the self-paced listening task in Experiment 4 were used,
but they were prepared in the written form. Each sentence again contained five
segments (temporal adverbial / subject / verb / object / prepositional phrase) and
closed comprehension questions were again created for a quarter of the sentences
(Table 17). All sentences used identical font size and style (font size: 24, font style:
Courier New) for presentation in E-Studio 2.0.

Table 17.

Experiment 5: Illustration of stimuli sentence presented using a visual moving-window

paradigm for the self-paced reading task.

Temp. Adverbial Subject Verb Object Prep. Phrase

Slide 1 Yesterday --- ---- ------- ---------- -- --- ----

Slide 2 --------- the girl ------- ---------- -- --- ----

Slide 3 --------- --- ---- painted ---------- -- --- ----

Slide 4 --------- --- ---- ------- sunflowers -- --- ----

Slide 5 --------- --- ---- ------- ---------- in the park

125
 Chapter 3

Design

The design was identical to Experiment 4.

Procedure

The experimenter introduced participants to the set-up of the experiment via the
same procedures as Experiment 4. All aspects of the self-paced reading task were
identical to the self-paced listening task except for the modality of presentation. For
the self-paced reading task, a visual moving-window paradigm (Ferreira et al., 1996),
was used where participants silently read 36 experimental and 36 filler sentences one
segment at a time by pressing the [SPACE BAR] (Table 17). At every press, all other
non-target segments became dashes to prevent interference. At the end of each trial,
participants either answered a comprehension question or continued with the next
trial as per the listening experiment. Participants had an optional break after every 12
trials to prevent fatigue.

Following the self-paced reading task, participants again completed the

Morphological Proficiency Test and were paid £5 for their time.

Coding and Scoring

The self-paced reading task produced RTs from the start of presentation to the
point of response for each segment. 2% of trial data were removed based on the
following exclusion criteria: 1) All trials with extreme RTs exceeding 3000 ms
(assumed to reflect a lack of concentration) and below 100 ms (assumed to be an
non-intentional response) in the temporal adverbial and verb segments. 2) Trials with
incorrect comprehension responses. Residualised RTs for each segment were
calculated using the word length (number of letters per segment; Ferreira & Clifton,
1986).

126
 Chapter 3

3.3.2. Results

Self-Paced Reading (SPR) Task

In order to analyse L1 Mandarin participants visual sensitivity to L2 English

inflectional omissions in different temporal contexts, GLMMs were again used with
Group (between-subjects), Temporal Context and Grammaticality (within-subjects)
as fixed-effect predictors. Identical to the SPL analyses in Experiment 4, Participant,
Item and Trial Order were included as random effects if they significantly improved
model fit. Prior to the construction of the GLMMs, fixed-effects predictors were
contrast-coded, and residualised RTs were centred around a mean of 0. I considered
log-transformations inappropriate for the current RT data, as a logarithmic scale may
obscure between-group differences, especially their interactions with key predictors
in the analyses (see Lo & Andrews, 2015, for discussion).

For the SPR task, I focused on the differences between RTs as affected by
Temporal Context and Grammaticality across L1 Mandarin and L1 English groups
on the critical (verb) and on the spill-over (object) segments. Assuming that L2
learners generally have slower reading times (e.g. Fraser, 2007), sensitivity to
grammatical violations at the critical (verb) segment could potentially be delayed and
reflected in the post-critical spill-over (object) segment.

Figure 17. Experiment 5: Mean residualised RTs for temporal adverbial, subject, verb,
object and prepositional phrase segments in the self-paced reading task for L1 Mandarin and
L1 English groups (N=61;57). Errors bars denote +/- 1 SE.

127
 Chapter 3

Figure 17 shows the residualised RTs for each of the five segments across
Present Habitual (3SG -s) and Past (-ed) temporal contexts for L1 Mandarin and L1
English groups. Highlighted regions contain the pre-critical, critical and post-critical
(spill-over) segments.

Critical (verb) segment

Overall, there was a significant main effect of Group on the critical verb
segment. The L1 Mandarin group produced significantly shorter RTs than the L1
English group after accounting for word length (M = -0.09 vs. M = 0.15). However,
there were no effects of Temporal Context (Present Habitual: M = 0.02 vs. Past: M =
0.02) nor Grammaticality (Grammatical: M = 0.02 vs. Ungrammatical: M = 0.02).
Importantly, there was a significant two-way interaction between Group and
Grammaticality, indicating that L1 Mandarin and L1 English groups responded to
inflectional omissions differently, also prompting subgroup analyses (Table 18).
Subgroup analyses revealed a marginal main effect of Grammaticality for the L1
Mandarin group, and a significant effect of Grammaticality for the L1 English group.
There was no significant main effect of Temporal Context, nor any interactions
between Temporal Context and Grammaticality in either group. This indicated that
within each group, participants did not process 3SG -s and past -ed omissions
differently.

Spill-over (object) segment

Examining the spill-over (object) segment, there was a significant main effect of
Group (Table 18). Unlike the critical verb segment, the L1 Mandarin group produced
longer RTs compared than the L1 English group, after accounting for word length
(M = 0.01 vs. M = -0.20). The main model did not show significant main effects of
Temporality nor Grammaticality, but did show a significant three-way interaction
between Group, Temporal Context and Grammaticality. Upon closer examination,
subgroup analyses revealed no significant main effects of Temporal Context nor
Grammaticality for either group, restricting the significant three-way interaction to
group differences alone.

128
 Chapter 3
Table 18.

Experiment 5: General linear mixed-effects model (GLMM) statistics for residualised RTs in
the critical (verb) segment and spill-over (object) segment in the self-paced reading task for
L1 Mandarin and L1 English groups (N=61;57).

Verb Segment Spill-over Segment

B (SE) p B (SE) p
Intercept 0.02 (0.02) .223 -0.09 (0.03) .002
Group (L1 Mandarin vs. L1 English) 0.24 (0.04) <.001 -0.21(0.04) <.001
Temporal Context -0.00 (0.02) .985 -0.01 (0.05) .772
(Present Habitual vs. Past)
Grammaticality 0.01 (0.02) .829 0.01 (0.05) .771
(Grammatical vs. Ungrammatical)
Group ✕ Temporal Context -0.03 (0.05) .574 0.07 (0.06) .238
Group ✕ Grammaticality 0.20 (0.05) <.001 -0.03 (0.06) .596
Temporal Context ✕ Grammaticality 0.03 (0.05) .518 -0.07 (0.10) .453
Group ✕ Temporal Context ✕ Grammaticality 0.05 (0.10) .578 0.28 (0.12) .026

L1 Mandarin
Intercept -0.09 (0.03) .009 0.01 (0.04) .749
Temporal Context 0.01 (0.04) .786 -0.05 (0.08) .519
Grammaticality -0.09 (0.04) .056 0.03 (0.08) .695
Temporal Context ✕ Grammaticality 0.01 (0.09) .940 -0.20 (0.15) .187
L1 English
Intercept 0.15 (0.02) <.001 -0.20 (0.03) <.001
Temporal Context -0.02 (0.03) .521 0.02 (0.05) .700
Grammaticality 0.11 (0.03) <.001 -0.00 (0.05) .969
Temporal Context ✕ Grammaticality 0.05 (0.05) .318 0.07 (0.09) .464

129
 Chapter 3

Between-experiment (Modality) Comparisons

As auditory (Experiment 4) and visual (Experiment 5) comprehension were

measured on different scales (RT for auditory comprehension and residualised RT
based on word length for visual / reading comprehension), direct model fitting using
Modality as a fixed effects predictor was not possible. Instead, an effect size measure
(Cohen’s d) was used as an indicator of cross-modality effects (Table 19). Crucially,
data from self-paced comprehension tasks showed a greater effect of Grammaticality
in the auditory (Experiment 4) than in the visual (Experiment 5) modality, indicating
that auditory stimuli elicited stronger sensitivity to inflectional omissions.
Interestingly, there was a greater between-group difference in the visual compared
with the auditory modality, with L1 Mandarin participants showing shorter response
RTs overall compared with L1 English participants in the visual modality.

Table 19.

Cohen's d effect size statistics across Experiment 4 and Experiment 5.

Cohen’s d
Experiment 4 Experiment 5
(N=61;57) (N=61;56)
Group 0.07 1.16
Temporal Context 0.05 -0.00
Grammaticality 0.32 0.01
Group ✕ Temporal Context 0.07 -0.02
Group ✕ Grammaticality 0.15 0.13
Temporal Context ✕ Grammaticality -0.09 0.02
Group ✕ Temporal Context ✕ Grammaticality -0.02 0.02

3.3.3. Interim discussion

Experiment 5 examined L2 temporal comprehension in L1 Mandarin learners of

L2 English in the visual modality. There were several findings of interest:
Surprisingly, the L1 Mandarin group did not show sensitivity to inflectional

130
 Chapter 3

omissions at a statistically significant level in the critical verb segment in the visual
modality, unlike the L1 English group who showed such sensitivity. The spill-over
region did not show any delayed Grammaticality effects. Similar to Experiment 4,
we did not see any effects of Temporal Context or its interaction with
Grammaticality for the L1 Mandarin group, indicating no differential sensitivity to
3SG -s and past -ed omissions in the visual modality. Interestingly, unlike their
performance in the self-paced listening task in Experiment 4, the L1 English group
did not show a significant main effect of Temporal Context nor an interaction with
Grammaticality, indicating no differential sensitivity to 3SG -s and past -ed
omissions.

Comparing across modalities, the L1 Mandarin group was not sensitive to

inflectional omissions to a statistically significant level in the visual modality
compared with the auditory modality, irrespective of inflectional type, unlike the L1
English group who did for both auditory and visual modalities.

3.4. General Discussion

Previous research has shown that L2 learners generally do not process L2

sentences in a native-like way. Given current evidence on L2 sentence processing, it
is unclear 1) whether L2 learners can consistently assimilate information from L1-
absent linguistic cues during real-time L2 comprehension, and therefore exhibit
sensitivity (or experience processing difficulties) when they encounter mismatches
between lexical and morphosyntactic cues; 2) whether processing of cues is
recognisably context dependent; 3) whether sensitivity to mismatches between
lexical and morphosyntactic cues during L2 sentence processing is affected by
comprehension modality in a significant way.

In Experiments 4 and 5, L1 Mandarin and L1 English participants either listened

to or read sentences with and without inflectional omissions on the main verb in a
self-paced moving-window paradigm (in which inflectional omissions were always
associated with ungrammaticality). Unsurprisingly, L1 Mandarin participants

131
 Chapter 3

exhibited non-native-like sensitivity towards inflectional omissions on the critical

verb segment in L2 comprehension. In auditory comprehension, whilst L1 Mandarin
participants clearly exhibited sensitivity towards inflectional omissions like L1
English participants, they did not show stronger sensitivity towards 3SG -s omissions
compared with past -ed omissions, which was evident among L1 English
participants. In visual (reading) comprehension, L1 Mandarin participants did not
exhibit sensitivity towards inflectional omissions on the critical verb segment at a
statistically significant level compared with L1 English participants who did. There
were no spill-over effects in the object segment either. Effect size measures for
Grammaticality were larger in the auditory modality compared with the visual
modality, which indicated that an auditory presentation of stimuli gave rise to
stronger sensitivity in detecting inflectional omissions.

3.4.1. L2 comprehension mechanism and application of implicit knowledge

Revisiting the aims of the current study, one key aspect of our investigation was
to see if L2 learners could acquire a native-like comprehension mechanism which
and apply L2 grammatical knowledge in real time despite grammatical properties of
the L1. In other words, whether L2 learners could exhibit real-time sensitivity if
mismatches between lexical and morphosyntactic cues occur (when the relevant
grammatical rules are violated). The theoretical question lies not with whether L2
learners understood the intended message (demonstrated via performance on
comprehension questions), nor whether they know the grammatical features on an
explicit level (demonstrated via performance on the morphological proficiency test),
but rather whether L2 learners incrementally assimilate information from multiple
linguistic cues and implicitly apply relevant grammatical knowledge in a native-like
way during L2 comprehension. If so, L2 learners should be sensitive to mismatches
between cues in a way similar to that in native-L1 speakers.

Our findings were mixed. Results from Experiment 4 provided convincing

evidence that L2 learners do have auditory sensitivity to missing inflections which
affected their processing of L2 sentences (as indicated by longer reaction times for

132
 Chapter 3

ungrammatical trials), indicating that they could incrementally assimilate information

from both lexical and morphosyntactic cues, and were able to apply grammatical
knowledge in real time. This evidence seemingly suggests that L1 Mandarin learners
of L2 English, with no inflectional morphology in their L1, can acquire a
comprehension mechanism (or implicit knowledge) that applies L2 grammatical
knowledge during real-time sentence processing. Contrary to Roberts and Liszka
(2013), the absence of the grammatical feature in the learners’ L1 did not prohibit
them from processing it in the L2. However, this did not necessarily mean that their
performance and their processing of L2 sentences were native-like. Note that whilst
L1 Mandarin participants exhibited sensitivity to inflectional omissions in general,
they did not exhibit stronger sensitivity when 3SG -s was omitted than when past -ed
was omitted in experimental sentences, which was evident in L1 English participants.
It is clear that L1 Mandarin participants did not process these inflections in a strictly
native-like fashion, and that the fact that 3SG -s requires more than one type of
agreement did not facilitate or hinder their sensitivity to its omission. Though it
could be argued that current findings are in favour of a fundamental difference
interpretation of L1 and L2 processing, current evidence does not conclusively rule
out a quantitative difference interpretation, especially when L2 learners are found to
be sensitive to both 3SG -s and past -ed inflectional omissions in the auditory
modality overall.

Let us return to Bates and MacWhinney’s (1989) claim of cue validity and cue
strength, that L2 learners must adjust processing priority when L2 grammar has a
different hierarchy of form-to-function mappings. We hypothesised if L1 Mandarin
learners of L2 English could acquire new grammatical features, and assign value and
processing priority to these newly acquired L2 grammatical features as linguistic
cues, then they should in principle experience processing difficulties if these cues
gave inconsistent or contradictory information. It is clear from Experiment 4, that L2
learners were indeed sensitive to inflectional omissions, indicating that L2 learners
have assigned value and processing priority to inflectional morphology as a linguistic
cue for temporal information, even if it does not exist in the L1. Importantly, this
pattern occurred when the task did not explicitly require participants to monitor the

133
 Chapter 3

grammatical acceptability of the L2 stimuli, which showed that this is an implicitly

driven process.

Let us also reconsider the concept of learned attention under the associative
learning theory (Ellis & Wulff, 2008), which claimed that L2 learners direct their
attention to the most salient cues to meaning during initial acquisition. Findings from
Experiment 4 showed that at the current proficiency level (intermediate to advanced),
L2 learners no longer prioritised cues based on surface level saliency, even when
experimental sentences had fronted temporal adverbials and inflections occurred later
in the sentence. Instead, they carried out incremental parsing using all relevant cues
as required by L2 grammar, including less salient cues such as inflectional
morphology.

One shouldn’t ignore one interesting finding in Experiment 4, that sensitivity to

inflectional omissions not only differed across inflectional endings in the L1 English
controls, but was stronger for 3SG -s than for past -ed. This asymmetry indicated that
native-like processing is different for 3SG -s and past -ed inflections. One
interpretation of this difference could be down to the number of features (or the
number of agreement) contained within the inflection. Specifically, the inflection
requiring more than one type of agreement (i.e. 3SG -s) was more salient when
absent during real-time sentence processing than the inflection requiring only one
type of agreement (i.e. past -ed), resulting in increased sensitivity for the former in
L1 English participants. It is possible that as subject number is an important cue for
grammaticality in addition to temporal context for 3SG -s, the shorter distance
between the subject and inflectional morpheme compared with the temporal
adverbial contributed to stronger sensitivity for 3SG -s omission on the critical verb
segment.

Interestingly, this finding shows resemblance to findings by Roberts and Liszka

(2013), where L1 English controls showed sensitivity to temporal mismatches in the
present perfect condition but not in the past simple condition. Roberts and Liszka
suggested that different degrees of grammaticality could be at play (i.e. past simple
ungrammatical trials were more acceptable than present perfect ungrammatical
trials). In the context of the current study, omitting 3SG -s could be considered ‘more

134
 Chapter 3

ungrammatical’ than the omitting past -ed, giving rise to stronger 3SG -s sensitivity
in L1 English learners.

3.4.2. Modality effects in L2 comprehension

Let us now examine the most interesting finding from the current study.
According to previous research, there is an auditory disadvantage in L2
comprehension imposed by additional speech segmentation (Johnson, 1992; Murphy,
1997). Specifically, L2 English learners have been found to show superior
performance in grammaticality judgement when the L2 stimuli is presented in the
visual rather than the auditory form. Findings from Experiment 5 seemingly
contradicted this claim. With identical stimuli to Experiment 4, the L1 Mandarin
group did not exhibit visual sensitivity to inflectional omissions at a statistically
significant level. This showed that the auditory nature of stimuli did not invariably
make L2 comprehension more difficult. In fact, phonological saliency facilitated
assimilation and integration of L2 linguistic cues.

How could we explain the different modality effects found in previous studies
and in the current study? It is possible that during real-time comprehension, visual
text could take longer to process compared with auditory stimuli. However, it should
be noted that as both stimuli from self-paced listening and self-paced reading tasks
are readily segmented, the argument of additional speech segmentation increasing
processing difficulty does not apply here (Anderson, 1980). Instead, current data
seem to suggest greater perceptual saliency for auditory stimuli in the context of L2
sentence comprehension, facilitating sensitivity to L2 inflectional omissions. With
regard to the facilitatory effect of auditory cues, L1 Mandarin participants might
have been facilitated by the presence of additional syllables created by inflectional
morphemes. However, given the fact that the verbs in these experiments contained
alveolar endings, they would only facilitate comprehension in the past -ed context
(syllabic endings, as in shouted), and not in the 3SG -s context (consonant cluster
endings, as in kicks). If phonological saliency was the only contributor, there should
have been significant differences between sensitivity to syllabic endings for past -ed

135
 Chapter 3

and consonant cluster endings for 3SG -s in L1 Mandarin participants. However, this
was not the case, hence we could not attribute the observed effects solely to syllabic
features in the auditory stimuli. Moreover, I argue that given the semantically driven
nature of the current experiments, it is also possible that L2 learners have adopted a
processing strategy to prioritise semantic information (i.e. verbs) over syntactic
features (i.e. inflectional morphemes) when L2 input is not perceptually salient (i.e.
in the visual modality). In other words, L2 learners could, implicitly or explicitly,
exhibit different levels of grammatical sensitivity depending on the purpose of L2
comprehension.

3.4.3. Limitations and remaining issues

These experiments provide clear evidence for the integration of linguistic cues
during temporal processing in auditory L2 comprehension, and that phonological
saliency could facilitate integration of linguistic cues compared with visual / reading
comprehension. However, it is not clear how phonological features facilitate the
detection of grammatical violations in the auditory modality. Do L2 learners find
auditory cues more salient across-the-board? Or do they show perceptual bias
towards phonological features which exists in the L1 (syllabic endings, as in [tɪd] in
shouted) compared with those which do not (consonant clusters, as in [ks] in kicks)?
The extent of phonological influence will be addressed in Chapter 4.

Given that multiple studies have shown L2 learners possess visual sensitivity to
L2 grammatical violations, I do not dismiss that this sensitivity exists amongst L2
learners. Rather, I propose that task demands play an important role in measures of
L2 learners’ grammatical sensitivity. In a semantically driven task without
perceptually salient cues, or under cognitive stress, L2 learners could prioritise
semantic over syntactic cues during real-time sentence processing. However, in a
syntactically driven task, where the primary aim is to monitor for grammatical
violations, L2 learners could exhibit superior sensitivity towards syntactic cues. This
possibility could be confirmed by repeating the current experiments where
comprehension questions are replaced with a grammaticality judgement task (e.g.
Jackson & Bobb, 2009; Jackson & Dussias, 2009). Taking the findings of the current

136
 Chapter 3

experiments, it is clear that L2 learners can assign processing priority to novel L2

grammatical features and integrate multiple linguistic cues during L2
comprehension, even if these cues do not exist in the learner’s L1. Differences across
auditory and visual modalities raise question about the level of uniformity in L2
grammatical processing, particularly, whether L2 learners’ sensitivity to grammatical
violations change, depending on the nature of stimuli and task demands.

Let us return to the point about the nativelikeness of grammatical (or

morphology) processing in L2 learners during L2 sentence processing and the extent
of L1 effects. Unlike previous studies which contrasted L2 learners from multiple L1
groups (Marinis et al., 2005; Papadopoulou & Clahsen, 2006; Roberts & Liszka,
2013 etc.), the current study only used a single group of L2 learners when
investigating L2 sentence processing. Whilst it is clear that sensitivity to inflectional
omissions was evident in the auditory modality, without other L1 groups for
comparison, the extent to which L1 effects affect real-time L2 sentence processing
cannot be comprehensively discussed.

To summarise, this chapter addressed the question of cue processing during real-
time L2 sentence processing, specifically, whether L2 learners could acquire an L2
comprehension mechanism to incrementally apply explicit grammatical knowledge
to linguistic cues when such features are absent in the L1. Current evidence in
auditory comprehension suggests that this is possible. However, L2 learners did not
behave like native-L1 speakers and process inflections with different numbers of
features differently. Critically, their sensitivity was not uniform across
comprehension modalities. In the current semantic-oriented task where both auditory
speech and visual text are readily segmented, auditory cues had a facilitatory effect
on L2 learners’ sensitivity to inflectional omissions.

137
 Chapter 4

Chapter 4

L1 phonological influence on L2 comprehension and

production

Phonological influence has been implicated as one of the key factors affecting
L2 comprehension and production accuracy (Best, 1995; Flege, 1995; Bayley, 1996;
Hawkins & Liszka, 2003). Specifically, L2 comprehension is contingent on the
accurate perception of L2 phonological features, and L2 production relies on the
correct generation of phoneme sequences for articulation. The current study seeks to
examine the extent of phonological influence on L1 Mandarin speakers acquiring
English inflectional morphology, a grammatical feature which is absent in Mandarin
and is phonologically variable in English. This study tested the following
predictions: 1) Perceiving distinctions in L2 speech sounds is more difficult if L1
experience induces perceptual biases which favour L1 over L2 phonological features;
2) Production of selective L2 speech sounds is more difficult if the learner’s L1 does
not allow such phoneme sequences. Using English 3rd person singular -s (3SG -s)
and past -ed inflectional morphemes, the findings revealed that L1 Mandarin learners
of L2 English did not show consistent perceptual biases towards 3SG -s and past -ed
inflections which were significantly different to native-L1 speakers under different
phonological contexts. Moreover, they processed information from inflectional
markings in in the absence of additional cues, just like native-L1 speakers. However,
L1 Mandarin learners of L2 English exhibited significantly less accuracy in the
production task (CTOPP-2; Phoneme Elision Task) where omission and adjunction
of L2 phonemes were required, indicating strong L1 phonological influence on
production. Overall, given the absence of certain L2 phonological features in the L1,
L1 phonological influence has been found to affect L2 production more than L2
comprehension.

138
 Chapter 4

4.1. Introduction

L1 phonological influence is an overarching factor affecting L2 comprehension

and production, but it is often unclear to what extent phonological factors affect
perception and overt articulation of L2 speech. In L2 comprehension, in order to
understand and interpret L2 grammatical features, L2 learners must perceive L2
phonological features correctly, as well as acquire the underlying grammatical
distinctions associated with these phonological features. The absence of grammatical
distinctions (e.g. tense for L1 Mandarin learners of L2 English) and / or phonological
features (e.g. consonant clusters) in the learner’s L1 may hinder L2 learners’ ability
to effectively comprehend L2 speech. Previous research has shown mixed findings
regarding the facilitatory effect of auditory cues in detecting L2 grammatical
violations (Johnson, 1992; Murphy, 1997; Chapter 3). This brings in the question of
whether differences between L1 and L2 phonological properties could affect L2
learners’ sensitivity to L2 grammatical features. Could L2 learners be more sensitive
to L2 speech sounds that are shared by their L1 than those which are not? If so, to
what extent does this sensitivity affect comprehension of novel L2 grammatical
features (e.g. inflectional morphology)?

In order to produce L2 grammatical features, L2 learners not only have to

acquire the underlying grammatical distinctions in the L2, but must also be able to
create the correct phonological structures to articulate these features. It is also well-
known that L2 learners often find it difficult to produce grammatical features which
do not exist in their L1 even after years of L2 immersion (Lardiere, 1998; 2000).
Previous research has attributed variations in L2 grammatical production to L1
phonological influences, where phonological structures not permissible in the L1 are
more likely to fail in production (see Goad, White & Steele, 2003 for a linguistic
explanation). I argue that in order to draw conclusions about L1 phonological
influences, one should also examine L1 phonological constraints outside of specific
grammatical contexts (e.g. omission of inflectional morphemes). Could L2 learners
also experience phonological difficulties in L2 production without such grammatical
features (e.g. in non-inflected words)?

139
 Chapter 4

Inflectional morphology carries both syntactic and semantic information (person,

number, tense etc.) and is suffixal in English. More importantly, the phonological
properties of inflectional morphemes vary depending on phonological context,
especially on the phonological properties of the verb. Therefore, in order to
successfully comprehend and produce L2 inflectional morphology, the L2 learner not
only has to acquire L2 inflectional morphology as a grammatical feature, but also
acquire the variety of L2 phonological features that realise these inflections, which
may or may not be phonologically permissible in the learner’s L1. Inflectional
morphology is particularly relevant to our discussion as it is absent in Mandarin
Chinese. Different from English, which uses inflectional morphology as well as
temporal adverbials to indicate temporal properties of events, Mandarin does not use
a system of inflectional morphology and relies on temporal adverbials and aspectual
markers to indicate temporal information (e.g. le, guo, see Chapter 1.5.1 for detailed
descriptions). Moreover, in contrast to English, which allows for word-final
consonant clusters (e.g. [ks] in kicks), Mandarin Chinese consists of mostly
monosyllabic morphemes, and word-final consonant clusters are rare.

To what extent do L1 phonological properties influence the comprehension and

production of L2 inflectional morphology? On the one hand, if there is a facilitatory
effect of phonological overlap between L1 and L2, then one would expect L2
learners to show superior sensitivity to phonological features shared by their L1 than
phonological features which are rare or absent in the L1. In the context of
comprehending L2 inflectional morphology, the implication would be that L2
learners would be more sensitive to inflectional omissions if the phonological
features of the inflections are shared by the L1 than if the phonological features are
rare or absent in the L1. On the other hand, if there is no facilitatory effect of L1
phonological features during L2 comprehension, L2 learners’ sensitivity to L2
phonological features would be unaffected by the phonological properties of the L1.
That is to say, L2 learners’ sensitivity to L2 inflectional omissions would be
independent of whether the phonological features of the inflections are shared by
their L1. If this is the case, any behavioural differences in sensitivity between L2
learners and native-L1 speakers during online L2 comprehension would be
attributable processes other than phonological overlap.

140
 Chapter 4

From a production perspective, there are three possibilities regarding the extent
of L1 phonological influence. First, if L2 learners cannot generate L2 phoneme
sequences which are not permissible in their L1, they could have problems
articulating grammatical features, where such phoneme sequences are necessary (e.g.
verbs with obligatory inflections; e.g. [kt] as in yesterday she walked). Alternatively,
they could have problems generating and articulating L2 phoneme sequences across
all contexts, including in non-grammatical contexts as well (e.g. adjunction of
phonemes inside a non-inflected word; e.g. [kt] as in cocktail). However, if L2
learners do not have problems generating L2 phoneme sequences in any way, then
errors in L2 inflectional production would again be attributable to problems other
than L1 phonological influences.

In this chapter, I examine the extent of L1 phonological influence on English

inflectional morphology in L1 Mandarin learners of L2 English. Whilst previous
research has shown that L1 Mandarin learners of L2 English exhibit auditory
sensitivity to inflectional omissions (Johnson & Newport, 1989; Chapter 3), it is not
clear how phonological features of inflectional morphology contribute to this
sensitivity. Here, I test whether L1 Mandarin learners of L2 English are more
sensitive to inflectional morphology with syllabic endings (permissible in L1
Mandarin) than consonant cluster endings (rarely permissible in L1 Mandarin).
Moreover, I test whether L1 Mandarin learners of L2 English exhibit sensitivity to
temporal information via inflectional morphology (absent in L1 Mandarin) without
other temporal cues (i.e. temporal adverbials) whilst controlling for phonological
features. Previous theories of L1 phonological influence focused on the constraints
the L1 imposes on the phonological hierarchy (or prosodic structure in linguistic
terms; see Goad et al. 2003), and their effects on the L2 learner’s ability to produce
specific L2 grammatical features. Here, I test the extent of L1 phonological
constraints on L2 production without implicating L2 grammatical features.
Specifically, I examine whether L1 Mandarin learners of L2 English experience
difficulties with phoneme adjunctions without inflectional morphemes (3SG -s or
past -ed).

141
 Chapter 4

To summarise, this study investigates the extent of L1 phonological influence on

L2 inflectional comprehension and production. This study examines not only
whether L1 Mandarin learners of L2 English can exhibit sensitivity to inflectional
morphemes with different types of phonological features, but also whether they
exhibit superior sensitivity to inflectional morphemes with L1 phonological features.
Additionally, this study examines whether phoneme adjunction affects L2 production
more generally by testing whether L1 Mandarin learners of L2 English could
experience production difficulties in non-inflected contexts.

4.1.1. L1 and L2 phonological development

Phonological factors play a critical role in language comprehension and

production. In order to understand auditory speech, the listener must first segment
auditory signals into smaller units, identify these units, and then map them onto
lexical concepts. Languages contain a variety of phonological properties. Relevant to
our discussion are phonological features (features of phonemes and phoneme
sequences) and phonotactic constraints (rules which phoneme sequences must obey).
In linguistic terms, phonological units are organised into a hierarchy, namely a
prosodic structure (Selkirk, 1980; McCarthy & Prince, 1995), on which language-
specific constraints are imposed.

The first step in acquiring the phonological properties of a language is learning

to perceive the phonological distinctions which mark semantic differences over those
which do not. With regard to the perceptual sensitivity of phonological features,
early research suggests that whilst L1 infants can make phonological distinctions not
present in their native language from birth (Eimas, Siqueland, Jusczyk., & Vigorito,
1971; Kuhl, 1987), they quickly lose this ability during the first year of life as they
become immersed in their L1 (Werker & Tees, 1984). More specifically, researchers
have found that infants exhibit stronger sensitivity to vowels which belong to their
native languages than those which do not (Kuhl, Williams, Lacerda, Stevens., &
Lindblom, 1992). This suggests that L1 infants’ perceptual system becomes more
attuned to relevant phonological distinctions as they are exposed to L1 linguistic

142
 Chapter 4

input. It is also well established that through immersive exposure, L1 speakers

become sensitive to phonotactic information of their native language at a very young
age (Friederici & Wessels, 1993). By using probabilistic information of phoneme
sequences, L1 speakers can distinguish words from non-words and establish word
boundaries when comprehending speech (Jusczyk et al., 1994; Mattys & Jusczyk,
2001). Later through formal language instruction, explicit knowledge of phonotactic
constraints on words, including rules of permissible phoneme sequences is also
applied to increase chances of successful comprehension.

As Cutler (2000) pointed out, language-specific processing requires L2 learners

to acquire a degree of listening competence in order to segment L2 speech
appropriately, so how do L2 learners acquire a new phonological system with
different phonological properties compared with their L1? On a perceptual level, it is
plausible that given the underlying principle of L1 phonological development, late
L2 learners could use L1 categories to interpret L2 phonological distinctions first,
and would only create new categories for the L2 after lengthy exposure (Flege,
1995). As a result, L2 learners may be insensitive to phonological distinctions which
are not semantically relevant in the learner’s L1. This is also consistent with the
claim that L2 learners interpret unfamiliar L2 speech sounds in terms of phonetic
(articulatory) similarity to their L1 (Best, 1995).

To what extent do L1 phonotactic regularities constrain the perception of L2

speech? Past research has suggested that L1 phonotactic constraints can lead to
misperception of L2 phoneme sequences. For example, Dupoux, Kakehi, Hirose,
Pallier and Mehler (1999) contrasted L1 French and L1 Japanese learners of English
in their perceptual sensitivity to English consonant vowel sequences, and found that
L1 Japanese learners are more likely to perceive ‘illusory’ vowels between
consonants than L1 French learners (presumably so that the sequence is compatible
with L1 phonotactics). Importantly, L1 Japanese learners experienced difficulties
discriminating consonant sequences with and without an intermittent vowel. Such
evidence suggests there is a fundamental perceptual deficit for some phoneme
sequences not permitted in the learner’s L1. In another study, Flege and Wang (1989)
found that L1 Chinese speakers’ sensitivity to English /t/ and /d/ contrasts were
affected by the variety of Chinese spoken. Specifically, L1 Mandarin speakers whose

143
 Chapter 4

L1 does not permit word-final stops (/t/ and /d/) performed significantly worse than
L1 Cantonese speakers11 whose L1 permits unreleased obstruents (/p, t, k/), though
sensitivity to /t/ - /d/ contrasts improved after training for both groups. Other findings
by Cutler and colleagues also showed that L2 learners’ speech segmentation strategy
is heavily contingent on the phonological properties of the L1. For example, as
English is not syllable based, English listeners do not use syllable structure as a
segmentation strategy to perceive French (Cutler, Mehler, Norris & Segui, 1986),
and similarly they do not use morae to perceive Japanese (Otake, Hatano, Cutler &
Mehler, 1993). However, Weber and Cutler (2006) also showed that L2 learners
could acquire L2 phonotactic probabilities when detecting embedded words in the
L2, though they were not able to prevent L1 interference entirely even at high L2
proficiency.

If we take this conclusion forward, then L2 learners should primarily use the
segmentation strategy of the L1 to perceive the L2, then by implication they should
find some phonological features more difficult to perceive than others, depending on
whether such features are permitted in the learner’s L1. For example, they may
divide syllables or consonant clusters into separate words when they form the suffix
of the same word. The main consequence of this type of segmentation error is that L2
learners may miss important semantic information at the site of error. Take the
sentence ‘She performed a dance’. If the L2 learner’s L1 does not allow for [md] as a
phoneme combination at the word final position, then [d] might not be interpreted as
an inflectional morpheme. Instead, [d] may be misallocated to the following word,
making the sentence sound more similar to ‘She perform the dance’. Consequently,
the temporal interpretation of the sentence would be fundamentally different from the
intended meaning (not accounting for insensitivity to 3SG -s omission).

If these findings are applied more extensively to consonant cluster and syllabic
endings created by other inflectional morphemes, one can postulate that L2 learners
could experience difficulties detecting the presence (or absence) of a phonological
feature which does not fit with the phonotactic constraints of their L1. For example,
in the sentence ‘The girl walks in the park’, if the learner’s L1 does not allow for [ks]

11 A spoken variety of Chinese which is phonologically distinct from Standard Mandarin.

144
 Chapter 4

as a phoneme sequence, then it is plausible that L2 learners would fail to detect the
difference between walks and walk. If we take a more lenient view, that L2 learners
could at least detect the presence of an inflectional morpheme with an L2
phonological feature, L2 learners could still be more sensitive to phonological
features which are more frequent in the learner’s L1 than those which are rare. For
example, L2 learners could exhibit more sensitivity to shouted (syllabic endings)
than walks (consonant cluster endings).

An argument to the contrary would be that L2 learners, regardless of the

phonotactic constraints of their L1, would show similar levels of sensitivity towards
different types of phonological features. One interesting study by Solt et al. (2004)
examined the L2 perception of the past -ed morpheme as [t], [d] and [ɪd] allophones
under different phonological contexts (as determined by phonological properties of
verbs). Verbs with past -ed morphemes were presented in a sentential context (e.g.
The girl walked in the park), where an L1-English ‘student’ repeated sentences (with
or without -ed on the verb) after an L1-English ‘teacher’. L2 learners were asked to
make a same vs. different response based on whether they perceived the verb
repetition by the student was the same or different to the ‘teacher’. Their findings
showed that L2 English learners from different L1 backgrounds were generally more
sensitive to -ed morphemes in syllabic contexts (e.g. [ɪd] in [ʃaʊtɪd] / shouted) than
as when they appeared as part of consonant clusters (e.g. [t] in [wɔːkt] / walked or [d]
in [kləʊzd] / closed). If these findings stand, that L2 learners perform similarly to
each other in L2 perception regardless of their respective L1s, one may speculate
whether L2 learners are fundamentally constrained by L1 phonological properties at
all, and whether their performance would be significantly different to native-L1
speakers in terms of perceptual sensitivity.

In production, Solt et al. (2004) found that L2 learners of English from multiple
L1 backgrounds were also most accurate in the condition with [ɪd] endings across
proficiency levels compared with [t] and [d] endings. This could suggest that the ease
of L2 production is determined more by intrinsic properties of L2 phonological
features rather than specific phonological constraints from the learners’ L1s.

145
 Chapter 4

To summarise, previous research commonly agree that L1 phonological

properties imposes perceptual and phonotactic constraints on L2 learners’ perception
and production of L2 speech as a result of L1 development and maturation. L2
learners typically exhibit perceptual biases against L2 phonological features which
are rare or absent in the L1. However, limited attempts have been made to explain
the behavioural evidence for these constraints. Specifically, the principles of
phonological acquisition and processing have not been extensively explored. In the
next section, I will briefly outline a few explanations of L1 phonological influence
on L2 comprehension and production from psycholinguistic and linguistic
perspectives.

4.1.2. Accounts of L1 phonological influence

How do psycholinguistic models explain phonological influences in L2

comprehension and production? On the establishment of single phonemes,
researchers from connectionist perspectives claim that L2 phonemes are acquired
based on the saliency of phonetic features (Hancin-Bhatt, 1994). In other words, the
most salient phonemes would be more easily perceived and learned. Whereas
Hancin-Bhatt (1994) defined saliency in terms of frequency of occurrence, we can
also consider L2 phonemes in terms of perceptual saliency. If L1 development
diminishes the L2 learners’ sensitivity towards phonological distinctions which do
not mark semantic distinctions, then L2 learners could, as a result, exhibit perceptual
biases against the relevant phonological distinctions in the L2. Alternatively, they
could use L1 phonological categories to process L2 sounds by default before
developing new phonological categories for the L2 (Best, 1995; Flege, 1995).

On the use of phonotactic information, connectionist models such as the

Bilingual Interactive Model of Lexical Access (BIMOLA; Grosjean; 1997; Léwy &
Grosjean, 2008) provides a detailed account of how bilinguals use phonotactic
information to recognise and process words from each of the two languages. The
model claims that whilst phonemes and words from each of the bilinguals’ languages
are stored as subsets of the same system, features are shared between the two

146
 Chapter 4

languages. Therefore, in order to identify the language of the input, the bilingual
must be able to use phonotactic information to identify and activate one subset and
inhibit another at phoneme and word levels. It is important to note that models such
as BIMOLA do not account for ‘unbalanced’ bilinguals, where the L2 learner may
still be acquiring the L2 phonotactic information through limited input. Given the
importance of phonotactic processing in the model, the lack of L2 exposure or a lack
of explicit knowledge of L2 phonotactic constraints could be detrimental to the
bilinguals’ ability to identify and breakdown L2 phonemes and words from auditory
signals. That is, if the L2 learner does not have enough phonotactic information from
the L2, he / she would have to rely on L1 phonotactics to interpret L2 phonological
features, which may result in comprehension errors.

The picture is more complicated for L1 phonological influence on L2

production. Specifically, existing theories propose different principles regarding the
form which L1 phonological constraints could take. Roelof’s (1997) WEAVER++ is
a dedicated model which details the stages of phonological encoding during speech
production. Important for our discussion is the process of syllabification, where
syllables from the phonological word take on the phonological context of the word or
phrase and are organised in accordance with the phonological rules of the language
spoken. If the same principles apply to L2 speech production, then the process of
syllabification is where L1 phonotactic constraints and rules of L1 phonology are
applied. This is a processing constraint where activations are facilitated or inhibited
depending on the rules of the language spoken. In a way, this is comparable to Goad
et al.’s (2003) Prosodic Transfer Hypothesis (PTH), which claimed that the L1
prosodic structure is transferred to L2 production and constrains L2 English learners’
ability to consistently produce inflectional morphology. More specific than claims
relating to syllabification in WEAVER++, PTH assumes that the generation of
phonological representations for speech must adhere to a theoretical hierarchy
imposed by the phonological rules of the language. For example, in order to produce
regular English inflectional morphology, the prosodic structure must allow for
adjunction to the prosodic word (see Figure 18).

147
 Chapter 4

Figure 18. Prosodic structure of Mandarin aspectual marking (3) and English inflectional
marking (4). Taken from Goad and White (2006).

In the example above, whereas the Mandarin perfective aspectual marker le

modifies the verb mai3 (to buy) inside the prosodic word to produce mai3 le5
(bought already), the English -ed inflection require adjunction outside the prosodic
word in order to produce helped. If the learner’s L1 (i.e. Mandarin) does not allow
for the adjunction structure for attaching inflectional morphemes to the prosodic
word, then the learner would not be able to generate the phonological representation
for an inflected form, leading to production failures. This stands in contrast with
irregular past forms where modification occurs inside the prosodic word (drink -
drank), which were found to be easier to produce for L1 Mandarin learners of L2
English (Wolfram, 1985; Bayley, 1996; Hawkins & Lizska, 2003). Different from
WEAVER++, Goad et al. (2003) stated that the transfer of L1 prosodic structures is a
representational issue, and that such constraints on acquiring L2 prosodic structure
originate with limited access to interlanguage representations. Note that the most
recent version of the PTH claims that prosodic transfer plays a constraining role in
L2 acquisition, but not a permanent one (see Goad & White, 2019, for review).
Moreover, whilst Goad et al. initially claimed prosodic constraints do not necessarily
affect L2 comprehension, recent evidence seems to suggest a role of L1 prosodic
transfer in comprehension errors, when no production is involved (see Lieberman,
2013).

Overall, it remains unclear whether L2 learners consistently exhibit perceptual

deficiency or bias for specific grammatical features as a result of L1 phonological
properties. Moreover, if L1 phonological influence affects L2 production more
generally, then difficulties with phoneme adjunction should be observable outside of
specific grammatical contexts (i.e. inflectional morphology). However, such
evidence has not yet been shown.

148
 Chapter 4

4.1.3. The current study

I present two experiments examining the extent of phonological influence on L2

learners’ perceptual sensitivity to L2 temporal inflectional morphemes in L2
comprehension, as well as adjunction of phoneme sequences in non-inflected
contexts in L2 production.

L1 Mandarin learners of L2 English (AoA > 5 years) and L1 English controls

participated in two experiments, each including two tasks: a Phonological
Discrimination (PD) Task using an auditory ABX paradigm (Clarks, 1982), and a
Phoneme Elision task (CTOPP-2; Wagner, Torgesen, Rashotte, & Pearson, 2013).
For the PD Task, participants identified test words (e.g. kicks) from a previous word
pair (bare verb - inflected verb; e.g. kick - kicks) in an auditory ABX paradigm, a
method often used to measure participants ability to discriminate between two
stimuli. Phonological Endings (consonant clusters vs. syllabic) were manipulated via
inflectional morpheme (3rd person singular -s; 3SG -s vs. past -ed) and verb. Note
that in an ABX paradigm, participants must identify which of the two stimuli was
presented as the test stimulus and respond (e.g. by pressing a key) as quickly as
possible. This task relies on participants’ ability to quickly discriminate between the
two stimuli in question. Hence, participants’ speed of response to the test stimulus
(reaction time, or RT in milliseconds) would be a measure of their sensitivity to the
differences between the two stimuli. In this paradigm, shorter RTs indicate stronger
sensitivity for the phonological difference in question, and longer RTs indicate
weaker, or absence of sensitivity to such differences. In this case, participants’ RTs
on the test words were taken as a measure of sensitivity to the phonological
distinctions between the bare and inflected verbs.

For the PE Task, participants completed the standardized Phoneme Elision Task
from CTOPP-2. For each of the 20 items in the test, participants deleted a specific
phoneme or phoneme sequence from an English word upon instruction and
articulated the post-elision word out loud.

149
 Chapter 4

Given current empirical evidence for the extent of L1 phonological influence,

separate predictions have been made regarding the influence of L1 phonology on L2
comprehension and production.

In L2 comprehension, if L1 Mandarin learners of L2 English are constrained by

L1 phonotactics and are insensitive to phonological features absent in L1 Mandarin,
then it is plausible that they have a fundamental perceptual deficit with regard to
temporal morphemes 3SG -s and past -ed when presented as an inflection of a verb.
In which case, they should experience difficulties distinguishing inflected from non-
inflected forms, especially for 3SG -s and past -ed inflections which form consonant
cluster endings (e.g. kicks from kick - kicks, and closed from close - closed). That is,
one would expect longer test word RTs in L1 Mandarin learners of L2 English
compared to L1 English controls. If they do not have a perceptual bias regarding 3SG
-s and past -ed inflectional morphemes more than their native-L1 counterparts, they
should exhibit similar sensitivity to L1 English controls on test words. That is, one
would expect no significant differences in test word RTs between L1 Mandarin and
L1 English participants. Finally, if L1 Mandarin learners of L2 English could
perceive L2 phonological features (consonant clusters vs. syllabic) created by
inflectional morphemes 3SG -s and past -ed, but are still influenced by L1
phonological properties, L1 Mandarin learners could exhibit perceptual biases
favouring sensitivity L1 phonological features and against L2 phonological features.
That is, they may be more sensitive to syllabic endings compared to consonant
cluster endings relative to L1 English controls. In which case, we should see L1
Mandarin learners exhibit significantly shorter test word RTs than L1 English
controls for syllabic endings irrespective of the inflectional morpheme (3SG -s or
past -ed).

In L2 production, if the constraint on phoneme adjunction is a general one

extending beyond inflectional morphemes, then similar difficulties should also be
found in non-inflected contexts. In the PE task, if L1 Mandarin speakers have
difficulty performing phoneme adjunctions within a word, then this would provide
evidence for general phonological constraint in L2 English production beyond
specific grammatical features.

150
 Chapter 4

4.2. Experiment 6

4.2.1. Methods

Participants

55 L1 Mandarin learners of L2 English aged 19-28 (M=23.52; SD=1.72) and 41

L1 English speakers aged 19-32 (M=23.50; SD=3.60) took part in Experiment 1. L1
Mandarin speakers were postgraduate students at the University of Edinburgh who
were within two years of first arrival in the UK. All L1 Mandarin participants
acquired English after the age of five (AoA = 6+), and had obtained at least a score
of 6.5 overall on the IELTS English proficiency exam, with no less than a score of 6
on the listening component (see Appendix L for summary). L1 English participants
were monolingual speakers who did not have extensive exposure to other L2s before
the age of five.

Materials

For the Phonological Discrimination (PD) Task, 20 English verbs were chosen
based on their phonetic properties when attached to 3SG -s and past -ed inflectional
endings (see Appendix M). Verb Set A consisted of ten verbs which formed
consonant cluster endings when attached to 3SG -s, and syllabic endings when
attached to past -ed; Verb Set B consisted of ten verbs which form consonant cluster
endings in both cases. For this task, all bare verbs were paired with their
corresponding inflected forms for each trial. Order of presentation for the two verb
forms in each trial was counterbalanced for 3SG -s and past -ed endings, then rotated
around two lists using a Latin square design. The verb form for the test word (bare or
inflected) was also counterbalanced for different presentation order across the two
lists. All verbs forms were recorded in a recording studio using the voice of a British
English speaker in .wav format. E-Prime was programmed for the auditory
presentation of stimuli, and a desktop computer with a 1920 x 1080 pixels screen and
a pair of stereo headphones were used to deliver the recordings.

151
 Chapter 4

The Phoneme Elision (PE) Task was taken from the Comprehensive Test of
Phonological Processing (CTOPP-2; Wagner et al., 2013) to assess adult
participants’ phonemic awareness (ability to manipulate phonemes) in English. The
test consisted five practice items and 20 test items. Of the 20 test items, nine items
required phoneme elision (or deletion) at word boundaries (the beginning or the end
of words), with another eleven items requiring phoneme elision in the middle of the
words (see Appendix N for details). The test was adapted for audio presentation
using the voice of a British English speaker as the instructor. For each item in the
test, the speaker provided instructions for each step in the task (word repetition
followed by phoneme elision), with feedback on the first ten items (five practice
items and five test items). Audio recordings were then edited, adding 2000 ms of
silence after each instruction to allow time for participant response. All recordings
were normalised for volume and stored in .wav format to preserve audio quality.

Design

The PE Task used a mixed-design with Group (L1 Mandarin vs. L1 English;
between-subjects), Temporal Context (Present Habitual vs. Past; within-subjects) and
Phonological Ending (consonant cluster vs. syllabic; within-subjects) as fixed-effects
predictors, and response RT on the test word (ms) as the outcome variable. The PE
Task used a mixed design also, with Group and Place of Elision (word boundary vs.
mid-word; within-subjects) as fixed effects predictors, and response accuracy
(correct vs. incorrect) as the outcome variable.

Procedure

Upon arrival, participants were first asked to read the experiment instructions
and provide consent for their data to be used. They also provided demographic
details and language background information, including details of L2 acquisition for
the L1 Mandarin group (IELTS scores, AoA for L2, length of stay in months, and
daily L2 exposure in hours). Afterwards, participants were asked to explain the
experimental tasks back to the experimenter before commencing the experiment.

152
 Chapter 4

Participants were reminded that they must respond as soon as possible after hearing
the test word.

Figure 19. Experiment 6: Trial procedure for the Phonological Discrimination Task (beep -
word 1 – 1000ms break – word two – 1000ms break – test word).

For each trial in the PD Task, participants first heard a beep, followed by two
words. One of the two words was then played as the test word, after which
participants must identify the test word by pressing one of two number keys ([1] for
the first word, or [2] for the second word). Participants were given a maximum of
2500 ms to respond, after which the next trial would commence (see Figure 19).
Participants listened to four blocks of 20 trials, with optional breaks between each
block. Trial order was randomised, with each verb appearing only once in each
block. Response RT for each test word was recorded by E-Studio 2.0 for subsequent
analyses.

For the PE Task, the experimenter first explained the principle of the task to the
participants with two emphases: 1) participants must focus on the pronunciation of
each word, omitting sounds, not letters; 2) all words were known English words (i.e.
participants are not expected to make up non-words). Participants were then given
several examples of phoneme elision, none of which featured in the test items. Audio
recordings were played to the participant one by one in the same order as in the
original CTOPP-2 PE Task. Recordings were paused after each line of instruction to

153
 Chapter 4

give participants time to respond. For each item in the test, participants must first
repeat an English word (e.g. instructor: say ‘text’). Then, the participants must omit a
specific phoneme from the given word, and pronounce the remaining parts of the
word according to instruction (e.g. instructor: say ‘text’ without saying /k/). Several
changes were made to the original PE test procedure: 1) participants listened to all
items via audio recording in British English instead of American English; 2) all
participants listened to all items regardless of response accuracy; 3) feedback was
provided on the first ten items regardless of response accuracy. Participants’ oral
response to each item in the task was recorded in wav. format for subsequent
analyses.

Each experimental session lasted 30-35 minutes. Participants were paid five
pounds in cash for their time.

Coding and Scoring

Phonological Discrimination (PD) Task

E-Studio 2.0 generated raw reaction times (RTs) from the end of the test word
recording to the point of response. 2% of data were excluded as participants
incorrectly identified the test word on these trials. RT data above 1000ms for this
task were considered as outliers, and RT data below 200ms could not be reliably
attributed to intentional behavioural responses. A further 4% of data were excluded
based on these criteria.

Phoneme Elision (PE) Task

Oral production data from all participants were coded using three separate
coding criteria. For each item in the PE task, participants’ repetition, omission and
adjunction accuracy were coded as binomial data (correct, incorrect). Repetition
accuracy was defined by whether the participants’ repetition of the word was
identifiable as the original word (accounting for non-native like pronunciations).
Omission accuracy was defined by whether the target phoneme, and only the target
phoneme was accurately deleted from the original word. For example, sit and spit

154
 Chapter 4

would both be scored as incorrect if the trial required participants to omit /p/ from
split (slit). Adjunction accuracy applied only to mid-word phoneme elisions, and was
defined by whether the two remaining parts of the word were blended together in
oral production with no audible gaps. For example, win and win - er would both be
scored as incorrect if the trial required participant to omit /t/ from winter (winner).

4.2.2. Results

Phonological Discrimination (PD) Task

For the PD Task, analyses were conducted on the effect of Group, Temporal
Context and Phonological Ending on participants’ reaction time (RT) on the test
word. General linear mixed effects models (GLMMs) were constructed using a
forward model building strategy with a maximal random effects structure (Barr et al.,
2013). Group (between-subjects) was included in the model with either Temporal
Context or Phonological Ending (within-subjects) as fixed-effects predictors, and
Participant, (Verb) Item and Trial Order were included as random effects if they
significantly improved model fit. All categorical predictors were contrast-coded and
centred before they were entered into the model. Note that only data from accurate
responses trials were included in these analyses (3% of response data had inaccurate
responses and were excluded).

There were two parts to the analyses. First, the effect of Phonological Ending on
test word RT was analysed with Group and Phonological Ending as fixed-effects
predictors, also controlling for (Past) Temporal Context. As past -ed produced
syllabic endings with Verb Set A (e.g. shouted) and consonant cluster endings with
Verb Set B (e.g. chased), only trials from the Past temporal context was suited to this
analysis. Second, the effect of Temporal Context on test word RT was analysed using
only one set of verbs with Group and Temporal Context as fixed-effects predictors,
controlling for (consonant cluster) Phonological Ending. Verb Set B contained verbs
which shared consonant cluster endings for both Present Habitual (3SG -s) and Past
(-ed) inflectional morphemes (e.g. attacks, attacked), and was therefore suited to this

155
 Chapter 4

particular analysis. From these two analyses, the effects of Temporal Context and
Phonological Ending can be isolated respectively.

Phonological Ending Effects

RTs on test words with consonant cluster (Verb Set B) or syllabic endings (Verb
Set A) in the Past temporal context were analysed for L1 Mandarin and L1 English
groups. Note that (Verb) Item effects were controlled by adding Item as a random
intercept in the GLMM.

Figure 20. Experiment 6: Average reaction Time (ms) on test words across Consonant
Cluster and Syllabic endings in the Past temporal context (-ed) for L1 Mandarin and L1
English groups (N=55;41).

There was a significant main effect of Group (Table 20). L1 Mandarin and L1
English groups differed significantly in their RTs on the test words overall (M=
360.39 vs. M= 393.13). There was also a significant main effect of Phonological
Ending. Test words with consonant cluster endings in the Past temporal context
produced significantly longer RTs than those with a syllabic ending in both L1
Mandarin and L1 English groups (M = 405.13 vs. M = 340.28; Figure 20).
Interestingly, there was also a significant interaction between Group and
Phonological Ending, indicating that L1 Mandarin and L1 English groups responded

156
 Chapter 4

differently to test words with consonant cluster and to words with syllabic endings.
This prompted further subgroup analyses on each participant group. In addition to
confirming the significant effect of Phonological Ending for both L1 Mandarin and
L1 English groups (Table 20), subgroup analyses also revealed that L1 Mandarin
participants exhibited greater differences in RT on tests words with a consonant
cluster ending and with a syllabic ending, indicating stronger sensitivity to test words
with syllabic endings (in the Past temporal context).

Table 20.

General linear-mixed effects model (GLMM) statistics for Phonological Ending Effects
analysis in the Phonological Discrimination task for L1 Mandarin and L1 English groups
(N=55;41).
B (SE) p
Intercept 375.92 (8.81) <.001
Group (L1 Mandarin vs L1 English) 34.42 (14.80) .022
Phonological Ending (Consonant Cluster vs Syllabic) -69.96 (10.95) <.001
Group ✕ Phonological Ending -30.22 (10.06) .003
L1 Mandarin
Intercept 362.16 (10.58) <.001
Phonological Ending -82.00 (9.98) <.001
L1 English
Intercept 396.29 (12.31) <.001
Phonological Ending -51.73 (13.66) .001

* bold italic indicates reference levels

Temporal Context Effects

Response RTs on test words involving a consonant cluster distinction for both
3SG -s (Present Habitual) and past -ed (Past) were analysed for L1 Mandarin and L1
English groups (Figure 21).

There was no significant main effect of Group. Overall, the L1 Mandarin group
had shorter RTs on the test word compared to the L1 English group (M= 386.84 vs.

157
 Chapter 4

M= 409.51), but the two groups did not differ significantly (Table 21). Importantly,
there was a significant main effect of Temporal Context (or inflectional ending).
Both L1 Mandarin and L1 English groups produced shorter RTs for 3SG -s (Present
Habitual) than for past -ed (Past) (Figure 19). Group and Temporal Context did not
show significant interaction, indicating Temporal Context did not affect the L1
Mandarin and L1 English groups differently.

Figure 21. Experiment 6: Average reaction time (ms) on test words with Consonant Cluster
endings across Present Habitual (3SG -s) and Past (-ed) temporal contexts for L1 Mandarin
and L1 English groups (N=55;41).

Table 21.
Experiment 6: General linear-mixed effects model (GLMM) statistics for Temporal Context
Effects analysis in the Phonological Discrimination task for L1 Mandarin and L1 English
groups (N=55;41).

B (SE) p
Intercept 401.88 (9.24) <.001
Group (L1 Mandarin vs L1 English) 25.21 (16.47) .129
Temporal Context (Present Habitual vs Past) 18.23 (4.56) <.001
Group ✕ Temporal Context -9.05 (9.27) .329

* bold italic indicates reference levels

158
 Chapter 4

Phoneme Elision (PE) Task

For the PE Task, separate analyses were conducted for participants’ accuracy on
each of the three response types: repetition, elision and adjunction. Generalised
logistic mixed-effects models (GLMMs) were constructed to analyse the likelihood
of each type of response. For repetition accuracy, Group was used as the only fixed-
effects predictor with Participant and Item as random intercepts. For omission
accuracy, both Group and Place of Elision (PoE) were used as fixed-effects
predictors, and Participant, Item as random intercepts if they improved model fit.
Note that whilst all 20 items of the PE task required phoneme elision, only eleven out
of 20 items from the PE task required phoneme adjunction. Therefore, analysis on
adjunction accuracy was only run on data for these eleven items.

Figure 22. Experiment 6: Repetition, Elision and Adjunction accuracy in the Phoneme
Elision Task for L1 Mandarin and L1 English groups (N=55;41).

For repetition accuracy, there was no main effect of Group (Table 22). Overall,
the L1 Mandarin group was not significantly less likely to make an accurate response
than the L1 English group when asked to repeat a word in the PE task (M= 0.93 vs.
M= 1.00; Figure 22). However, there were several items which were consistently
misperceived in the L1 Mandarin group. Their phonological properties will be
discussed later on.

159
 Chapter 4
Table 22.

Experiment 6: Generalised logistic mixed-effects model (GLMM) statistics for the Phoneme
Elision task across L1 Mandarin and L1 English groups (N=55;41).
Repetition Elision Adjunction
B (SE) p B (SE) p B (SE) p
Intercept 14.20 .598 2.38 <.001 0.75 .058
(0.32) (0.40)
(26.94)
Group 21.64 .722 2.18 <.001 2.28 <.001
(0.28) (0.26)
(L1 Mandarin vs. (60.92)
L1 English)
Place of Elision - - -3.01 <.001 - -
(0.63)
(Word Boundary vs.
Mid-word)
Group ✕ - - 0.21 .677 - -
(0.50)
Place of Elision

* bold italic indicates reference levels.

For elision accuracy, there was an effect of Group (Table 22). Overall, the L1
Mandarin group was significantly less likely to delete the correct phoneme than the
L1 English group for items in the PE task (M= 0.69 vs. M= 0.90; Figure 22).
Interestingly, there was also an effect of PoE. Participants were significantly more
likely to produce accurate elision responses for items with word boundary elisions (at
the beginning or at the end of the word) than for items with mid-word elisions (M=
0.95 vs. M= 0.65). There was no significant interaction between Group and PoE.

For adjunction accuracy, there was an effect of Group (Table 22). Overall, the
L1 Mandarin group was significantly less likely to produce accurate adjunction
responses than the L1 English group (M= 0.46 vs. M= 0.81; Figure 22).

160
 Chapter 4

4.2.3. Interim Discussion

In Experiment 6, there were two key findings regarding the effect of

phonological feature on perceptual sensitivity to inflected verb forms. First,
controlling for temporal context (Past), both L1 Mandarin and L1 English
participants exhibited stronger sensitivity on test words with a syllabic ending than
those with a consonant cluster ending, indicating that consonant clusters were more
difficult to process irrespective of the listener’s L1. Critically, data showed that L1
Mandarin participants exhibited superior sensitivity towards syllabic endings
compared with the L1 English controls. Controlling for Phonological Ending
(consonant cluster), both L1 Mandarin and L1 English participants exhibited stronger
sensitivity to test words in the Present Habitual temporal context (3SG -s) compared
with the Past (-ed) context with no between-group differences. This indicated that L1
Mandarin learners of L2 English assimilated temporal (and subject number)
information from inflectional morphemes like L1 English controls, even when the
resulting phonological feature was rarely permissible in their L1.

The PE Task showed L1 Mandarin learners of L2 English have substantial

difficulties performing phoneme adjunctions within non-inflected words, pointing
towards a generalised difficulty with processing and articulating L2 phoneme
sequences.

Experiment 6 is not without problems. The choice of verbs for the PD Task
meant that the analyses could not have a set of balanced conditions for Temporal
Context and Phonological Ending. Hence, the Temporal Context effect analysis was
only conducted on one set of verbs (Verb Set B) but not the other, and the
Phonological Ending effect analysis was only conducted for the Past but not for the
Present Habitual temporal context. For each result, a point of comparison was not
available to eliminate temporal context specific or item specific effects.
Consequently, for Experiment 7, Verb Set B was replaced by another set of ten verbs
(Verb Set C) to counterbalance Verb Set A.

161
 Chapter 4

4.3. Experiment 7

4.3.1. Methods

Participants

42 newly recruited L1 Mandarin learners of L2 English aged 18-31 (M=21.69;

SD=2.71) and 43 newly recruited L1 English speakers aged 18-37 (M=21.523;
SD=3.84) took part in Experiment 7. All recruitment criteria were identical to those
in Experiment 6.

Materials

Verb Set A was taken from Experiment 6, but Verb Set B was replaced by Verb
Set C, which consisted of ten verbs which form syllabic endings when attached to
3SG -s inflections and consonant cluster endings when attached to past -ed
inflections (chases, chased; see Appendix M for full list). Materials for the PE task
were identical to those used in Experiment 6. No changes were made.

Design

Aside from the balancing of experimental conditions across Temporal Context

and Phonological Ending, other aspects of experimental design were identical to that
of Experiment 6.

Procedure

All aspects of the experimental procedure were identical to Experiment 6.

Coding and Scoring

All coding and scoring procedures were identical to Experiment 6 for both PD
and PE Tasks.

162
 Chapter 4

4.3.2. Results

Phonological Discrimination (PD) Task

For the PD Task in Experiment 7, Group (between-subjects), Temporal Context

and Phonological Ending (within-subjects) were used as fixed effects predictors, and
Participant, Item and Trial Order as random intercepts in the main GLMM. Our
model building strategy was identical to that of Experiment 6. As Experiment 7
contained a balanced experimental design, a single GLMM was constructed to
examine both Temporal Context and Phonological Ending effects on response RT.
Only RTs from accurate responses from the PD Task in Experiment 7 were
examined.

Figure 23. Experiment 7: Average reaction time (ms) for Consonant Cluster and Syllabic
endings across Present Habitual (3SG -s) and Past (-ed) temporal contexts in the
Phonological Discrimination Task for L1 Mandarin and L1 English groups (N=42;43).

There was no significant main effect of Group (Table 23). L1 Mandarin and L1
English groups did not significantly differ in their response RT on the test word (M=
352.52 vs. M= 356.58; Figure 23). Similar to Experiment 6, there was a significant
main effect of Temporal Context (or inflectional ending). Irrespective of Group,
participants produced shorter RTs in the Present Habitual (3SG -s) context than in
the Past (-ed) temporal context (M= 347.18 vs. M= 361.62). Critically, similar to
results from Experiment 6, there was a significant effect of Phonological Ending.

163
 Chapter 4

Participants from both L1 Mandarin and L1 English groups produced longer RTs on
test words with consonant cluster endings compared with words with syllabic
endings (Figure 23). A three-way interaction was found between Group, Temporal
Context and Phonological Ending.

Table 23.
Experiment 7: General linear-mixed effects model (GLMM) statistics for test word reaction
time in the Phonological Discrimination task for L1 Mandarin and L1 English groups
(N=42;43).
B (SE) p
Intercept 357.954 (7.71) <.001

Group (L1 Mandarin vs. L1 English) 2.596 (14.57) .859

Temporal Context (Present Habitual vs. Past) 17.420 (3.55) <.001

Phonological Ending (Consonant Cluster vs. -61.72 (3.55) <.001

Syllabic)
Group ✕ Temporal Context 3.79 (7.08) .593

Group ✕ Phonological Ending -5.98 (7.09) .399

Temporal Context ✕ Phonological Ending 20.61 (12.33) .112

Group ✕ Temporal Context ✕ Phonological Ending 28.41 (14.17) .045

* bold italic indicates reference levels.

164
 Chapter 4

Phoneme Elision (PE) Task

Identical analyses from Experiment 6 were carried out for participants’

repetition, elision and adjunction accuracy on the PE task. All fixed effects predictors
and random intercepts were identical to the PE analyses in Experiment 6.

Figure 24. Experiment 6: Repetition, Elision and Adjunction accuracy in the Phoneme
Elision Task for L1 Mandarin and L1 English groups (N=42;43).

For repetition accuracy, there was no significant main effect of Group (Table
24). Similar to Experiment 1, both L1 Mandarin and L1 English groups performed at
ceiling levels, and the L1 Mandarin group was not significantly less likely to produce
accurate responses when asked to repeat a word in the PE task (M=0.98 vs. M=1.00;
Figure 24).

For elision accuracy, there was a significant main effect of Group (Table 24).
The L1 Mandarin group was significantly less likely to delete the correct phoneme
than the L1 English group for items in the PE task (M= 0.72 vs. M= 0.85; Figure 24).
There was also an effect of PoE. Similar to Experiment 6, phonemes at word
boundaries were significantly more likely to be deleted accurately compared to mid-
word phonemes (M= 0.95 vs. M= 0.66).

165
 Chapter 4

For adjunction accuracy, there was an effect of Group (Table 24). Once more,
the L1 Mandarin group was significantly less likely to produce an accurate
adjunction than the L1 English group (M= 0.53 vs. M= 0.76; Figure 24)

Table 24.

Experiment 7: Generalised logistic mixed-effects model (GLMM) statistics for the Phoneme
Elision task across L1 Mandarin and L1 English groups (N=42;43).

Repetition Elision Adjunction

B (SE) p B (SE) p B (SE) p
Intercept 20.23 .755 2.38 <.001 0.75 .058
(0.32) (0.40)
(64.93)
Group 23.10 .857 2.18 <.001 2.28 <.001
(0.28) (0.26)
(L1 Mandarin vs. (128.29)
L1 English)
Place of Elision - - -3.01 <.001 - -
(0.63)
(Word Boundary vs.
Mid-word)
Group ✕ - - 0.21 .667 - -
(0.50)
Place of Elision

* bold italic indicates reference levels.

4.3.3. Interim Discussion

Experiment 7 set out to replicate the findings from Experiment 6 with a balanced
design. The results were mixed. First, consistent with Experiment 6, Experiment 7
confirmed the finding that participants’ perceptual sensitivity to 3SG -s and past -ed
inflectional morphemes did not differ significantly, irrespective of the listener’s L1
(L1 Mandarin or L1 English). However, inconsistent with Experiment 6, L1
Mandarin participants did not exhibit superior sensitivity to L1 phonological features
(syllabic endings) than L1 English controls, which pointed to a lack of perceptual
biases contingent on L1 phonological properties. Second, Experiment 7 showed that
L1 Mandarin participants, like L1 English controls, were more sensitive to Present

166
 Chapter 4

Habitual 3SG -s than past -ed in syllabic phonological contexts as well as consonant
cluster contexts. This indicated that the L1 Mandarin participants in particular
processed temporal information via L2 inflectional morphology irrespective of
phonological context. Regarding the PE Task, consistent with Experiment 6,
Experiment 7 confirmed difficulties in phoneme adjunction in non-inflected contexts,
indicating a generalised difficulty with articulating L2 phoneme sequences not
restricted to adjunction of inflectional morphemes.

4.4. General Discussion

Previous research has proposed L1 phonological influence as a key factor during

L2 comprehension and production. However, it is unclear how L1 phonological
properties affect the perception and production of L2 speech. Given previous
research evidence, it is unclear whether L2 learners are more sensitive to
phonological features that are shared by their L1 than those that are not, and whether
perceptual sensitivity interacts with the processing of grammatical features.
Moreover, we do not know whether the principles involved in adjoining grammatical
features in L2 production would also apply outside specific grammatical contexts.

In two experiments, L1 Mandarin learners of L2 English and L1 English controls

completed Phonological Discrimination Tasks, where they differentiated bare verbs
from inflected verbs (and vice versa). They also performed Phoneme Elision Tasks,
where they deleted specific phoneme(s) from a non-inflected word. The findings
showed that L1 Mandarin learners of L2 English performed on apar with L1 English
controls when perceiving phonological features created by L2 inflectional
morphemes. This indicated that L1 Mandarin learners of L2 English did not have
perceptual biases for L2 phonological features that were different to L1 English
controls. Although Experiment 6 showed L1 Mandarin learners of L2 English
exhibited superior sensitivity on syllabic phonological endings, this was not
replicated in Experiment 7 with the addition of a different set of verbs, indicating no
reliable effect. In L2 production, L1 Mandarin learners of L2 English consistently
performed less accurately than L1 English controls when instructed to delete specific

167
 Chapter 4

phonemes from an L2 English word. Importantly, L1 Mandarin learners of L2

English were significantly less accurate than L1 English controls performing
phoneme adjunctions within a non-inflected word. This indicated a generalised
difficulty with articulating L2 phoneme sequences among L1 Mandarin participants,
which is not restricted to adjoining inflectional morphemes to verbs.

4.4.1. L2 perceptual biases and comprehension

Current findings show no reliable effect of L1 induced perceptual biases among

L2 learners. More specifically, as L2 learners did not exhibit phonological sensitivity
which were significantly different to their L1 counterparts on both phonological
features, these findings suggest that L1 phonological properties do not fundamentally
affect how L2 learners perceive L2 phonological features.

Looking back on existing theories on how L2 sounds are acquired (e.g. Best,
1995; Flege, 1995), current findings do not suggest that L2 learners interpreted L2
sounds based solely on L1 phonological distinctions. Rather, L2 learners were either
not consistently affected by L1 phonological distinctions, or have acquired the ability
to interpret L2 sounds in terms of L2 phonological distinctions (perhaps through L2
exposure). On a phonotactic level, L2 learners could detect the presence or absence
of a phonological feature (e.g. a consonant cluster), even if it is rarely permitted in
the learner’s L1. Note that L1 Mandarin participants showed a greater degree of
perceptual bias than L1 English participants towards syllabic -ed endings in
Experiment 6 but not in Experiment 7. Hence, one cannot categorically rule out L1
induced perceptual biases favouring L1 over L2 phonological features, only that
there was no consistent evidence for such biases.

To view the current findings in the context of psycholinguistic models of

bilingual lexical access (i.e. BIMOLA, Léwy & Grosjean, 2008), current findings
show that (late) L2 learners were able to appropriately use L2 phonotactic
information to access higher-level representations, even if the L2 learners do not
have equal mastery of L1 and L2. One of the key motivations for examining the
extent of L1 phonological influence is how it affects L2 learners’ perception to L2

168
 Chapter 4

grammatical features (i.e. inflectional morphology). One may argue that perhaps L2
learners simply detected a surface-level phonological difference without having
decomposed the words in terms of verb and inflectional morpheme. However, the
fact that L2 learners, like native-L1 speakers, exhibited sensitivity to temporal
contexts irrespective of phonological feature in the current study, pointed towards
higher-level temporal and / or syntactic information processing, indicating L2
comprehension beyond surface-level phonological sensitivity.

Moreover, current findings crucially showed that the idea of superior sensitivity
towards syllabic endings over consonant cluster endings was similar across native-L1
speakers and L2 learners, indicating that this perceptual bias is independent of the
listener’s L1. Consistent with Solt et al. (2004), these findings not only confirmed a
general processing advantage for the syllabic endings regardless of L1 background,
but also extended the scope of the finding beyond a single inflectional morpheme
(i.e. past -ed).

4.4.2. Extent of L1 phonological constraints on L2 production

How can we interpret adjunction failures within non-inflected words? The

Prosodic Transfer Hypothesis claimed that L1 prosodic structure constrains L2
learners’ ability to produce the phonological representations for inflected words
(Goad et al., 2003; Goad & White, 2009). According to PTH, inflectional omissions
are caused not by an absence of L2 grammatical representations among L1 Mandarin
learners of L2 English, but rather an inability to adjoin inflectional morphemes to the
prosodic word (according to L1 prosodic structure), which results in difficulties
forming phonological representations for production. Current evidence suggests the
problem with adjunction extends beyond adjunction to the prosodic word. However,
the generalisability of this claim is a theoretically tricky issue to resolve. In
particular, as PTH makes specific claims about how the phonological representation
of a word is structured, and psycholinguistic frameworks (e.g. WEAVER++,
Roelofs, 1997) have not made clear how language specific syllabification rules could

169
 Chapter 4

be applied in L2 learners, direct comparisons are not currently possible across

theories. Unfortunately, resolving this issue is beyond the scope of this paper.

One long-lasting issue with interpreting language production data has been the
task of disentangling representational from processing problems. In other words, it is
unclear whether production errors are due to representational or processing failures.
The same issue applies here. Although the current study was only concerned with
how L1 Mandarin learners of L2 English could perform phoneme adjunctions in a
non-inflected context, one should not ignore the fact that adjunction accuracy was
contingent on participants’ ability to correctly isolate the target phoneme(s). Looking
at one specific item in the PE Task, where L1 Mandarin participants consistently
failed to produce stain after omitting /r/ from strain (see Appendix O and Appendix
P), one may strongly suspect that L1 Mandarin learners of L2 English do not
represent the phonological make-up of L2 words in the same way as native-L1
speakers. I argue that this lack of awareness of specific phonemes is a reflection of a
representational problem, which may be compounded by processing difficulties in
assembling L2 phonological segments or syllables according to L2 phonological
rules. This is not to say that L1 Mandarin learners of L2 English lack the
phonological category for /r/, rather it was not perceived as part of the phonological
representation for strain during L2 perception. This interpretation would be
compatible with Best (1995), where she claimed L2 learners initially use
phonetically similar L1 phonological categories to interpret L2 speech (which may
not include /r/ for strain).

One other interesting finding was that some English words were consistently
misperceived across L1 Mandarin participants in the PE task. For example, the word
bold was consistently misperceived as boat (therefore producing oat instead of old
after omitting /b/). Misperception of /t/ and /d/ could be related to insensitivity to
voicing in obstruents (see Flege & Wang, 1989). Another error frequently made by
L1 Mandarin participants was the tendency to omit entire syllables from words when
the omission of a single consonant was required. For example, producing win instead
of winner when asked to omit /t/ from winter. Such tendencies also resulted in some
L1 Mandarin participants producing non-words like ‘pow’ when omitting /d/ from
powder. Again, one can speculate whether this was caused by L1 Mandarin

170
 Chapter 4

participants misperceiving the phoneme, or simply being unable to isolate single

consonants from syllable structures.

4.4.3. Limitations and remaining issues

The current study is not without problems. For example, in the PD Task, the
syllabic vs. consonant cluster distinction may be considered arbitrary for categorising
phonological features. The implication being that phonological variability within
each category may affect perceptual saliency, and therefore perceptual sensitivity
towards these features. In other words, factors other than the syllabic natures of [tɪd]
and [sɪz] could confound the current results. Moreover, allophones for 3SG -s and
past -ed were counterbalanced as much as possible within each verb set, but
allophonic variations across verbs sets with different phonological features were
difficult to control. A more detailed examination of specific allophones within each
phonological feature may prove useful as an extension of the current analyses (e.g.
sensitivity to variations of -ed as [t] in chased and [d] in saved). Note that the
CTOPP-2 PE Task was a standardised test used to test phonemic awareness in
children and young adults and was not designed specifically to test the adjunction of
‘inflection-like’ segments (e.g. [sɪz] [kt]) in non-inflected words. Therefore, the
generalisability of the production data from this task should be viewed with caution.
Future studies should also take a more targeted approach towards specific phoneme
adjunctions in the L2 (e.g. matching phoneme sequences in inflected and non-
inflected contexts), accounting for L1-L2 phonological similarities.

From a theoretical perspective, the precise mechanisms of how phonological

rules or constraints are applied to L2 speech remains unclear, particularly in the
psycholinguistic literature. It was therefore difficult to test specific claims about how
L2 phonological segments are organised and articulated according to L2
phonological rules. This is an unresolved theoretical shortcoming. Consequently,
linguistic theories on phonological structure becomes crucial to our understanding of
L2 phonological processing. However, as the two approaches make different claims

171
 Chapter 4

about how L2 phonology is acquired, direct theoretical comparisons are not always
possible.

Nonetheless, I argue that the current set of findings does make a unique
contribution to the understanding of L1 phonological influence on L2 comprehension
and production. Whilst plenty of research studies have examined perceptual saliency
of isolated L2 sounds, L2 researchers rarely focus on the perceptual saliency of
meaningful linguistic units. Findings from the current study are not only useful in
examining L2 perceptual sensitivity to allophones of inflectional morphemes, but
also have implications for perception of L2 grammatical features in the wider context
of L2 comprehension. By showing no reliable perceptual biases in L2 learners, these
findings to a degree favour accounts which point to grammatical processing errors in
non-native-like L2 comprehension. In addition, current findings partly support the
comprehension-production asymmetry discussed by Goad et al. (2003) and Goad and
White (2006), which claimed that prosodic constraints only affect production but do
not acts as a filter for comprehension and interfere with acquisition of grammatical
features.

172
 Chapter 5

Chapter 5

Discussion and summary

Second language learners frequently exhibit errors in both production and

comprehension in real-time interactions, despite having relevant grammatical
knowledge. Why do they make such errors? What do these errors tell us about the
state of L2 acquisition and human language processing?

In the opening chapter, I discussed in general terms the potential problems for
L2 learners during L2 production and comprehension. Specifically, I first considered
why they might experience difficulties producing L2 grammatical features absent in
their L1. For example, why L1 Mandarin speakers of English might fail to produce
English inflectional morphology (e.g. 3SG -s or past -ed) consistently according to
L2 temporal context. Moreover, I also discussed whether L2 learners can go beyond
knowing L2 grammatical rules in the abstract form and integrate information from
new L2 grammatical features during real-time L2 comprehension. For example, L1
Mandarin learners of English could learn to extract temporal information from
inflectional morphemes despite the fact that Mandarin does not have a system for
inflectional morphology. Lastly, I discussed whether phonological factors,
particularly perceptual deficiencies or biases could influence comprehension of L2
grammatical features. For example, L1 Mandarin learners of English could find some
inflectional morphemes perceptually more salient than others depending on their
phonological contexts, and whether they share phonological features with the
learners’ L1.

In this chapter, I will first revisit the key findings from the three sets of
experiments presented in this thesis and discuss their implications with regard to
theories of L2 production and comprehension. Discussion of Experiments 1, 2 and 3
will centre around sources of error underlying L2 inflectional errors, specifically,
whether inconsistent L2 inflectional production is caused by representational deficits
or processing breakdowns. Discussion of Experiment 4 and 5 will centre around

173
 Chapter 5

integration of linguistic cues during L2 comprehension, specifically, whether L2

learners could extract and integrate information from grammatical cues which are
absent in their L1. Discussion of Experiments 6 and 7 will centre around effects of
phonological saliency on comprehension of L2 grammatical features.

Subsequently, I will discuss some overarching themes running through

explanations of L2 production and comprehension. In particular, I will discuss how
psycholinguistic frameworks can be viewed in conjunction with linguistic theories in
explaining issues in L2 production and comprehension, where they fall short, and
ways they can complement each other. Moreover, I will discuss some general
methodological issues in L2 research, including issues I have attempted to address,
and remaining issues which require further research. Lastly, I will make some
tentative proposals regarding further research studies which may add to the
theoretical scope of current findings. To round up, I will highlight the key
contributions of this thesis to second language research.

5.1. Inflectional errors in L2 production

5.1.1. Theoretical motivations and findings (Chapter 2)

Past research has shown that L2 learners frequently exhibit difficulties

producing grammatical features which do not exist in their L1. Particularly, research
studies have shown that L2 learners of English often experience difficulties
producing L2 inflectional morphology in a consistent manner appropriate to temporal
context, especially amongst learners whose L1 does not use a system of inflectional
morphology (Lardiere, 1998a; 1998b; 2000; 2003; Hawkins & Liszka, 2003).
However, researchers currently have not reached a consensus over the underlying
cause for these errors (i.e. representational deficits or processing breakdowns). With
reference to existing psycholinguistic frameworks of language production (Levelt,
1989; Levelt et al., 1999; Bock & Levelt, 1994; de Bot, 1992; 2003), and theories of
L2 inflectional errors (Hawkins & Chan, 1997; Hawkins & Liszka, 2003; Haznedar
& Schwartz, 1997; Prevost & White, 2000; Hawkins, 2007), I examined whether

174
 Chapter 5

there are consistent error patterns which could point to the nature of inflectional
errors in L2 production. In addition to the debate over whether inflectional errors are
caused by representational deficits or processing breakdowns, I considered whether
L2 learners process relevant abstract level information during conceptualisation of
utterances, as well as the effect of articulation on inflectional accuracy. In this set of
production experiments, L1 Mandarin (L2 English) and L1 English participants
produced event descriptions in a scene description task under distinct temporal
contexts in spoken (Experiments 1 and 2) and written (Experiment 3) modalities.
Using a fixed set of regular English verbs, the paradigm elicited 3SG -s and past -ed
inflections in Present Habitual and Past contexts in L1 Mandarin and L1 English
participants.

Overall, in keeping with previous research, the results showed that L1 Mandarin
participants produced 3SG -s and past -ed inflections inconsistently across temporal
contexts compared with L1 English control participants. Particularly interestingly,
despite the fact that L1 Mandarin participants were significantly more likely to
produce the correct inflection morphology in the appropriate temporal contexts
compared with inappropriate contexts, they made substantially more errors in the
temporal context which required 3SG -s than those which required past -ed.
Moreover, L1 Mandarin participants produced significantly fewer inflectional
omissions in the written compared with the spoken modality, unlike L1 English
participants who did not show such differences.

5.1.2. Theoretical implications

Returning to the debate between representational deficits and processing

breakdowns, these results demonstrated that, in psycholinguistic terms: 1) L2
learners can process temporal information relevant to L2 inflectional morphology
during conceptualisation of the message, even if the learners’ L1 does not use this
information; 2) L2 learners can acquire the appropriate lemma level representations
(diacritic features) for inflectional morphology, even if their L1 does not have such
representations; 3) L2 learners can retrieve the appropriate inflectional forms from

175
 Chapter 5

the mental lexicon. Thus, given that L1 Mandarin learners can accurately but
inconsistently produce appropriate inflections, the findings pointed towards a
processing account of L2 inflectional errors. As such, the most likely explanations
for the current set of L2 inflectional error data are processing breakdowns in
consistently activating appropriate diacritic features or breakdowns in retrieving
inflectional morphology. Moreover, consistent error patterns across spoken and
written production indicated that although the absence of articulation significantly
decreased the number of inflectional omissions in L1 Mandarin participants, it did
not change the asymmetrical patterns of production accuracy observed for 3SG -s
and past -ed inflections. Therefore, the error patterns observed could not be solely
attributed to articulation errors.

5.1.3. Limitations and future directions

Current findings provide convincing evidence against representational deficits at

the syntactic level, but do not tease apart different types of processing breakdowns.
Particularly, current production data cannot distinguish between inconsistent
activation of diacritic features and inconsistent retrieval of inflectional forms.
However, given the asymmetrical pattern of error between inflections with different
degrees of featural complexity (i.e. more errors for the featurally complex 3SG -s
requiring tense and subject number, compared with past -ed, requiring only tense),
there is a case for arguing that inappropriate activation of diacritic features is the
main cause for inconsistent inflectional production.

Additionally, current data do not address possible phonological constraints on

L2 inflectional production. Of the representational deficit accounts of L2 inflectional
errors, the current findings do not test the notion of L1 prosodic constraints on L2
learners’ ability to form the phonological word (Goad et al., 2003). One way of
addressing this issue would be to contrast L2 English learners from multiple L1
backgrounds, with different degrees of overlap between L1-L2 grammatical and
phonological properties. For example, Japanese has the past tense feature, but does
not allow successive consonants in the word final position. Therefore, if L1 Japanese

176
 Chapter 5

learners of English are significantly more accurate than L1 Mandarin learners of

English in producing inflections in the form of consonant clusters, then it would be a
case against L1 prosodic constraints. However, as grammatical and phonological
features are intrinsic in languages and often confound each other, between-group
production data should be interpreted with care (see Amaro et al., 2018, for
discussion).

Another way to test the extent to which L2 inflectional production is affected by

L1 prosodic constraints is to examine whether L2 learners have problems producing
similar phoneme sequences in non-inflectional contexts. One of the main claims of
the Prosodic Transfer Hypothesis is that the phonological make-up of adjoining
inflectional morphemes to verbs (adjunction of inflections to the prosodic word) is
somehow ‘special’, and different to other forms of phonological adjunctions (Goad et
al., 2003; Goad & White, 2006). If future production data show these phoneme
sequences are also difficult for L2 learners to produce in non-inflectional contexts,
then it could argue for a generalised difficulty with phonological processing not
restricted to inflected words. On a methodological level, data coding from all three
production experiments could have benefitted from triangulation across multiple
coders to improve reliability.

5.2. L2 morphosyntactic processing

5.2.1. Theoretical motivations and findings (Chapter 3)

Past research has repeatedly shown non-native-like syntactic processing in L2

learners, which can manifest in the form of behavioural or neurological insensitivity
to syntactic violations during real-time L2 processing (Chen et al., 2007; Jiang et al.
2004; 2007 and more). Whilst much of L2 sentence processing research has focused
on whether L2 learners can make use of L2 syntactic parsing strategies (Marinis et
al., 2005; Felser et al., 2005; Papadopoulou & Clahsen, 2006), very few have
explicitly looked into whether L2 learners can acquire a comprehension mechanism
which can make use of newly acquired L2 grammatical knowledge, specifically,
those that are absent in the L2 learners’ L1. In addition, it is unclear to what extent

177
 Chapter 5

do L2 learners process L2 temporal cues in a native-like way. Moreover, cross-

modality comparisons between auditory and visual comprehension has been rare.

Using English inflectional morphology as a target, I examined how L2 learners,

with no inflectional morphology in their L1, can acquire a comprehension
mechanism which could apply grammatical knowledge relevant to inflectional
morphology. If they can, L2 learners should be sensitive when such features are
omitted in obligatory contexts during incremental sentence processing (e.g. if past -
ed inflections were omitted after a past temporal adverbial). In addition, I focused on
whether L2 learners make processing distinctions when inflections require more than
one type of agreement (i.e. 3SG -s requires both subject number and tense
agreement), and to what extent this resembles native-L1 processing. Moreover, given
that previous studies have shown an auditory disadvantage for sensitivity to L2
grammatical violations (Johnson & Newport, 1989; Johnson, 1992; Murphy, 1997), I
examined whether comprehension modality (auditory or visual) could affect the
integration of L2 grammatical cues. Specifically, I examined whether L2 learners
would find stimuli presented in the auditory modality more difficult to process than
stimuli in the visual modality.

In Experiments 4 and 5, I examined whether L1 Mandarin learners of English,

with no inflectional morphology in their L1, could integrate temporal information
from temporal adverbials (e.g. every day, last week) and temporal inflectional
morphology (3SG -s, past -ed). L1 Mandarin and L1 English participants either
listened or read English sentences in a self-paced comprehension paradigm, and their
sensitivity to inflectional omissions was taken as evidence for cue integration.

The findings showed that L1 Mandarin participants in general do not process L2

inflectional morphology in a native (English) like way. Specifically, in the auditory
modality, whilst L1 Mandarin participants showed sensitivity to inflectional
omissions in general, they did not show differential sensitivity towards 3SG -s and
past -ed omissions, which was evident among L1 English participants. With regard to
the effect of comprehension modality, the findings showed greater sensitivity for
inflectional omissions in auditory versus visual comprehension across L1 Mandarin
and L1 English participants (see effect sizes on Grammaticality), with greater

178
 Chapter 5

between-group differences in self-paced reading (Experiment 5) than in self-paced

listening (Experiment 4). In fact, whereas L1 Mandarin participants exhibited
auditory sensitivity to inflectional omissions, they did not exhibit visual sensitivity to
a statistically significant level, unlike L1 English participants who did across both
auditory and visual modalities.

5.2.2. Theoretical implications

These findings showed that L1 Mandarin participants, though in a non-native

like way, could acquire a comprehension mechanism that applies L2 grammatical
knowledge during real-time L2 comprehension, even when the relevant grammatical
features are absent in the L1. In connectionist terms, these findings supported the
idea that L2 learners can adopt language-specific processing priorities (or cue
validity) during L2 comprehension. However, as L1 English data showed, 3SG -s
and past -ed processing are inherently different, most likely due to the addition of
subject number information which enhanced sensitivity for 3SG -s. In other words, it
is possible that L1 English participants found grammatical 3SG -s trials easier to
process due to the proximity of the subject to the verb, augmenting the reaction time
difference between grammatical and ungrammatical trials (see Figure 19 in 3.2.2.).
The fact that L1 Mandarin participants did not show this difference could suggest
that they do not process subject number information for 3SG -s in a native-like way,
much like the featural complexity effect found in previous production data.

Returning to the issue of comprehension modality, previous studies have shown

an auditory disadvantage for L2 comprehension in the context of grammatical
sensitivity. Specifically, researchers suggested difficulties in auditory comprehension
could be due to additional speech segmentation (Anderson, 1980). With the element
of speech segmentation removed, L1 Mandarin participants showed greater
sensitivity to inflectional omissions in the auditory than in the visual modality. This
seems to show that auditory comprehension is not invariably more difficult. In fact,
auditory stimuli can facilitate grammatical processing (and therefore sensitivity to
grammatical violations).

179
 Chapter 5

Another way in which the current data could explained, is by task-modality

interactions. As discussed in 3.4.3, L2 learners could intentionally or unintentionally
adopt comprehension strategies which suited the demands of the task, but such
strategies could also be confounded by comprehension modality. For example,
auditory stimuli could facilitate L2 grammatical sensitivity in a task centred around
semantic interpretation, whereas visual stimuli could facilitate L2 grammatical
sensitivity in a grammaticality judgement task. One could argue that the facilitatory
effect of auditory stimuli among L2 learners in the current study was contingent on
the fact that they were (sometimes) required to answer comprehension questions, for
which inflectional morphemes (phonologically salient) were important to the
interpretation of the sentence (though none of the questions concerned temporal
context). On the other hand, the auditory disadvantage found in previous studies
could be due to the fact that the task required L2 learners to explicitly monitor for
grammatical violations. As not all grammatical features can be differentiated
auditorily, one may speculate whether these grammatical violations are easier
detected visually.

5.2.3. Limitations and future directions

The first potential limitation concerns the statistical reliability of cross-modality

comparisons. As far as I am aware, few studies have directly contrasted sensitivities
to grammatical violations across different comprehension modalities, possibly due to
difficulties in standardising and comparing reaction time measures. In this case,
auditory comprehension used raw reaction times (accounting for duration of audio
file per segment) and visual (reading) comprehension used residualised reaction
times (accounting for number of characters per segment). Cohen’s d, a commonly
used effect size measure for between-experiment comparisons (see Brysbaert, 2018,
for discussion), was used for the cross-modality contrast. However, it could be
argued that the reliability of this method will need further confirmation. Moreover, as
the cross-modality comparison was conducted across different participant groups
(using identical stimuli), it could be argued that individual differences in L2
proficiency and cognitive functioning could have confounded some of our data.

180
 Chapter 5

Another potential limitation is the interpretation of differential sensitivity to 3SG

-s and past -ed omissions. Currently, I can only infer an effect of featural complexity
given the number of features contained within 3SG -s and past -ed (i.e. L1 Mandarin
participants lack subject number integration). However, in order to demonstrate this
effect unambiguously, subject number violations must be isolated from temporal
violations in the experimental stimuli. This could be implemented by moving the
temporal adverbial to the sentence-final position so one can observe: 1) whether L2
learners (such as L1 Mandarin learners of English) would respond to subject-verb
agreement errors at the verb segment, and 2) whether L2 learners would also respond
to the inconsistent temporal information provided by inflectional morphemes and
temporal adverbials at the post-object temporal adverbial segment (see Table 25).
Moreover, contrasting the ‘subject violation’ condition with the ‘subject + temp
violation’ condition in this design could potentially demonstrate a two-stage
integration process for subject-verb and temporal agreements. Such a study would
need pre-tests to establish the native-L1 pattern for subject number and temporal
sensitivity.

Table 25.

Sample stimuli for morphological violations relating to subject number and temporal
context. Critical segments are marked in red.
Error Type Example

- The girls / paint / sunflowers / every Saturday / in the park.

Subject violation The girls / paints / sunflowers / every Saturday / in the park.
- The girls / painted / sunflowers / yesterday / in the park.
Subject + Temp violation The girls / paints / sunflowers / yesterday / in the park.

The last potential limitation of the current findings concerns the facilitatory
effect of auditory stimuli. This claim lacks detail and is incomprehensive.
Particularly, it is unclear whether there was a general facilitatory effect of auditory
stimuli across inflections, or whether some inflections are more readily facilitated by
their corresponding phonological features in the L2. In other words, could L2

181
 Chapter 5

learners find some L2 sounds easier to perceive, and therefore their corresponding
inflections easier to detect than others? This is an important question as the
phonological features of inflectional morphemes are contingent on their phonological
contexts. To address this question, one could contrast L2 sounds which share L1
phonological features with L2 sounds which do not and establish if L2 learners are
equally sensitive to the same inflections across phonological contexts, or if they
show some degree of perceptual bias towards L1 phonological features. This
question will be addressed in the next section (5.3.).

5.3. L1 phonological effects on L2 comprehension and production

5.3.1. Theoretical motivations and findings (Chapter 4)

Past research into L2 perception suggested that it is possible for L2 learners to

have different perceptual biases in the L2, contingent on the phonological properties
of the L1 (e.g. phonotactic restrictions; see Flege & Wang, 1989). However, this
effect has rarely been examined with respect to meaningful units of language and
specific phonological features. Taking the set of results from the previous
comprehension experiments, it is unclear if L2 learners’ sensitivity to 3SG -s and
past -ed inflectional morphemes were selectively affected by the perceptual saliency
of their resulting phonological features. Note that some phonological features are
intrinsically more salient than others (e.g. syllabic endings are more easily perceived
than consonant clusters), but perceptual saliency can also be exacerbated by
experience of L1 phonological properties (e.g. Mandarin contains mostly
monosyllabic morphemes, and word-final consonant clusters are rare).

For Experiment 6 and 7, I examined whether L2 learners could selectively

exhibit perceptual biases toward L2 grammatical features which share L1
phonological features over those which do not. Again, using English inflectional
morphemes 3SG -s and past -ed, I examined whether L1 Mandarin learners of
English would be more sensitive to 3SG -s and past -ed inflectional morphemes
forming syllabic endings (common in Mandarin Chinese) compared with consonant
cluster endings (rare in Mandarin Chinese). L1 Mandarin and L1 English participants

182
 Chapter 5

discriminated between bare and inflected forms of verbs in an auditory ABX

paradigm (Clark, 1982). The findings showed that, given both L1 Mandarin and L1
English participants processed temporal information from inflectional morphemes,
L1 Mandarin participants showed no consistent perceptual biases toward those
forming syllabic endings over consonant cluster endings compared with L1 English
participants. This was true for both 3SG -s and past -ed.

As discussed in 5.1.3, it is important to establish whether there is a generalised

phonological difficulty in L2 production that goes beyond the adjunction of
inflectional morphemes to verbs. With reference to linguistic theories of L2
inflectional errors, I considered whether difficulties generating prosodic structure to
produce inflectional morphemes could also extend to non-inflected words. For
Experiments 6 and 7, L1 Mandarin and L1 English participants also performed a
phoneme deletion task where they articulated English words with specific
phoneme(s) taken away. This task selectively required participants to adjoin
phonemes after mid-word phoneme deletions. Most distinctly, the findings showed
that L1 Mandarin participants were significantly less accurate in performing
phonemes adjunctions on non-inflected words compared with L1 English speakers.

5.3.2. Theoretical implications

Returning to the question of selective perceptual biases in L2 learners, these

findings suggested that whilst L2 learners showed some perceptual biases towards
inflectional morphemes with syllabic phonological feature, the extent of this bias was
not consistently different from L1 speakers. Moreover, L2 learners processed
grammatical features irrespective of the phonological contexts they appeared in. In
theoretical terms, this suggested that L1 phonological properties did not constrain
perception of L2 phonological features nor the underlying grammatical feature.
Moreover, the fact that L1 speakers and L2 learners did not perceive L2 phonological
features differently has important implications for interpreting data from
comprehension experiments. Current findings pointed toward a general facilitatory
effect of auditory stimuli, with no selective perceptual biases which favour L1 over

183
 Chapter 5

L2 phonological features. Therefore, the findings that L2 learners showed non-

native-like sensitivity towards inflectional omissions in the auditory modality could
not be attributed solely to phonological saliency effects.

Turning to the extent of L1 phonological effects on production, Goad et al.

(2003) theorised that Mandarin prosodic structure constrains the learners’ ability to
adjoin inflectional morphemes to the prosodic word for regular English verbs. This
process is fundamentally different for irregular verbs, where the phonological change
occurs inside the prosodic word. My findings suggest that difficulties in adjoining
phonemes extend beyond inflectional morphemes in L2 learners, such that they
would have problems in non-inflectional contexts as well.

5.3.3. Limitations and future directions

As mentioned in 4.4.4., there may be an issue with imposing arbitrary

phonological feature categories when considering phonological effects for different
inflectional morphemes. L2 learners may have different levels of sensitivity towards
allophones of the same morpheme. For example, consider the phonological context
for 3SG -s in waits and cooks: whereas a digraph12 is formed in waits, a strict
consonant cluster is formed in cooks. One could argue that the perception of 3SG -s
in these two instances could differ even though they represent the same inflectional
morpheme and belong to the same category of phonological features. As allophones
of 3SG -s and past -ed were counterbalanced within each set of verbs (see Appendix
M), adding an allophone manipulation could be an interesting extension to the
current analyses.

The other limitation concerns the items used in the Phoneme Elision task. The
task was taken from a standardised test for phonological processing (i.e. CTOPP-2,
Wagner et al., 2013), and was not specifically designed for L2 learners on their
ability to make adjunctions inside a non-inflectional word. For this reason, I cannot
make specific claims about the extent of phonological influence for 3SG -s and past -

12
a combination of two letters representing one sound

184
 Chapter 5

ed production in Experiments 1 and 2. Conclusive evidence for this effect would

require carefully designed items for the phoneme adjunction part of the task.

5.4. Further theoretical and methodological considerations

5.4.1. Theoretical considerations

Returning to the question of how L2 language data should be examined, I have

taken the view that researchers should seek to take an integrated approach,
accounting for explanations from multiple disciplines. Whilst theories from different
areas of research take distinct perspectives and make different assumptions, they do
not always have to be seen as opposing or contradictory. In fact, the production
experiments showed that a cross-disciplinary approach could provide valuable
insight and highlight some key common grounds in explaining L2 language
production.

Whilst a psycholinguistic framework provides mechanistic explanations for how

a spoken message is generated from the point of conceptualisation to articulation,
linguistic theories of second language learning provide targeted explanations for why
specific language phenomena might occur (i.e. inconsistent production of L2
inflectional morphology). In tackling this specific L2 phenomenon, I have taken
explanations from both approaches and interpreted them in representational or
processing terms, drawing comparisons where appropriate. For example, recall
Levelt and colleagues had outlined the role of diacritic features at the lemma level in
grammatical encoding. Particularly, diacritic features (e.g., tense, number) specify
the syntactic content of lexical representations, based on which grammatical features
relevant to inflectional morphology could be encoded for production. This can be
viewed in conjunction with the concepts of functional category (functional roles of
lexical units) and feature (abstract features such as number, person, tense) in
linguistic terminology. Although these concepts differ in their definition, the
acquisition of these theoretical constructs are acknowledged to be essential for
production by both approaches. Similarly, whereas the psycholinguistic framework
of language production assumes a process of spreading activation for activating

185
 Chapter 5

syntactic features and retrieving morphological forms, linguistic approaches refer to

morphology-syntax mapping and access to morphology within the interlanguage
grammar. Despite differences in their assumptions, they point to similar loci of
processing breakdowns.

L2 comprehension faces a different issue. Whilst psycholinguistic explanations

(connectionist theories) focus on comprehension mechanisms and real-time
integration of linguistic information, linguistic theories explain language in terms of
its hierarchical structure. Thus, there is an inevitable bias towards using
psycholinguistic models in examining the principles of real-time information
integration. This is not to say that linguistic theories do not offer insights into L2
comprehension processes. In fact, they offer detailed descriptions of states of
acquisition in terms of linguistic structures, but do not offer explanations to real-time
processing. Again, these approaches tackle different aspects of the language
acquisition and processing and can be viewed as complementary. However,
equivalences between these approaches cannot be easily drawn.

Theories of L2 phonological processing also suffer from an imbalance of

theoretical explanations from psycholinguistic and linguistic approaches. As pointed
out by Kormos (2006), current psycholinguistic theories do not provide enough
support for the acquisition of L2 phonology beyond individual phonemes (e.g.
syllables), nor the precise nature of L1 influence over L2 phonology. However,
linguistic theories have provided a detailed hierarchical structure for how
phonological units could be combined. For example, the Prosodic Transfer
Hypothesis claims that a specific prosodic structure hindered L1 Mandarin speakers
from producing inflectional morphemes, and is able to describe in structural terms
how phonological properties across languages differed. Such explanations
accounting for phonological rules or properties of languages are currently lacking in
the psycholinguistic arena, leading to a bias towards linguistic theories when
explaining phonological influences in language processing.

To summarise, I have highlighted the importance of using theories of multiple

disciplines in examining L2 production and comprehension data. Unsurprisingly,
both psycholinguistic frameworks and linguistic theories contain shortfalls in terms

186
 Chapter 5

of aspects of language representation and processing they address, which require

further theoretical development.

5.4.2. Methodological considerations

Recall 1.2., where we discussed the importance of investigating real-time

language processing across production and comprehension modalities, and that
research studies should not generalise findings from one specific modality to another
without providing convincing evidence for them. I argue that the current experiments
adequately addressed possible effects of modality in L2 production and
comprehension in a well-controlled manner. For L2 spoken and written production,
the scene description task (Experiments 1, 2 & 3) simulated real-time production
processes by imposing time restrictions on participant responses whilst controlling
for content and structure of the message. This element of the task, though seemingly
trivial, adds significant time pressure for what is a semantically driven task (i.e.
describing an action or event), much like real-life scenarios of L2 production.
Moreover, this task accommodated the slower speed of ‘writing’ by allowing
participants more time, but not the option to edit or rewrite responses. This element
was added to imitate the real-time nature of spoken production as much as possible
(see Gardner et al., 2018, for further discussion). For L2 comprehension,
Experiments 4 and 5 used auditory and visual moving-window paradigms for
sentence presentations. The key advantage of a visual moving-window paradigm is
that it parallels the listening process by not allowing participants to revisit previous
segments of sentence, therefore isolating the processing target to one specific
segment. The self-paced nature of these paradigms ensured controlled exposure to
the experimental stimuli in both modalities.

More methodological issues remain, especially with regard to experimental

design and the choice of languages for second language acquisition studies. As
discussed in previous literature, it is often difficult to infer the extent of L1 influence
by contrasting learner performance from L1s with different degrees of grammatical
and phonological overlap with the L2. As these properties are intrinsic to each

187
 Chapter 5

language, it is often not possible to manipulate these aspects of language according

to the research question. The other difficulty, as evident with the interpretation of
current data, is the inability to isolate the processing of individual grammatical
features when multiple features are marked together (e.g., in English). If one were to
examine whether L1 Mandarin can integrate subject number and temporal
information separately, one solution could be to use a morphologically rich L2 where
these features are marked by different inflectional morphemes. Having said this,
there are benefits to using a language like English, given it is one of most commonly
learned languages, providing possibilities for contrasts across multiple L1s.

5.5. Further research questions

5.5.1. L2 grammatical features in L1-L2 mixed-speech

One interesting extension to the current findings on L2 production could be to

investigate syntactic and morphological processing during L1-L2 mixed-speech, a
frequently observed phenomenon amongst bilingual speakers. Having shown that L2
learners could acquire the relevant representations for inflectional production and
(inconsistently) produce inflectional morphology in appropriate temporal contexts, it
would also be valuable to examine the circumstances which facilitate or hinder the
activation of L2 grammatical features in L1-L2 mixed speech. Such investigations
would shed light on the way shared and unique grammatical representations in the L1
and L2 are organised and activated within the bilingual speech system.

One relevant theoretical framework which was only briefly mentioned in

Chapter 1 is the Matrix Language framework (Myers-Scotton, 1993; 2006). The
central claim by Myer-Scotton and colleagues is that one of the speaker’s two
languages would be dominant (the ‘base’ language) during production and provides
the grammatical framework for syntactic and morphological processing. Although
words from the other language (‘guest’ language) are ‘inserted’ into speech, the
grammatical features of the two languages would not mix (see examples of Dutch-
English code-switching in Poulisse & Bongaerts, 1994). I argue that this

188
 Chapter 5

phenomenon is highly language-specific and contingent on the degree of overlap

between the grammar of the two languages. The question remains open whether the
grammatical features from the ‘guest’ language would also play a part in the abstract
structure of the message if the ‘base’ language does not have an equivalent
grammatical feature (see discussions on composite code-switching in Myers-Scotton,
2006). For example, in a Mandarin-English mixed-speech scenario, where English
verbs are inserted into Mandarin speech (where no suitable Mandarin translation
exists or is not accessible by the speaker), are L1 Mandarin learners of L2 English
likely to adopt the inflectional markings with the inserted English verbs (as activated
by an English lemma), or do they use non-finite verb forms in English with Mandarin
aspectual markers? In Levelts’ terms, this would bring into question whether lemma
level activation of diacritic features (e.g., for tense) for an inflectional language like
English would still proceed if the ‘base’ language does not require such features. An
additional question could be whether grammatical features in production would
differ if the ‘inserted’ word belonged to a different word class (e.g. nouns). In which
case, are grammatical markings (e.g. plural -s) likely to be taken from the ‘base’
language or the ‘guest’ language? Does this tendency vary depending on language-
specific properties (e.g. countability of nouns), language proficiency, L1-L2 balance
and exposure? To answer these questions, an elicited production paradigm can again
be used, controlling for grammatical context and the content of mixed production.

5.5.2. Syllabification of L2 grammatical features

Another extension of current research briefly mentioned in 4.4.4., is whether L2

learners could consistently apply L2 syllabification rules in speech, especially
regarding phonological variable grammatical features. Remember that Experiments 6
and 7 showed via the phoneme elision task, that L1 Mandarin learners of English
have significant difficulty performing adjunctions within non-inflected words after a
specific phoneme is taken away. I considered two possibilities which have not yet
been teased apart. The first possibility was that L1 Mandarin participants did not
represent the items in the task as a sequence of L2 English phonemes. Hence, they
were unable to identify the correct phoneme for deletion. The second possibility was

189
 Chapter 5

that L1 Mandarin participants, despite the correct representations for the words, were
not able to adjoin the remaining phonemes together. The latter possibility concerned
the process of L2 syllabification in psycholinguistic models. Existing research
claimed that L2 syllabification rules could be learned simultaneously with L2
phonological feature distinctions (Archibald, 1998). However, the precise
mechanism via which this takes place is far from clear. How do phonological
processing mechanisms distinguish L1 from L2 syllabification rules when they
supposedly share the same (or similar) phonological representations?

In the context of L2 inflectional production, this also brings into question how
knowledge of L2 syllabification rules is acquired and applied in L2 speech
production, especially when inflectional morphemes can have multiple allophones
depending on phonological contexts. Therefore, it would be valuable to examine the
interaction between the morphology and phonology in the syllabification process.
More specifically, we might consider whether higher-level linguistic knowledge
drives the application of syllabification rules in L2 learners (i.e. whether L2 learners
could syllabify phonemes differently depending on whether they constitute an
inflectional morpheme).

5.6. Summary and conclusions

In the opening chapter of this thesis, I discussed the variety of problems an L2

learner could experience during real-time L2 production and comprehension. More
specifically, having learned L2 grammar, why do L2 learners still violate
grammatical rules during real-time production (e.g. omit inflectional morphemes in
required contexts)? Were these errors due to representational problems, such as
acquiring the abstract representation for inflectional morphology, or were they due to
processing breakdowns, such as activating representations of inflectional
morphology or retrieving inflectional forms. In broad terms, my findings from
Experiments 1, 2 and 3 showed that a processing breakdown explanation for L2
inflectional errors is most likely, given that L2 learners not only processed temporal
contexts under which inflectional morphemes are obligatory, but also activated the

190
 Chapter 5

necessary abstract representations to overtly produce inflectional morphemes.

Moreover, my findings supported the theory of featural complexity, where the
production of 3SG -s (containing subject number and tense features) was consistently
less accurate than past -ed (containing tense feature only) across spoken and written
modalities. Furthermore, modality affected overall accuracy but not the asymmetrical
pattern of inflectional errors, thus indicating processing errors primarily occur at the
grammatical level and cannot be attributed solely to articulation of speech.

The idea of language processing problems carries over to real-time L2

comprehension. The key question here was that whether L2 learners could direct
their attention to and prioritise information from L2 grammatical features during L2
comprehension, even when such features do not exist in the L1. If so, they should
exhibit sensitivity if these features are missing. My findings from Experiments 4 and
5 indicated that L2 learners can indeed use L2 grammatical features (i.e. inflectional
morphology) as a linguistic cue to semantic information in the L2, even when such
cues are not used during L1 comprehension. However, it appeared that L2 learners
could not integrate information from inflectional morphemes in a native-like manner,
as they did not show greater sensitivity towards inflectional morphemes with
multiple features (i.e. 3SG -s), which was observed among L1 English participants.
My findings also showed that integration of information was more successful from
L2 inflectional morphemes when the stimuli were perceptually more salient (in the
auditory modality). However, questions remained as to whether L2 learners
experienced a general facilitatory effect of auditory stimuli, or whether the effect was
selective depending on whether similar phonological features existed in the learner’s
L1.

For the final part of the thesis, I investigated the extent of L1 phonological
influence on perception of inflectional morphemes. Specifically, whether L2 learners
would have perceptual biases towards L2 grammatical features (i.e. inflectional
morphemes) if they share phonological features with the learner’s L1. My findings
from Experiments 6 and 7 showed no consistent between-group effect in
phonological sensitivity between L1 and L2 participants, indicating no reliable
perceptual biases towards phonological features shared with the learner’s L1.

191
 Chapter 5

One prominent question which remained was whether possible phonological

effects played a part in L2 inflectional errors. Although I attempted to explore
possible effects of L2 phonological processing in L2 production in the last set of
experiments, the data were far from conclusive. Current findings from Experiments 6
and 7 seemed to indicate general difficulties with adjoining phonemes according to
L2 phonological rules, rather than a specific difficulty with adjoining L2 inflectional
morphemes to verbs. However, direct comparisons between phoneme adjunctions in
and outside inflectional contexts were not carried out, and this finding was therefore
not conclusive. Further research would also require detailed examinations of the
relevant psycholinguistic accounts on L2 phonological processing. For example, one
would need to examine accounts which explain the L2 syllabification processes and
morphology-phonology interactions.

Overall, this thesis used a multi-disciplinary approach to examine the acquisition

of L2 grammatical features and its processing during L2 production and
comprehension. Using both psycholinguistic frameworks of production and
comprehension and relevant linguistic accounts, I examined possible loci of
inflectional errors in L2 production, and assimilation of information from L2
inflectional morphemes. Current results favour successful acquisition of grammatical
representations but point to processing breakdowns, such as integration of features
during production and comprehension. L1 phonological properties were not found to
fundamentally bias the perception of comprehension of L2 grammatical features, but
the question of L1 phonological influence on L2 production remains open.

192
 Appendices

Appendices
Appendix A.
Experiments 1, 2 and 3: L1 Mandarin (L2 English) participant language background
information.

Experiment 1 Experiment 2 Experiment 3

(N=16) (N=37) (N=48)
M SD M SD M SD

IELTS Overall 7.22 0.41 6.93 0.43 7.05 0.42

IELTS Spoken (Exp. 1 & 2) - - 6.28 0.55 - -
Written (Exp. 3) - - - - 6.92 3.07
AoA for L2 English (years) 10.81 2.32 9.51 2.39 8.75 2.43
Length of Stay (months) 4.31 2.47 8.51 1.35 6.11 3.17
L2 Contact (hours) 3.88 3.30 3.20 2.57 4.55 2.54

Appendix B.
Experiments 1, 2 and 3: Experimental and filler verbs used in the scene description task.

Experimental Filler
Shout Watch
Wait Cook
Load* Write*
Start Listen
End Drink
Applaud* Go*
Need Run
Print* Speak*
Paint Sit
*not included in Exp. 3

193
 Appendices
Appendix C.
Experiments 1, 2 and 3: Image stimuli for experimental and filler trials in the scene
description task.
Experimental stimuli

SHOUT

WAIT

LOAD

194
 Appendices
Appendix C (continued)

START

END

APPLAUD

195
 Appendices
Appendix C (continued)

NEED

PRINT

PAINT

196
 Appendices
Appendix C (continued)

Filler stimuli

WATCH

COOK

WRITE

197
 Appendices
Appendix C (continued)

LISTEN

GO

RUN

198
 Appendices
Appendix C (continued)

DRINK

SPEAK

SIT

199
 Appendices
Appendix D.
Experiments 1, 2 and 3: Experimental legend and vocabulary list for the scene description
task.

Calendar Legend

Today Every day Yesterday

Vocabulary List

Policeman Farmer Builder Firefighter Dancer

Clown Waiter Chef Teacher Doctor
Robber Receptionist Parents Children Fishing
Ice skating Fair Bird Duck Dog
Cat Swan Butterfly Fish Basketball
Football Tennis Photograph Laptop Printer
Vegetable Tomato Pizza Hamburger Cake
Coffee Fire hose Petrol Rocket Chess
Camera Money Flower Map Paint brush
Popcorn Note Letter Chalk Blackboard
Tap Soda Wine Treadmill Track
Swing Sofa Fence Pond Park
House School Restaurant Supermarket Post office
Stage Bus Truck Car

*Each item was presented with the corresponding images from Appendix C during
vocabulary training in Exp. 3.

200
 Appendices
Appendix E.

Experiment 1: Multiple-choice section of Oxford Placement Test with answers (Allan, 1992)

Oxford Placement Test (OPT)

The entire OPT should not take you more than 10 minutes, i.e. don’t think too long and hard
about your answers but keep a steady pace filling out the form. Now please choose the
option that you think matches the sentences best in parts I to III below.

Part I of III (1-50)

Look at these examples. The correct answer is highlighted.

a In warm climates people / like / likes / are liking / siting outside in the sun.

b If it is very hot, they sit / at / in / under / the shade.

Circle the correct answer.

1 Water / is to boil / is boiling / boils / at a temperature of 100oC.

2 In some countries / it is / is / there is / very hot all the time.

3 In cold countries people wear thick clothes / for keeping / to keep / for to keep / warm.

4 In England people are always talking about / a weather / weather / the weather /.

5 In some places / it rains / there rains / it raining / almost every day.

6 In deserts there isn’t / the / some / any / grass.

7 Places near the Equator have / a warm / the warm / warm / weather even in the cold
season.

8 In England / coldest / the coldest / colder / time of year is usually from December to
February.

9 / The most / Most of / Most / people don’t know what it’s really like in other countries.

10 Very / less / little / few / people can travel abroad.

11 Mohammed Ali / has won / won / is winning / his first world title fight in 1960.

12 After he / had won / have won / was winning / an Olympic gold medal he became a
professional boxer.

13 His religious beliefs / have made him / made him to / made him / change his name
when he became champion.

14 If he / has / would have / had / lost his first fight with Sonny Liston, no one would have
been surprised.

15 He has travelled a lot / both / and / or / as a boxer and as a world-famous personality.

201
 Appendices
16 He is very well known / all in / all over / in all / the world.

17 Many people / is believing / are believing / believe / he was the greatest boxer of all
time.

18 To be the best / from / in / of / the world is not easy.

19 Like any top sportsman Ali / had to / must / should / train very hard.

20 Such is his fame that people / would / will / did / always remember him as a champion.

21 The history of / aeroplane / the aeroplane / an aeroplane / is

22 / quite a / a quite / quite / short one. For many centuries men

23 / are trying / try / had tried / to fly, but with

24 / little / few / a little / success. In the 19th century a few people

25 succeeded / to fly / in flying / into flying / in balloons. But it wasn’t until

26 the beginning of the / this / next / last / century that anybody

27 / were / is / was / able to fly a machine

28 / who / which / what / was heavier than air, in other words, in

29 / who / which / what / we now call a ‘plane’. The first people to achieve

30 ‘powered flight’ were the Wright brothers. / His / Their / Theirs / was the machine which
was the forerunner of the jumbo jets

31 that are / such / such a / so / common sight today.

32 They / could / should / couldn’t / hardly have imagined that in 1969,

33 / not much / not many / no much / more than half a century later,

34 a man / will be / had been / would be / walking on the moon.

35 Already / a man / man / the man / is taking the first step towards the stars.

36 Space satellites have now existed / since / during / for / around

37 half a century and we are dependent / from / of / on / them all for all

38 kinds of / informations / information / an information /. Not only

39 / are they / they are / there are / being used for scientific research in

40 space, but also to see what kind of weather / is coming / comes / coming /.

41 By 2018 there / would / must / will / have been satellites in space for sixty

42 years and the ‘space superpowers’ will be / having / making / letting /

43 massive space stations built. When these / will be / are / will have been /

44 completed it will be the first time / when / where / that / astronauts will be

45 able to work in space in large numbers. / Apart / For / Except / all that,

46 in many ways the most remarkable flight / of / above / at / all was

47 / it / that / that one / of the flying bicycle, which the world saw on television,

202
 Appendices
48 / flying / to fly / fly / across the Channel from England to France, with nothing

49 / apart / but / than / a man to power it. As the bicycle-flyer said,

50 ‘It’s the first time / I realize / I’ve realized / I am realizing / what hard work it is to be a
bird!’

Part II of III (51-90)

51 Many teachers / say to / say / tell / their students should learn a foreign language.

52 Learning a second language is not the same / as / like / than / learning a first language.

53 It takes / long time / long / a long time / to learn any language.

54 It is said that Chinese is perhaps the world’s / harder / hardest / more hard / language
to master.

55 English is quite difficult because of all the exceptions / who / which / what / have to be
learnt.

56 You can learn basic structures of a language quite quickly, but only if you / are wanting
/ will to / are willing to / make an effort.

57 A lot of people aren’t used / to the study / to study / to studying / grammar in their
own language.

58 Many adult students of English wish they / would start / would have started / had
started / their language studies earlier.

59 In some countries students have to spend a lot of time working / on / by / in / their own.

60 There aren’t / no / any / some / easy ways of learning a foreign language in your own
country.

61 Some people try to improve their English by / hearing / listening / listening to / the
BBC World Service.

62 / Live / Life / Living / with a foreign family can be a good way to learn a language.

63 It’s no use / to try / trying / in trying / to learn a language just by studying a dictionary.

64 Many students of English / would rather not / would rather prefer not / would rather
not to / take tests.

65 Some people think it’s time we all / learn / should learn / learnt / a single international
language.

66 Charles Walker is a teacher at a comprehensive school in Norwich. He / has joined /

joined / joins /

67 the staff of the school in 1998 and / has been working / worked / works / there ever
since.

68 Before / move / to move / moving / to Norwich, he taught in Italy and in Wales,

69 and before that he / has been / was / was being / a student at Cambridge

70 University. So far he / isn’t / wasn’t / hasn’t been / in Norwich for as long as

203
 Appendices
71 as he was in Wales, but he likes the city a lot and / should / would / could /

72 like to stay there for at least another two years, or, / how / which / as / he

73 puts it, until his two children / have / will have / will be / grown up a bit.

74 He met his wife, Kate, in 1992 while he / was to live / was living / had been living /
abroad for a while, and they got married in 1996.

75 Their two children, Mark and Susan, / are / were / have been / both born in Norwich.

76 The Walkers’ boy, / who / which / he / is five, has just started

77 at school, but / his / their / her / sister

78 / shall stay / stays / will be staying / at home for another couple of years,

79 because she is nearly two years / younger / more young / the younger /

80 than him. Charles and Kate Walker / are used / use / used / to live in the

81 country, but now that they have children, they / have moved / move / moved /

82 into the city. Charles wanted a house / next / near / close / the

83 school / in order / for / to / get to work easily. Unfortunately

84 / the / a / that / one the two of them really wanted was too expensive,

85 so they / must / should / had to / buy one a bit further away. By the time the

86 children / go / will go / will have gone / to secondary school.

87 / that / which / what / Charles and Kate hope will be in Norwich, the

88 Walkers / will have been / have been / will be / living there for at least fifteen years.

89 They can’t be sure if they / stay / do stay / will stay /, but if they

90 / don’t / didn’t / won’t /, their friends won’t be too surprised.

Part III of III (91-100)

Look at the following examples of question tags in English. The correct form of the tag is
highlighted.

a He’s getting the 9.15 train, / isn’t he / hasn’t he / wasn’t he /?

b She works in a library, / isn’t she / doesn’t she / doesn’t he /?

c Tom didn’t tell you, / hasn’t he / didn’t he / did he /?

d Someone’s forgotten to switch off the gas, / didn’t one / didn’t they / haven’t they /?

Now circle the correct question tag for the following 10 items:

91 John’s coming to see you, / hasn’t he / wasn’t he / isn’t he /?

92 It’s been a long time since you’ve seen him, / hasn’t it / isn’t it / haven’t you /?

93 He’s due to arrive tomorrow, / won’t he / isn’t he / will he /?

204
 Appendices
94 He won’t be getting in till about 10.30, / isn’t he / is he / will he /?

95 You met him while you were on holiday, / didn’t you / weren’t you / haven’t you /?

96 I think I’m expected to pick him up, / aren’t I / don’t I / are you /?

97 No doubt you’d rather he stayed in England no, / didn’t you / wouldn’t you / shouldn’t
you /?

98 Nobody else has been told he’s coming, / is he / has he / have they /?

99 We’d better not stay up too late tonight, / didn’t we / have we / had we /?

100 I suppose it’s time we called it a day, / didn’t we / isn’t it / don’t /?

Answers:

1 boils 21 the 41 will 61 listening to 81 have moved

aeroplane
2 it is 22 quite a 42 having 62 living 82 near
3 to keep 23 had tried 43 are 63 trying 83 to
4 the weather 24 little 44 that 64 would rather not 84 the
5 it rains 25 in flying 45 for 65 learnt 85 had to
6 any 26 last 46 of 66 joined 86 go
7 warm 27 was 47 that 67 has been working 87 which
8 the coldest 28 which 48 flying 68 moving 88 will have been
9 most 29 what 59 but 69 was 89 will stay
10 few 30 theirs 50 I’ve realized 70 hasn’t been 90 don’t
11 won 31 such a 51 say 71 would 91 isn’t he
12 had won 32 could 52 as 72 as 92 hasn’t it
13 made him 33 not much 53 a long time 73 have 93 isn’t he
14 had 34 would be 54 hardest 74 was living 94 will he
15 both 35 man 55 which 75 were 95 didn’t you
16 all over 36 for 56 are willing to 76 who 96 aren’t I
17 believe 37 on 57 to studying 77 his 97 wouldn’t you
18 in 38 information 58 had started 78 will be staying 98 have they
19 had to 39 are they 59 on 79 younger 99 had we
20 will 40 is coming 60 any 80 used 100 isn’t it

205
 Appendices
Appendix F.
Experiment 1: Descriptive and inferential statistics (t-test) on the multiple-choice
section of Oxford Placement Test for L1 Mandarin and L1 English groups
(N=16;18).

M SD t (df)
L1 Mandarin 71.90 8.14 -10.55 (32) ***
L1 English 94.06 3.46
*** p <.001 sig. level against L1 English group.

Appendix G.
Experiments 4 and 5: L1 Mandarin group language background information.
Experiment 4 Experiment 5
(N=61) (N=61)
M SD M SD
IELTS Overall 7.09 0.40 7.02 0.40
IELTS Listening (Exp. 4) 7.55 0.75 - -
Reading (Exp. 5) - - 7.60 0.65
AoA for L2 English (years) 9.05 3.08 8.27 2.15
Length of Stay (months) 9.72 3.86 5.20 2.42
L2 Contact (hours) 3.30 2.09 4.34 2.59

206
 Appendices
Appendix H.
Experiments 4 and 5: Experimental sentences with comprehension questions for the
self-paced listening and self-paced reading tasks. Forward slashes (/) denote
segment boundaries. Condition labels: PH – Present Habitual. P – Past. G –
Grammatical. UG – Ungrammatical.

Group A.

1. In the morning / the gardener / shouts / at the cat / in the house. (PH - G)
Q: Does the gardener shout at the cat in the house? (Yes)
2. Every weekend / the food critic / shout / at the waiter / in the restaurant. (PH -
UG)
3. Last weekend / the boy / shouted / at the cat / in the garden. (P - G)
4. Last night / the chef / shout / at the waiter / in the kitchen. (P - UG)
Q: Did the chef shout at the waiter by the bar? (No)
5. On Mondays / the man / waits / for a flight / at the airport. (PH - G)
6. In the afternoon / the customer / wait / for the car / at the restaurant. (PH -
UG)
7. Yesterday evening / the patient / waited / for the doctor / at the hospital. (P -
G)
8. Yesterday afternoon / the lady / wait / for the elevator / at the shopping mall.
(P - UG) Q: Did the lady wait for the elevator at the shopping mall? (Yes)
9. On Fridays / the builder / loads / the bricks / onto the lorry. (PH - G)
10. Every morning / the builder / load / the rocket / with fuel. (PH - UG)
11. Yesterday afternoon / the man / loaded / the washing machine / with clothes.
(P - G)
Q: Did the man load the washing machine with clothes? (Yes)
12. Yesterday morning / the fireman / load / the hose / onto the truck. (P - UG)
13. Every morning / the manager / starts / the meeting / with a joke. (PH - G)
14. At noon / the man / start / a conversation / with his friends. (PH - UG)
Q: Does the man start a conversation with his friends?(Yes)
15. Yesterday / the girl / started / a tennis game / in the schoolyard. (P - G)
Q: Did the girl start a football match in the schoolyard? (No)

207
 Appendices

16. Last Sunday / the teenager / start / a game of chess / at home. (P - UG)
17. Every week / the fireman / ends / a fire / in the café. (PH - G)
Q: Does the fireman end a fire in the café? (Yes)
18. Every weekend / the waiter / end / an argument / between the customers. (PH
- UG)
19. Yesterday morning / the doctor / ended / a conversation / with a patient. (P -
G)
20. Last Monday / the policeman / end / a fight / in the pub. (P - UG)
21. On weekends / the director / applauds / the dancer / on the stage. (PH - G)
22. On Fridays / the woman / applaud / the girls / at school. (PH - UG)
23. Yesterday evening / the teacher / applauded / the children / on stage. (P - G)
24. Last night / the athlete / applaud / the tennis game / on tv. (P - UG)
Q: Did the athlete applaud the tennis game on tv? (Yes)
25. Every day / the chef / needs / vegetables / for the restaurant. (PH - G)
26. Every month / the manager / need / leaflets / for the reception. (PH - UG)
Q: Does the manager need posters for the reception? (No)
27. Last Friday / the chef / needed / glasses / for wine. (P - G)
Q: Did the chef need glasses for juice? (No)
28. Last summer / the boy / need / a camera / for his trip. (P - UG)
29. Every Saturday / the woman / prints / flyers / for the concert. (PH - G)
30. On Tuesdays / the architect / print / a building plan / in the office. (PH - UG)
Q: Does the architect print a building plan on the building site? (No)
31. Yesterday afternoon / the boy / printed / his homework / at school. (P - G)
32. Last week / the teacher / print / a map / of the Old Town. (P - UG)
33. Every summer / the artist / paints / butterflies / in the garden. (PH - G)
Q: Does the artist paint butterflies in the studio? (No)
34. On weekends / the girl / paint / vases / at home. (PH - UG)
35. Last week / the woman / painted / swans / in the park. (P - G)
36. Last month / the girl / paint / sunflowers / at school. (P - UG)

208
 Appendices

Group B.

1. Every day / the chef / shouts / at the waiter / in the kitchen. (PH - G)
Q: Does the chef shout at the waiter by the bar?(No)
2. In the morning / the gardener / shout / at the cat / in the house. (PH - UG)
3. Last night / the food critic / shouted / at the waiter / in the restaurant. (P - G)
4. Last weekend / the boy / shout / at the cat / in the garden. (P - UG)
Q: Does the gardener shout at the cat in the house?(Yes)
5. Every afternoon / the lady / waits / for the elevator / at the shopping mall. (PH
- G)
Q: Does the lady wait for the elevator at the shopping mall? (Yes)
6. Yesterday evening / the patient / wait / for the ambulance / at the hospital.
(PH - UG)
7. Yesterday afternoon / the customer / waited / for the car / at the restaurant. (P
- G)
8. On Mondays / the man / wait / for a flight / at the airport. (P - UG)
9. In the morning / the fireman / loads / the hose / onto the truck. (PH - G)
10. On Fridays / the builder / load / the bricks / onto the lorry. (PH - UG)
11. Last Tuesday / the builder / loaded / the rocket / with fuel. (P - G)
12. Yesterday afternoon / the man / load / the washing machine / with clothes. (P
- UG)
Q: Did the man load the washing machine with clothes? (Yes)
13. Every Sunday / the teenager / starts / a game of chess / at home. (PH - G)
14. Every morning / the manager / start / the meeting / with a joke. (PH - UG)
15. Last Monday / the man / started / a conversation / with his friends. (P - G)
Q: Did the man start a conversation with his friends? (Yes)
16. Yesterday / the girl / start / a tennis game / in the schoolyard. (P - UG)
Q: Did the girl start a football match in the schoolyard? (No)
17. Every weekend / the policeman / ends / a fight / in the pub. (PH - G)
18. Every week / the fireman / end / a fire / in the café. (PH - UG)
Q: Does the fireman end a fire in the café? (Yes)

209
 Appendices

19. Last weekend / the waiter / ended / an argument / between the customers. (P -
G)
20. Yesterday morning / the doctor / end / a conversation / with a patient. (P -
UG)
21. In the afternoon / the athlete / applauds / the tennis game / on tv. (PH - G)
Q: Does the athlete applaud the tennis game on tv? (Yes)
22. Yesterday evening / the teacher / applaud / the children / on stage. (PH - UG)
23. Last Friday / the woman / applauded / the girls / at school. (P - G)
24. On weekends / the director / applaud / the dancer / on the stage. (P - UG)
25. Every summer / the boy / needs / a camera / for his trip. (PH - G)
26. Last Friday / the chef / need / glasses / for wine. (PH - UG)
Q: Did the chef need glasses for juice? (No)
27. Last month / the manager / needed / leaflets / for the reception. (P - G)
Q: Did the manager need posters for the shop? (No)
28. Every day / the chef / need / vegetables / for the restaurant. (P - UG)
29. Every year / the teacher / prints / a map / of the Old Town. (PH - G)
30. Every Saturday / The woman / print / flyers / for the concert. (PH - UG)
31. Last Tuesday / the architect / printed / a building plan / in the office. (P - G)
Q: Did the architect print a building plan on the building site? (No)
32. Yesterday afternoon / the boy / print / his homework / at school. (P - UG)
33. Every week / the girl / paints / sunflowers / at school. (PH - G)
34. Last week / the woman / paint / swans / in the park. (PH - UG)
35. Last weekend / the girl / painted / vases / at home. (P - G)
36. Every summer / the artist / paint / butterflies / in the garden. (P - UG)
Q: Does the artist paint butterflies in the studio? (No)

Group C.

1. Every afternoon / the boy / shouts / at the cat / in the garden. (PH - G)
2. Every day / the chef / shout / at the waiter / in the kitchen. (PH - UG)
3. Yesterday evening / the gardener / shouted / at the cat / in the house. (P - G)
Q: Did the gardener shout at the cat in the house? (Yes)

210
 Appendices

4. Last night / the food critic / shout / at the waiter / in the restaurant. (P - UG)
5. Every morning / the patient / waits / for the doctor / at the hospital. (PH - G)
6. Every afternoon / the lady / wait / for the elevator / at the shopping mall. (PH
- UG)
Q: Does the lady wait for the elevator at the shopping mall? (Yes)
7. Last Monday / the man / waited / for a flight / at the airport. (P - G)
8. Yesterday afternoon / the customer / wait / for the car / at the restaurant. (P -
UG)
Q: Did the customer wait for the car at the hotel? (No)
9. Every day / the man / loads / the washing machine / with clothes. (PH - G)
Q: Does the man load the washing machine with clothes? (Yes)
10. In the morning / the fireman / load / the hose / onto the truck. (PH - UG)
11. Last Tuesday / the builder / loaded / the bricks / onto the lorry. (P - G)
12. Last Tuesday / the builder / load / the rocket / with fuel. (P - UG)
13. In the afternoon / the girl / starts / a tennis game / in the schoolyard. (PH - G)
14. Every Sunday / the teenager / start / a game of chess / at home. (PH - UG)
15. Yesterday morning / the manager / started / the meeting / with a joke. (P - G)
16. Last Monday / the man / start / a conversation / with his friends. (P - UG)
Q: Did the man start a conversation with his friends? (Yes)
17. At the end of the day / the doctor / ends / a conversation / with a patient. (PH
- G)
Q: Does the doctor end a conversation with a nurse? (No)
18. Every weekend / the policeman / end / a fight / in the pub. (PH - UG)
19. Last Friday / the fireman / ended / a fire / in the café. (P - G)
Q: Did the fireman end a fire in the café? (Yes)
20. Last weekend / the waiter / end / an argument / between the customers. (P -
UG)
21. Every week / the teacher / applauds / the children / on stage. (PH - G)
22. In the afternoon / the athlete / applaud / the tennis game / on tv. (PH - UG)
Q: Does the athlete applaud the tennis game on tv? (Yes)
23. Last week / the director / applauded / the dancer / on the stage. (P - G)
24. Last Friday / the woman / applaud / the girls / at school. (P - UG)

211
 Appendices

25. Every evening / the chef / needs / glasses / for wine. (PH - G)
Q: Does the chef need glasses for juice? (No)
26. Every summer / the boy / need / a camera / for his trip. (PH - UG)
27. Yesterday morning / the chef / needed / vegetables / for the restaurant. (P - G)
28. Last month / the manager / need / leaflets / for the reception. (P - UG)
29. Every term / the boy / prints / his homework / at school. (PH - G)
30. Every year / the teacher / print / a map / of the Old Town. (PH - UG)
Q: Does the teacher print maps of the New Town? (No)
31. Last Saturday / The woman / printed / flyers / for the concert. (P - G)
32. Last Tuesday / the architect / print / a building plan / in the office. (P - UG)
Q: Did the architect print a building plan on the building site? (No)
33. Every Sunday / the woman / paints / swans / in the park. (PH - G)
34. Every week / the girl / paint / sunflowers / at school. (PH - UG)
35. Last summer / the artist / painted / butterflies / in the garden. (P - G)
Q: Did the artist paint butterflies in the studio? (No)
36. Last weekend / the girl / paint / vases / at home. (P - UG)

Group D.

1. Every weekend / the food critic / shouts / at the waiter / in the restaurant. (PH
- G)
2. Every afternoon / the boy / shout / at the cat / in the garden. (PH - UG)
3. Last night / the chef / shouted / at the waiter / in the kitchen. (P - G)
4. Yesterday evening / the gardener / shout / at the cat / in the house. (P - UG)
Q: Did the gardener shout at the cat in the house? (Yes)
5. In the afternoon / the customer / waits / for the car / at the restaurant. (PH - G)
6. Every morning / the patient / wait / for the doctor / at the hospital. (PH - UG)
7. Yesterday afternoon / the lady / waited / for the elevator / at the shopping
mall. (P - G) Q: Did the lady wait for the elevator at the shopping mall? (Yes)
8. Last Monday / the man / wait / for a flight / at the airport. (P - UG)
9. Every morning / the builder / loads / the rocket / with fuel. (PH - G)
10. Every day / the man / load / the washing machine / with clothes. (PH - UG)

212
 Appendices

Q: Does the man load the washing machine with clothes? (Yes)
11. Yesterday morning / the fireman / loaded / the hose / onto the truck. (P - G)
Q: Did the fireman load the hose into the van? (No)
12. Last Tuesday / the builder / load / the bricks / onto the lorry. (P - UG)
13. At noon / the man / starts / a conversation / with his friends. (PH - G)
Q: Does the man start a conversation with his friends? (Yes)
14. In the afternoon / the girl / start / a tennis game / in the schoolyard. (PH - UG)
15. Last Sunday / the teenager / started / a game of chess / at home. (P - G)
16. Yesterday morning / the manager / start / the meeting / with a joke. (P - UG)
17. Every weekend / the waiter / ends / an argument / between the customers.
(PH - G)
18. At the end of the day / the doctor / end / a conversation / with a patient. (PH -
UG)
19. Last Monday / the policeman / ended / a fight / in the pub. (P - G)
20. Last Friday / the fireman / end / a fire / in the café. (P - UG)
Q: Did the fireman end a fire in the café? (Yes)
21. On Fridays / the woman / applauds / the girls / at school. (PH - G)
22. Every week / the teacher / applaud / the children / on stage. (PH - UG)
Q: Does the teacher applaud the children on the stage? (Yes)
23. Last night / the athlete / applauded / the tennis game / on tv. (P - G)
Q: Did the athlete applaud the tennis game on tv? (Yes)
24. Last week / the director / applaud / the dancer / on the stage. (P - UG)
25. Every month / the manager / needs / leaflets / for the reception. (PH - G)
26. Every evening / the chef / need / glasses / for wine. (PH - UG)
Q: Does the chef need glasses for juice? (No)
27. Last summer / the boy / needed / a camera / for his trip. (P - G)
28. Yesterday morning / the chef / need / vegetables / for the restaurant. (P - UG)
29. On Tuesdays / the architect / prints / a building plan / in the office. (PH - G)
30. Every term / the boy / print / his homework / at school. (PH - UG)
31. Last week / the teacher / printed / a map / of the Old Town. (P - G)
Q: Did the teacher print maps of the New Town? (No)
32. Last Saturday / The woman / print / flyers / for the concert. (P - UG)

213
 Appendices

33. On weekends / the girl / paints / vases / at home. (PH - G)

34. Every Sunday / the woman / paint / swans / in the park. (PH - UG)
35. Last month / the girl / painted / sunflowers / at school. (P - G)
Q: Did the girl paint sunflowers at school? (Yes)
36. Last summer / the artist / paint / butterflies / in the garden. (P - UG)
Q: Did the artist paint butterflies in the studio? (No)

Appendix I.
Experiments 4 and 5: Filler sentences with comprehension questions for the self-
paced listening and self-paced reading tasks (same for Groups A, B, C and D).
Forwards slashes (/) denote segment boundaries. Error labels: V- verb form. A –
agreement. PP – preposition. D – determiner.

1. In the theatre / the boys / are watching / the clowns / perform on stage.
Q: Are the boys watching the clowns perform on stage? (Yes)
2. At school / the teachers / watched / the children / to play football. (V)
Q: Did the teachers watch the children play basketball? (No)
3. At the park / the children / watched / the ducks / play.
4. In the theatre / the girls / are watching / the dancers / to perform on stage. (V)
5. In the restaurant / the chefs / are cooking / fish / for the food critic.
6. At home / the boys / cooked / spaghetti / with a tomatoes. (A)
Q: Did the boys cook spaghetti with tomatoes? (Yes)
7. At the park / the chefs / cooked / hamburgers / with potatoes.
8. In home / the girls / are cooking / vegetables / in the garden. (D)
9. In the car park / the policemen / are writing / parking tickets / by the cars.
10. In the library / the students / wrote / in her notebooks / with pencil. (A)
11. In the bedroom / the girls / wrote / their homework / with pen.
Q: Did the girls write their homework with pen?(Yes)
12. In the library / the ladies / are writing / a letter / to her friend. (A)
13. In the car / the teenagers / are singing / to music / on their iPods.
14. At school / the children / sang / carol / by the Christmas tree. (D)
15. In the park / the clowns / sang / for the children / on the stage.
Q: Did the clowns sing for the children in the school hall?

214
 Appendices

16. In the living room / the girls / are singing / music / on the sofa. (PP)
Q: Are the girls singing to music on the sofa? (Yes)
17. In the emergency room / the patients / are drinking / water / from the tap.
Q: Are the patients drinking water from the tap? (Yes)
18. In the kitchen / the girls / drank / glass / of orange juice. (PP)
19. In the restaurant / the chefs / drank / wine / with the waiters.
20. In the garden / the boys / are drinking / water / from glass. (PP)
21. In the kitchen / two cats / are going / out of the house / through a window.
22. At the railway station / the women / went / to the shop / buy lunch. (V)
Q: Did the women go to the shop at the railway station? (Yes)
23. On the beach / the children / went / for a walk / with the dog.
24. In the restaurant / the chefs / are going / see / the guests. (V)
25. On the pavement / the children / are running / to the park / with a dog.
26. In the stadium / the athletes / ran / on the track / to win medal. (D)
27. In the park / the boys / ran / on the grass / with the teacher.
Q: Did the boys run on the track with the teacher? (No)
28. In the gym / the athletes / are running / the treadmill / with weights. (PP)
Q: Are the athletes running on the treadmill at home? (No)
29. In class / the boys / are speaking / to the girl / about their homework.
30. At the golf club / the guests / spoke / to the receptionist / in the phone. (PP)
31. At the restaurant / the customers / spoke / to the waiter / at the entrance.
32. At the hotel / guests / are speaking / to the chef / about the menu. (D)
Q: Are the guests speaking to the chef about the waiter? (No)
33. At the park / the girls / are sitting / on the swing / with a dog.
Q: Are the girls sitting on the grass at the park? (No)
34. In the café / customers / sat / at the table / for an hour. (D)
35. In the restaurant / the guests / sat / by the bar / with some wine.
36. At the hotel / the women / are sitting / at table / by the window. (D)

215
 Appendices
Appendix J.
Experiment 4 and 5: Morphological Proficiency Test (with answers)

This grammar test consists of 50 questions. Part I consists of 30 multiple choice

questions (MCQs), and Part II consists of 20 gap filling exercises. For each MCQ, you must
mark the most appropriate option out of the 3 that are given. For each gap-filling exercise,
you must write down a suitable verb in its correct form in order to form a coherent sentence.
This test should take no longer than 10 minutes overall to complete.

Part I (Multiple-choice questions - MCQs)

1. Court is in session, the lawyers is making / are making / was making a case for the
victims on the TV broadcast.
2. The photographer often are coming / come / comes to the studio by taxi.
3. Daniel is apologising / was apologising / were apologising for his mistakes at work
when the manager arrived.
4. The children are running / will have run / will be running to the buses despite being
instructed to walk.
5. Sarah and her boyfriend are engaged; they is living / are living / has lived in
Nottingham.
6. During the ceremony, the winning athletes has stood / will have stood / will stand on
the podium.
7. Emma sees that the customer is arguing / was arguing / argued with the shop
assistant.
8. The workers have told / has told / was telling the man to stay away from the crime
scene.
9. Michael expect / were expecting / expected his train at seven o'clock last night.
10. The engineer hold / has held / were holding the pipe for at least an hour now.
11. Every fortnight, Benjamin is having / has / were had an appointment with the doctor
12. The meeting will start / will have started / has started by the time I get there.
13. The ferry depart / departs / is departing from the port right now.
14. The criminal have shot / has shot / were shooting the victim multiple times in the
back.
15. I were having / was having / has a shower when she called.
16. Yesterday, the reporter will explain / have explained / explained her intentions at the
interview.
17. John have eaten / eats / was eating a ham sandwich with coffee for lunch every day.

216
 Appendices
Appendix J (continued)
18. Elizabeth is writing / were writing / was writing a letter to her mother at the desk
when Will entered.
19. The dancers was performing / were performing / perform at the opera house last
Sunday.
20. Jane missed / had missed / have missed her flight to New York at 2pm yesterday.
21. Katie are borrowing / will borrow / will have borrowed a dress from Jane if she can't
go home tonight.
22. The prime minister has introduced / have introduced / had introduced his secretary
before the meeting started.
23. I had seen / have seen / am seeing her mother twice since this morning.
24. She hid / has hidden / was hiding behind a bush when we found her.
25. The children buys / is buying / buy sweets from the shop every weekend.
26. Sophie was leaving / left / have left the cat on the table before going to work.
27. The football fans chose / choose / had chosen their favourite team before the game
had started.
28. The gentleman are defending / is defending / were defending his argument in a
debate.
29. Will believes that he failed / will fail / has failed the blood test if he eats too much.
30. The judges must decide / decided / is deciding on the outcome of the trial.

Part II (Gap-Filling Exercise)

1. Mark ___________ (find) his watch on the kitchen table this morning.
2. The professor __________ (make) his decision by the time the committee met again.
3. The passenger ___________ (appear) 5 minutes before take-off.
4. As the girl plays with her doll, the boy ___________ (feed) the cat some biscuits from a
jar.
5. Chris ___________ (catch) the 8.30 train if he cannot get up early tomorrow.
6. As of today, my friend and I __________ (know) each other for exactly ten years.
7. I ___________ (write) a letter when my friend knocked on my door.
8. At this moment, Jessica _________ (prepare) for her friend’s party.
9. As she sat down, the woman ___________ (remember) her time working at the
hospital.
10. The volleyball team __________ (win) ten games in a row by the time they were
beaten.
11. The driver ___________ (spend) 3 pounds on his lunch every day last week.
12. I ___________ (arrive) in London by six tomorrow evening.

217
 Appendices
Appendix J (continued)
13. David wasn't sure if he ________ (order) already when the waiter came back.
14. The players ___________ (think) there will be a delay to the start of the game.
15. She believes that Kevin ___________ (sing) on his way to work every day.
16. The team ___________ (build) a skyscraper which attracted many visitors.
17. The artist ___________ (draw) a portrait yesterday by the sea.
18. Catherine can hear that her dog ___________ (snore) in the living room.
19. Until you arrive, Jeremy _________ (wait) for you at the station.
20. I __________ (work) for 2 organisations since I came back from Spain.

Answers

Part I Part II
1. are making 1. found
2. comes 2. will have made
3. was apologising 3. appeared
4. are running 4. feeds
5. are living 5. will catch
6. will stand 6. have known
7. is arguing 7. was writing
8. have told 8. is preparing
9. expected 9. remembered
10. has held 10. had won
11. has 11. spent
12. will have started 12. will have arrived
13. is departing 13. had ordered
14. has shot 14. think
15. was having 15. sings
16. explained 16. built
17. eats 17. drew
18. was writing 18. is snoring
19. were performing 19. will wait
20. missed 20. have worked
21. will borrow
22. had introduced
23. have seen
24. was hiding
25. buy
26. left
27. had chosen
28. is defending
29. will fail
30. decide

218
 Appendices
Appendix K.
Experiments 4 and 5: Descriptive and inferential statistics on the Morphological Proficiency
Test for L1 Mandarin and L1 English groups.

Experiment 4 Experiment 5
(N=61;56) (N=61;57)
M SD t (df) p M SD t (df) p
L1 Mandarin
MCQs 27.61 1.64 -5.77 (114.95) *** 27.00 2.54 -5.48 (78.16) ***
Gap-filling 12.70 1.97 -2.83 (115.39) ** 12.15 1.92 -4.32 (108.40) ***
L1 English
MCQs 29.04 1.08 - 28.92 0.97 -
Gap-filling 13.73 2.05 - 13.86 2.35 -
** p <.01 sig. level against L1 English group.
*** p <.001 sig. level against L1 English group.

Appendix L.
Experiments 6 and 7: L1 Mandarin group language background information.

Experiment 6 Experiment 7
(N=61) (N=61)
M SD M SD

IELTS Overall 7.20 0.36 6.99 0.40

IELTS Listening 7.86 0.70 7.95 3.14
AoA for L2 English (years) 8.79 2.80 8.39 2.19
Length of Stay (months) 10.50 3.83 7.48 2.34
L2 Contact (hours) 3.33 2.17 3.98 2.87

219
 Appendices
Appendix M.
Experiments 6 and 7: Experimental (Sets A, B and C) and filler verbs used in Phonological
Discrimination Task with phonetic properties of the final phoneme: Voicing, Place of
Articulation (POA), and phonetic realisations of 3SG -s and past -ed endings (using
International Phonetic Alphabet transcription).

Experimental Verbs (regular).

Final Phoneme 3SG -s ending -ed ending
Set A Voicing POA realisation feature realisation feature
cons. syllabic
Shout voiceless Alveolar [s] [ɪd]
cluster
cons. syllabic
Wait voiceless Alveolar [s] [ɪd]
cluster
cons. syllabic
Applaud voiced Alveolar [z] [ɪd]
cluster
cons. syllabic
Start voiceless Alveolar [s] [ɪd]
cluster
cons. syllabic
End voiced Alveolar [z] [ɪd]
cluster
cons. syllabic
Load voiced Alveolar [z] [ɪd]
cluster
cons. syllabic
Need voiced Alveolar [z] [ɪd]
cluster
cons. syllabic
Print voiceless Alveolar [s] [ɪd]
cluster
cons. syllabic
Paint voiceless Alveolar [s] [ɪd]
cluster
cons. syllabic
Avoid voiced Alveolar [z] [ɪd]
cluster

Final Phoneme 3SG -s ending -ed ending

realisatio
Set B Voicing POA realisation feature feature
n
Attack voiceless Velar [s] cons. [t] cons.
cluster cluster
Arrive voiced Labiodental [z] cons. [d] cons.
cluster cluster
Bake voiceless Velar [s] cons. [t] cons.
cluster cluster
Pack voiceless Velar [s] cons. [t] cons.
cluster cluster
Move voiced Labiodental [z] cons. [d] cons.
cluster cluster
Disturb voiced Bilabial [z] cons. [d] cons.
cluster cluster
Escape voiceless Bilabial [s] cons. [t] cons.
cluster cluster

220
 Appendices
Kick voiceless Velar [s] cons. [t] cons.
cluster cluster
Save voiced Labiodental [s] cons. [d] cons.
cluster cluster
Approve voiced Labiodental [z] cons. [d] cons.
cluster cluster

Final Phoneme 3SG -s ending -ed ending

Set C Voicing POA realisation feature realisation feature
syllabic cons.
Chase voiceless Alveolar [z] [t]
cluster
syllabic cons.
Crash voiceless Palatal [z] [t]
cluster
syllabic cons.
Damage voiced Palatal [z] [d]
cluster
syllabic cons.
Encourage voiced Palatal [z] [d]
cluster
syllabic cons.
Finish voiceless Palatal [z] [t]
cluster
syllabic cons.
Kiss voiceless Alveolar [z] [d]
cluster
syllabic cons.
Manage voiced Palatal [z] [t]
cluster
syllabic cons.
Please voiceless Alveolar [z] [d]
cluster
syllabic cons.
Stretch voiceless Palatal [z] [t]
cluster
syllabic cons.
Wish voiceless Palatal [z] [t]
cluster

Filler Verbs (irregular).

Break Hold
Buy Make
Choose Meet
Drink Sleep
Eat Spend
Feed Stand
Fly Take
Forget Tell
Give Wear
Go Write

221
 Appendices
Appendix N.
Experiments 6 and 7: Practice and test items from the Phoneme Elision Task in
order of presentation (CTOPP-2; Wagner et al., 2013). Table includes initial word,
phoneme omitted, target word and additional information: place of (phoneme)
elision and task type.

Phoneme(s) Place of
Initial word Target word Task Type
omitted elision
- airplane plane air - -
- doughnut dough nut - -
- cup /k/ up - -
- meat /t/ me - -
- farm /f/ arm - -
1. popcorn corn pop word boundary omission
2. baseball base ball word boundary omission
3. spider der spy word boundary omission
4. bold /b/ old word boundary omission
5. mat /m/ at word boundary omission
6. tan /t/ an word boundary omission
7. mike /k/ my word boundary omission
8. time /m/ tie word boundary omission
9. tiger /g/ tire mid-word adjunction
10. powder /d/ power mid-word adjunction
11. winter /t/ winner mid-word adjunction
12. snail /n/ sail mid-word adjunction
13. faster /s/ fatter mid-word adjunction
14. sling /l/ sing mid-word adjunction
15. driver /v/ dryer mid-word adjunction
16. silk /l/ sick mid-word adjunction
17. flame /f/ lame word boundary omission
18. strain /r/ stain mid-word adjunction
19. split /p/ slit mid-word adjunction
20. fixed /k/ fist mid-word adjunction

222
 Appendices
Appendix O
Experiments 6 and 7: Sample transcription of trials from the Phoneme Elision Task
(based on instructions from CTOPP-2 manual, Wagner et al., 2013) with coding
examples (1- correct; 0 - incorrect).
---
Experimenter: Now we are going to play a word game.

I am going to play you some English words, and the recording will ask you to repeat the
word, and then to repeat the word with a certain sound taken away.

Take the word ‘Window’, it might say, say ‘Window’ without saying ‘ow’. ‘Window’ then
becomes ‘Wind’. Is that clear?

Participant: Yes.
Experiment: Okay. I am going to record your voice for analysis. Can you try and speak as
clearly as possible?

Participant: Okay.

(Practice items)
Recording: Let’s play a word game: Say airplane.
Participant: Airplane.
Recording: Now, say airplane without saying plane.
Participant: Air.
Recording: It’s Air.
Experimenter: Good, well done.
…
Recording: Now, let’s take away smaller parts of words - say cup.
Participant: Cup.
Recording: Now, say cup without saying /k/.
Participant: pa-
Recording: It’s up.
…
(Test items)
Recording: Say bold.

223
 Appendices
Participant: Boat.
Recording: Now say bold without saying /b/.
Participant: Oat.
Coding: repetition – 0 / omission - 1.
…
Recording: Say winter.
Participant: Winter.
Recording: Now, say winter without saying /t/.
Participant: Win.
Coding: repetition – 1 / omission – 0 / adjunction - 0
…
Recording: Say strain.
Participant: Strain.
Recording: Now, say strain without saying /r/.
Participant: S..rain? Sorry, I don’t know.
Coding: repetition – 1 / omission – 0 / adjunction - 0
…
Recording: Say split.
Participant: Split.
Recording: Now, say split without saying /p/.
Participant: Spit?
Coding: repetition – 1 / omission – 0 / adjunction - 0
…
Recording: Say fixed.
Participant: Fixed.
Recording: Now, say fixed without saying /k/.
Participant: f.. fist?
Coding: repetition – 1 / omission – 1 / adjunction - 1

Experimenter:
That’s great. Thank you.

224
 Appendices
Appendix P.
Experiments 6 and 7: Average response accuracy for individual items in the
Phoneme Elision Task across L1 Mandarin and L1 English groups (CTOPP-2,
Wagner et al., 2013).

Experiment 6.

Experiment 7.

225
 References

References

Allan, D. (1992), Oxford Placement Test. Oxford: Oxford University Press.

Amaro, J. C., Campos-Dintrans, G., & Rothman, J. (2018). The role of L1 phonology
in L2 morphological production: L2 English past tense production by L1
Spanish, Mandarin, and Japanese Speakers. Studies in Second Language
Acquisition, 40(3), 503-527.
Anderson, J. R. (1980). Cognitive psychology and its implications. Worth publishers.

Archibald, J. (1998). Second language phonology, phonetics, and typology. Studies

in Second Language Acquisition, 20(2), 189-211.
Bardovi-Harlig, K. (2000). Tense and aspect in second language acquisition: Form,
meaning, and use. Language Learning: A Journal of Research in Language
Studies, 50, 1.
Bates, E., & MacWhinney, B. (1989). Functionalism and the competition model. The
Crosslinguistic Study of Sentence Processing, 3, 73-112.
Bayley, R. (1996). Competing constraints on variation in the speech of adult Chinese
learners of English. In R. Bayley, & D.R. Preston, (Eds.), Second language
acquisition and linguistic variation. Amsterdam: John Benjamins, 97–120.
Best, C. T. (1994). Learning to perceive the sound pattern of English. In Lewis P.
Lipsitt & Carolyn K. Rovee-Collier (eds.), Advances in infancy research, 217–
304. Hillsdale, NJ: Ablex Publishers.
Best, C. T. (1995). A direct realist perspective on cross-language speech perception.
In W. Strange (Ed.), Speech perception and linguistic experience: Theoretical
and methodological issues in crosslanguage speech research (pp. 167–200).
Timonium, MD: York Press.

Birdsong, D. (2005). Interpreting age effects in second language

acquisition. Handbook of Bilingualism: Psycholinguistic approaches, 109, 127.
Bley-Vroman, R. (2009). The evolving context of the fundamental difference
hypothesis. Studies in Second Language Acquisition, 31(2), 175-198.
Bley-Vroman, R. (1988). The fundamental character of foreign language learning.
In Rutherford, W. & Sharwood Smith, M. (Eds.), Grammar and Second
Language Teaching (pp. 19–30). Rowley, MA: Newbury House.
Boatwright, C. (1999). On-line processing of time reference: Meaning before
morphology. Paper presented at the Second Language Research Forum,
Minneapolis, MN.

226
 References

Bock, J. K. (1982). Toward a cognitive psychology of syntax: Information

processing contributions to sentence formulation. Psychological Review, 89(1),
1-47.

Bock, K. (1995). Sentence production: From mind to mouth. In J. L. Miller & P. D.

Eimas (Eds.), Handbook of perception and cognition (2nd Ed.). Vol. 11. Speech,
Language, and Communication (pp. 181-216). San Diego, CA, US: Academic
Press.
Bock, K., & Levelt, W. J. (1994). Language production: Grammatical encoding (pp.
945-984). Academic Press.

Boersma, P. & Weenink, D. (2015). Praat: doing phonetics by computer. Version 6.0
Booij, G. (2005). The grammar of words: An introduction to linguistic morphology.
Oxford University Press.

Bradlow, A., Pisoni, D., Akahane-Yamada, R., & Tohkura, Y. (1997). Training
Japanese listeners to identify English / r / and / l /: IV. Some effects of
perceptual learning on speech production . The Journal of the Acoustical Society
of America, 101(4), 2299-2310.
Chan, M. (1980). Temporal reference in Mandarin Chinese: An analytic-semantic
approach to the study of the morphemes le, zai, zhe and ne. Journal of the
Chinese Language Teachers' Association 15 (3), pp. 33–79.
Chen, C. C. (2009). Ambiguity of le in Chinese: the perfective as well as
imperfective. Journal of Chinese Linguistics, 37(1), 108-129.
Chen, L., Shu, H. U. A., Liu, Y., Zhao, J., & Li, P. (2007). ERP signatures of
subject–verb agreement in L2 learning. Bilingualism: Language and Cognition,
10(2), 161-174.
Chenoweth, N. A., & Hayes, J. R. (2003). The inner voice in writing. Written
Communication, 20(1), 99-118.
Chondrogianni, V., & Marinis, T. (2012). Production and processing asymmetries in
the acquisition of tense morphology by sequential bilingual children.
Bilingualism, 15(1), 5-21.
Clahsen, H., & Felser, C. (2018). Some notes on the shallow structure
hypothesis. Studies in Second Language Acquisition, 40(3), 693-706.

Clahsen, H., Felser, C., Neubauer, K., Sato, M., & Silva, R. (2010). Morphological
structure in native and nonnative language processing. Language
Learning, 60(1), 21-43.
Clark, D. (1982). High-resolution subjective testing using a double-blind
comparator. Journal of the Audio Engineering Society, 30(5), 330-338.

227
 References

Comrie, B. (1976). Aspect: An introduction to the study of verbal aspect and related
problems (Vol. 2). Cambridge university press.
Comrie, B. (1985). Tense. Cambridge University Press.

Cutler, A. (2000). Listening to a second language through the ears of a first.

Interpreting, 5(1), 1-23.
Cutler, A., & Clifton, C. (1999). Comprehending spoken language: A blueprint of the
listener. The Neurocognition of Language, 123-166.
Cutler, A., Mehler, J., Norris, D., & Segui, J. (1986). The syllable’s differing role in
the segmentation of French and English. Journal of Memory and Language, 25,
385-400.
de Bot, K. (1992). A bilingual production model; Levelt's 'speaking' model adapted.
Applied Linguistics, 13, 1-24.
de Bot, K. (2003). A bilingual production model: Levelt’s ‘speaking’ model adapted.
In L. Wei (Ed.), The Bilingualism Reader (pp. 420-442). London; New York:
Routledge.

de Bot, K., & Schreuder, R. (1993). Word production and the bilingual lexicon. The
Bilingual Lexicon, 191, 214.
DeKeyser, R. (2000). the Robustness of Critical Period Effects in Second Language
Acquisition. Studies in Second Language Acquisition, 22(4), 499-533.
Dell, G. S. (1986). A spreading-activation theory of retrieval in sentence
production. Psychological review, 93(3), 283.
Dell, G. S., Chang, F., & Griffin, Z. M. (1999). Connectionist models of language
production: Lexical access and grammatical encoding. Cognitive Science, 23(4),
517-542.
Dell, G. S., Schwartz, M. F., Martin, N., Saffran, E. M., & Gagnon, D. A. (1997).
Lexical Access in Aphasic and Nonaphasic Speakers. Psychological Review,
104, 801– 838.
Dijkstra, T., & van Heuven, W. (2002). The architecture of the bilingual word
recognition system: From identification to decision. Bilingualism: Language and
Cognition, 5(3), 175-197.
Duanmu, S. (2007). The Phonology of Standard Chinese. OUP Oxford.
Dupoux, E., Kakehi, K., Hirose, Y., Pallier, C., & Mehler, J. (1999). Epenthetic
vowels in Japanese: A perceptual illusion?. Journal of Experimental
Psychology: Human Perception and Performance, 25(6), 1568.

228
 References

Dussias, P. E., & Piñar, P. (2010). Effects of reading span and plausibility in the
reanalysis of wh-gaps by Chinese-English second language speakers. Second
Language Research, 26(4), 443-472.

Eimas, P. D., Siqueland, E. R., Jusczyk, P., & Vigorito, J. (1971). Speech perception
in infants. Science, 171(3968), 303-306.
Ellis, N. C. (2006). Selective attention and transfer phenomena in L2 acquisition:
Contingency, cue competition, salience, interference, overshadowing, blocking,
and perceptual learning. Applied linguistics, 27(2), 164-194.
Ellis, N. C., & Wulff, S. (2008). Usage–based approaches to SLA1. Theories in
second language acquisition: An introduction, 1, 75.
Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A
psychometric study. Studies in Second Language Acquisition, 27(2), 141-172.

Ellis, R. (2006). Modelling learning difficulty and second language proficiency: The
differential contributions of implicit and explicit knowledge. Applied
Linguistics, 27(3), 431-463.

Felser, C., Roberts, L., Marinis, T., & Gross, R. (2003). The processing of
ambiguous sentences by first and second language learners of English. Applied
Psycholinguistics, 24(3), 453-489.

Ferreira, F., Henderson, J. M., Anes, M. D., Weeks, P. A., & McFarlane, D. K.
(1996). Effects of lexical frequency and syntactic complexity in spoken-
language comprehension: Evidence from the auditory moving-window
technique. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 22(2), 324.
Flege, J. E. (1995). Second language speech learning: Theory, findings, and
problems. Speech perception and linguistic experience: Issues in cross-language
research, 92, 233-277.
Flege, J. E., & Wang, C. (1989). Native-language phonotactic constraints affect how
well Chinese subjects perceive the word-final English/t/-/d/contrast. Journal of
Phonetics, 17(4), 299-315.
Flynn, S. (1986). Production vs. comprehension: Differences in underlying
competences. Studies in Second Language Acquisition, 8(2), 135-164.
Foygel, D., & Dell, G. S. (2000). Models of Impaired Lexical Access in Speech
Production. Journal of Memory and Language, 43(2), 182-216.
Fraser, C. A. (2007). Reading rate in L1 Mandarin Chinese and L2 English across
five reading tasks. The Modern Language Journal, 91(3), 372-394.

229
 References

Frazier, L. (1978). On comprehending sentences: Syntactic parsing strategies.

Doctoral dissertation, University of Connecticut.
Frenck-Mestre, C., Foucart, A., Carrasco-Ortiz, H., & Herschensohn, J. (2009).
Processing of grammatical gender in French as a first and second language:
Evidence from ERPs. Eurosla yearbook, 9(1), 76-106.
Frenck-Mestre, C., Osterhout, L., McLaughlin, J., & Foucart, A. (2008). The effect
of phonological realization of inflectional morphology on verbal agreement in
French: Evidence from ERPs. Acta Psychologica, 128(3), 528-536.
Friederici, A. D., & Wessels, J. M. (1993). Phonotactic knowledge of word
boundaries and its use in infant speech perception. Perception &
Psychophysics, 54(3), 287-295.
Friederici, A. D., Schoenle, P. W., & Goodglass, H. (1981). Mechanisms underlying
writing and speech in aphasia. Brain and Language, 13, 212-222.
Gardner, Q., Branigan, H., & Chondrogianni, V. (2018). Investigating the production
of L2 inflectional morphology using a picture description paradigm. SAGE
Research Methods Cases
Garrett, M. (1975) The analysis of sentence production. in G. Bower Psychology of
Learning and Motivation (Ed.), pp. 133–177, Academic Press.

Goad, H., & White, L. (2019). Prosodic effects on L2 grammars. Linguistic

Approaches to Bilingualism, 9(6), 769-808.
Goad, H., White, L., & Steele, J. (2003). Defective syntax or L1-constrained
prosodic representations? Canadian Journal of Linguistics, 48(3), 243–263.
Goto, H. (1971). Auditory perception by normal Japanese adults of the sounds" L"
and" R.". Neuropsychologia.
Grosjean, F. (1997). Processing mixed language: Issues, fmdings and models. In A.
de Groot & J. Kroll (Eds.), Tutorials in bilingualism: Psycholinguistic
perspectives (pp. 225-254). Majwah, NJ: LEA
Hahne, A. (2001). What's different in second-language processing? Evidence from
event-related brain potentials. Journal of Psycholinguistic Research, 30(3), 251-
266.

Hahne, A., & Friederici, A. D. (2001). Processing a second language: Late learners'
comprehension mechanisms as revealed by event-related brain
potentials. Bilingualism: Language and Cognition, 4(2), 123-141.

230
 References

Haig, J. 1991: Universal grammar and second language acquisition: the influence of
task type on late learner’s access to the subjacency principle. TESL monograph,
McGill University.

Hancin-Bhatt, B. (1994). Segment transfer: A consequence of a dynamic

system. Second Language Research, 10(3), 241-269.
Hauck, W. W., & Donner, A. (1977). Wald’s test as applied to hypotheses in logit
analysis. Journal of the American Statistical Association, 72, 851–853.
Hawkins, R. (2007). Emergent and innate sources of knowledge in the early L2
acquisition of English verbal morphology. In Martin Atkinson – The Minimalist
Muse (pp. 135–156).
Hawkins, R., & Chan, C. Y. H. (1997). The partial availability of Universal
Grammar in second language acquisition: The ‘failed functional features
hypothesis.’ Second Language Research, 13(3), 187-226.
Hawkins, R., & Liszka, S. (2003). Locating the source of defective past tense
marking in advanced L2 English speakers. Language Acquisition and Language
Disorders, 30, 21-44.
Haznedar, B. & Schwartz, B.D. (1997). Are there Optional Infinitives in child L2
acquisition? In E. Hughes, M. Hughes & A. Greenhill (Eds.), Proceedings of the
21st Annual Boston University Conference on Language Development, 21, 257-
68. Somerville, MA: Cascadilla Press.
Hill, A.A. (1958). Introduction to linguistic structures: From sound to sentence in
English. New York.
Hopp, H. (2010). Ultimate attainment in L2 inflection: Performance similarities
between non-native and native speakers. Lingua, 120(4), 901-931.

Hsieh, F.-t. (2009). The acquisition of English agreement / tense morphology and
copula be by L1-Chinese-speaking learners. Lancaster University LEL
Postgraduate Conference, 3, p. 45.
Jackson, C. N., & Bobb, S. C. (2009). The processing and comprehension of wh-
questions among L2 German speakers. Applied Psycholinguistics, 30(4), 603.
Jackson, C. N., & Dussias, P. E. (2009). Cross-linguistic differences and their impact
on L2 sentence processing. Bilingualism: Language and Cognition, 12(1), 65-
82.
Jiang, N. (2004). Morphological insensitivity in second language processing. Applied
Psycholinguistics, 25(4), 603-634.
Jiang, N. (2007). Selective integration of linguistic knowledge in adult second
language learning. Language learning, 57(1), 1-33.

231
 References

Johnson, J. S. (1992). Critical period effects in second language acquisition: The

effect of written versus auditory materials on the assessment of grammatical
competence. Language Learning, 42(2), 217-248.

Johnson, J. S., & Newport, E. L. (1989). Critical period effects in second language
learning: The influence of maturational state on the acquisition of English as a
second language. Cognitive Psychology, 21(1), 60-99.

Juffs, A., & Harrington, M. (1995). Parsing effects in second language sentence
processing: Subject and Object Asymmetries in wh-extraction. Studies in second
language acquisition, 483-516.

Juffs, A., & Harrington, M. (1996). Garden path sentences and error data in second
language sentence processing. Language learning, 46(2), 283-323.
Jusczyk, P., Luce, P., & Charles-Luce, J. (1994). Infants′ Sensitivity to Phonotactic
Patterns in the Native Language. Journal of Memory and Language, 33(5), 630-
645.
Just, M. A., Carpenter, P. A., & Woolley, J. D. (1982). Paradigms and processes in
reading comprehension. Journal of Experimental psychology: General, 111(2),
228.

Kilborn, K. (1992). On-line integration of grammatical information in a second

language. In Advances in Psychology (Vol. 83, pp. 337-350). North-Holland.
Kormos, J. (2006). Speech production and second language acquisition. Routledge.

Kroll, J. F. (1993). Accessing conceptual representation for words in a second

language. In R. Schreuder & B. Weltens (Eds.),The Bilingual Lexicon (pp. 53–
81). Philadelphia: John Benjamin.
Kuhl, P. K. (1987). Perception of speech and sound in early infancy. Handbook of
Infant Perception, 2, 275-382.
Kuhl, P. K., Williams, K. A., Lacerda, F., Stevens, K. N., & Lindblom, B. (1992).
Linguistic experience alters phonetic perception in infants by 6 months of
age. Science, 255(5044), 606-608.
Lardiere, D. (1998a). Dissociating syntax from morphology in a divergent L2 end-
state grammar. Second Language Research, 14(4), 359–375.
Lardiere, D. (1998b). Case and Tense in the ‘fossilized’ steady state. Second
Language Research, 14(1), 1–26.
Lardiere, D. (2000). Mapping features to forms in second language acquisition.
Second language acquisition and linguistic theory, 102-129.

232
 References

Lardiere, D. (2003). Second language knowledge of [±past] vs.[±finite].

In Proceedings of the 6th Generative Approaches to Aecond Language
Acquisition Conference (GASLA 2002) (pp. 176-189).

Lardiere, D. (2008). Feature assembly in second language acquisition. In J. M.

Liceras, H. Zobl, & H. Goodluck (Eds.), The role of formal features in second
language acquisition (pp. 106-140). New York: Lawrence Erlbaum Associates.

Lee, J.F., Cadierno, T., Glass, W., and VanPatten, W. (1997). The effects of lexical
and grammatical cues on processing past temporal reference in second language
input. Applied Language Learning, 8(1), 3-14.

Lenneberg, E.H. (1967). The Biological Foundations of Language. New York:

Wiley.
Levelt, W. J. (1989). Speaking: From intention to articulation. Cambridge, MA, US:
The MIT Press.
Levelt, W. J. (2001). Spoken word production: a theory of lexical access.
Proceedings of the National Academy of Sciences of the United States of
America, 98(23), 13464–13471.
Levelt, W. J. M., Roelofs, A., & Meyer, A. S. (1999). A theory of lexical access in
speech production. Behavioral and Brain Sciences, 22(1), 1-38.

Lewy, N., & Grosjean, F. (2008). The Lewy and Grosjean BIMOLA model. In
Grosjean, F. (ed.), Studying Bilinguals, pp. 201–210. Oxford: Oxford University
Press.

Li, C. N., & Thompson, S. A. (1989). Mandarin Chinese: A functional reference

grammar (Vol. 3). University of California Press.
Lieberman, M. (2013). The importance of comprehension to a rounded view of
second language acquisition. Paper presented at the Department of Linguistics,
University of Utah, Salt Lake City, UT.
Lin, J. W. (2010). A tenseless analysis of Mandarin Chinese revisited: A response to
Sybesma 2007. Linguistic Inquiry, 41(2), 305-329.
Lin, T. H. J. (2015). Tense in Mandarin Chinese sentences. Syntax, 18(3), 320-342.
Liu, H., Bates, E., & Li, P. (1992). Sentence interpretation in bilingual speakers of
English and Chinese. Applied Psycholinguistics, 13(4), 451-484.
Lo, S., & Andrews, S. (2015). To transform or not to transform: Using generalized
linear mixed models to analyse reaction time data. Frontiers in psychology, 6,
1171.

233
 References

MacWhinney, B. (1987). The Competition Model. Mechanisms of language

acquisition, 249-308.
Marinis, T., Roberts, L., Felser, C., & Clahsen, H. (2005). Gaps in second language
sentence processing. Studies in Second Language Acquisition, 53-78.
Marslen-Wilson, W. D. (2007). Processes in language comprehension. Oxford
Handbook of Psycholinguistics, 495-524.

Mattys, S. L., & Jusczyk, P. W. (2001). Phonotactic cues for segmentation of fluent
speech by infants. Cognition, 78(2), 91-121.
Maurer, D., & Werker, J. F. (2014). Perceptual narrowing during infancy: A
comparison of language and faces. Developmental Psychobiology, 56(2), 154-
178.
McCarthy, J. & Prince A. (1995). Prosodic morphology. In John Goldsmith (ed.),
Handbook of Phonological Theory. Cambridge, MA: Blackwell.
McClelland, J. L., & Elman, J. L. (1986). Interactive processes in speech perception:
The TRACE model. In Parallel distributed processing: Explorations in the
Microstructure of Cognition, Vol. 2: Psychological and biological models (pp.
58-121).
McDonald, J. L. (2006). Beyond the critical period: Processing-based explanations
for poor grammaticality judgment performance by late second language learners.
Journal of Memory and Language, 55(3), 381-401.
McQueen, J. M., & Cutler, A. (1998). Morphology in word recognition (pp. 406-
427). Blackwell.
Meyer, A. S., Huettig, F., & Levelt, W. J. (2016). Same, different, or closely related:
What is the relationship between language production and comprehension?
Journal of Memory and Language, 89, 1-7.
Murphy, V. A. (1997). The effect of modality on a grammaticality judgement task.
Second Language Research, 13(1), 34-65.
Myers-Scotton, C. (1992). Comparing codeswitching and borrowing. Journal of
Multilingual and Multicultural Development, 13(1-2), 19-39.
Myers-Scotton, C. (2006) Natural codeswitching knocks on the laboratory door.
Bilingualism, Language and Cognition, 9 (2), 203-212.
Osterhout, L., & Mobley, L. A. (1995). Event-related brain potentials elicited by
failure to agree. Journal of Memory and Language, 34(6), 739-773.
Otake, T., Hatano, G., Cutler, A., & Mehler, J. (1993). Mora or syllable? Speech
segmentation in Japanese. Journal of Memory and Language, 32, 258-278.

234
 References

Paradis, J., Tulpar, Y., & Arppe, A. (2016). Chinese L1 children's English L2 verb
morphology over time: Individual variation in long-term outcomes. Journal of
Child Language, 43(3), 553-580.
Papadopoulou, D., & Clahsen, H. (2003). Parsing strategies in L1 and L2 sentence
processing: A study of relative clause attachment in Greek. Studies in Second
Language Acquisition, 501-528.
Poulisse, N., & Bongaerts, T. (1994). First language use in second language
production. Applied Linguistics, 15(1), 36-57.
Prévost, P., & White, L. (2000). Missing Surface Inflection or Impairment in second
language acquisition? Evidence from tense and agreement. Second Language
Research, 16(2), 103–133.
Rainey, C. (2016). Dealing with separation in logistic regression models. Political
Analysis, 24(3), 339-355.
Roberts, L., & Felser, C. (2011). Plausibility and recovery from garden paths in
second language sentence processing. Applied Psycholinguistics, 299-331.
Roberts, L., & Liszka, S. A. (2013). Processing tense/aspect-agreement violations
on-line in the second language: A self-paced reading study with French and
German L2 learners of English. Second Language Research, 29(4), 413-439.

Roberts, L., Mackey, A., & Marsden, E. (2016). Self-paced reading and L2
grammatical processing. Advancing Methodology and Practice. The IRIS
repository of instruments for research into second languages, 58-72.

Roelofs, A. (1997). The WEAVER model of word-form encoding in speech

production. Cognition, 64(3), 249-284.
Roelofs, A. (2015). Modeling of phonological encoding in spoken word production:
From G ermanic languages to Mandarin Chinese and J apanese. Japanese
Psychological Research, 57(1), 22-37.
Romberg, A. R., & Saffran, J. R. (2010). Statistical learning and language
acquisition. Wiley Interdisciplinary Reviews: Cognitive Science, 1(6), 906-914.

Rumelhart, D. E., & McClelland, J. L. (1982). An interactive activation model of

context effects in letter perception: II. The contextual enhancement effect and
some tests and extensions of the model. Psychological Review, 89(1), 60–94.

Schilder, F. (1997). Temporal relations in English and German narrative discourse.

Schlesinger, I. M. (1977). The role of cognitive development and linguistic input in
language acquisition. Journal of Child Language, 4(2), 153-169.

Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-Prime (Version 2.0).
Pittsburgh, PA: Psychology Software Tools Inc.

235
 References

Segalowitz, N. S., & Segalowitz, S. J. (1993). Skilled performance, practice, and the
differentiation of speed-up from automatization effects: Evidence from second
language word recognition. Applied Psycholinguistics, 14(3), 369-385.
Selkirk, E. (2008). The Prosodic Structure of Function Words. Optimality Theory in
Phonology: A Reader(February 1993), 464-482.
Silva, R., & Clahsen, H. (2008). Morphologically complex words in L1 and L2
processing: Evidence from masked priming experiments in
English. Bilingualism: Language and cognition, 11(2), 245-260.

Slabakova, R. (2015). Acquiring temporal meanings without tense morphology: The

case of L2 Mandarin Chinese. The Modern Language Journal, 99(2), 283-307.
Slobin, D. I. (1987). Thinking for speaking. In Annual Meeting of the Berkeley
Linguistics Society (Vol. 13, pp. 435-445).
Smith, C. S. (1991). The aspectual system of Mandarin Chinese. In The parameter of
aspect (pp. 343-390). Springer, Dordrecht.
Smith, C. S. (1994). Aspectual viewpoint and situation type in Mandarin
Chinese. Journal of East Asian Linguistics, 3, 107–146.
Solt, S., Pugach, Y., Klein, E. C., Adams, K., Stoyneshka, I., & Rose, T. (2004). L2
perception and production of the English regular past: Evidence of phonological
effects. The CUNY Graduate Centre: New York.
Song, H., & Schwartz, B. (2009). Testing the fundamental difference hypothesis: L2
adult, l2 child, and l1 child comparisons in the acquisition of Korean Wh-
constructions with negative polarity items (Vol. 31).
Steinhauer, K., & Ullman, M. T. (2002). Consecutive ERP effects of morpho-
phonology and morpho-syntax. In Brain and Language (Vol. 83, No. 1, pp. 62-
65).
Strange, W., & Dittmann, S. (1984). Effects of discrimination training on the
perception of /r/- /l/ by Japanese adults learning English. Perception &
psychophysics, 36(2), 131-145.
Tang, S. (2016). Aspectual system. In C. Huang & D. Shi (Eds.), A Reference
Grammar of Chinese (Reference Grammars, pp. 116-142). Cambridge:
Cambridge University Press.
Tolentino, L., & Tokowicz, N. (2011). Across languages, space, and time. Studies in
Second Language Acquisition, 33(1), 91-125.
Ullman, M. T. (2004). Contributions of memory circuits to language: The
declarative/procedural model. Cognition, 92, 231–270.

236
 References

Ullman, M. T. (2005). A cognitive neuroscience perspective on second language

acquisition: The declarative/procedural model. Mind and context in adult second
language acquisition, 2005, 141-78.

Van Galen, G. P. (1991). Handwriting: Issues for a psychomotor theory. Human

movement science, 10(2-3), 165-191.
VanPatten, B., Keating, G. D., & Leeser, M. J. (2012). Missing verbal inflections as
a representational problem: Evidence from self-paced reading. Linguistic
Approaches to Bilingualism, 2(2), 109-140.
Vitevitch, M., Luce, P., Charles-Luce, J., & Kemmerer, D. (1997). Phonotactics and
Syllable Stress: Implications for the Processing of Spoken Nonsense Words.
Language and Speech, 40(1), 47-62.
Wagner, R. K., Torgesen, J. K., Rashotte, C. A., & Pearson, N. A.
(2013). Comprehensive Test of Phonological Processing: CTOPP-2. Austin,
TX: Pro-ed.
Weber, A., & Cutler, A. (2006). First-language phonotactics in second-language
listening. The Journal of the Acoustical Society of America, 119(1), 597-607.
Weber-Fox, C. M., & Neville, H. J. (1996). Maturational constraints on functional
specializations for language processing: ERP and behavioral evidence in
bilingual speakers. Journal of cognitive neuroscience, 8(3), 231-256.

Werker, J. F., & Tees, R. C. (1984). Phonemic and phonetic factors in adult cross‐
language speech perception. The Journal of the Acoustical Society of
America, 75(6), 1866-1878.

White, L. (2009). Some questions about feature re-assembly. Second Language

Research, 25(2), 343-348.
Williams, J. N. (2006). Incremental interpretation in second language sentence
processing. Bilingualism Language and Cognition, 9(1), 71.
Williams, J. N., Mobius, P., & Kim, C. (2001). Native and non-native processing of
English wh-questions: Parsing strategies and plausibility constraints. Applied
Psycholinguistics, 22(4), 509-540.
Wolfram, W. (1985). Variability in tense marking: A case for the obvious. Language
Learning, 35(2), 229-253.

Zhang, Q., & Damian, M. F. (2010). Impact of phonology on the generation of

handwritten responses: Evidence from picture-word interference tasks. Memory
and Cognition, 38(4), 519-528.

237
 References

Zorn, C. (2005). A Solution to separation in binary response models. Political

Analysis, 13(2), 157-170.

238